1927 Thorndike

THE MEASUREMENT OF INTELLIGENCE
THE MEASUREMENT OF
INTELLIGENCE
By
EDWARD L. THORNDIKE , E. O. BREGMAN , M. V. COBB ,

ELLA WOODYARD , and the Staff of the Division of Psychol-
ogy of the Institute of Educational Research of Teachers Col-
lege , Columbia University
The investigations and results reported

in this volume were made possible by
a grant from the Carnegie Corporation
Bureau of Publications
TEACHERS COLLEGE , COLUMBIA UNIVERSITY
NEW YORK
Medical
426
PREFACE
This volume represents the fruits of three years of investigation
(from July 1 , 1922 , to July 1 , 1925 ) by the Division of Psychology
of the Institute of Educational Research . It attempts to answer
the essential questions concerning the nature and meaning of the
measurement of a mental fact in the sample case of intelligence , or
rather of a defined segment thereof . Its conclusions , in so far
as they are warranted , should become the basis of sound practice
in the construction and calibration of scales for use in mental mea
surement . According to them , the present theory and practice of
measurement of mental abilities are justified to a remarkable de
gree in certain respects , but in others should be almost recreated .
Some of the most important of these conclusions were reached
only in the last six months of the inquiry and are consequently
presented with less adequate evidential support than is desirable .
The concept of area of intellect in particular needs more experi
mentation to make it clear , and still more to demonstrate its sound
ness and worth .
We had intended to add a long chapter reviewing the literature
on the topics dealt with in this volume , but it seemed more impor
tant to exemplify and apply the results of our conclusions in a
concrete series of tasks selected and scaled according to the prin
ciples described ; and there was not time to do both . We hope to
be able to publish such a review later , and in particular to do jus
tice to the notable contribution of Kelley ( ' 23a ) , which deserves
most careful study by everyone who is concerned with the general
logic of mental measurements .
We had intended also to include full treatment of the method
of obtaining a group of approximately known forms of distribution
in respect of a mental trait measured in truly equal units , by taking
the members of an array in that trait who have identical scores in
a second trait correlated with the trait in question . This method
was abandoned in favor of a better one , but nearly a third of our
time and effort was spent in exploring its possibilities . The results
should be made known , both because of their intrinsic interest , and
because otherwise someone will surely be tempted to do again what
has already been done by us . The material is , however , highly
technical and elaborate ; and it seemed best not to include it in
this volume .
V
vi PREFACE
The general responsibility for the work rests upon the senior
author , who planned and directed the various inquiries , organized
the results , and wrote this book , with the exception of Appendix
III. It would , however , have been utterly impossible for him to
have carried the work through without the financial assistance of
the Carnegie Corporation and the Trustees of Teachers College ,
and without the loyal cooperation of the staff of the Division of
Psychology of the Institute of Educational Research , and many
scientific workers in all parts of the country . Dr. Bregman col-
lected and organized most of the facts which are used in Chapter
VII and Appendix III
, and some of those used in Chapter VIII
.
Miss Cobb devised many of the tasks of levels A, B , C , D , E , and
F, and , with the aid of Dr. Murdoch , Dr. Tilton and Miss Robin-
son , measured 180 imbeciles of mental age 3 to 5 and 100 of mental
age 6. Dr. Woodyard has arranged and supervised most of the
testing and scoring in grades 4 to 9 , and has shared in the evalua-
tion of the difficulty of the thousands of tasks which have been
used in our experiments . Dr. Murdoch made all the tests with the
fifty feeble - minded at Polk . Mrs. Miner has computed most of the
correlations . Miss Robinson , Dr. Hunsicker , Dr. Tilton , and Mr.
Upshall have given expert and painstaking service in testing and
scoring .
Dr. Toops and Mrs. Ruger worked up the data which provided
the first set of tasks graded in difficulty from which the final scale
eventually developed . Miss Hanson , Mrs. Work and Miss Wilcox
have had a large share in the arrangement and tabulation of the
results .
We are indebted , for most courteous and efficient cooperation ,

to all the psychologists on the staff of Teachers College , to fifty
members of the American Psychological Association who made
various ratings for us , to Dr. Raymond Franzen and Dr. Grace A.
Taylor who supplied valuable records , to Miss Elizabeth E. Far-
rell , Inspector of Ungraded Classes , New York City , Mr. George
Melcher , Assistant Superintendent of Schools , Kansas City , Mis-
souri , Dr. E. H. Nifernecker , Director of the Bureau of Educa-
tional Research of New York City , Dr. Howard W. Potter , Clinical
Director of Letchworth Village , Dr. Louise M. Poull , Psychologist
at the Randall's Island Institution , Mr. Lionel J. Simmons , Super-
intendent of the Hebrew Orphan Asylum of New York City , and
to the many principals and teachers who have facilitated our ex-
perimentation .
CONTENTS
CHAPTER I. - The Present Status .
PAGE
Ambiguity in content ..
7 3 1
Arbitrariness of units....
Ambiguity in significance .
Measurements of intelligence are measures of intellectual
products 11
Measurements of intelligence imply valuation . 12
Truth 14
Development with age . 16
...
Ability to learn 17
...
Other attempted simplifications of the process of valuation
18
Relational thinking 19
The content or data of tests of intellect 20
..
The form of tests of intellect 21

..
Scoring the products intellect 22

of
Further facts concerning difficulty

.
Width or extent or range ..... 31
82283
Speed
importance and quickness of
of
The relative altitude extent

,
333
intellect
CHAPTER The Measurement of Difficulty

-
II
.
The present status 37

..
difficulty by way
of
Measurement differences knowledge

in
of
of the form of distribution of the variations of an individual

in level of intellect 40
...
The relation of the variability of an individual his amount

to
of ability 43
.
Measurement by way of the form of distribution intellect

of
in
some defined group 51

.
Measurement by way of the form distribution an array

of
of
in correlation table.... 54
a
The defects of the Measurements so far described 56

..
CHAPTER III The Measurement the Intellectual Difficulty

-
of
.
of Tasks and of Level of Intellect More Rigorous and

:
Exact Methods
.
Intellectual difficulty 62
...
3939
Intellect CAVD 65
...
vii
viii CONTENTS
PAGE
The relation of intellect CAVD to the abilities measured by
ordinary intelligence examinations .. 96
The homogeneity of difficulty CAVD ... 101

The inference from the form of distribution of a grade popu
lation in Standard Intelligence Examinations scores to the
form of its distribution in level or altitude of intellect
CAVD
-
104
CHAPTER IV. The Measurement of the Intellectual Difficulty of

a single Brief Task .
The problem in the case of single tasks , each of which measures
intellect plus a mere sampling error .. 109
The problem in the case of such single tasks as are used in
...
CAVD or in Standard Intelligence Examinations 114
... ...
The solution by the use extensive composite tasks 118
of
The correlations of single tasks with measures of intellec 122

Summary
-
132
CHAPTER The Measurement of the Intellectual Difficulty

V.
of
Tasks by Consensus of Expert Opinion
a
The Experiments . 135

..
The Ratings 138

The Validity of the Consensus 141
..
-
Summary 156
CHAPTER VI Levels of Intellect

.
Composite tasks 160

...
The construction of composite tasks 175

...
10 composites word knowledge or 179

in
V
"
."
-
The construction composite tasks sentence completion

of
10
in
,
-
arithmetical problems and the understanding of sentences

,
and paragraphs 193

...
The difficulty of the 10 composites 208

...
-
The combination of 10 composites into 40 composites 211
-
..
-
CHAPTER VII The Transformation of the Scores Standard

of
.
Intelligence Examinations into Terms of Scales with

Equal Units
.
The method of transformation illustrated by the Thorndike

,
Examination and Army Alpha .......... 224

The National Intelligence Examination 239
A
..
The Otis Advanced Examination ............ 245

CONTENTS ix
PAGE
The Haggerty Examination , Delta 2. 247
The Terman Group Test 250
....
The Myers Mental Measure 254
...
The Pintner Non Language Test 254
.
The I.E.R. Test of Selective and Relational Thinking Gen-
,
eralization and Organization 257
..
The Brown University Psychological Examination 260
The Army Examination 264
...
a
CHAPTER VIII The Form of Distribution Intellect Man
-
in
of
.
.
General considerations 271
..
The evidence 274

The form of distribution ages up fifteen 285
at
to
..
The form of distribution in adults 287
...
CHAPTER IX for Measuring Altitude Intellect

-
Scale
P of
A
.
.
The difficulty of composites and 295
M
K
Q.
O
J
I,
,
,
L,
,
,
Estimating from ot- 297
σ
Expressing the of each group terms of common unit

in
303
o₁
.
Expressing the measures difficulty distances from com-
as
of
a
mon point of reference 314
...
The difficulty of composites and ... 321

H
E
B
A
G
C
,
,
,
,
F,
,
Estimating from 323

ot
σ
Expressing the each group 328

of
terms
of
in
o19-
o,
Expressing the measures difficulty distances from com-

as
of
mon point of reference..... 331

The Scale.......
-
336
CHAPTER X. The Absolute Zero Intellectual Difficulty

of
Locating zero difficulty by experiment 340

...
program of tasks use measuring tasks of very little in-

in
to
A
tellectual difficulty 341

Locating zero difficulty by consensus 342
a
..
CHAPTER XI The Measurement the Altitude an Individ-

-
of
of
.
ual Intellect
.
The form of the curve of percent correct relation difficulty 351

to
in
Estimating the CAVD altitude of an individual 365

..
CHAPTER XII The Measurement of Width Area of

-
and
.
Intellect
.
Width of intellect the case of truly intellectual task- 373

in
X CONTENTS
PAGE
Width of intellect in the of
the number of single short
sense
tasks mastered , any one of these tasks being only a very
partial representation of intellect 376
...
Area of intellect........... 378
-
Proportional counts 383
CHAPTER XIII The Relations Altitude Width Area and
of
to
,
Speed .
.
The relation between altitude and W 10C 10A 10V
+
+
+
(
10D i.e. number of 40 omposite CAVD tasks succeeded
,
a ,
)
-c
with given level difficulty 388
at
of
.
The relation between altitude and W 1C 1A or 1V or 1D
or
(
),
i.e. the number of single tasks succeeded with given
at
a
,
level 390
The relation between altitude and area of intellect 397
.
The relation of altitude level of intellect speed
or
to
400
.
CHAPTER XIV The Meaning Scores Obtained Standard
in
-
of
.
Intelligence Examinations
.
The meaning of the Binet Mental Ages 402

..
The meaning of scores obtained Standard group examina-

in
tions 403
The meaning of scores obtained tests of the ability learn
in
to
and to improve...... 408
Mean Square Error of CAVD Altitude Units of the CAVD
in
a
Equalling
-
Scale 411
19
1.00
(
.)
CHAPTER XV The Nature of Intellect

.
working definition intellect 413

of
A
...
The hypothesis that quality intellect depends upon quantity

of
of connections 415
Experimental verification the quantity hypothesis 422
of
..
Summary 430
CHAPTER XVI The Measurement of Original Intellectual

- .
Capacity and of Acquired Intellectual Ability

.
The present status of opinion 433

...
General principles 435

...
The use of novel tasks 437

...
The use of familiar tasks 439

..
The use of series graded for susceptibility environmental

to
a
influences 440
The test and results of Burt 447
CONTENTS xi
PAGE
The use of altitude and width of intellect 458
...
Other methods separating original capacity from acquired
of
ability 460
Summary 462
CHAPTER XVII Changes the Altitude and Area of Intellect

-
in
.
With Age
.
Altitude 463
Area 467
General considerations 468
...
CHAPTER XVIII Summary Results and Applications

of
the
-
to
.
Measurement of Human Abilities General

in
.
Summary of results 469
..
Applications the measurement human abilities general 476

of
in
to
...
APPENDIX The Form of Distribution of an Individual's Varia-
I.
tions in Intellect 491
APPENDIX The Relation of an Individual's Variability His

II
to
.
Ability in Tests of Intelligence ....... 497
APPENDIX III On the Form of the Distribution of Intellect the

in
.
Sixth Grade the Twelfth Grade and Among

,
College Freshmen 521

...
APPENDIX IV The Homogeneity of Intellect CAVD All Levels

at
.
of Difficulty 556
.
APPENDIX V. The Adequacy Tasks of Any One Level of Diffi-

of
culty Measure of All of Intellect CAVD 565

as
.
a
APPENDIX VI The Estimated Form of Distribution of Various

.
Groups 572
UNIVERSITY OF MICHIGAN
Original from
Digitized by Google
3)/6o06-pd#¥asn ssad0e/buo0'1snuityuiey* mMMM//:d11y / pazt1Tbtp-a16009 ‘utewog 2T1qNd
8ZOSIZOOOSTOGE ‘Apw//Z70Z/}EU* a pueYy’ pYy//:sdiI7y / IWD 0Z:Z@ OZ-vO-EZOZ UO pazeusua
LIST OF TABLES
TABLE 1 - Variations of the Scores of Thirteen ( or Fewer ) 30
Minute Trials with Part I
of the Thorndike Intelli
PAGE
gence Examination for High School Graduates from

the Median Score for the Individual in All Thirteen
Trials . 20 Gifted Pupils , a , c , d , e , · . . u. 13 Dif
-
ferent Days 45
TABLE 2 The Relation of the Variability of an Individual to His

Amount of Ability in Fifteen Tests or Amalgama
-
tions of Tests , Using Eight Levels of Ability . 47
TABLE 3 The Variability of Four Individuals in Intellect Accord
ing to a Certain Additive Combination of Factors All
4-
Positive 49
- Four Forms of Distribution

TABLE 52
....
TABLE Approximate Percentages Which the Differences Diffi
in
5
culty between Task Task T₂ Task

T¸
etc. are
T₁
of
,
the Difference between According the ,

T.
and
to
-
T
5,
Form of Distribution of the Group ………………………… 52

…
...
TABLE The Correlation of Success Task 281 with Average

in
6
-
Score in Total Series of Intellectual Tasks... 55
a
TABLE The Correlation between CAVD Summation Score and

7
Stanford Binet Mental Age the Case of 178 Imbe

in
ciles Sixteen Years Old Older of Mental Age 28

or
-
,
Months to 59 Months 98
..
TABLE The Correlation between Score in the Thorndike Exami

8
nation for High School Graduates Average of Two

(
Forms and an Incomplete Sampling of Intellect

)
-
CAVD 100
TABLE The Correlations between Scores in Stock Intelligence
9
Examinations and Level Scores in Arithmetical Prob

lems and Sentence Completions 180 Pupils Grades
in
.
to 12. Data from Clark '24 107

7
).
TABLE The Effect of Decreasing the Error of Estimating the

-
10
Difficulty of the Median Task of Composite Ten
“ of
a
by the Use of the Percent of Group Scoring

or
a
More Right out of Ten Place of the Median of the

in
"
Ten Percents of the Group Scoring Right the

in in
"
"
Tasks Taken One Time Vocabulary Tasks the

at
of a
Case of 250 Pupils Grade 113

8
xiii
xiv LIST OF TABLES
-
PAGE
TABLE 11 The Difficulty and Intellectualness of 30 Single Tasks in
Understanding Sentences , Measured by the Percent of
668 11th Grade Pupils Succeeding with Each , and by
the Correlations of Success in Each with the Average
Score in Two Forms of the I.E.R. Sel . Rel ., Gen. Org .
Examination 121
TABLE 12 Percentages Succeeding and Correlations with a Criterion

in the Case of 24 Reading Tasks and 52 Vocabulary
Tasks : Grade 11 : n = 668 for the Reading Tasks and
...
454 for the Vocabulary Tasks 123
...
Percents Succeeding and Correlations with Criterion of
-
TABLE 13
a
Intellect the Case 240 College Graduate
of
in
124
.
,
TABLE 14 The Correlations Bi Serial of Each of 99 Reading

(
-
r)
and Vocabulary Tasks with Intellect I.E.R. Sel Rel
.,
(
.
Gen. Org Grouped According the Percent Suc
to
,
)
.
-
ceeding with the Task 125
.
TABLE 15 Overlappings and Bi Serial r's for 35 Elements...........................

-
The Correspondence between Success Single Small

16
TABLE
-
in
a
Task and Intellect Measured by the Overlapping
as
,
of the Score Intellect of Those Failing with the

in
Task Past the Median Score in Intellect of Those

Succeeding with the Task Compiled from the Origi
.
nal Data of Vincent 131

..
TABLE The Correspondence between the Sum of the Ratings of

-
17
Ten Judges and the Sum of the Ratings of the Other
Ten 137
The Probable Divergence Difficulty Rating by Ex

of
TABLE
18
-
20
a
perts from the Average of an Infinite Number of

Ratings of the Task Each by 20 Experts 139
.
TABLE Measures of the Difficulty of Composite Tas............. 144

- -
10
19
-
TABLE Differences Difficulty of Various Composite Tasks and

in
20
of the Median Sums of 20 Expert Ratings of the
Single Tasks of These Composites Which Were Rated
.
Each Difference Expressed Percent of the Dif

as
Is
ferences between the and the Composite of Its

D
A
Kind
-
145
TABLE 21 Form of Distribution Used in the Calculations of Tables

19 and 20. Relative Frequencies Equal Successive
at
Intervals 146
LIST OF TABLES XV
TABLE 22 - Measures
sumed
of Difficulty If
Is Form A or
the Form of Distribution As-
a Rectangle . Distance from
PAGE
/ 25......
-
C.T. in Terms of σ or Q 148
TABLE 23 Differences in Difficulty of Various Composite Tasks and

of the Median Sums of 20 Expert Ratings of the Sin-
gle Tasks of
These Composites Which Were Rated .
Each Difference Is Expressed as a Percent of the Dif-
ference between the A and the D Composite of Its
Kind 149
TABLE 24 Difficulty of Twelve Composites by the Results with 240

College Graduates , and 189 Candidates for College
Entrance , in Distances + and - from the Median for
the 240 , in Terms of the Sigma
Composite Con- of the
cerned .Also the Median Ratings by the Consensus
...
-
of Such Tasks in Each Composite As Were Rated 150
-
TABLE 25 Measures of Difficulty 152
....
TABLE 26 Measures of Difficulty 153

..
27 Difficulty of CAVD 40 omposite

-
TABLE The Differences

in
-C
Tasks by Experiment and by the Consensus of 20

Experts 157
TABLE 28 Percents Correct for Each Single Word of Seven 10-

Word Composite Tasks Each Various Groups of
of
in
Individuals 185
TABLE 29 Percents Correct the Single Tasks of Word Knowl-

- - -
in
edge 10 omposite Tasks 10 and ...................

190
11
8,
9,
:
.
-C
TABLE 30 Percents Correct for Each Single Word of the Seven 10-
Word Composite Tasks la ......
2a
3a
5a
6a
7a
4a and 191
,
TABLE 31 The Percentages Obtaining Five or More Right Out of

Ten the Vocabulary Composites
2a
3a
in
1a
1,
,
2,
,
3,
etc. 193
TABLE Percents Correct the Single Tasks of Word Knowl-

in
32
edge Composite Tasks and Adult
D.
180
A
B
C
,
,
:
Imbeciles 194
TABLE 33 Percents Correct the Single Tasks of Word Knowledge

-
in
and H. 195
G
F,
,
E,
TABLE 34 Percents Succeeding with Each Single Task of Various

Two Groups Adult Imbecil ..............
of
omposites
10
196
in
-C
Succeeding with Each Single Task of Various

-
TABLE 35 Percents
Four Groups 100 Adults Mental
of
Composites
10
in
:
-
xvi LIST OF TABLES
PAGE
Age 6, 50 Feeble - Minded of Class 3 in an Institution .
Pupils in Special Classes in a Large City , and Pupils
in Grade 4 ( second half ) .. 197
TABLE 36 - The Permilles Succeeding with Each Single Task of

199
Various 10 - Composites of Sentence Completion .......
TABLE 37 - The Permilles Succeeding with Each Single Task of Vari-
202
ous 10 - Composites of Arithmetical Problems
...
TABLE 38 The Permilles Succeeding with Each Single Task Vari-
- -
of
omposites and Reading 205
of
ous 10 Directions
..
-C
TABLE 39 The Difficulty of Composites Co. and Ar

10
D
A
B
C
;
,
,
-
, .
and V. and D. and
A
A
B
B
D
D
C
D
C
,
;
,
,
in the Case of 180 Adult Imbeciles 210
..
The Difficulty Composites Co. Ar
-
TABLE 40 and
of
10
, E
I;
,
F,
,
-
.
and V. and D. and
H
H
E
E
G
G
;
;
F,
,
F,
,
F,
Various Groups Distances Are Omitted from
in
o
(
.
This Table 210

.
).
TABLE 41 The Difficulty of omposites Co. and

-
K
10
, H
, G
F, F
J
;
,
I,
,
-C
Ar and V. and
H
H
G
G
J
% 4
;
;
F,
,
,
I,
2,
.
D. and 22 Various Groups 211

in
in
of ½
H
G
s
F,
,
,
1,
.
TABLE 42 The Difficulty Composites Measured by the Percents
10
-
of 147 Pupils Grade 52 Succeeding with Five
or or
in
-More of the Ten Single Tasksand by Distances +

,
from the Median Difficulty for Grade 52 Units

in in
,
of the Mean Square Variation Grade 5½ Level

of
of Whatever Ability the 10 Composite Measures

in
-
Each Case Similar Facts for 205 Pupils and 200
......
.
Pupils Grade 52. The 147 Pupils Are Those Who

in
Were Included in Both the 205 and the 200 212
The Difficulty Various Composites

of
TABLE 43 the Case 44

of
10
in
-
Adults Recruits the United States Army 213

in
..
:
44 Difficulty of 10 Composites Measured by the Permilles of

-
TABLE
-
Successes and by Distances - from the Median

or
+
Difficulty for Grade 82 Units of the Mean Square

in
,
Variation of Grade 812 in the Ability Measured by the

Composite 213
Difficulty Composites Measured by the Percents
of
TABLE
-
of
45
10
-
Two Groups 246 Pupils Grade and 264 Pupils

in
or 9
(
Grade Succeeding with Five More of the Ten

in
9
)
Single Tasks and by Distances or from the Median

-
+
,
Difficulty for the Group Units of the Mean Square

in
LIST OF TABLES xvii
PAGE
Variation of the Group in the Altitude of Whatever
Ability the 10 - Composite Measures . 214
TABLE 46 - or -
Difficulty of 10 - Composites Measured by the σ Distances +
from the Median Difficulty of a Given Grade in
Units of the Mean Square Variation of the Population
of That Grade in the Ability Measured by the 10
215
-
Composites
TABLE 47 The Difficulty of Various 10 - Composites Measured by the

Percent Succeeding by the Distance from the
and
Median in Terms of the Mean Square Variation of the
Group : 422 Normal School Seniors . The form of dis
tribution is assumed to be " normal ." The division
into two groups of 150 and 185 is approximately at
random . The group of 87 represented a somewhat
superior selection and took certain additional tests....... 216
TABLE 48 - Difficulty of 10 - Composites Measured by the Percents of

( a ) 240 College Graduates and ( b ) 100 Students of
Education ( College or Normal School Graduates )
Succeeding , and by Distances + or - from the Median
Difficulty for the College Graduates in Question , in
Units of the Mean Square Variation of the College
Graduates in Question in Altitude of Whatever Ability
the 10- Composite Measures in Each ...... 217
-
Cas
TABLE 49 Difficulty of 10 - Composite Measured by the Percents of

53 Adult Students Succeeding , and by Distances + or
- from the Median Difficulty for the Group , in Units
of the Mean Square Deviation of the Group in the
Ability Measured by the Composite .. 218
TABLE 50 - Difficulty of 10 - Composites Measured by the Percents of

63 University Students Succeeding 218
...
TABLE 51 Summary ofthe Facts Concerning the Difficulty of the

- -
Four 10 Composites Constituting Each 40 Composite 220

...
-
TABLE 52 Thorndike Examination for High School Graduates

.
Part Forms and N. Grade 12. 1527................ 225

D
n
=
I,
TABLE 53 Thorndike Examination for High School Graduates

.
Part Forms and N. Scores from 54 129 Cor

to
-
D
I,
rected to Be in Truly Equal Units 226

..
-
TABLE 54 Army Alpha Grade 1721 227
n
=
9
..
:
TABLE 55 Army Alpha Grade 12 1387 228

=
n
:
.
2
xviii LIST OF TABLES
-
PAGE
TABLE 56 Army Alpha : College Freshmen ( Ohio ) : n = 2545 229
...
TABLE 57 Army Alpha Grades 12 and 13 College Freshmen
9,
(
:
).
Values of Successive Point Intervals of the Original
5
-
-
Equal Uni............... 230
in
Scores
,
TABLE 58 Army Alpha Grade ............. 232
26
5
:
:
-
TABLE 59 Army Alpha Grade 232
-
281
n
=
6
..
:
:
TABLE 60 Army Alpha Grade 321 233
=
n
7
:
.
TABLE Army Alpha Grades and Values Successive
-
of
61
7.
6
5,
:
Point Intervals of the Original Scores Equal
in
5
-
,
Units 234
TABLE Army Alpha Grades 12 and 13. Supplemental Values

-
62
:
of Successive Point Intervals of the Original Scores

5
,
-
Equal Units............. 235

in
TABLE Final of of Army Alpha

- -
Estimate
63
Relative Values Scores

Equal Unit ........... 236
in
TABLE 64 Equivalents for Army Alpha Scores from 20 170
to
in
a
Scale with Equal Units This Scale Equalling
of
1
,
89/90 of the Difference between 60 and 150 of the

Original Alpha Scores Approximately 1/100 of the
or
,
Difference between and 150 the Original Alpha

50
of
Scores 237
TABLE 64a Provisional Equivalents for Army Alpha Scores from

-- --- - -
170 to 209 Scale in Table 64 238

as
;
TABLE 65 Army Alpha Distribution of Scores of 216 College

Graduates 239
TABLE National 240

66
Grades and
A A
1668 494
=
=
n
n
6
9
:
).
TABLE 67 National Grades 1679 and 482 241

=
=
n
n
7
8
(
)
:
).
National Summary of Determinations of

in
TABLE 68 Values
A
:
Equal Units …………………… 242

...
TABLE 69 National Grades 1677 and 243

A
2487
=
=
4
n
n
5
(
:
).
TABLE Equivalents for National Scores from

70
20
in
to
170
A
a
,
Scale with Equal Units Approximately 1/50 of

=
1
.
the Difference between 100 and 150 of the Original
-
Scale 244
TABLE 71 Otis Advanced Distributions 246

..
:
TABLE Otis Advanced Equivalents for Each 10 Point Interval

-
72
-
:
of the Original Scale Equal Units.... 247

in
LIST OF TABLES xix
TABLE 73 - Equivalents for Otis Advanced Scores from 30 to 200 in

PAGE
a Scale with Equal Units . 1 = 1 / 120 of the Difference
-
between 50 and 170 of the Original Scores 248
...
TABLE 73a Provisional Values for Otis Advanced Scores from 10 to
29 249
TABLE 74 Haggerty Delta Distributions 249
2
:
.
TABLE 75 Haggerty Delta Values Equal Units 250
- --
in
...
2
:
:
TABLE 76 Equivalents for Haggerty Delta Scores from 50
to
160
,
in Scale with Equal Units 251
a
..
TABLE 77 Terman Group Tests 252
..
TABLE 78 Terman Group Tests Sample of the Six Sets of

(
)
.
Values Equal Units Whence the General Transmu-

in
tation Table Is Derived... 253
TABLE 79 Terman Group Test of Mental Ability Values Equal

-
in
:
Units of Each Point on the Original Scale from 35

to 193 255
TABLE 80 Myers Mental Measure Grade and Grade

-
724
=
n
6
(
)
.
Values of Intervals Terms of Equal

in
311
=
n
9
(
)
.
Units Expressed Multiples 1/45 Difference

... as
of
x
,
-
between 36 and 81 256
)
TABLE Equivalents Scores from 21 for Myers Mental

to
of
81
86
Measure In Equal Units 256
....
:
TABLE Pintner Non Language Mental Test Original

- -
82 Scores
-
and Values of Intervals Equal Units 257

....
in
TABLE 83 Equivalents for Pintner Non Language Scores from 100

-
Scale with Equal Units refer

in
380 and
to
C
a
,
to the original scores and the scores transmuted into
-
scale with equal units..... 258
a
TABLE 84 I.E.R Tests of Selective and Relational Thinking Gen-

,
.
eralization and Organization Distributions Grade

in
-
:
and Grade 12 259

9
..
TABLE I.E.R Tests Selective and Relational Thinking Gen-

of
85
,
,
.
eralization and Organization Values of Intervals of

.
-
Original Scale Equal Units 260
in
..
TABLE 86 Transformation Table I.E.R. Tests of Selective and

.
Relational Thinking Generalization and Organization 261

,
TABLE 87 Brown University Psychological Examination Grades

-
12 and 13. N 3333 +2118 262

...
=
XX LIST OF TABLES
-
PAGE
TABLE 88 Equivalents of Scores from 20 to 80 for the Brown Uni-
versity Psychological Examination , in Equal Units...... 262
TABLE 89 - Army Examination a : Distributions of Pupils in Grades

5 , 6, 7 , 8 , 9 and 13.
-
4, 263
TABLE 90 Army Examination a : Equivalents for Each 10 - Point

Interval of the Original Scale in Equal Units . Re-
sults from Grades 6 , 7 and ............. 264
TABLE 91 - Army Examination a : Equivalents for Certain 10 - Point

Intervals of the Original Scale in Equal Units . Re-
sults from Grades 4 and 5 .. 265
TABLE 92 - Army Examination a : Equivalents in Equal Units . Re-

sults from Grades 6 , 7 and 8 ; 4 and 5 ; 13 ; and Com-
posite from All... 266
- Table for Army Examination a . 267
-
TABLE 93 Transmutation
TABLE 94 Equivalents for Army Examination a : Scores from 10

to 360 in a Scale with Equal Units . 1-1 / 80 of the
difference between 130 and 210 of the original scale...... 268
TABLE 95 - National Intelligence Examination

for White Pupils , Age
: Distribution of
273
-
Scores 11 .
TABLE 96 National Intelligence Examination : Distribution of

for White Pupils ,
-
Scores Age 12 .. ......... 275
TABLE 97 National Intelligence Examination A : Distribution of

Scores for White Pupils , Age 13 .. 278
TABLE 98 - National A : Data for Surface of Frequency in Equal
-
Units 279
-
TABLE 99 Otis Advanced : Distribution of Scores : Ages 11 and 12 282
...
-
TABLE 100 Otis Advanced Distribution of Scores Ages 13 and 14 284
...
:
:
TABLE 101 Otis Advanced Data by Which the Surfaces of Fre-

:
quency Are Constructed 285

..
Haggerty Delta Distribution of Scores 286

--
TABLE 102
2
.
:
:
TABLE 103 Haggerty Delta Data for Surface of Frequency

in
2
:
Equal Units 288

...
TABLE 104 The Effectof Correlation between Status and Gain When
Gain Increases in Geometric Rati 289
-
a
TABLE 105 The Effect of Correlation between Status and Gain When
AS 290
B
+
G
=
LIST OF TABLES xxi
-
PAGE
TABLE 106 Percents of Various Groups Succeeding with 20 or More
.............. 295 I .
-
Single Tasks of CAVD 40 - Composites to Q
TABLE 107 The Difficulty of Composites I

to Q in Various Groups
Expressed As a Deviation from the Difficulty for the
Median of That Group , in Terms of the σ of that
Group in the Ability Measured by Success with the
Composite in Question . - Is Easier , + Is Hard............... 296
TABLE 108 - rt₁t2 As Estimated

of Single Tasks
from Correlations between Number
Correct in One Half of a 40 - Com-
posite and Number of Single Tasks Correct in the
Other Half ; and Also As Estimated from Correlations
between Number Correct in a 40 - Composite and Num-
-
ber Correct in a Neighboring 40 - Composit ........... 300
TABLE 109 Values of rt , t₂ Derived from Table 108 , and the Values
of Vrt , t₂ Used to Obtain Table 110 from Table 107.....
-
301
TABLE 110 The Intellectual Difficulty of Composites to Q in I

Groups 52 , 9 I , 9 II , 13 and 17 Expressed in Terms
of 15 , 1919 19119 13 or o117 ; As Derived by the
Use of Table 302
-
109 ..
TABLE 111 Values of rt , Estimated from Correlations between

Number of Single Tasks Correct in a 40 - Composite
-
and Number Correct in a Long CAVD Series ..............303
...
TABLE 112 The Intellectual Difficulty of Composites Terms

to
of in
Q
I
As Derived by the Use Table

of
151 191 etc.

,
,
,
111 304
The Intellectual Difficulty of Composites Aver-

-
TABLE 113
to
Q.
I
ages of the Determinations of Table 110 and Table
-
112 305
TABLE 114 Data for Computing Relative Variabilities of Different
....
Grades in Intellect and for Computing Distances
;
between Medians of Different Grades in Intellect 307
TABLE 115 The Relative Variability of Different Grade Populations 311

...
TABLE 116 The Intellectual Difficulty of Composite Tasks in

to
-
Q
I
-
Terms of 316
19
TABLE 117 Difference between Grades in Scores Attained in Various
-
Intelligence Examinations 317
...
TABLE 118 The Intellectual Difficulty of Tasks Expressed

to
in
Q
I
Each Case As Difference from the Median Difficulty

a
for Group Units 321

of
in
19
9,
xxii LIST OF TABLES
TABLE 119 - Percents Succeeding with Various Composites in Groups

PAGE
im 3 , im 6 , f, sp , 4 , 5 and ad 323
...
TABLE 120 The Difficulty Composites
to Various Groups
K
of
in
A
,
Expressed As Deviation from the Difficulty for the
a
Median of That Group Terms of the of That
in
σ
,
Group the Ability Measured by Success with the
in
Composite in Question 324
TABLE 121 Raw Intercorrelations of Composites and

-
E
F
G
D
C
,
,
in the Case of 100 Individuals Chronologically Sixteen
Over and Mentally Six 325
or
...
,
The Difficulty Composites to
in
TABLE 122 Terms
K
of
of
- -
1m3
A
89 etc. 327
ig
ime
,
,
TABLE 123 The Difficulty Composites Terms
to
K
of
of
in
…………………
330
.
The Intellectual Difficulty Tasks Expressed
of
TABLE 124
—
K
A
to
As Difference from the Median Difficulty for Group
in a
Units of 335
19
9
The Estimated Differences Difficulty

-
TABLE 125 Intellectual
of
in
Tasks from Zero Difficulty the Difficulty of Com

to
posite ......... 350
The Number Psychologists out of Judging

of
TABLE
-
126 Cer
40
a
(
tain Task To Be More Difficult Intellectually than
a
Certain Other and the Number Judging the Two Tasks
,
To Be Equally Difficult The table reads Task

to 36
"
:
.
was judged harder than Task 37 by 13 and equal

difficulty by Task 36 was judged harder than
in
it
7,
Task 38 by and equal difficulty by Task

to
in
it
2
3,
36 was judged harder than Task 27 by 10 and equal

difficulty by and on 362
in
to
so
it
...
"
2,
TABLE Median of the Scores Number Right out for

-
127
of
40
(
Each of Twenty Nine Groups with Each of Four

or
-
-
More Neighboring Composites 370

...
TABLE 128 Altitudes Corresponding Any Number Correct from

to
35 out of 40 for Tasks 371

to
to
Q
5
I
..
TABLE 129 Approximate Provisional Altitudes Corresponding Any

-
to
Number Correct from to 35 out of 40 for Tasks

5
-
A to H
... 372
TABLE 130 Correlations Raw and Corrected for Attenuation between

,
Rate and Level After Hunsicker '25 Table 401

V
,
,
(
)
.
TABLE for Attenuation of Sen

-
131 Intercorrelations Corrected

(
tence Completion Co Vocabulary Arith

V
(
),
,
(
)
LIST OF TABLES xxiii
PAGE
metical Control ( Ac ) , Arithmetical Association ( Aa ) ,
Analogies ( An ) , and Information ( Inf ) in 250
Pupils of Grade 82. ( Compiled from Tables of
-
Tilton , 25 , in press . ) 426
TABLE 132 of Four Tests of the Higher and

The Intercorrelation
Two Tests of Associative Thinking . 100 University
Students . P by Pearson formula , S by Sheppard
-
formula 428
TABLE 133 The Intercorrelations of Three Tests of the Higher and

Two Tests of the Lower or Associative Thinking . 126
Pupils in Grade 52. The correlations are all raw
correlations by the Sheppard formula 429
...
TABLE 134 Observed and Partial Correlations between Age Intelli-
,
Table XX gence School Attainments and the Results of the
(
,
,
of Burt Binet Simon Tests 452

...
)
TABLE 135 The Intercorrelations of One Binet Test One Burt

-
B
,
(
)
Reasoning Test One Measure of School Work
I
,
(
)
and Age by Certain Assumptions Concern-

A
S
ing the Intercorrelations If an Infinite Number

(
(
),
of
-
Such Tests Had Been Used 455
.
TABLE 136 Observed and Partial Correlations between the Binet-

Table XXI Simon Tests and Attainments in the Several School
(
of Burt Subjects 457

)
TABLE 137 — The Average Scores the Original Units and Equal
in
in
Units in Various Intelligence Examinations Various

at
-
Ages and the Differences between Successive Years.......465
;
TABLE 138 Variations of the Scores of Thirteen Fewer 30-

or
)
(
Minute Trials with Part of the Thorndike Intelli-

I
gence Examination for High School Graduates from

the Median Score for the Individual in All Thirteen
Trials 20 Gifted Pupils 13 ...
.
u
,
a,
c,
d,
e,
.
.
.
Different Days 498
TABLE 139 The Relation between an Individual's Ability and His

Variability The variability that of one trial 30
is
(
.
minutes of the Thorndike Test Part from the

)
I,
average of an infinite number of such trials The two

.
trials were taken on the same day 501

.
TABLE 140 Thorndike Intelligence Examination for High School

-
Graduates '19 -23 Series Part Trial Arrayed

= 2
,
I,
,
.
under Trial Test of Feb. '22 30 30 to 34

1.
,
.
35 35 to 39 etc. 507
=
,
xxiv LIST OF TABLES
-
PAGE
TABLE 141 Thorndike Intelligence Examination for High School
Graduates , '19 -23 Series . Part I
, Trial 2 , Arrayed
under Thorndike Examination , Total Score . Women

Students in High School , Normal School , College and
-
University . 30 = 30-34 ; 35 = 35-39 , etc. 508
TABLE 142A Thorndike Intelligence Examination for High School

Graduates . Part I. Average Difference between
Two Trials ( Single Session ) in Relation to the Average
Sum of Trial 1 and Trial 21
Score 12 ) 509
-
2
TABLE 142B Same as Table 142A , except that the Difference is be-
tween Trials on Different Days and that the Average
Score Is from Four Trials . Normal School Students...... 510
TABLE 143 - Thorndike Intelligence Examination for High School

Graduates . Part I. Variability of Score in One Trial
Arrayed under Score in Another Trial or under Total
Score in the Entire Examination . 10 = 10 to 14 ;
15 = 15 to 19 , etc. 511
TABLE 144 The Summaries of Table 142B and Table 143 , with
-
Coarser Grouping 513

Amount of Ability in Fifteen Tests or Amalgamations
of Tests . The upper number is the measure of varia-
bility ; the lower number ( in italics ) is the weight
-
attached to it . 516
of
-
TABLE 146 The Effect the Selection of Task............. 518

Amount of Ability in Fifteen Tests or Amalgamations
-
of Tests , Using Eight Levels of Ability .. 519
-
TABLE 148 The Closeness of Fit of Six Test Scores , Taken Singly........ 531
TABLE 149 The Closeness of Fit of Six Test Scores , Taken Two or
- - -
More at a Time and Averaged .. 533
TABLE 150 Goodness of Fit of Observed Distributions Groups 1 to
-
6 to Normal Curve 552
TABLE 151 Goodness of Fit of Composite Distributions to Normal
-
Curve 552
TABLE 152 Self- and Inter - Correlations of Four 40 - Composites of

CAVD , Each Divided into Two Random Halves ( and I
II). 98 Imbeciles . ( P means Pearson Coefficient ,
S means Sheppard Coefficient . ) 557
LIST OF TABLES XXV
PAGE
TABLE 153 — The Inter - Correlations of Four CAVD Composite Tasks
Like A , B , C , and D in Constitution and Difficulty , but
Each Consisting of an Infinitely Large Number of
Single Tasks . The Inter - Correlations of Table 152
Corrected for Attenuation 558
TABLE 154 Self- and Inter - Correlations of Four 40 - Composites of

CAVD , Each Divided into Two Random Halves ( I
and II
). 121 High School Graduates . ( P means
-
Pearson , Sh means Sheppard .) . 558
TABLE 155 of Four CAVD Composite Tasks

The Inter - Correlations
Like NY6a42 , OZ , 75 , PZ , 86 , and 97 , in Con- Q.
struction and Difficulty , but Each Consisting of an
Infinitely Large Number of Single Tasks . 121 High
-
School Graduates 559
TABLE 156 The "Raw" Inter - Correlations of Five CAVD 40 - Com-
-
posites in 246 Pupils of Grade 9. 560
TABLE 157 The Inter - Correlations of Five CAVD Composite Tasks ,

Like Those of Table 156 in Constitution and Difficulty ,
but Each Consisting of an Infinitely Large Number
of Single Tasks 560
...
TABLE 158 The Raw Inter Correlations of Four CAVD Compos-

- -
"
"
Pupils 561
in
ites Grade
of
192
...
9
TABLE 159 The Inter Correlations of Four CAVD Composites Like

-
Those of Table 158 Constitution and Difficulty but

in
Each Consisting of an Infinitely Large Number of

Single Tasks 561
TABLE 160 Summary of the Inter Correlations Corrected for At-

-
-
-
tenuation 562
TABLE 161 The Correlations between the Number of Single Tasks

Responded Correctly Various 40 Composites and
in
to
to -
the Number of Tasks Responded Correctly

in
a
Long Series of CAVD Tasks Ranging from Tasks

Very Easy for the Group Question Tasks Very
to
in
Hard for the Group Question Groups 9I and

II
in
566
9
TABLE 162 The Correlations of Single Tasks

-
between the Number

Responded Correctly Various 40 Composites and
in
to
to -
Correctly
Responded
of
the Number Tasks

in
a
Long Series of CAVD Tasks Ranging from Tasks

Very Easy for the Group Question Tasks Very
to
in
Hard for the Group Question Groups 13 and 17...... 568

in
:
xxvi LIST OF TABLES
TABLE 163 - The Assumed Magnitude of the Error Whereby a Stan

PAGE
ford Binet Mental Age Differs from the Mental Age

Which Would Be Found by a Perfect Measurement of
Altitude of Intellect 578
TABLE 164 Form of Distribution Assumed in Obtaining Measures of

the Difficulty of Various Composites for the Groups
-
of 50 Feeble - Minded 580
...
TABLE 165 The Probable Form of Distribution of Altitude of Intel
lect in Group 17 Law Students 597
...
(
)
TABLE 166 — Data for Estimating the Form of Distribution of Alti
tude of Intellect the Group Ad 44 Recruits 599
in
....
(
.
)
CHAPTER I
THE PRESENT STATUS¹
Existing instruments for measuring intellect² developed
from three roots , the interview , the school examination , and
the ' tests ' of sensory acuity, memory , attention , and the
like , devised during the early history of psychology . The
Stanford Binet , for example , is an improved , systematized
and standardized interview . The Army Alpha is in part an
improved school examination and in part an improved bat
tery of tests like those used before 1900 by Galton , Ebbing
haus , Cattell , Jastrow , and others .
Existing instruments represent enormous improvements
over what was available twenty years ago , but three funda
mental defects remain . Just what they measure is not
known ; how far it is proper to add , subtract , multiply ,
divide , and compute ratios with the measures obtained is not
known ; just what the measures obtained signify concerning
intellect is not known . We may refer to these defects in
order as ambiguity in content , arbitrariness in units , and
ambiguity in significance .
AMBIGUITY IN CONTENT
If we examine any of the best existing instruments , say

the Stanford Binet , the Army Alpha or the National Intel
ligence Test , we find a series of varied tasks . Some concern
words , some concern numbers , some concern space relations ,
some concern pictures , some concern facts of home life .
Some seem merely informational ; some are puzzle - like .
Some concern mental activities which will be entirely famil
iar to almost all of the individuals to be tested ; some con
1 This chapter is reprinted with some alterations from the Psychological
Review , Vol . 31 , pp . 219 to 252 .
2 We shall use ' intellect ' and ' intelligence ' as synonyms throughout this
book .
1
2 THE MEASUREMENT OF INTELLIGENCE
cern novelties . Some are irrespective of speed ; in some

speed is a large element in success . In particular , as we
shall see later , the score attained is a composite in variable
proportions whereby A is rated as more intelligent than B
-first , if he can do certain hard tasks with which B fails ,
second , if he can do a greater number than B can of tasks
of equal difficulty , and third , if he can do more rapidly than
B tasks at which both succeed . The only sure statement of
what abilities the Army Alpha measures is to show the test
itself and its scoring plan.
To this it may be retorted that this variety is not really
an ambiguity , that one of these tests is a representative
sampling of tasks for intellect , and that the scoring plan is
one which weights each response according to its importance
as a symptom of intellect . Unfortunately this is not true . We
may cherish the hope that these tests approximate to such
representativeness of sampling and suitability of weights .
In fact , however , nobody has ever made an inventory of
tasks , determined the correlation of each with intellect ,
selected an adequate battery of them , and found the proper
weight to attach to each of these . Such a procedure was
carried out in part by the Committee responsible for the
construction of the National Intelligence Test , but limita
tions of time and funds restricted it to a very small fraction
of what would be adequate . If anybody did this wisely , a
large fraction of his labor would be precisely to find out
what abilities our best present instruments did measure ,
and how these abilities were related to intellect ; or to find
out what abilities constituted intellect , and how these abili
ties were measured by our present instruments.3
One of the main lines of work in the improvement of
instruments for measuring intellect is then to find out what
abilities our best present instruments do measure .
3 The balance of his labor might be expended upon experimentation with
tasks that seemed promising as symptoms , even though we did not know what
abilities they required .
THE PRESENT STATUS 3
ARBITRARINESS OF UNITS
The score obtained by using the instrument to measure

an intellect is in present practice either a number represen
ting a summation of credits and penalties or , more rarely ,
a number representing the grade of difficulty of the tasks
which the person can respond to with some assigned per
centage of correct responses . Thus in Army Alpha he may
score by summation from 0 to 212 ; in the first suggestion of
Binet he could score 5 or 6 , or 7 , or 8 , or 9 , according as he
was able to do correctly all but one of the tasks set as 5 - year
tasks , 6- year tasks , 7 -year tasks and so on .*
In neither case ( even supposing the measurement to be
a perfect representation of the person's abilities ) can the
numbers be taken at their face value . If A scores 50 on
Alpha , B , 75 , and C , 100 , we do not know that the difference
between A and B in the abilities tested by Alpha is the same
as the difference between B and C , nor that C has twice as
much of these abilities as A. If
D scores mental age 4, E
mental age 6 and F mental
, age 8 by the Binet , we do not
know that in the abilities tested
, by the Binet , F excels E as
much as E excels D , or that F has one and one third times as
much of these abilities as E has . The numbers , 1 , 2 , 3 , 4 , etc. ,
designating the scores made by individuals , do not represent
a series of amounts of intellect progressing by equal steps .
The difference in intellect between Army Alpha 10 and
Army Alpha 20 may indeed conceivably be as great as the
difference between Alpha 100 and Alpha 150. From Stan
ford Binet 40 months to 60 months may be as great a
difference in intellect as from 140 months to 180 months .
The value of what is called a difference of 1 on the scale is
not known , and its value may fluctuate greatly as we move
along the scale .
This suggestion was , however , abandoned in favor of a procedure which
mixes two sorts of measure . The procedure is , " Take for point of departure
the age at which all tests are passed ; and beyond this age count as many fifths
of a year as there are tests passed . " [ ' The Development of Intelligence , ' Eng .
trans . of Kite , 1916 , p . 278. ]
We have then no right to add , subtract , multiply , or

divide with these scores of A , B , C , D , E , and F in the way
that we do with their heights or weights . Suppose that A
scores 100 ; B , 110 ; C , 90 ; and D , 120. We cannot say that
the average intellect of A and B equals the average intellect
of C and D. If E changes from 60 to 70 , while F changes
from 70 to 80 , we cannot say that they have made equal
gross gains .
The numbers designating the scores made by individuals
are usually not even approximately related to any true zero
point . Consequently , even if the scores 1 , 2 , 3 , 4, did
represent an equal - interval series of amounts or degrees of
the ability in question , they would properly be treated as
x + 1 , x + 2 , x + 3 , x + 4. The ' times as ' or ratio judg
ment is thus not surely applicable and the relations of the
scores to anything else are thus undetermined . For ex
ample , we cannot say whether the intellect of the average
twelve -year - old is one and a quarter times that of the aver
age six - year - old or twice it , or ten times it .
The second main problem in improving measurements of
intellect is thus to attach fuller and more definite meanings
to these credit summations and difficulty levels , and if pos
sible to find their equivalents on absolute scales on which
zero will represent just not any of the ability in question ,
and 1 , 2 , 3 , 4 and so on will represent amounts increasing
by a constant difference .
We have to estimate equivalents of this sort somehow
before we can make much use of ratings by either credit
summations or difficulty levels ; before , for example , we can
conveniently compare individuals or groups , or the changes
made by individuals or by groups , or the effects of different
environments . The commonest method at present is to take
as the equivalent for any score by any instrument , the age
whose average achievement is that score , and to assume that
5 Attempts have been made to define ' zero ' or ' just not any ' ability and to
assign scores in relation to zero in the case of knowledge of English words ,
ability to understand sentences , handwriting , drawing , and English composition .
the increments in average ability are equal for equal differ

ences in age up to some limit such as 192 months , and are
zero thereafter . This of course is purely hypothetical in
general and is almost certainly in error for the ages near
the point where the age change suddenly turns from its
full amount to zero . The curve of ability in relation to age
is almost always smooth as in the continuous line of Fig.
FIG . la. The probable form of the curve of intellect in relation to age.
FIG . 1b. The form of the curve of intellect in relation to age , if annual gains
are equal up to some stated age , and are zero thereafter .
1a , but notwith a sharp turn as in the dash - line of Fig . 1b .

The competent thinkers who use the method know this and
are cautious in inferences based upon its application to the
higher ages ; but they use it rather freely for the lower ages ,
because some method must be used , because it is easy to

understand and apply , and because we do not know what
method is really right .
It may be objected that equality of units is an unneces
sary refinement , for present practical purposes , since the
mental age defines the status of an individual sufficiently ,
' as able as the average ten - year- old , ' ' as able as the average
twelve - year- old . ' These , it may be said , are better measure
ments for practical purposes than some absolute scale in
terms of equal ' mentaces ' or ' intels . ' The convenience , in
telligibility and realism of the mental age scale up to about
12 or 13 years are indeed great advantages , but after 13 or
14 it is neither convenient nor intelligible nor realistic . It
is not convenient because the computation of intelligence
quotients becomes very troublesome for the higher ages . It
is not readily intelligible because mental ages 14 , 15 , 16 , etc. ,
are not ' as good as the average ' 14 -year old , 15 - year old , etc.
The average 25 - year old for example is about the mental age
of 14 by one of the best instruments . It is not realistic
because we have no clear or vivid sense of what the average
person is intellectually at fifteen , or at sixteen , and do not
even know whether he improves in the next two or three
years . A mental age of 15 or 16 or 17 is in fact as arbitrary
a quantity as an Alpha ability of 123 .
A rarer but more promising procedure than that of trans
forming test scores into ' ages ' is to transform them into
units of ability on the assumption that the distribution of
ability in all adults 21-30 , or in all twelve - year - olds , or in all
pupils in grade six of a certain city , or in some other speci
fied group , is approximately that given by
y - 1 -x2
€202
σν2π
For example , the Alpha scores from 0 to 212 were not
used in the army at their face value , which would give a
distribution of the form shown in Fig . 2 , but were trans
muted into seven letter measures by the following scheme ,
which assumed an approximately ' normal ' distribution for

a random sampling of 128,747 of the literate white draft :
0 100 200
-=
FIG . 2. The form of distribution of the literate white draft if Army Alpha
are taken at their face value .
-
135-212 A
105-134 B
75-104 C +
=C
44C
45-74
25
15- 24
0-14 =
-D D
――――――――
The score used in the Thorndike - McCall test of para

graph reading is not the number of correct answers , but a
transmutation on the assumption that the real ability con
cerned is distributed ' normally ' amongst twelve - year - olds
in American cities .
We know very little concerning the permissibility of the
assumption of the so - called normal distribution for adults
or for an age , or for a school grade . The search for evi
dence pro and con is one important feature of the attempt
to obtain units of mental ability which shall be at least ap
proximately equal .
AMBIGUITY IN SIGNIFICANCE
The test score measures directly only the measurer's
impression from the subject's performance , or the summa
tion in a more or less capricious fashion , of credits and
penalties for the subject's responses to the different ele
ments of the tests , or a combination of these . What this
3
score signifies about the subject's intellect depends upon

the intuition of the measurer , or upon the correlation be
tween the summation and intellect , or upon both . When we
assert that a child is found by measurement with the Stan
ford Binet to have the intellect of a child 10½ years , all that
is really asserted is that the child does as well in that par
ticular standardized interview as did the average of the
children of 10½ years of age tested by Terman in making
his standards . We do not know what the average intellect
of these children was , nor how closely the Stanford Binet
score represents or parallels or signifies it .
When we assert that a man is found by measurement
with the Army Alpha to have the intellect of an average
recruit in the draft , all that is really asserted is that he does
as well in that particular battery of tests scored and sum
mated in a particular way , as the average recruit did . Just
what the intellects of recruits were and how closely their
Alpha scores paralleled their intellects , we do not know.
The measurement is one thing , the inference to intellect is a
different thing .
This is of course true of many measurements . The
amount of silver deposited in one second by an electric cur
rent is not the amount of current . The dividend rate on
stock during any one year is not the worth of the stock . The
amount of silver is , under proper conditions , of perfect sig
nificance as an indicator of the amount of current , since the
correlation between it and a perfect criterion of amount of
current is perfect . The dividend rate is of very imperfect
significance , since the correlation between it and a perfect
criterion of the worth of the stock is far from perfect .
We do not know how closely the rating or score in the
Stanford Binet or the Army Alpha or any other instrument
correlates with a perfect criterion of intellect , because we do
not know what such a criterion is , much less its correlations
with these tests . One great task of the measurement of
intellect is to obtain such a criterion , or a closer approxima
tion to it than we now have , and to use it to improve the
selection and weighting of the elements of our testing in

struments .
The present status of such instruments as the Binet or
Army or National tests is roughly as follows : We have
chosen tests where the judgment of sensible people in gen
eral is that correct response or speed of correct response is
characteristic of intellect . Such is the case with directions
tests , arithmetical problems , common sense questions ( as
in Alpha 3 ) , and the like . We have chosen tests using the
judgment of psychologists in the same way. Such is the
case with the completion tests devised by Ebbinghaus , the
mixed relations or analogies test devised by Woodworth ,
and the like . We have tried these or other tests with chil
dren secluded in institutions because of imputed intellec
tual inferiority and with children of like age who are in
ordinary schools ( as by Norsworthy ) , with adult males of
good health and morals who were found in a Salvation
Army home , glad to work for a dollar a day , and with
adults of the professional classes ( as by Simpson ) , with
children in general of different ages ( as by Binet and
Terman ) , with various groups of children ranked for im
puted intelligence by teachers , fellow pupils , school ad
vancement , and other symptoms ( as by Spearman , Burt ,
Terman , Whipple , Yerkes , and others ) , with children of
alleged superior intelligence in comparison with others ( as
by Whipple and Terman ) , with soldiers in the National
Guard and regular army in connection with ratings for
intelligence given by their officers ( as by the Psychology
Committee of the National Research Council ) , with stu
dents whose success in high school and college studies was
also measured ( as by Colvin , Wood , and many others ) ,
with individuals who were tested with a very long series
of tests ( as by Terman and Chamberlain , Stenquist , and
others ) , and in other ways .
As a general result we know that certain systematized
interviews and batteries of tests measure somewhat the
same trait , since they correlate somewhat one with an
other ; and that this trait has to some extent the same con
stitution as the trait which sensible people , psychologists ,
and teachers rate as intellect .
The failure of perfect correlation between the amount
of intellect a person has , as revealed by the criterion , and
the amount indicated by the instrument is due , as has been
said , partly to the imperfection of the criterion , but partly
also to the imperfection of the instruments . They ( at
least all but one of them ) are demonstrably imperfect , since
no two of them correspond perfectly in their findings for
the same intellects . Since it is extremely unlikely that ,
out of a dozen instruments devised with about equal care
by a dozen individuals or committees at about the same
date one should be very much superior to all the others ,
we may assume , until there appears proof to the contrary ,
that all are imperfect .
The imperfection may be of two sorts . First , the re
sponses measured by the instrument may not be represen
tative of the whole intellect and nothing but intellect ; the
score obtained may not give enough weight to certain fac
tors or elements of intellect and may give weight to others
which really deserve less or even zero weight . The instru
ment is then like a wattmeter which gives only half weight
to the voltage of the current or adds two watts for every
time that the current is turned on or turned off . Second ,
the same person may receive a different score when re
measured by the instrument . In so far as such differences
are due to the ' accidental ' ups and downs in the person's
achievements , they are taken care of by measuring him at
enough different times ; but in so far as they are due to ac
quaintance with the instrument itself or with instruments
like it , they are a very serious imperfection . For example ,
a given score with Army Alpha represents a very different
status according as it is from a first , a second , or a third
trial . The case here is as if a thermometer tended after
subjection to a temperature of 200 ° once to register 220 °
when 200 ° was next encountered . The provision of means
for distinguishing between that part of the score due to

certain general characteristics of the person measured and
that part of the score which is due to certain special train
ing that he has had with the tasks of the tests , or with
tasks like them , is thus an important part of the work of
making the measurements more fully and exactly signifi
cant of intellect .
Ingeneral , all our measurements assume that the indi
vidual in question tries as hard as he can to make as high a
score as possible . None of them can guarantee that the
scores would correspond at all with a perfect criterion if
the individuals measured tried to appear as dull as they
could . The correlation would indeed then probably be in
verse , the more intelligent persons being more successful
in their efforts to appear dull ! It is theoretically possible
to arrange a system of incentives such that each person
measured by an instrument would put forth approximately
his maximum effort , and in scientific testing of the instru
ments this can often be done . In general practice , how
ever , we rarely know the relation of any person's effort to
his possible maximum effort . Since , however , the disturb
ances due to differences in effort on the part of those tested
require in study and treatment procedures which have little
or nothing to do with the procedures by which the instru
ments are made to give better measurements of those who
do try their best , we shall disregard the former and shall
limit our inquiry to the latter sort of procedures .
MEASUREMENTS OF INTELLIGENCE ARE MEASURES OF

INTELLECTUAL PRODUCTS
All scientific measurements of intelligence that we have
at present are measures of some product produced by the
person or animal in question , or of the way in which some
product is produced . A is rated as more intelligent than B
because he produces a better product , essay written , answer
found , choice made , completion supplied or the like , or pro
duces an equally good product in a better way , more quickly
or by inference rather than by rote memory , or by more in

genious use of the material at hand .
We can conceive of states of affairs such that a man's A
intellect could be measured without consideration of the

products he produces or the ways in which he produces
them . Intellect might be exactly proportionate to the ac
tivity of the thyroid gland , or to the proportion of the brain
weight to body weight, or to the number of associative neu
rones in the frontal lobes or to the complexity of the fibril
lary action of certain neurones , or to the intensity of a
certain chemical process , and hence be measurable by ob
servations of the thyroid's action , or estimates of the
brain's volume , or by a count or measurement of neurones ,
or by a chemical analysis .
Psychologists would of course assume that differences
in intelligence are due to differences histological or physio
logical , or both , and would expect these physical bases of
intelligence to be measurable . At present , however , we
know so little of the neural correlates of intellect that if
twenty college freshmen were immolated to this inquiry ,
ten being the most intellectual of a hundred , and ten
being the least intellectual of the hundred , and their
brains were studied in every way by our best neurologists ,
these could probably not locate sixteen out of the twenty
correctly as at top or bottom . Moreover , what we do know
of neural correlates is of little avail during life , the living
neurones being extremely inaccessible to present methods
of observation .
Even if one aimed at discovering the physiological basis
of intellect and measuring it in physiological units , one
would have to begin by measuring the intellectual products
produced by it . For our only means of discovering physio
logical bases is search for the physiological factors which
correspond to intellectual production .
MEASUREMENTS OF INTELLIGENCE IMPLY VALUATION

Our present measurements of intelligence rest on human
judgments of value , judgments that product A is ' better '
or ' truer ' or ' more correct ' than product B , that method C
is ' preferable ' to method D , or that C is ' right ' while D is
' wrong , ' and the like .
In some cases this is so clear that everyone must admit
it. Thus in three of our best tests of intelligence , giving
the opposites of words , completing sentences by supplying
omitted words , and answering questions about a paragraph
read , we make elaborate keys assigning credits to the dif
ferent responses . These keys are obviously made by
human judgments of the value of each response .
The credits given may represent valuations by the truth
fulness or wisdom of the answers or sentences , by their
grammatical form , by their rhetorical excellence , by their
originality , by the rate of producing them , or by a subtle
sense of their significance as evidence of intelligence .
6 For example , the task being to complete ,
' God made . . . and . . . let him pass for a man , ' we find among the
responses of high -school graduates :
him therefore
him SO
him then
him will
him they
I
him he
him
him let
man always
man then
man God
man has
man he
man therefore
man please
Adam then
Adam Eve
Adam he
animal wouldn't
Eve God
us we
heaven earth
and must assign some value to each , or make a dividing line between full value
and no value somewhere .
In some cases the value is assigned so easily ( as a simple

deduction from , or following of , a general rule ) that we may
thoughtlessly assume that the response indicates intelli
gence regardless of any process of valuation . For ex
ample , we may consider that in a test in arithmetical com
putation or problem solving , the right answers are signs of
intelligence , regardless of what anybody thinks . A little
thought will convince us , however , that in such tests the
human judgment acts as truly as in a completion or para
graph - reading test. The main difference is that , having
once for all decided that right answers are better than
wrong answers , we do not raise the issue about any par
ticular answer . We simply assume or make a general rule
of valuation . The valuation becomes obvious if we col
lect all the responses made to an arithmetical task and ask
whether all the different ' rights ' are equally good or right ,
and whether all the different ' wrongs ' are equally undesir
able ."
One criterion of value , truth , is so widely used in fram
ing , keying , and scoring tests of intelligence that it deserves
comment , especially since there may be in the case of truth
an objective criterion , power in prediction , by which our
judgments of value are or should be determined . Two
other criteria of value also need comment because they have
been suggested explicitly or implicitly as direct criteria
for intelligence . They are development with age and abil
ity to learn .
TRUTH
Probably over half of our present tests of intelligence
are tests where the response is given credit as a symptom
of intelligence in proportion to its truthfulness . Such is
the case , for example , with eight out of ten tests of the Otis
7 In the special case where we arrange for Yes and No answers valuation is
doubly active . We arrange so that a Yes or a No will be ' good ' as a response .
Then , since some of the correct ' Yeses ' or ' Noes ' may be due to chance , and
since chance answers are deemed of no value , we plan our scoring so as to give
the chance ' Yeses ' and ' Noes ' zero value .
Advanced ; and with Army Alpha , 2 , part of 3 , 4 , 6, 7, and 8 .

It is more or less the case with Stanford Binet III , 5 ; IV ,
1, 2, 3 ; V, 1 , 2 , 3, 4 ; VI , 1, 2, 3, 5 ; VII , 1, 2, 5 ; VIII , 4, 5, 6 ;
IX , ; X, 1 , 2 ; XII , 1 , 2 , 8 ; XIV , 1 , 2 , 3 , 5 , 6 ; and with
1, 2, 3
National Intelligence A , 1 , 3 , 4 , and B , 1 , 2 , 3 , and 5 .
One could make an attractive theory of intelligence and
its measurement somewhat as follows : Intellect is con
cerned with facts , being the ability to see and learn the
truth , to get true knowledge and use it to the best advan
tage . Truth is insight into the real world , the evidence
that knowledge is true is its predictive power . Measures
of intelligence are then ultimately measures of a man's
mastery of prediction , that 2 and 2 will be 4 , or that it will
be profitable to buy such and such a stock , or that a planet
will be found having such and such a path . More immedi
ately , they are measures of certain abilities which contrib
ute to , or accompany , or indicate the existence of, the abil
ity to get and use the truth .
By this theory we should rest our valuations of truth.
all on the ultimate test of power of prediction . One truth
would be better than another in proportion as it predicted
more facts , or more important facts , or predicted the same
facts more acccurately , or helped more in the acquisition of
other truths . Our valuations of abilities as evidences of
intellect would rest on their significance as symptoms of
ability to get and use truth .
It seems sure , however , that people in general , psychol
ogists , and framers of intelligence tests , alike mean by
intellect something more than ability in truth - getting to
improve prediction . They mean what Pericles and Wash
ington and Gladstone had as well as what Aristotle and
Pasteur and Darwin had . In the oral interview of the
business man or physician to test intelligence , in such tests
as Ebbinghaus ' completions , and in such a battery of tests
as Army Beta , there is little obvious reference to predic
tion or truth getting . In the first case , the aim is rather
to see how the person fits his thoughts and acts to little
problems or emergencies ; in the second , it was rather to

give him a chance to use all the so - called higher mental
powers ; in the third , many tasks were selected in which
people who were regarded as intelligent could do better
than people regarded as dull , and those of them which most
conveniently distinguished the alleged bright from the al
leged dull were kept as the final choice . If
these instru
ments do really measure ability at truth getting , it is only
indirectly and more by accident than by design .
It may be that truth -getting is what we unwittingly do
measure by our intelligence tests , or what we ought to try
to measure , but very few of those who devise or apply the
tests think so . And it is surely wise to find out what we
do measure before deciding that it is or ought to be truth
getting .
DEVELOPMENT WITH AGE
Binet had it in mind to discover those intellectual abili
ties which six - year - olds had that five -year - olds did not have ,
those which seven -year - olds had that five - year- olds and
six -year- olds did not have , and so on . It might seem that ,
except for the one judgment that abilities were ' better ' or
represented ' greater intelligence ' the later they came in
this series of normal chronological process , the Binet mea
surement would be free from valuation .
However , valuation came in from the start because Binet
tried only abilities which he valued as intellectual . He did
not take all the psychological features of five- , six- , and
seven -year - olds and choose as his series of tests those which
separated the ages most distinctly . In revising Binet's
series Terman and others have paid less and less attention
to lateness of development and more and more to signifi
cance as valued symptoms of intelligence in their choice of
tasks .
This is well . For if Binet or they had collected a series
of tasks such as showed the least overlapping of one chrono
logical age on the next , the resulting series would be in
ferior as a measure of intellect to the series as it stands .
For example , quality of handwriting , rate of tapping , and
ability in checking A's on a mixed sheet of capitals would

probably show less overlapping with age than vocabulary ,
rate of reading, and ability in completing sentences . But
they would be far less effective in diagnosing amount of
intelligence .
Development with age would be a poor and partial crite
rion for intellect of any sort or degree , and for the higher
ranges of intellect , say those above the 70 - percentile intel
lect of the average of the white draft , or above the average
ninth -grade pupil , it would be well - nigh worthless . It has
never been so used . The Terman mental ages above 14 ,
for example , are not functions of development with age , but
of differences between individuals , regardless of age .
ABILITY TO LEARN
An obvious hypothesis , often advanced , is that intellect

is the ability to learn , and that our estimates of it are or
should be estimates of ability to learn . To be able to learn
harder things or to be able to learn the same thing more
quickly would then be the single basis of valuation . Suc
cess in solving arithmetical problems , or defining words , or
completing sentences would then be good , simply and solely
because it signified that the person had greater ability to
learn .
If greater ability to learn means in part ability to learn
harder things , we have excluded the vague general valua
tion of certain products and ways of producing only to in
clude it again. For we shall find ourselves selecting or
defining A as harder to learn than B on the ground that
only the more intellectual persons can do it , or on the
ground that it requires a higher type of intellect , and shall
find ourselves using those vague general valuations to pick
the persons or describe the type of intellect required .
If greater ability to learn means only the ability to learn
more things or to learn the same things more quickly , we
have a view that has certain advantages of clearness and
approximate fitness to many facts . Even less than in the
case of truth - getting , however , do our present actual instru
ments for measuring intelligence measure directly a per

son's ability to learn more things than another person can ,
or to learn the same things more quickly . The substitution
test included in Army Beta , in the National Intelligence
Examination and in some others , is about the only test of
speed of learning that is used ; and it is more than a learn
ing test.
Much evidence will therefore be required before we can
wisely replace our present multifarious empirical valua
tions by the formula that intellect is the ability to learn
more things or to learn the same things more quickly .
The reduction of all valuations of response to valuation
as symptoms of ability to learn more and more quickly
thus seems too narrow a view . It has other defects . Were
it true , we ought , other things being equal , to get better
correlations with a criterion of intellect from tests in learn
ing something new and from tests deliberately framed to
measure how much one has learned in life so far , than from
the existing batteries of miscellaneous tasks .
This does not seem to be the case . Quantitative data
concerning individual differences in learning under experi
mental conditions are rather scanty , and their correlations
with a criterion of intellect are scantier still ; but what facts
we have been able to gather do not show that , per hour of
time spent , tests in learning predict the criterion as well
as do the tests now in use . Tests framed to measure how
much one has learned in life so far , such as vocabulary
tests , information tests , or such Binet elements as ' Knows
whether he is a boy or a girl , ' and the like , are valuable ,
but not , so far as we can determine , more valuable than a
composite containing also tests primarily of selective , rela
tional , generalizing , and organizing abilities .
OTHER ATTEMPTED SIMPLIFICATIONS OF THE PROCESS

OF VALUATION
Response to Novelty
one way allied to the doctrine just described and in
In
one way sharply contrasted with it , is the doctrine that a
person's intellect is measured by his ability to respond well

to new situations , to do ' originals . ' The importance of
some such ability as this will , of course , be admitted . How
ever , in view of the great difficulty of deciding just what
situations are ' new ' for any given individual ; in view of
the fact that ' to respond well ' is likely to bring in many
or all of our vague general valuations again ; in view of the
fact that distinctions among novel situations as ' harder '
( that is , making greater demands on intellect ) will have the
same effect ; and in view of the fact that our most approved
present instruments include many tasks which seem as
fittingly called responses to the familiar as to the new ―――
in view of all this it seems best at present not to try to

narrow our valuations to fit this theory .
Relational Thinking
Spearman has argued that intellect equals the appre
hension of experience , the eduction of relations and the
eduction of correlates . The two processes are defined as
"
follows : The mentally presenting of any two or more
characters ( simple or complex ) tends to evoke immediately
a knowing of relation between them . " [ 23 , p . 63. ] " The
presenting of any character together with any relation
tends to evoke immediately a knowing of the correlative
character . " [ 23 , p . 91. ]
There is no doubt that the appreciation and manage
ment of relations is a very important feature of intellect ,
by any reasonable definition thereof . Yet it seems hazard
ous and undesirable to assume that the perception and use
of relations is all of intellect . In practice , tests in para
graph reading, in information , and in range of vocabulary ,
seem to signify intellect almost as well as the opposites and
mixed relations tests . In theory , analysis ( thinking things
into their elements ) , selection ( choosing the suitable ele
ments or aspects or relations ) , and organizing ( managing
many associative trends so that each is given due weight
in view of the purpose of one's thought ) , seem to be as
deserving of consideration as the perception and use of

I
relations . Moreover , fear that , in all four cases , we need
other valuations to decide which are the better relations
or more abstract relations , or the more essential elements ,
or the more sagacious selection , or the more consistent
organization , or the more desirable balance of weights , and
the like .
However this may be , our present tests of intelligence
are not merely instruments to measure how little stimulus
is required to produce a perception of a relation , or how
many relations will be perceived from a given constant
stimulus , or how quickly . And we may best study them as
they are before dismissing the valuations on which they
are based , in favor of any simpler and more objective sys
tem .
We shall then accept for the present the status of mea
surements of intellect as measures of different products
produced by human beings or of different ways taken by
them to produce the same product , each of these products
and ways having value attached to it as an indication of
intellect by a somewhat vague body of opinion whether
popular or scientific .
THE CONTENT OR DATA OF TESTS OF INTELLECT

Presumably a man can use intellect and display the
amount of it which he possesses in operations with any
sort of material object , any living plant or animal , includ
ing himself , any quality or relation that exists in reality or
in imagination , any idea or emotion or act . Our tests
might draw upon anything for their material .
They have , in fact , greatly favored words , numbers ,
space -forms , and pictures , neglecting three - dimensional ob
jects and situations containing other human beings . How
far this has been due to convenience , and how far intellect
is really best measured by its operations with words , num
bers , space -forms , and pictures , is a matter that obviously
deserves investigation . Our choices of test material have
certainly been somewhat determined by convenience . They

have also favored ideas , general notions , abstractions , sym
bols and relations , to the relative neglect of percepts and
particulars . This has been in the main deliberate , our
general scheme of valuation attaching on the whole more
intellectual worth to operations with generals and facts in
relations than to particulars and facts in isolation .
The nature and extent of the specialization of intellect ,
according to the content or material operated on , has been
and still is a matter of dispute ; and the difference of
opinion carries over into the practice of measurement .
Some psychologists would be fairly well satisfied to mea
sure intellect by a series of mazes alone ; or by a series of
sentence completions alone . Others , the great majority ,
attach much more confidence to a battery of tests including
surely both words and numbers , probably also some space
forms and perhaps some more concrete pictorial material .
THE FORM OF TESTS OF INTELLECT

Whether we consider the external appearance of the
tasks or the internal nature of the processes in the person
doing them , there is a great variety in respect to form ,
that is , to the operations performed with the words , num
bers , pictures , and other content . Externally , there ap
pear questions to be answered , sentences or pictures to be
completed , errors to be found and corrected , definitions to
be given and to be chosen , items to be matched , directions
to be followed , disarranged parts to be put together , dis
arranged events to be put in proper sequence , keys or codes
to be learned , true statements to be distinguished from
false , items to be checked as fit by various criteria , items
to be crossed out as unfit , and so on .
Internally , the individual finds himself striving to at
tend to certain matters , to fix others in memory , to recall
what he knows about others , to select from many things or
ideas the one which best satisfies certain requirements , to
define the relation between two terms , to discover an ele
ment common to three or four given facts , to hold in mind .

many different facts and use them to some specified pur
pose , and to inhibit customary habits in view of some rule .
He also finds himself in some cases ( such as many elements
of information tests , vocabulary tests , and arithmetical
computations ) utilizing a wide range of knowledge and
skills .
Any system of units of measurement that is to be ade
quate must then apparently be flexible enough to apply to
a wide variety of operations such as we may call attention ,
retention , recall , recognition, selective and relational think
ing , abstraction, generalization , organization , inductive and
deductive reasoning , together with learning and knowledge
in general .
SCORING THE PRODUCTS OF INTELLECT
In the great majority of instruments for measuring in

tellect the score or rating is determined in part by the de
gree of difficulty of the tasks the individual can do success
fully . Thus ' There are three main differences between a
president and a king ; what are they ? ' ( Stanford Binet
XIV 3) is harder than ' Are you a little boy or a little
,
girl ? ' ( Stanford Binet ,III
...
, 4) . To complete 3 6 8 16 18
36 · ( Alpha 6 , 20 ) is harder than to complete 10 15
20 25 30 35 . . . . ( Alpha 6 , 2 ) . Psychologists and scien
tific and sensible people in general readily rank tasks as
easy or hard for intellect and would accept the principle
that , ' Other things being equal , the harder the tasks a per
son can master, the greater is his intelligence . ' The con
cept of hardness or difficulty in intellectual tasks , as now
used , is somewhat vague and variable . Its outstanding
characteristic is that among a large group of persons vary
ing in intelligence , the harder the task , the fewer will be
the persons who can do it , and the more intelligent they
will be. Sometimes , however , tasks are called hard which
really are only recondite , familiar to few ; and sometimes
tasks are called hard which really are only long .
CHAPTER I
THE PRESENT STATUS¹
Existing instruments for measuring intellect² developed
from three roots , the interview , the school examination , and
the ' tests ' of sensory acuity , memory , attention , and the
like , devised during the early history of psychology . The
Stanford Binet , for example , is an improved , systematized
and standardized interview . The Army Alpha is in part an
improved school examination and in part an improved bat-
tery of tests like those used before 1900 by Galton , Ebbing-
haus , Cattell , Jastrow , and others .
Existing instruments represent enormous improvements
over what was available twenty years ago , but three funda-
mental defects remain . Just what they measure is not
known ; how far it is proper to add , subtract , multiply ,
divide , and compute ratios with the measures obtained is not
known ; just what the measures obtained signify concerning
intellect is not known . We may refer to these defects in
order as ambiguity in content , arbitrariness in units , and
ambiguity in significance .
AMBIGUITY IN CONTENT
If
we examine any of the best existing instruments , say
the Stanford Binet , the Army Alpha or the National Intel-
ligence Test , we find a series of varied tasks . Some concern
words , some concern numbers , some concern space relations ,
some concern pictures , some concern facts of home life .
Some seem merely informational ; some are puzzle - like .
Some concern mental activities which will be entirely famil-
iar to almost all of the individuals to be tested ; some con-
1 This chapter is reprinted with some alterations from the Psychological
Review , Vol . 31 , pp . 219 to 252 .
2 We shall use ' intellect ' and ' intelligence ' as synonyms throughout this
book .
1
he does . We must then consider as a possible principle

'Other things being equal , the greater number of tasks of
equal difficulty that a person masters , the greater is his
intelligence . ' This principle would not be accepted so
readily as the principle about greater difficulty , and per
haps would not be accepted at all unanimously . ‘ Knowing
more things than someone else , and being able to do more
things than someone else ' is not so clearly and surely hav
ing more intelligence as ' being able to do harder things
than some one else can do . '
The two things have been somewhat confused in general
discussions and in the construction of measuring instru
ments because , by and large , a person increases the num
ber of things he can do in large part by adding on harder
ones , and also because the person who can do the harder
can on the average learn those which the duller person can
learn more quickly than he , and so learns more of them .
Consequently what we may call the level or height or alti
tude of intellect and what we may call its extent or range
or area at the same level are correlated and either one is
an indicator of the other . It will be best , however , to keep
them separate in our thinking .
Inmany of the instruments for measuring intellect a
person's score is determined partly by the speed with which
he can do the tasks . Even in batteries of tests where all
candidates attempt all the tasks , speed may count , since the
persons who do the easier tasks more quickly may have time
to review some of the tasks and perfect their work . If
speed deserves any weight in determining the measures of
intellect it is by virtue of the principle that , ' Other things
being equal , the more quickly a person produces the correct
response , the greater is his intelligence . ' Giving much
weight to speed arouses decided objections in the laity and
among some psychologists , and the principle just stated
certainly would not be accepted as axiomatic . By and
large , however , if A can do harder things than B can , A will
do those things which B can do more quickly than B can
A certain moderate weight attached to speed will not then

much decrease a test's significance ; and , per hour of time
spent on testing and scoring , an even greater significance
may perhaps be obtained by giving a liberal weight for
speed than by giving none .
For the practical purposes of estimating intellect , a
battery of tests in which level , extent , and speed combine in
unknown amounts to produce the score may be very useful .
For rigorous measurements , however , it seems desirable to
treat these three factors separately , and to know the exact
amount of weight given to each when we combine them .
We shall try to make the concepts of intellectual prod
uct , difficulty of producing an intellectual product , range
of products produced , and speed of producing a product ,
more definite and precise , but without so altering them as
to lose the elements which have given them practical value
in the best current practice in measurement , or to weaken
in any way their usefulness in measuring intellects as we
actually find them by the tests which we have so far de
veloped .
We shall start with certain first approximations . For
a first approximation , let intellect be defined as that quality
of mind ( or brain or behavior if one prefers ) in respect to
which Aristotle , Plato , Thucydides , and the like , differed
most from Athenian idiots of their day , or in respect to
which the lawyers , physicians , scientists , scholars , and edi
tors of reputed greatest ability at constant age , say a dozen
of each , differ most from idiots of that age in our asylums .
Let an intellectual product , i.e. , a product or response
requiring , or depending on , intellect for its production , be
defined as a product or response which , given the same ex
ternal situation , the intellects in the half toward Aristotle
are more likely to make than the intellects in the half to
ward the idiot . For example , if, when all Athenians of
age forty were confronted by the question ' Is a straight
line the shortest distance between two points ? ' the growth
of the white blood corpuscles was equal for the Aristotelian
We have then no right to add , subtract , multiply , or

divide with these scores of A , B , C , D , E , and F in the way
that we do with their heights or weights . Suppose that A
scores 100 ; B , 110 ; C , 90 ; and D , 120. We cannot say that
the average intellect of A and B equals the average intellect
of C and D. If E changes from 60 to 70 , while F changes
from 70 to 80 , we cannot say that they have made equal
gross gains .
The numbers designating the scores made by individuals
are usually not even approximately related to any true zero
point . Consequently , even if the scores 1 , 2 , 3 , 4 , did
represent an equal - interval series of amounts or degrees of
the ability in question , they would properly be treated as
x + 1 , x + 2 , x + 3 , x + 4. The ' times as ' or ratio judg-
ment is thus not surely applicable and the relations of the
scores to anything else are thus undetermined . For ex-
ample , we cannot say whether the intellect of the average
twelve - year- old is one and a quarter times that of the aver-
age six- year - old or twice it , or ten times it.
The second main problem in improving measurements of
intellect is thus to attach fuller and more definite meanings
to these credit summations and difficulty levels , and if pos-
sible to find their equivalents on absolute scales on which
zero will represent just not any of the ability in question ,
and 1 , 2 , 3 , 4 and so on will represent amounts increasing
by a constant difference .
We have to estimate equivalents of this sort somehow
before we can make much use of ratings by either credit
summations or difficulty levels ; before , for example , we can
conveniently compare individuals or groups , or the changes
made by individuals or by groups , or the effects of different
environments . The commonest method at present is to take
as the equivalent for any score by any instrument , the age
whose average achievement is that score , and to assume that
5 Attempts have been made to define ' zero ' or ' just not any ' ability and to
assign scores in relation to zero in the case of knowledge of English words ,
ability to understand sentences , handwriting , drawing , and English composition .
the increments in average ability are equal for equal differ-

ences in age up to some limit such as 192 months , and are
zero thereafter . This of course is purely hypothetical in
general and is almost certainly in error for the ages near
the point where the age change suddenly turns from its
full amount to zero . The curve of ability in relation to age
is almost always smooth as in the continuous line of Fig .
FIG . la . The probable form of the curve of intellect in relation to age .
FIG . 1b . The form of the curve of intellect in relation to age , if annual gains
are equal up to some stated age , and are zero thereafter .
1a , but not with a sharp turn as in the dash - line of Fig . 1b .

The competent thinkers who use the method know this and
are cautious in inferences based upon its application to the
higher ages ; but they use it rather freely for the lower ages ,
because some method must be used , because it is easy to

understand and apply , and because we do not know what
method is really right .
It may be objected that equality of units is an unneces-
sary refinement , for present practical purposes , since the
mental age defines the status of an individual sufficiently ,
'as able as the average ten - year- old , ' ' as able as the average
twelve -year - old . ' These , it may be said , are better measure-
ments for practical purposes than some absolute scale in
terms of equal ' mentaces ' or ' intels . ' The convenience , in-
telligibility and realism of the mental age scale up to about
12 or 13 years are indeed great advantages , but after 13 or
14 it is neither convenient nor intelligible nor realistic . It
is not convenient because the computation of intelligence
quotients becomes very troublesome for the higher ages . It
is not readily intelligible because mental ages 14 , 15 , 16 , etc. ,
are not ' as good as the average ' 14- year old , 15- year old , etc.
The average 25 - year old for example is about the mental age
of 14 by one of the best instruments . It is not realistic
because we have no clear or vivid sense of what the average
person is intellectually at fifteen , or at sixteen , and do not
even know whether he improves in the next two or three
years . A mental age of 15 or 16 or 17 is in fact as arbitrary
a quantity as an Alpha ability of 123 .
A rarer but more promising procedure than that of trans-
forming test scores into ' ages ' is to transform them into
units of ability on the assumption that the distribution of
ability in all adults 21-30 , or in all twelve - year- olds , or in all
pupils in grade six of a certain city , or in some other speci-
-σ2π
fied group , is approximately that given by
1
y e
202
For example , the Alpha scores from 0 to 212 were not

used in the army at their face value , which would give a
distribution of the form shown in Fig . 2 , but were trans-
muted into seven letter measures by the following scheme ,
which assumed an approximately ' normal ' distribution for

a random sampling of 128,747 of the literate white draft :
100 200
FIG . 2. The form of distribution of the literate white draft if Army Alpha
are taken at their face value .
135-212 = A
105-134 = B
75-104 C= +
45-74 = C
25-44C-
15- 24 -
0-14 = D-
D
The score used in the Thorndike - McCall test of para-

graph reading is not the number of correct answers , but a
transmutation on the assumption that the real ability con-
cerned is distributed ' normally ' amongst twelve - year - olds
in American cities .
We know very little concerning the permissibility of the
assumption of the so - called normal distribution for adults
or for an age , or for a school grade . The search for evi-
dence pro and con is one important feature of the attempt
to obtain units of mental ability which shall be at least ap-
proximately equal .
AMBIGUITY IN SIGNIFICANCE
The test score measures directly only the measurer's
impression from the subject's performance , or the summa-
tion in a more or less capricious fashion , of credits and
penalties for the subject's responses to the different ele-
ments of the tests , or a combination of these . What this
3
score signifies about the subject's intellect depends upon

the intuition of the measurer , or upon the correlation be-
tween the summation and intellect , or upon both . When we
assert that a child is found by measurement with the Stan-
ford Binet to have the intellect of a child 10½ years , all that
is really asserted is that the child does as well in that par-
ticular standardized interview as did the average of the
children of 102 years of age tested by Terman in making
his standards . We do not know what the average intellect
of these children was , nor how closely the Stanford Binet
score represents or parallels or signifies it .
When we assert that a man is found by measurement
with the Army Alpha to have the intellect of an average
recruit in the draft , all that is really asserted is that he does
as well in that particular battery of tests scored and sum-
mated in a particular way , as the average recruit did . Just
what the intellects of recruits were and how closely their
Alpha scores paralleled their intellects , we do not know.
The measurement is one thing , the inference to intellect is a
different thing .
This is of course true of many measurements . The
amount of silver deposited in one second by an electric cur-
rent is not the amount of current . The dividend rate on
stock during any one year is not the worth of the stock . The
amount of silver is , under proper conditions , of perfect sig-
nificance as an indicator of the amount of current , since the
correlation between it and a perfect criterion of amount of
current is perfect . The dividend rate is of very imperfect
significance , since the correlation between it and a perfect
criterion of the worth of the stock is far from perfect .
We do not know how closely the rating or score in the
Stanford Binet or the Army Alpha or any other instrument
correlates with a perfect criterion of intellect , because we do
not know what such a criterion is , much less its correlations
with these tests . One great task of the measurement of
intellect is to obtain such a criterion , or a closer approxima-
tion to it than we now have , and to use it to improve the
selection and weighting of the elements of our testing in-

struments .
The present status of such instruments as the Binet or
Army or National tests is roughly as follows : We have
chosen tests where the judgment of sensible people in gen-
eral is that correct response or speed of correct response is
characteristic of intellect . Such is the case with directions
tests , arithmetical problems , common sense questions ( as
in Alpha 3 ) , and the like . We have chosen tests using the
judgment of psychologists in the same way . Such is the
case with the completion tests devised by Ebbinghaus , the
mixed relations or analogies test devised by Woodworth ,
and the like . We have tried these or other tests with chil-
dren secluded in institutions because of imputed intellec-
tual inferiority and with children of like age who are in
ordinary schools ( as by Norsworthy ) , with adult males of
good health and morals who were found in a Salvation
Army home , glad to work for a dollar a day , and with
adults of the professional classes ( as by Simpson ) , with
children in general of different ages ( as by Binet and
Terman ) , with various groups of children ranked for im-
puted intelligence by teachers , fellow pupils , school ad-
vancement , and other symptoms (as by Spearman , Burt ,
Terman , Whipple , Yerkes , and others ) , with children of
alleged superior intelligence in comparison with others ( as
by Whipple and Terman ) , with soldiers in the National
Guard and regular army in connection with ratings for
intelligence given by their officers ( as by the Psychology
Committee of the National Research Council ) , with stu-
dents whose success in high school and college studies was
also measured (as by Colvin , Wood , and many others ) ,
with individuals who were tested with a very long series
of tests ( as by Terman and Chamberlain , Stenquist , and
others ) , and in other ways .
As a general result we know that certain systematized
interviews and batteries of tests measure somewhat the
same trait , since they correlate somewhat one with an-
other and that this trait has to some extent the same con-
;
stitution as the trait which sensible people , psychologists ,
and teachers rate as intellect .
The failure of perfect correlation between the amount
of intellect a person has , as revealed by the criterion , and
the amount indicated by the instrument is due , as has been
said, partly to the imperfection of the criterion , but partly
also to the imperfection of the instruments . They ( at
least all but one of them ) are demonstrably imperfect , since
no two of them correspond perfectly in their findings for
the same intellects . Since it is extremely unlikely that ,
out of a dozen instruments devised with about equal care
by a dozen individuals or committees at about the same
date one should be very much superior to all the others ,
we may assume , until there appears proof to the contrary ,
that all are imperfect .
The imperfection may be of two sorts . First , the re-
sponses measured by the instrument may not be represen-
tative of the whole intellect and nothing but intellect ; the
score obtained may not give enough weight to certain fac-
tors or elements of intellect and may give weight to others
which really deserve less or even zero weight . The instru-
ment is then like a wattmeter which gives only half weight
to the voltage of the current or adds two watts for every
time that the current is turned on or turned off . Second ,
the same person may receive a different score when re-
measured by the instrument . In so far as such differences
are due to the ' accidental ' ups and downs in the person's
achievements , they are taken care of by measuring him at
enough different times ; but in so far as they are due to ac-
quaintance with the instrument itself or with instruments
like it , they are a very serious imperfection . For example ,
a given score with Army Alpha represents a very different
status according as it is from a first , a second , or a third
trial . The case here is as if a thermometer tended after
subjection to a temperature of 200 ° once to register 220 °
when 200 ° was next encountered . The provision of means
for distinguishing between that part of the score due to

certain general characteristics of the person measured and
that part of the score which is due to certain special train-
ing that he has had with the tasks of the tests , or with
tasks like them , is thus an important part of the work of
making the measurements more fully and exactly signifi-
cant of intellect .
In general , all our measurements assume that the indi-
vidual in question tries as hard as he can to make as high a
score as possible . None of them can guarantee that the
scores would correspond at all with a perfect criterion if
the individuals measured tried to appear as dull as they
could . The correlation would indeed then probably be in-
verse , the more intelligent persons being more successful
in their efforts to appear dull ! It is theoretically possible
to arrange a system of incentives such that each person
measured by an instrument would put forth approximately
his maximum effort , and in scientific testing of the instru-
ments this can often be done . In general practice , how-
ever , we rarely know the relation of any person's effort to
his possible maximum effort . Since , however , the disturb-
ances due to differences in effort on the part of those tested
require in study and treatment procedures which have little
or nothing to do with the procedures by which the instru
ments are made to give better measurements of those who
do try their best , we shall disregard the former and shall
limit our inquiry to the latter sort of procedures .
MEASUREMENTS OF INTELLIGENCE ARE MEASURES OF

INTELLECTUAL PRODUCTS
All scientific measurements of intelligence that we have
at present are measures of some product produced by the
person or animal in question , or of the way in which some
product is produced . A is rated as more intelligent than B
because he produces a better product , essay written , answer
found , choice made , completion supplied or the like , or pro-
duces an equally good product in a better way, more quickly
or by inference rather than by rote memory , or by more in-

genious use of the material at hand .
We can conceive of states of affairs such that a man's
intellect could be measured without consideration of the
products he produces or the ways in which he produces
them . Intellect might be exactly proportionate to the ac-
tivity of the thyroid gland , or to the proportion of the brain
weight to body weight , or to the number of associative neu-
rones in the frontal lobes or to the complexity of the fibril-
lary action of certain neurones , or to the intensity of a
certain chemical process , and hence be measurable by ob-
servations of the thyroid's action , or estimates of the
brain's volume , or by a count or measurement of neurones ,
or by a chemical analysis .
Psychologists would of course assume that differences
in intelligence are due to differences histological or physio-
logical , or both , and would expect these physical bases of
intelligence to be measurable . At present , however , we
know so little of the neural correlates of intellect that if
twenty college freshmen were immolated to this inquiry ,
ten being the most intellectual of a hundred , and ten
being the least intellectual of the hundred , and their
brains were studied in every way by our best neurologists ,
these could probably not locate sixteen out of the twenty
correctly as at top or bottom . Moreover , what we do know
of neural correlates is of little avail during life , the living
neurones being extremely inaccessible to present methods
of observation .
Even if one aimed

at discovering the physiological basis
of intellect and measuring it in physiological units , one
would have to begin by measuring the intellectual products
produced by it . For our only means of discovering physio-
logical bases is search for the physiological factors which
correspond to intellectual production .
MEASUREMENTS OF INTELLIGENCE IMPLY VALUATION
Our present measurements of intelligence rest on human

judgments of value , judgments that product A is ' better '
a minute each , and leave the best available intellect to put

its time on tasks far above level D3 .
Common sense recognizes the greater importance of alti
titude . It rates a Pasteur far above the most widely com
petent general practitioner . It does not ask how quickly
Milton could give opposites , or turn out doggerel rhymes .
Probably Pasteur was very much above the average in
extent of intellect ; probably Milton could have written as
good poetry as A can write very much faster than A can .
But common sense considers extent and quickness as unim
portant in comparison with reaching a level far above the
average .
From the economic and philanthropic points of view ,

altitude is enormously more important . If
an intellect
could be hired from Mars of so high level that it could learn
how to prevent war as easily as Jenner learned how to pre
vent smallpox , a million dollars a day would be a cheap
wage for the earth to pay him .
Our analysis of the measurement of intelligence may be
represented by space and number as follows :
Let one sixteenth of a square inch represent one intellec
tual task . Let those equal in difficulty be placed in the same
row across the page ; let the order of the rows from the
bottom to the top of the page be the increasing order of
difficulty ; let the square be shaded if the individual in ques
tion cannot do it ; if he can do it , let it bear a number repre
senting the time he requires to do it . For illustration , we
have assumed that there are 320 tasks and that they are of
20 levels of difficulty , 16 at each level .
Figures 3 and 4 then represent the measurements or in
ventories of two specimens of intellect
. Such measurements
or inventories may by using a random sam

be abbreviated
pling of tasks at each level , or by using only every other
level or every third or every fourth level , or in other ways .
Only one thing is needed to make such measurements sub
missible to the arithmetic and calculus of science in general .
That is the expression of the altitude of each level ( now
Insome cases the value is assigned so easily ( as a simple

deduction from , or following of, a general rule ) that we may
thoughtlessly assume that the response indicates intelli-
gence regardless of any process of valuation . For ex-
ample , we may consider that in a test in arithmetical com-
putation or problem solving , the right answers are signs of
intelligence , regardless of what anybody thinks . A little
thought will convince us , however , that in such tests the
human judgment acts as truly as in a completion or para-
graph - reading test . The main difference is that , having
once for all decided that right answers are better than
wrong answers , we do not raise the issue about any par-
ticular answer . We simply assume or make a general rule
of valuation . The valuation becomes obvious if we col-
lect all the responses made to an arithmetical task and ask
whether all the different ' rights ' are equally good or right ,
and whether all the different ' wrongs ' are equally undesir-
able .'
One criterion of value , truth , is so widely used in fram-
ing , keying , and scoring tests of intelligence that it deserves
comment , especially since there may be in the case of truth
an objective criterion , power in prediction , by which our
judgments of value are or should be determined . Two
other criteria of value also need comment because they have
been suggested explicitly or implicitly as direct criteria
for intelligence . They are development with age and abil-
ity to learn .
TRUTH
Probably over half of our present tests of intelligence
are tests where the response is given credit as a symptom
of intelligence in proportion to its truthfulness . Such is
the case , for example , with eight out of ten tests of the Otis
7 In the special case where we arrange for Yes and No answers valuation is
doubly active . We arrange so that a Yes or a No will be ' good ' as a response .
Then , since some of the correct ' Yeses ' or ' Noes ' may be due to chance , and
since chance answers are deemed of no value , we plan our scoring so as to give
the chance ' Yeses ' and ' Noes ' zero value .
Advanced ; and with Army Alpha , 2 , part of 3 , 4 , 6 , 7, and 8 .

It is more or less the case with Stanford Binet III , 5; IV ,
1 , 2, 3 ; V, 1 , 2, 3, 4 ; VI , 1, 2 , 3, 5 ; VII , 1, 2, 5 ; VIII , 4, 5 , 6 ;
IX, 1, 2, 3 ; X, 1, 2 ; XII , 1 , 2, 8 ; XIV , 1, 2 , 3, 5 , 6 ; and with

National Intelligence A,
4 , and B , 1 , 2 , 3, and 5 .
1, 3,
One could make an attractive theory of intelligence and
its measurement somewhat as follows : Intellect is con-
cerned with facts , being the ability to see and learn the
truth , to get true knowledge and use it to the best advan-
tage . Truth is insight into the real world , the evidence
that knowledge is true is its predictive power . Measures
of intelligence are then ultimately measures of a man's
mastery of prediction , that 2 and 2 will be 4 , or that it will
be profitable to buy such and such a stock , or that a planet
will be found having such and such a path. More immedi-
ately , they are measures of certain abilities which contrib-
ute to , or accompany , or indicate the existence of, the abil-
ity to get and use the truth .
By this theory we should rest our valuations of truth
all on the ultimate test of power of prediction . One truth
would be better than another in proportion as it predicted
more facts , or more important facts , or predicted the same
facts more acccurately , or helped more in the acquisition of
other truths . Our valuations of abilities as evidences of
intellect would rest on their significance as symptoms of
ability to get and use truth .
It seems sure , however , that people in general , psychol-
ogists , and framers of intelligence tests , alike mean by
intellect something more than ability in truth - getting to
improve prediction . They mean what Pericles and Wash-
ington and Gladstone had as well as what Aristotle and
Pasteur and Darwin had . In the oral interview of the
business man or physician to test intelligence , in such tests
as Ebbinghaus ' completions , and in such a battery of tests
as Army Beta , there is little obvious reference to predic-
tion or truth getting . In the first case , the aim is rather
to see how the person fits his thoughts and acts to little
evidence is the number and nature , in respect of intellect ,

of those who succeed with each task . The harder the task ,
the fewer the persons who succeed at it , and the more intel
lect they have . The argument implicitly involved is ( 1 ) that
whether a person succeeds or fails in such tasks is deter
mined largely by the amount of intellect which he possesses ,
and not greatly by anything other than intellect , and ( 2 )
that in the hardest tasks which a person masters , he uses
in general nearly all the intellect which he has .
The argument is sound enough to justify such a rank
order as the A, B , C . . . G order shown above , or the
order of a series made of Stanford Binet tests for Mental
Age 10 , Mental Age 12 , Mental Age 14 , Mental Age Adult ,
and Mental Age Superior Adult , but we shall find trouble
if we try to make a very close ordering , or to use the per
centages of successes for other than approximate rankings .
The exact determination of a rank order of test elements
for intellectual difficulty requires that the individuals in the
group be tested with each of the tasks under similar condi
tions , including interest and effort , which is a matter of
general scientific care that needs no further discussion here .
It requires also that each of the tasks in the series shall be
' intellectual ; ' and this requirement will eventually need
very elaborate discussion . We shall , indeed , find that it is
desirable to define an intellectual task as one in which the
person tested uses all the intellect he then has ; and in which
he differs from other persons in nothing save the amount
of intellect used .
If, however , we applied any such rigorous definition now ,
we should be unable to deal with any elements of any tests
ever used in measuring intellect , since not a single one of
them is a task which depends on intellect in its entirety ,
and differentiates individuals with no disturbance by any
thing other than intellect . A test element which did so
would correlate 1.00 with a perfect criterion . In order to
maintain continuity with previous work , we shall first treat
each test element as if correct response to it was caused by
intellect intact and uncontaminated by aught else .
ability in checking A's on a mixed sheet of capitals would

probably show less overlapping with age than vocabulary ,
rate of reading , and ability in completing sentences . But
they would be far less effective in diagnosing amount of
intelligence .
Development with age would be a poor and partial crite-
rion for intellect of any sort or degree , and for the higher
ranges of intellect , say those above the 70 -percentile intel-
lect of the average of the white draft , or above the average
ninth - grade pupil , it would be well - nigh worthless . It has
never been so used . The Terman mental ages above 14 ,
for example , are not functions of development with age , but
of differences between individuals , regardless of age .
ABILITY TO LEARN
An obvious hypothesis , often advanced , is that intellect
is the ability to learn and that our estimates of it are or

,
should be estimates of ability to learn . To be able to learn

harder things or to be able to learn the same thing more
quickly would then be the single basis of valuation . Suc-
cess in solving arithmetical problems , or defining words , or
completing sentences would then be good , simply and solely
because it signified that the person had greater ability to
learn .
If greater ability to learn means in part ability to learn

harder things , we have excluded the vague general valua-
tion of certain products and ways of producing only to in-
clude it again . For we shall find ourselves selecting or
defining A as harder to learn than B on the ground that
only the more intellectual persons can do it , or on the
ground that it requires a higher type of intellect , and shall
find ourselves using those vague general valuations to pick
the persons or describe the type of intellect required .
If greater ability to learn means only the ability to learn
more things or to learn the same things more quickly , we
have a view that has certain advantages of clearness and
approximate fitness to many facts . Even less than in the
case of truth - getting , however , do our present actual instru-
ments for measuring intelligence measure directly a per-

son's ability to learn more things than another person can ,
or to learn the same things more quickly . The substitution
test included in Army Beta , in the National Intelligence
Examination and in some others , is about the only test of
speed of learning that is used ; and it is more than a learn-
ing test.
Much evidence will therefore be required before we can
wisely replace our present multifarious empirical valua-
tions by the formula that intellect is the ability to learn
more things or to learn the same things more quickly .
The reduction of all valuations of response to valuation
as symptoms of ability to learn more and more quickly
thus seems too narrow a view . It has other defects . Were
it true , we ought , other things being equal , to get better
correlations with a criterion of intellect from tests in learn-
ing something new and from tests deliberately framed to
measure how much one has learned in life so far , than from
the existing batteries of miscellaneous tasks .
This does not seem to be the case . Quantitative data
concerning individual differences in learning under experi-
mental conditions are rather scanty , and their correlations
with a criterion of intellect are scantier still ; but what facts
we have been able to gather do not show that , per hour of
time spent , tests in learning predict the criterion as well
as do the tests now in use . Tests framed to measure how
much one has learned in life so far , such as vocabulary
tests , information tests , or such Binet elements as ' Knows
whether he is a boy or a girl , ' and the like , are valuable ,
but not , so far as we can determine , more valuable than a
composite containing also tests primarily of selective , rela-
tional , generalizing , and organizing abilities .
OTHER ATTEMPTED SIMPLIFICATIONS OF THE PROCESS

OF VALUATION
Response to Novelty
In one way allied to the doctrine just described and in
one way sharply contrasted with it , is the doctrine that a
person's intellect is measured by his ability to respond well

to new situations , to do ' originals . ' The importance of
some such ability as this will , of course , be admitted . How-
ever , in view of the great difficulty of deciding just what
situations are ' new ' for any given individual ; in view of
the fact that ' to respond well ' is likely to bring in many
or all of our vague general valuations again ; in view of the
fact that distinctions among novel situations as ' harder '
( that is , making greater demands on intellect ) will have the
same effect ; and in view of the fact that our most approved
present instruments include many tasks which seem as
fittingly called responses to the familiar as to the new-
in view of all this it seems best at present not to try to
narrow our valuations to fit this theory .
Relational Thinking
Spearman has argued that intellect equals the appre-
hension of experience , the eduction of relations and the
eduction of correlates . The two processes are defined as
follows : " The mentally presenting of any two or more
characters ( simple or complex ) tends to evoke immediately
a knowing of relation between them . " [ 23 , p . 63. ] " The
presenting of any character together with any relation
tends to evoke immediately a knowing of the correlative
character . " [ 23 , p . 91. ]
There is no doubt that the appreciation and manage-
ment of relations is a very important feature of intellect ,
by any reasonable definition thereof . Yet it seems hazard-
ous and undesirable to assume that the perception and use
of relations is all of intellect . In practice , tests in para-
graph reading , in information , and in range of vocabulary ,
seem to signify intellect almost as well as the opposites and
mixed relations tests . In theory , analysis ( thinking things
into their elements ) , selection ( choosing the suitable ele-
ments or aspects or relations ) , and organizing ( managing
many associative trends so that each is given due weight
in view of the purpose of one's thought ) , seem to be as
deserving of consideration as the perception and use of

relations . Moreover , I fear that , in all four cases , we need
other valuations to decide which are the better relations
or more abstract relations , or the more essential elements ,
or the more sagacious selection , or the more consistent
organization , or the more desirable balance of weights , and
the like .
However this may be , our present tests of intelligence
are not merely instruments to measure how little stimulus
is required to produce a perception of a relation , or how
many relations will be perceived from a given constant
stimulus , or how quickly . And we may best study them as
they are before dismissing the valuations on which they
are based , in favor of any simpler and more objective sys-
tem .
We shall then accept for the present the status of mea-
surements of intellect as measures of different products
produced by human beings or of different ways taken by
them to produce the same product , each of these products
and ways having value attached to it as an indication of
intellect by a somewhat vague body of opinion whether
popular or scientific .
THE CONTENT OR DATA OF TESTS OF INTELLECT

Presumably a man can use intellect and display the
amount of it which he possesses in operations with any
sort of material object , any living plant or animal , includ-
ing himself , any quality or relation that exists in reality or
in imagination , any idea or emotion or act . Our tests
might draw upon anything for their material .
They have , in fact , greatly favored words , numbers ,
space - forms , and pictures , neglecting three - dimensional ob-
jects and situations containing other human beings . How
far this has been due to convenience , and how far intellect
is really best measured by its operations with words , num-
bers , space -forms , and pictures , is a matter that obviously
deserves investigation . Our choices of test material have
certainly been somewhat determined by convenience . They

have also favored ideas , general notions , abstractions , sym-
bols and relations , to the relative neglect of percepts and
particulars . This has been in the main deliberate , our
general scheme of valuation attaching on the whole more
intellectual worth to operations with generals and facts in
relations than to particulars and facts in isolation .
The nature and extent of the specialization of intellect ,
according to the content or material operated on , has been
and still is a matter of dispute ; and the difference of
opinion carries over into the practice of measurement .
Some psychologists would be fairly well satisfied to mea-
sure intellect by a series of mazes alone ; or by a series of
sentence completions alone . Others , the great majority ,
attach much more confidence to a battery of tests including
surely both words and numbers , probably also some space-
forms and perhaps some more concrete pictorial material .
THE FORM OF TESTS OF INTELLECT

Whether we consider the external appearance of the
tasks or the internal nature of the processes in the person
doing them , there is a great variety in respect to form ,
that is , to the operations performed with the words , num-
bers , pictures , and other content . Externally , there ap-
pear questions to be answered , sentences or pictures to be
completed , errors to be found and corrected , definitions to
be given and to be chosen , items to be matched , directions
to be followed , disarranged parts to be put together , dis-
arranged events to be put in proper sequence , keys or codes
to be learned , true statements to be distinguished from
false , items to be checked as fit by various criteria , items
to be crossed out as unfit , and so on .
Internally , the individual finds himself striving to at-
tend to certain matters , to fix others in memory , to recall
what he knows about others , to select from many things or
ideas the one which best satisfies certain requirements , to
define the relation between two terms , to discover an ele-
value , it appears that the variability of an individual whose

median score is about 105 ( from 100 to 113 ) is very nearly
the same as the variability of an individual whose median
score is about 128 ( 125 to 132 ) . If
, however , the units of
the scoring scale from 90 to 120 really represent smaller in

crements of ability than the units from 120 to 145 , the real
variability of an individual of ability 105 is less than the
real variability of an individual of ability 128 , and con
versely , if the units of the scoring scale from 90 to 120
really represent larger increments of ability than the units
from 120 to 145 .
We thus record the face - value - score results for many
different sorts of tests of intelligence , noting in each case
any facts about the construction of the tests which concern
the probability that its units progressively swell or shrink
in ' real ' value over any considerable fraction of the range
we are concerned with . We note especially the results in
those cases where there is no reason to expect swelling
more than shrinking . The average relations between varia
bility and ability found in these cases may be taken to rep
resent approximately the real relation , until some one pro
duces evidence that , in all or nearly all tests for the ability
in question , there are forces leading psychologists , quite
without intention , to devise scoring plans which make for
progressive swelling or shrinking of units .
The general drift of the facts is shown in Table 2 which
gives the variability ( in face - value - score units ) of an indi
vidual from day to day in intellect as a percent of the varia
bility of a person whose amount of intellect is that repre
sented by an Army Alpha first - trial score of about 100 .
4 We have secured extensivedata concerning Army Alpha , Examination A ,
Army Beta , Stanford Mental Age , the National Intelligence Test , the Otis
Advanced Test , the Haggerty Delta 2 , the Myers Mental Measure , the Kelley
Trabue , the Stanford Binet , the Terman Group Test , the I.E.R. Test of Selec
tive and Relational Thinking , the I.E.R. Test of Generalization and Organiza
tion , the Thorndike Non - Verbal Test , the Thorndike Examination for High
School Graduates , series of 1919 to 1930 , and the Toops Clerical Test . See
Appendix II.
We shall presently define this concept of the intellectual

difficulty of a task , so as to make it more useful in science ,
but for the present we may leave it vague , the principle
stated above being true for any reasonable definitions of
, ' and ' intelligence . '
' difficulty
In many of the instruments for measuring intellect
there are tasks which are of equal difficulty ( or at least
tasks so nearly equal that which of them is hardest is not
certain ) . In the Binet series the tasks for any one year of
age were supposed to be equally hard . In Alpha 7 only
by statistical inquiry could one decide which of these is
hardest , which next hardest and so on .
- -
- -
6 love hatred :: friend
lover mother need enemy .
7 wrist
- -
bracelet :: neck collar leg foot giraffe
...
sailor navy
- soldier gun private army fight
-
-
::
10 9 8
..
67890
carpenter house
....
shoemaker hatmaker wax shoe leather
:: ::
shoestring button coat catch bell hook 10

-
shoe
- -- -
.
11 quinine bitter sugar cane sweet salt beets.... 11

::
tiger wild cat dog mouse tame pig 12

--
12
:: ::
- -
.
legs man spokes carriage go tire 13

13
wheels
-
.
14 north south east north west south east 14

... ...
feather float rock ages hill sink break 15

-
15
-
::
- --
16 grass cattle bread man butter water bones.... 16
::
fin fish wing feather air bird sail 17
-
-
17
...
::
-
paper wall carpet tack grass sweep floor 18
- - - -
18
..
19 food man fuel engine burn coal wood 19

...
::
runner buggy horse carriage harness wheel 20

- -
20 sled
::
..
21 poison death food eat bird life bad 21

::
...
- -
22 Japanese Japan Russia China Japanese pigtail
-
Chinese
::
-
..
23 angels heaven men earth woman boys Paradise 23

::
-
..
22222
24 Washington Adams first contrast best second last 24

...
::
25 prince princess king palace queen president kingdom 25

..
::
Now test includes dozen tasks absolutely equal

if
in
a
difficulty for people general any one person who gets

in
some right will by no means always get them all right and
,
any one person who gets some wrong will by no means

always get them all wrong So person's score partly
is
a
.
determined by the number tasks equal difficulty

of
that
of
4
It appears from Table 2 , and still more clearly from the

consideration of the detailed facts in Appendix II which
Table 2 barely summarizes , that if we had scales for
intellect whose units were really equal , the variability of an
individual from day to day would be the same , regardless of
whether the average amount of intellect possessed by him
was that of a ' low grade ten - year - old ' or of a ' superior
adult , ' that of an Army Alpha score of 25 or that of a score
of 175 .
This result is so important , if true , that we have sought

for facts and probabilities in real or apparent opposition
toit .
First , there are the obvious opposing facts of range of
variability in intellectual or similar production . Keats may
have written " On Reading Chapman's ' Homer ' " in one
hour , and have written nothing in some other hour when he
tried as hard , whereas an average twelve -year - old varies at
the most from nothing up to a composition scoring 50 on
the Hillegas scale . A gifted stock - exchange trader who in
transactions of 10,000 shares a day, averages $ 100 profit ,
may vary from a profit of $ 25 to one of $ 2,500 , whereas a less
gifted trader who averages $ 10 a day on 100 shares in the
same market , it is said , varies over a much narrower range .
Such apparently opposing facts as these are , however ,
not so simple as they seem . If we had a full record of all of
Keats ' hours of equal effort , the production called zero
might turn out to be far above zero . The ideas he had then
might rank in poetic value far above those of the best hours
of the average man . The less gifted trader may vary over
just as wide a range . For example , a still less gifted trader
losing $ 100 on the average , may lose in two days the $ 25
and the $ 2500 that the gifted trader gains . Furthermore ,
we have to consider the alleged common observation that
as one increases his expertness in acting , music , dancing , or
athletic feats , he seems to reduce his variability . Thus a
sprinter who can on the average run 95 yards in 10 seconds
almost never runs less than 90 yards or more than 98 yards
A certain moderate weight attached to speed will not then

much decrease a test's significance ; and , per hour of time
spent on testing and scoring, an even greater significance
may perhaps be obtained by giving a liberal weight for
speed than by giving none .
For the practical purposes of estimating intellect , a
battery of tests in which level , extent , and speed combine in
unknown amounts to produce the score may be very useful .
For rigorous measurements , however , it seems desirable to
treat these three factors separately , and to know the exact
amount of weight given to each when we combine them .
We shall try to make the concepts of intellectual prod-
uct , difficulty of producing an intellectual product , range
of products produced , and speed of producing a product,
more definite and precise , but without so altering them as
to lose the elements which have given them practical value
in the best current practice in measurement , or to weaken
in any way their usefulness in measuring intellects as we
actually find them by the tests which we have so far de-
veloped .
We shall start with certain first approximations . For
a first approximation , let intellect be defined as that quality
of mind ( or brain or behavior if one prefers ) in respect to
which Aristotle , Plato , Thucydides , and the like , differed
most from Athenian idiots of their day , or in respect to
which the lawyers , physicians , scientists , scholars , and edi-
tors of reputed greatest ability at constant age , say a dozen
of each , differ most from idiots of that age in our asylums .
Let an intellectual product , i.e. , a product or response
requiring , or depending on , intellect for its production•, be
defined as a product or response which , given the same ex-
ternal situation , the intellects in the half toward Aristotle
are more likely to make than the intellects in the half to-
ward the idiot . For example , if, when all Athenians of
age forty were confronted by the question ' Is a straight
line the shortest distance between two points ? ' the growth
of the white blood corpuscles was equal for the Aristotelian
and the idiotic halves , whereas the answer Yes was more
prevalent in the Aristotelian half , we should rate the latter
as a product depending on intellect , and the former as a
product not depending on intellect .
Let the intellectual difficulty of producing a given intel-
lectual product in response to a given external situation be
defined as follows : Enough time being allowed for produc-
tion so that an increase in time would not increase the num-
ber producing it , the difficulty for Athenians of forty is
approximately greater the smaller the number of them who
produce it , provided that the ranking of those who do pro-
duce it differs from the ranking of those who do not by
greater nearness to the Aristotelian end . We could be
much more rigid here by supposing a population to vary
from the idiots to the Aristotles in amount of intellect only ,
being identical in all else . Then , if all conceivable pro-
ductions of intellectual products in response to given ex-
ternal situations were ranked for difficulty , the order would
be very closely that of rarity and of the nearness to Aris-
totle of those who achieved it . We could omit the ' ap-
proximately , ' and the ' provided that . ' Our definition has
deliberately been left loose , since we do not know exactly
what it is in which Aristotle differs most from the idiot ,
much less can we know in the case of any group of actual
individuals that they are identical in all else than it .
The range of products produced at any one level , i.e.
of products which are equally hard to produce , is defined
simply by their number . What we may call the relative
range at any level may be defined as the percent or fraction
of the products at that level which can be produced by the
intellect in question . The speed of producing any given
product is defined , of course , by the time required .
It will be convenient to use the word task to mean the
production of a given product in response to a given exter-
nal situation , and to speak of the difficulty of tasks , the
number of tasks of a given difficulty that can be done , and
the speed of doing a given task .
We now have intellect defined by a ranking of men whose

differences therein are roughly appreciated as we appre-
ciate the differences of the world's varied objects in volume
(only much more roughly . ) We have intellectual tasks and
products defined in a catholic way that would , for instance ,
probably include every task in all the stock instruments in
use by psychologists to -day . We have difficulty defined
objectively so that a series of tasks could be approximately
ranked as to their respective amounts of difficulty for any
specified group .
If welist all tasks , find the difficulty of each , apply an
intellect to them , observe which it can do , and how long it
requires to do each , we have measured how hard tasks it
can do , how many it can do at each level , and how quickly
it can do them . If we use in place of a complete list of
tasks a fair sampling from them , we have attained the same
end , subject to the
error of our sampling .
The new problems of theory and technique in the mea-
surement of intellect that is , the problems not soluble by
,
the general methods of measurement in any science , con-

cern the measurement of difficulty of task . Extent and
are -
speed are measurable in two of the most perfect units there
number and time . In the case of difficulty , however ,
we have so far provided only for an inventory of intellec-
tual tasks and their arrangement in an order of difficulty .
Their differences in amount of difficulty and the dif-
ferences between the amount of difficulty of any one of
them and some zero point of difficulty ( some task which is
just below a task of infinitesimal difficulty ) , are not deter-
mined . To find ways of determining these will be our
main work .
Before attempting it , however , we may best consider
certain further facts about difficulty , extent and speed in
the production of intellectual products , and certain conse-
quences of our analysis of a measurement of intellect into
this three -fold determination .
intellect that would otherwise act , so that the more of them 1
that acted the less intellect there would be , the relation

between amount and variability would be reversed , the
variability of a man's intellect being as the square root of
the amount by which the man was below the maximum intel
lect ! There may be , and probably is , some combination of
additive and inhibitive factors making the average intel
lects of men vary up and down from an amount typical of
the human species ; and this may result in equal variability
for A, who is much below the average , and B , who is much
above it . For example , suppose there are 6 factors , a , b , c ,
d, e , and f, each contributing -1 , and 6 factors , A, B , C ,
D , E , and F , each contributing +1 ; and that every intellect

is constituted by 6 factors chosen from the 12 ; and that the
momentary conditions of each intellect represent the chance
whose averages range from -

combinations of its six factors . Then we have intellects
3 to +3, according to
whether they are constituted by six minus causes , or by 5
minus and 1 plus , or by 4 minus and 2 plus , or by 3 minus
and 3 plus , and so on . All will have the same variability ,
however , the frequencies being in the proportions 1 , 6 , 15 ,
20 , 15 , 6 , 1 , with a mean square deviation of 1.2247 .
A consideration of the relative probabilities of various
types of constitution of intellect out of positive and negative
factors would be interesting , but is too speculative to be
profitable for our present purpose . The attainment of
greater intellect by the lack or suppression of negative
factors as well as by the possession and use of positive
factors is at least a possibility ; and will seem highly prob
able to many .
On the whole , then , we do not need to be especially

skepticalof the experimental findings that the variability in
tests of a half hour from time to time is approximately
equal over the range from , say , the ten - percentile adult in
tellect to the ninety -five percentile adult intellect .
appreciably by growth or training . We should not leave

individuals to strive for ten hours to complete : " The body
gives light. the is the
...
sun because once in ten thousand times some child who
',
,
failed during nine hours succeeded the tenth This would
in
.
be valuable experiment but we have far more valuable
a
,
ways of using ten hours of his time
.
What we are really concerned about avoid rating
to
is
one task as harder than another merely because longer
it
is
,
that the poorer intellects do less quickly than the
so
it
others and within too short time limit show spuri-
so
a
a
,
,
ously greater percentage failures
of
.
We have made the requirement that the intellectual
ranking of those who do produce the response shall be
higher than that those who fail Usually this require-
of
ment unnecessary can that usually be assumed

It
is
is
,
,
.
that the good correct response will be obtained by the

or
better intellects more often than by the poorer inserted

It
is
.
provide against cases where the better intellects are sub-

to
ject some constant error that they give fewer correct

so
to
responses than the dull do or where other factors than in-

,
tellect distort the percent rights from what would be

of
if
it
everything but intellect were equalized For example

is
it
,
.
conceivable that and below were given ran-

to
if
a
a
)
,
(
)
(
dom sampling of intellects

,
Underline the right answers

:
1
equals 41
3
5
b a
4 4
( (
) )
¹-
equals 31 412
8
2
ratings for difficulty by the percents correct would be very

much error The percent for would probably be
in
a
)
(
.
lower than for because lacking knowledge exponents

of
b
(
)
the more intelligent one was the more likely one would be
4-1
,
report for valid 4-1 means and report

( to
to
if
a
3
X
) ,
(
),
valid means halves or for

if
4
as 4
2
b
1
)
(
).
We have treated the task being produce certain

as to
product scored consequently done not done

or
It
is
,
.
,
success or failure , right or wrong . Now when any task for

intellect is set there are often many different responses
varying in ' goodness ' or correctness . In such cases , our
method requires that in determining the difficulty of the
task , a dividing line be set somewhere . Our method will
not , however , prevent us from later using different credit
values in a scoring plan for such a task and taking full ad-
vantage of whatever added value these more detailed credit
values may have in estimating an individual's intellect .
It may be noted further that a task may consist of vari-
ous combinations and complications of other tasks . Thus
the task may be to get the right answer to 8 + 3 , or to get
the right answer to 11 + 7 , or to get the right answers to
both 83and 11 + 7, or to get the right answers to 8 + 3
and 11 + 7 and also 18+ 4 , or to get the right answer to :
Find the sum
94738
which ordinarily involves the above plus knowledge
of
229
,
of certain words and procedures and control over

,
certain habits such as holding numbers mind and adding

in
,
seen to thought of number

a
We are now in position to state one theorem of the

a
measurement intellect Let difficulty be defined above

of
as
,
.
then
:
Theorem Other things being equal intellect can

A
if
1
:
do correctly all the tasks that intellect can do save one

B
and place of that one can do one that harder than

in
is
it
intellect has the higher level

is A
One tempted go further and assume that other

to
things being equal and can do correctly the same

A
if
B
,
number tasks has the higher level the average diffi-

of
if
,
What seems be one task the person tested may be used as two or
to
to
8
more tasks by scoring first with the dividing line one place and second
at
at it
with the dividing line another

.
culty of the tasks he can do is greater than the average diffi-

culty of the tasks B can do . This cannot , as yet , be wisely
assumed first because we do not know that we have any
right to average measures of difficulty, and secondly , be-
cause , even if we could , it is not safe to assume that as much
intellect is required to do 10 tests each of difficulty 20 as to
do one task of difficulty 200 .
On the other hand one is tempted to suggest the measure-
ment of an intellect by the hardest things it can do , assuming
that since it can do these , it could do all easier , as we assume
that one who can jump over a bar 6 feet high could surely
jump over bars at 5 ft . 10 in . , 5 ft . 8 in . , and so on . The
possible variety and specialization of intellectual tasks
makes this uncertain .
WIDTH OR EXTENT OR RANGE
Our definition of greater difficulty enables us also to

define equal difficulty and so to make a fairly rigorous defini-
tion of width or extent or range by making it separately at
each level of difficulty . For any specific group G and any
specific time t those tasks are equally difficult which are done
correctly by equal percentages of intellects .
Consider then all the tasks which are of a certain diffi-
culty D. Some intellects will fail with all of them . Among
the intellects which succeed with some of them we may make
comparisons according to the number succeeded with . Such
a statement as ' N tasks , of equal difficulty D , being given ,
with t time allowed per task , A did 0.1N while B did 0.2N
and C did 0.3N , ' is clear and useful . We can say that B
did twice as many as A, that C exceeded B in the number
done as much as B exceeded A, and that the average for A
and C was the same quantity as the score for B. Where the
problem concerns the extent of an ability , as in the number
of certain facts that are known in history or science or the
⁹ We have provided for determinations of which one of two or more tasks is
the more difficult , but not , as yet , for determinations of how much more diffi-
cult it is.
features of previous work which up till now has been taken

on faith . Appendices I and II
are perhaps of greater theo
retical importance , but Appendix III
is fundamental for
present and future practice in mental measurement .
We can then measure the difficulty of any intellectual
task for pupils in any one of these grades by the percent of
the group succeeding with it , as shown in the illustration
that follows :
3190 pupils in grade 9 were tested with four tasks in com
pleting sentences . The percentages succeeding were re
spectively 60 , 30.5 , 46.1 , and 37.1 . We assume that these are
intellectual tasks , that is , that success with each depends
upon intellect .
The form of distribution of the intellects of the group
-
being Form A , a percentage correct of 60 corresponds to a
division of the group at -
.25330 , that is , at — .25330 of the
mean square deviation of the group ( in the ability mea
sured in truly equal units by that task ) below the average
or median of the group ( in the ability measured by that
task ) .
.51010 +.09790 , and +.32920 have similar meanings
,
for the difficulties of tasks 21 , 22 , and 23 .

The differences in difficulty between the tasks are
21-20 = .7634 , 21-22 = .4122 , and so on , in truly equal units ,
unity being taken arbitrarily as the mean square deviation
of the group in intellect .
MEASUREMENT BY WAY OF THE FORM OF DISTRIBUTION OF AN

ARRAY IN A CORRELATION TABLE
The fourth method of attacking our problem uses , as the
group whose form of distribution is to be determined , the
population comprising one array in a correlation table of
the sort shown in Table 6 , where the individuals are ar
rayed under their scores in some examination symptomatic
of intellect . Each array consists of two compartments rep
resenting the two scores ( Failure and Success ) attainable
in the intellectual task whose difficulty we wish to measure .
number of correct responses and the speed of producing

them .
In the instruments that are actually used , it is customary
to have the time a mixture of ( 1 ) the time spent in doing
some tasks correctly , ( 2 ) the time spent in doing other tasks
incorrectly and ( 3 ) the time spent in inspecting other tasks
and deciding not to attempt them . This confusion may
be permissible , or even advantageous , in the practical work
of obtaining a rough measure of intellect at a small expense
of time and labor and skill , but for theory at present and for
possible improvement of practice in the future we need to
separate the speed of successes from the speed of failures .
To the number of tasks correctly done at each level we
may add a record of the time for each or of the average
time for all at that level .
Since to save time in intellectual production is a ‘ good , '
we may frame Theorem III as follows : Other things being
equal , if intellect A can do at each level the same number of
tasks as intellect B , but in a less time , intellect A is better .
To avoid any appearance of assuming that speed is com-
mensurate with level or with extent , we may replace ' better '
by ' quicker . '
THE RELATIVE IMPORTANCE OF ALTITUDE , EXTENT , AND

QUICKNESS OF INTELLECT
Each of these three factors is essential . If it required
an infinite time per task , an intellect would produce no prod-
uct at any level no matter how high its potentialities as to
altitude and extent might be . If it had zero extent at all
levels , it would not matter how high its potentialities as to
altitude or how quickly it could do nothing . In the ordinary
sense of the word , however , altitude or level is by far the
most important . The chief evidence for this is that it alone
is indispensable , irreplaceable by anything save itself . If
the best available intellect can do only things of level D19 ,
we cannot get things of level D20 done at all . If the best
available intellect can do only 72 things at level D. and we
intellectual task ; and even if we did , the results would be

hard to interpret because of possible effects of practice . It
is possible to find a hundred individuals who are substan
tially identical in their average performance at intellectual
tasks , and test them all once with any given task .
task in terms of the distance + or -

The measurements of the difficulty of one intellectual
from the average of
one such array , expressed as a multiple of the variability
of that array , can be made approximately commensurate

with measurements of the difficulty of another intellectual
from the average of
the corresponding array , expressed as a multiple of its
variability . For we have shown that the variability of an
individual ( and so of such an array ) in intellect is approxi
mately the same regardless of his average amount of in
tellect . Consequently the two multiples are of approxi
mately the same unit and the distance between the two aver
ages of overlapping arrays can be measured in terms of this
same unit . If two arrays do not overlap , we can bridge
the gap by inserting data from intermediate arrays which
do form a series of overlapping arrays .
THE DEFECTS OF THE MEASUREMENTS SO FAR DESCRIBED
We have determined the approximate form of distribu

tion of a grade population , from Grade 6 to Grade 12 , in re
spect of level of intellect at one time , if that were measured
in truly equal units . We have done the same for a popula
tion ( an array ) characterized by identity in average of in
tellect measured by a random selection of times . By an ex
tension and refinement of the methods which we have used ,
this could be done with greater precision .
If all that we require for the measurement of the intel
lectual difficulty of tasks is to secure a group of known form
of distribution in intellect when measured in truly equal
units , whose members we may test with the tasks in ques
tion , the problem is solved . Unfortunately more is required .
The chief defect in our procedures is that the difficulty which
a minute each , and leave the best available intellect to put

its time on tasks far above level D3 .
Common sense recognizes the greater importance of alti-
titude . It rates a Pasteur far above the most widely com-
petent general practitioner . It does not ask how quickly
Milton could give opposites , or turn out doggerel rhymes .
Probably Pasteur was very much above the average in
extent of intellect ; probably Milton could have written as
good poetry as A can write very much faster than A can .
But common sense considers extent and quickness as unim-
portant in comparison with reaching a level far above the
average .
From the economic and philanthropic points of view,

altitude is enormously more important . If an intellect
could be hired from Mars of so high level that it could learn
how to prevent war as easily as Jenner learned how to pre-
vent smallpox , a million dollars a day would be a cheap
wage for the earth to pay him .
Our analysis of the measurement of intelligence may be
represented by space and number as follows :
Let one sixteenth of a square inch represent one intellec-
tual task . Let those equal in difficulty be placed in the same
row across the page ; let the order of the rows from the
bottom to the top of the page be the increasing order of
difficulty ; let the square be shaded if the individual in ques-
tion cannot do it ; if he can do it , let it bear a number repre-
senting the time he requires to do it . For illustration , we
have assumed that there are 320 tasks and that they are of
20 levels of difficulty , 16 at each level .
Figures 3 and 4 then represent the measurements or in-
ventories of two specimens of intellect . Such measurements
or inventories may be abbreviated by using a random sam-
pling of tasks at each level , or by using only every other
level or every third or every fourth level , or in other ways .
Only one thing is needed to make such measurements sub-
missible to the arithmetic and calculus of science in general .
That is the expression of the altitude of each level ( now
only a fraction of intellect and is influenced by other forces

than intellect . That is , any one short task measures intel
lect plus an error . The nature and amount of this error
must be considered in connection with any procedure for
estimating the intellectual difficulty of a task from the per
centage of individuals who succeed with it."
There are other hidden assumptions and weak or even
missing links in the argument by which we proceed from
knowledge of who and how many can do a task , to a meas
ure of its intellectual difficulty . In the next chapter we shall
expose these , subject the entire argument to a much more
rigorous treatment , and seek to remedy the defect noted
above and such others as are found .
9 The exposure of this defect should not diminish our use of the general
procedure of inferring degree of difficulty from percentage of failures in a dis
tribution of known form . On the contrary , now that we are aware of the defect ,
we can make much better use of the procedure than when we were ignorant of it .
As we shall elsewhere show in detail , if we replace a single task by a composite
99
of forty tasks , and use twenty or more right as our mark of " success , we can
use the procedure with better results than have ever been obtained hitherto .
CHAPTER III
THE MEASUREMENTS OF THE INTELLECTUAL DIFFICULTY OF
TASKS AND OF LEVEL OF INTELLECT : MORE
RIGOROUS AND EXACT METHODS
In the two previous chapters we have operated with pro

visional and somewhat vague definitions and inexact as
sumptions , largely in order to maintain continuity with
what has been done to date in the measurement of intellect .
It is now necessary to treat the whole matter of intellectual
difficulty and level of intellect more rigorously .
We have assumed ( 1 ) that there is such a quality or
characteristic of man as altitude or level of intellect ; ( 2 )
whose amount or degree is measured by the height at which
it can attain success with a series of intellectual tasks
ranked for difficulty ; ( 3 ) that the same individual differs
in the amount or degree of it which he has available from
time to time ; and ( 4 ) that different individuals differ in
the amounts or degrees of it which they have available on
the average . (5)We have defined intellectual tasks only
loosely and vaguely as those in which men esteemed very
intelligent differ most from men esteemed very unintelli
gent . ( 6 ) We have defined intellectual difficulty only
loosely and vaguely as that characteristic of a task, an in
crease in which reduces the number of intellects who can
succeed with it , eliminating those esteemed unintelligent
more rapidly than those esteemed intelligent .
Since we are treating intellect as the ability to perform
intellectual tasks , our primary need is a clearer and more
exact notion of intellectual tasks . We can reach this in
either of two ways . The first is by assuming that certain
abilities , such as to understand directions , or to know facts ,
or to use relations of likeness , part and whole , actor and
acted upon , genus and species , and the like , or to use facts
59
evidence is the number and nature , in respect of intellect ,

of those who succeed with each task. The harder the task,
the fewer the persons who succeed at it , and the more intel-
lect they have . The argument implicitly involved is ( 1 ) that
whether a person succeeds or fails in such tasks is deter-
mined largely by the amount of intellect which he possesses ,
and not greatly by anything other than intellect , and ( 2 )
that in the hardest tasks which a person masters , he uses
in general nearly all the intellect which he has .
The argument is sound enough to justify such a rank
order as the A , B , C . . . G order shown above , or the
order of a series made of Stanford Binet tests for Mental
Age 10 , Mental Age 12 , Mental Age 14 , Mental Age Adult,
and Mental Age Superior Adult , but we shall find trouble
if we try to make a very close ordering , or to use the per-
centages of successes for other than approximate rankings .
The exact determination of a rank order of test elements
for intellectual difficulty requires that the individuals in the
group be tested with each of the tasks under similar condi-
tions , including interest and effort , which is a matter of
general scientific care that needs no further discussion here .
It requires also that each of the tasks in the series shall be
' intellectual ; ' and this requirement will eventually need
very elaborate discussion . We shall , indeed , find that it is
desirable to define an intellectual task as one in which the
person tested uses all the intellect he then has ; and in which
he differs from other persons in nothing save the amount
of intellect used .
If, however , we applied any such rigorous definition now ,
we should be unable to deal with any elements of any tests
ever used in measuring intellect , since not a single one of
them is a task which depends on intellect in its entirety ,
and differentiates individuals with no disturbance by any-
thing other than intellect . A test element which did so
would correlate 1.00 with a perfect criterion . In order to
maintain continuity with previous work , we shall first treat
each test element as if correct response to it was caused by
intellect intact and uncontaminated by aught else .
THE MEASUREMENT OF DIFFICULTY 61
perfect correlation is due entirely to the fact that the intel

lect of the same individual varies from time to time , not to
any intrinsic inadequacy or irrelevance .
If we take the latter way , we may attach the term " intel
lectual task " to any task or collection of tasks , the score in
which correlates +1.00 with the ranking of individuals in
an order for intellect by the consensus , or correlates as
closely with that ranking as its own self- correlation permits .
By collecting such tasks we may obtain a total series which
may then be used as a criterion in the same manner as a
series derived by the other method .
These two procedures are more definite and systematic
and rigorous forms of what has been done in test construc
tion . Psychologists have on the one hand taken tasks which
they assumed to be intellectual and have put samplings of
them into teams of tasks . On the other hand they have as
sumed that a certain validity attached to rankings by
teachers , by the sifting process of advancement in school ,
or by other forms of consensus , and have selected those col
lections of tasks which showed high correlations with such
a criterion .
What abilities and tasks shall be treated as intellectual
is essentially a matter of arbitrary assumption or choice at
the outset , either directly , of the abilities or tasks them
selves , or indirectly , of the consensus which provides the
criterion . After the first choice is made , tasks not included
in it , and even not known , may be found to correlate per
fectly with the adopted total , and so to be " intellectual " ;
but their intellectualness is tested by and depends on the
first arbitrary choice . Had a different first choice been
made , they might not be intellectual . This arbitrariness is
a sign of weakness , but it is at present unavoidable . We
have to define intellectual tasks as best we can , and trust
that future scientific uses of the definition will improve it .
We shall see later that the arbitrariness is greatly tempered
by certain guiding principles and facts , and that a total
series of intellectual tasks can be defined so as to represent
Using these results at their face value , we should con-

clude that within this group No. 6 can be done by intellects
of level x , and perhaps of lower levels ; Nos . 18 and 19 re-
quire level x + k ; Nos . 15 and 20 require level x + 2k ; No.
22 requires level x + 5k ; No. 28 requires level x + 7k ; No.
29 requires level x + 8k ; No. 30 requires level x + 9k ; Nos .
31 and 33 require level x + 10k or higher . Nos . 31 and 33
are thus 1k harder than 30 , which in turn is 1k harder than
29 , which in turn is 1k harder than 28 , which in turn is 2k
harder than 22 , which is 3k harder than 15 or 20 , and so on .
THE MEASUREMENT OF DIFFERENCES IN DIFFICULTY BY WAY

OF KNOWLEDGE OF THE FORM OF DISTRIBUTION OF
THE VARIATIONS OF AN INDIVIDUAL IN
LEVEL OF INTELLECT¹
An individual does not display the same level of intellect

at all times and seasons . He varies around his average
status . If we know the real form of distribution of his
variations in level , we can use it to compare the differences
of tasks in real difficulty , just as we use knowledge of the
form of distribution of a group .
1 We began our search for of measuring differences in difficulty by
means
inquiring whether the real form of distribution of the real abilities of the indi-
viduals represented in a single array in a correlation table , might not be de-
termined with greater certainty than the form of distribution of the group as a
whole . This is indeed often the case ; and the use of a group sorted into arrays
has much to recommend it . The consideration of the factors which do influence
the form of distribution of the real ability of the individuals in an array , led
us to a broader view of the means of scaling difficulty of task and level of
intellect .
The form of distribution of the real abilities in an array is determined by
three causes : ( 1 ) The form of distribution of an individual's variations around
his own average ; ( 2 ) the relation of an individual's variability to his amount
of ability , and ( 3 ) the form of distribution of the entire group from which the
array is sorted out by its correlation .
It will be shown that if we can determine the facts for any one of these ,
we can transmute certain differences in rank into differences in amount . The
transmutations by ( 1 ) and ( 2 ) are almost , ifnot quite , independent of those
by ( 3 ) in respect of facts and assumptions , and so provide a check of great
value . The use of an array insteadof a total group utilizes all three methods
together in a way that has many advantages . We shall not , however , make
use of this method in the main body of our work .
arated points of the success - frequency ranking . This is a

matter for experimental determination after the total series
has been chosen and the group of intellects in respect of
which difficulty has been defined has been chosen . The re
sults of such experiments are of great significance , inform
ing us of the degree to which amounts of intellect as defined
by the total series do represent increases in the same kind
of thing , and are amounts of some unified , coherent fact in
nature which can properly be isolated in thought from non
intellectual factors . Our experiments on this matter will
be reported elsewhere , but we may note now that they indi
cate that intellect has a rather high degree of unity and
consistency and independence of non - intellectual factors ;
and consequently permits a fairly close approximation to
sub - series of tasks which , as total sub - series , do approxi
mate to perfect " intellectualness , " while differing enor
mously in difficulty . We proceeded by an approximation
to the first method , and later checked our choice by an ap
proximation to the second method . In our choice we were
guided by the following considerations :
( 1 ) Of psychological theory :-(

a ) that responding to parts
or elements or aspects of situations is more " intellec
tual " than responding to gross total situations ; ( b )
that responding to parts or elements or aspects which
do not present themselves separately to sense but must
be abstracted is more intellectual than responding to
those which do ; ( c ) that responding to relations be
tween objects is more intellectual than responding to
objects ; ( d ) that , in particular , responding to so - called
subjective or logical relations , such as likeness and dif
ference , is more intellectual than responding to the so
called objective relations of space and time ; ( e ) that
organizing several mental connections or habits to se
cure a certain result , " thinking things together , " as
James put it , is more intellectual than using one habit
4 In Appendix IV and Appendix V.
at a time ; ( f ) that responses to novel situations are

likely to be more " intellectual " than responses to
familiar situations .
( 2 ) Of the theory of measurement :-(
a ) that the tasks
representing any one ability should be capable of very
fine gradation from very easy to very hard ; ( b ) that
they should be capable of very wide extension by alter
nates at any degree of difficulty ; ( c ) that , so far as pos
sible , any one ability should represent in some real and
useful sense something varying only in amount , so that
the different degrees of it might properly be repre
sented by numbers .
(3) Of common sense :-(
a ) that the tasks should be from
among those which had high standing on the basis of
correlations with reasonable criteria ; ( b ) that they
should be convenient for use in the actual measurement
of intellect ; ( c ) that they should be tasks concerning
which subjects for experiment were obtainable .
Over and above the narrowness due to these considera

tions , our choice is also deliberately narrow . We have not
included any tasks involving responses to actual human be
ings or to material objects present to sense tasks of what
has been called social intelligence and mechanical intelli
-
gence . Our tasks all concern responses to ideas and sym
bols , especially words and numbers . The reasons for this
need not detain us here . Also we limit ourselves to tasks
which are intellectual for a group of persons bred in the
5
United States and aged twelve or over . The reasons for
this narrowness may also in general be omitted .
5 The tasks will very probably serve to measure intellect for younger ages
even more accurately than tasks now in use , but we have not demonstrated this
to be so .
The chief reason was that the measurement of intellect in children up to
twelve or fourteen and the definition of the measurement by an age -scale are
in a far more satisfactory condition than the measurement of intellect at older
ages and at higher levels .
INTELLECT CAVD
In view of these elaborate and arbitrary restrictions , the

intellectualness of our total inventory of tasks , and the
intellect whose level or altitude , range or width , and facility
or quickness it measures , will be called hereafter Intellec
tualness CAVD and Intellect CAVD ( the symbol CAVD
refers to the four series of tasks which constitute it com -
pletions , arithmetical problems , vocabulary and directions ) .
The total series of tasks concerns four lines of ability : "
C. To supply words so as to make a

statement true and sensible .
A. To solve arithmetical problems .
V. To understand single words .
D. To understand connected dis
course as in oral directions or
paragraph reading .
The arrangementof scoring is such as to attach equal

weight to each of these four varieties of tasks .
The whole series is put into a rough approximate order
of intellectual difficulty by the methods described on pages
39 to 56 of the previous chapter . Consequently all the
single tasks or task elements of any one sub - series are of
somewhere nearly equal intellectual difficulty .
Each single task is scored 1 ( right ) or 0 ( wrong or
omitted ) . The number right at each level , that is in each
subseries , is recorded . The time required for each task
may be recorded , if desired . Selections of forty single - task
elements from each of certain sub - series of the total series
are shown below , making composite tasks A, B , C, D, N, O ,
P , and Q.
-
7 We shall sometimes use also Intellect CAVDI , which is constituted
by
including a fifth sort of task to understand and answer questions which re
quire information about such facts as are considered by the world to - day worthy
of study in school and of record in encyclopedias ; plus organization thereof
and sagacious inference therefrom .
value , it appears that the variability of an individual whose

median score is about 105 ( from 100 to 113 ) is very nearly
the same as the variability of an individual whose median
score is about 128 ( 125 to 132 ) . If
, however , the units of
the scoring scale from 90 to 120 really represent smaller in-

crements of ability than the units from 120 to 145 , the real
variability of an individual of ability 105 is less than the
real variability of an individual of ability 128 , and con-
versely, if the units of the scoring scale from 90 to 120
really represent larger increments of ability than the units
from 120 to 145 .
We thus record the face - value - score results for many
different sorts of tests of intelligence , noting in each case
any facts about the construction of the tests which concern
the probability that its units progressively swell or shrink
in ' real ' value over any considerable fraction of the range
we are concerned with . We note especially the results in
those cases where there is no reason to expect swelling
more than shrinking . The average relations between varia-
bility and ability found in these cases may be taken to rep-
resent approximately the real relation , until some one pro-
duces evidence that , in all or nearly all tests for the ability
in question , there are forces leading psychologists , quite
without intention , to devise scoring plans which make for
progressive swelling or shrinking of units .
The general drift of the facts is shown in Table 2 which
gives the variability ( in face - value - score units ) of an indi-
vidual from day to day in intellect as a percent of the varia-
bility of a person whose amount of intellect is that repre-
sented by an Army Alpha first - trial score of about 100 .
4 We have secured extensive data concerning Army Alpha , Examination A ,
Army Beta , Stanford Mental Age , the National Intelligence Test , the Otis
Advanced Test , the Haggerty Delta 2 , the Myers Mental Measure , the Kelley-
Trabue , the Stanford Binet , the Terman Group Test , the I.E.R. Test of Selec-
tive and Relational Thinking , the I.E.R. Test of Generalization and Organiza-
tion , the Thorndike Non -Verbal Test , the Thorndike Examination for High
School Graduates , series of 1919 to 1930 , and the Toops Clerical Test . See
Appendix II.
VOCABULARY A
(A row of 5 small pictures like those shown in Figure 6
is put before the person who is being measured . He is told
to " Show me the horse , " or " Put your finger on the
horse . " ) The words in A are :
21. pitcher 26. baseball

22. man 27. girl
23. string 28. train
24. apple 29. socks
25. violin 30. dog
DIRECTIONS , ORAL A
31. " Make a ring , like this , " showing act .
31. " Make a line , like this , " showing act .
33. " Make a cross , like this , " showing act .
34. " You can write , can't you ? Show me how you can
write . " ( Credit if S imitates effect of writing . )
35. " Put the cover on the box . " ( Credit if S attempts
to do so , turning cover to correct axis . )
" Turn
99
the box upside down . '
"
36.
37. Put the pennies in the box and then shake the box . "
( Have 4 pennies ; credit even if cover is not put on ,
if box is shaken . )
38. " Stand on that paper ' .
99
( A sheet of paper is left
on the floor .)
39. " Put your hands behind you . " ( Give while S is
standing . )
40. " Make a ring . ' 99
( If S fails , show again , but do not
credit . )
" How many ? " and the 2 pennies are slipped under the card . Another penny
is shown . " How many ? ''
is answered ( rightly or wrongly ) ; this penny also is
slipped under the card , and he is asked , " How many are under here now ? ''
In subtracting , as for instance 2 minus 1 , proceed as follows : The 2 pennies
are shown and the subject answers to " How many ? " The pennies are then
slipped under the card . One is then taken out as he watches , and the question
asked , " How many under here now ? " It is necessary to make sure that he
watches what is done .
It appears from Table and still more clearly from the

2,
consideration of the detailed facts in Appendix II which
Table 2 barely summarizes , that if we had scales for
intellect whose units were really equal , the variability of an
individual from day to day would be the same, regardless of
whether the average amount of intellect possessed by him
was that of a ' low grade ten - year - old ' or of a ' superior
adult , ' that of an Army Alpha score of 25 or that of a score
of 175 .
This result is so important , if true , that we have sought
for facts and probabilities in real or apparent opposition
to it.
First , there are the obvious opposing facts of range of
variability in intellectual or similar production . Keats may
have written " On Reading Chapman's ' Homer ' " in one
hour , and have written nothing in some other hour when he
tried as hard , whereas an average twelve - year - old varies at
the most from nothing up to a composition scoring 50 on
the Hillegas scale . A gifted stock - exchange trader who in
transactions of 10,000 shares a day, averages $ 100 profit ,
may vary from a profit of $ 25 to one of $ 2,500 , whereas a less
gifted trader who averages $ 10 a day on 100 shares in the
same market , it is said , varies over a much narrower range .
Such apparently opposing facts as these are , however ,
not so simple as they seem. If we had a full record of all of
Keats ' hours of equal effort , the production called zero
might turn out to be far above zero . The ideas he had then
might rank in poetic value far above those of the best hours
of the average man . The less gifted trader may vary over
just as wide a range . For example , a still less gifted trader
losing $ 100 on the average , may lose in two days the $ 25
and the $ 2500 that the gifted trader gains . Furthermore ,
we have to consider the alleged common observation that
as one increases his expertness in acting , music , dancing , or
athletic feats , he seems to reduce his variability . Thus a
sprinter who can on the average run 95 yards in 10 seconds
almost never runs less than 90 yards or more than 98 yards
VOCABULARY B
(The method is as in A above . ) The words are :
21. soup 26 comb
22. bag 27. locomotive
23. window 28. door
24. wings 29. cradle
25. envelope 30. sun
DIRECTIONS , ORAL B
Set 1. ( with paper and pencil )
( Unless otherwise specified , the tasks of Directions Oral B

are those of set 1. )
31. " Make a line . " ( If
S fails , show again , but do not
credit . )
32. " Make a cross . " ( If S fails , show again , but do not
credit . )
33. " Turn the paper over and make a ring on the other
side . "
34. " Turn the paper back again , and make a line on the
other side . "
35. " Make two rings down here , " pointing .
36. " See the lines ? Make one more line . " (Credit if
one or two lines are drawn anywhere .)
37. "Make two crosses , like these two . Make one here
and one here , " pointing .
38. "Make the other arm on this man , " pointing .
39. "Make the other leg on this man , " pointing .
40. "Make 2 lines , like these two , " pointing .
Fig . 7 shows the pictures used in connection with tasks
37 , 38 , and 39 , reduced to half size . For task 36 , three
parallel lines two inches long and half an inch apart one
from another , drawn parallel to the side of the sheet , are
shown in the lower left -hand corner of a letter - size sheet.
For task 40 , two such parallel lines are shown , at the top of
a sheet otherwise blank .
7000
LA
FIG . 6. Six rows of pictures such as were used in the Picture Vocabulary
tests : reduced to three -fourths of the original dimensions .
Such hypotheses as this can be nearly reconciled with

our results if the difference between the intellect of the level
of Alpha 25 and Alpha 175 is due to an increase in the
number of factors which is large absolutely , but small in
comparison with the number involved in Alpha 25. Thus if
the difference is 500,000 , but an ability of Alpha 25 involves
5,000,000 then the variabilities around levels of 25 and 125
will be as 1,250,000 and V1,350,000 , or as 1118 and 1162 ,

the second being only 4 percent greater . The reasonable-
ness of this depends upon the location of the absolute zero
of intellect . If
that is ten times as far below Alpha 25 as
Alpha 25 is below Alpha 175 , it is perfectly reasonable .
Another way out of the difficulty is to deny the validity
of the theory that intellect is constituted by the addition of
positive factors only . If the factors in the above illustra-
tions were inhibitive against some maximum amount of
TABLE 3
THE VARIABILITY OF FOUR INDIVIDUALS IN INTELLECT ACCORDING TO A CERTAIN

ADDITIVE COMBINATION OF FACTORS ALL POSITIVE
Amount of Frequencies at 1000 Random Periods

Intellect Α B C D
0 31 1
1 156 10
2 313 44 3
3 313 117 14
4 156 205 42 15
5 31 246 92 15
6 205 153 37
7 117 197 74
8 44 197 120
9 10 153 160
10 1 92 176
11 42 160
12 14 120
13 3 74
14 37
15 15
16 5
17 1
Average 2.5 5.0 7.5 10.0

intellect that would otherwise act , so that the more of them

that acted the less intellect there would be , the relation
between amount and variability would be reversed , the
variability of a man's intellect being as the square root of
the amount by which the man was below the maximum intel-
lect ! There may be , and probably is , some combination of
additive and inhibitive factors making the average intel-
lects of men vary up and down from an amount typical of
the human species ; and this may result in equal variability
for A , who is much below the average , and B , who is much
above it . For example , suppose there are 6 factors , a, b , c ,
d , e , and f, each contributing 1 , and 6 factors , A , B , C ,
D , E , and F , each contributing + 1 ; and that every intellect

is constituted by 6 factors chosen from the 12 ; and that the
momentary conditions of each intellect represent the chance
combinations of its six factors . Then we have intellects
whose averages range from 3 to +3, according to
whether they are constituted by six minus causes , or by 5
minus and 1 plus , or by 4 minus and 2 plus , or by 3 minus
and 3 plus, and so on. All will have the same variability ,
however , the frequencies being in the proportions 1 , 6 , 15 ,
with a mean square deviation of 1.2247 .
20 , 15 , 6 , 1 ,
A consideration of the relative probabilities of various
types of constitution of intellect out of positive and negative
factors would be interesting , but is too speculative to be
profitable for our present purpose . The attainment of
greater intellect by the lack or suppression of negative
factors as well as by the possession and use of positive
factors is at least a possibility ; and will seem highly prob-
able to many .
On the whole , then , we do not need to be especially
skeptical of the experimental findings that the variability in
tests of a half hour from time to time is approximately
equal over the range from , say , the ten - percentile adult in-
tellect to the ninety - five percentile adult intellect .
MEASUREMENT BY WAY OF THE FORM OF DISTRIBUTION OF

INTELLECT IN SOME DEFINED GROUP
If
Ka Ka
T1 , T2 , T3 , and T. , etc. , are intellectual tasks with
K + b K+a+b+ c
which " etc. , individ-
n n " n n
uals of a group of individuals succeed respectively ( K, a,
b, c , etc. , all being positive , K being greater than 0 and the
largest percentage being under 100 ) , we can measure the
differences in difficulty for intellect between T1 , T2 , T3 , T4 ,
C C reversed
FIG . 5. Four surfaces of frequency : A rectangle , Form A , Form C , and
Form C reversed .
etc. , in terms of amount , if we know the form of distribu-

tion of intellect in the group . If , for example , n is 100 , K
6
is 5 , and a , b, c , d , and e are each 10 , the differences in diffi-

culty will be in the proportions shown in Table 5 , according
as the form of distribution of the group is a rectangle , a
surface like A , a surface like C , or a surface like C reversed ,
shown in Table 4 and Figure 5 .
6 Our measures will approximate perfection in proportion as T1 , T2 , T3 , T4 ,
etc. , dependupon all of intellect and nothing but intellect . As has been noted ,
we are assuming this for the present , reserving for full treatment later the
influence of failures of certain tasks to utilize intellect fully , and the influence
of admixture of other factors than intellect .
40. " Make two squares out of these . " ( Two ğ in .

squares are shown , one with the right - hand side
lacking , the other with the lower side lacking . )
FIG . 11 .
FIG . 12 .
FIG . 13 .
Jul FIG . 14 .
SUB - SERIES D
SENTENCE COMPLETION , ORAL D

There being only eight tasks , each is counted as 114 .
1. Boys baseball
("Playing "
.
and " play ball " are called wrong . )

2. The stars and the will shine tonight .
3. Two and one make …………………………
.
4. A boy has and legs .

5. The bird sings ; the barks .
6. Men are than boys .
7. The pulls the cart .
8. Horses are big and
with any reasonable criterion of intellect ; ( 2 ) is different

from the others ; ( 3 ) was constructed without any depend-
ence of the selection of elements or of the scoring system ,
upon the assumption that the distribution of intellect in the
group in question approximates Form A. Find by each test
by actual experiment the form of distribution for the group ,
using the scoring system for each test at its face value .
Find the form of distribution which best fits all these vary-
ing forms . Observe the effect ( upon the form of distribu-
tion ) of reducing the chance error in the scores by obtaining
the form of distribution for the group when two or more
trials with the same instrument are combined for each indi-
vidual . If the best fit distribution is of Form A , and if the
reduction of the chance error does not produce divergence
from this form , we may conclude that Form A represents
closely the form of distribution of the real ability in the
group , as measured by a scale of equal units of difference
in that ability . The general argument is that nothing in the
instruments themselves or their scoring favors this form of
distribution for this group , and that it can not be due to the
chance error , since reducing that leaves it unimpaired .
The details of the argument and the evidence are pre-
sented in Appendix III. They demonstrate that for Grades
from 6 to 12 , and probably for freshmen in colleges of equal

standards of admission , the form of distribution of the pop-
ulation of a grade , when perfectly measured in respect of
the ability required for success with standard types of intel-
ligence tests , in truly equal units , will be unimodal , sym-
metrical , and very closely of Form A , the ' normal ' proba-
bility surface , the equation of whose bounding curve is the
1 -x2
exponential curve y = e where σ is the mean
σν2π
square deviation.
The critical reader should examine Appendix III with
especial care . The method of measuring the intellectual
difficulty of tasks which we adopted for our actual scale con-
struction is based on it . It also provides support for certain
features of previous work which up till now has been taken

on faith . Appendices I and IIare perhaps of greater theo-
retical importance , but Appendix III
is fundamental for
present and future practice in mental measurement .
We can then measure the difficulty of any intellectual
task for pupils in any one of these grades by the percent of
the group succeeding with it, as shown in the illustration
that follows :
3190 pupils in grade 9 were tested with four tasks in com-
pleting sentences . The percentages succeeding were re-
spectively 60 , 30.5 , 46.1 , and 37.1 . We assume that these are
intellectual tasks , that is , that success with each depends
upon intellect.
The form of distribution of the intellects of the group
-
being Form A,
a percentage correct of 60 corresponds to a
division of the group at -
.25330 , that is , at — .25330 of the
mean square deviation of the group ( in the ability mea-
sured in truly equal units by that task ) below the average
or median of the group ( in the ability measured by that
task ) .
.51010 +.09790 , and +.32920 have similar meanings
,
for the difficulties of tasks 21 , 22 , and 23 .

The differences in difficulty between the tasks are
21-20.7634 , 21-22.4122 , and so on , in truly equal units ,
unity being taken arbitrarily as the mean square deviation
of the group in intellect.
MEASUREMENT BY WAY OF THE FORM OF DISTRIBUTION OF AN

ARRAY IN A CORRELATION TABLE
The fourth method of attacking our problem uses , as the
group whose form of distribution is to be determined , the
population comprising one array in a correlation table of
the sort shown in Table 6 , where the individuals are ar-
rayed under their scores in some examination symptomatic
of intellect . Each array consists of two compartments rep-
resenting the two scores ( Failure and Success ) attainable
in the intellectual task whose difficulty we wish to measure .
For example , we might

have data concerning success with
the task in question from 1000 persons each scoring 30 in
Army Alpha , from 1000 persons each scoring 35 in Army
Alpha , from 1000 persons each scoring 40 in Army Alpha ,
and so on . Or we might have data concerning success with
the task in question from 1000 persons scoring Mental Age
8.0 in the Stanford Binet , from 1000 scoring Mental Age
8.5 in the Stanford Binet , and so on .
If both the total score and success in the task depend
upon intellect , and nothing but intellect , the latter being
one of the varying manifestations of intellect of which the
former represents the average condition , the form of dis-
tribution of the intellects measured in an array in such a
TABLE 6
THE CORRELATION OF SUCCESS IN TASK 281 WITH AVERAGE SCORE IN A

TOTAL SERIES OF INTELLECTUAL TASKS ."
Score in the total series

Score in
task 281. 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31
Wrong 14 44 72 94 151 213 259 274 281 187 75 55 6 5 1

Right 2 4 14 19 38 60 111 203 265 302 223 260 90 39 12
correlation table , measured in truly equal units , will be

symmetrical and approximately ' normal . ' For they are
a random sampling from the combined distribution of cer-
tain individuals closely alike in average intellect , when all
the variations of each individual from time to time are
taken ; and we have shown that each of these individuals '
distributions is symmetrical and approximately " normal . "
The use of such an array is in fact a convenient means
of applying our knowledge of the form of distribution of
the variations of an individual in intellect . It is imprac-
ticable to obtain a hundred trials of an individual with an
7 The entries of Table 6 are genuine , but the total series is not a series
representing all of intellect , nor is the score in it an average of many trials .
Such data are not available . The " Score in the total series " in Table 6 is in
fact the score in one trial of a one - half hour test of certain features of intellect .
8 We shall later see uses of other sorts of arrays .
6
intellectual task ; and even if we did , the results would be

hard to interpret because of possible effects of practice . It
is possible to find a hundred individuals who are substan-
tially identical in their average performance at intellectual
tasks , and test them all once with any given task .

The measurements of the difficulty of one intellectual
from the average of
one such array , expressed as a multiple of the variability
of that array , can be made approximately commensurate
with measurements of the difficulty of another intellectual
task in terms of the distance + or from the average of
the corresponding array , expressed as a multiple of its
variability . For we have shown that the variability of an
individual ( and so of such an array ) in intellect is approxi-
mately the same regardless of his average amount of in-
tellect . Consequently the two multiples are of approxi-
mately the same unit and the distance between the two aver-
ages of overlapping arrays can be measured in terms of this
same unit . If two arrays do not overlap , we can bridge
the gap by inserting data from intermediate arrays which
do form a series of overlapping arrays .
THE DEFECTS OF THE MEASUREMENTS SO FAR DESCRIBED
We have determined the approximate form of distribu-

tion of a grade population , from Grade 6 to Grade 12 , in re-
spect of level of intellect at one time , if that were measured
in truly equal units . We have done the same for a popula-
tion ( an array ) characterized by identity in average of in-
tellect measured by a random selection of times . By an ex-
tension and refinement of the methods which we have used ,
this could be done with greater precision .
If all that we require for the measurement of the intel-
lectual difficulty of tasks is to secure a group of known form
of distribution in intellect when measured in truly equal
units , whose members we may test with the tasks in ques-
tion , the problem is solved . Unfortunately more is required .
The chief defect in our procedures is that the difficulty which
we measure by the percentages of our group which succeed

is not pure intellectual difficulty. Any such task as solving
an arithmetical problem or completing a sentence or obey-
ing a command is deficient by not involving all of intellect ,
and often also by involving other factors than intellect .
From the percentage of a group of known distribution in
respect of intellect , which succeed with it , we can derive a
close measure of its difficulty , but not of its intellectual diffi-
culty . Although this has not been understood in the past ,
it can easily be realized by considering cases like the fol-
lowing : A group of known distribution in respect of intel-
lectual level measured in truly equal units , is tested with ( a )
leaping over a certain hurdle , ( b ) distinguishing a certain
pitch from one higher , ( c ) spelling a certain word , ( d ) giv-
ing the opposite of a certain word , and ( e ) giving the oppo-
site of a certain other word . The percent of success is
equal for a , b, c , d , and e , being , let us say , 40 , so that each
of the five tasks is +.2533 S.D. The five are not equal in
intellectual difficulty , however . Common sense tells us this ;
and the verdict of common sense is a crude intimation of
the scientific fact that for ( a ) the + .2533 S.D. means .2533
times the S.D. of the group in ability to leap that hurdle
above the mean of the group in ability to leap that hurdle ,
whereas for (d ) , the +.2533 S.D. means .2533 times the S.D.
of the group in ability to think of the first opposite above
the mean of the group's ability to think of that opposite .
Ability to think of the second opposite may conceivably
differ from ability to think of the first opposite by involving
much more of intellect , or much less of non - intellect , or both ,
in the same way that the ability to think of the first opposite
differs from the ability to leap a hurdle . If
we take the
tasks chosen as intellectual tasks and put in any of the stock
intelligence examinations , they will so differ . This has been
abundantly proven by investigations which will be reported
in Chapter IV . Moreover , no one of them will measure all
of intellect and nothing but intellect .
In fact , no one short task does or can involve all of intel-
lect and nothing but intellect . Any one short task measures
only a fraction of intellect and is influenced by other forces

than intellect . That is , any one short task measures intel-
lect plus an error . The nature and amount of this error
must be considered in connection with any procedure for
estimating the intellectual difficulty of a task from the per-
centage of individuals who succeed with it ."
There are other hidden assumptions and weak or even
missing links in the argument by which we proceed from
knowledge of who and how many can do a task , to a meas-
ure of its intellectual difficulty . In the next chapter we shall
expose these , subject the entire argument to a much more
rigorous treatment , and seek to remedy the defect noted
above and such others as are found .
The exposure of this defect should not diminish our use of the general
procedure of inferring degree of difficulty from percentage of failures in a dis-
tribution of known form . On the contrary , now that we are aware of the defect ,
we can make much better use of the procedure than when we were ignorant of it .
As we shall elsewhere show in detail , if we replace a single task by a composite
" ""
of forty tasks , and use twenty or more right as our mark of success , we can
use the procedure with better results than have ever been obtained hitherto .
CHAPTER III
THE MEASUREMENTS OF THE INTELLECTUAL DIFFICULTY OF
TASKS AND OF LEVEL OF INTELLECT : MORE
RIGOROUS AND EXACT METHODS
In the two previous chapters we have operated with pro-

visional and somewhat vague definitions and inexact as-
sumptions , largely in order to maintain continuity with
what has been done to date in the measurement of intellect .
It is now necessary to treat the whole matter of intellectual
difficulty and level of intellect more rigorously .
We have assumed ( 1 ) that there is such a quality or
characteristic of man as altitude or level of intellect ; ( 2 )
whose amount or degree is measured by the height at which
it can attain success with a series of intellectual tasks
ranked for difficulty ; (3 ) that the same individual differs
in the amount or degree of it which he has available from
time to time ; and ( 4 ) that different individuals differ in
the amounts or degrees of it which they have available on
the average . ( 5 ) We have defined intellectual tasks only
loosely and vaguely as those in which men esteemed very
intelligent differ most from men esteemed very unintelli-
gent . ( 6 ) We have defined intellectual difficulty only
loosely and vaguely as that characteristic of a task, an in-
crease in which reduces the number of intellects who can
succeed with it , eliminating those esteemed unintelligent
more rapidly than those esteemed intelligent .
Since we are treating intellect as the ability to perform
intellectual tasks , our primary need is a clearer and more
exact notion of intellectual tasks . We can reach this in
either of two ways . The first is by assuming that certain
abilities , such as to understand directions , or to know facts ,
or to use relations of likeness , part and whole , actor and
acted upon , genus and species , and the like , or to use facts
59
together , and certain tasks which represent them , are as a

whole intellectual . We must then describe these tasks , and
the credit or weight to be attached to each , precisely , and
put them in a total series in such form that an individual
intellect can attempt them all .
The second is by assuming that the ranking of individ-
uals in an order from idiots to Aristotles for amount of in-
tellect¹ by some defined consensus of opinion is valid . We
must then describe this consensus and the method of its
operation .
If
we take the former way , we may attach the term
" intellectual task " to any selection from the total series
which, when treated in the same way , gives measures for in-
dividuals which correlate +1.00 with measures from the
total series . This task would be just as intellectual as the
total series , would involve just the same abilities as it in-
volved and no others . Everything would be rigorous and
precise after the selection of the tasks and arrangement for
scoring them . In practice a selection which gave a correla-
tion slightly under 1.00 might be accepted as substantially
an intellectual task . '
Further , if any selection from the total series , when
treated in the same way as the total series , correlates as
closely with the total series as its own self correlation per-
mits , that selection is an intellectual task . Its failure of
1 Amount of intellect means here the average amount which the individual
has available over a period long enough to be representative of him .
2 Let the group be measured a number of times by the total series i and
by the task t . Let any two measures by i and
r₁₁¹2 be the correlation between
let
rtit2 be the correlation between any two measures by t . Let
Itill be the
correlation between any measure by t and any measure by . Let It i
be the
correlation between the average measure by t repeated indefinitely , and i re-
peated indefinitely .
Then by the well -known attenuation formula ,

It, 14
Ttwi » —
VI₁₁₂¹¹2
and if is no less than the geometric mean of It₁t2 and tool will be
1112
1.00 or perfect correlation . Since r₁₁₂ in a group of wide range in intellect will
approximate unity , we may set as the requirement that , in a group of wide range
in intellect , It ' should be little if any less than Vrt₁₂
perfect correlation is due entirely to the fact that the intel-

lect of the same individual varies from time to time , not to
any intrinsic inadequacy or irrelevance .
If we take the latter way , we may attach the term " intel-
lectual task " to any task or collection of tasks , the score in
which correlates +1.00 with the ranking of individuals in
an order for intellect by the consensus , or correlates as
closely with that ranking as its own self - correlation permits .
By collecting such tasks we may obtain a total series which
may then be used as a criterion in the same manner as a
series derived by the other method .
These two procedures are more definite and systematic
and rigorous forms of what has been done in test construc-
tion . Psychologists have on the one hand taken tasks which
they assumed to be intellectual and have put samplings of
them into teams of tasks . On the other hand they have as-
sumed that a certain validity attached to rankings by
teachers , by the sifting process of advancement in school ,
or by other forms of consensus , and have selected those col-
lections of tasks which showed high correlations with such
a criterion .
What abilities and tasks shall be treated as intellectual
is essentially a matter of arbitrary assumption or choice at
the outset , either directly , of the abilities or tasks them-
selves , or indirectly , of the consensus which provides the
criterion . After the first choice is made , tasks not included
in it , and even not known , may be found to correlate per-
fectly with the adopted total , and so to be " intellectual " ;
but their intellectualness is tested by and depends on the
first arbitrary choice . Had a different first choice been
made , they might not be intellectual . This arbitrariness is
a sign of weakness , but it is at present unavoidable . We
have to define intellectual tasks as best we can , and trust
that future scientific uses of the definition will improve it .
We shall see later that the arbitrariness is greatly tempered
by certain guiding principles and facts , and that a total
series of intellectual tasks can be defined so as to represent
a fairly clear , unified , coherent feature of human life , suit-

able for theoretical treatment and of great practical im-
portance .
INTELLECTUAL DIFFICULTY
We are now in a position to make the concept of intel-
lectual difficulty more rigorous and exact . Having , by
either method , derived a series of tasks ( N in number )
which as a whole are intellectual , we define the group in
respect of whose members the difficulty of the task is to be
determined , ( for example
as ten thousand taken at random
,
from all living human beings twenty years of age ) , test each
individual of the group with each of the N tasks , rank the
tasks in order by the percents succeeding with each , and
divide them into x sub - series³ ( called D1 , D2 , D3 , etc. ) in
accord with the ranking , D , containing the " easiest , " D₂
the next " easiest , " and so on .
If the score in eachof these sub - series of tasks gives cor-
relations of 1.00 ( or as high as its self - correlation permits )
with the total series , we can define intellectual difficulty as
that feature , which D1 , D2 , D3 , etc. , have in increasing
amounts . They differ in nothing else of consequence to our
inquiry , the score in each being determined by all of the
intellect defined by our total series and nothing but that
intellect .
The attainment of such sub - series may conceivably be

an impossibility . It may be that, no matter how large N
(and consequently N/x ) is , the sub - series of tasks at some
points in the success - frequency ranking may fail of perfect
correlation with the total series . The kind of tasks chosen
as intellectual may , for example , vary in such manner that
all of even a " small " intellect can not be utilized without
tasks from the very hard end ; or in such manner that non-
intellectual factors can not be eliminated or equalized for
all twenty -year - old individuals without tasks at widely sep-
3 For convenience of exposition we will assume that the number of tasks
in each sub -series is the same , though the argument will hold regardless of the
size of the sub -series .
arated points of the success - frequency ranking . This is a

matter for experimental determination after the total series
has been chosen and the group of intellects in respect of
which difficulty has been defined has been chosen . The re-
sults of such experiments are of great significance , inform-
ing us of the degree to which amounts of intellect as defined
by the total series do represent increases in the same kind
of thing , and are amounts of some unified , coherent fact in
nature which can properly be isolated in thought from non-
intellectual factors . Our experiments on this matter will
be reported elsewhere , but we may note now that they indi-
cate that intellect has a rather high degree of unity and
consistency and independence of non - intellectual factors ;
and consequently permits a fairly close approximation to
sub - series of tasks which , as total sub - series , do approxi-
mate to perfect " intellectualness , " while differing enor-
mously in difficulty . We proceeded by an approximation
to the first method , and later checked our choice by an ap-
proximation to the second method . In our choice we were
guided by the following considerations :
( 1 ) Of psychological theory :-(

a ) that responding to parts
or elements or aspects of situations is more " intellec-
tual " than responding to gross total situations ; ( b )
that responding to parts or elements or aspects which
do not present themselves separately to sense but must
be abstracted is more intellectual than responding to
those which do ; ( c ) that responding to relations be-
tween objects is more intellectual than responding to
.
objects ; ( d ) that , in particular , responding to so - called
subjective or logical relations , such as likeness and dif-
ference , is more intellectual than responding to the so-
called objective relations of space and time ; ( e ) that
organizing several mental connections or habits to se-
cure a certain result , " thinking things together , " as
James put it , is more intellectual than using one habit
In Appendix IV and Appendix V.
at a time ; ( f ) that responses to novel situations are

likely to be more " intellectual " than responses to
familiar situations .
( 2 ) Of the theory of measurement :-(
a ) that the tasks
representing any one ability should be capable of very
fine gradation from very easy to very hard ; ( b ) that
they should be capable of very wide extension by alter-
nates at any degree of difficulty ; ( c ) that , so far as pos-
sible , any one ability should represent in some real and
useful sense something varying only in amount , so that
the different degrees of it might properly be repre-
sented by numbers .
(3 ) Of common sense :-( a ) that the tasks should be from
among those which had high standing on the basis of
correlations with reasonable criteria ; ( b) that they
should be convenient for use in the actual measurement
of intellect ; ( c ) that they should be tasks concerning
which subjects for experiment were obtainable .
Over and above the narrowness due to these considera-

tions , our choice is also deliberately narrow . We have not
included any tasks involving responses to actual human be-
ings or to material objects present to sense tasks of what-
has been called social intelligence and mechanical intelli-
gence . Our tasks all concern responses to ideas and sym-
bols , especially words and numbers . The reasons for this
need not detain us here . Also we limit ourselves to tasks
which are intellectual for a group of persons bred in the
United States and aged twelve or over. The reasons for
this narrowness may also in general be omitted .
5 The tasks will very probably serve to measure intellect for younger ages
even more accurately than tasks now in use , but we have not demonstrated this
to be so .
6 The chief reason was that the measurement of intellect in children up to
twelve or fourteen and the definition of the measurement by an age -scale are
in a far more satisfactory condition than the measurement of intellect at older
ages and at higher levels.
INTELLECT CAVD
In view of theseelaborate and arbitrary restrictions , the

intellectualness of our total inventory of tasks , and the
intellect whose level or altitude , range or width , and facility
or quickness it measures , will be called hereafter Intellec-
tualness CAVD and Intellect CAVD ( the symbol CAVD
refers to the four series of tasks which constitute it com- -
pletions , arithmetical problems , vocabulary and directions ) .
The total series of tasks concerns four lines of ability :"
C. To supply words so as to make a

statement true and sensible .
A. To solve arithmetical problems .
V. To understand single words .
D. To understand connected dis-
course as in oral directions or
paragraph reading .
The arrangement of scoring is such as to attach equal

weight to each of these four varieties of tasks .
The whole series is put into a rough approximate order
of intellectual difficulty by the methods described on pages
39 to 56 of the previous chapter . Consequently all the
single tasks or task elements of any one sub - series are of
somewhere nearly equal intellectual difficulty .
Each single task is scored 1 ( right ) or 0 ( wrong or
omitted ) . The number right at each level , that is in each
sub - series , is recorded . The time required for each task
may be recorded , if desired . Selections of forty single - task
elements from each of certain sub - series of the total series
are shown below , making composite tasks A, B , C, D, N, O,
P, and Q.
-
7 We shall sometimes use also Intellect CAVDI , which is constituted
by
including a fifth sort of task to understand and answer questions which re-
quire information about such facts as are considered by the world to -day worthy
of study in school and of record in encyclopedias ; plus organization thereof
and sagacious inference therefrom .
SUB - SERIES A
SENTENCE COMPLETION , ORAL A
1. You are sitting on a
2. We take a ride on the
3. At night you sleep in
4. You like to drink
5. We get up in the
6. Mary has a ring on her
7. You wear gloves on your
8. The snow is on the
9. We go to church on
10. You wear a on your head .
ARITHMETIC , ORAL A
11. Counts 2 pennies . ( Binet procedure , but credit for
success 2 of 3 trials . )
12. Counts 3 pennies . ( Credit if successful in 2 of 3
trials . )
13. "
Show me 2 pennies . " (Credit if successful in 2 of
3 trials . )
14. "
Show me 2 pennies . " ( Credit if successful in 3 of
3 trials . )
15. Recognizes 2 fingers . ( Credit if successful in 3 of 5
trials . )
16. " Show me the littlest pencil ; show me the littlest
one of all , " showing 3 . ( Credit if successful in 2 of
3 trials . )
17. " Show me the littlest square ; show me the littlest
one of all , " showing 3. ( Credit if successful in 2 of
3 trials . )
18. " Show me the biggest square ; show me the biggest

one of all , " showing 3.
19. Adds unseen , 1 plus 1.8 ( Credit if successful in 2 of
3 trials . )
20. Subtracts unseen , 2 minus 1.8 ( Credit if successful
in 2 of 3 trials . )
8 In adding unseen , as 2 plus 1 , for instance , the procedure is as follows :
2 pennies are shown , the subject answers ( rightly or wrongly ) the question ,
VOCABULARY A
( A row of 5 small pictures like those shown in Figure 6

is put before the person who is being measured . He is told
to " Show me the horse , " or " Put your finger on the
horse . " ) The words in A are :
21. pitcher 26. baseball

22. man 27. girl
23. string 28. train
24. apple 29. socks
25. violin 30. dog
DIRECTIONS , ORAL A
31. " Make a ring , like this , " showing act .
31. " Make a line , like this , " showing act .
33. " Make a cross , like this , " showing act .
34. " You can write , can't you ? Show me how you can
write . " (Credit if S imitates effect of writing . )
35. " Put the cover on the box . " ( Credit if S attempts
to do so , turning cover to correct axis . )
36. " Turn
the box upside down . "
37. "
Put the pennies in the box and then shake the box . "
(Have 4 pennies ; credit even if cover is not put on ,
if box is shaken . )
38. " Stand on that paper . ' ( A sheet of paper is left
on the floor . )
39. " Put
your hands behind you . " ( Give while S is
standing . )
40. " If
Make a ring . " ( S fails , show again , but do not
credit . )
" How many ? ''

and the 2 pennies are slipped under the card . Another penny
is shown . " How many ? ''
is answered ( rightly or wrongly ) ; this penny also is
slipped under the card , and he is asked , " How many are under here now } ''
In subtracting , as for instance 2 minus 1 , proceed as follows : The 2 pennies
are shown and the subject answers to " How many ? " The pennies are then
slipped under the card . One is then taken out as he watches , and the question
asked , " How many under here now ? " It is necessary to make sure that he
watches what is done .
SUB - SERIES B
SENTENCE COMPLETION , ORAL B

1. We put stamps on a
2. We cut meat with a
3. When we are sick , we call the
at night .
I can
4. We go to
5. with a pencil .
6. The rug is on the
7. One and one make
8. A dog has four

9. Apples are to
10. Chairs are made of
ARITHMETIC , ORAL B
trials . )
trials . )
12. " One and one make " Add " what ?" if neces-
sary .
14. "Which is the biggest pile ? " showing 13 and 2 pen-
nies . ( Credit if successful in 3 of 3 trials . )
trials . )
16. " Which is the longest of these three lines ? " (Credit
if successful in 3 of 3 trials . )
17. "
Which is the biggest , a baby or a man ? " (Credit
18. Adds unseen , 1 plus 2. ( Credit if successful in 2 of
3 trials . )
19. Subtracts unseen , 3 minus 2. ( Credit if successful in
2of 3 trials . )
20. Subtracts unseen , 3 minus 1. ( Credit if successful in
2 of 3 trials . )
VOCABULARY B
( The method is as in A above . ) The words are :
21. soup 26 comb

22. bag 27. locomotive
23. window 28. door
24. wings 29. cradle
25. envelope 30. sun
DIRECTIONS , ORAL B
Set 1. (with paper and pencil )
( Unless otherwise specified , the tasks of Directions Oral B

are those of set 1. )
31. " Make a line . " ( If S fails , show again , but do not
credit . )
32. " If
Make a cross . " ( S fails , show again , but do not
credit . )
33. " Turn the paper over and make a ring on the other
side . "
34. " Turn the paper back again , and make a line on the
other side . "
35. " Make two rings down here , " pointing .
36. " See the lines ? Make one more line . ( Credit if "
one or two lines are drawn anywhere . )
37. " Make two crosses , like these two . Make one here
"
and one here , pointing .
38. " Make the other arm on this man , " pointing .
39. " Make the other leg on this man , " pointing .
40. " Make 2 lines , like these two , " pointing .
Fig . pictures used in connection with tasks
7 shows the
and
37 , 38 , reduced to half size . For task 36 , three
39 ,
parallel lines two inches long and half an inch apart one
from another , drawn parallel to the side of the sheet , are
shown in the lower left - hand corner of a letter - size sheet .
For task 40 , two such parallel lines are shown , at the top of
a sheet otherwise blank .
‫مرل‬
ha
FIG Six rows of pictures such were used in the Picture Vocabulary
as
6.
.
tests reduced three fourths of the original dimensions

to
.
-
:
SUB - SERIES C
SENTENCE COMPLETION , ORAL C
1. Clouds are in the
2. We send children to school , because they
must
3. We burn in the stove .
4. The is barking at the cat .
5. We wash clothes with and water .
6. Grass is
7. is sweet .
8. We see with our
9. Roses and daisies are
Х Х
10. The eats the mouse .
‫ره‬
FIG . 7. The pictures used with Directions Oral B , 37 , 38 and 39 : reduced to

one -half the original dimensions .
7
ARITHMETIC , ORAL C
trials . )
trials . )
13. " Show me 3 pennies . " ( Credit if successful in 2 of
3 trials . )
14. " Which is the biggest pile ? " showing 10 and 5 pen-
nies . ( Credit if successful in 3 of 3 trials . )
15. " Two and one make "
. (Add " what ? "
if necessary .
trials . )
"
17. Which is the biggest , a chair or a cup ? " ( Credit
18. Subtracts unseen , 5 minus 4. ( Credit if successful
VOCABULARY C
The method is as before . The words used are :
21. camera 26. pistol
22. stationery 27. vase
23. hole 28. stamps
24. corn 29. tiger
25. puppy 30. kennel
Three of the rows of pictures used in this type of test

are shown in Figures 8 , 9 and 10 . It will be observed that
the task sometimes involves a considerable degree of ability
1§ H
in interpreting the pictures .
FIG . 8. Picture used with " lamp . "

B
THE MEASUREMENT OF INTELLIGENCE 73
‫ر‬฀฀
FIG . 9. Picture used with " pond . "
O A
FIG . 10. Picture used with " cork . "
DIRECTIONS , ORAL C
In the actual tasks the drawings have twice the dimen-
sionsof those shown here :
31. " See the square ? " ( A 1 inch square is shown at
the top of a sheet 11 by 81. )
" Make a ring in the square . "
99
32. " Now make another ring in the square . 99

33. " See the ring ? Make a cross in the ring . 99 ' (A
circle 2 inches in diameter is shown near the middle
of the sheet . )
34. " See thecup . Draw a line around the cup . " ( Fig .
11 is shown at the bottom of the sheet . )
35. " Make a ring and a cross up here , " pointing .
36. " Make a cross where the line is . " (A line 21
inches long is shown , parallel with the bottom of the
sheet . )
37. " Draw a line to finish the square . ' ( A half-inch
99
square with the left - hand side omitted is shown . )

38. " Make a cross in here , " pointing to a triangle which
is printed with a square on one side of it and a
circle on the other . The square is 1 in .; the tri-
angle has a base of 1½ in .; the circle has a diameter
of 1 in .
39. " Make a cross X in the square . " ( Fig . 12 is
shown . )
40. " Make two squares out of these . " (Two § in.
squares are shown , one with the right - hand side
lacking , the other with the lower side lacking . )
FIG . 11 .
FIG . 12 .
FIG . 13 .
FIG . 14 .
SUB - SERIES D
SENTENCE COMPLETION , ORAL D

There being only eight tasks , each is counted as 114 .
1. Boys baseball .
"
( Playing " and " play ball " are called wrong . )
2. The stars and the will shine tonight .
3. Two and one make
4. A
boy has and legs .
5. The bird sings ; the barks .
6. Men are than boys .
7. The pulls the cart .
8. Horses are big and
ARITHMETIC , ORAL D
trials . )
12. Recognizes 4 fingers . (Credit if successful in 3 of 5
trials . )
13. " Show me 4 pennies . " ( Credit if successful in 3 of 3
trials . )
14. " How many fingers have you on one hand ? ”
trials . )
trials . )
17. " Which is biggest , 3 or 1 ? " (Credit if successful
3 trials . )
3 trials . )
VOCABULARY D
The method is as heretofore . The words are :
21. tools 26. trumpet

22. fuel 27. cube
23. screw 28. cork
24. angel 29. blade
25. cartridge 30. arrow
DIRECTIONS , ORAL D
The illustrations shown here all have dimensions half
those used in the actual tasks . Each row is also in the
actual tasks separated from the one above and from the one
below it by from 1 to 3 inches .
31. " Make a cross inside the little square . (Fig. 13 is

shown . )
32. " Draw a line to make this a cross , " pointing . (A

thick line in . long, parallel to the side of the sheet
is shown . )
33. " Make a ring on the cup . " ( Fig. 11 is shown . )
34. " See the ring . Make 2 crosses in the ring . " ( A
circle 2 inches in diameter is shown . )
35. " Make a cross on top of the boy's head . " ( Fig . 13
is shown , on a new sheet . )
36. " Draw a line around the big hand . " ( Fig . 14 is
shown . )
37. " Make a cross on the horse . " ( Fig . 12 is shown , on
a new sheet . )
38. " Make a cross outside the big square . " ( A second
copy of Fig. 13 is shown . )
39. "
Make this a circle , " pointing . ( An incomplete
circle with a diameter of inch , lacking the right-
hand quarter , is shown . )
40. " Make a line outside the ring . " ( A circle 2 inches
in diameter is shown . )
The sub - series N , O , P , and Q which follow presuppose
ability to read in the individuals measured by them .
SUB - SERIES N
SENTENCE COMPLETION
Write words on the dotted lines so as to make the whole
sentence true and sensible . Write one word on each inch
of dots.
1. At time was progress
rapid during the last half of the nine-
teenth
2. He will come to the meeting
... the fact .
he rather stay quietly at home .
3. His friends , wished to dissuade him
from this undertaking , asserted that
he followed their advice would with-
draw their support .
4. It would several pages to

contain the list .
5. Standing beside the grave
...
great Englishman
enough for us to know
lived and died and made the his
,
heirs
.
You may safely conclude that you in

6.
yourself the means of at the truth
.
the fact that you disagree with me
7
.
I
,
shall continue aid you
to
At ancient banquets the of the day

.
8.
seems have the chief

of conversation .
9. As the treasure he had come to seek
,
probably existed in his own
it
10. The Declaration affirms

that the Creator all men with certain
inalienable
ARITHMETIC
11. camp has food enough last 300 men months
to
A
How long will

.
last 200 men

it
12. watch was set correct noon Wednesday At M.

P.
at
A
At that rate
.
on Thursday was 15 seconds fast

it it
how much will gain half an hour

in
13. Five sixths equal how many thirds

?
How many quarters quarter equal half half

of
of
14.
a a
How long will take man walk 14 miles the

to
15.
at
it
rate of miles an hour while walking he makes

if
3
three stops of an average length of 10 minutes each

?
A for 5c for 5c for 1.00

F E
6
4 8
2
JI
$
.
B for 5c for 10c for 25c

4
4 3
.
.
for
.
5c 40c per lb. 14c each

C
LK
G
.
for H
.
D 25c 10c each 31c each

3
.
.
etc. are articles costing shown above 1A

as
B
D
C
A
,
,
,
means of 2A means 3A means Sup-

of
of
etc.
A
A
A
1
3
,
,
ply the missing numbers in lines 16 to 20 as shown in lines , I

II, and III. Use the bottom of the page or another sheet
of paper to figure on .
I. 3 E cost as much as 3 B.
II.
22
3 D cost just as much as 4 J.
III . 1 H cost 4 times as much as 1 A.
I J.
J
16. 3 costs as much as 2
17. 1 66 66
1 D.
18. 6 F 66 66
8 A.
19. 2 B 66 66
4 C.
20. 5 oz . G 66 66
2 F
VOCABULARY
Look at the first word in line 21 . Find the other word in
the line which means the same or most nearly the same .
Write its number on the line at the right side of the page .
Do the same in lines 22 , 23 , 24 , etc. Lines A , B , C , and D
show the way to do it . Do all the lines you can . Write only
one number for each line .
4. beast 1 afraid.......2 words ......3 large.......4 animal.....5 bird 4
B. baby 1 cradle...... 2 mother ......3 little child..... 4 youth..... 5 girl
C. raise 1 lift .....3
up..... 2 drag sun.....4 bread..... 5 deluge 1
D. blind ......3 game......4 unhappy......5 eyes

1 man...... 2 cannot see 2
21. sexton 1 cube......2 janitor .....3 compass ..... 4 archbishop .....5 six singers
22. buckler 1 keel ....2 servant.... 3 stag4 shield.....5 scraper
23. animosity 1 hatred.......2 animation ... 3 disobedience ......4 diversity .5 friendship
24. conflagration 1 carnival.....2 celebration..... 3 decoration with flags.....4 contagion......5 fire
25. confidential 1 respectable ......2 secure......3 sensitive......4 secret..... 5 confident
26. scrivener 1 searcher ......2 forger......3 chaplain......4 elerk..... 5 sceptic
27. beaker 1 cup...... 2 binnacle.......3 beak .... 4 slanderer..... 5 bottle
28. emanate 1 populate.....2 free.....3 prominent.....4 rival 5 come ....
29. landau 1 pier......2 coach...... 3 postern .
4 gable.......5 headdress
30. amaranthine 1 jubilant.... 2 bitter......3 maritime......4 ungracious......5 purple
DIRECTIONS AND COMPREHENSION OF SENTENCES

AND PARAGRAPHS
In each set of sentences , check the two which mean most
nearly the same as the sentence printed in heavy type .
THE MEASUREMENT OF DIFFICOLTY 79
31. Show me the man you honor . I know by that

symptom better than any other , what you are
-
,
yourself . (Carlyle . )
A man is known by the company he keeps .
Tell me what you've done and I
will tell you
what you are .
A man is known by his idols .
...
I
Show me your chips and will tell you whether
you are a good woodsman .
32. It is one thing to see that a line is crooked and
another thing to be able to draw a straight one .
It is one thing to see the mote in our neighbor's
eye and another to see the beam in our own .
Those who see mistakes cannot always correct
them .
As the eye is trained to accuracy the hand de-
velops skill .
We may recognize faults that we are unable to
overcome .
33. If we agree that morality is what is social and im-
morality , anti - social , we shall be led to inquire
of any course of action how it affects the welfare
of society . (Pearson . )
To judge whether an action is good , we must
investigate its results on society .
An act is moral or immoral regardless of its ef-
fects .
We must judge a man's deeds by his motives .
Acts which are socially harmful are immoral .
34. There was a painter became a physician , where-
upon a citizen said to him : You have done "
well ; for before the faults of your work were
seen but now they are unseen . ” — ( Bacon . )
The citizen indicates that long training is neces-
sary to appreciate true art .
He implies that science is more exacting than art .
He mans to make sport of the medical profession .
The items on this page and all similar items in this volume are from a test
issued by the Carnegie Institute of Technology . We are greatly indebted to
their author , Dr. L. L. Thurstone , for permission to use them .
He implies that doctors may make mistakes

which remain undetected .
Itis easy to be virtuous when one's own conveni-

.-
35.
ence is not affected ( Stevenson . )
Virtue is its own reward .
It is easier to preach virtue than to practice it .
It is difficult to do right when it conflicts with
our inclinations .
We would all be virtuous if virtue were merely
a matter of doing what we enjoy .
36. Don't cross the bridge before you come to it .
Look before you leap .
Don't borrow trouble .
Don't lock the barn after the horse is gone .
Take care of today and tomorrow will take care
of itself .
Read this and then write the answer . Read it again if
you need to .
There is an old saying , " As harmless as a fly ; " and
until recently the fly has been regarded only as an unpleas-
ant but harmless nuisance . Had our forefathers known as
much about flies as we now know, they might have made the
proverb , " As dangerous as the fly . " His origin and his
habits are of the worst sort .
37. Copy the words which mean the same as proverb .
Read this and then write the answer . Read it again if

you need to.
EVERY HOME NEEDS A GARDEN

A MAGAZINE published to promote real gardening .
Most people do not think much about their gardens at this
time of the year , but if more people did , there would be
more good gardens . If you live in the city where space is
at a premium , we provide pleasure for you by suggesting
how to grow flowers indoors . If you live in the country
and have a garden and do not experience the satisfaction of
seeing things grow as a result of your own efforts then
you need the X.Y.Z. magazine .
-
At what time of year do you think this advertise

38.
ment appeared ?
Read this and then write the answers . Read it again if
you need to.
However certain it may seem to be that men work only
because they must , and would avoid labor except for the
food , clothing and luxuries that are its rewards , the facts
may well be to the contrary . It can hardly be the case that
men dislike work because they wish to be utterly idle . For
mere rest , mere inactivity , is not commonly enjoyed . To
have nothing to do is not what men seek . Were that so , we
should envy the prisoner shut up in his cell . men had If
to choose between a life spent at eight hours of work daily
in a factory and a life spent at eight hours of sitting on a
throne without moving hand or foot , many of them would ,
after trying both , choose the former . Activity of body or
mind , at which a man can succeed , is , in and of itself , rather
enjoyed than disliked .
39. What , according to the paragraph , has no appeal
per se?
40. What is it the author of the paragraph suspects men
might choose unless they had tried it ?.
SUB - SERIES O
COMPLETIONS
1. India is rich in of scenery and climate ,
the mountains to vast
deltas raised a few
above sea
2. Undue consciousness often the flow of
expression diffuseness is detrimental
to a clear and exposition of our ideas .
3. Knighthood and Chivalry are words
are nearly not
synonymous .
4. Throughout the river plains of northern India , two

harvests , and , some provinces ,
.are each
82
5. a man time sufficient for

all laudable pursuits , and sufficient for
all generous purposes , he is free
shadow of blame or reproach .
6. Maize contains small a proportion of
nutritious matter it
not for horses which
fast work is
7. The drafting a measure depends
the pains and skill ex-
erted by its
8. is natural that being dissatisfied with
the we should
-9 form a too
estimate of the past.
9. He believed in hard things
because hard .
10. Not do living things grow themselves ,
they produce life
like
ARITHMETIC
11. A factory earns $70 a day for its owner when it is
working full capacity and $ 15 a day when it is work-
ing to half capacity. In how many days will it earn
$ 1,000 if two days out of every three are only half
capacity ?
12. A company marched 120 miles in 5 days . How many
times as fast must they march to cover 90 miles in
three days ?
13. A man started with $ 12,500 and doubled his capital
every year for five years . How much had he at the
end of the fifth year ?
14. An airplane went 60 mi . at the rate of 90 mi . per hour .
It made a stop of 30 minutes . On the return trip it
went half the distance at 100 mi . and half the dis-
tance at 80 mi . per hour . How long was the total
time ?
15. If the dividend were multiplied by 4 , and the divisor
divided by 2 , the quotient would be 40. What is the
quotient ?
16. How long will enough food for 400 men for 120 days
last 250 men ?
17. How many times as big an area has a circle with radius
of 9 as a circle with radius of 3 ?
18. At an average rate of 12 miles per hour for the first
half of the time spent and 10 miles per hour for the
last half of the time spent , how long will it take a
truck to cover 110 miles ?
19. Three fourths equals how many thirds ?
20. A push - cart man buys eggs at 15c . per dozen and sells
them at 15 for 25c . How many eggs must he sell to
gain $ 1.80 ?
VOCABULARY
The directions and samples are the same as in Sub

Series N.
21. gainsay 1 persuade ...... 2 beshrew...... 3 deny......4 profit..... 5 imprint
22. eclogue 1 obituary.....2 a poem ..... 3 carousal..... 4 epigram ......5 portrait
23. cloistered 1 miniature...... 2 bunched 3 arched...... 4 malady..... 5 secluded
24. reciprocal 1 saturnine...... 2 mutual.....3 receptive ......4 morose...... 5 careless
25. accolade 1 salutation...... 2 anchovy .... 3 procession ......4 bivouac ..... 5 acolyte
26. benighted 1 fraudulent .....2 weary......3 insuperable ..... 4 ignorant.....5 venal
27. madrigal 1 song..... 2 montebank.....3 lunatic.....4 ribald...... 5 sycophant
28. pinnace 1 a boat.....2 doublet.....3 pinnacle......4 hold fast..... 5 forfeiture
29. broach 1 dodge ..... 2 clasp ....
3 open.....4 top..... 5 edify
30. nectarine 1 bouillon...... 2 a fruit......3 a jewel.....4 a drink......5 diurnal

AND PARAGRAPHS
There are only six tasks in place of the usual ten . So

each is counted as 13.
31. Better be a big frog in a little puddle than a tadpole

in a lake .
Better the head of an ass than the tail of a
horse .
I had rather be a door - keeper in the house of

my God than to dwell in the tents of wicked
ness .
Better to reign in hell than serve in heaven .

Better to be a beggar in Rome than a prince
in a village .
32. Don't cross the bridge before you come to it .
Don't borrow trouble .
Don't lock the barn after the horse is gone .
Take care of today and tomorrow will take
care of itself .
The paragraph for questions 33 and 34 is the last para
graph in Sub - Series N.
33. What choice is described as an argument that work ,
merely as such , is not always avoided ?.
a million dollars ?...

34. In what respect is a prisoner in his cell like a man with
Read this and then write the answers . Read it again if

you need to.
THE AMERICAN STATE
He who looks at a map of the Union will be struck by
the fact that so many of the boundary lines of the States
are straight lines . Those lines tell the same tale as the
geometrical plans of cities like Leningrad or Washington ,
where every street runs at the same angle to every other .
The States are not areas set off by nature . Their boun
daries are for the most part not natural boundaries fixed
by mountain ranges , nor even historical boundaries due to
a series of events , but boundaries , purely artificial , deter
mined by an authority which carved the national territory
into strips of convenient size , as a building company lays
out its suburban lots . Of the States subsequent to the origi
nal thirteen , California is the only one with a genuine nat
ural frontier , finding it in the chain of the Sierra Nevada
on the east and the Pacific Ocean on the west . No one of
these later States can be regarded as a naturally developed
political organism . They are as trees planted by the for
ester , not self- grown with the help of the seed - scattering
wind .
36. To what may we attribute the similarity between the

plans of certain cities and the arrangement of the
States ?
SUB - SERIES P
COMPLETIONS
1. The monuments of Persepolis
the use incense
as in ancient Persia
as Babylonia .
2. Ever since the hearing before him the
governor giving
spare moment a
of the case .
3. So far the displeasure of
the people by the will of their repre
sentatives , a President generally gains ..
by the bold use of his veto power . It conveys the
firmness ; it shows
has a view and does
to give effect to it.
4. The of character is its ability to
liberty from license .
5. Judicial decisions are of or less author
ity as precedents to circumstances .
6. The deepest difference practically , in the moral

,
of is the difference
the easy -going and the
mood .
7. Ibsen's whole problem it has well been
stated , is the of the individual to his
social and personal
8. In the sixteenth century , was not more
decidedly the land of the fine arts ,
was not more the land of bold theo
logical speculation , Spain was the
of statesmen and soldiers .
9. The human race may be as parcelled

into a of distinct
groups societies ,
greatly in size and circumstances .
10. Farmers brought up in the traditions of the
of New England , on going where close
association and cooperation were to
carry on irrigated agriculture , found that it took a
long and involved
waste to learn to act
ARITHMETIC
Write the numbers and signs in each line below in the 1
proper order , so that they make a true equation as shown
in the three sample lines . Use the bottom of the page to
figure on if you need to .
3 3 6 =+ 3 + 3= 6
Sample lines 4 7 8 20 = + +8 x 7X420
2 3 3 7 18 = + −X ( ) 7 + 2 = 18— (3X3 )
1. 1 3 3 3 3 21 =+ -X ÷ ()
23 XX ( )
.
12. 5 33 =+
13. 13 2 2 2 8 12 + XXX ( )
14. 2 =+ X ÷÷ ( )
= -XX ( )
2 5 10 70
15
J
2 1 4 4 20
A 2 for 5c . E 3 for 10c . for 3 25c .
B 31c . per lb. F 4 for 10c . K 4 for 25c .
C 4 c. per lb. G 50c. per lb. L 6 for 25c .
D 3 for $ 1.00 . H 2c . each . M 6 for $ 1.00 .
A, B , C , D , etc. , are articles costing as shown above . 1A

means 1 of A , 2A means 2 of A , 3A means 3 of A , etc. Sup-
ply the missing numbers in lines 16 to 20 as shown in lines
I, III , and III . Use the bottom of the page or another sheet
of paper to figure on .
I. 2 A cost as much as 3 E.
II
. 1 E costs 13 as much as 1 F.
III. 3D cost just as much as 12 J.
16. 3 F costs as much as 2 J.

17. 3 K 66 66
5 L.
18. 2 K 66 66
1 dozen H.
19 . 2J + 1 M " 66
10 H.
20. 3 A 66 66
1 K.
VOCABULARY
Directions and Samples as in Sub - Series N.
21. monomania 1 flying machine .......2 conceit ......3 one -colored.......4 endogen...... 5 aberration
22. saturnalian 1 reptilian..... 2 impertinent.....3 gloomy..... 4 impregnated.....5 riotous
23. pristine 1 flashing.....2 earlier.....3 primeval.....4 bound 5 green ....
24. quaternion 1 officer...... 2 fourfold system.....3 four - line stanza......4 tremolo ......5 geologic age
25. predatory 1 hasty.......2 ante -dated.....3 rapacious..... 4 foretold...... 5 four - footed
26. persiflage 1 camouflage ......2 wit......3 banter......4 vivacity......5 metaphor
27. encomium 1 repetition...... 2 friend.....3 panegyric..... 4 abrasion...... 5 expulsion
28. abattoir 1 usurpation..... 2 cessation ..... 3 legal desertion ..... 4 slaughterhouse .....5 nuisance
29. meticulous 1 partial .....2 spacious .....3 finical.....4 melodic ......5 tiny
30. largess 1 enormity......2 present ......3 monstrosity......4 amiability......5 size

AND PARAGRAPHS
Only six are included instead of the usual ten , so each
one is counted as 13.
Read this paragraph . Then read the questions . Make

a ( V) check before the best answer to each question . Read
the paragraph again as much as is necessary .
War ship and merchant ship alike clung to the coast-

or if they ventured out to sea , they did so for a voyage to
be counted by the hour , as , for example , from the southwest
of Sicily to the opposite coast of Africa or they relied on
regular trade winds , like the seamen who sailed from the
-
Red Sea to the coast of Malabar going and coming with the
monsoons . In spite of exceptions , more apparent perhaps
than real , such as the voyages of Irish anchorites to Ice-
land , and of the Norsemen to that island , and to Greenland ,
seamanship continued to be the art of the coaster till the
close of the middle ages . Chaucer's sailor has hardly lost
sight of the coast . Such treatises as were written for sea-
8
men were books of pilotage . Examples will be found at the

end of the Hakluyt Society's edition of Hues Tractatus de
globis . The war - ships , Phœnician , Greek , Roman , Norse ,
Byzantine and Italian , throughout the middle ages , used
sails only when not in action . They were rowed in battle ,
and the mast was lowered , or left on shore . Whenever
they could they avoided passing the night at sea . Their
galleys were beached or anchored close to the shore and
the men landed . We know from Thucydides ' narrative of
the expedition to Syracuse that the crews were landed even
for their meals ; from the chronicle of Ramon de Muntaner
we know that this was also the case with the best Mediter
ranean at the end of the 13th century . The
squadrons
Athenians , clinging to the coast , spent two months in going
from Athens to Syracuse . Roger di Lauria , the admiral of
Aragon , when coming from Sicily in circumstances of great
urgency to Catalonia , went round by the coast of Africa
and Spain . When under sails the ships of war and of com
merce alike had , at the outside , very few sails , and gener
ally only one great course , square and slung by the middle
of the yard . It could be trained fore and aft by bowlines ,
so as to enable the vessel to sail on the wind . Under these
restrictions seamanship was necessarily a limited art .
From Marco Polo we learn that the seamen of the China
Sea and of the Indian Ocean were coasters like their Euro
pean contemporaries .
Put a check before two of the following statements

31.
which make it almost certain that the Spanish sailors of the
middle ages were afraid to venture far from land . Check
only two .
Crews landed even for their meals .
They did not stay on the sea at night .
They were afraid of lack of wind .
Lauria went from Sicily to Catalonia by the coast of
Africa and Spain .
Ships had few sails .
Marco Polo's statement .

Treatises on seamanship were chiefly about pilotage .
32. Put a check before the two of the following which

best show the fear of the open sea . Check only two .
Books were on pilotage .
Chaucer's sailor .
Hues Tractatus de globis .
In sailing away from land they relied on regular
trade winds .
Scant supply of food .
The sailors landed at night .
The use of primitive sails .
33. How was the single sail trained so that the ship
would go in the same direction as the wind ?
At right angles to the long axis of the ship .
By bowlines .
Parallel to the long axis of the ship .
How was the single sail trained so that the ship

34.
would go at right angles to the wind?
At right angles to the long axis of the ship .
By bowlines .
Parallel to the long axis of the ship .
35. What fact stated in the paragraph gives a measure
of how near the ships of the middle ages kept to the shore ?
They sailed by the hour .
Chaucer's sailor hardly lost sight of the coast .
Masts were left on shore during a battle .
They returned at night .
The Athenians ' trip to Syracuse .
The crews landed for meals .
Read this paragraph . Then read the question . Make
(
a V ) check before the best answer to the question . Read
the paragraph again as much as is necessary .
The church cantata , solo or chorale is indistinguishable
from a small oratorio or portion of an oratorio . In Bach's
case many of the larger cantatas are actually called ora-

torios .
Many of Bach's greatest cantatas begin with an elabor-
ate chorus followed by a couple of arias and recitatives ,
and end with a plain chorale . This has often been com-
mented upon as an example of Bach's indifference to
artistic climax in the work as a whole . But no one will
maintain this who realizes the place which the church can-
tata occupied in the Lutheran church service . The text was
carefully based upon the gospel or lessons for the day ;
unless the cantata was short the sermon probably took place
after the first chorus or one of the arias , and the congrega-
tion joined in the final chorale . Thus the unity of the ser-
vice was the unity of the music ; and , in the cases where all
the movements of the cantata were founded on one and the
same chorale - tune , this unity has never been equalled , ex-
cept by those 16th - century masses and motets which are
founded upon the Gregorian tunes of the festival for which
they are written .
36. What feature is stated as giving some of Bach's

cantatas extraordinary unity ?
They begin with an elaborate chorus followed by a
couple of arias , and end with a plain chorale .
If the cantata was short, the sermon took place after
the first chorus .
His founding all the movements of the cantata on
the same chorale tune .
The text was based on the gospel of the day .
The congregation joined in the final chorus .
The unity of the service was the unity of the music .
SUB - SERIES Q
COMPLETIONS
1. It must seem to the wisest .
men , when brought into contact with the great things
of nature that they is
nothing to the infini-

tude of they are ignorant .
2. It is a maxim that man
ever written out of reputation by
himself .
3. The American press
above the moral level of the average
good -in
citizen , no country
either expect or find it
so ,-but it is
of the machine politi-
cians in the cities .
4. David Hume founded the

literary school of English historical writing , and
of the more important
doctrines of modern political economy , but also
a paramount influence on the philo-
sophic
eighteenth
5. Queen Anne was much to horseracing ,
and not only royal plates to be

for , ran
for them
6. The mere practical man regards favorably
the results of science ,
the through which these results are
quite superfluous .
7. happens
relations of the Senate and the President are seldom
cordial , confidential ,
he and the majority of
the Senate belong to the same party ,
the Senate and the President are rival powers jeal-
ous
Francis Bacon in his will , " For my
I
8.
name and memory , it to
92
charitable speeches ,
foreign nations , the
next ages . "
9. wonderfully little genu-
ine inventiveness in the and perhaps
of all has been shown in ..
of political institutions .
10. The florid of the debating club or
pomp of the funeral is
frequently used by orators when but
of exposition is desirable .
MATHEMATICS
There being only five single mathematical tasks in Sub-

series Q , each is counted as 2 .
11. Let pp " mean any flat surface enclosed by straight

lines , the n denoting the number of sides it has . Let E
mean equiangular . What is the common name for
Epp¹ ?
12. I
Express in brief form , using , B and D : " The illumi-
nation varies directly as the brightness of the light
and inversely as the square of the distance . " Use
" K times " for " varies as . "
13. Let n = any number
= 1 divided by
66
nr n
66
NR ==10 66 66
n
=the
1
66
ns number raised to the same power as
itself .
S
AR
What does equal ?
ar
14. Let m , m2 , m3 , etc. , be any numbers .

Let n be their number , that is , n tells how many m's
there are .
Let S ( ) mean " the sum of ” .

S ( m's ) ?
What name will you give to n
15. Let A. D. = the average of the deviations of a set of

numbers from their average , disregarding the signs
of the deviations .
Find the A. D. of 6 , 9, 10 , 11 , 14.
VOCABULARY
Directions and samples as in Sub - Series N.
21. radial 1 light......2 agitator......3 straight line......4 root......5 ray
22. sequestrate ..... 22 petition...... 3 horseman......4 confiscate ...... 5 redwood
1 follow
23. tactility 1 tangibility ....2 grace.... 3 subtlety .....4 extensibility......5 manageableness
24. apogee 1 orbit...... 2 nadir.... 3 ellipse
......4 culmination ..... 5 zodiac
25. nugatory 1 candy
......2 belittling ... 3 inoperative ..... 4 lump of gold.....5 hades
26. sedulous 1 muddied .....2 sluggish.....3 stupid.....4 assiduous ......5 corrupting
27. umbel 1 cluster ......2 canopy..... 3 shadow..... 4 pigment......5 ribbing
28. asseveration ......2 oath......3 continuance..... 4 partition.....5 cleverness
1 pluck
29. abjure 1 swear......2 recant..... 3 refuse ...
4 degraded ....
5 illegal
30. auricular 1 golden
......2 heard .....3 jointed...... 4 distinct..... 5 clear

AND PARAGRAPHS .
The paragraph for questions 31 , 32 , 33 and 34 is " The

American State " used in Sub - Series O. There being only
six tasks instead of the usual ten , each is counted as 13 .
31. Two words are used several times to indicate com

parison . Which are they ?
32. Name three states which are like plants which have
grown from seeds spread by the wind .
33. What states may properly be thought of as being what

they are as a result of ordinary political growth ?
34. What part did nature play in deciding that Montana

should comprise certain territory ?
Read this and then write the answers . Read it again if you
need to .
DIRGE IN WOODS
A wind sways the pines ,
And below
Not a breath of wild air ;
Still as the mosses that glow
On the flooring and over the lines
Of the roots here and there .
The pine tree drops its dead ;
They are quiet , as under the sea .
Overhead , overhead
Rushes life in a race ,
As the clouds the clouds chase ;
And we go ,
And we drop like the fruits of the tree ,
Even we,
Even so.
35. What is as still as the mosses ?
36. Three words in the poem indicate comparison . What

are they ?
Some of the sub - series intermediate between D and N
may be found in Chapter VI.
We could have improved this series at the beginning if
our resources for work had been more extensive . We could
improve it still more now with the knowledge which we have
already gained from using it ; and the reader should con-
sider it more as an illustration of the method than as an
ideal series of intellectual tasks . It is , however , a reason-
ably satisfactory series for its purpose , as will be seen .
One possible criticism we may mention , as it concerns an
important general question of method which we have not

yet discussed .
We have defined intellect as that which produces intel-
lectual products , succeeds with intellectual tasks . We thus
include not only the native , inherent capacity which a per-
son has for such successes , but also whatever education has
added thereto , and whatever increment of success with in-
tellectual tasks he has by virtue of working with better in-
tellectual tools . For example , if A can succeed with the
tasks of the first four sub - series shown whereas B can suc-
ceed with only the first three , we credit A with a higher
altitude or level of intellect than B , even though we may
be confident that if B had had the advantages of A, he
would have surpassed him . We are measuring available
power of intellectual achievement without any specification
as to its genesis . A person who has acquired the intellec-
tual tool , reading, probably has a considerable advantage
over one of equal original capacity who has not acquired
that tool , in the harder completions and directions . One
who has studied arithmetic surely has a notable advantage
in many of the arithmetical problems of our series over one
of equal original capacity who has not studied it . This
procedure would be open to criticism if we should assume
that the score made in the series is a measure of original
capacity to grow into or acquire intellect , without proving
that it did so . We shall not ; nothing about the causation
of the ability measured by the series will be taken for
granted .
The procedure will be criticized by others as a failure
to separate original capacity from the circumstances of
training and to select tasks which would measure the former
alone . This is an attractive enterprise , but not , in our
judgment, so important as the measurement of intellect as
it actually exists and works . We also doubt whether it can
be achieved until the latter has been . There is also danger
that , if we include in a series of intellectual tasks only those
in whose accomplishment differences of education can make
little or no difference , we shall have a collection of freakish

puzzles , irrelevant to the actual operations of intellect by
persons twelve years or older in the United States to -day
or possibly have nothing at all .
Whatever be its defects , our series defines intellectual
tasks and provides us with a rank order of sub - series , each
of which represents all of Intellect CAVD and nothing but
Intellect CAVD nearly enough so that its intellectual diffi
culty can be measured by the methods of the previous chap
ter . The number right out of 10C + 10A + 10V + 10D
tasks at any one level correlates almost as closely with the
number right out of twenty times as many tasks represent
ing twenty levels from very low to very high as its own
self - correlation permits . Our proof of this statement is
given in Appendices IV and V. It has to be somewhat ir
regular and roundabout , since we have been unable to ob
tain records from any individuals attempting the entire
series . But it is conclusive .
We have now to consider what theoretical or practical
significance this Intellect CAVD has .
THE RELATION OF INTELLECT CAVD TO THE ABILITIES MEASURED
BY ORDINARY INTELLIGENCE EXAMINATIONS
The ability which it measures is very much the same as

that which is measured by the Stanford Binet , or by the
Otis Self - Administering Group Test , or by the Terman
Group Test for Grades 7 to 12 , or by a combination of these
three , or by the Thorndike Intelligence Examination for
High School Graduates , or by the I. E. R. Tests of Selective
and Relational Thinking , Generalization and Organization .
This last is a selected team of tests representing a general
consensus of psychological opinion concerning symptoms
of intellect . That is , Intellect CAVD is very much the same
as that which is measured by representative examinations
for so - called general intelligence . The evidence for this is
the correlations obtained . We report these briefly .
Using persons sixteen years old or older , with Stanford
Mental Age of from 28 months to 59 months ( all in asy
lums for the mentally deficient ) , we find a correlation of

.73 ( .68 Pearson , .78 Sheppard ) between a summation score
for CAVD and Stanford Mental Age , the details being as
shown in Table 7. This is in fact probably as close as the
self - correlation of the Stanford Binet will permit . ( In the
62 cases for which we have a second trial of the Stanford
Binet , the self - correlation is .53 . ) If
the range were ex
tended to include all persons sixteen years old or older ,
this correlation of .73 would rise to about .98.⁹
At the other extreme we measured twenty adults , all
high - school graduates , chosen from professional and cler
ical workers , with the Thorndike Intelligence Examination
for High School Graduates ( average of two forms ) , and
with an incomplete sampling of Intellect CAVD . The cor
relation is about .95 , the facts being as shown in Table 8.
The self - correlation of the Thorndike Examination score
for this group would be only about .97 , the correlation
of one form with the other being .95 . So Intellect CAVD
is nearly identical with the ability measured by the Thorn
dike Examination .
Clark [ 25 ] , using 180 pupils in Grades 7 to 12 of the
Lincoln School of New York City , finds the intercorrela
tions stated below among ( 1 ) a score based on a selection
of tasks from the arithmetical and sentence completion and
information sections of Intellect CAVDI , ( 2 ) the Otis Self
Administering Group Test , ( 3 ) the Terman Group Test ,
and ( 4 ) the Stanford Binet Mental Age . Raw
Correlation
Part of Intellect CAVDI with Otis S. A. Test .87
Terman Group Test
66 66 66 66 66
.94
66 66 66 66 66
Stanford M. A. .78
Otis S. A. Test with Terman Group Test .88
Stanford M. A.
66 66 66 66 66
.77
Terman Group Test with Stanford M. A. .77
The mean square variation of the random sample of the Army in Stan
ford Mental Age was over 34 months ( Memoirs , p . 392 ) ; that of our group of
178 cases was under 8 months .
98
.7
TABLE
-
THE CORRELATION BETWEEN CAVD SUMMATION SCORE AND STANFORD BINET MENTAL AGE - IN THE CASE OF 178 IMBECILES
,
SIXTEEN YEARS OLD OR OLDER OF MENTAL AGE 28 MONTHS TO 59 MONTHS
.
Stanford- CAVD Score
5
0
Binet M.A. 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95 100 105 110 115 120
.
25-28 mos
""
2
1
1
1
1
1
29-32
""
3
1
1
2
2
2
1
3
2
1
1
4
THE MEASUREMENT
33-36
""
2
1
3
2
1
1
1
2
4
37-40
""
1
2
4
2
6
3
3
1
1
1
41-44
""
2
2
2
3
1
3
5
1
1
3
1
2
1
1
45-48
""
1
1
4
3
3
3
3
5
1
3
1
1
1
1
3
49-52
OF INTELLIGENCE
""
2
1
2
1
1
1
6
3
2
2
2
53-56
""
1
2
2
1
1
57-60
The selection of tasks from the three fifths of Intellect

CAVDI was meagre , so that even these three fifths were
far from perfectly measured . Nor does one trial with an
Otis or Terman or Stanford examination measure per
fectly what would be found if a dozen alternative examina
tions of the same type were used . The correlations of In
tellect CAVD or CAVDI with the ability as measured by a
dozen alternative Otis Tests or Terman Tests would ap
proximate to perfect correlation for the group in question .
The obtained correlations are higher than the obtained
intercorrelations of Otis , Terman and Stanford , and about
as high as their self- correlations.¹
10
One hundred and forty- six pupils at the very beginning

of Grade 6 in one school were tested with a fairly extensive
selection from CAVD , each being allowed time enough to
do all that he could . A summation score was given with
approximately equal weight to C , A , V , and D. These same
pupils had been tested some months earlier with the Na
tional A and B. The correlation between the summation
score in CAVD and the score in the National was about .76
( .81 by Pearson's , and .71 by Sheppard's formula ) . This
is about as close as the correlation between the National
and a repetition of itself . If
we assume that the variabil
ity for a group of constant chronological age 12 is two times
the variability of this selected grade population , the corre
lation for the former would be .92 .
Dr. M. A. May measured a group of about 650 pupils in
Grades 5 to 8 with a composite of our Vocabulary , Arith
metic , Completions , Information and Reading tasks . Pint
ner had measured the same pupils with the National Intel
ligence Examination . The correlation between a rough
summation score for the former and the score in the latter
was .84 . This again is about as high as the self - correlations
of the two would permit .
Sixty - one college sophomores were measured with Army
Alpha and with 70 CAVD tasks . The correlation was .71
10 Just what these self -correlations would be for the group in question is not
known ; but they would certainly not be on the average much above .90 .
100
.8
TABLE
(
)
THE CORRELATION BETWEEN SCORE IN THE THORNDIKE EXAMINATION FOR HIGH SCHOOL GRADUATES AVERAGE OF TWO FORMS
.
AND AN INCOMPLETE SAMPLING OF INTELLECT CAVD
Score in
Score Sampling Intellect CAVD
Thorndike
in 8
2
8
2
4
6
8
2
4
6
8
2
4
of 4
6
8
2
6
4
50
.
Exam 30 40 60 70
1
40
1
45
THE MEASUREMENT
1
50
11
1
55
60
1
65
70
1
75
80
2
1
85
1 11
1
90
2
1
95
1
100
OF INTELLIGENCE
105
1
110
by the Pearson and .81 by the Sheppard formula . The self

correlation of so short a selection from CAVD will not be
above .80 within this group , and the self - correlation of
Alpha will be little , if any , above .80 . So the correlation of
.71 or .76 is about as high as the self - correlations permit .
Three hundred and eighty - eight pupils in Grade 6 were
tested with a composite of stock intelligence tests ( the I.
E. R. Tests of Selective and Relational Thinking , General
ization and Organization ) and with a sampling of about 40
of the completions and about 40 of the arithmetical prob
lems of CAVD , the score for these latter being a summation
of credits . The correlation was .81 . The self - correlation
for the I. E. R. tests in such a group will be not over .85 ;¹¹
it will probably be about .80 for the sampling of C and A.
So the composite of C and A correlates nearly as closely
with the stock test as the reliabilities permit .
Intellect CAVD or CAVDI is then no more limited or
unreal or remote from the practical management of intellect
than the " intellects " which are measured by the scores in
examinations representative of the best present practice .
It is so nearly the same thing as they that what we learn
about it will have an application nearly or quite as broad as
present practice is .
THE HOMOGENEITY OF DIFFICULTY CAVD
We have cured the main defect in the methods of mea

suring the intellectual difficulty of tasks which was brought
forward in the previous chapter . We can make sure that a
task ( always now a composite of many single tasks ) mea
sures all of Intellect CAVD and nothing but Intellect CAVD
by correlating it therewith , and then measure its difficulty .
We have now to consider or reconsider a number of
other questions . The first is whether the tasks whose dif
ferences in difficulty we thus measure do really differ in the
possession of varying amounts of some one thing which are
11 It is .82 for a group of 1,039 boys in Grades 9 , 10 , and 11 , and .86 for
a group of 16 -year - old boys in Grades 9 , 10 , and 11.
properly represented by cardinal numbers . Do the sub

series or levels ( A , B , C , . . . Q ) form a progressive , homo
geneous series enough like a series of lengths or weights or
temperatures to be subject to the laws of mathematics ?
We show elsewhere¹2 that a sub - series of CAVD tasks
at any level of difficulty measures closely the same ability
as the total series ( provided the sub - series contains enough
single tasks fairly to sample the tasks at that level ) . In that
sense Intellect CAVD is nearly or quite the same from its
lowest to its highest levels . It is obviously the same in the
sense that the tasks are throughout to supply words to com
plete sentences , to solve arithmetical problems , to under
stand single words , and to understand connected discourse .
It is also the same in the sense that any person who is
accustomed to think scientifically about intellectual diffi
culty progresses up this series without any sense of shock
or shift or qualitative change . The progress is , of course ,
not so obviously an increase in the amount of one character
istic which does not change its nature , as when one looks at
straight lines of increasing length or cubes of increasing
volume or lifts graded weights . But it seems logically
fairly comparable to one's experience who looks at very
irregular and differently shaped solids , such as cups , shoes ,
babies , wheels , steam radiators , chairs , and motor cycles ,
which form a series of increasing volumes .
Finally , there is a very close correlation between level
-
or altitude of Intellect CAVD and range or width of intel
lect between the degree of difficulty at which a person can
succeed with CAVD tasks and the number of CAVD tasks
that he can succeed with at any specified degree of diffi
culty.13 Range at a given level is entirely measured by
number , is a variable varying in nothing whatsoever save
amount . Whatever correlates so closely with it may be ex
pected also to exist as varying amounts of some one quality
or characteristic .
12 Appendix IV and Appendix V.
13 The measurements of this are reported in Chapter XIII .
These facts probably warrant us in using the one term

Difficulty CAVD to designate the variable , with cardinal
numbers to designate the varying amounts of degrees or in
tensities of it .
Difficulty CAVD is not the same throughout its varia
tions in the sense that any one unit of it can replace or be
interchangeable with any other unit of it , in the way in
which one inch or one cent can replace any other . We can
not put three tasks , each of difficulty 4 , together and have a
task of difficulty 12 , nor can we subtract some part of any
task of difficulty 10 from it so as to leave a task of difficulty
9 and add that fraction of the actual task to a task of diffi
culty 5 so as to make its difficulty 6. The meaning of arith
metical operations upon numbers representing degrees of
difficulty of intellectual tasks must be considered with refer
ence to the realities which these numbers represent .
This is the case , also , with numbers representing many
variables such as volts , degrees of temperature , wave
lengths or ages , to which arithmetic is none the less usefully
applied . Dividing a temperature of 300 degrees ( above the
absolute zero ) by 10 gives 30 degreesin a certain real and
useful sense , but not in the same sense that dividing 300 dol
lars by 10 gives 30 dollars . We cannot take 5 years of age
from 40 years old and use it to make a five - year - old into a
ten -year - old . Multiplying short vibrations will not give
long vibrations .
We shall return to a consideration of the applicability
of arithmetic to the numbers representing different degrees
of difficulty CAVD later , after we have gained more knowl
edge of them and shown more facts concerning the realities
for which they stand . For the present it may serve to note
that the numbers representing difficulty CAVD are roughly
comparable to the numbers representing temperatures in
respect of the meaning and use of arithmetical operations
performed upon them .
9
THE INFERENCE FROM THE FORM OF DISTRIBUTION OF A GRADE

POPULATION IN STANDARD INTELLIGENCE EXAMINATION
SCORES TO THE FORM OF ITS DISTRIBUTION IN
LEVEL OR ALTITUDE OF INTELLECT CAVD
The next matter that needs consideration concerns our
use of the scores in stock intelligence examinations in the
investigations of the form of distribution of the variations
of an individual ; and of the variations among individuals
in Grade 6 , Grade 9 , and Grade 12 ; and of the relation of
the variability of an individual to his average degree of
intellect. The scores in any one of these stock examinations
represent the composite influence of level or altitude , range
or width , and facility or speed , in unknown proportions . In
strict logic we should have used for our purpose a large
number of examinations , each concerned with level or alti-
tude alone , but made by different experts and without prej-
udice concerning the form of distribution of intellect , as
was the case with the material which we did use . The form
of distribution which we really needed for our argument is
the form of distribution in respect of how hard things the
varying intellect of an individual and the varying intellects
of a group can succeed with . It would , however , have been
utterly impracticable to have attempted to have experts
make seventeen such examinations , and to have applied
these to the large number of individuals needed to make the
argument valid . The cost in time and labor would have been
prohibitive . We therefore used the stock examinations
which were available ; and set up experiments to ascertain
how closely the ability measured by these examinations is
correlated with altitude or level of intellect as measured by
the hardest intellectual tasks at which a person attains a
given percentage of successes . Are we justified in infer-
ring the form of distribution of level or altitude of Intel-
lect CAVD from the form of distribution of the ability
measured by these stock intelligence examinations ?
It is impossible to answer this question by a straight
forward experiment in which a group of several hundred
pupils of, for example , Grade 9 should be given the graded

tasks of Intellect CAVD and also be tested with a dozen or
more stock tests of intelligence . The form of distribution
found for the group in level or altitude measured by the
graded tasks of Intellect CAVD depends upon the real dif
ferences in difficulty between the tasks . This is precisely
what we are trying to determine . We must argue indi
rectly , as by the correlation between level or altitude of
Intellect CAVD and score in a stock test of intelligence .
We have already shown that , in general , summation
scores in Intellect CAVD measure very closely the same
ability as the stock intelligence examinations do . Conse
quently we may correlate a level or altitude score in Intel
lect CAVD with a summation score in Intellect CAVD .
Such correlations will have closely the same meaning as
correlations between a level score in Intellect CAVD and
the official score in the stock intelligence examination . If
the correlations are close , the abilities are similar and will
have , in any given group , similar forms of distribution
when measured in truly equal units .
The correlations between altitude or level score in
CAVD and summation score in CAVD are reported in
Chapter 13. They are well above .90 even in groups of
rather narrow range . The correlations between altitude
or level score in CAVD and official score in standard intelli
gence examinations are reported here . They show in gen
eral that if an individual is tested with any sufficiently ex
tensive collection of intellectual tasks representing wide
variations in difficulty , the level which he reaches is closely
correlated with the score which he obtains by a summation
of credits for work done within a certain time , after the
fashion common in the stock examination .
In the case of 146 pupils of Grade 5 , the correlation
between altitude or level score in CAVD and score in the
National Examination ( A + B ) was .72 by one determina
tion of altitude and .665 by an independent determination
of altitude . The two independent measures of altitude or
level correlated .755 . The level score in CAVD thus corre

lates with the National Intelligence score nearly as closely
as with itself ( .691 compared with .75 ) .
Twenty individuals were measured in respect of a sam
pling from Intellect CAVD by the level at which they could
succeed with fifty percent of the task elements , that is , could
do correctly 20 of the 40 single tasks making up a composite
task . Each was measured also by two alternative forms of
the Thorndike Examination for High School Graduates .
The correlations between the level score and the two stock
examination scores are .92 and .90 . One of the stock ex
amination scores correlates with the other .95 ; and the level
score would probably not correlate with that from another
similar sampling from Intellect CAVD more than .97 . So
perfectly measured level or altitude CAVD and perfectly
measured ability in this sort of stock examination would
probably correlate about .95 .
In this same group the correlation between the level
score in CAVD and a summation score in a composite of
C , A , V , and D , including the tasks used in determining the
level score and many others , was .96 . The correlations
between this summation score and the scores in the two
Thorndike Examinations were both .95 . Thus level score
in CAVD correlates with summation score in CAVD nearly
or quite up to its probable self- correlation , and correlates
with summation score in the Thorndike 91/95 as high as
does the summation score from much more extensive
testing .
Dr. John R. Clark secured measurements of one hundred
eighty pupils in Grades 7 to 12 of the Lincoln School in
the Stanford Binet , Terman Group Test and Otis Self - Ad
ministering Group Test , and in rough measures of level or
altitude , range or width , and speed or facility in samplings
from the arithmetic and completion tasks of Intellect
CAVD . He found the correlations with level to be as
shown in Table 9 .
The three stock examinations correlate on the average
.80 one with another , and correlate on the average .76 with
arithmetic level score if the latter is perfectly measured ,

and .64 with completion level score if that is perfectly mea-
sured . They may fairly be expected to correlate almost
perfectly with a level score for the sort of tasks which they
themselves contain .
TABLE 9
THE CORRELATIONS BETWEEN SCORES IN STOCK INTELLIGENCE EXAMINATIONS

AND LEVEL SCORES IN ARITHMETICAL PROBLEMS AND SENTENCE COMPLE-
TIONS . 180 PUPILS IN GRADES 7 TO 12 . DATA FROM CLARK ( '24 ) .
measured
measured
if
.
if
.
Level
Level
Co. Level
Co Level
perfectly
perfectly
Terman
Arith
Arith
Otis
Stanford Binet .74 .77 .57 .57 .65 .65

..
Otis Self Adm .88 .74 .55 .83 .61

&
..
Terman Group .74 .59 .80 .66

A
.
Arith Level .46

.
Arith Level perfectly

.
measured .55
We may then assume with risk only moderate error

of
that achievement with the series CAVD composite tasks

of
represents an important ability closely allied that which

to
such stock intelligence examinations as the Stanford Binet

,
National Otis Terman and Thorndike measure that the

,
level or altitude score attained in Intellect CAVD may

properly be expressed cardinal number that this level
as
a
score measured truly equal units will show rather

in
if
a
,
close approach Form the probability surface

to
in
(
Grade Grade Grade any intervening grade and

or
12
6,
;
9,
that the correlation between person's true level alti-

or
a
tude Intellect CAVD and his score any one of the sub-
in
in
series or composite tasks will be very

B
A
Q
C
(
,
.
.
)
.
nearly as close as the self correlation of the sub series will

-
-
permit
.
We can then measure the differences in difficulty be-

tween any two of these composite tasks for any grade group
between 6 and 12 which is so constituted that all of its indi-
viduals do not succeed , neither do all fail , with either of
the two composite tasks in question .
CHAPTER IV
THE MEASUREMENT OF THE INTELLECTUAL DIFFICULTY OF A
SINGLE BRIEF TASK
In the previous chapter we have shown that if intel
lectual difficulty is defined as the difficulty of a perfectly
representative sub - series of a total series of concretely spe
cified tasks , it can be measured , at least over a range from
tasks which the stupidest children in Grade 6 can do to
tasks which only the brightest of college freshmen can do .
In the present chapter we have to consider the difficul
ties which confront us when the sub - series is only imper
fectly representative of the total . As the title of the chap
ter indicates , we shall emphasize the extreme cases where
the task is a very partial representative ; but the discussion
will provide also for the treatment of any degree of par
tiality and incompleteness .
We have seen that when we know that k percent of a
group succeed with a task t₁ (k being > 0 and < 100 ) we
may express the difficulty of the task as M₁₁ + Aot₁ where
M. is the central tendency of the group in the ability mea
sured by t, σt is the variability of the group in the ability
measured by t, and A is a factor dependent for its sign and
absolute value on k. We have seen that we cannot , without
further knowledge to that effect , assume that M₁₁ is equal
to the central tendency of the group in intellect or anything
else save the ability measured by t₁ ; or that σt is equal to
the variability of the group in general intellect or anything
else save the ability measured by t₁ .
THE PROBLEM IN THE CASE OF SINGLE TASKS , EACH OF WHICH

MEASURES INTELLECT PLUS A MERE SAMPLING ERROR
We have now to consider the possibility of such further

knowledge . Consider it first for cases where t₁ is a repre
sentative sample of intellectual tasks , and the measurement
afforded by t, is a compound of perfectly measured intel
lect and error of sampling , and the error is of the same
109
magnitude for any one t, as for any other , so that the aver-
age of a sufficient number of t₁'s would be a perfect measure
of intellect , so that the correlation of any one t, with any
other will be a constant . For such cases the knowledge
needed is available and M , and o₁ , the central tendency and
variability of the group in intellect , can be computed from
Mt1 and ot₁ , when the amount of the error is known .
Since M , is the average of the N individuals of the

group , each measured in ( t₁ + t₂ + ts + tą •
t )/n , M ,
will , if N is large , approximate closely to Mt1 , Mt2, Mt3 , etc. ,
and any one of these will approximate closely to any other
of them . The effect of the error whereby the estimate of
intellect by any t differs from that by the average of all
the t's is as often plus as minus , and is negligible for our
purposes so far as concerns the central tendency of a large
group . M,
may be taken as equal to Mt₁ .
Because of the sampling error , σ , will always be smaller
than σt Ot will equal Vo ,2 +0.2 , where Ốp
σt σ, is the variabil-
ity of the N individuals each measured by ( t₁ + t₂ + tз + ts
t₁ ) / n , and σ. is the variability of the sampling error ,
dependent upon the variations of t₁ , t₂ , ts , etc. , in any indi-
vidual from the average of t₁ , t2 , ts , etc. , for that individual .
σ. may be computed in various ways from various measures
of the unlikeness of t₁ , t₂, in the same individual ,
tз , etc. ,
such as the correlation of t , with t in the group , or the cor-
,
relation of t , or t₂ or t, with ( t₁ + t₂ + tз t ) n , or the /
differences between t₁ and t₂ in individuals or the variabil-
,
ity of an individual in intellect as estimated first by t₁ , then

by t₂ , then by t , and so on .
Thus o σt Vrta tb or ot = [ Kelley, '23 , formula

Vrta tb
166 , p . 213 ] and since rti = Vrtath [ Kelley , '23 , formula 160 ,
p . 206 ] , σt = 01,1 where

Iti
1 This second formula is presented because it lends itself better to much of
the material at our disposal . It is derived by Kelley directly from Spearman's
formulas for the correlations of sums or averages . The reader of less mathe-
THE DIFFICULTY OF A SINGLE TASK 111
σ₁ = the variability of the group in intellect , which is

here identical with σ,, the variability of the group in ( t₁ +
to + ts
•
t₂ ) n . /
Iti = the correlation between the estimate of intellect by
any one t in question and the estimate of intellect by ( t, +
t₂ + tz
·
t )/n in the group in question .
σ, = the variability of the group in the ability measured
σt
by the one t in question .
rtat the correlation between the estimate of intellect
by any one t and that by any other t .
For
example , assume that the completion task 22 is one
taken at random from a number of completions , each of
which measures intellect plus a similar sampling error , the
average of all of them measuring it exactly .
We found the difficulty of task 22 to be .098 times the
022 of the ninth grade group harder than the M22 of that
group . In accord with our assumptions , we may replace
M22 by M₁ . The correlation between score in task 22 and
intellect may be taken as approximately .40 for the group
in question , since the obtained correlation with a fairly
close representation of intellect is .37 . In place of .098 022
σι
we then put .098
.40 °
The purely intellectual difficulty
of task 22 , freed from the effect of the sampling error , is
now measured as .245 ( σi and can be compared
for
9th grade
)
with that of any other task representing intellect plus the

effect of sampling error for which we have the percent
of
correct responses in this group Thus task 20 which

,
.
matical ability may easily derive from the more familiar formula for the
it
correction for attenuation as follows

,
Consider the ordinary Spearman attenuation formula for our case

,
Iti
It∞
=
i
perfect measures = 1.00 Itool is 1.00

=
Let and intellect Then

of
be
1112
.
i,
1
i,
by hypothesis
.
So Vitt₂
It
=
бр бр Οι
Substituting in we have σt = in this case
ot
Iti
=
Γει
showed 60 % of correct responses in this group , giving

.2533020 , has a correlation with the same fairly close rep-
resentation of intellect just mentioned , of .18 . Its corre-
lation with intellect may be taken as approximately .20 .
Assuming that it is a random sample from a set of tasks
whose average measures intellect perfectly , ' and each of
which suffers an error of equal magnitude , we transmute
" .2533 020 easier than M20 in Grade 9" into

" 2533 019
.20
easier than M , in Grade 9 , " or - 1.27 019. Task 20 is then
1.51 easier than task 22 , the unit of measure being the mean
square deviation of intellect in Grade 9 .
If the error
whereby the ability measured by a task dif-
fers from intellect is a random sampling error , so that per-
fectly measured intellect can be got by merely increasing
the number of tasks strictly comparable to it drawn in the
sample , we can then correct for it , the correction being a
further application of the facts shown by Spearman [ '04 ,
'07, '10 , and '13 ] , Boas [ '06 ] , Thorndike [ '13 ] , and Kelley
[ '19 , '21 and '23 ] .
If thesingle tasks whose intellectual difficulty we wish
to determine measured intellect perfectly , except for such a
random sampling error , we could and should compute
rti ( or Ft (CAVD ) ) for each of them in each group used , and
apply the corrections .
The effect of the correction may be illustrated by cases
where we have reduced the sampling error empirically by
using ten tasks in place of one .
Thus for 250 pupils in Grade 81 , the median of the ten
percents correct for the ten single word tasks and the
percent scoring five or more correct responses out of the
ten was as shown in Table 10 for each of the fourteen 10-
word composites in the I. E. R. A - 2 and B - 2 . Table 10
2 This average will have to be computed from a larger number of t's than
would be needed in the case of the tasks from which task 22 was drawn as a
random sample , since the error is here larger , making the correlation with
i smaller .
also reports the σ values and the σ10 values which corre-
spond to these percents . The percent is more remote
from 50 when we shift from one right out of one to five or
more right out of ten ; and the value in terms of σ10 is more
remote from the median of the group .
TABLE 10.
THE EFFECT OF DECREASING THE ERROR OF ESTIMATING THE DIFFICULTY OF THE

MEDIAN TASK OF A COMPOSITE OF TEN BY THE USE OF THE PERCENT OF A
GROUP SCORING 5 OR MORE RIGHT OUT OF TEN " IN PLACE OF THE
MEDIAN OF THE TEN PERCENTS OF THE GROUP SCORING
" RIGHT " IN THE TASKS TAKEN ONE AT A TIME.
VOCABULARY TASKS IN THE CASE OF 250
PUPILS OF GRADE 82 .
Distance from the Median

Median of Percent Ability of the Group
the ten Scoring In Terms In Terms
Percents 5 or more of o , of 010
Composite 1 93.2 98.0 - 1.49 - 2.05
"" la 93.4 97.6 - 1.51 - 1.98
"" 2 82.6 92.0 - 0.94 - 1.41
"" 2a 87.4 97.2 - 1.15 - 1.91
62.8 74.0 - 0.32 - 0.64
"" 3a 72.3 86.4 -0.60 - 1.10
""
4 55.2 64.4 -0.13 - 0.37
"" 4a 55.6 61.6 - 0.14 -0.30
"" 10 43.0 + 0.15
44.4 + 0.18
"" 5a 43.2 44.8 + 0.17 + 0.13
6 23.4 15.2 + 0.73 + 1.03
""
6a 25.8 19.6 + 0.65 + 0.86
7 15.8 4.4 + 1.00 +1.71
7a 10.8 .8 +1.24 + 2.41
It may be realized more exactly by applying the formula

to a few representative cases . Thus , tasks A, B and C ,
each being done correctly by the same percent ( 80 ) of a
group ( of normal form of distribution ) , but correlating
with intellect to the extent of .20 , .35 and .50 , respectively ,
in that group , will be of intellectual difficulty - 4.208 01 , -
-2.405 01σ1 and -1.683 σ1 , respectively . Tasks C , D and E ,
although done correctly by very different percents of the
group , are of equal intellectual difficulty , their differences
in difficulty being counterbalanced by reverse differences in

intellectualness .
% suc- Correlation Intellectual

Task cessful with intellect Difficulty Difficulty
A 80 .20 -.84160 - 4.20801

B 80 .35
-.84160B - 2.4050,
C 80 .50 - 1.6830 ,
-
-.8416øc
D 60.1 .20 - .25590D 1.2800 ,
E 67.3 .35
-.44820E - 1.2800₁
F 73.9 .50 -.64030 - 1.2800 ,
THE PROBLEM IN THE CASE OF SUCH SINGLE TASKS AS ARE USED

IN CAVD OR IN STANDARD INTELLIGENCE EXAMINATIONS
Unfortunately we cannot be sure that a single task will
measure intellect save for such a sampling error . This may
be best realized by taking our Intellect CAVD as intellect
for the moment , and considering a task made up of 20 com-
pletions , 20 arithmetical problems , 20 words and 20 direc-
tions , all of equal difficulty . The ability measured by such
an 80 - element task , if the
elements are well selected , is ap-
proximately perfectly representative of ability CAVD .
Now if we take one of the eighty tasks at random , we
do not have something which measures what the eighty
together do plus an ordinary error of sampling . One
word -knowledge test does not differ from one arithmetical
problem test in the same way that one arithmetical prob-
lem test differs from another . The total is too varied a
synthesis and the single task is too small a sample for the
latter to represent the former plus an ordinary sampling
error . In the eighty are four different sorts of tasks ; in
the twenty completions there may be four or five which
require knowledge of specialized facts ; amongst these four
or five , there may be one which is very much easier for in-
tellects which have lived in the country than for intellects ,
otherwise similar , which have lived in the city ; and another
of which the reverse is true .
The case is not so much like measuring a man's height
a dozen times and taking one of the dozen to represent
their average , as like measuring his head , his neck , his

trunk , his legs to the knees , his shins , and his feet and
adding the results to get his height . Our measures of in
tellect are inventories ; we combine C , A, V and D as we
might combine a man's real estate , ships , stocks , bonds , ac
counts receivable , merchandise , materials , and cash on
hand . What he happened to own in the way of real estate
in Boston would not in a useful sense represent his total
wealth plus a sampling error .
Assume for the purpose of illustration that : ( 1 ) intel
lect is composed of C and A in equal parts , and is perfectly
measured at the level in question by a task composed of 20
completions and 20 arithmetical problems , the two twenties
having equal weight ; ( 2 ) a task comprising the 20 comple
tions will correlate perfectly with a task comprising 100
completions from which the 20 are a random sample ; ( 3 )
a task comprising the 20 arithmetical problems will corre
late perfectly with a task comprising 100 problems from
which the 20 are a random sample ; ( 4 ) the 20 completions
or the 100 completions will correlate 0 with the 20 arith
metical problems or the 100 arithmetical problems .
If now N individuals composing a group distributed
" "
normally are measured in respect of their success with
a task composed of 40 completions , and if a given percent
succeed with the task ( that is , have 20 or more of the 40
right ) , the difficulty of that 40 - completion task is M40c +
X040c . The correlation between the score in 40C and the
score in intellect , or C + A measured by 20C + 20A , is .707 .
The correlation between the score in 40C and the score in
another 40C is 1.00 .
By our assumptions
020C + 20A
=
= √σ²20c + σ²20A since 120c 20A
- 0.
020C + 20A
= V20200 since 200 = 020A⚫
020C + 20A
2
040C since 040c = 2020c , since r20C 20C
is 1.00 .
020C + 20A = .707 040C .
Οι
But by the formula o .
(0. Vrtitz
we should have
040C
- 020C + 20A
9 giving
V20c 200
020C + 20A
040C
Vī
or 020C + 20A 040C⚫
We see the reason for the discrepancy if we consider

the attenuation formula
TC40 with C20 + A20
TC with (C + A ) » =
VTC40 C40 (C20 + A20 ) (C20 + A20 )
With our present assumptions , rc with (C + A) ∞

is not 1.00 ,
because , no matter how many C's we take , we do not get
all of intellect and nothing but intellect . It is in fact .707 .
So we do not have
VFc40 with C40 C40 with C20 + A20 ,

but
Vrc40 with C40

= Tc40 with C20 + A20
.707
If we substitute rc40 with C20 + A20 for Vгc40 C40 in the

Οι
formula σ = " we have again the erroneous result
Vrtita
020C + 20A
= 040C⚫
Now the correlation between C. and ( C + A + V +

D ) 00or any other form of perfectly measured intellect is not
perfect ; and the correlation between C. and either A. or
Vor D. or Picture Completions or Geometrical Rela
tions 00 is not in fact perfect . In general , if we sample by
taking one small task, it has to be so limited that if we take
a thousand tasks closely like it , the score therein need not
correlate perfectly with the score in intellect , or with the
score in a thousand tasks closely like any other one task
with which we might begin . In particular , no single com-

pletion , or word to be defined , or problem in arithmetic , or
sentence to be comprehended can safely be regarded as dif-
fering from intellect only by a sampling error such as
may be adequately corrected for by
σε Οι
σt = cor σt •
Vrtitz Iti
A single task, t₁ , measures not a large part of intellect
plus a small error due to the action of a large number of
factors of about equal magnitude , but a small part of in-
tellect plus a large error . The latter is due to the action
of factors some of which , like residence in the city , access
to books formal training with arithmetical problems , spe-
,
cial acquaintance with the particular word or sentence or

problem , may be of very great magnitude in comparison
with others .
More generally , o , =σt rt or = is , as Kelley's dis-

σt
Iti
cussion [ p . 213 ] makes clear , true for a case where i is
simply the average of many t's , each of which has closely
the same σ as any other and closely the same rt as any
other . It is not true when we fail to get i by a collection
of tasks however extensive . And no matter how many
completions we take , we shall never get an i made up of
completions and arithmetical problems unless the corre-
lation between sentence completion and solving arithmeti-
cal problems is perfect .
The quantitative importance of having a varied as well
as a large sample may be illustrated by measurements of
the correlation between i , as represented by the summation
score in CAVD ( 40C + 40A + 40V + 40D ) , and Composites
of 10 made up all of C or A or V or D on the one hand , and
on the other , composites of 10 made up of 2C + 3A + 2V
+ 3D or of 3C + 2A + 3V + 2D . In the case of 240 col-
lege graduates the average of the former sort was .59 with
,
a P.E. of ± .028 ; the average of the latter sort was .72

with a P.E. of ± .022 .
It should also be noted that , even if

the correlation be
tween the score in an infinite number of completions and
the score in an infinite number of arithmetical problems
were perfect , so that we got all of intellect and nothing but
intellect as well by a sampling of one type of task as by a
mixed sampling (the reduction of rt , below 1.00 being due
purely to sampling error ) , still the practical difficulties in
the way of applying the correction would make it far wiser
first to construct composite tasks . It is very laborious to
computer for each element . It will be low ( roughly
from .20 to .60 for a group of individuals in the same school
grade ) , and the probable error of a low biserial r is such
that an enormous number of individuals must be tested to
obtain r with a precision such that the probable error is
less than .01 ( from 5,000 to more than 10,000 for r = .40 ,
according as the split of successes and failures is near .50 ,
.50 , or remote therefrom ) .
THE SOLUTION BY THE USE OF EXTENSIVE COMPOSITE TASKS

The only safe and wise course is , then , to make sure that
the tasks whose difficulty we are to measure are alike in the
amount of intellect which each involves , and in the amount
of non - intellect by which each is contaminated , by using
composite tasks each containing many single tasks , repre
senting with proper weight the various aspects or constitu
ents of intellect . The nearer we come to having each of
them measure all of intellect and nothing but intellect , the
safer our course will be .
With composites which differ from i only by the sam
pling error the correction formulas are appropriate . In
proportion as the composite is made to include a large
sampling , the labor of computing given de
rti
or
to
rt
a
t,
,
gree of precision reduced and the reliability the cor

of
is
rection increased With forty element CAVD com

is
-
.
posites for example safe infer from either by

to
= is
,,
it
ot
σ
,
Vrt or by lot
=
σ₁
ot
σ₁
rt
,.
,
,
,
t
In constructing composite tasks whose difficulty will be

truly intellectual difficulty freed from the sampling error
,
by having many tasks , and freed from the constant error

by having a proper representation of all the elements of
intellect, it may be desirable , other things being equal , to
include in any one composite only tasks which would show
approximately the same intellectual difficulty if, by a mir-
acle , all of intellect , and nothing but intellect , could in each
case be utilized for success . The measure of the difficulty
of a composite of n tasks would be more reliable if this
could be the case . The construction of composites of speci-
fied amounts of difficulty would be less a matter of trial and
correction .
So , for this purpose , we may need to measure approxi-
mately something which , for lack of a better name , we may
call the “ intellectual difficulty ” of single tasks , and to know
how close the approximations are .
The facts which we shall present in this connection are
also of importance in estimating the errors in scales³
which have been constructed on the assumption that ot ,,
Ot , ot , ot , etc. , are equal . They are also of importance in
connection with the general technique of selecting single
tasks to make a composite , even if we make no attempt to
select them to be of the same intellectual difficulty , rather
than of the same difficulty .
These facts are the percents of some group succeeding
with the several tasks ( t1 , t2, ts , etc. ) whence we may com-
pute M measures of their difficulty ( Mt , + C , σt ,, Mt , +
C , ot , Mt , + C , σt ,, etc. ) ; and the correlations ( rt , 1, rt , l ,
rt31 ,etc. , between each of many single tasks and intellect
( CAVD or some other defined intellect ) , whence we may
compute the extent to which t₁ , t₂ , tз , etc. represent intellect ,
and so estimate their " intellectual difficulty . "
We have
seen that with a large group , Mt ,, Mt ,, Mt ,, etc. , will be
closely equal . In proportion as rt ,, rt ,, rt , i , etc. , are ap-
proximately equal , σt ,, σt ,, σt ,, etc. , will be approximately
equal , and o , will be approximately the same fraction of
each of them , equalling respectively Vo , + E , Vo₁² + E₂² ,
3 Such as the Buckingham Spelling Scale , Trabue Completion Scales , Van
Wagenen History Scales .
10
Vo₁2 + E²
, where E , or E₂ or E , is the " error " by which
the estimate of intellect by the single task diverges from the
estimate of intellect from a properly weighted sum of all
tasks . E1 , E2 , E3 , etc. , will be approximately equal , if they
produce approximately equal reductions from perfection in
the correlations rt , i , rt , 1, rt , i , etc.
If
, then , we select single tasks which are done by equal
percents of a large group , and also are approximately

equally closely correlated with intellect , we shall have
equality in the sort of intellectual difficulty which we are
discussing . For example , Table 11 shows , in the case of 30
reading tasks , the percents succeeding and the correlations
(biserial r ) with the combined score in two forms of a
standard intelligence examination given a year apart ( the
I. E. R. Tests of Selective and Relational Thinking , Gener
alization and Organization * ) . The facts are given for 668
pupils in Grade 11. Using the facts of Table 11 as our
guide , tasks 10 , 15 , and 24 may be expected to be of approxi
mately equal " intellectual difficulty . " They are approxi
mately equally difficult because the percents succeeding
are respectively 66 , 65 , and 67. They are approximately
equally intellectual because the rt's
are , respectively , .40 ,
.41 , and .38 . We can also balance low degrees of % s ( per
cent successful ) against high degrees of rti so as to get
tasks that would be of equal intellectual difficulty in so far
as the formula is applicable .
Even if it is not desirable to spend time in choosing

tasks which are alike in the + — values of o , as inferred .
from σrtot, it will be very useful to know how much
difference will be shown in the r's
of single tasks in com
pleting sentences , solving arithmetical problems , knowing
word - meanings , following directions or answering ques
tions about a paragraph , giving opposites , possessing and
using information , completing pictures , supplying or se
lecting the proper related term as in the analogies test , and
other stock forms of tasks used in instruments for measure
1.70
4 The self -correlation of this combined score is approximately 1.85 or .92 ,
in this group .
ment of intellect
. For , other things being equal , the higher
rt is the more suitable the task is for inclusion in a com
posite to measure i .
TABLE 11
THE DIFFICULTY AND INTELLECTUALNESS OF 30 SINGLE TASKS IN UNDERSTAND

ING SENTENCES , MEASURED BY THE PERCENT OF 668 11TH GRADE PUPILS
SUCCEEDING WITH EACH , AND BY THE CORRELATIONS OF SUCCESS
IN EACH WITH THE AVERAGE SCORE IN TWO FORMS OF
THE I.E.R. SEL . REL . GEN . ORG . EXAMINATION .
% Suc Unreliability
Task ceeding Iti of rti
or
(
)
DIA 93 .43 ± .07
94 .30 .075
"
4 123
91 .29 .065
' '"
84 .55 .04
76 .36 .045
" "
9 8 7 56
83 .45 .045
75 .43 04
82 .53 .04
82 .54 .04
10 66 .40 .04
11 81 .52 .04
12 69 .48 .04
13 77 .45 .04
" "
14 57 .28 .045
15 65 .41 04
16 75 .48 .04
17 66 .45 .04
18 65 .46 .04
" "
19 75 .40 .045
20 70 .49 .04
21 69 .36 .045
"
22 56 .33 .04
23 54 .35 .04
24 67 .38 .04
" "
25 70 .32 .045
26 66 .52 .04
27 64 .46 .04
28 64 .49 .04
"
29 57 .45 .04
30 64 .53 .035
'"
THE CORRELATIONS OF SINGLE TASKS WITH MEASURES OF

INTELLECT
We have made the computations in the case of 24 other
reading tasks and 55 vocabulary tasks with the results
,
shown in Table
12. We also give in Table 13 the facts for
10 completion tasks , 10 arithmetical tasks , and 10 vocabu
lary tasks , using a group of 240 college graduates . These
data , together with those of Vincent (to be described
shortly ) , make possible a general estimate of how much r
may be expected to vary in the case of single tasks selected
or devised by psychologists as suitable elements of an in
telligence examination .
The obtained correlations vary very widely , but some
of this variability is due to the unreliability of the deter
minations ; and allowance must be made for this in order
to estimate the true variation in r , due to the differences
among single tasks in the amount of i which each involves
and the amount of non - i by which it is contaminated .
Consider first the facts from 99 tasks in reading and
vocabulary , where the percent is between 5 and 95 , in the
case of 668 and 454 pupils , respectively , in Grade 11 , shown
in Table 14. We omit the very , very easy and very , very
hard tasks , since we should measure their difficulty by a
duller and by a brighter group , respectively .
It is obvious to inspection that the correlations vary
more than can be accounted for by their unreliabilities . In
the .40 to .60 group , we have a range from .45 to .52 , in
the .60 to .80 group , a range from .11 to .56 , and in the .80
-
to .95 group , a range from — .02 to .67.5
The variation which we should obtain with the unrelia
bilities cut to 0 by a sufficiently large group is to be found
from σtrue = Vo²obt .
-
o'error . Using medians as central ten
dencies , the facts are :
Otrue .40 -.60 -
V.0217 .0024 or .139 ( n = 20 ) .
Otrue .60 - .80
= V.0092 -.0022 or .084 (n = 35 ) .
= V.0246 or .143
38
.0041
.95
.80
=
n
Otrue
-
).
-
The serial r's Tables and were computed by an approxi

11
12
13
bi
in
5
,
-
mate method They will diverge from r's computed accurately by not over
.
.005 which not of consequence comparison with the variations which we

in
is
,
are considering
.
TABLE 12
PERCENTAGE SUCCEEDING AND CORRELATIONS WITH A CRITERION IN THE CASE OF

24 READING TASKS AND 52 VOCABULARY TASKS : GRADE 11 : n = 668
FOR THE READING TASKS AND 454 FOR THE VOCABULARY TASKS .
Task % 8 Iti Or Task % Tit Or
8
DIA 66 72 .39.045 VB 45 87 .29.07
67 81 .28.045 46 85 .39 " .06
68 83 .31 .05 47 85 .45 " .06
""
69 88 .39.055 48 97 .40.11
70 80 .24.045 49 20 .15 " .06
71 94 .30.075 50 66 .33 .055
"
72 42 .13.05 51 77 .24 " .055
73 80 .23.045 52 73 .52 " .05
74 81 .41 .045 53 26 .13 66 .06
"
-
75 53 .39.04 54 65 .27 .055
" "
77 69 .35.045 55 84 -.02 .07
78 90 .31.065 56 96 .58 " .085
79 73 .32.045 57 44 .48.05
58 80 .48 " .05
DII A 1 47 .17.045 59 90 .36 .07
41 .35.045 60 82 "
.50 66 .05
23 34 .13.05 61 87 .54.06
4 54 .22 .045 62 66 .29 .055
""
""
5 51 .25.045 63 43 .35 " .05

17 22 .31.045 64 85 .04.07
18 .44 .005 65 95 .31 " .095
' "
2
19 49 .45.04 66 91 .34 .07

" "
20 14 .37.055 67 91 .67 .07

21 .12.08 68 74 .11.06
9 4
2222
.61.05 69 43 .05.055
70 45 .36 .05
"
VB 31 92 -.01.085 76 69 .45 " .05

32 92 .27 .075 77 53 .00.055
'"
33 88 .34.055 78 85 .47.06
34 98 .07.135 79 63 .36 " .05
35 83 .14.06 80 78 .40.055
36 60 .18.055 81 43 .18 " .055
37 95 .20.095 82 56 .40 .05
""
38 92 .24.08 83 65 .56 " .045

39 97 .14 .115 84 66 .38 " .05
"
"
-
40 89 .53.065 85 50 .52 .045

"
41 92 .31.075 86 91 .63 " .06

42 91 .29 .075 87 96 .45 .090
"
'"
43 91 .34.07 88 52 .45 " .045

89 70 .41.05
44 97 .50.105 90 73 .37 .055
"
Inthe case of the 30 tasks done by the 240 college grad

uates , the correlations vary from .18 to .90 .
Otrue
= V.0208 -.0059 or .122 ( n = 30 ) .
TABLE 13
PERMILLES SUCCEEDING AND CORRELATIONS WITH A CRITERION OF INTELLECT ,

IN THE CASE OF 240 COLLEGE GRADUATES .
Task Permille's Iti Orti
CO 1 600 .42 ± .07

754 .38 .08
""
775 .55 .07
""
5 234
567 .57 .06
'
L 654 .43 ❝.07
458 .31 .07
" '
521 .18 .08
฀฀฀฀
66 .07
729 .54
6789
733 .24 .08
""
฀฀
2
10 575 .49 .06
' "
AZ 792 .39
66 .08
1232 1
" '
779 .45 .08
฀฀฀
642 .23 ❝.08

467 .20 .08
"
+54
฀฀
421 .33 ❝.07
679 .30 " .08
฀
฀฀฀
671 .29 .08

" "
642 .42 .07

67899
700 .43 .07

' "' "'
10 600 .47 .07
VA2 61 671 .59 .06

"
62 496 .46 ❝.07

❝.07
฀฀฀
63 775 .51
64 983 .90 ❝.09
65 308 .48 .07
"
66 396 .42 ❝.07

67 650 .41 ❝.07
68 292 .45 ❝.07
69 650 .31 .08
"
70 192 .28 .09

TABLE 14
THE CORRELATIONS ( BI - SERIAL R ) OF EACH OF 99 READING AND VOCABULARY

TASKS WITH INTELLECT ( I.E.R. SEL . REL ., Gen. Org .) , GROUPED
ACCORDING TO THE PERCENT SUCCEEDING WITH
THE TASK .
5 to 20 20 to 40 40 to 60 60 to 80 80 to 95
Iti Orti Iti Orti Iti Orti Iti Orti Iti Orti
.37.055 .13.05 .28.045 .36.045 .43.07

66
.61 " .05 .31 " .045 .33 66 .04 .43 66 .04 .30 .075
.15 " .06 .35 " .04 .40 " .04 .29 66 .065
-.13 " .06 .45 " .04 .48 " .04 .55 " .04
- .13 " .05 .45 " .04 .45 66 .045
.39 " .04 .41 " .04 .53 " .04
.17 " .045 .48 " .04 .54 " .04
.35 " .045 .45 " .04 .52 " .04
.22 " .045 .46
66 .04 .28 " .045
- .25 66 .045
.40 " .045 .31 " .05
-.45 " .04 .49 " .04 .39 66.055
.48 " .05 .36 " .045 .24 " .045
.35 " .05 .38 " .04 .30 " .075
.05 "
.055 .32 " .045 .23 "
.045
.36
66
.05 .52 " .04 .41 66 .045
.00 " .055 .46 " .04 .31 " .065
.18 " .055 .49 " .04 -.01 " .085
.40 " .05 .53 " .035 .27 " .075
.52 " .045 .39 " .045 .34 " .055
.45 " .045 .35 " .045 .14 " .06
.32 " .045 .24 " .08
.18 " .055 .53 "$ .065
.33 " .055 .31 " .075
.24 " .055 .29 " .075
.52 66 .05 .34 " .07
.27 " .055 .29 " .07
.29 " .055 .39 " .06
.11 " .06 .45 " .06
.45 " .05 -.02 " .07
.36 " .05 .48 " .05
.40 " .055 .36 " .07
.56 " .045 .50 " .05
.38 " .05 .54 " .06
.41 " .05 .04.07
.37 " .055 .34 " .07
.67 " .07
.47 " .06
.63 " .06
It is
thus clear that only a small part ( about one fifth )
of the variation in the r's
is due to the limitation to 668
or 454 or 240 cases .
Suppose now that we take the mere difficulty of a task
as a measure of its " intellectual difficulty . How large an "
error do we make by such neglect of any correction for the
magnitude of r .. ? Such a procedure is equivalent to treat-
ing as equal σt1 , Ot2, Otз , Ot49 Otn, which after the correc-
tion would be respectively
σ₁ Οι Οι
9 " " and so on .
rtil rt21 rtzi
Since r,
has a median of about .38 and a mean square vari-
ation of about .12 for Grade 11 when n is ∞ , the sigmas
which we treat as equal and which will in reality not all
' vary from about ; in about

Οι Οι Οι
equal to about a sixth
.38 .04 .74
Οι
of the tasks will be below and in about a sixth of
.26 '
σt
Οι
them it will be above The sigmas will vary around
.50 *
2.630 with a mean square variation of .830 ,. If the vari-

,
ability of an eleventh - grade population is one - fourth of the

variability of all 17- year- olds , this equals .21o1 of
all
year olds
17
-
or nearly one thirtieth the entire range of adult human

of
-
intellect.฀
It may be well call attention to the effect of the variability of the
to
6
group upon such correlations as we have presented As well known

is
a a
,
.
correlation of .99 between two measures of intellect for group composed of

a
random selection year olds will shrink greatly by selection for some
20
of
if
,
-
characteristic closely related intellect we have group varying only one-

to
a
,
fifth much as the random group

as
This means of course that the error of single small task by its failure
to
a
utilize all of intellect and its adulteration by factors other than intellect may
,
be small fraction of the total range of say adult human intellect but large
a
a
,
fraction of the range of collegiate intellect

.
The rise in the correlations between score single task and intellect
in
a
with wider range of the group used does not impair the validity of anything
hitherto stated this chapter as an inference from the correlations found for
in
any group The correction for the error assuming act as

to
chance
it
a
,
.
Obviously , we do not obtain a very close resemblance

in " intellectual difficulty " within a composite of single
tasks of equal difficulty . This , however , impairs the value
of such a composite only slightly , since the variations of
individuals ' intellects in respect to how difficult any given
single task is to each of them are so great . We use mea
sures of intellectual difficulty to measure the level or height
of individual intellects . If we did have a score of single
tasks that were of absolutely equal intellectual difficulty
for , say, eleventh -grade intellects all taken together , they
would vary greatly in intellectual difficulty for those intel
lects taken one at a time , and they would also vary greatly
in intellectual difficulty in the case of a thousand of these
eleventh -grade intellects chosen to be all of identical intel
lectual level . If, for example , each of such a thousand in
tellects gave ten correct responses out of twenty , they would
not all answer any one task correctly , nor all fail com
pletely on any one . The single task does not measure all
of intellect and nothing but intellect , and so may utilize a
large fraction of A's intellect and a small fraction of B's ;
it may be solved by C largely by factors other than intel
lect , while in D there exist non - intellectual factors which
prevent him from solving it .
Consequently a score of single tasks , all with , say , 50 %
,
of successes , which after correction would be represented

by twenty values ranging down even to Median grade 11
201 grade 11, and even up to Median grade 11 +201 grade will
119
nevertheless be very serviceable composite

a
The two paragraphs preceding the last one are how

,
ever really fallacious Since single word be defined

to
a
,
sentence be understood does not have any genuine

or
in
to
tellectual difficulty the sense of difficulty for all of intel

in
lect and nothing but intellect we cannot properly attribute

,
sampling error terms the variability of the group for which the corre
of
in
is
In absolute units the correction will be the same the in

,
lation found
in is
,
.
crease the variability of the group exactly counterbalancing the effect of the
increase in the correlation
.
any amount of intellectual difficulty to it . We should not

impute purely intellectual difficulty to any save purely in
tellectual tasks . A task cannot be purely intellectual when
it correlates only .40 with intellect in a grade population or
only .80 in an age population . We are back at the familiar
point . We cannot measure the intellectual difficulty of a
single brief task.
Moreover , in putting tasks together in a composite , we
should pay attention to their equality in purely intellectual
difficulty , if it could be measured , only after much more im
portant desiderata had been provided .
In making up composites of tasks our chief aims are to
make composites which will correlate highly with intellect
and which will be of specified difficulty . This means that
after the first element of the composite has been chosen , the
merit of the next depends largely upon its partial correla
tion with intellect ; and after two have been chosen , the
merit of the next depends largely upon r₁3.12 , its partial cor
relation with intellect ( after elimination of the influence of
the first and second elements ) . And so on with the others .
Getting high partials means getting different aspects of in
tellect represented and getting different non - intellectual
factors counteracted . A moderate amount of wisdom in
predicting what a given task will do in these respects will
save much labor in computing rt's .
So far as concerns the first aim , equality in the correla
tions (r's ) is valueless . Among equally difficult tasks we
would prefer those with the highest correlations and partial
correlations .
So far as concerns obtaining a composite of precisely a
certain specified difficulty , there is no practicable way of
guaranteeing this beforehand . In practice , however , the
matter is easy to arrange . We make up each composite
from tasks of equal difficulty , and then measure the diffi
culty of the composites . If we have enough single tasks
and over a wide enough range , we shall have a great num
ber of composites differing progressively by small amounts
of difficulty , and can usually find among them one close

enough to the specified difficulty to serve . If such a one is
not found , it can be created by combining two neighboring
composites , taking half of the tasks from each , or four
tenths from one and six tenths from the other , or whatever
proportions are likely to give a composite of the desired
degree of difficulty .
In the
actual selection of elements for a composite , then ,
the main desiderata are to have the percents of successes
equal and the r's
either high or with high partials . The
equality of the r's
is an altogether minor matter . "
The facts about variation in the r .. ' s retain their im-
portance because we do need to get high correlations with i ,
both for their intrinsic value and because one of the best
practical ways to get high partial correlations with i is to
find tasks which measure intellect with different data or
different operations and still show high total correlations .
It is a sound rule not to use any single task in a composite
unless its r , or the average r , for it or for tasks like it is
above .30 for a school - grade population of the level for
which it is intended . "
Any information about the r's of representative tasks
is therefore of general value ; and we quote here the results
obtained by Vincent ( '24 ) . Using data furnished by us ,
she measured the correlation between the score attained
with a single sentence to be completed and the total score
attained in an intelligence examination of two and a half
hours , whose reliability coefficient is about .85 for such
far as concerns boredom from too easy or irritation and discourage-
7 So
ment from too hard single tasks within the same composite , they are due
chiefly to variations in difficulty , not in intellectual difficulty . They are pre-
vented chiefly by including tasks which are equally difficult . The tasks should
also be fairly free from environmental influences ; and high " total " correla-
tions with i are one symptom of this .
8 More precisely , for any two minutes of work we should obtain r ฀฀ above
.40 for Grades 3 to 5 , .35 for Grades 6 to 8 , .30 for Grades 9 to 11 , and .20
for Grades 12 to 14. It is harder to get high correlations in the higher grades ,
where the range of intellect may be narrower and where the specialization of
the environment is greater .
groups as she used . The first group consisted of candidates

for entrance to college ; the second consisted of sixth - grade
pupils . The examination for the former was the Thorndike
Intelligence Examination for High School Graduates ; that
TABLE 15
OVERLAPPINGS AND BI - SERIAL R'S FOR 35 ELEMENTS
Bi -serial r's
Overlappings with P.E.'s
0.0 .826.046
5.2 .752 " .049
6.4 .726 " .046
8.9 .588 " .057
9.1 .639 " .054
10.0 .406 " .082
11.7 .431 " .083
12.3 .558 " .061
12.4 .559 " .068
12.5 .463 " .068
12.6 .559 " .060
15.2 .551 " .048
15.8 .597 " .061
16.5 .542 " .041
18.7 .665 " .031
20.7 .520 " .037
21.2 .475 .043
21.5 .480 " .051
21.6 .444 " .044
22.9 .427 " .045
23.3 .398 " .045
27.9 .416 66 .053
28.1 .352 " .045
28.6 .163 " .047
30.4 .160 " .083
31.7 .377 " .055
31.8 .372 " .081
31.9 .343 " .073
35.1 .359 " .076
35.6 .276 " .068
36.2 .168 " .064
36.7 .145 " .066
36.9 .261 " .073
37.7 .323 " .062
40.9 .307 " .083
for the latter was made up of Sentence Completions , Arith-

metical Problems Vocabulary , Sentence Comprehension ,
,
and a battery of stock intelligence tests . The correlation

(using the biserial r ) varies from .70 or higher to near 0 .
TABLE 16
THE CORRESPONDENCE SUCCESS IN A SINGLE SMALL TASK AND INTEL-

BETWEEN
LECT , AS MEASURED BY THE OVERLAPPING OF THE SCORE IN INTELLECT
OF THOSE FAILING WITH THE TASK PAST THE MEDIAN SCORE IN
INTELLECT OF THOSE SUCCEEDING WITH THE TASK .
COMPILED FROM THE ORIGINAL DATA OF VINCENT .
Candidates
Sixth Grade Pupils
for College Entrance
-
% of Frequencies
Over-
C. A. D. All C. A. All
lapping
0-4
2
3
4 3
5-9 11 13
3
245
156
10-14 11 14 30 12 18
15-19 15 25 19 59 ++4 28275 12
22 55
20-24 40 22 19 81 10
25-29 21 21 18 60
30-34 19 17 17 53 6
35-39 11 11 31
1 2
2
2
9
40-44 10 17
2
1
1
5
1 2
45-49
2 2 5 5
50-54
32
1 3
55-59
1
60-64
1
1
65-69
1
70-74
1
n 122 126 117 365 27 43 70

Median 24 24 25 24 162 14 15
614
7
฀
An overlapping of corresponds correlation coefficient of from

to
25
%
about .30 about .42 according the percent succeeding from

or
to
remote
as
is
near to 50
.
An overlapping of 15 corresponds correlation coefficient of about

to
.45
%
about .60 according the percent succeeding remote from near

or
to
to
50
as
is
The facts for Table 15. We may use

35 tasks are shown
in
as measure of the correspondence between score for one

a
element and score intellect measured perfectly nearly

or
in
perfectly , the smallness of the overlapping ( in the total ex

amination scores ) of the failures past the median of the
successes ( success and failure referring to the single task
in question ) . This ranges from 0 to nearly 50 % , the latter
figure corresponding to a correlation of zero .
From Vincent's original data we have compiled Table
16, which shows the percentages of overlapping for various
tasks in completing sentences , understanding paragraphs ,
and solving arithmetic problems .

The overlappings of Table 16 would all be somewhat
smaller if the measure of intellect were perfect . They
would be less variable if the number of individuals used in
the determinations were larger . This number ranged from
50 to 499 in the college group ( two - thirds of the sentences
being taken by fewer than 175 individuals ) and was either
about 175 , or about 240 , or about 375 in the sixth - grade
group . They would become both smaller and less variable
in proportion as any mistakes in scoring were eliminated .
Those for the college group would become less variable if
the tasks had been done with no limitations of time . The
sixth -grade groups were allowed to use as much time as
they wished ; the college group were instructed to work as
fast as they could without making mistakes , and were sub
ject to a time limit which was rather generous in the case
of the completion and the reading tasks , but rather limited
in the case of the arithmetical tasks . The smaller overlap
pings ( that is , higher correlations ) for the sixth grade
group are due probably partly to this fact and partly per
haps to a greater variability of the sixth - grade group in
intellect .
The eleven cases of negative correlation in the college
group are in some cases due to badly chosen tasks ; in
others they are due probably to some constant error in the
scorers ; in others they are due to the chance error attached
to a determination from a limited number of cases .
SUMMARY
The sum and substance of this chapter is an emphatic

conclusion that for every theoretical and practical purpose
in the measurement of intellectual difficulty , we should use
collections of tasks rather than single small tasks . We
ought to measure the difficulty of single tasks ; but we can
profitably measure intellectual difficulty only in the case of
composites which contain enough kinds of tasks to repre
sent a fair sampling of all of intellect as it operates at that
level , and enough tasks to make the error closely the same
for any one composite as for any other with which we wish
to compare it in respect of difficulty .
When such composite tasks are attained we can infer
the difficulty values in terms of o , from the values in terms
of σt1 , σt2 , σtg , etc. , or , since ot1 , t2 , Otз , etc. , are closely
equal , we can use σ , as the unit .
CHAPTER V
THE MEASUREMENT OF THE INTELLECTUAL DIFFICULTY OF

TASKS BY A CONSENSUS OF EXPERT OPINION
No one doubts that a certain validity attaches to human

judgments of the difficulty of intellectual tasks ,- that , for
example , it is harder to find an opposite of " government "
beginning with " a " than an opposite of " below " beginning
" "
with a, or to answer correctly, " How many quarters of
a quarter equal half of a half ? " than to answer correctly ,
" How many cents are three cents and one cent ? "
If a thousand psychologists or others who are ac
quainted with intellectual tasks are required to state which
of two tasks is harder , the amount of agreement is a mea
sure , or at least a symptom , of the magnitude of the real
difference . If 910 of the thousand rank A as harder than
K, whereas only 510 of the thousand rank B as harder than
K , the difference A-K will be supposed by all sensible per
sons to be greater than the difference B -K , except for
tasks in respect of which the thousand suffer from some
illusion or constant error .
Given the truth of certain assumptions about the judges
and the process of judging , the magnitude of the real dif
ference may be determined from the percent of judges
discerning it . These assumptions have been used as a pro
visional way to determine the magnitudes of differences in
the general merit of handwriting and drawings by Thorn
dike [ '10 and '13 ] and Kline and Carey [ '23 ] ; the general
merit of compositions by Hillegas [ '12 ] ; and the beauty of
designs by Thorndike [ '16 ] . They have been used widely
by Hollingworth , Strong and others in measuring various
features of advertisements .
This method of deriving units of measure is more appro
priate in the case of certain esthetic and ethical values ,
134
MEASUREMENT OF DIFFICULTY BY EXPERT OPINION 135
where to be more beautiful ( amusing , sublime , refined , rev
erent , patriotic ) is to be thought so , than in the case of in

tellectual difficulty , but it has obvious advantages in econ
omy and ease of application , and , at the worst , it utilizes
human judgments without exaggerating their intrinsic
errors and without introducing any new errors .
What the intrinsic errors of human judgments of the
difficulty of intellectual tasks are has not been known be
cause hitherto there has been no extended or systematic
study of such judgments , and no criterion against which to
check their validity and precision .
THE EXPERIMENTS
We have carried out two experiments with these impres
sionistic judgments of the difficulty of tasks . The first ,
which was reported in the Journal of Educational Research ,
February , 1924 , [ Thorndike , Bregman and Cobb , '24 ] used
a hundred tasks as the material to be judged , and forty stu
dents of psychology and education as the judges . The raw
correlation of the ranking for difficulty by the consensus
with the ranking by the percentages of a group succeeding
with the respective task elements was .88 .
The second experiment used some twelve hundred tasks
and twenty sets of judgments , these being made by Dr. E.
M. Bailor , Dr. E. O. Bregman ( 2 ) , M. V. Cobb ( 2 ) , Dr. A.
I. Gates , Z. F. Miner , Dr. R. Pintner , E. E. Robinson ( 2 ) ,
G. J. Ruger , Dr. L. S. Hollingworth , Dr. Godfrey Thomson ,
Dr. L. M. Vincent ( 2 ) , J. W. Tilton , Dr. B. D. Wood , Ella
Woodyard ( 2 ) , and E. L. Thorndike ( 2 ) .
The number 2 in parenthesis means that the person in
question made two sets of judgments . The instructions for
the ratings and a few sample tasks including some near
both extremes are quoted below .
INSTRUCTIONS FOR GRADING

Each slip is a task in Arithmetic , Sentence Completion , Vocabulary ,
Directions , Reading , Information , or giving Opposites . The nature of each
task will be apparent , if you remember that :
11
1. A single word preceded by a letter means that the task is to give an

opposite to the word beginning with that letter , e.g. , " b ...... white " requires
" black "
as the response .
2. A word followed by 5 other words means that the task is to select that
one of the five which means most nearly the same as the first word , e.g. ,
" powerful .....holy.....strong......during...... sad..... old " requires " strong " as the re
sponse .
3. A word followed by 4 or 5 pictures means that the task is to select

that one of the pictures which best fits the word .
4. If
the task is preceded by the word " oral , " the task is not to be shown
to the person in print , but put to him clearly orally , and repeated once , and a
second time if he desires .
5. If
the task is not preceded by the word " oral , " the person doing the
task is supposed to have the opportunity to read and re -read it . he has If
difficulty in reading , he is supposed to have the task stated to him orally in
whole or in part as often as he wishes .
The tasks are to be rated in 200 or more groups , in respect of their intel
lectual difficulty , for a group of persons twenty years old brought up in the
United States , with an opportunity to go to school for at least 7 years , unless
they were so dull as to be unable to learn at school . At one end will be the
tasks which you think only the best intellects would do correctly ; at the other
end will be those which all save the lowest imbeciles would do correctly .
You should assume that the general nature of the task of giving an oppo
site , or of completing a sentence , or of selecting the word most nearly of the
same meaning , has been stated in very simple language and illustrated by five
easy samples , and that the tasks of any one sort are given at one time and in
an order beginning with the easiest .
In all ratings pay no attention to the possibility of chance successes .
Think of the difficulty of the task in every case as the difficulty of succeeding
with it by real knowledge or ability .
In about
one case out of 200 there was an omitted or am
biguous rating .
To simplify later computations , an esti
mate was made of the probable intent of the judge in such
cases , by consideration of his ratings of four tasks of ap
proximately the same sort .
The basis for the judgments doubtless varies from one
judge to another and from one task to another for the same
judge , and for the same task for the same judge at different
times . It would be interesting and perhaps valuable to dis
cover what qualities in a task and what facts or fancies
about it make any given judge regard it as hard . We shall ,
however , limit our inquiry to the ratings themselves re
gardless of how they were caused .
.
TABLE 17
.
THE CORRESPONDENCE BETWEEN THE SUM OF THE RATINGS OF TEN JUDGES AND THE SUM OF THE RATINGS OF THE OTHER TEN
Sum of 1st ten

Sum of 2nd
1760
1360
1600
1200
1520
1680
1840
1040
1280
1440
1120
640
320
560
240
400
480
08
160
ΟΙ
ten 10
10 to 79 12
4
1
24
80 159 20 22
MEASUREMENT
32-1
7
160 239 15 10 12
244
5 4
79
"" "
240 319 16 21 19
OF
5
320 " 399 15 20 16
"
400 479 20 13 13 13
2335
5
1 1
6 1
66 559 14 16 15
480
8 3
639 10 29 18
560
2 5 6
1 7
640 719 30
9 6 3
3
3 11
' " " ('

720 799 10 17 14
5
1
7 8 5
879 22 12 13
800 "
8
959 16 11
880
4
16 14
""
960 1039
66 1119 15 16
1040
433
6 1 1 2
3
7 7
1120 1199 14 21
2
6 6 4
2
""
1200 1279 13 18
9
2 4
"
1280 1359 17 11
1360 1439 10 16 15 11
""
1 4 9
1440 1519 10 24 11
DIFFICULTY BY EXPERT OPINION
42
8 7 1 2 1
1520 1599 18 16
265
3 2
6 6
""
1600 1679 11 19
5
1680 66 1759 13
137
1270
5 1
3
1
"
1760 1839 10
THE RATINGS
The ratings were combined by simple addition , the re-
sult being a series of arbitrary numbers from 32 to over
3,600 which represent accurately enough for all our pur-
poses an order of difficulty by the consensus . Its statistical
reliability is fairly high . The sum of the ratings by ten of
the judges ( Br . , Mi. , Ro . ( 2 ) , Thom . , Thor . ( 2 ) , Vi . and
Wo . ( 2 ) ) corresponds closely with the sum of the ratings by
the other ten . The facts appear in Table 17 .
The reliability is about the same for any one sort of
task , such as sentence completion , or arithmetical prob-
lem or word knowledge , as for the entire series . That is ,
the judges agreed about as closely when they compared two
tasks of different sorts as when they compared two tasks
of the same sort .
The correlations between the two sums of ten are as fol-
lows :
Completion tasks .973
Arithmetic tasks .988
Vocabulary tasks .954
Directions tasks .996
Information tasks .979
Opposite tasks .978
The average of the six is .978 . The correlation when all

are mixed together is .984 .
This material is unsuitable for the computation of co-
efficients of correlation , the distributions being of very ir-
regular form . The correlations given above are used only
as rough indicators of the closeness of agreement between
the two groups of ten judges .
The mean square error and the median or " probable "
error of the sum of the twenty ratings for any task are as
shown in Table 18. The error varies , increasing in general
1 These measures of unreliability are computed from the mean square devia-
tions of the differences between the sum of the ratings of the first ten and
the sum of the ratings of the second ten judges . The mean square error for
1
a sum of ten equals Odiff .. The mean square error for the average
V2
with the difficulty of the tasks , but also decreasing at the

two extremes of the set of tasks used . On the average , it is
about one thirtieth of the difference between task and I
task II
, shown at the bottom of Table 18 , for the mean
square error , and about one forty - fifth thereof for the prob-
able error .
TABLE 18
THE PROBABLE Divergence OF A DIFFICULTY RATING BY 20 EXPERTS FROM THE

AVERAGE OF AN INFINITE NUMBER OF DIFFICULTY RATINGS OF THE TASK ,
( EACH RATING BEING THE AVERAGE OF THE RATINGS OF 20 EXPERTS ) .
The unit being the The unit being one The unit being one
same as that of the hundredth of the hundredth of the
difficulty ratings by difference between difference between
the 20 experts Level A and difficulty rating for
Level O Task Iand Task II
Tasks rated under S.D. P.E. S.D. P.E. S.D. P.E.
400 ( approx .) 44 29 1.5 1.0 1.3 .8
400 to 799 81 55 2.8 1.9 2.3 1.6
800 " 1199 97 66 3.3 2.2 2.8 1.9
1200 1599 99 67 3.4 2.3 2.9 1.9
1600 1999 129 87 3.7 2.5 3.2 2.1
2000 2399 110 74 4.4 3.0 3.7 2.5
24002799 117 79 4.0 2.7 3.4 2.3
2800 3199 108 73 3.7 2.5 3.1 2.1
3200 3599 86 58 2.9 2.0 2.4 1.7
1. Hold up your hand .

2. Show me your nose . Put your finger on your nose .
3. Show me your mouth . Put your finger on your mouth .
1 1
of two sums of ten equals X X Oairt . Since , however , we are using
V2 V2
the sum of twenty in place of the average of two sums of ten , our numbers
are all twice as large as they would be for the average of two sums of ten.
That is , the mean square error for a sum of twenty equals :
1 1
2 x X Xσaff . Or simply σaitt ..
V2 V2
2 Level A is the ability of adults of mental age a little under 36 months , and
so with I.Q.'s of about 20 .
Level O is approximately the ability of the average graduate of American

colleges of high requirements .
II
1. Read this and then write the answers . Read it again if you need to .
COLERIDGE
I pine like her in golden story
see thee
The gates thrown open -

Who , in her prison , woke and saw, one day ,
saw the sunbeams play
- -
With only a web ' tween her and summer's glory ;
Who , when the web so frail , so transitory ,
It broke before her breath had fallen away ,
Saw other webs and others rise for aye ,
Which kept her prisoned till her hair was hoary .
Those songs half - sung that yet were all divine
That woke Romance , the queen , to reign afresh—
Had been but preludes from that lyre of thine ,
Could thy rare spirit's wings have pierced the mesh
Spun by the wizard who compels the flesh ,
But lets the poet see how heav'n can shine .
Copy the first word of the line which implies there had not been a con
tinuous stream of like songs .
2. Supply the missing words to make this a true and sensible sentence .
Speech , gesture and form of human
action are in run resolvable
contraction .
3. Arrange these numbers and signs to form a true equation .

2/3 2 3 15 15 = X
So much of these unreliabilities as is due to the small

number of judges can be reduced to any desired extent by
increasing the number of judges . The crude summations
of ranks can also be replaced by more precise and refined
uses of the differences between the rankings for any two
tasks . The general value of the method can , however , be
studied well enough for our purposes with the sums of the
twenty ranks as they stand .
The meaning of these sums of the twenty ranks in terms
of the percentage of the judges who judge the direction of
the difference correctly may be realized from the following
facts :
Taking 618 pairs of tasks at random from those pairs

which differ in the " sum of the twenty " by approximately
100 ( 95 to 105 ) , we find that , in 263 , eleven and a half or
fewer of the twenty judges ' judged correctly ; in 114 , twelve
judged correctly ; and in 241 , twelve and a half or more
judged correctly . A difference of 100 in the " sum of the
twenty ranks " thus corresponds to a percentage of judges
a little under 60 .
Taking 853 pairs of tasks at random from those pairs
which differ in the " sum of the twenty " by approximately
200 ( 195 to 205 ), we find that , in 404 , thirteen or fewer of
the judges judged correctly ; in 49 , thirteen and a half

judged correctly ; in 400 , fourteen or more judged cor-
rectly . A difference of 200 in the sum of the twenty ranks
thus corresponds almost exactly to a percentage of 674 .
A percentage of 67 correct means a difference of .673
times the median deviation of the judges in ability to judge
the intellectual difficulty of tasks , and 60 % means a differ-
ence of .375 times it . So we may regard the median devia-
tion ( or difference observable by 75 % of these judges ) as a
bit over 300 in the units of the " Sums of twenty . The "
entire range is thus only about twelve times the amount of
difference which 75 % of these judges recognize , which
means , of course , that our judgments of the intellectual dif-
ficulty of tasks are not acute .
THE VALIDITY OF THE CONSENSUS
The important matter is , of course , the validity of the

-
consensus , its correspondence with intellectual
difficulty
when that is objectively determined . How far we have a
right to use a consensus of expert opinion to measure the
difficulty of a task depends upon the freedom of the con-
sensus from systematic or " constant " errors , such as a
3 When any judge assigned the same rank to the two tasks which the twenty
put as 100 apart , he was scored as half right and half wrong .
4 It will be a little less than that at the two extremes , and more than that in
the middle , the agreement of the judges being closer for a difference of 100 or
200 at the extremes than for the same numerical difference in the middle ranges .
tendency to over- estimate the intellectual difficulty of com

pletions in comparison with arithmetical problems , or to
underestimate differences in intellectual difficulty at the
hard end of the series , or to fail to see differences over a
certain range of the scale .
We can determine intellectual difficulty objectively for
certain composite tasks . Thus , for very dull adults the re
spective intellectual difficulties of the composites A, B , C ,
and D are determined by the facts that 159 , 87 , 23 , and 1 out
of 180 such dull adults succeed with these composites . That
the composites measure Intellect CAVD is elsewhere
proved . From the form of distribution of the group of 180 ,
the differences between A , B , C , and D in intellectual diffi
culty may be found .
If now we examine the average consensus estimate of
difficulty of the forty elements of composite A , and simi
larly for B , C , and D , we may easily compute the average
differences between A , B , C , and D in intellectual difficulty
as estimated by the consensus . The closeness of correspon
dence of the objectively and subjectively determined sets
of differences in intellectual difficulty may be measured in
various ways . A similar procedure can be carried out for
composites used with a group of college graduates , or for
any other set of composites , whose intellectual difficulty is
objectively measured in suitable units . In proportion as
the consensus agrees closely with the objective results in
the case of composites where we have such objective re
sults , we can trust the consensus " in the case of tasks where
we lack objective results . It is therefore of great impor
tance to inquire how close this correspondence is , how free
the consensus is from errors other than the variable errors
due to the small number of experts .
It will in a summary manner the out
be useful to state
come of the inquiry before presenting its details .
Such a consensus , even from a thousand experts , will not
be trustworthy throughout . It will make blunders , suffer
5 Subject to due consideration of its variable errors , and within the range
of difficulty and sort of task where it has been proved valid .
from "constant errors ,

99
some of them regrettably large .
We shall see , for example , that it overestimates the diffi
culty of easy sentence - completions in comparison to that of
directions or vocabulary or arithmetic tests , and that it
fails to observe genuine differences in difficulty within a
rather wide range of vocabulary tests . If composite tasks
were made up on the basis of its estimates of the intellec
tual difficulty of tasks , the difficulty of these composites as
composites would need to be carefully measured objectively .
The consensus would, on the other hand , often be near
the truth and rarely be greatly in error . There will be a
substantial correlation with objectively determined results .
If such a consensus alone had been used to estimate the dif
ficulty of single tasks , and CAVD composites of forty had
been constructed on the basis of its estimates , they would
have been serviceable composites , forming a gradation in
intellectual difficulty , and containing in any one composite
few single tasks which would appear puerile on the one
hand or mystifying on the other to the individuals who
could succeed with half of the tasks in that composite .
After being evaluated as composites by objective methods ,
these composites would be not much inferior to those which
we have constructed at enormous cost of time and labor
spent in experimentation . Consequently , the use of esti
mates by a suitable consensus may well replace measure
ments of the percents succeeding in the case of single tasks ,
in the preliminary work of making composite tasks .
The evidence that the consensus is in certain respects
definitely wrong is as follows :
We have four composites each made up of ten sentence
completions ; four , each made up of ten arithmetical tasks ;
four , each made up of ten vocabulary tasks ; four , each
made up of ten directions , and four , each made up of ten
information tasks . Each of 180 very dull adults was tested
with each of these twenty 10- composite tasks , which we
shall designate hereafter as
and
D
A
B
C
,
A
,
66 66 66 66 66
V 66 66 66 66 66
D 66 66 66 66 66
66 66 66 66 66
I
The results were as shown in Table 19 the 180 individuals
,
being divided into two groups one institution
of
100
at
,
,
and 80 at another
.
TABLE 19
.
MEASURES OF THE DIFFICULTY OF 10 COMPOSITE TASKS
.
By Experiment By the Consensus
Distances from
Percent the Median Median Sum of No. of
In terms of 010
.
Succeeding 20 Expert Ratings Tasks Rated
n = 100 80 100 80
=
n
CA 84 8212 1.19 1.13 800

-
-
-
2 8
65 56 .51 .20 830
-
44 4 ฀
35 272 .45 .68 970
BOAC
D +1.59 high 1023

3
A 69 80 - .65 1.03 309

-
45 49 .15 .03 536

+
+ +
322
15 21 +1.05 .87 458

ABCA
с
D 1.47 +1.47 858

5
+
5
V 80 81 1.03 1.07 292

-
- -
49 572 .03 .25 562

+
14 19 +1.09 .93 925

ABCD
с
9466
6 7
+
1.78 +1.47 848

1
D A
-
90 86 1.45 1.27 not over 300
-
-
45 672 .15 .59

+ +
19 272 .93 .68 529

ABCD
с
+ +
฀ 12 14 +1.16 1.09 668 10
A 76 .89 1.17 341

-
8312
-
I
-
3 253335
51 592 .03 .32 618

212 .81 .86 733
ABCD
7476
+ +
+
D +1.59 1.53 792

4
In the same table are shown the median summation

scores for the single tasks each these composites
in
of
so
far as they were included in the 1200 rated , and the number
so included . "
There is an obvious " constant error " in the direction
of overestimating the difficulty of the sentence - comple
tions , especially the easier ones . To be in line with the
other tasks , the figures for them should be , respectively ,
about 500 , 300 , 350 , and 200 lower than they are . There is
a failure to distinguish the Arithmetic B's from the Arith
metic C's . There is a similar failure with the Vocabulary
C's and D's .
TABLE 20 .
DIFFERENCES IN DIFFICULTY OF VARIOUS COMPOSITE TASKS AND OF THE
MEDIAN SUMS OF 20 EXPERT RATINGS OF THE SINGLE TASKS OF THESE
COMPOSITES WHICH WERE RATED . EACH DIFFERENCE IS Ex
PRESSED AS A PERCENT OF THE DIFFERENCE BETWEEN THE
A AND THE D COMPOSITE OF ITS KIND .
Differences in Difficulty Disagreements
100 100 80
By Experiment By the with with with
n = 100 80 Consensus
80 Consensus Consensus
- A
C B C 24 30 34 20
6
฀ ฀
* *
CC- CB 35 28 30 37
671
CD - CC 41 42 25 16 17
1
*
A B -A A 38 42 41
4
A C-A B
1
42 34 -14 56 48
38885
A D -A C
8
20
22 24 73 49
4
฀฀฀
V B -V A 38 32
฀฀
49 11 17
4 8 6
VC-V B 38 46 65 27 19
8886
V D -V C
225
25 21 -14 39 35
I B-I A 35 31 612 2612 302

4
IC-IB 82 182
I D-I C
34 44 2512 10
31 25 13 18 12
6
Sum of disagreements 68 322 304
The difficulty of CD estimated 2.00

as
is
+
*
The probable errors of these medians will be approximately 50 when

= =
n n
6
7, 2,
40 when 35 when 31 when 28½ when 26½ when

= =
= =
=
n n
n n
n
8, 3,
9, 4,
5,
6,
25 when 23 when and 22 when 10

=
n
.
If welist the differences between the A and the B , the

B and the C , and the C and the D tasks and express each as
a percent of the difference between the A and the D tasks ,
we can observe more readily how closely the consensus par-
allels the objective results , when the kind of task is kept
constant . The disagreement between the consensus and
TABLE 21 .
FORM OF DISTRIBUTION USED IN THE CALCULATIONS OF TABLES 19 AND 20 .
RELATIVE FREQUENCIES AT EQUAL SUCCESSIVE INTERVALS .
Interval Frequency
0 to .99 5.5
1 66 1.99 5.5
24 2.99 11
3 "" 3.99 11
4" 4.99 17
5" 5.99 17
6 " 6.99 22
7 "" 7.99 28
8 " 8.99 33
9 "" 9.99 33
10 " 10.99 34
11 "" 11.99 39
12 " 12.99 38
13 13.99 39
14 " 14.99 39
15 " 15.99 39
16 16.99 39
17 " 17.99 39
18 " 18.99 44
19 " 19.99 45
2020.99 44
21 21.99 45
22 22.99 44
2323.99 44
24 " 45
"
24.99
25 25.99 44.5
26 " 26.99 38.5
27 " 27.99 33.7
28 28.99 27.8
29 " 29.99 22.3
30 30.99 16.6
31 " 31.99 12.2
32 " 32.99 5.5
the result of either experiment is about four times as great

as the disagreement between the results of the two experi-
ments . The facts are shown in Table 20 .
In the calculations of Table 19 and Table 20 , the form
of distribution of these low imbeciles is taken to be that
shown in Table We ask the reader to take the validity
21.
of this form of distribution on faith for the present , or to
turn to Appendix VI and examine the facts given in con-
nection with its derivation there . To show that the present
conclusion does not depend for its validity upon the par-
ticular form of distribution used , we have carried through
the computations supposing it to be Form A ( the " nor-
mal" form ) and supposing it to be a rectangle . The results
appear in Table 22 and Table 23.
The resemblances may also be measured crudely by cor-
relation coefficients , after first expressing the estimates of
difficulty as a rank order . The correlations ' using
-—
—
6ΣD2
P 1 are
n (n² 1)
.97 for the experiment with 100 with the experiment with
80 ;
.62 for the experiment with 100 with the consensus ;
.70 for the experiment with 80 with the consensus .
We have records from 240 college graduates and from
189 candidates for college entrance with the composites of
ten tasks listed in Table 24. We have also computed the
medians of the 20 - expert sums of ratings of such tasks in
each composite as were rated by the experts . These and
the measures of difficulty from the experiments with the 240
and the 189 are entered in Table 24. In this case the form
of distribution of intellect in the 189
.
is known to be ap-
7 We may here use for r without transmuting , since the form of distribu-
tion of these twenty composites in respect of difficulty is probably better repre-
sented by a rectangle than by a surface of Form A. If
transmuted , all would
be a trifle higher . In the ranks for the consensus , D A is put as 1 and D B as
5 on the basisof the ratings by the consensus of certain tasks closely resembling
the tasks of D A and D B.
proximately of Form A ; and that of the 240 may be treated

8
as such with no damage to the present argument .
A general inspection of Table 24 reveals notable irregu
larities in the measures by the consensus , as when AO is
rated 178 points easier than AN , and D5 is rated 242
points easier than D41 . The consensus seems to fail to dis
criminate well in general among the D ( reading ) tasks .
TABLE 22 .
MEASURES OF DIFFICULTY IF THE FORM OF DISTRIBUTION ASSUMED IS FORM A
OR A RECTANGLE . DISTANCE FROM C.T. IN TERMS OF σ AND OF Q 25 . /

In termsof In terms 25
of
o
a Q
/
Percent assuming assuming
Succeeding Form A rectangle
100 80 100 80 100 80
CA 84 822 - .995 - .935 34 -322
- -
65 56 .385 .158 15
6
-
3 3385
C 272 .385 .598 +15 +222

+
+
BOA
+1.881 high +47 high

฀
A A 69 80 - .496 - .842 19 30
-
45 49 .126 .031
--
5
+ +
+ +
+ +
1
15 21 +35 29
ABCD
1.036 .798
с
+1.645 +1.645 +45 +45

฀
5
5
VA - 30 31
-
80 81 .842 .883
- -
-
49 572 .050 .189 72

+
1
+ + +
1 8145
19 1.080 .887 +36 +31

ABGA
с
D 2.326 +1.645 +49 +45

5
DA -40 36
-
90 86 1.282 1.092
- -
-
45 671/2 .126 .454 172

+
+ + +
19 272 .598 +31 +222

ABCD
.878
+ +
с
12 14 1.175 1.092 +38 +36
A 76 832 - .974 26 -33½

-
.706
-
-
I
51 592 .025 .240 1 92

-
23 212 .739 .789 +27 +282

ABCD
+ +
с
4 +1.881 1.751 +47 +46

3
By the best treatment which we are able make of the available evi
to
8
dence the form of distribution of level intellect the 240 college graduates
of
in
,
diverges from Form only the manner and the extent shown Table
to
in
in
A
165 of Appendix VI
.
TABLE 23
IN DIFFICULTY OF VARIOUS COMPOSITE TASKS AND OF THE MEDIAN SUMS OF 20 EXPERT RATINGS OF THE SINGLE TASKS
DIFFERENCES
.
OF THESE COMPOSITES WHICH WERE RATED EACH DIFFERENCE IS EXPRESSED AS PERCENT OF THE
A
DIFFERENCE BETWEEN THE AND THE COMPOSITE OF ITS KIND
in
Differences difficulty between tasks Disagreements
:
By experiment By experiment
By the 100 100
Assuming the Form Assuming the form
:
:
of Distribution of Distribution Consensus from 80 from 80 100 from Consensus
to
a
to be Form A be Rectangle
100 80 100 80 Form A Rectangle Form A Rectangle
21 20 232 32 10 812 11 132

CB CA
27 19 37 34/2 65 21/2 38 28
CC CB
1 8 9
6
* * *
52 61 392 3312 25 27 142
CD CC
6
4
A AA 29 35 372 412 41 12 3/2
-
31 47 372 14 11/2 912 562 61
A A 422
B C
A
A 2812 34 1512 21 73 512 512 442 5712
28 31 49 912 21 82
1/2
V VA 2712 402
43 65 10 72 322 22
VC V 322 422 502
** **2
-
-
V
B C
392 30 162 182 14 912 532 302
D DA 57 29 572 252
DC 31 48 332 5512
B C
D
D 12 23 19
4
IB A 26 27 34 30 6112 352 272
IC IB 30 38 382 48 2512 912 42 13

1809
C
885
- - - -B -C -D -B - V-D -B D- D- I- - I-
I
D 44 35 272 22 13 512 31 142
367 294
,
disagreements omitting the Directions Tests 79 74
Sum of
as
+
.
is
*
The difficulty of CD estimated 2.00
6ΣD2
The correlations , using Pp = 1— with the rank
n ( n² — 1 )
orders , are .94 for the 240 experiment with the 189 experi-
ment , .79 and .72 for the correlations between experimental
TABLE 24 .
DIFFICULTY OF TWELVE COMPOSITES BY THE RESULTS WITH 240 COLLEGE GRADU-

ATES AND 189 CANDIDATES FOR COLLEGE ENTRANCE , IN DISTANCES + AND
FROM THE MEDIAN FOR THE 240 , IN TERMS OF THE σ OF THE COM-
POSITE CONCERNED . ALSO THE MEDIAN RATINGS BY THE CONSEN-
SUS OF SUCH TASKS IN EACH COMPOSITE AS WERE RATED .
Estimated
Difficulty Probable
By the con- Error of the
By the By the sensus : Consensus
240 189 Median N Median
CN - 1.47 - 1.29 3084 35

0
-
- .87 .77 3279 2
69 60
P .29 - .00 3482 3 50
Q + .21 + .86 3314 1 95
AN - 2.03 - 2.07 2855 6 35

O - 1.23 - 1.18 3047 3 50
Р - .92 - .67 2869 4 42
Q - .45 - .36 3338 3 50
D 42 -1.79 - 2.07 3191 4 42

- 1.08 - 1.54
-
5 2949 3 50
6 .64 - 1.07 3258 3 50
7 + .30 + .16 3291 1 85
Differences Disagreements
By the By the By the 240- 240- 189-
240 189 Consensus 189 Cons . Cons .
CO-N 36 24 85 12 49 61
P-O
-
34 36 88 2 54 52
Q-P 30 40 73 10 103 113
A O- N 502 52 392 12 11 122

P-O 192 30 - 362 102 56 662
Q- P 30 18 97 12 67 79
D 5-42 34 24 - 242 10 276 266

6-5 21 21 309 0 288 288
7-6 45 55 33 10 12 22
Sum of disagreements 68 916 960

results and the consensus . A treatment of the differences

in terms of percents like that of Table 20 is presented in the
lower half of Table 24. The disagreements between ex
periment and consensus are 13 or 14 times as large as the
disagreements between the two experiments . The disagree
ments between experiment and consensus would , however ,
be reduced if we had ratings of five or six tasks instead of
one in the CQ and D7 composites .
We have extensive experiments with the vocabulary
composites 1a , 2a , 3a , 4a , 5a , 6a , and 7a with pupils in
Grades 9 , 10 , and 11 ; and four of the tasks of each of these
composites were rated by the consensus . The essential
facts appear in Table 25. The consensus is badly in error
in putting la much too low , or 2a much too high , or in doing
both . It also crowds 4a , 5a , and 6a close together , failing
to distinguish fully the large differences which exist be
tween these . It makes the sum of the differences between
la and 2a and between 6a and 7a nearly four times as large
as the entire difference between 2a and 6a , though the ex
perimental results make the latter nearly twice the former .
The disagreements between the consensus and any experi
ment are about seven times as large as the disagreements
between any one experiment and any other . In spite of
these notable errors , there remains a general correspon
dence between consensus estimates and experimental re
sults . The rank - order correlation is indeed almost unity ,
Table 25 showing no reversals .
9 It would not be fair to make this statement on the basis of the facts
of Table 25 alone , since both V la and V 2a are so easy for pupils in
Grade 9 and above that results from these grades are not suitable to measure
the difficulty of either at all accurately . We have evidence from a group of
200 pupils in Grade 5 % , however , to the effect that the difference between
Vla and V 2a is less than the difference between V 2a and V 3a .
The percents correct are : 99.5 , 89.5 , and 58.0 Taken at their face value ,
these give differences of .880 for la to 2a and 1.500 for 2a to 3a . The .88
may be too small because the one pupil in 200 who failed to get five of the 10
words in la right may have been extremely careless . Very , very low percents
are of course unreliable for n = 200 , for many reasons . It is extremely un
likely , however , that the true o value for V la will be below -3.20 , so as
to make the difference between 1a and 2a actually greater than the difference
between 2a and 3a .
12
We have also very extensive experiments with the D

( reading ) composites 1 to 7 , the results from which are pre-
sented in Table 26. Unfortunately , the median consensus
ratings for composites 1 and 7 are from only one task
each . Even after liberal allowance for the large probable
errors of these medians , there is a clear failure of the con-
TABLE 25 .
MEASURES OF DIFFICULTY .
Grade 9 Grade 10 Grade 11 Consensus

n = 1041 n = 700 n = 752 Medians n
V la n = 10 - 2.457 - 2.457 - 3.090 2139 4

""
2a 66 -2.576 - 2.366 - 3.090 2622 ""
""
3a "" - 1.366 - 1.468 - 1.995 2679 ""
66 4a "" - .719 - - 1.483 66
- -
.990 2787
66 66 66
5a + .068 .264 .845 2809
"" 6a 66
+ .904 + .527 + .050 2854 "6
"฀"
7a + 2.120 1.866 + 1.259 3227
Differences
2a 1a- .119 .091 .000 483
3a 2a- 1.210 .898 1.095 57
-
-4a
4a 3a .647 .478 .512 108
5a .787 .726 .638 22
6a 5a- .836 .791 .895 45
7a 6a- 1.216 1.339 1.209 373
Differences Divided by the 7a - 1a Difference

2a la- .026 .021 .000 .444
3a-2a .264 .208 .252 .052
4a-3a .141 .111 .118 .099
5a 4a- .172 .168 .147 .020
6a 5a- .183 .183 .206 .041
7a 6a- .266 .310 .278 .342
Discrepancies
9-10 9-11 10-11 -
9 Con . -
10 Con . -
11 Con .
2a- 1a .005 .026 .021 .418 .423 .444
-
3a 2a .056 .012 .044 .212 .156 .200
-
4a 3a .030 .023 .007 .042 .012 .019
-
5a 4a .004 .025 .021 .125 .148 .127
6a -5a .000 .023 .023 .142 .142 .165
7a -6a .044 .012 .032 .076 .032 .064
Sum .139 .121 .148 1.042 .913 1.019
sensus to distinguish differences in difficulty accurately

amongst these reading tasks . The range from D1 to D7
represents the range from what two thirds of pupils in
Grade 6 can do to what not one in twenty - five high - school
seniors can do . By the experimental results , it is clear that
TABLE 26 .
MEASURES OF DIFFICULTY .
Grade 10 Grade 11 Grade 12

n = 1185 n = 1053 n = 742 Consensus B
D1 - 2.409 - 2.878 - 2.652 2713
1
D2 - 1.398 - 1.695 - 1.812 2743 4
- .690 - 1.170
142
D3 -
-
.999 3118
-
2
D4 .055 .333 .516 3177
43 5
D5 + .542 + .306 .065 2949
D6 +1.243 + .966 + .824 3255
147 4
D7 + 2.170 +1.896 + 1.774 3291
Differences
D 2-1 1.011 1.183 .840 30
D 3-2 .708 .696 .642 375
D 4-3 .635 .664 .654 59
D 5-4 .597 .641 .451 - 228
D 6-5 .701 .660 .889 306
D 7-6 .927 .930 .950 36
Differences Expressed Fractions of the D7 D1 Difference

as
D 2-1 .221 .248 .190 .052

D 3-2 .155 .146 .145 .649
D 4-3 .139 .139 .148 .102
D 5-4 .130 .134 .102 - .394
D 6-5 .153 .138 .201 .529
D 7-6 .202 .195 .214 .062
Disagreements
10-11 10-12 11-12 -
10 Con . 11 Con 12 Con
-
-
.
D 2-1 .027 .031 .058 .169 .196 .138

D 3-2 .009 .010 .001 .494 .503 .504
D 4-3 .000 .009 .009 .037 .037 .046
D 5-4 .004 .028 .032 .524 .528 .496
D 6-5 .015 .048 .063 .376 .391 .328
D 7-6 .007 .012 .019 .140 .133 .152
Sum .062 .138 .182 1.740 1.788 1.664
each step is a substantial increase in difficulty . But by the

consensus we have great irregularities and reversals , and
the differences 1 to 2 and 6 to 7 , which should be the largest,
are specially small ( if we trust the single - task medians ) .
The comparisons of the last 10 pages are all subject to
the criticism that the experimental results do not measure
the truly intellectual difficulty of the composites in ques
tion , but rather their difficulty for whatever ability each
involves , and thatif by a miracle we could know how well
people would succeed with these composites if each individ
ual could use all of his intellect and nothing but intellect
with each composite , the results thus obtained might corre
spond more closely with the consensus estimates than our
actual experimental results do .
For this criticism to have force the r , correlations of
the tasks which are put as unduly hard by the consensus
would have to be lower than the others . For example , r
for the sentence - completions should be much lower than r₁
for the arithmetic , vocabulary , or directions . This is not
the case . We have computed the correlations of each with
a composite made up of completions , arithmetic , vocabu
lary, directions , information and opposites with approxi
mately equal weights , using 176 of the 180 imbeciles . They
are :10
Completions with CAVDIO .90
Arithmetic with CAVDIO .80
Vocabulary with CAVDIO .68
Directions with CAVDIO .92
Information with CAVDIO .85
Opposites with CAVDIO .92
The range is here very restricted , all the individuals

being within a range of 28 to 58 months of mental age , with
a σ of about 8 months . If
the correlations were for all
10 There is some spurious correlation in each of these , but this does not
seriously damage the argument , since the amount is not large and is approxi
mately equal for all .
twenty- year - olds or for adults, they would all be very much
higher . If we take the variability of adults as σ = 2.85
years , as computed from the army data [ Memoirs , p . 391 ] ,
and apply the correction " for restricted range , we have :
Completions with CAVDIO .994

Arithmetic with CAVDIO .985
Vocabulary with CAVDIO .971
Directions with CAVDIO .995
Information with CAVDIO .990
Opposites with CAVDIO .995
In the case of 240 college graduates , the correlations of

10 - composites made up of completions , arithmetical prob
lems , vocabulary , and reading , respectively , with a CAVD
summation score from 160 tasks averaged as follows :
с .69
A .49
V .51
D .56 or higher.¹2
Nowhere in fact do we find any inferiority of C to A, V ,

and D in closeness of correlation with i or anything ap
proximating to i .
Wherever and however we estimate it , the r ,, for ten of
our sentence - completions will not be below the average of
the correlations for ten of our A's or V's or D's .
We can think of no good reason why the discrepancies
in the case of the vocabulary and reading tasks should be
any less on the whole if we should compare them with the
difficulties found by experiment corrected for differences
in re , instead of with the mere difficulties . We may , it is
true , hope that the experts ' estimates of difficulty will dis
regard some of the sources of error to which the experi
01 I12
11 R12 See [ Kelley , '23 , p . 225. ]
Σ. V1 12+ r²12( 21/01 ) ².
− r²₁2
12 Some of the D composites had fewer than 10 elements , so that the aver
age of .56 is somewhat too low .
mental determinations are subject , and so deviate from the

experimental determinations toward the difficulty for intel
lect uncontaminated by non - intellectual factors . But the
particular constant errors which we have described do not
seem alleviated in this way .
SUMMARY
On the whole it is certain that we cannot trust any con
sensus of present opinion to provide an accurate measure
of the difficulty or of the intellectual difficulty of a single
brief task. Psychologists do not as yet know enough about
intellect and intellectual difficulty to avoid occasional large
constant errors , such as the over - estimation of the difficulty
of easy completions , or to distinguish well amongst vocabu
lary or reading tasks . The psychologist cannot as yet know
from inspecting a task what fraction of intellect it will call
into action , how high degree of intellect will be needed to
succeed with it , and what effect non - intellectual factors will
have upon its solution , so as to answer the question of how
hard it will be in an actual experiment or how hard it would
be if each person in the group used all of his intellect and
was entirely uninfluenced by non - intellectual factors . A
consensus of experts cannot , in the present status of psy
chology , either relieve us from the need of experimental
tests of difficulty or provide an escape from our previous
conclusion that the measurement of intellectual difficulty
may best limit itself to composites , varied enough to utilize
all of intellect and to equalize non - intellectual factors .
On the other hand , the consensus estimates are in no
sense fortuitous . The correlations of estimates with ex
perimental results are always positive and fairly high , even
within the very narrow range of low- grade imbeciles , or of
college graduates . Over a wide range the correlations will
of course be much higher . The correspondence of opinion
with experiment is not close enough to justify us in accept
ing estimates of the difficulty or of the intellectual difficulty
of single brief tasks as always even approximately true , or
MEASUREMENT OF DIFFICULITY BY EXPERT OPINION 157
in leaving any result of any such estimates unchecked by

experiment . But it is not so slight as to justify us in mak
ing no use of it. On the contrary , if we free it from its over
and under - estimation of the difficulty of certain types of
tasks , it will give a serviceable first rough approximation
to an order of intellectual difficulty . Even without any cor
rection or amendment , composites formed by taking ten of
C , ten of A , ten of V , and ten of D , all forty of which had
TABLE 27 .
THE DIFFERENCES IN DIFFICULTY OF CAVD 40 -COMPOSITE TASKS BY

EXPERIMENT AND BY THE CONSENSUS OF 20 EXPERTS .
Percent which the Stated Difference

is of the Difference P A -
Difference
By the Median
of the Four
Consensus Medians By Experiment Discrepancy
-
B A 16.8 8.0 + 8.8
C- B 11.0 4.8 + 6.2
-
D C 7.4 4.1 + 3.3
-
E D 5.8 6.2 .4
F -E 2.6 6.1 - 3.5
G- F 3.7 7.7 - 4.0
H- G 4.2 9.4 - 5.2
I-H - 9.1
J-I
7.4 16.5
- 4.6
J
10.0 14.6
K- 5.8 3.5 + 2.3
L-K 5.3 -.9 + 6.2
M -L 5.3 6.0 - .7
N-M 4.2 2.1 + 2.1
-
O N 6.3 5.7 + .6
P -O 4.2 5.4 - 1.2
identical consensus estimates , would be useful composite

tasks . We have not had time actually to make and test
such , but we have carried out the converse procedure of
computing the median consensus estimates for the tasks of
our experimentally determined composites . The results
appear in Table 27.
The number of single tasks from the different com
posites which were rated by the consensus , ranged from 6
of P to 29 of F , being for A , B , C , D , E , F, G , H , I , , J K,
etc. , in order 20 , 13 , 18 , 23 , 18 , 29 , 20 , 20 , 13 , 15 , 17 , 8 , 12 , 16 ,
13 , and 6.
The derivation of the experimental results for com
posites A to P is given in Chapter IX .
CHAPTER VI¹
LEVELS OF INTELLECT
We measure the level or altitude of an intellect by the

difficulty of the intellectual tasks which it can perform suc
cessfully , or more exactly , by the difficulty of the tasks a
certain defined percent of which it can perform successfully .
Such tasks , to be truly intellectual , have to be com
posites of a number of single tasks . In the case of Intellect
CAVD , each should represent C , A , V , and D with approxi
mately equal weight to each . Success with such a composite
task may be taken to mean getting all of its single tasks
right , or 99 percent or more of them , or any other defined
fraction of them. For several reasons , the most useful
1 This chapter should properly be preceded by a chapter presenting the
facts concerning the difficulty of the single tasks used in constructing com
posites with which to measure altitude or level of intellect , and concerning
their intellectualness as measured by rt , or some approximation thereto , where
that information has been obtained .
In constructing the tests for intellectual level or altitude , we have made
measurements of the difficulty of over three thousand tasks . The number of
individuals concerned in one of these measurements varies from a hundred to
over four thousand .
These measurements are of great value quite apart from the uses which we
have made of them . They will assist future workers in the field to extend
and refine the selection of tests for altitude of intellect . They provide a sub
stantial beginning for the construction of tests of mental growth in its later
and higher stages , including alternative forms . They provide material for
many scientific studies , for example , of judgments of intellect , of the organi
zation of intellect , of the nature of intellectual difficulty . They may be used
in many ways in the practical work of examining for intellect . The publication
of such an inventory of intellectual tasks with a rough measurement of the
difficulty of each will encourage others to add to it , so that after some years
we shall have a standard source of supply of intellectual tasks of any kind ,
at any desired level of difficulty .
The expense of ordinary publication is , however , prohibitive . Consequently ,
we have prepared a hundred sets in the form of volumes of mimeographed
sheets . These will be sold at cost by the Bureau of Publications of Teachers
College , Columbia University .
159
meaning to take is getting 50 percent or more of the single

tasks right . We shall use this meaning , unless some other
is specified .
COMPOSITE TASKS
The sub - series presented in Chapter III

are samples of
the composite tasks which we have constructed ; and addi-
tional ones are shown below .
LEVEL I
of dots.
1. Hot weather comes in the and ........
weather the winter .

2. The first after June is
3. Children are rude not easily
win friends .
4. The dog a useful because
his intelligence and faithfulness .
5. The rose is a favorite because of

fragrance and
6. The poor little has nothing
to - ; he is hungry .
7. He will come he is not ill .
8. Not persons are eager to work hard .
9. Divisor times quotient will dividend , if
the is done correctly .
10 . you look, will see flowers .
Write the answers to these problems . Use the empty space

to figure on.
11. What will 4 eight - cent stamps and 1 three - cent stamp
cost ?
12. How many inches are there in 2 feet and 7 inches ?
13. How many stamps are there in a sheet 8 stamps wide
and 5 stamps long ?
LEVELS OF INTELLECT 161
14. What does a pound of candy cost when you pay 10

cents for a quarter of a pound ?
15. How much longer is 100 minutes than an hour ?
16. 32 plus what number equals 36 ?
17. How much more is 7 X 6 than 2 × 20 ?
18. How long is it from seven o'clock in the morning to two
o'clock in the afternoon ?
19. The sum of two numbers is 40. One of the numbers is
14. What is the other number ?
20. What number added to 16 gives a number 4 less than
27 ?
Look at the first word in line 1. Find the other word in

Do the same in lines 2 , 3 , 4 , etc. Lines A , B , C, and D show
the way to do it . Do all the lines you can . Write only one
number for each line.²
4. beast 1 afraid.....2 words..... 3 large......4 animal......5 bird :3
4
B. baby 1 cradle...... 2 mother ......3 little child.......4 youth..... 5 girl

C. raise 1 lift up......2 drag
.....3 sun......4 bread.....5 deluge
D. blind 1 man .....2 cannot see......3 game......4 unhappy ...... 5 eyes
21. confess 1 agree......2 mend.... 3 deny......4 admit.....5 mingle

22. backward 1 downwards. 2 after 3 toward the rear 4 defense .... 5 arrears
23. advertise 1 detain...... 2 explore ..... 3 give notice of......4 adverse...... 5 newspaper
24. combat 1 fight..... 2 dismay ...
......3 club 4 expedition...... 5 comb
25. blond 1 polite......2 dishonest ..... 3 dauntless ...
4 coy......5 fair
26. broaden 1 efface.......2 make level .....3 elapse......4 embroider ......5 widen
27. chubby 1 indolent......2 obstinate...... 3 irritable......4 plump..... 5 muscular
28. concern 1 see clearly . 2 engage.... 3 furnish...... 4 disturb......5 have to do with
29. cargo ..... 4 draught......5 vehicle
1 load..... 2 small boat......3 hem
30. clutch 1 exploit....2 nest . 3 flit 4 grasp
...... 5 cane
need to .
Then , upon one knee uprising ,
Hiawatha aimed an arrow ;
Scarce a twig moved with his motion ,
2 Ifit has not been given previously , practice or supervision should be given
to insure that the individual tested understands these directions .
Scarce a leaf was stirred or rustled ;

But the wary roebuck started ,
Stamped with all his hoofs together ,
Listened with one foot uplifted ,
Leaped as if to meet the arrow ;
Ah ! the singing , fatal arrow ;
Like a wasp it buzzed and stung him !
31. What was Hiawatha trying to kill ?
32. What word is used to describe the roebuck ?
33. What is the arrow said to resemble ?
need to .
There is an old saying , " As harmless as a fly " ; and

until recently the fly has been regarded only as an unpleas
ant but harmless nuisance . Had our forefathers known as
much about flies as we now know, they might have made the
"
proverb , As dangerous as the fly . " His origin and his
habits are of the worst sort . He is , in short , a disgusting
and dangerous pest .
The scientists have told us also how to keep clear of the
Houses and grounds should be kept free of decaying
flies .
organic matter , and stables should be screened so as to cut
them off from their breeding places . Our houses should be
carefully screened and food kept free from their dangerous
feet and mouths . Fly paper and fly traps can be bought
everywhere . Your teacher , also , can probably tell you
other means of protection . But don't forget that the
" harmless fly " of the proverb is the dangerous fly of fact .
34. Did our great grandparents know as much about flies

as men do now ?
Should flies be prevented from reaching their breeding
places ?
Is it desirable for a girl to be so gentle that she cannot
bear to kill a fly ?
[ All three answers must be right in 34. ]
There are only four single D Tasks in level I, so each receives a credit
of 21.
LEVEL J.
Write words on the dotted lines so as to make the whole sen-
tence true and sensible . Write one word on each inch of
dots .
1. The way to is by airplane .
2. There is no on earth cannot
bear misfortune .
3. Two pounds of silver are more than two
pounds of iron .
4. He cheerful will make friends
is
.
A body of entirely surrounded by
...
5.
...
is called an
..
6. The think about th -per-

I
it
,
plexed am
It
I
.
7. strength heavy weight

to
....
...
.
8. When....... ………………lines are each perpendicular
to
.
other they form right

a
,
9. One
...... times one half equals one fourth
.
10. The of five and ten fifteen
is
Write the answers these problems Use the blank sheets

to
figure on
to
11. What number minus equals 23

7
12. What number minus 16 equals 20

?
of
If 12
13.
a is
g
.
.
.
.
present costing be paid for by men

27
14. 9.45
to
is
$
contributing equal amounts what one man's

is
,
share
?
15. Dick started from his house walked two miles north
,
then two miles west then two miles south How far
,
away from his house was he then

?
16. man bought land for 400 He sold for 445 gain-
A
it
$
,
.
ing 15 an acre How many acres were there

X ...
$
?
.
17. 12
is
18. Counting that 100 lb. will last men for week how
15
3 a
much will be required last 30 men for weeks

to
girl had 20 quarters dimes nickels and pen-

12
19.
16
A
8
,
nies She made four piles Pile Pile Pile and

A
C
,
,
.
Pile D. In Pile A she put half the quarters , one

fourth of the dimes , one third of the nickels , and all
of the pennies . How much money had she in Pile A ?
20. At the rate of $ 2.25 per week how long will it take to
save $90.00 ?
Directions and samples as on page 161 .

21. awe 1 lamb...... 2 fear......3 tool.....4 mound ..... 5 opera
22. aged ...
1 years ......2 active...... 3 old 4 merciful ..... 5 punctual
23. arrive 1 answer ......2 rival .....3 enter .....4 force ......5 come
24. blunt 1 dull..... 2 drowsy 3 deaf .
4 doubtful ...
5 ugly
25. accustom 1 disappoint..... 2 customary ......3 encounter .....4 get used......5 business
26. bade 1 gaze...... 2 a tool.....3 fetched 4 wait 5 ordered
27. bog 1 ebb......2 disorder. 3 swamp....4 field.....5 difficulty
28. cascade 1 hat ......2 waterfall firmament disaster box
...
...
5
3
29. bray cry of an ass..... bowl..... cry of an ox...... frustrate...... raven's cry
5
4
2
3
1
30. disembark unearth..... ashore..... dislodge..... disparage..... strip

1
5
In each set of sentences check the two which mean most
,
nearly the same as the sentence printed heavy type

in
.-
.
31. weigh the man not his title Wycherley
I
)
(
,
.
Tis not the king's stamp can make the metal
'
better
.
Fine feathers make fine birds

.
Titles are the marks of honest men and wise

.
.........
The rank but the guinea stamp the man's the
is
gold for that

a
,
.
'
32. Anyone can hold the helm when the sea calm
is
Sail when the wind blows

.
Untempted virtue easily retained

is
The pilot cannot mitigate the billows or calm

the winds
.
An unassaulted castle easily held

is
33. In the presence of the greater malady the lesser

is
,
forgot
.
We see not the moon be shining

the candle
if
An ounce of prevention worth pound of cure

is
The greater glory dims the less

.
There are some remedies worse than the disease

.
LEVELS OF INTELLECT . 165
34. What is failure ? It is only

a spur to the one who
receives it in the right spirit .
Every rebuff is a stepping stone to higher
things .
To reach the port of heaven we must sail , and

not drift , nor lie at anchor .
Failure makes the spirit within stir to go in once
more and fight .
Not failure , but low aim is crime .
The paragraph for task 35 is the Hiawatha paragraph
on page 161 .
35. What two words are used to tell the noise the arrow
made ?
The paragraph for tasks 36 , 37 , and 38 is the “ Fly ”
on page 162 .
36. Where does the paragraph say the fly is born ?
37. Who or what informs us how to avoid the dangerous

pest described in the paragraph printed above ?
38. Name three devices which protect us from the disgust-

ing pest .
EVERY HOME NEEDS A GARDEN

A MAGAZINE published to promote real gardening .
Most people do not think much about their gardens at this
time of the year , but if more people did , there would be
more good gardens . If
you live in the city where space is
at a premium , we provide pleasure for you by suggesting
how to grow flowers indoors . If
you live in the country
-
and have a garden and do not experience the satisfaction
of seeing things grow as a result of your own efforts then
you need the X.Y.Z. magazine .
39. What is recommended for persons who fail to make

things grow in their gardens ?
40. Which one of these words could best be used instead of

at a premium ? Draw a line under it .
space flowers valuable extension extensive
cheap noble
LEVEL K
Write words on the dotted lines to make the whole sentence
true and sensible
. Write one word on each inch of dots .
1. When a man is .. of sight,.
2. No.
also very soon out ..
is powerful.
two and two be five .
mind .
--to
3. you wish me to help you......
Latin , please .. me by telephone .
4. He is genteel who does deeds
..
-----------
5. It may
.
effort and long
a but
...
.
the result is sure
.
This magazine the of new and pro- a

is
6.
..
gressive movement
.
7. Four..... two is more .seven

...
.
8. No what happens wrong right
is
..
in Denmark
.
9. The source wealth
...
..
agriculture
In ..........
.
10. to maintain health one

.
should have nourishing

..
Write the answers these problems Use the blank sheets

to
figure on
to
man spent two thirds his money and had left

of
11.
A
$
8
How much had he at first

.
?
bought yards cloth gave the clerk and re-

of
12.
4
$
I
2
,
ceived 20 cents as correct change What was the

.
price of the cloth per yard

?
13. dealer bought some mules for 800 He sold them

A
for 1,000 making 40 on each mule How many

$
$
mules were there

?
14. How much more the sum of and than the sum
is
of 24 and 31
?
15. 20 how many times 12 ?

16. How many times must you add
1 to 6 to have 15 as
a result ?
17. How many times as long as 8 feet is 12 yds . ?
18. 20 = X
19. 8 is 13 X
20. 20 is 14 X
Directions and samples the same as on page 161 .

21. conspire 1 plot...... 2 breathe .
......3 rely 4 die...... 5 outrun
22. check 1 error......2 stop..... 3 flash 4 rude...... 5 haste
23. cherish 1 dedicate ...... 2 happy.....3 covet......4 hold dear...... 5 marry
24. chirrup 1 aspen......2 joyful .....3 capsize ......4 chirp...... 5 incite
25. accessible 1 indefatigable ......2 successful.... 3 limpid..... 4 easy to reach.....5 liable
26. dingy 1 afraid.....2 hostelry.....3 small bell.....4 midget.....5 dirty
27. edible 1 auspicious......2 eligible ......3 eat..... sagacious ..... able speak
fit
to
to
5
28. confound discovered ..... fulfill.3 establish..... mix up...... expire
1
5
29. concur agree...... race...... mongrel...... pounce...... ramble
5
3
4
2
1 1
30. contact tactful ..... hate...... injunction .... touch......oversight M.COM

2

,
nearly the same the sentence printed heavy type

as
in
31. Today worth two tomorrows

is
Time is an herb that cures all diseases

.
A bird in the hand worth two in the bush

to is
To speed today be set back tomorrow

is
There no time like the present

is
heart never won fair lady

.
32. Faint
.
Nothing venture nothing gain

;
Married haste we repent leisure

in
at
Fools rush where angels fear tread

in
to
Fortune favors the brave

.
33. Fight fire with fire

.
Set thief to catch thief

a
Knavery the best defence against knave

is
Sow the wind reap the whirlwind

,
Fire that's closest kept burns fiercest

.
34. One sorrow never comes but brings an heir

Two in distress makes sorrow less
.
.
never rains but pours

It
it
13
On horror's head horrors accumulate .

Sorrow's crown of sorrow is remembering hap-
pier things .
The paragraph for task 35 is " Every Home Needs a
Garden," on page 165 .
35. What does the advertisement say would be the result
if people thought more about their gardens in the
time of year referred to ?
The paragraph for task 36 is the paragraph on Work
on page 81 .
36. According to the paragraph what even would a prisoner
welcome ?
The paragraph for task 37 is " The American State ,"

on page 84.
37. In what respect are some of the original thirteen states
and California unlike all the others ?.
The paragraph for tasks 38 , 39 , and 40 is " Dirge in

Woods ," on page 94 .
38. To whom does " we " refer ?..
39. What veins the moss carpet ?..
40. What event of man's career is like the falling of the
fruits of the pin
LEVEL L
of dots.
1. Many new are printed every year , but
some wise prefer to the
old ones .
2. Telephone and were means of
unknown in the seventeenth
3. Much of the débris upon the
is valuable .
4. Most men themselves more

kindly than their judge them .
5. Power is generated -, gaso-
line , and several things .
6. A of ease is preparation for
achievement .
7. Sailors fear most and snowy

because then there is most of a collision
between
8. She was to fashion fine
from the cloth she had learned to
9. The of the river are being constantly
by the of the water .
10. Any will stick to a master is
and kind to
In the lines below , each number is gotten in a certain

way from the numbers coming before it . Study out what
this way is in each line , and then write in the space left for
it the number that should come next . The first two lines
are already filled in as they should be .
2 4 6 8 10 ......12.......
SAMPLES {{ 11 12 14 15 17 ......18......
11 . 85 79 73 67
12 . 90 81 72 63
13 . 76.3 85.3 94.3 103.3
14 . 64 32 16 8
15 . 240 120 60 30
16 . 12 16 22 26 32 36
2234
17 . 7 11 15 16 20 24 25 29
18 13 12 121 12
.
19 46 45 45 45
.
20 11 13 14 17
2
8
7
.
Directions and samples on page 161

as
21. downcast thrown down ... neutral judicious sad broken

...
...
1
22. pact puissance .... remonstrance ...... agreement skillet ...... pressure
1
23. audible festive ...... easy ..... audit..... heard ..... downy
1
24. solicitor lawyer ..... chieftain..... watchman ..... maggot ..... constable
1
25. beguile entreat ......delight...... dispense ...... deceive...... foster

1
26. dominate abide...... goad... threaten..... control... dissuade

1
27. average level..... count ..... evident..... ordinary..... distinct

1
5
28. behave act...... own...... keep still..... enable...... entitle
|||
29. comely ignoble ..... handsome ..... disagreeable ..... enter...... time
in
1
5
30. cycle scythe...... cyclone ...... circle...... ode...... junction
5
,
nearly the same the sentence printed heavy type
in
as
.
Man's evil manners live brass their virtues we
.-
31.
in
;
write in water Shakespeare
)
(
.
Some rise by sin and some by virtues fall
.
The evil that men do lives after them the good
;
is oft interred with their bones
.
He lives in fame that died in virtue's cause
.
The memory of vices lives longer than the mem-
ory of virtues
.
32. In this world man must either be anvil or hammer

a
.
To get along man must be knocker
a
a
,
.
Man must either do or be done
.
Man cannot be neutral he must accomplish
;
something or lose out

.
Might right
is
No greater grief than remember days joy when

of
33.
to
misery at hand Dante

is
)
.
.
Misery loves company

Sorrows remembered sweeten present joy
.
To recall past pleasures but aggravate our

to
is
present miseries
.
sorrow's crown sorrow remembering

of
A
is
happier things
It
.
34. consolation to the wretched have com-

to
is
a
panions in misery
.
Society shipwreck comfort all

to
in
is
a
Misery acquaints man with strange bedfellows

a
Company misery makes light

in
it
When misery highest help nighest

is
is
,
35. Nothing emboldens sin much mercy

as
so
Spare the rod and spoil the child

.
.
Pardon one offense and you encourage the com-

,
mission of many
.
He that has charity merits no mercy .

no
Let the punishment fit the crime .
36. He counsels best who lives best .
Practice what you preach .
A poor cask may hold good wine.
A good example is the best sermon .
An ounce of prevention is worth a pound of cure .
37. The fault is not in our stars but in ourselves , that
we are underlings .-
( Shakespeare . )
Every man is the architect of his own fortune .
Man is the master of his destiny .
We can't read our fates from the stars .
If we are underlings , it is not our fault .
38. Every white will have its black , and every sweet
its sour.
It never rains but it pours .
Birds of a feather flock together .
Every cloud has a silver lining .
You cannot pluck roses without thorns .
"
The paragraph for task 39 is the Fly " on page 162 .
39. What is meant by " his origin " ?
The paragraph for task 40 is " Every Home Needs a
"
Garden on page 165 .
40. What feeling is usually said to be experienced by people
who see things grow as a result of their efforts ?
LEVEL M
Write words on the dotted lines so as to make the whole sen
tence true and sensible . Write one word on each inch of
dots .
1. Modern of communication should
closer to each other .
2. Astronomers are uncertain the planet
Mars is
3. Cleanliness is a item in securing and
good health .
4. More were killed , more houses

more money . during the Great -----9
than during any equal number of in

history .
5. In the time squirrels store
for food in the when the
is such that they cannot for things to
eat .
6. Columbus America , but it was
for another Italian ,
7. A boy on a farm things about
animals and which a city
usually does not
8. The wind the streets of
every flake of
9. The benefit to the from the
of science is incalculable .
10. The old days are often with
the present .
Tasks 14 and 15 must both be right to secure credit .

There are thus only 8 tasks ; and each counts as 14 .
A 3 for 5c . E 6 for 5c . 4 for 25c . J
B 3 for 10c . F 80c . per peck . K 2 c . each .
C 3 for 25c . G 40c. per lb. L 14c . per lb.
D 48c . per 50c per lb. per lb.
H
lb.
M
4
c
.
for 1.00
8
$
I
are articles costing shown above 1A

as
etc.
, D
B
A
C
,
,
means of 2A means of 3A means of etc. Sup-

A
A
1
3
,
ply the missing numbers lines 15 as shown lines

11
in
in
of to
and III
Use the empty parts the page figure on
II
4 8 3 to
I
,
A
.
cost as much as B.
I.
II
3
lb. cost just as much as

J. I.
III
G
2
.
lb. H costs times as much as

2
1
.
11 lb. H costs much as lb.

G.
as
1
.
1 1 4
12 lb. 66 66
lb. G.
F ฀
1
. .
lb. H.
BRE
66 66
13 peck
1
14 . 2 lbs . M 66 66
1 lb. L.
15. 1 lb. L 66 66
1 lb. M.
Write the numbers and signs in each line below in the
proper order , so that they make a true equation as shown
in the three sample lines . Use the bottom of the page to
figure on if you need to .
(3 3 6 = + 3 + 3=6
= X
7 X 4 = 20 +8
33
Sample lines 4 7 8 20 +
2 3 3 7 18 + −X ( ) 7 + 2 = 18 ― ( )
16 . 2 2 3 5 15
-
17 . 1 1 4 4 16 ) (
18 . 2 5 6 7 10
19 . 1 4 8 15 20 +
20. Counting that 25 dozen sheets of paper are worth ten
cents , how many sheets of paper are worth a fifth
of a cent ?
Directions and samples the same as on page 161 .
21. action 1 play......2 dee .......3 mention ..... 4 opinion......5 crime
22. avarice 1 ordinary......2 various ....3 empress......4 frailty
......5 greed
23. bearing 1 a large ring .... 2 behavior......3 cub... 4 commendation .......5 destination
24. allusion 1 aria..... 2 illusion .... 3 eulogy 4 dream ...
...
5 reference
25. dynasty 1 davenport..... 2 very unpleasant ... 3 framework 4 ruling family......5 engine
26. habitat 1 dweller......2 bodice.....3 prodigality......4 habit......5 home
27. adversity 1 ...
ill fortune..... 2 dialogue 3 advertisement ... 4 dislike...... 5 distemper
28. caprice 1 value...... 2 a star......3 grimace......4 whim
......5 inducement
29. ignominious 1 seductive ......2 not guilty....3 incontestable......4 ignorant..... 5 shameful
30. chastity 1 dissension ..... 2 pursuit......3 eminence ......4 purity.....5 punishment

31. What a man has , so much is he sure of .
There's many a slip ' twixt the cup and the lip .
He who hesitates is lost .
A bird in the hand is worth two in the bush .
32. Tho the knowledge they ( the ancients ) have left
us be worth our study , yet they exhausted not
all its treasures ; they left a great deal for the
industry and sagacity of after ages . ( Locke . )
Worth is wholly dependent on long use .
Build the present on a knowledge of the past .
in admiration of the
Do not neglect the present
past.
There is nothing new under the sun .
33. Cowards die many times before their death.-
( Shakespeare .)
Fortune favors the brave .
Discretion is the better part of valor .
The valiant never taste of death but once .
They suffer more who fear than they who die .
34. Some books are to be tasted , others to be swal-
-
lowed , and some few to be chewed and digested .
( Bacon . )
Reading is profitable to every one .
One should read only parts of some books , while
others should be carefully studied .
Only a few books repay one for painstaking
effort .
People's tastes differ in books .
35. Write it on your heart that every day is the best
day of the year .-
( Emerson . )
There is no time like the present .
Never do today what you can put off until to-
morrow .
Anticipation is better than realization .
A common delusion is that the present hour is
not the critical , decisive hour .
Our virtues disappear when put in competition
.-
36.
with our interests ( La Rochefoucauld . )
A dog with a bone knows no friend .
My teeth are nearer than my kindred .
Virtue is its own reward .
A good friend is my nearest relation .
If
men wish to be held in esteem , they must asso-
.-
37.
ciate with those only who are estimable ( La
Bruyere . )
What a man does shows what he is .
You cannot always judge a man by his sur-
roundings .
He who comes from the kitchen smells of its

smoke .
If you always live with those who are lame , you
will yourself learn to limp .
38. We too often forget that not only is there a soul of
goodness in things evil , but very generally also a
soul of truth in things erroneous .-
( Spencer . )
Falsity frequently has a nucleus of reality .
Beliefs that are shown to be untrue may , never
theless , be based on some element of truth .
Benevolence sometimes has evil consequences .
Evil is commonly due to error .
39. They build too low who build beneath the stars .
Not failure , but low aim is crime .
Hitch your wagon to a star .
He that strives to touch a star often stumbles at
a straw .
Wouldst thou reach stars because they shine on
thee ?
The paragraph for task 40 is " Every Home Needs a Gar
den ," on page 165 .
40. Copy the four words which most fully state the pur
pose of the X. Y. Z. magazine .
THE CONSTRUCTION OF COMPOSITE TASKS

With in the course of our investi
the knowledge gained
gations , we could now construct composite tasks for use in
measuring altitude or intellect which would be much supe
rior to these . But these will serve reasonably well .
If
we had begun our work with the knowledge which we
now have , we should also have proceeded somewhat differ
ently in their construction . The procedures which we did
use be reported here only very briefly .
will consequently
We shall preface them by a description of a more efficient
and economical method of construction of such composite
tasks , which we recommend for the future .
It is as follows
Select the special abilities which to-
:
gether constitute the sort of intellect ( call it intellect
abc . . . n ) for which composite tasks are to be constructed .
Select a sufficient number of single tasks to provide one
hundred for each special ability that is included at each
twentieth of the total range of intellect abc . . . n from the
lowest thousandth of human adults to the highest thou-
sandth ( or the proper segment of such a collection , if the
tasks are to cover only a part of this range ) . In this selec-
tion you trust your own knowledge and judgment . Have
twenty or more competent judges rank these tasks for intel-
lectual difficulty for the group whose intellect abc • n
you plan to measure by the tasks . Let them use as fine a
scale as is convenient up to two hundred compartments ,
and require the use of approximately the same number of
compartments by each judge ( say , 150 to 200 , or 75 to 100 ,
or 60 to 75 , or 45 to 60 , or 32 to 45 , or 25 to 32 , or 18 to 25 ) .
Express the results of this consensus by simple summing .
Arrange the single tasks in order of difficulty as estimated
by the consensus , and in series representing each the same
special ability ( unless some better way is found to insure
that persons to be tested understand the general nature of
the tasks , and do not fail because of misunderstanding
directions ) .
Test with a cross - section of these tasks from fifteen hun-
dred to twenty -five hundred individuals , taking about two
hundred from each of ten groups selected to represent dif-
ferent altitudes of intellect abc . . . n , such as , college
graduates , pupils in grade 12 , pupils in grade 9 , . . . adults
of mental age 4. Let the tasks used always begin at a point
where 95 % of the group of two hundred can succeed with at
least four out of five of the tasks . Be sure that each indi-
vidual has sufficient time . It will be found most convenient
to have each individual in the group attempt all of the tasks
used with that group .
Enter the score as c, x , or · (correct , wrong or omitted )
for each individual in each group for each task . Find the
percent of successes for each task in each group . Make

up composites containing 2 tasks of a, 2 tasks of b, 2 tasks
of c ...
2 tasks of n , putting in one such composite tasks
most nearly alike in difficulty . Call such a composite a 2n
composite . Find the percent of successes for each 2n - com
posite in each group which was tested by all its tasks . Plot
the successes and failures in each 2n - composite in at least
one group against the total score ( number of tasks cor
rect) , and compute the overlapping of the failures past the
median of the successes in that group . Compute the bi
serial r.
Combine the 2n - composites into 4n or 6n or 8n or 10n or
12n composites , using 2n composites which are neighbors
in difficulty , and making each composite large enough so
that its rt will be at least .90 for a grade population or
other group of approximately the variability of a grade
population . How large composites will be needed can be
judged from the size of r for the 2n composites , the self
correlations of the 2n composites , and the self - correlation
of the measure of i . This last should be approximately
1.00 .
The resulting composites should be nearly or quite as

satisfactory for measuring intellect abc . . . n as the 40
composites described in this chapter are for measuring In
3 A success in a 2n -composite is a case which has n or more right . A
failure is a case which has fewer than n right .
4 Use the group which most nearly approximates 50 % of successes with
the 2n -composites .
5
Let It!1 = theaverage r from the 2n composites .
""
Tt1 = the average self -correlation for a 2n composite .
66 the self - correlation of the measure of i .
11
66
Itx = the average self -correlation of a composite necessary to produce
an rt of .90 .
Ttl1
Then .90 =
VTtx 11
and n , the number of 2n composites necessary to produce a self -correlation of
Ttx can be computed from
Tt1
( = nrt1
2
,
)
.81r11 r₁₂
+
−
n
1
1
(
)
tellect CAVD . The rt , for any one of them should be very

close to 1.00 for all adults , or for any group of the same
chronological age . All the tasks in any one of them will be
enough alikein difficulty to seem neither much too easy nor
much too hard to those for whom the composite as a whole
is suitable .
The same procedure may be followed in constructing

levels for any ability which has what we have termed " alti
tude , " that is , which has to master tasks varying in diffi
culty . The difficulty may be in words that are harder to
spell , that is , require a higher altitude of spelling ability
for success ; or in temptations to dishonesty that are harder
to resist , that is , require a higher altitude of honesty to
pass ; or in hundreds of other sorts of tasks . But wherever
the concepts of difficulty and altitude are applicable , this
method of constructing measuring instruments is appli
cable .
At the outset of our studies , we lacked the knowledge
of how often and how far a consensus of expert judges
could be trusted in its estimates of intellectual difficulty ,
and the knowledge of how many single elements are needed
to give a reliable measure of intellectual difficulty , and the
knowledge of the essential impossibility of measuring the
intellectual difficulty of any single small task . So we did
not proceed in the way outlined above , but began with single
small tasks , estimated their difficulty by the percent of vari
ous groups which succeeded with each , combined these into
composites by special abilities , that is , into sets of ten or
twenty completions of approximately equal difficulty ; sets
of ten or twenty arithmetical problems of approximately
equal difficulty , and so on . The 40 element composites were
made by putting together a 10 completion composite , a 10
arithmetic composite , a 10 word knowledge composite , and
a 10 sentence - comprehension composite , which were , as
composite tasks , as nearly equal in difficulty as could be
found in our material .
This method does have the advantage that we have
means of conveniently measuring the difficulty of tasks in
these four abilities separately , and have made many such

measurements of value (these are reported in Chapter
VIII ) . The disadvantages are that our composite tasks do
not represent as narrow segments or slices of difficulty as
they might have done ; are not spaced apart as evenly as
they might have been , and required much more labor in
their construction than would have been the case by the
other method .
We shall describe briefly the derivation of the word

knowledge composites of ten single tasks as a sample to
show the nature and validity of the selection and the extent
of the experimentation involved . In the case of the others
we shall simply present the evidence that the elements of
each composite of ten ( occasionally fewer ) , do belong fairly
in that rather than in an easier or harder composite . We
shall then even more briefly relate samples of the evidence
by which these composites of ten were put into composites
of forty. Finally we shall state the facts concerning the
value of the composites of forty as intellectual tasks the
difficulty of which we shall later measure .
10 - COMPOSITES IN WORD KNOWLEDGE OR V

Consider the tasks shown below . Each ' Level ' or 10
Composite is , by our definition of difficulty , harder than
the preceding for such a group as persons twelve to twenty
years old or older who have lived in the United States five
years or more , since a smaller percentage of them will get
five or more of the ten elements right . The difficulty is
' intellectual ' to the extent that within any sub -group of
equal age the greater intellects will show higher percents
correct than the smaller intellects in the case of any word .
It may seem far - fetched and forced and an unhappy
consequence of our definitions to argue thus that it requires
more intellect to know such words as cloistered , madrigal
and ignominious , than to know such words as confess , ad
vertise and combat . A dull person , it may be said , could
learn the former as well as the latter ; and it is a matter of
range rather than level that he does not . There is much

force in this criticism , and we chose the case of Word
Knowledge as one illustration of the measurement of dif-
ficulty , in order to state the answer to the criticism .
Word Knowledge is representative of many tasks of an
informational character where many of the harder tasks
might have been in the repertory of the dull so far as the
essential difficulty of mastering them is concerned , but sim-
ply are not as a matter of observed fact . They are not
there because the greater intellect can learn more per unit
of time and has learned more at equal age ; range is posi-
tively correlated with level . Also there is , for any locality
and epoch , a certain rough order of acquisition , whereby
people usually do not progress to learn certain things until
they have learned certain other things . The former are
then ' harder ' by our definition although , if customs had
been reversed , they might have been easier .
Look at the first word in line Find the other word in

1.
Do the same in lines 2 , 3, 4 , etc. Lines A, B, C, and D show
the way to do it . Do all the lines you can . Write only one
number for each line .
A. beast 1 afraid......2 words .....3 large......4 animal......5 bird
4
B. baby 4 youth .....5
1 cradle.......2 mother ......3 little child....... girl
C. raise 1 lift up .....2 drag......3 sun...... 4 bread .....5 deluge 1
D. blind 1 man...... 2 cannot see......3 game......4 unhappy......5 eyes
2
LEVEL V1
Begin :
1. await 1 pace...... 2 slow .... 3 wait for......4 tired..... 5 quit
2. beautify 1 make beautiful ....2 intrude.... 3 exaggerate ....4 insure ......5 blessed
3. bug 1 insect 2 a vehicle 3 fiber.......4 abuse ......5 din
4. arrange 1 put in order......2 hasten......3 distance ......4 frighten......5 charge
5. different 1 not the same......2 quarrelsome ......3 better ......4 complete .…… not here
..5
.…
6. cotton cloth...... small bed..... hut flour..... herd
1
7. blacken fern...... interpose ...... impel..... make black..... slack

a
1
8. ablaze on fire...... slightly..... loaf about ..... urbane

………
ostensible ......
1
9. avenue justice...... arrival street.... jury...... library

...
1 1
5
.
10. bench tool..... pull ashore...... opinion...... seat...... pond *****

2
5
LEVEL V2
11. confess 1 agree.......2 mend.... 3 deny.... 4 admit......5 mingle
12. backward 1 downwards . 2 after 3 toward the rear.......4 defense.......5 arrears
13. advertise 1 detain ......2 explore ......3 give notice of......4 adverse...... 5 newspaper
14. combat 1 fight......2 dismay.....3 club expedition ... comb
...
5
4
15. blond polite...... dishonest...... dauntless ...... coy...... fair
5
3
2
1 1
16. broaden efface...... make level ..... elapse..... embroider ...... widen
2
5
17. chubby indolent....... obstinate....... irritable...... plump...... muscular
1
5 5
see clearly.... engage...... furnish..... disturb...... have do with
to
18. concern
4
3
2
load...... small boat..... hem... draught...... vehicle

1 1
19. cargo
...
2
5
20. clutch exploit...... grasp..... *******
nest..... flit cane
1
5
LEVEL V3
21. awe lamb...... fear... tool mound..... opera
1
2 2
.4
5
22. aged years active...... old...... merciful ..... punctual
....
3 3
5
1 1
23. arrive answer ..... rival enter force...... come

2
5
.
24. blunt dull... drowsy...... deaf.... doubtful ..... ugly

1
5
25. accustom disappoint..... customary encounter ...... get used..... business
5
4
3
wait...
2
1 1
26. bade gaze....... tool...... fetched... ordered

5
4
3
2
a
27. bog ebb...... disorder...... swamp...... field difficulty

1 1
2 2
3 3
28. cascade hat waterfall firmament disaster ..... box

...
4
29. bray cry of an ox ... frustrate....... raven's cry

.....
cry of an ass..... bowl

5
4
2
3
1
30. disembark unearth ..... go ashore..... dislodge ..... disparage...... strip

1
LEVEL V4
31. conspire plot..... breathe ...... rely..... die...... outrun
1
5 5
32. check error...... stop... flash..... rude...... haste

4
2
3
1 1
33. cherish dedicate ...... happy ..... covet..... hold dear...... marry
2
34. chirrup aspen..... joyful ..... capsize...... chirp..... incite

4
5
3
2
1 1
35. accessible indefatigable ...... successful .... limpid...... easy reach..... liable
to
2
36. dingy afraid..... hostelry...... small bell ..... midget...... dirty

1
37. edible auspicious....... eligible....... fit eat...... sagacious ...... able speak
to
to
1 1
2 2
38. confound discovered ..... fulfill ..... establish...... mix up..... expire
3
39. concur
4.4
...................... mongrel ..... pounce..... ramble

agree
1 1
40. contact tactful ..... hate...... injunction...... touch...... oversight

2
LEVEL V5
41. downcast thrown down .... neutral .... judicious sad ... broken
1
5
.
42. pact puissance ...... remonstrance ...... agreement ...... skillet..... pressure
1
43. audible festive....... easy..... audit....... heard....... downy

1
44. solicitor lawyer...... chieftain....... watchman ....... maggot..... constable

1
45. beguile entreat..... delight...... dispense ...... deceive ..... foster

1
46. dominate abide...... goad...... threaten...... control... dissuade

1
47. average level..... count...... evident ...... ordinary ..... distinct

1
48. behave act own .... keep still..... enable...... entitle

1
49. comely ignoble..... handsome ...... disagreeable ..... enter..... time

in
1
50. cycle scythe...... cyclone ...... circle....... ode...... junction ***

1
5
||
||
LEVEL V6
51. action 1 play..... 2 deed......3 mention ....4 opinion....5 crime
52. avarice 1 ordinary......2 various......3 empress ......4 frailty..... 5 greed
||
53. bearing 1 a large ring .....2 behavior......3 cub......4 commendation ...... 5 destination
54. allusion 1 aria 2 illusion .....3 eulogy ..... 4 dream...... 5 reference
55. dynasty 1 davenport......2 very unpleasant.....3 framework......4 ruling family....
|
5 engine
56. habitat 1 dweller ......2 bodice ..... 3 prodigality......4 habit.....5 home
57. adversity 1 ill fortune......2 dialogue...... 3 advertisement......4 dislike...... 5 distemper
58. caprice 1 value......2 a star......3 grimace......4 whim inducement
...
5
59. ignominious seductive ..... not guilty..... incontestable..... ignorant...... shameful
1
5
60. chastity dissension ..... pursuit.... eminence ...... purity..... punishment
1
5
LEVEL V7
61. gainsay persuade ..... beshrew ..... deny..... profit..... imprint
1
5
62. eclogue obituary..... poem..... carousal...... epigram ..... portrait
1
5
a
63. cloistered miniature..... bunched...... arched....... malady ...... secluded

1
5
64. reciprocal saturnine..... mutual..... receptive ...... morose...... careless …………………
1
5
65. accolade salutation..... anchovy...... procession ..... bivouac .... acolyte
1
5
66. benighted fraudulent weary .... insuperable..... ignorant...... venal
1
5
67. madrigal song.... mountebank ...... lunatic ..... ribald..... sycophant ***********
1
5
68. pinnace boat..... doublet..... pinnacle...... hold fast..... forfeiture
a
1
5
69. broach dodge....... clasp....... open...... top....... edify
1
70. nectarine bouillon...... fruit...... jewel..... drink...... diurnal

a
1
5
a
Intellectual tasks range in this respect between two ex

tremes At one extreme the tasks are in and of themselves
,
.
,
almost quite impossible for the dull person regardless
or
of which things the world tries teach him At the other

to
the tasks are such he can master nearly or quite easily

as
as
as he can master any intellectual tasks the question being

,
rather how many dull person can master given age or

at
of a
with given set opportunities For example two our

of
a
,
.
very hard word tasks are

:
....
reciprocal saturnin ................ mutual.......... receptive
.
moros careless
nectarine bouillon.......... fruit jewel
...
a
a
.
drink................. diurnal
person twenty years old with mental age four not

of
A
only would not know the meaning reciprocal but also

of
probably never could be taught The idea involves think

it
.
ing of things by aspects and in relationships in a way that

is probably beyond his degree of intellect . He would not ,
save in rare instances , know nectarine ; but with proper
training he could know nectarine instead of some word , say
apple , which he does know.
Theoretically it is best to measure level or altitude " "
of intellect by tasks that lie toward the former extreme ;
and for practical purposes also , we may , in general , expect
better results per hour of time spent from using such .
They are likely to involve more of intellect , and to be less
adulterated by other influences than intellect , and to be
more representative of level and less of width or range .
However , the standard tests used for measuring intelli
gence contain tasks that range far toward the other ex
treme , and it is obviously desirable to measure the diffi
culty of these tasks and ascertain how much of it is due to
intellect pure and simple , and how much of it is due to other
factors .
Word Knowledge is a specially suitable case for study ,
because it has been approved by Terman as one of the very
best single measures of intellect , and is involved to some
degree in many of our better tests , such as oral and printed
directions , paragraph reading or comprehension , sentence
completion , opposites , and other tests of relations pre
sented in words .
We began with four hundred words chosen originally
to make an instrument for measuring word knowledge with
out regard to the merits or demerits of any one of them as
a measure of intellect .
The selection amongst these was made solely on grounds
of the percentages right in certain groups , the end sought
being to have for any one level word - tasks which were ap
proximately equally hard in the sense of being done cor
rectly by approximately equal percents of the group ; and
6 These matters will be treated in connection with new experimental data ,
to be presented in Chapter XV . We shall there see that the theoretical and
practical advantages are much less than has been supposed .
14
to have , at the next higher level , words which were done by

fewer of the group .
The procedure was as follows : 400 words , ranging from
very common words to words far outside the first ten thou-
sand as listed in the Thorndike Teachers Word Book , were
used in the case of 278 pupils in grade nine . On the basis
of the percents correct , 110 of the tasks were chosen ,
10 done correctly by 276 or 277 or 99.3 to 99.6 % of the pupils
10 66 "" ""
271 to 273 or 97.1 to 97.8 % " 66 ""
15 " " 66 " 257 to 261 or 92.4 to 93.9% 66 "" ""
15 66 66
228 to 236 or 82.1 to 84.9 % " 66 ""
15 66 ""
" 185 to 194 or 66.6 to 69.8 % " "" 66
15 "" 66
" 134 to 143 or 48.2 to 51.5 % " 66 ""
15 66 66 79 to 66 " " 66
90 or 28.4 to 32.4 %
15 66 66 37 to
51 or 13.3 to 18.3 % " "" ""
These 110 tasks were experimented with in the case of

430 pupils in grades 11 or 12 , 500 pupils in grades 9 or 10 ,
250 pupils in grade 81 , and 514 pupils in grade 6 , and
smaller groups of college students .
From them were chosen the seven ' Levels ' of ten tasks
each shown above . Levels 1 and 2 were constructed chiefly
on the basis of the results with the 514 pupils of grade 6.
Levels 3 , 4 , and 5 were constructed chiefly on the basis of
the results with pupils of grades 9 to 12. Levels 6 and 7
were constructed chiefly on the basis of the results with
pupils in grades 11 , 12 , 13 , and 17. The tasks within any
one level vary in difficulty somewhat widely and it is pos-
sible that results from as many thousands as we have hun-
dreds might show some tasks in adjacent levels which actu-
ally should be transposed .
Greater equality within and distinctness between levels
could have been attained by reducing the number from ten
to eight or fewer , but this did not , on the whole , seem de-
sirable . The order of difficulty of these tasks varies so
much from group to group , and so enormously from one
individual to another that , at levels where a person gets
from 20 % to 80 % right , the percent which an individual has
correct from one of our sets of ten is probably a more re-
liable measure of the percent which he would have correct

from a hundred tasks each of exactly the same difficulty as
the median task of the ten than is the percent which he
would have correct of the middle eight of the ten . Diffi-
culty is taken in the above to be difficulty for the sort of
persons who get about half right at the level in question .
The essential facts concerning the percentages correct
for each of the 110 tasks are shown in Table 28 .
TABLE 28
PERCENTS CORRECT FOR EACH SINGLE WORD OF SEVEN 10 -WORD COMPOSITE

TASKS IN EACH OF VARIOUS GROUPS OF INDIVIDUALS
Grade 6a 812 9 9 + 10 11 + 12 12 12
City N. Y. N. Y. Mix . K K K₁ K₂
Number of
Individuals 514 250 278 500 430 200 200
1 await 94.9 90.0 99.6 92.8 97.9 95.5 96.5

3 beautify 94.3 93.2 99.3 94.6 94.2 97.5 94.0
6 bug 94.6 94.4 99.6 97.8 99.5 99.5 99.0
7 arrange 96.0 95.6 99.3 97.2 99.3 98.5 98.5
9 different 94.5 93.0 99.3 97.4 100.0 99.5 96.5
10 cotton 93.4 93.2 99.6 96.4 98.8 98.0 96.5
12 blacken 95.2 94.4 97.5 98.0 99.3 98.5 98.0
13 ablaze 89.9 95.6 97.8 94.6 99.3 99.0 97.0
18 avenue 94.6 93.2 97.8 98.0 99.5 98.5 98.5
21 bench 92.0 90.8 93.5 92.2 96.3 92.0 95.5
22 confess 62.4 86.0 93.9 92.2 96.7 99.0 98.0

25 backward 70.9 88.4 92.4 87.6 90.5 95.0 94.5
26 advertise 69.0 82.0 93.1 79.6 88.8 89.0 89.0
28 combat 59.6 88.4 92.4 89.2 97.4 99.0 99.0
30 blond 62.4 63.2 92.8 87.2 96.0 97.5 98.0
31 broaden 62.9 83.2 93.1 94.6 99.1 97.5 98.0
32 chubby 64.6 78.8 93.5 92.4 95.8 97.5 98.5
33 concern 65.1 . 74.0 93.5 87.6 94.7 97.0 95.5
34 cargo 67.1 89.2 93.9 84.0 89.1 92.5 95.5
35 clutch 60.2 80.4 92.4 89.8 94.0 97.5 97.5
36 awe 29.4 62.8 82.4 69.0 83.5 89.0 86.0

37 aged 53.8 69.6 83.5 73.8 85.8 88.5 90.0
39 arrive 43.0 68.8 83.8 63.8 68.8 73.5 74.5
40 blunt 41.0 66.8 84.5 85.8 92.3 96.5 94.0
41 accustom 45.6 62.4 82.4 52.0 68.6 70.0 68.0
42 bade 45.1 84.9 72.6 82.8 84.5 83.5
43 bog 40.2 56.8 84.2 66.6 79.3 87.5 88.0
44 cascade 39.9 56.8 82.1 65.6 75.3 87.5 92.5
* Omitted because of a misprint in test .

Grade 6a 812 9 9 + 10 11+ 12 12 12

City N. Y. N. Y. Mix . K K K₁ K₂
Number of
Individuals 514 250 278 500 430 200 200
46 bray 54.7 62.0 84.2 79.4 85.8 93.0 95.5

50 disembark 50.4 66.0 82.4 65.4 81.4 89.5 95.0
51 conspire 29.6 70.8 69.1 61.0 85.8 94.5 93.0

54 check 28.0 67.6 69.8 50.6 71.4 80.5 77.0
56 cherish 22.9 40.4 68.4 48.4 70.5 72.0 79.0
57 chirrup 39.4 56.8 68.0 66.0 70.2 73.0 71.0

58 accessible 29.8 58.8 68.4 54.8 82.3 94.0 94.0
59 dingy 27.9 53.6 66.6 71.0 87.2 93.5 92.0
61 edible 30.2 42.4 69.1 57.2 73.0 91.5 87.0
62 confound 27.1 47.2 68.4 43.2 56.0 52.5 52.5
63 concur 40.8 56.4 68.0 61.6 79.3 83.5 80.0
64 contact 21.1 54.0 66.6 57.2 81.9 85.0 88.5
66 downcast 21.3 44.0 51.5 42.4 61.4 64.0 61.0
67 pact 19.2 38.8 49.3 29.2 63.7 77.5 77.0

69 audible 5.0 52.8 49.3 38.8 60.7 83.0 80.5
70 solicitor 45.0 57.2 49.3 39.8 47.0 71.0 68.5
71 beguile 45.6 48.6 44.4 59.1 47.5 51.0
73 dominate 42.0 49.3 42.2 70.7 79.0 82.5
75 average 26.8 48.2 52.1 61.6 72.5 67.5
78 behave 35.6 48.2 39.8 43.0 70.0 64.5
79 comely 45.2 48.6 39.0 42.8 62.5 64.5
94 cycle 40.4 31.7 37.6 50.7 64.5 67.0
81 action 24.0 29.5 23.4 33.3 46.5 42.0

84 avarice 31.6 29.2 31.0 42.3 60.5 58.5
86 bearing 34.8 32.4 29.0 34.9 54.0 43.5
87 allusion 17.2 31.7 22.8 32.1 43.0 43.0
90 dynasty 22.8 32.1 23.8 56.5 70.5 77.5
91 habitat 15.6 32.4 26.0 44.9 54.0 49.5
92 adversity 25.2 28.8 22.6 41.6 67.5 67.5
93 caprice 22.0 29.2 21.2 40.5 55.0 61.5
105 ignominious 17.2 17.6 17.6 30.7 41.5 42.0
107 chastity 26.0 16.9 25.2 38.4 64.0 64.0
88 gainsay 22.0 32.1 18.8 24.9 30.0 37.5

89 eclogue 40.4 30.9 23.8 23.3 33.0 35.0
97 cloistered 12.0 * 10.8 14.2 31.0 24.0
98 reciprocal 10.0 13.3 11.0 20.0 26.0 25.5
99 accolade 16.0 13.3 11.8 12.3 15.0 17.0
100 benighted 11.2 14.7 7.5 13.3 16.0 17.0
102 madrigal 22.4 17.6 8.2 11.4 21.0 28.0
104 pinnace 15.6 13.3 8.4 10.7 14.5 13.5
106 broach 16.4 18.3 14.6 27.4 39.0 34.5
110 nectarine 5.6 18.3 6.8 14.9 13.5 12.0
* Omitted because of a misprint in test .

Ninety tasks were chosen to represent harder words

than level 7 , and were used with one hundred college gradu
ates . From these ninety , four composites of ten each were
chosen to be most alike in difficulty within a ten and most
widely apart between tens . These four sets of ten were
used with 240 college graduates who were also tested with
levels 6 and 7. The results are shown in Table 29. We
thus obtain level 8 of about the same difficulty as 7 , and
levels 9, 10 , and 11 progressively harder . These levels
from 1 to 11 are competent to measure word knowledge
from below the level of the average ten -year - old to far
above the level of the average college graduate .
Composites 1a , 2a , 3a , 4a , 5a , 6a , and 7a , of approxi
mately the same difficulty as 1 , 2 , 3 , 4 , 5 , 6 , and 7 , were con
structed by testing many pupils in grades 6 , 8 , 9 , 10 , 11 ,
12 , and 100 college graduates with composites 1 to 7 and
also with 240 new tasks , obtaining the percents succeeding
with each of the 310 and selecting sets of ten from the 240
to match sets 1 , 2 , 3 , 4, 5 , 6 , and 7 , respectively . The facts
are shown in Tables 30 and 31 .
At the low end of the ability , the four sets A, B , C , and
D shown below were constructed by selection from about
twice as many on the basis of trials with 180 individuals
16 years old or older of mental age from 2 to 4. The facts
are shown in Table 32 .
Composites of ten intermediate between D and were I
constructed on the basis of the ratings of about 160 single
tasks by the consensus of twenty experts , and trials of these
with a hundred adults of mental age 6.0 to 7.0 , with 50
feeble - minded individuals in the same educational " class "
in an institution for the feeble - minded , with 101 pupils fif
teen years old or over in special classes in a large city , and
with 162 pupils in grade 4B ( second half ) . The facts con
cerning these word - knowledge tasks appear in Table 33 .
These composites intermediate in difficulty between
I
V D and V are imperfect in three respects . The dif
ficulty of each single task element is not determined from
enough cases . The oral picture selection tests are not

equated accurately enough with the oral word - selection
tests . The difficulty of written word - selection tests has not
been equated accurately against the difficulty of the same
sort of test given orally .
In general , we have devoted most of our work in the
preparation of composite tasks to making effective instru-
ments to measure altitude of Intellect CAVD from an alti-
tude corresponding roughly to a mental age of ten up to
very high levels . Our work with composites at lower levels
has been aimed first at demonstrating that Intellect CAVD
can be measured at the altitude of low imbecility , and that
we can , subject to certain limitations , locate an absolute
zero point for intellect and so , by later studies which will
bridge the interval between imbecility and our level I , at-
tach approximate absolute values to all the levels . We have
not been able to give adequate attention to the construction
of CAVD composites to bridge this interval and our com-
posites between D and I
are not so well made as the easier
and harder ones .
LEVEL 1A
Begin :
1. boyhood 1 childhood
..... 2 mischief 3 hardihood.... 4 cap..... 5 cherub .1
2. churchman 1 janitor ...... 2 member of a church......3 elector ......4 disciple..... 5 steeplejack 2
3. boyish 1 naughty..... 2 male......3 impudent......4 like a boy .....5 informal

4. cocoa chocolate ..... drug ..... chrysalis..... biscuit..... trivial 4
a
1
5. bottomless 1 artless..... 2 deeper.....3 unreasonable ..... 4 ultimate..... 5 without bottom .5

6. assistant orator...... perseverant ...... progressive ...... hand..... helper 6
at
1
7. chauffeur 1 carter 2 stove.......3 hot water 4 coachman automobile driver .....

...
7
5
8. dine sprawl ..... visit..... make noise..... have dinner...... bespeak .8

22
a
1
9. blouse whisk...... storm..... below..... pouch...... waist

9
1
10. cafe chaperon ..... theater...... restaurant..... flask..... festivity .....10

1
LEVEL 2A
11. dandruff ruffle...... scamp...... bald ....
dastard...... disease of the scalp .11
1
12. abashed ashamed...... overpowered ..... overlooked ... bruised ..... lowered ..12
1
13. bethink dream..... forget...... ascertain .... call

molest..... mind 13
to
......
1
14. comical funny...... coming ..... placid..... typical ... alert ......14
1 1
2 2
3 3
15. apology excuse..... verdict tribulation conclusion .... disease 15

4
...
.
16. clung held fast... part of wheel .... stung part .... nestled ..16
1
5
a
17. amidst among .... drenched ..... middle ..... lost...... partly ......17
1
5
18. baste 1 sew ...... 5 dump

2 list......3 calico..... 4 wallow .....18
2 without reason .....3 ineffective......4 highway......5 faultless 19
... ...
19. causeless 1 eventual
20. aster 1 flower ......2 bitter......3 matin...... 4 star......5 guilder 20
LEVEL 3A
21. ballot song ..... vote...... ammunition..... dance ..... award 21
5
4
3
2
1
22. rinse scald..... wash ... smear...... wrench...... grin 22

1
5
23. barge seaport..... ..... tonnage..... expansive ......
knock boat 23
... ...
5
3
4
2
1 1 1
24. acquit do...... free of blame leave...... aquatic.... pipe 24
...
2
5
25. cambric brittle... linen ..... moccasin ...... leather ...... crochet 25
......
2
5
26. brawn strength brood..... brine...... burnt.... bolster ......26
5
4
3
2
1
27. appreciation forbearance..... accomplishment ..... speech..... sympathetic

1
....
recognition..... sermon 27
5
28. alliance league...... enchantment .... slander ...... hypocrisy...... assembly 28

......
1
5
29. deceiver spy..... 29
... ...
detective ...... illusion ..... cavalier...... cheat
1
5
30. calculate marvel ..... administer...... plaster..... reckon..... convene 30
1
5
LEVEL 4A
31. childlike innocent...... saucy..... foolish... piteous .... affectionate 31
1 1
2 2
5
32. betwixt confused ..... braided..... between...... bewitched ..... pinched 32
3
5
33. crafty meager ...... difficult ..... adjacent sly artistic 33
...
4
5
3
2 2 2
1 1 1
34. outstrip subside...... outer edge...... outskirt...... satiate...... out run ..34
3
-
35. available hidden at hand economical lamentable ......5 useful 35
......
3
.4
.
36. certify exhort ascertain boast fuse ......5 assure ..36

1
37. annihilate crucify...... enamor ..... nihilist..... destroy 37
...
dead......
1 1
2 2
38. contentedly fully..... heretofore..... without stop...... cheerfully..... massy ......38

3
39. carcass mold....... body cargo.... rind..... hold of ship 39
...
...
a
1
40. console alone...... qualify ..... visit...... thin sole...... soothe ..40
1
LEVEL 5A
41. amen be ... hymn..... proverb
...... farewell..... communion 41
so
it
at 5
3
2
1
42. brawl pouch..... roast... hoot quarrel... lie length 42
...
1
5
.
43. debase degrade ..... base..... chastise ..... blaspheme ...... unfounded .43
1
44. adventurous clamorous ..... casual ..... bold travel .... advancing 44
1
4 5
45. adequate capricious..... conscientious ..... enough added..... water supply 45

...
......
1
46. amiable tractable...... trusty...... passionate ..... pleasing..... odious .46

1
47. ally league ...... associate ...... council..... factor...... navigator 47

1
48. benefactor patron..... churchman tourist..... sexton ..... advantage 48

.....
....
1
49. bethought perhaps ...... credulous ...... forget bewildered ... considered ..49
5
4
3
2
1
50. aperture through..... precipice..... opening .... raiment...... opportunity ......50

1
LEVEL 6A
51. ascribe attribute...... pertain...... clerk...... write...... upbraid ......51
1 1
52. default defeat...... blame...... failure..... libel5 displace 52

...
2
53. apparition ghost ...... insurrection...... apparent..... farce...... apparel 53

1
.… 2
.……5
54. appliance request adjustment...... conformity..... device pliant 54

..2
...
.…
.…
...
1
55. churlish craven ..... rude contemptible...... envious ......55

....
reckless .....
1
56. sexton cube...... janitor ..... compass ..... archbishop ..... six singers ..56
1 1
2 2
3 3
57. buckler keel servant stag shield....... scraper ....57

4
58. animosity hatred... animation...

disobedience ...... diversity ..... friendship .......58
1
59. conflagration carnival..... celebration..... decoration with flags...... contagion...... fire ..59
1
60. confidential respectable ..... secure...... sensitive...... secret...... confident ......60

5
4
2
3
1
LEVEL 7A
61 scrivener 1 searcher 2 forger......3 chaplain ... 4 clerk
..... 5 sceptic ......61
bottle ..62
....
62. beaker 1 cup.......2 binnacle...... 3 beak.....4 slanderer
5
63. emanate populate ...... free...... prominent rival come ......63
5
...
4
3
2
1 1 64
64. landau pier...... coach...... postern gable...... headdress
5
4
3
2
65. amaranthine jubilant ..... bitter..... maritime...... ungracious...... purple ..65
5
4
3
2
1
66. athwart alongside ...... above...... alert across thwarted ......66
5
4
3
2
1
67. conscientious guilty..... cautious ...... efficient ... good..... knowing ..67
.. 5
3
2
....
1
……
68. ingenuous ungenerous ...... unselfish dull..... frank ..5 unthinking ......68
4
3
2
1 1
69. betimes hereby ..... sometimes ...... meantime ... early..... now and then ......69
5
4
3
2
70. lambrequin knapsack...... drapery...... raw wool..... matting..... chandelier 70

......
5
4
3
2
1
TABLE 29
PERMILLES CORRECT IN THE SINGLE TASKS OF WORD KNOWLEDGE
- Composite Tasks and
10
11
10
9,
8,
T.C. Grad L. Grad T.C. Grad L. Grad
.
.
100 240 n 100 240
=
n
=
=
n
n
10
8
1. monomania 550 392 shrievalty 250 283

1.
2. saturnalian 520 375 2. sessile 210 179

pristine 510 421 teleological 210 221
6. 5. 4. 3.
7. 6. 5. 4. 3.
quaternion 540 346 peccancy 210 358

predatory 520 571 cacophony 240 413
persiflage 500 521 pediment 250 254
7. encomium 480 600 licentiate 190 154
8. abattoir 480 613 ambulatory 220 317
8.
9. meticulous 510 658 9. murrain 230 133
10. largess 500 429 10. cantilena 230 288
11
9
radial 400 408 saltatory 190 121

1.
1.
sequestrate 350 529 2. amerce 110 154

3. 2.
tactility 360 204 3. distrain 130 458
4. apogee 320 363 4. besom 090 154

nugatory 320 525 5. rhodolite 090 138
5.
6. sedulous 350 363 6. rune 130 112

7. umbel 350 129 7. hermeneutic 100 021
8. asseveration 340 254 devolution 070 046
8.
abjure 340 342 palindromic 100 112

9.
9.
10. auricular 320 321 10. carmagnole 120 120

TABLE 30 .
PERMILLES CORRECT FOR EACH SINGLE WORD OF THE SEVEN 10 -WORD COMPOSITE
TASKS 1A , 2A , 3A , 4A , 5A , 6A , 7A IN EACH OF VARIOUS
GROUPS OF INDIVIDUALS .
6b 6c 812 9k 10k 11k 12k

n = 139 n = 105 250 306 311 224 195
1 boyhood 777 990 904 917 927 933 933

2 churchman 820 971 928 933 960 964 970
3 boyish 777 942 932 891 940 946 964
4 cocoa 805 990 848 911 921 937 949
5 bottomless 683 933 935 968 982 982 979
7 assistant 640 895 952 952 960 982 985

9 chauffeur 604 942 976 968 976 991 979
10 dine 626 933 952 952 972 996 990
13 blouse 604 914 956 968 966 996 990
15 cafe 546 933 932 981 976 991 990
11 dandruff 590 790 896 965 969 991 990

16 abashed 661 628 752 757 828 812 872
17 bethink 460 752 892 863 886 875 923
22 comical 554 809 952 964 985 996 995
23 apology 446 834 964 912 921 937 954
24 clung 496 866 928 967 966 991 970

31 amidst 446 781 856 843 892 914 923
32 baste 446 743 640 824 857 914 913
34 causeless 410 790 820 819 914 954 970
39 aster 417 657 532 889 950 946 659
33 ballot 424 514 756 771 824 825 816

35 rinse 388 581 676 637 683 749 852
42 barge 395 638 836 752 737 888 831
45 acquit 453 457 724 523 647 852 841
47 cambric 460 343 504 706 747 861 887
58 brawn 316 486 736 569 700 834 846

59 appreciation 374 571 708 676 786 847 821
61 alliance 244 447 728 667 728 830 826
64 deceiver 252 609 732 775 721 812 826
86 calculate 093 371 720 598 728 843 836
36 childlike 496 324 356 500 528 602 718

46 betwixt 230 343 464 572 631 772 785
52 crafty 273 457 752 542 583 669 657
60 outstrip 244 324 660 539 670 727 713
67 available 173 257 504 494 715 852 852
TABLE 30 — Continued .
6b 6c 812 9k 10k 11k 12k

n = 139 n=
= 105 520 306 311 224 195
68 certify 108 447 532 543 570 683 677

78 annihilate 201 171 507 667 825 841
80 contentedly 302 486 588 549 686 754 785

94 carcass 144 352 580 549 615 731 881
113 console 065 228 556 509 663 785 821
50 amen 237 466 564 425 480 629 559

54 brawl 266 305 568 350 441 598 636
79 debase 273 267 432 294 438 665 682
84 adventurous 122 257 432 399 486 500 584
89 adequate 187 114 336 363 425 598 657
93 amiable 209 238 448 363 460 624 667

100 ally 201 219 416 366 441 611 672
103 benefactor 173 314 384 355 409 558 652
108 bethought 137 152 520 359 502 549 616
109 aperture 093 133 416 342 460 566 616
63 ascribe 345 324 256 275 316 317 416

69 default 108 219 256 271 316 352 390
85 apparition 151 124 360 164 219 415 605

88 appliance 165 162 224 157 267 406 498
101 churlish 230 162 292 229 283 312 359
107 sexton 216 162 300 228 332 379 462

112 buckler 165 228 220 211 267 526 374
125 animosity 065 124 364 176 264 388 482
137 conflagration 022 048 260 160 293 459 451

138 confidential 124 057 216 121 216 357 457
C 53 scrivener 058 146 158 185

C 73 beaker 106 158 231 431
C 76 emanate 067 091 098 154
C 79 landau 080 101 133 190
C 83 amaranthine 102 126 150 159
C 88 athwart 067 101 197 113

C 89 conscientious 061 126 115 195
C 90 ingenuous 128 154 171 195
C 93 betimes 054 032 051 082
C95 lambrequin 096 066 098 149
* Omitted because of misprint in test .

THE CONSTRUCTION OF 10-COMPOSITE TASKS IN SENTENCE COM

PLETION , ARITHMETICAL PROBLEMS , AND THE UNDER
STANDING OF SENTENCES AND PARAGRAPHS
The for C , A , and D were constructed by

10 - composites
the process of trying many single tasks with various groups
and selecting tasks of similar difficulty , which has been de
scribed and illustrated in the case of V. Only the main re
sults will be presented here . They are in the form of tables
TABLE 31 .
PERMILLES OBTAINING FIVE OR MORE RIGHT OUT OF TEN IN THE VOCABULARY

COMPOSITES 1 , 1a , 2 , 2A , 3 , 3A , ETC.
Grade 52 81/2 9 10 11 12
N. Y. N. Y. K K K K
n= 148 250 1089 723 769 643
V 1 993 980 993 996 997 994

V la 993 976 993 993 999 997
V 2 905 920 989 989 996 994

V 2a 959 972 995 991 999 1000
V 3 615 740 913 924 967 975

V 3a 601 864 914 929 977 987
V 4 645 749 801 936 946

V 4a 618 764 839 931 962
V 5 440 428 560 748 824

V 5a 448 473 604 801 846
V 6 152 129 290 473 560

V 6a 236 183 299 480 593
V 7 044 017 030 061 107

V 7a 060 017 031 104 137
giving the percent of successes for each single task in each

group . The constitution of the group sometimes varies
within a table , because sometimes in a certain group some
tasks would be assigned to only a part of the group . Where
this is the case , the fact is noted by printing the new n in
the body of the table . The n at the top of a column applies
to all entries in that column unless a second n appears in
the column . If
a second n appears in the column , it applies
to all entries below it unless a third n appears ; and so on .
Percents are strictly comparable only where they are for
the same n.
The sentence - completion 10 - composites are A, B , C , D ,
J
E , F , G, I , , K , L , M , N , O , P , and Q. The main facts con
cerning these are shown in Tables 34 , 35 , and 36. We also
TABLE 32 .
PERCENTS CORRECT IN THE SINGLE TASKS OF WORD KNOWLEDGE : COMPOSITE

TASKS A, B , C AND D. 180 ADULT IMBECILES .
100 n = 80 n 100 80
=
n
n
=
A 1 76 81 26 36
C
6 5 4 3 2 1
2 71 79 25 15
3 74 75 24 31
4 76 80 24 49
5 76 89 23 19
6 73 85 22 2212
7 72 73 21 30
∞ 7
8 72 79 20 24
8 8
67 772 21 25
9
10
90 67 76 10 21 30
B 48 55 D 12 21
8 7 6 5 4 1232 1
1
49 522 15 9
8 7 6 5 4 123
51 61 11
9
46 622 14 16
44 422 15 122
47 54 122
9
40 36 14 172
43 50 72
9
41 572
9
6
6
10 39 56 10 122
have certain provisional completion composites and

10
J
I-
-
which will be useful until better ones are constructed

R
Some of these composites and also some of the arith

metic and directions composites be presently described
to
could probably be improved by transfers of some elements

.
We have not made these transfers because the gain would

,
not be great and the labor of recomputing the composite

TABLE 33 .
PERCENTS CORRECT IN THE SINGLE TASKS of Word KNOWLEDGE E , F, G, AND H.
n = 100 n = 50 n = 101 n = 162

M.A.6 F. M. Spec . 4B
VE 1 gasoline ( picture selection )

"" 66
59 100
2 crayon 51 64
3 tresses "" 55 20
4 refrigerator " 66 48 70
plume 66 66 48 96
5
6 entrance 66 66 46 32
porridge 66 "" 44 74
7
8 hide 42 76
9 drummer "" "" 35 68
10 ram "" 66 38 60
VF 1 rock ( verbal selection ) 26 94 94.1 98.1

"" "" 25 92 97.5
2 people 85.1
66 ""
3 large 23 90 90.1 96.9
4 "" 66
heaven 23 88 88.1 92.6
"" "" 22
5 speak 88 84.2 95.7
6 mountain "" "" 22 92 96.0 95.1

7 dark 66 66 23 94 80.2 93.8
8 kind "" "" 21 94 77.2 96.9
""
9 quiet 20 80 77.2 90.1
10 short "" 66 79.2 93.2
90
VG 1 good
66 "" 70 69.3 87.7
still
220
2 66 "" 16 70 68.3 95.1

warm "" 21 72 81.2 95.1
"
"
4 3
walk 17 72 69.3 95.1

""
""
behind 17 76 73.3 94.4

"
"
""
5
near 66 16 58 72.3 92.0

""
9 8 7 6
fast 66 21 74 71.3 95.7

" "" 12 62 80.2 98.1
once
6
sweet 66 12 58 71.3 90.1

""
66
10 bring 70 71.3 87.7
7
VH love 66 66 18 62 61.4 84.6

1
lift " 58 61.4 83.3

"
"
"
2
8 6
great " 70 63.4 91.4

"
"
"
5 4 3
66 58
nation 53.5 71.6
""
"
65
space 38 44.6 73.5

"
"
"
none 66 72.3 90.7

60
""
8 7 6
66
44
every 38 36.6 70.4

before 11 56 66.3 84.6
"6
""
pair 48 47.5 74.1

""
"
"
7 0
9
today 66
17
10 38 49.5 75.3
task scores and correlations which have been obtained from

these 10 - composites would greatly outweigh the gain .
The arithmetical 10 - composites were constructed in the
same way by trial , selection , and retrial . It is difficult to
secure large groups to take long experimental tests , and
TABLE 34 .
PERCENTS SUCCEEDING WITH EACH SINGLE TASK OF VARIOUS 10 -COMPOSITES

IN TWO GROUPS OF ADULT IMBECILES .
n = 100 n = 80 n = 100 n = 80 n = 100 n = 80

CA 1 78 84 A A 9 8 7 6 5 4 3 2 1 67 64 DA 83 872
1
77 79 53 572 84
78 81 75 73 69
82 81 56 672 80
9 8 1234567
68 722 67 65 81 86
85680628
59 71 64 73 84
1234567∞00
69 74 67 722 72 70
67 622 61 672 86
9 8
~
10
63
81
66
73 10
72
60
64
71 *
73
63
75
73
CB 1 46 56 A B 28 50 DB 59 65
1
9 8 7 6 5 4 3 2 1
65 56 55 46 43 572
7 6 5 4 3 2
57 38 49 42 58 66
56 55 49 44 47 50
9 8 HASHDON
53 50 40 54 40 44
50 45 55 51 43 75
1234567DO
45 42 422
********
56 49 44
51 49 27 49 42 59
9 8
48 56 44 49 40 74
10 40 39 10 57 40 10 472
CC 1 26 26 AC 26 24 DC 35 34
9 8 7 6 5 4 3 2 1
9 8 7 6 5 4 3 2 1
37 30 20 24 24 29
9 8 7 6 5 4 3 2
34 35 34 24 27 36
44 34 23 29 17 322
36 322 38 17 23 19
37 31 20 19 272
9
HABIBONDO
HOSHONDO
27 26 16 30 20 222
31 222 21 32 30 35
30 272 15 22 20 24
10 22 172 10 21 19 10 22 25
CD 1 35 A D 24 DD 16 26
5
7 6 5 4 3 2 1
9 8 7 6 5 4 3 2 1
17 19 18 16 10
8
9 8 7 6 5 4 3 2
12 11 10 16 25
19 15 11 11 10 16
9 8 HASHBGT
13 14 16 24
2
HASHDOTADO
23 11 2212
8
-0
22 14 13 30
1 9
22 12 12 14 172
64241
13 21 10 16 172
10 11 10 17 10 22
5
9
.
TABLE 35
10
PERCENTS
:
SUCCEEDING WITH EACH SINGLE
,6
TASK OF VARIOUS COMPOSITES IN FOUR GROUPS 100 ADULTS OF MENTAL AGE 50
-
3
-,
A
,
FEEBLE MINDED OF CLASS IN AN INSTITUTION PUPILS IN SPECIAL CLASSES IN LARGE CITY AND
4
(
)
PUPILS IN GRADE SECOND HALF
F.
.
F.
.
M.A.6 M.
F.
Spec 4B M.A.6
.
M. Spec 4B M.A.6 M. Spec 4B
=
n
100 50 101 162 100 50 101 162 100 50 101 162
CE 50 78 91
1
96.3 AE 46 86 DE 37 86 85
12
12
49 90 92 94.4 45 84 31 86 88
3
58 92
2 3
93 98.8 49 92 91 99.4 48 84 80
46 78 92 98.8 48 92 100 99.4 53 88 88
60 82 85 96.9 57 59 88 82
3 4 5 6
49 98 86 95.7 51 38 84 79
41 82 83 96.9 48 56 90 96
43 90 97 93.2 49 38 96 97
8 9
45
78 9
82 86 72.8 59 38 90 97 85.2
10 47 100 99.4 10 47 10 50 80 94 93.2
C
F 31 80 83 96.3 AF 33 78 89 96.9 D F 35 60 71
31 82 79
1 2
98.1 23 76 87 97.5 34 80 74
1 2 3
123
24 90 81 87.7 26 88 93 100.0 27 68 90
20 84 89 97.5 22 76 99 97.5 27 80 95 95.1
5
24 74 73 77.2 24 82 97 96.9 25 76 87 85.8

36 56 82 84.0 20 90 96 97.5 23 50 67 82.7
37 64 65 80.2 35 86 98 97.5 19 68 86 80.2
8
27 82 87.7 22 78 91 98.8 29 94 92 90.7
5 6 7 8 9
8 9
34 80 92.6 22 82 92 98.1 12 70 85 82.7

10 25 76 82.1 10 23 84 96 100.0 10 17 82 85 75.3
197
198
(
TABLE 35 continued
F.
.
F.
.) .
M.A.6 M. Spec 4B
.
M.A.6 F.M. Spec 4B M.A.6 M. Spec 4B
50 101 162 100 50 101 162
=
n
100 50 101 162 100
G
1
D
48 69 64.2
1
85 92.0
1
CG 10 38 47 87.7 AG 17 64
56 51 63.0
6
130
36 62 76.5 11 48 84 93.2
64 90 95.7 34 71 67.9
13 44 57.4
76 90 94.4 42 70 76.5
11 60 60.5
64 87 95.7 60 68 67.9
21 44 66.0
68 93 95.1 50 63 65.4
38 66.7
56 85 94.4 56 58 71.6
40 67.9
THE MEASUREMENT
54 93 90.7 56 64 41.4
56 50.0
54 75 59.3
8 9
68 88 95.1
1234567899
YE57486877
36 63.0
1234567890
1234567000
72 85 95.7 10 50 52 69.1
36 67.9
1
42 45 60.5
1
AH 11 56 62 79.6 DH
62 81 90.1 34 54 63.6
28 64 80.2 50 55 66.0
60 65 92.0 34 45 56.8
34 54 86.4 38 49 58.0
46 58 74.1 54 60 43.8
0
34 54 90.1 38 48 40.1
OF INTELLIGENCE
48 54 93.2 42 47 64.2
81.5 56 51 48.1
16134GOLBO
40 69
1234567890
1234567890
0 28 59 75.3 30 45 59.9
TABLE 36 .
THE PERMILLES SUCCEEDING WITH EACH SINGLE TASK OF VARIOUS

10 -COMPOSITES OF SENTENCE COMPLETIONS .
Grade 52 812 6 (1) 6 ( 2) 6 (3 ) 6 (4 ) 17 or +

n= 205 250 61 100 107 140 60
CI
654 787 490 738 650 933
654 672 610 682 593 950
693 443 530 748 564 900
746 812 607 460 757 578 1000
123456
634 804 639 380 682 564 950

634 824 623 460 626 585 800
n 162 80 75 104
=
n
=
=
n
n
649 840 667 362 787 510 933
639 796 543 500 773 779 983
580 800 617 413 667 519 891
7880
9
10 751 832 630 587 667 693 983

J
C
541 580 630 437 720 712 941

1
49 52 140 59
=
=
n
n
n
605 696 531 692 690 893

2
375 177 186 = 248

=
=
=
n
n
n
498 780 501 429 613 564 783

532 776 481 350 600 539 983
345
493 704 925
341 736 443 260 495 436 958

463 664 377 290 542 364 717
358 708 278 150 547 443 891
6780
294 681 279 280 402 450 900

9
10 206 673 352 79 414 314 925
K
C
116 n == 59
=
n
376 364 379 890

1
184 49 53 126
=
=
n
=
n
=
n
225 432 386 122 547 238 825

270 644 288 41 679 262 867
234
196 336 255 184 453 333 917
n 49 52 140
=
=
n
377 564 61 442 155 861

279 568
353 580 123 250 293 202 808
56789
225 336 185 137 280 192 683

152 476 246 110 299 150 983
10 191 476 180 90 280 186 908
15
TABLE 36 Continued .
CL
This is the least satisfactory 10 -composite . It was used because , as a com
posite , it filled a certain place . The 10 single tasks in order showed percents
correct of 60 , 59 , 60 , 54 , 42 , 41 , 15½ , 11 , 5½ and ½ in a group of 200 pupils
in grade 9 .
C M
The 10 single tasks in order showed percents correct in a group of 200
pupils in grade 9 of 22 , 20 , 36 , 38 , 30 , 25 , 28 , 20 , 26 , and 25 .
Grade 5½ 8½ NS ( 1 ) NS ( 2 ) NS NS 17or 17or 17or S.Sch 10,11,12
.
n = 205 250 100 100 135 87 60 28 17 35 82
CN
1 020 200 470 530 482 678 830 821 882 857 350
2 083 160 690 530 467 609 817 821 824 771 386
3 005 080 530 420 297 483 830 964 882 886 446
4 167 240 600 570 526 713 733 786 647 914 349
5 010 108 410 470 341 540 667 964 647 829 277
6 020 164 450 500 356 506 770 893 706 714 578
7 010 92 540 550 400 575 746 893 647 857 602
8 025 100 420 400 259 506 627 964 824 600 747
9 034 192 380 580 326 609 686 893 824 771 482
10 108 148 680 710 511 759 885 893 882 943 482
со
090 160 126 253 600
290 220 200 264 577
430 380 363 391 551
12345
110 150 097 172 433

280 210 297 368 442
260 330 259 448 596

250 230 297 253 619
410 490 445 586 636
67899
350 430 304 425 593

10 210 190 208 345 610
P
C
040 020 067 000

5 4 3 2 1
150 170 193 207

050 040 030 034
100 060 059 161
100 050 082 149
5
TABLE 36 — Continued .
Grade 5½ 8½ NS ( 1 ) NS ( 2 ) NS NS 17or + 17or + 17or + S.Sch . 10,11,12

n = 505 520 100 100 135 87 60 28 17 35 82
120 090 119 138

67 260 240 297 310
8 100 100 082 149
9 030 090 052 080
10 110 100 111 057
C Q
080 080 082 115 750 353 171
2 1
000 050 015 000 429 235 114

000 020 008 023 571 765 229
512345
030 030 015 023 500 765 257

4
020 040 030 023 500 824 257
000 040 000 000 429 588 171

020 020 015 011 536 529 229
020 030 000 000 607 471 229
6789
000 010 000 000 571 706 171

10 000 010 000 000 393 706 086
CR
000 000 000 000 321
000 000 000 000 464
000 010 000 000 429
12345
000 000 000 000 250

000 000 000 000 500
5
000 000 000 000 250

000 000 000 000 393
6797
000 000 000 000 250

9 8
000 000 000 000 286

10 000 000 000 000 536
when they have been secured especially hard obtain

to
is
it
,
time enough exhaust abilities arithmetical problems

to
in
patient adult may work for half an hour single prob-

at
A
to a
lem We fear that the tasks which happen come late in

.
the series first printed showed fewer successes our

as
in
returns than they would have shown they had been at-
if
tempted first In general we feel less security that the per-

,
.
cents of successes correspond closely with degrees dif

in of
ficulty the case the arithmetical problems than the

in
of
TABLE 37 .
THE PERMILLES SUCCEEDING WITH EACH SINGLE TASK OF VARIOUS 10 -COMPOSITES

OF ARITHMETICAL PROBLEMS .
I
Grade Sp Sp 52 82 91 II
9
p= 50 52 189 126 246 264
260 346 751

5+54 1232 1
340 192 682

340 288 661
340 404 762
380 308 857
620 673 831

300 365 857
440 423 788
67890
280 250 815

10 140 135 656
A J
302 646 792
233 626 739
2
296 818 672 708

12345
545 778 846 799

4
217 627 512 538

5
370 690 667 633

217 643 715 610
344 603 768 527
6789
217 540 732 564

10 328 611 821 652
AK
317 532
143 643
175 571
12345
307 603
206 524
058 429
228 540
196 587
67899
139 579
10 105 341
TABLE 37- ( Continued ) .
Grade 52 82 91 9II NS ( 1 ) NS (2) NS NS 17
n= 189 250 246 264 100 100 135 87 240
AL
296 544
5123453 2 1
296 484
190 448
206 424
185 396
153 416
238 516
148 436
67899
127 376
10 201 484
A M
059 174 680 570 437 563
093 246 730 650 481 598
102 220 700 650 452 644
12345
089 133 720 650 496 621

4
065 140 680 630 504 736
053 171 580 430 341 609

8780 6
057 129 650 600 407 701

057 095 600 510 348 667
8
AN
630 480 252 391 762
400 390 185 322 725
2
420 290 185 379 762

512345
310 260 222 230 662

4
470 370 185 333 742
530 420 378 517 829

470 450 326 437 683
410 430 326 471 817
67890
340 300 178 310 742

10 340 260 200 379 712
TABLE 37- (Continued ) .
Grade NS NS NS NS 17
2
(
)
(
)
n= 100 100 135 87 240
AO
120 140 104 161 792
120 130 081 230 779
130 090 104 149 642
12345
030 040 022 034 467

000 040 022 057 421
5
030 040 022 115 679

020 030 067 092 671
010 040 044 057 642
67849
020 040 052 057 700

10 020 050 059 092 600
AP
546
617
400
12345
504
650
579
629
562
67800
612
9
10 675
A
Q
519
343
5+14 123
423
502
218
case of or or D. However the errors this respect

in
V
C
are probably very small comparison with the difference

in
difficulty from Arithmetic Arithmetic

to
in
P.
A
The main facts for the arithmetical composites

10
A
in B
,
-
and are given

M
K
H
N
E
, G
D
Q
C
J
I
F,
,
,
,
,
L,
,
P,
Tables and 37. We have also certain provisional

34
35
,
TABLE 38 .
THE PERCENTS SUCCEEDING WITH EACH SINGLE TASK OF VARIOUS 10 -COMPOSITES OF
DIRECTIONS AND READINGS .
Grade 4 5 Ad 91 911 10 11 ( 1 ) 11 ( 2 ) 11 ( 3 ) 11 ( 4 ) 12 ( 1 ) 12 ( 2 ) 12 ( 3 ) Ad
n= 162 311 44 246 236 100 100 100 100 63 100 100 84 44
D 2
1 51.9 71.1 75.0 95.1 96 100 96 97 98.4
2 26.5 46.0 63.6 91.9 95 99 92 97 96.8
3 5.6 20.3 56.8 81.3 96 98 96 99 100.0
4 34.6 50.2 77.3 91.5 92 99 98 94 98.4
D 1
1 37.7 51.4 65.9 88.2 85 84 94 84 88.9
2 56.2 62.4 65.9 80.9 91 94 89 94 90.5
3 6.2 11.3 38.6 82.9 89 95 90 88 87.3
4 35.2 37.9 59.1 72.0 92 92 92 89 96.8
5 40.1 47.6 63.6 82.9 91 95 94 91 98.4
6 26.5 32.5 52.3 65.0 82 85 93 86 98.4

7 24.1 35.7 34.1 76.4 92 90 90 80 96.8
8 41.4 60.1 65.9 91.9 95 96 93 86 95.2
888888
9 6.2 17.7 50.0 79.4 83 98 93 90 94.0

10 46.9 55.3 47.7 58.9 81 94 91 87 90.5
D
2
17.3 20.3 50.0 62.2 52.6 75 82 81 79 79.4

5 4 3 2 1
5.6 8.7 38.6 47.6 45.8 69 85 83 82 81.0

25.9 27.7 52.3 57.7 53.8 68 88 77 73 79.4
6888
13.6 9.6 47.7 41.9 39.0 70 79 79 82 79.4

2290
5.6 15.4 45.5 42.7 79 782 85 73 88.9
16.0 23.1 47.7 42.7 77 83 77 75 74.6

9 8 7 6
4.3 16.1 52.3 78.5 74 84 76 75 84.1

19.8 30.2 43.2 65.4 78 78 83 76 81.0
78
28888
9.3 22.2 40.9 69.5 83 72 78 76.2

10 16.7 16.4 40.9 44.3 66 77 77 76 76.2
D 22
8.6 13.8 52.4 48.3 74 72 78 73 79.4 72 71 77.4
4 3 2 1
338
11.7 15.4 32.5 38.6 64 77 72 70 77.8 66 69 79.8

15.4 19.6 58.1 58.1 63 76 76 68 73.0 85 69 76.2
3.7 3.9 39.4 24.2 66 71 74 72 69.8 79 76 85.7
9.3 15.8 59.3 71 72 68 78 82.5 76 68 75.0
88
5
TABLE 38 continued
.)
Grade 4 Ad 91 9II 10 11 11 11 11 12 12 12
5
3
)
(
(
)
(
)
(
)
(
)
(
)
(
)
n= 162 311 44 246 236 100 100 100 100 63 100 100 84
53.8 67 77 74 64 67.7 81 79 83.3

20.8 56 71 71 72 69.2 68 69 60.7
42.0 54 62 73 64 69.2 68 67 70.2
6789
33.3 61 66 64 68 69.2 68 68 70.2
6888
10 34.1 48 63 66 61 60.0 71 67 61.9
3
43.1 37.7 67 66 65 57 55.6 70 73 61.9
31.7 23.7 64 73 71 65 77.8 69 66 66.7
2
40.7 22.5 64 67 66 54 58.7 60 61 63.1

12345
64 68 63 63.5 76 78 65.5
62
46.7 42.0
48.8 38.6 65 62 65 62 69.8 74 74 72.7
5
43.5 42.8 60 64 69 66 65.1 72 80 71.4

¢
65
28865
49.6 39.4 64
฀฀฀฀ 69 66 61.9 67 74 73.8
22.4 14.0 69 ฀
64 64 67 74.6 63 71 63.1
g
67899
86868
39.8 56 69 70 65 77.8 79 72 78.6

10 41.1 17.0 59 ฀฀
71 72 67 76.2 74 64 72.7
D
4
38.6 30.5 44 41 54 55 57.1 52 53 47.6

1
41.5 28.8 58 50 52 51 55.6 48 68 47.6

2
42.3 28.4 45 46 49 52 57.1 58 73 58.3

5 4 3
25.2 19.9 47 49 60 56 58.7 65 59 65.5

42.3 25.0 47 49 44 54 52.4 49 ∞
48 47.6
36.2 25.4 43 55 51 53 52.4 56 56 54.8

45.1 27.1 46 54 51 54 44.4 66 64 65.5
5555555
40.2 25.9 43 48 54 44.4 61 52 60.7

67890
19.5 15.3 40 45 52 52 50.8 60 45 61.9

10 32.9 12.6 41 49 43 42 42.9 41 47 38.1
42
D
41 42 42 46 60.3 41 59.5
51 58 45 48 47.7 45 51.2
123
63 50 52 42 47.7 67 59 58.3
3
4
64568
47 68 57 51 58.5 60 67.9
45
42 59 62 61 56.9 63 62 54.8
5
TABLE 38 continued
).
10 11 11 11 11 12 12 12
3
)
(
(
)
(
)
(
)
(
)
(
)
(
)
236 100 100 100 100 63 100 100 84
42 67 57 64 52.3 60 71 66.7
30 52 39 37 46.2 55 49 46.4
33 62 48 37 43.1 59 51 55.9
67890
49 52 61 63.1 52 44 48.8
559
10 50 53 61.5 49 58 59.5
฀฀
D
5
+2332
44 46 33 40 49.2 54 46 55.9
39 39 37 25.4 34 32 32.1
43 47 35 47 46.0 55 52 64.3
12345
31 43 39 38 36.5 43 41 46.4
27 30 35 32 33.3 53 35 42.9
27 37 31 34 34.9 38 22 44.0
28 40 37 25 49.2 45 41 40.5
25 43 30 46 46.0 38 41 38.1
67890
26 42 41 37 36.9 47 37 44.0
10 41 30 38 41 33.8 33 33 40.5
D
6
16 17 24 16 14.3 25 22 28.6
17 21 17 18 23.8 21 16 21.4
30 26 21 21 32.3 27 24 23.9
08 16 17 09 15.4 17 08 13.1
123456
19 21 20 22 16.9 17 24 21.4
20 24 17 05 26.2 27 18 28.6
6
D
7
00 03 02 0.15 00 03 02.4
3220
12 06 10 40.0 15 21 11.9
2
02 03 03 05 06.1 11 01 04.8
4 3
00 03 03 01 01.5 16
6C 1235
07 04.8
09 08 05 04 16.9 13 10 09.5
05 05 08 08 12.3 16 09 09.5
TABLE 38 continued
).
N.S. N.S. N.S. N.S.
2
)
(
(
)
100 100 135 87
52
D
1 31 37 27.4 46.0
2
2 27 34 26.7 29.9
3 52 44 40.7 39.1
4 54 49 34.8 43.7
5 20 40 27.4 23.0
6 37 46 38.5 28.7
7 42 42 34.8 31.0
8 41 45 34.1 36.8
9 47 40 39.3 26.4
10 31 26 25.9 19.5
62
D
1 29 31 23.7 23.0
2 10 5.9 5.1
20
3 24 21 18.5 17.2
3.0 5.7
*** 7 8
10 10.4 6.9
456
32 27 23.7 11.5
arithmetical composites IJ and which will

10
L
I
I
JI
,
,
-
be useful
.
In the case comprehension directions and para

of
of
graphs we have composites from six ten elements

of
to
A
,
and
11
31
51
61
21 41
B
D
E
G
C
,
F,
,
1,
,
2,
,
3,
H 4,
,
,
5,
,
6,
7,
and have provisional composites and The facts con

.
cerning the difficulty of the constituent elements of these

composites are given Tables 34 35 and 38
in
,
,
THE DIFFICULTY OF THE 10 COMPOSITES

-
After the composites of ten have been obtained for sen

tence completions arithmetical tasks and understanding
,
,
sentences by such experimentation and selection has

as
,
been described for the word knowledge tasks the difficulty

-
of each composite comparison with one or more others

in
was measured several groups individuals As many

in
of
different composites were used each group of individuals

in
as was feasible . It is hard to secure the cooperation of

large groups in taking such long examinations as are neces
sary to put a large number of these composites in compari
son for the same group ; but in one way or another , we have
accumulated a very large body of facts ( shown in Tables
39 to 50 ) .
We use these 10 - composites to make 40- composites each
containing 10C , 10A, 10V , and 10D . They will also be
available for special scales for sentence completion , arith
metical reasoning , vocabulary knowledge , and comprehen
sion of sentences and paragraphs . The arithmetical series ,
for example , is unquestionably a better instrument for mea
suring arithmetical ability of the problem- solving sort than
has hitherto been available .
In Tables 39 to 50 there are sometimes two forms of
entry : " % s " ( percent successes ) means the percent of the
group in question having 50 % or more of the single tasks
"
right ; o distance " means the difference in difficulty be
tween the 10 - composite in question and a 10 - composite of
the same kind which exactly half of the group would suc
ceed with in the sense of having 50 % or more of the single
tasks correct . o distance is in terms of the mean square
deviation of the group in the ability measured by the 10
composite in question . The o distances even for the same
group are not then strictly comparable , since the mean
square deviation of the group in the ability measured by,
say , C Imay not be identical with its mean square devia
J
,
tion in the ability measured by C , or by A , etc. Two I

10 - composites of identical o distances will not , however , be
-
far apart in difficulty .
Minus ( ) means easier than the median difficulty de
fined by the 10 - composite which exactly half of the group
succeed with ; plus (+ ) means harder than it . The form of
distribution is arbitrarily assumed to be " normal " in the
case of all the abilities in all the groups . This is often
7 In some cases the number is less than 10. Each single task is then given
a weight so that a perfect score would count 10.
erroneous and always doubtful , but will do no harm if its

arbitrariness is kept in mind . The o distances are rough
approximate measures convenient for comparison . The
actual fact is always the % s.
-
TABLE 39.
THE DIFFICULTY
V - OF 10 -COMPOSITE C— A ,
A, B ,
C , AND D ; AND D -A
B , C , AND D ; A A , B , C , and D ;
CASE OF 180 ADULT IMBECILES .

, B , C , AND D , IN THE
% distance % distance
σ
8
σ
8
n 100 80 100 80 n 100 80 100 80
CA VA
-
84 822 1.19 1.13 80 81 1.03 1.07
-
-
- -
-
65 56 .51 .20 49 572 .03 .25
+
C 35 272 .45 .68 14 19 +1.09 .93
ABGA
ABGA
с
+ +
+ +
D 1.59 hign D 1.78 1.47
+
5
3
A A 69 80 - .65 1.03 D A 90 86 1.45 1.27
-
-
-
45 49 .15 .03 45 6712 + .15 ..59

+
+ +
15 21 .87 19 272 .93 .68

ABCD
ABCD
+1.05
с
+ +
+
D 1.47 +1.47 ฀ 12 14 1.16 +1.09
5
+
5
- -
TABLE 40
-
.
THE DIFFICULTY — and

10
OF COMPOSITES AND
H
A
G E
G
C
,
F,
I,
E,
F,
,
-
AND AND IN VARIOUS

AND GROUPS
V
H
( F
o G
D
E,
,
E,
F,
DISTANCES ARE OMITTED FROM THIS TABLE

)
.
% %
8
100 50 101 100 50 101

M.A. Class Special M.A. Class Special
3
3
6
Class Class
CE 56 94 93 VE 94
฀
25 94 88 23 96 96
฀
42
FFGH
74 85
FEGT
15 84
7
20 18 64 62
I
AE DE
฀
55 92 100 45 96 98
F 20 90 96 23 92 97
PROGE
฀
PFGE
74 93 58 75
฀฀
5
H 88 73 H 44 57
The Sentence Completion and Arithmetical tasks were

done June 1924 the Vocabulary and Directions Reading
in
;
,
tasks were done September 1924. We treat them

to
in
,
gether , since the differences due to an interval of less than

three months are small .
The following composites had fewer than 10 single
tasks : Directions - Reading , which had 4 , 24 , which had 5 ,
and 3 , which had 6. 2 right , 3 right , and 3 right are used
respectively , in these cases , instead of 5 right .
-
TABLE 41 .
J - J
-
THE DIFFICULTY OF 10 -COMPOSITES C F , G , I , , and K , A F , G, H , I, ,
AND K , V F, G , H , 2 , 3 , AND 4 , AND D — F, G , H , ½ , 1 , 2 , AND

22 , IN VARIOUS GROUPS IN % S.
4B 5A 5B 4B 5A 5B
n= 162 125 186 162 125 186
C F 100.0 V F 98.8
83.3 98.1
฀
J 35.2 47.2 67.6 H 87.0 93.6 100.0

Ι FGTIE
FGE234
22.2 25.6 44.6 50.6 57.6 75.3

K 1.2 9.6 6.4 25.9 35.2 53.8
5.6 5.6 9.7
A F 100.0 D F 93.8
98.8 G 82.1
H 96.3 95.2 98.4 H 75.9 80.8 91.9
J 72.8 77.6 86.6 2 36.4 52.8 68.3

Ι FGRISE
6.8 8.0 25.3 3.1 5.6 10.2

2
K 3.2 15.6 22 1.9 0.8 5.4

0
Of theadult students only 45 were measured with

53
,
the completion tasks and only 28 of these attempted the

Q
,
and composites We have estimated as well we can

as
R
how the individuals question would have succeeded

in
if
they had attempted all

.
THE COMBINATION OF 10 COMPOSITES INTO 40 COMPOSITES

-
-
These 10 composites were combined into 40 composites

-
by putting together an and which from the

A
D
C
a
a
a
,
,
,
,
data hand seemed of nearly equal difficulty com-

as
at
posites Some the 10 composites were constructed espe-

of
-
.
cially fit others this way Into the history of the pro-
in
to
cedure by which the final arrangement of the 40 composites

-
.
TABLE 42
-
FIVE OR MORE OF THE
212
DIFFICULTY OF 10 COMPOSITES MEASURED BY THE PERCENTS OF 147 PUPILS IN GRADE 5½ SUCCEEDING WITH
%
,
5
+
-
,
TEN SINGLE TASKS AND BY DISTANCES OR FROM THE MEDIAN DIFFICULTY FOR GRADE IN UNITS OF THE MEAN
%
5
SQUARE VARIATION OF GRADE IN LEVEL OF WHATEVER ABILITY THE 10 COMPOSITE MEASURES IN EACH CASE
SIMILAR FACTS FOR 205 PUPILS AND 200 PUPILS IN GRADE 52. THE 147 PUPILS ARE THOSE WHO WERE
- .
INCLUDED IN BOTH THE 205 AND THE 200
8
-
-
10 Composite % distance 10 Composite % distance
147 205 147 205 147 200 147 200
1
CI 83.1 80.6 .96 - .86 100 99.5 2.58
J
43.2 44.4 .17 0.14 la 100 99.5 2.58
*
*
2
-
K 20.9 20.0 .81 0.84 93.2 89.5 1.49 1.25
+ + +
+ + +
----
N 1.4 1.0 2.20 2.33 2a 95.5 1.692
3
THE MEASUREMENT
0.0 0.0 55.4 ** 60.5 .14 ** .40

3a 58.0 .20
DG -4 100.0
-
A IK 91.9 93.2 1.40 1.49 -3 100.0
IJ 74.3 68.3 - .65 - .48 -2 100.0
+
ΚΙ 34.5 35.1 .40 .38 92.5 1.44
)))))
( (H (I 0(K (J
J 26.8 .62 -1 95.9 1.74
---
+
K 19.6 18.5 .86 .90 1/2 88.5 1.20
+ + + +
OF INTELLIGENCE
L 15.6 1.01 74.3 .65

1 2 30.4 .51
22 14.9 1.04
+ + +
6.1 1.55
Average of the results for and 2a
..
* **
23
Average of the results for and 3a
TABLE 43 .
THE DIFFICULTY OF VARIOUS 10 -COMPOSITES IN THE CASE OF 44 ADULTS :

RECRUITS IN THE UNITED STATES ARMY .
%8 %8 %8 %8
C F 100.0 AF 100.0 V F 100.0 D F 100.0

G 86.4 G 100.0 G 100.0 G 95.5
H
I
JK I
H 100.0 H 100.0 88.6
2 1 1/2
77.3 95.5 1 100.0 79.6
70.5 70.5 90.9 61.4
I-J
3 2
50.0 81.8 47.7
61.4 22 38.6
45.5 38.6
11.4 18.2
5 4
7 6
2.3 09.1
TABLE 44
.
DIFFICULTY OF 10 COMPOSITES MEASURED BY THE PERMILLES OF SUCCESSES

-
AND BY DISTANCES OR FROM THE MEDIAN DIFFICULTY FOR GRADE

+
82 IN UNITS OF THE MEAN SQUARE DEVIATION OF GRADE 8½

,
IN THE ABILITY MEASURED BY THE COMPOSITE
10 Composites Permille .
σ distance
8
-
CI 972 1.91
-
876 -1.152
J
K 472 .07
+ +
N 72 1.46
Α ΚΙ 564 - .16
L 440 .15
+
V 980 2.05
- - - - -
1
la 976 1.98
920 1.41
2
2a 972 1.91
740 .64
3
3a 864 -1.10
644 .37
-
4
4a 618 .30
440 .15
5
+ +
5a 448 .13
152 +1.03
6
6a 236 .72
+
44 +1.71
7
7a 60 +1.55
Inf 984 2.14

- -
8 7 6 5 4 3
)6 )5 )4 )3 )2 )1
.
( ( ( (
716 .57
500 .00
356 .37
+ + +
324 .46
( (
200 .84
TABLE 45 .
DIFFICULTY OF 10 -COMPOSITES MEASURED BY THE PERCENTS OF TWO GROUPS

( 246 PUPILS IN GRADE 9 AND 264 PUPILS
IN GRADE 9 ) SUCCEEDING WITH
FIVE OF MORE OF THE TEN SINGLE TASKS , AND BY DISTANCES + OR -
FROM THE MEDIAN DIFFICULTY FOR THE GROUP IN UNITS OF THE
MEAN SQUARE DEVIATION OF THE GROUP IN THE ALTITUDE OF
WHATEVER ABILITY THE 10 -COMPOSITE MEASURES .
9 I 911
10 -Composite Permille s σ distance Permille s σ distance
CI - 1.84
IJ
967
- 1.65
J
951
967 - 1.84 936 - 1.52
K 805 .86 689 - .49
L1 350 + .39 295 + .54
M1 191 + .88 178 + .92
N 30 +1.88
O 0
A I
I -J
1000
- 2.05
JJ1 980
886 - 1.21 773
-
- .75
943 - 1.58 784 .79
฀
K1
฀
545
671
439
- .11
- .44
333
500
258
+ .43
.00
+ .15 . + .65
-
฀
L1
z
629
167 + .97
.33
72 + 1.46
V 1 996 - 2.65 996 - 2.65

la
12 967 - 1.84 977 - 2.00
2a
38 866 - 1.11 826 - .94
За
4
8+ 703 - .53 678 -- .45
4a
5 492 + .02 405 + .24
5a
6 150 +1.04 144 +1.06
62a 49 + 1.65
7 20 + 2.05 23 + 2.00
7a 11 + 2.29
1 In the case of the L and M completions with those pupils of Group 9 I

who did not have time to do everything to their satisfaction , an estimated
score was derived on the basis of what they did as far as they went and of
what they did with the completions of K.
TABLE 45 — Continued
.
91 911
10 -Composite Permille distance Permille distance
o
σ
s
s
D 2 992 2.41
- -
1 951 1.65
2 683 .48
22 600 .25
*
362 .35 205 .82
+ + +
+
32 293 .54
334
289 .56 133 +1.11

4
42
5т
126 +1.15 27 +1.93

6 С
23 2.00
+
0
7
TABLE 46
.
DIFFICULTY OF 10 COMPOSITES MEASURED BY THE DISTANCES OR FROM
+
σ
-
-
THE MEDIAN DIFFICULTY OF GIVEN GRADE IN UNITS OF THE MEAN

A
SQUARE DEVIATION OF THE POPULATION OF THAT GRADE IN

THE ABILITY MEASURED BY THE 10 COMPOSITES
-
.
Grade Grade 10 Grade 11 Grade 12
9
1089 723 769 643

=
=
n
n
=
=
n
n
V 2.457 2.652 2.748 2.512

- - - -
- - - -
1
--
- - -
2.290 2.290 2.652 2.512
-
1.360 1.433 1.838 1.960
- -
-
.671 .845 1.522 1.607
.182 .151 .668 .931
+
1234567
+1.131 .553 .068 .151

+ +
+ +
2.120 1.881 +1.546 1.243

+
1041 700 752 637

=
=
=
n
n
V la 2.457 2.457 3.090 2.748

- - - -
-
- - -
- - - -
฀
2a
฀฀
2.576 2.366 3.090
฀ -
1.366 1.468 1.995 2.226
- -
4a .719 .990 1.483 1.774

-
-
.068 .264 .845 1.019
+ + +
-
-
฀
฀฀ .904 + .527 .050 .235
+ +
2.120 +1.866 1.259 1.094

+
1185 1053 742

=
n
=
=
n
D 2.409 2.878 2.652

1
- - -
- - - -
- - - -
1.398 1.695 1.812
-
.690 .999 1.170
.055 .335 .516
5 4
.542 .306 .065

+
+ + +
+1.243 .966 .824

+ +
1237
6
2.170 1.896 1.774

+
71.5 had right out 48.4 had right out of

of
5.
%
%
2
5
*
16
TABLE 47 .
THE DIFFICULTY OF VARIOUS 10 -COMPOSITES MEASURED BY THE PERCENT SUC

CEEDING AND BY THE DISTANCE FROM THE MEDIAN IN TERMS OF THE MEAN
SQUARE DEVIATION OF THE GROUP . 422 NORMAL SCHOOL SENIORS . THE
FORM OF DISTRIBUTION IS ASSUMED TO BE " NORMAL . '' THE DIVISION
INTO TWO GROUPS OF 150 AND 185 IS APPROXIMATELY AT RANDOM .
THE GROUP OF 87 REPRESENTED A SOMEWHAT SUPERIOR SELECTION
AND TOOK CERTAIN ADDITIONAL TESTS .
%8 σ distances
n= 150 185 87 422 150 185 87 422
C M2 98.7 96.2 98.9 97.6 - 2.23 - 1.77 - 2.29 - 1.98

- 1.41 - 1.58 - 1.15
- -
MNS 92.0 80.5 94.3 87.4 .86
N
-
-
-
65.3 45.9 75.9 59.0 .39 + .10 .70 .23
Na 54.0 53.0 69.0 56.6 .10 -- .08 - .50 .17
No 17.3 19.5 20.7 19.0 + .94 + .86 + .82 + .88
O 18.7 20.5 35.6 23.0 + .89 + .82 + .37 + .74
OP 04.7 08.1 04.6 06.2 +1.67 + 1.40 +1.68 + 1.54
P 00.7 04.3 03.4 02.8 + 2.46 +1.72 +1.83 +1.91
Pa 13.0 04.9 03.4 03.3 + 2.23 + 1.65 +1.83 +1.84
฀ 00.0 00.5 00.0 00.2 ∞ + 2.58 ∞ + 2.88
Qa 00.0 00.0 00.0 00.0 ∞ ∞ ∞ ∞
A - 1.30
-- -
X1-8 90.7 73.5 83.9 81.8 - .63 .99 .91
X9-18 77.3 51.9 75.9 65.9 .75 - .05 ..70 .41
X. 94.7 85.4 96.6 91.0 - 1.62 - 1.05 - 1.83 - 1.34

..99 - - .78 -
-
Xb 84.0 62.7 79.3 73.7 .32 .63
X, 72.7 44.9 69.0 59.7 .60 + .13 .50 .25
Y 46.7 22.2 36.8 33.9 + .08 + .77 + .34 + .42
Ya 45.3 26.5 42.5 36.5 + .12 + .63 + .19 + .35
Yz 11.3 06.5 24.1 11.8 +1.21 +1.51 + .70 +1.19
Z₁ 00.7 03.2 06.9 03.1 + 2.46 +1.85 +1.48 +1.87
-
Z2 05.3 02.2 14.9 05.9 +1.62 + 2.01 +1.04 +1.56
Z. Not taken off so few got it .
V - 2.46 - 1.72 - 2.29 - 1.98
- .86 - 1.48 - 1.13
4a 99.3 95.7 98.9 97.6
-- .60 -- .06 - .63 - .36

5a 91.3 80.5 93.1 87.0 - 1.36
6a 72.7 52.4 73.6 64.0
612 44.7 35.7 50.6 41.9 + .13 + .37 .02 + .20
7
-
18.0 16.2 20.7 17.8 + .92 + .99 + .82 + .92
D 423 96.7 81.6 90.8 88.9 - 1.84 .90 - .87 - 1.22

5 66.7 .43
512 38.7 22.7 24.1 28.7 + .29 + .75 + .70 + .56
62 54.7 42.2 34.5 45.0 .12 + .20 + .40 + .13
6124 07.3 04.3 04.9 05.7 +1.45 +1.72 +1.65 +1.58
74 10.7 04.3 08.0 07.3 + 1.24 +1.72 +1.41 +1.45
2 Only six of the ten single tasks were used .

3 Only nine of the ten single tasks were used . A person thus had to
succeed with 552 % to score a success ; hence the % is too low.
4 This composite has only six single tasks .
TABLE 48 .
THE DIFFICULTY OF 10 -COMPOSITES MEASURED BY THE PERCENTS OF 240
COLLEGE GRADUATES AND 100 STUDENTS OF EDUCATION ( COLLEGE OR

NORMAL SCHOOL GRADUATES ) SUCCEEDING , AND BY DISTANCES + OR -
FROM THE MEDIAN DIFFICULTY FOR THE COLLEGE GRADUATES IN QUES
TION , IN UNITS OF THE MEAN SQUARE DEVIATION OF THE COLLEGE
GRADUATES IN QUESTION IN ALTITUDE OF WHATEVER ABILITY
THE 10-COMPOSITE MEASURES IN EACH CASE .
% distance % distance
σ
o
n= 240 100 240 100 240 100 240 100
-
--1.47 - .912
C N 92.9 95 1.64½ Vocab 99.6 100 -2.65
5
.
O 66
-
80.8
-
82 .87 5a 100.0 100
P 61.3 56 .29 .15 66 99.6 100 -2.65
6
66 -1.61
-
44
-
฀. 41.7 .21 .15 ба 94.6 99 2.33
-
+ +
฀
77 -
18.3 .90 68.4 75 .48 .6712
""
66 -
56 -
7a 69.6 .51 .74
66 53.3 .08 .15
8
66 29.6 31.54 .50

+
9
A N 87.5 91 -1.15 -1.34 10 14.2 10 +1.07 +1.28

""
-
82.1 79 .92 .81 66 11 1.7 +2.02 +1.75
O
P 67.5 80 .45 .84

OFO
39.2 .27
+
♦
Inf 100
7 6 5 4 3
) )3 ) )1
( ( ( (2 (
.
100
""
42 96.3 1.79 66 100

D
5 66
-
85.9 1.08 97 1.88
-
-
)5 4
52 74 .64 66 81 .88
.
73.8 .64 60 .25

-
""
6
8
)6
(
62 33 .44
+
38.3 .30
7
As given these groups contained only tasks contained only

D
to
A
Q
5
5
contained only 6½ contained only and contained only

D
D
D
6
6
8,
6,
6,
.
TABLE 49 .
DIFFICULTY OF 10 -COMPOSITES BY THE PERCENTS OF 53 ADULT STU
-
MEASURED
DENTS SUCCEEDING , AND BY DISTANCES + OR FROM THE MEDIAN DIFFICULTY
FOR THE GROUP , IN UNITS OF THE MEAN SQUARE DEVIATION OF THE
GROUP IN THE ABILITY MEASURED BY THE COMPOSITE .
Permilles distance
σ
M 1000
C
962 1.77
-
887 -1.19
774 .75
MNOPQR
-
฀ Р
453 .12
+ +
฀ 245 .69
A 1000
J
K 1000
962 1.77
-- - - -
M 943 1.58
N
TALWROA
887 1.21
717 .572
P 736 .63
TABLE 50
.
DIFFICULTY OF 10 COMPOSITES MEASURED BY THE PERCENTS OF 63 UNIVERSITY

-
STUDENTS SUCCEEDING
.
CN 98.4 V 62a 98.4

95.2 82.6
8 7
74.6 57.1
NOPOR
33.3 20.6
9
฀
A M D 93.7
5
81.0 77.8
6
69.9 62 27.0
MNOPQ
60.3 39.7
7
14.3
฀
was arrived at , we need not enter . The combinations made

are as shown below :
40 -Composite 10 Composite
-
A V D
Q
Α
B
с
с
с
฀ ฀
AACDEFO
G
ABCDEFGHIT
ABCDEFGHS
…………
..
J
I
1/2
ABCDEFGHY223456
J
4 3
K K ABCDEFGHIKLMNOP 212
L L
M
62a ฀
฀
฀UDHAENOPER
9 56678
MNOPQR
฀ Р
612
&
7
฀
10
฀
The chief facts concerning the likeness difficulty

of
in
the four composites making each composite are sum

40
10
-
marized Table which gives the percents succeeding

51
of in
for each the four for each group which was tested with
all four and some other data
,
Not all of the facts of Tables 39-50 were available when

the process putting the composites together make
If 10
to
of
40 composites was begun they had been better com

,
-
binations could have been made In some cases

.
10 com
a
-
.
posite which itself inferior used because its difficulty

is
is
as composite fits certain place Such happened with

a
Co In some cases composites which seemed nearly

L.
10
-
enough equal difficulty belong well together were

to
in
found from wider experience not be very closely alike

to
but the discovery was made too late allow amendments

to
to be made
.
TABLE 51 .
SUMMARY OF THE FACTS CONCERNING THE DIFFICULTY OF THE FOUR

10 -COMPOSITES CONSTITUTING EACH 40 -COMPOSITE .
с A V D Group n
A 100
88888
84 80 90 im
3
6889
" .
822 80 81 86 80
"
B 65 45 49 45 66 100
56 49 572 672 80
""
C 35 15 14 19 100
::
272 21 19 272 80
D 12 100
5
80
15
::
30 14
5
E 56 55 45 im
6
.
28895
94 92 94 96
93 100
f.
sp
.
F 25 20 23 23 im
6
.
94 90 96 92
f.
88 96 96 97 sp
.
100 100 99 94
4
100 100 100 100 ad

.
G 15 im
5
6
6
74 84 58
f.
93 85 75 sp
0878
7248
*฀
83 99 98 82
4
H im
4
฀
.
44
฀
f.
73 62 57 sp
.
18885
96 76 4b
95 81 5a
I 98 ฀
100 92 5b
฀
20 14 ฀฀฀
24 12
f.
20 37 sp
฀
.
35 73 51 36 4b
47 78
฀
฀
58 53 5a
216488
87 75 68 5b
2288888
฀ 52
83 92 93 89
97 100 97 99 91
TABLE 45 - Continued .
91 911
10 -Composite Permille s σ distance Permille s σ distance
D 2 992 - 2.41
951 1.65
-
683 .48
22 600 - .25
*
362 .35 205 .82
+ + +
+
32 293 .54
12233445
289 .56 133 1.11
+
42
126 +1.15 27 +1.93
23 2.00
+
7 6
0
TABLE 46
.
DIFFICULTY OF 10 COMPOSITES MEASURED BY THE DISTANCES OR FROM
+
σ
-
-
THE MEDIAN DIFFICULTY OF GIVEN GRADE IN UNITS OF THE MEAN

A
SQUARE DEVIATION OF THE POPULATION OF THAT GRADE IN

THE ABILITY MEASURED BY THE 10 COMPOSITES
-
.
Grade Grade 10 Grade 11 Grade 12
9
1089 723 769 643

=
=
n
=
n
n
=
n
V 2.457 2.652 2.748 2.512

-
- - - - -
- - -
1
- - -
2.290 2.290 2.652 2.512

- - -
-
1.360 1.433 1.838 1.960
-
- -
.671 .845 1.522 1.607
+ .182 .151 .668 .931
234567
1.131 .553 .068 .151

+ +
+ +
+
+ +
2.120 1.881 1.546 +1.243
1041 700 752 637

=
=
n
=
n
V la 2.457 2.457 3.090 2.748

- - -
-
- - -
- - - -
2a 2.576 2.366 3.090
-
3a 1.366 1.468 1.995 2.226
- - -
฀฀฀฀
4a .719 .990 1.483 1.774

-
-
.068 .264 .845 1.019

+ + +
-
฀฀
฀
.904 .527 .050 1 .235
+
+ +
7a 2.120 +1.866 1.259 +1.094

1185 1053 742
=
=
=
n
n
n
D 2.409 2.878 2.652

- - -
- - - -
- - -
7 6 5 4 3 2 1
1.398 1.695 1.812

.690 .999 1.170
- .055 .335 .516
-
.542 .306 .065

+ + +
+
+1.243 .966 .824

+ +
2.170 1.896 1.774

+
71.5 had right out of 48.4 had right out of

5.
%
%
2
5
*
16
In general , however , these 40 - composites are satisfac

tory . All the single tasks within any one of them are
nearly enough alike in difficulty to prevent an individual
for whom the composite as a whole is a suitable test , from
being either bored or bewildered by any item of it . They
rise in difficulty by steps which are small enough so that ( as
will be demonstrated later ) a very finely graduated and rea
sonably precise measurement of altitude of Intellect CAVD
is possible . In particular , composites J
to Q furnish a very
convenient series of defined and graded tasks . Each of
them measures very nearly the same ability as the one
1
below it or above it , and correlates with the score in the

entire series of CAVD tasks nearly or quite as closely as
its own self - correlation permits .
The methods used for obtaining CAVD 40 - composite
tasks , each of which measures all of Intellect CAVD and
nothing but it , save for a small chance error which can
safely be corrected for by o σ , = σt rt , or by o₁ = σt√rtit2 , are I
applicable to any other form of intellect . We could have
taken , in place of CAVD , Intellect OGANS ( O standing for 1
the ability to give or select opposites , G standing for the

ability to perceive the common element in n facts and select
other facts containing it , An standing for the ability to per
ceive and apply relations as in an Analogies test , and S
standing for the ability to put together to make a sensible
total , a series of elements as in a dissected sentences test .
Or we could have taken in place of CAVD , a non - verbal
Intellect CRPF made up of Picture Completions , Geomet 1
rical Relations , Picture Sequences , and Forms to be cut up

so as to produce specified parts . Or we could have taken an I
Intellect NIL made up of novel problems of the sort used
in Burt's Reasoning Test , " informational tasks , and tasks
in learning . Or we could have taken any combination of
all these .
8 Shown in Chapter XVI .
Such as the substitution test of the Army Beta , The National Intelli
gence Test , The Pintner - Non Verbal , and other examinations .
Inplace of attaching approximately equal weight to

each sort of tasks , we may use any specified weights . We
may also weight the single tasks of the same sort ( C or A
or V or D ) differently one from another , provided the sta
tistical consequences are allowed for consistently .
We may use the same situation and task as an element
in composites at different levels according to the quality of
the performance . Thus to give " mean " as the opposite of
" grand " may count as a right in levels a , ak , and
a + 2k , whereas to give " little " may count as a right for a

and a + k , but as a wrong for level a + 2k .
The essentials are that whatever intellect we claim to
measure be defined by an actual series of tasks , and that
each composite task measure the ability .
The difficulty of these CAVD 40 - composites A , B , C ,
•
Qis intellectual and it can be measured for each of
them by means of the facts stated in Chapters IIand IV ,
together with certain others . We shall carry through these
measurements . But in order to do so , certain facts will be
needed ; and it will be most instructive to present these
facts as parts of a general discussion of the meaning of the
scores assigned in standard intelligence examinations . So
the next two chapters will be in some respects a digression .
CHAPTER VII
THE TRANSFORMATION OF THE SCORES OF STANDARD EXAMINA

TIONS INTO TERMS OF SCALES WITH EQUAL UNITS
THE METHOD OF TRANSFORMATION : BY THE

ILLUSTRATED
THORNDIKE EXAMINATION AND ARMY ALPHA
We have shown that the form of distribution of altitude
of intellect measured in truly equal units is approximately
Form A for any grade population from Grade 6 to college
freshmen . We can then transmute the scores in any ex
amination which is closely symptomatic of intellect , into
terms of truly equal units , if we have the distributions of
scores in grade populations . Consider , for example , the
facts of Table 52 , which shows the essential steps in such a
transformation and two further steps by which the mea
sures in truly equal units are adapted to the general scale
of the original scoring .
The first column is a series of points on the scale of the
original scores 54 , 57 , 60 , 63 , and so on . The second column
gives the permille of individuals in the grade in question
whose scores are below the point at which they are entered .
Thus up to 54 , there were 23 ; up to 57 , there were 30 ; up to
60 , there were 43 , and so on . The third column gives the
distance below or above the central tendency for the grade
in terms of the mean square deviation of the grade . The
sign is — until the entry in column 2 passes 500 , when it
changes to +. The fourth column gives the successive dif
ferences , the entry on a line with 57 giving the true differ
ence between the score of 54 and the score of 57 ; the entry
on a line with 60 giving the true difference between the
score of 57 and the score of 60 ; and so on . These are the
essentials .
Ifwe desire to put these differences in such shape as to
compare each readily with the difference which it replaces ,
224
THE TRANSFORMATION OF STANDARD SCORES 225
TABLE 52 .
THORNDIKE Ex . H.S. GRADS . PART I , FORMS D AND N. GRADE 12. n = 1527 .
1 2 3* 4 5 6 7
X
Points
Per - or + Δ
mille Deviation Value of Corrected
A /.05400 **
on Column 6
up to from the
Original Interval Points on - Column 1
Scale
Stated Median in Equal Scale
Point in Equal Units
Units
54 23 - 1.995393 54.77 .77

57 30 - 1.880794 .114599 2.12 56.89 - .11
60 43 - 1.716886 .163908 3.04 59.93 - .07
63 59 - 1.563224 .153662 2.84 62.77 .23
66 75 - 1.439531 .123693 2.29 65.06 .84
69 95 - 1.310579 .128952 2.39 67.45 - 1.55
72 136 - 1.098468 .212111 3.93 71.38 - .62
75 166 .970093 .128365 2.38 73.76 - 1.24
78 214 - .792619 .177474 3.29 77.05 - .95
81 263 .634124 .158495 2.93 79.98 - 1.02
84 330 .439913 .194211 3.60 83.58 .42
87 391 - .276714 .163199 3.02 86.60 .40
90 463 - .092878 .183836 3.40 90 . .00
93 536 + .090361 .183239 3.39 93.39 .39
96 600 + .253347 .162986 3.02 96.41 .41
99 656 + .401571 .148224 2.75 99.16 .16
102 720 + .582841 .181270 3.36 102.52 .52
105 770 + .738847 .156006 2.89 105.41 .41
108 807 .866894 .128047 2.37 107.78 - .22
111 852 + 1.045050 .178156 3.30 111.08 .08
114 885 +1.200359 .155309 2.88 113.96 .04
117 916 + 1.378659 .188300 3.49 117.45 .45
120 934 + 1.506262 .127603 2.36 119.81 - .19
123 954 + 1.684941 .178679 3.31 123.12 .12
126 966 + 1.825007 .140066 2.60 125.72 - .28
129 980 + 2.053749 .228742 4.24 129.96 .96
always 3 in this case , we may divide the entries of column 4

by whatever o σ value is on the average equal to 1 in the orig
inal scores . The result of such a division appears as col
umn 5. If we desire to use the true differences to form a
* Computed from the Kelley -Wood table [ Kelley , '23 , p . 373 ff .] .
** .054 is the average o value corresponding to a difference of 1.00 in the
original scale .
true scale of the same general location and extent as the

original scores , we can do so by taking any point of the
original scale arbitrarily and replacing all the other points
by the values derived from the true differences . Column 6
shows the result of this operation , 90 being taken as the
point of coincidence of the original and true values .
Column 7 is simply a convenient way of showing how
and how far the measures in truly equal units diverge from
TABLE 53 .
THORNDIKE INT . EX . H.S. GRADS . I

PART , FORMS D AND N ; SCORES FROM 54
TO 129 CORRECTED TO BE IN TRULY EQual Units .
Original Corrected Original Corrected Original Corrected
54 54.8 80 79.0 106 106.2

55 55.5 81 80.0 107 107.0
56 56.2 82 81.2 108 107.8
57 56.9 83 82.4 109 108.9
58 57.9 84 83.6 110 110.0
**
59 58.9 85 84.6 111 111.1
60 59.9 86 85.6
61 60.9 87 86.6 112 112.1
62 61.8 87.8 113 113.1
88888
63 62.8 89 88.9 114 114.0

64 63.6 90 90.0 115 115.2
64.4 91 91.2 116 116.4

66 65.1 92 92.3 117 117.5
67 65.9 93 93.4 118 118.3
68588
66.7 94 94.4 119 119.1

69 67.5 95 95.4 120 119.8
70 68.8 96 96.4 121 120.9

71 70.1 97 97.4 122 122.0
72 71.4 98 98.3 123 123.1
73 72.2 99 99.2 124 124.0
74 73.0 100 100.3 125 124.9
75 73.8 101 101.4 126 125.7

76 74.9 102 102.5 127 127.1
77 76.0 103 103.5 128 128.5
78 77.0 104 104.5 129 130.0
79 78.0 105 105.4
those by the original scores . In the illustration the diver

gences are small and irregular , so that for many purposes
the original scores can be taken as approximately the scores
in truly equal units . Finally , we may by interpolation ar
range a table giving the true - unit equivalents to any desired
detail . When interpolating , one may also smooth the re
sults if he has good reason to believe that the irregularities
TABLE 54 .
ARMY ALPHA GRADE 1721
=
n
9
:
.
:
X
Points on Permille
or - +
Δ
Deviations Value of
Original up to
A
from the Interval in .0399
/
Scale Stated Point
Median in Equal Units
Equal Units
55 30 1.880794
-
60 47 1.674665 .206129 5.17

1866
- - - -
65 76 1.432503 .242162 6.07

70 117 1.190118 .242385 6.07
75 174 .938476 .251642 6.31
80 238 .712751 .225725 5.66
85 307 - .504372 .208379 5.23
-
90 398 .258527 .245845 6.16
95 473 .067731 .190796 4.78
100 564 .161119 .228850 5.73
+ + + + + + +
105 630 .331853 .170734 4.28

110 700 .524401 .192558 4.83
115 755 .690309 .165908 4.16
120 809 .874217 .183908 4.61
125 856 1.062519 .188302 4.72
130 902 1.293032 .230513 5.78
135 936 +1.522036 .229004 5.74
140 959 1.739198 .217162 5.44
+ + +
145 971 1.895698 .156500 3.92

150 980 2.053749 .158051 3.96
are due the small size the population rather than

to
to
of
inequalities alleged units In the case of the illustration

in in
,
.
we present Table the results interpolation with

of
53
(
out smoothing
.
We have made the calculations necessary obtain values

to
truly equal units for scores Army Alpha National

in
in
A
,
,
In generalhowever , these 40 - composites are satisfac-

,
tory . All the single tasks within any one of them are
nearly enough alike in difficulty to prevent an individual
for whom the composite as a whole is a suitable test , from
being either bored or bewildered by any item of it . They
rise in difficulty by steps which are small enough so that ( as
will be demonstrated later ) a very finely graduated and rea-
sonably precise measurement of altitude of Intellect CAVD
is possible . In particular ,
composites J
to Q furnish a very
convenient series of defined and graded tasks . Each of
them measures very nearly the same ability as the one
below it or above it , and correlates with the score in the
entire series of CAVD tasks nearly or quite as closely as
its own self - correlation permits .
The methods used for obtaining CAVD 40 - composite
tasks , each of which measures all of Intellect CAVD and
nothing but it , save for a small chance error which can
safely be corrected for by o , =
σt rt , or by σι
σε o₁ = σt Vrt1t2 , are
applicable to any other form of intellect . We could have
taken , in place of CAVD , Intellect OGANS ( O standing for
the ability to give or select opposites , G standing for the
ability to perceive the common element in n facts and select
other facts containing it , An standing for the ability to per-
ceive and apply relations as in an Analogies test , and S
standing for the ability to put together to make a sensible
total , a series of elements as in a dissected sentences test .
Or we could have taken in place of CAVD , a non -verbal
Intellect CRPF made up of Picture Completions , Geomet-
rical Relations , Picture Sequences , and Forms to be cut up
so as to produce specified parts . Or we could have taken an
Intellect NIL made up of novel problems of the sort used
in Burt's Reasoning Test , " informational tasks , and tasks
in learning . Or we could have taken any combination of
all these .
8 Shown in Chapter XVI .
Such as the substitution test of the Army Beta , The National Intelli-
gence Test , The Pintner - Non Verbal , and other examinations .
In place of attaching approximately equal weight to

each sort of tasks , we may use any specified weights . We
may also weight the single tasks of the same sort ( C or A
or V or D ) differently one from another , provided the sta-
tistical consequences are allowed for consistently .
We may use the same situation and task as an element
in composites at different levels according to the quality of
the performance . Thus to give " mean " as the opposite of
" grand " may count as a right in levels a, a + k, and
a + 2k , whereas to give " little " may count as a right for a
and a + k , but as a wrong for level a + 2k .
The essentials are that whatever intellect we claim to
measure be defined by an actual series of tasks , and that
each composite task measure the ability .
The difficulty of these CAVD 40 - composites A, B , C ,
•
Qis intellectual and it can be measured for each of
them by means of the facts stated in Chapters II and IV,
together with certain others . We shall carry through these
measurements . But in order to do so , certain facts will be
needed ; and it will be most instructive to present these
facts as parts of a general discussion of the meaning of the
scores assigned in standard intelligence examinations . So
the next two chapters will be in some respects a digression .
CHAPTER VII
THE TRANSFORMATION OF THE SCORES OF STANDARD EXAMINA-
TIONS INTO TERMS OF SCALES WITH EQUAL UNITS
THE METHOD OF TRANSFORMATION : BY THE

ILLUSTRATED
THORNDIKE EXAMINATION AND ARMY ALPHA
We have shown that the form of distribution of altitude
of intellect measured in truly equal units is approximately
Form A for any grade population from Grade 6 to college
freshmen . We can then transmute the scores in any ex-
amination which is closely symptomatic of intellect , into
terms of truly equal units , if we have the distributions of
scores in grade populations . Consider , for example , the
facts of Table 52 , which shows the essential steps in such a
transformation and two further steps by which the mea-
sures in truly equal units are adapted to the general scale
of the original scoring .
The first column is a series of points on the scale of the
original scores 54 , 57 , 60 , 63 , and so on . The second column
gives the permille of individuals in the grade in question
whose scores are below the point at which they are entered .
Thus up to 54 , there were 23 ; up to 57 , there were 30 ; up to
60 , there were 43 , and so on . The third column gives the
distance below or above the central tendency for the grade
in terms of the mean square deviation of the grade . The
sign is until the entry in column 2 passes 500 , when it
changes to +. The fourth column gives the successive dif-
ferences , the entry on a line with 57 giving the true differ-
ence between the score of 54 and the score of 57 ; the entry
on a line with 60 giving the true difference between the
score of 57 and the score of 60 ; and so on . These are the
essentials .
If we desireto put these differences in such shape as to

compare each readily with the difference which it replaces ,
224
-
58 , 59 , and These differences are then divided by
60.
( X85 X25 ) /60 . The results of the divisions appear as the
three last columns of Tables 58 , 59 , and 60 , and again in
Table 61 , where they are averaged for each 5 - point interval
of the original scores . We thus have measures in truly
equal units for all the range from 20 to 85 , though not so
reliable ones as we should like to have . As an addition to
these determinations , we have used in a similar manner the
records from 242 12th - grade boys , 393 12th - grade girls , and
400 college freshmen . The results appear in Table 62 .
We have combined all these results from these three
determinations , giving weights of 3 to the determination
from 5,653 individuals ( 1,721 +1,387 +2,545 ) , 1 to the de
termination from 865 individuals ( 263 + 281 +321 ) , and 1
to the determination from 1,035 individuals ( 242 +393 +
400 ) . The results appear in Table 63 .
In Table 64 we give the results of interpolation with
rather liberal smoothing . In the smoothing , we have been
guided by three facts . First , the ups and downs from about
55 to about 145 are quite irregular and , in view of the prob
able errors , may well be chances of the sampling . Second ,
the data of Table 61 for the interval 15 to 20 and the gen
eral drift of the Army results support the hypothesis that
original scores from 10 to 20 will have a true - unit value
even higher than our 13.67 for 20 to 30 ; so that the increase
from 10.81 for 40 to 50 , to 11.67 for 30 to 40 , and 13.67 for
20 to 30 may be considered real . Third , the facts of Table
62 and the facts of Table 65 for Alpha scores of a college
graduate group show that the true - unit values per interval
of 5 score - points continue , above 175 , the rise shown in our
table from 150 to 170. We take 100 as the point of coinci
1 Before combining them we have multiplied all the results from the 865
determination ( Grades 5 , 6 , and 7 ) by 1.08 , which puts their total value from
65 to 95 on a parity with the total value from 65 to 95 of the 5623 series ; we
have also multiplied all the results from the 1035 determination by 1.03 , which
puts their total value from 110 to 155 on a parity with the total value from
110 to 155 of the 5623 series . These multiplications serve to keep our systems
of values consistent .
17
true scale of the same general location and extent as the

original scores , we can do so by taking any point of the
original scale arbitrarily and replacing all the other points
by the values derived from the true differences . Column 6
shows the result of this operation , 90 being taken as the
point of coincidence of the original and true values .
Column 7 is simply a convenient way of showing how
and how far the measures in truly equal units diverge from
TABLE 53 .
THORNDIKE INT . EX . H.S. GRADS . PART I,

FORMS D AND N ; SCORES FROM 54
TO 129 CORRECTED TO BE IN TRULY EQUAL UNITS .
Original Corrected Original Corrected Original Corrected
54 54.8 80 79.0 106 106.2

55 55.5 81 80.0 107 107.0
56 56.2 82 81.2 108 107.8
57 56.9 83 82.4 109 108.9
58 57.9 84 83.6 110 110.0
59 58.9 85 84.6 111 111.1
60 59.9 86 85.6
61 60.9 87 86.6 112 112.1
62 61.8 87.8 113 113.1
88888
63 62.8 89 88.9 114 114.0

64 63.6 90 90.0 115 115.2
65 64.4 91 91.2 116 116.4

66 65.1 92 92.3 117 117.5
67 65.9 93 93.4 118 118.3
68 66.7 94
฀฀฀฀ 94.4 119 119.1
8888
69 67.5 95 95.4 120 119.8
70 68.8 96 96.4 121 120.9

71 70.1 97 97.4 122 122.0
2b2
72 71.4 98 98.3 123 123.1

73 72.2 99 99.2 124 124.0
FINEN
74 73.0 100 100.3 125 124.9
75 73.8 101 101.4 126 125.7

76 74.9 102 102.5 127 127.1
77 76.0 103 103.5 128 128.5
78 77.0 104 104.5 129 130.0
79 78.0 105 105.4
those by the original scores . In the illustration the diver-

gences are small and irregular , so that for many purposes
the original scores can be taken as approximately the scores
in truly equal units . Finally , we may by interpolation ar-
range a table giving the true - unit equivalents to any desired
detail . When interpolating , one may also smooth the re-
sults if he has good reason to believe that the irregularities
TABLE 54 .
ARMY ALPHA : GRADE 9: n = 1721 .
-
X
or + Δ
A/.0399
Points on Permille
Deviations Value of
Original up to from the Interval in
Scale Stated Point
Median in Equal Units
Equal Units
55 30 - 1.880794
60 47 1.674665 .206129 5.17
- - - -
65 76 1.432503 .242162 6.07

70 117 1.190118 .242385 6.07
฀฀฀
75 174 .938476 .251642 6.31

80 238 .712751 .225725 5.66
85 307 - .504372 .208379 5.23
90 398 .258527 .245845 6.16
95 473 .067731 .190796 4.78
-
100 564 .161119 .228850 5.73

+ + + + +
105 630 .331853 .170734 4.28

110 700 .524401 .192558 4.83
115 755 .690309 .165908 4.16
120 809 .874217 .183908 4.61
125 856 +1.062519 .188302 4.72
130 902 1.293032 .230513 5.78
+ +
135 936 1.522036 .229004 5.74

140 959 +1.739198 .217162 5.44
145 971 1.895698 .156500 3.92
+ +
150 980 2.053749 .158051 3.96
are due the small size the population rather than

of
to
to
inequalities alleged units In the case of the illustration

in in
,
.
we present Table the results interpolation with-

53
of
(
out smoothing
.
We have made the calculations necessary obtain values

to
truly equal units for scores Army Alpha National

in
in
A
,
,
Otis Advanced Examination , Haggerty Intelligence Exami-

nation Delta 2 , Terman Group Test of Mental Ability ,
Myers Mental Measure , Pintner Non - Language Mental
Test , I. E. R. Tests of Selective and Relational Thinking ,
Generalization and Organization , Army Examination a , the
Trabue Mentimeter, and the Brown University Psycholog-
ical Examination . The results appear in Tables 54 to 94 .
TABLE 55 .
ARMY ALPHA : GRADE 12 : n= 1387 .
X
Points on Permille -or +
Δ
A/.0411
Value of
Original up to Deviations
Interval in
Scale Stated Point from the
Equal Units
Median
฀
75 18 - 2.096927
฀
80 40 - 1.750686 .346241 8.42
85 52 - 1.625763 .124923 3.04
90 74 - 1.446632 .179131 4.36
95 102 - 1.270237 .175395 4.27
100 142 - 1.071377 .198860 4.84
105 195 .859617 .211760 5.15
110 267 ..621912 .237705 5.78
115 343 .404289 .217623 5.29
120 422 - .196780 .207509 5.05
125 497 - .007520 .189260 4.60
130 581 + .204452 .211972 5.16
135 647 + .377234 .172782 4.20
140 711 + .556308 .179074 4.36
145 779 + .768820 .212512 5.17
150 838 + .986271 .217451 5.29
155 883 + 1.190118 .203847 4.96
160 924 + 1.432503 .242485 5.90
165 951 + 1.654628 .222125 5.41
170 968 + 1.852180 .197552 4.81
175 986 + 2.197286 .345106 8.40
In of these cases we have material from several

each
grade populations, which it is necessary to combine . The
methods of combination may be illustrated by Army Alpha .
For the upper range of scores 75 to 175 or higher , we have
three large grade populations : 1,721 in Grade 9 , 1,387 in
Grade 12 , and 2,545 college freshmen ( Ohio ) . Each of
TABLE 62.
ARMY ALPHA : GRADES 12 AND 13 ; SUPPLEMENTAL VALUES OF SUCCESSIVE

5 -POINT INTERVALS OF THE ORIGINAL SCORES , IN ' EQUAL UNITS .
Values in Equal Units

Interval Grade 12 Grade 12 Yale
Original Boys Girls Freshman
Score n = 242 n = 393 n = 400 Average Οι - ο
75 to 79 8.05 8.05
80 to 84 2.20 3.00 2.60 .28
85 to 89 2.13 1.08 1.61 .37
90 to 94 5.36 3.80 4.58 .55
95 to 99 1.80 5.16 3.48 1.19
100 to 104 5.56 5.84 5.70 .10
105 to 109 4.89 6.05 5.47 .41
110 to 114 4.76 5.83 7.13 5.91 .56
115 to 119 5.01 5.70 3.52 4.74 .53
120 to 124 3.29 4.56 4.06 3.97 .30
125 to 129 6.14 5.06 5.47 5.56 .26
130 to 134 2.24 5.57 4.76 4.19 .82
135 to 139 4.26 4.11 2.94 3.77 .34
140 to 144 7.19 4.71 5.45 5.78 .60
145 to 149 5.35 4.25 4.66 4.75 .26
150 to 154 5.23 3.09 4.99 4.44 .55
155 to 159 6.64 5.63 6.13 5.68 .35
160 to 164 3.37 6.52 5.00 4.96 .74
165 to 169 6.58 4.95 5.86 5.80 .39
170 to 174 10.29 7.96 9.13 .89
175 to 179 7.64 7.64
180 to 184 7.95 7.95
185 to 189 6.67 6.67
190 to 194 6.01 6.01
The values for the two intervals next below the median are :
170 to 174 .31360

175 to 179 .25850
These last and the values for 180 to 189 will not be very
different if any reasonable assumption is made about the
form of distribution .
Expressing the values for 190 and above in terms of the
170 to 190 difference , which is 1.3276260 , we have the fol
lowing results :
190 to 194 = .272
195 to 199 = .421
200 to 204 = .513
TABLE 63 .
FINAL ESTIMATE OF RELATIVE VALUES OF ARMY ALPHA SCORES IN EQUAL

UNITS .
Weighted Weighted
Average : Average :
Interval Intervals of 5 Intervals of 10
20 to 24 7.67
25 to 29 5.70 13.67
30 to 34 6.32
35 to 39 5.35 11.67
40 to 44 5.24
45 to 49 5.57 10.81
50 to 54 5.92
55 to 59 4.54 10.46
60 to 64 5.18
65 to 69 4.96 10.14
70 to 74 4.78
75 to 79 6.28 11.06
80 to 84 4.26
85 to 89 4.23 8.49
90 to 94 4.81
95 to 99 4.76 9.57
100 to 104 5.06
105 to 109 5.31 10.37
110 to 114 5.03
115 to 119 5.09 10.12
120 to 124 4.56
125 to 129 5.15 9.71
130 to 134 4.88
135 to 139 4.55 9.43
140 to 144 5.33
145 to 149 4.94 10.27
150 to 154 4.89
155 to 159 6.17 11.06
160 to 164 5.37
165 to 169 5.86 11.23
170 to 174 7.91
175 to 179 7.87 15.78
180 to 184 8.19
185 to 189 6.87 15.06
190 to 194 6.19
( X85 -
58 , 59 , and 60. These are then divided by
differences
X25 ) / 60 . The results of the divisions appear as the
three last columns of Tables 58 , 59 , and 60 , and again in
Table 61 , where they are averaged for each 5 - point interval
of the original scores . We thus have measures in truly
equal units for all the range from 20 to 85 , though not so
reliable ones as we should like to have . As an addition to
these determinations , we have used in a similar manner the
records from 242 12th- grade boys , 393 12th - grade girls , and
400 college freshmen . The results appear in Table 62 .
We have combined all these results from these three
determinations , giving weights of 3 to the determination
from 5,653 individuals ( 1,721 +1,387 +2,545 ) , 1 to the de-
termination from 865 individuals ( 263 + 281 +321 ) , and 1
to the determination from 1,035 individuals ( 242 +393 +
400 ) . The results appear in Table 63 .
In Table 64 we give the results of interpolation with
rather liberal smoothing . In the smoothing , we have been
guided by three facts . First , the ups and downs from about
55 to about 145 are quite irregular and , in view of the prob-
able errors , may well be chances of the sampling . Second ,
the data of Table 61 for the interval 15 to 20 and the gen-
eral drift of the Army results support the hypothesis that
original scores from 10 to 20 will have a true - unit value
even higher than our 13.67 for 20 to 30 ; so that the increase
from 10.81 for 40 to 50 , to 11.67 for 30 to 40 , and 13.67 for
20 to 30 may be considered real . Third , the facts of Table
62 and the facts of Table 65 for Alpha scores of a college-
graduate group show that the true - unit values per interval
of 5 score - points continue , above 175 , the rise shown in our
table from 150 to 170. We take 100 as the point of coinci-
1 Before combining them we have multiplied all the results from the 865
determination ( Grades 5 , 6 , and 7 ) by 1.08 , which puts their total value from
65 to 95 on a parity with the total value from 65 to 95 of the 5623 series ; we
have also multiplied all the results from the 1035 determination by 1.03 , which
puts their total value from 110 to 155 on a parity with the total value from
110 to 155 of the 5623 series . These multiplications serve to keep our systems
of values consistent .
17
TABLE 58 .
ARMY ALPHA : GRADE 5 : n = 263 .
X Δ
- or + Value of
A/.0588
Points Permille
on up to Deviation Interval
Original Stated from the in Equal
Scale Point Median Units
15 34 - 1.825007
- 1.248085
-
20 106 .576922 9.81
25 190 .877896 .370189 6.29
30 293 - .544642 .333254 5.67
35 4.73
-
395 .266311 .278331
40 494 .015040 .251271 4.27
45 589 + .224973 .240013 4.08
50 692 + .501527 .276554 4.70
55 768 + .732276 .230749 3.92
60 848 +1.027893 .295617 5.03
65 905 + 1.310579 .282686 4.81
70 939 + 1.546433 .235854 4.01
75 970 +1.880794 .334361 5.69
80 989 2.290370 .409576 6.96
+ +
1868
85 996 2.652070 .361700 6.15
TABLE 59
.
ARMY ALPHA GRADE 281

=
n
6
:
.
:
X Δ
Points Permille - or + Value of
on up to Deviation Interval
A
.0551
/
Original Stated from the in Equal

Scale Point Median Units
20 14 2.197286
- - - --
25 39 1.762410 .435876 7.91

30 89 1.346939 .415471 7.54
35 132 1.116987 .229952 4.17
40 214 .792619 .324368 5.89
-
45 299 .527279 .265340 4.81
-
-
50 427 .184017 .343252 6.23
55 566 .166199 .350216 6.36
60 630 .331853 .165654 3.01
+ + + + +
65 737 .634124 .302271 5.49

70 797 .830953 .196829 3.57
75 836 .978150 .147197 2.67
80 886 1.205527 .227377 4.13
85 939 +1.546433 .340906 6.19
90 964 +1.799118 .252685 4.58
95 985 2.170090 .370972 6.73
+
TABLE 64 .
EQUIVALENTS FOR ARMY ALPHA SCORES FROM 20 TO 170 IN A SCALE WITH EQUAL
/
UNITS , 1 OF THIS SCALE EQUALLING .89 90 OF THE DIFFERENCE BETWEEN
60 AND 150 OF THE ORIGINAL ALPHA SCORES , OR APPROXIMATELY 1/100
OF THE DIFFERENCE BETWEEN 50 AND 150 OF THE ORIGINAL
ALPHA SCORES .
Orig . Cor . Orig . Cor . Orig . Cor . Orig . Cor . Orig . Cor .
20 14.1 50 50.4 80 81.0 110 110 140 140

21 15.5 51 51.5 81 81.9 111 111 141 141
22 16.9 52 52.6 82 82.8 112 112 142 142
23 18.3 53 53.6 83 83.7 113 113 143 143
24 19.7 54 54.7 84 84.6 114 114 144 144
25 21.1 55 55.8 85 85.5 115 115 145 145
26 22.5 56 56.8 86 86.4 116 116 146 146
27 23.9 57 57.9 87 87.3 117 117 147 147
28 25.2 58 58.9 88 88.2 118 118 148 148
29 26.5 59 60.0 89 89.1 119 119 149 149
30 27.8 60 61 90 90 120 120 150 150
31 29.0 61 62 91 91 121 121 151 151.1
32 30.2 62 63 92 92 122 122 152 152.2
33 31.4 63 64 93 93 123 123 153 153.3
34 32.6 64 65 94 94 124 124 154 154.4
35 33.8 65 66 95 95 125 125 155 155.5
36 35.0 66 67 96 96 126 126 156 156.6
37 36.1 67 68 97 97 127 127 157 157.7
38 37.2 68 69 98 98 128 128 158 158.8
39 38.3 69 70 99 99 129 129 159 159.9
40 39.4 70 71 100 100 130 130 160 161.0
41 40.5 71 72 101 101 131 131 161 162.1
42 41.6 72 73 102 102 132 132 162 163.2
43 42.7 73 74 103 103 133 133 163 164.3
44 43.8 74 75 104 104 134 134 164 165.4
45 44.9 75 76 105 105 135 135 165 166.6
46 46.0 76 77 106 106 136 136 166 167.7
47 47.0 77 78 107 107 137 137 167 168.8
48 48.2 78 79 108 108 138 138 168 170.0
49 49.3 79 80 109 109 139 139 169 171.1
170 172.2
In our series as constructed from the extensive data in

Grades 5, 6, 7 , 9 , 12 , and 13 the 170 to 190 difference is .3084 .
So in the terms of that series 190 to 1948.39 , 195 to 199
= 12.98 , and 200 to 204-15.82 .
TABLE 64a .
PROVISIONAL EQUIVALENTS FOR ARMY ALPHA SCORES FROM 170 TO 209 ; SCALE
AS IN TABLE 64 .
Orig . Cor . Orig . Cor . Orig. Cor . Orig . Cor .
170 172.2 180 187.2 190 203 200 223

171 173.7 181 188.8 191 205 201 225
172 175.2 182 190.4 192 207 202 228
173 176.7 183 192 . 193 209 203 231
174 178.2 184 193.6 194 211 204 234
175 179.7 185 195.2 195 213 205 237
176 181.2 186 196.7 196 215 206 240
177 182.7 187 198.3 197 217 207 243
178 184.2 188 199.8 198 219 208 246
179 185.7 189 201.4 199 221 209 249
Averaging the two determinations for 190 to 194 ( 6.19

and 8.39 ) we have 7.28 as the best estimate for that . We
,
have 12.98 as the best estimate for 195 to 199 , and 15.82 as
the best estimate for 200 to 204. This would give us 20.26
for the interval with a probability that the in
190 to 199 ,
terval from 200 to 209 would be still larger .
We have not incorporated these determinations with
the others , because they are less secure . There can be no
doubt , however , that anyone will be much nearer the truth 1
by using them than by using the original score values for

Army Alpha . There can be little doubt also that the units
of Alpha represent in general progressively greater incre
ments of ability from 130 to 210. Subject to further inves
tigations we offer Table 64a as a scale of provisional
equivalents in truly equal units for Army Alpha scores of
170 to 209 .
The values of the letter - grade intervals used in the
Army reports are as follows , in equal units :
Original Corrected
209 to 135 High End of A
- to High End of B 249 to 135
134 to 105 High End of B to High End of C + 134 to 105
104 to High to High End of C 104 to 76
End of
-
75 C +
74 to 45 High End of C to High End of C 75 to 45
44 to High End of C - to High End of D 44 to 21
D
25
High D to High End of
D
24 to 15 End of 20 to 9
14 to 0 High End of D -- to Low End of ? to ฀
THE NATIONAL INTELLIGENCE EXAMINATION A

Consider next the case of the National Intelligence Ex
amination A. We use first the scores from 1,668 pupils in
Grade 6 and 494 pupils in Grade 9. The results of the
derivation of values in equal units are shown in Table 66 .
The procedure by which these are obtained is just the same
as that shown in detail in the case of Army Alpha . The two
sets of values are combined with weights of 3 and 1 .
TABLE 65 .
ARMY ALPHA DISTRIBUTION OF SCORES OF 216 COLLEGE GRADUATES .
100 to 104 1
105 to 109
110 to 114 2
2
115 to 119 2
120 to 124
125 to 129
3
130 to 134
135 to 139
9 8 35146
140 to 144
4
145 to 149
150 to 154
155 to 159
160 to 164
6
165 to 169 15
170 to 174 24
2253
175 to 179
180 to 184 38
185 to 189 22
190 to 194 20
195 to 199 18
200 to 204
2 8
205 to 209
We next use scores from 1,679 pupils Grade and 482

in
pupils Grade the results weighted by and being

in
3
8,
1,
as shown in Table 67. The values here are in the last col
,
umn divided by factor such that the total difference 70

a
,
to 170 is the same as for the determinations from Grades

and
6
.9
TABLE 63 .
FINAL ESTIMATE OF RELATIVE VALUES OF ARMY ALPHA SCORES IN EQUAL

UNITS .
Weighted Weighted
Average : Average :
Interval Intervals of 5 Intervals of 10
20 to 24 7.67
25 to 29 5.70 13.67
30 to 34 6.32
35 to 39 5.35 11.67
40 to 44 5.24
45 to 49 5.57 10.81
50 to 54 5.92
55 to 59 4.54 10.46
60 to 64 5.18
65 to 69 4.96 10.14
70 to 74 4.78
75 to 79 6.28 11.06
80 to 84 4.26
85 to 89 4.23 8.49
90 to 94 4.81
95 to 99 4.76 9.57
100 to 104 5.06
105 to 109 5.31 10.37
110 to 114 5.03
115 to 119 5.09 10.12
120 to 124 4.56
125 to 129 5.15 9.71
130 to 134 4.88
135 to 139 4.55 9.43
140 to 144 5.33
145 to 149 4.94 10.27
150 to 154 4.89
155 to 159 6.17 11.06
160 to 164 5.37
165 to 169 5.86 11.23
170 to 174 7.91
175 to 179 7.87 15.78
180 to 184 8.19
185 to 189 6.87 15.06
190 to 194 6.19
TABLE 64 .
EQUIVALENTS FOR ARMY ALPHA SCORES FROM 20 TO 170 IN A SCALE WITH EQUAL
/
UNITS , 1 OF THIS SCALE EQUALLING .89 90 OF THE DIFFERENCE BETWEEN
60 AND 150 OF THE ORIGINAL ALPHA SCORES , OR APPROXIMATELY 1/100
OF THE DIFFERENCE BETWEEN 50 AND 150 OF THE ORIGINAL
ALPHA SCORES .
Orig . Cor . Orig . Cor . Orig . Cor . Orig. Cor . Orig . Cor .
20 14.1 50 50.4 80 81.0 110 110 140 140

21 15.5 51 51.5 81 81.9 111 111 141 141
22 16.9 52 52.6 82 82.8 112 112 142 142
23 18.3 53 53.6 83 83.7 113 113 143 143
24 19.7 54 54.7 84 84.6 114 114 144 144
25 21.1 55 55.8 85 85.5 115 115 145 145
26 22.5 56 56.8 86 86.4 116 116 146 146
27 23.9 57 57.9 87 87.3 117 117 147 147
28 25.2 58 58.9 88 88.2 118 118 148 148
29 26.5 59 60.0 89 89.1 119 119 149 149
30 27.8 60 61 90 90 120 120 150 150
31 29.0 61 62 91 91 121 121 151 151.1
32 30.2 62 63 92 92 122 122 152 152.2
33 31.4 63 64 93 93 123 123 153 153.3
34 32.6 64 65 94 94 124 124 154 154.4
35 33.8 65 66 95 95 125 125 155 155.5
36 35.0 66 67 96 96 126 126 156 156.6
37 36.1 67 68 97 97 127 127 157 157.7
38 37.2 68 69 98 98 128 128 158 158.8
39 38.3 69 70 99 99 129 129 159 159.9
40 39.4 70 71 100 100 130 130 160 161.0
41 40.5 71 72 101 101 131 131 161 162.1
42 41.6 72 73 102 102 132 132 162 163.2
43 42.7 73 74 103 103 133 133 163 164.3
44 43.8 74 75 104 104 134 134 164 165.4
45 44.9 75 76 105 105 135 135 165 166.6
46 46.0 76 77 106 106 136 136 166 167.7
47 47.0 77 78 107 107 137 137 167 168.8
88888
48 48.2 78 79 108 108 138 138 168 170.0

49 49.3 79 80 109 109 139 139 169 171.1
170 172.2
In our series as constructed from the extensive data in

Grades and the 170 190 difference
12
13
.3084
to to
is
5,
6,
7,
9,
So in the terms of that series 190 1948.39 195 199

to
= 12.98 = 15.82
,
and 200 204

to
=
.
TABLE 64a .
PROVISIONAL EQUIVALENTS FOR ARMY ALPHA SCORES FROM 170 TO 209 ; SCALE
AS IN TABLE 64.
Orig . Cor . Orig . Cor . Orig . Cor . Orig . Cor .
170 172.2 180 187.2 190 203 200 223

171 173.7 181 188.8 191 205 201 225
172 175.2 182 190.4 192 207 202 228
173 176.7 183 192 . 193 209 203 231
174 178.2 184 193.6 194 211 204 234
175 179.7 185 195.2 195 213 205 237
176 181.2 186 196.7 196 215 206 240
177 182.7 187 198.3 197 217 207 243
178 184.2 188 199.8 198 219 208 246
179 185.7 189 201.4 199 221 209 249
Averaging the two determinations for 190 to 194 ( 6.19

and 8.39 ) , we have 7.28 as the best estimate for that . We
have 12.98 as the best estimate for 195 to 199 , and 15.82 as
the best estimate for 200 to 204. This would give us 20.26
for the interval with a probability that the in-
190 to 199 ,
terval from 200 to 209 would be still larger .
We have not incorporated these determinations with
the others , because they are less secure . There can be no
doubt , however , that anyone will be much nearer the truth
by using them than by using the original score values for
Army Alpha . There can be little doubt also that the units
of Alpha represent in general progressively greater incre-
ments of ability from 130 to 210. Subject to further inves-
tigations we offer Table 64a as a scale of provisional
equivalents in truly equal units for Army Alpha scores of
170 to 209 .
The values of the letter - grade intervals used in the
Army reports are as follows , in equal units :
Original Corrected
209 to 135 High End of A to High End of B- 249 to 135
134 to 105 High End of B -
to High End of C + 134 to 105
104 to 75 High End of C + to High End of C 104 to 76
High of High C 75 to
-
74 to 45 End C to End of 45
44 to 25 High End of C to High End of D 44 to 21

24 to 15 High End of D to High End of D- 20 to 1
14 to 0 High End of D- to Low End of D- ↑ to ฀
THE NATIONAL INTELLIGENCE EXAMINATION A

Consider next the case of the National Intelligence Ex-
amination A. We use first the scores from 1,668 pupils in
Grade 6 and 494 pupils in Grade 9. The results of the
derivation of values in equal units are shown in Table 66 .
The procedure by which these are obtained is just the same
as that shown in detail in the case of Army Alpha . The two
sets of values are combined with weights of 3 and 1 .
TABLE 65 .
ARMY ALPHA DISTRIBUTION OF SCORES OF 216 COLLEGE GRADUATES .
100 to 104 1
105 to 109
110 to 114 2
115 to 119 22
120 to 124
125 to 129
3
130 to 134
135 to 139
4 1
to
9 8 35146
140 144
145 to 149
150 to 154
155 to 159
160 to 164
6
165 to 169 15
170 to 174 24
175 to 179 22
180 to 184 38
185 to 189 22
190 to 194 20
195 to 199 18
200 to 204
2 8
205 to 209
We next use scores from 1,679 pupils Grade and 482

in
pupils Grade the results weighted by and being

in
in 3
8,
1,
as shown in Table 67. The values here are the last col-
,
umn divided by factor such that the total difference 70

a
,
to 170 is the same as for the determinations from Grades

and
6
9
.
TABLE 70 .
EQUIVALENTS FOR NATIONAL A SCORES FROM 20 TO 170 , IN A SCALE WITH

EQUAL UNITS 1 = APPROXIMATELY 1/50 OF THE DIFFERENCE BE
TWEEN 100 AND 150 OF THE ORIGINAL SCALE.
20 30.2 50 57.8 80 82.4 110 110 140 139

1 31.2 1 58.6 1 83.2 1 111 1 140.1
2 32.2 2 59.4 2 84.1 2 112 1 141.2
3 33.2 3 60.2 3 85.0 3 113 3 142.3
4 34.1 4 61.0 4 85.8 4 114 4 143.4
5 35.1 5 61.8 5 86.6 5 115 5 144.5
6 36.1 6 62.6 6 87.4 6 116 6 145.6
7 37.0 7 63.4 7 88.2 7 117 7 146.7
8 38.0 8 64.2 8 89.0 8 118 8 147.8
9 38.9 9 65.0 9 89.8 9 119 9 148.9
30 39.8 60 65.7 90 90.7 120 120 150 150.0
1 40.8 1 66.5 1 91.6 1 120.9 1 151.3
2 41.8 2 67.3 2 92.6 2 121.8 2 152.6
3 42.7 3 68.1 3 93.6 3 122.7 3 154.0
4 43.6 4 68.9 4 94.6 4 123.6 4 155.3
5 44.6 5 69.7 5 95.6 5 124.5 5 156.6
6 45.5 6 70.5 6 96.5 125.4 6 158.0
7 46.4 7 71.3 7 97.5 7 126.7 7 159.3
8 47.3 8 72.1 8 98.5 8 127.2 8 161.6
9 48.2 9 73.0 9 99.5 9 128.1 162.0
40 49.1 70 73.8 100 100.5 130 129.0 160 163.4
1 50.0 1 74.6 1 101.4 1 130 1 164.8
2 50.9 2 75.5 2 102.4 131 3 166.2
3 51.8 3 76.3 3 103.3 3 132 3 167.6
4 52.7 4 77.2 4 104.3 4 133 4 169.0
5 53.6 5 78.0 5 105.2 5 134 5 170.5
6 54.5 6 78.9 6 106.2 6 135 6 172.0
7 55.3 7 79.8 7 107.1 7 136 7 173.5
8 56.2 8 80.6 8 108.1 8 137 8 175.0
9 57.0 9 81.5 9 109.0 9 138 9 176.5
170 178.0
We combine these results with that from Grades 6, 7 , 8,

and 9 , allowing equal weight to each of the two , and so have ,
as provisional values for these low intervals , the following :
20-29 9.61
30-39 9.30
40-49 8.63
50-59 7.92
60-69 8.12
Using these values up to 70 and that of Table 68 from

70 on , and making the original scale and the scale in equal
units coincide at 120 , we have Table 70 as our transmuting
table .
THE OTIS ADVANCED EXAMINATION
In the case of the Otis Advanced Examination we have
the distributions shown in Table 71. We obtain the σ values
in equal units for each interval for each group , as shown in
the case of Army Alpha . In Groups , I II
, and III
we then
divide each of these by the difference between 70 and 140
( in equal units ) ; averageI and IIwith respective weights
of 2 and 1 ; combine this average with III
, giving equal
weight to Grade 6 and to Grade 9. In Group IV we divide

each of the σ values by the difference between 100 and 170
( in equal units ) and then multiply each by a factor which
makes the 100 to 170 difference for Group IV equal to that
I II III
for the , , weighted average . The , I II III
, weighted
average and the IV result are then averaged with weights
of 3 and 1 , respectively . The essential steps in these com
putations are shown in Table 72 , the last column of which
shows the combined estimate of the relative values of the
10 - point intervals from 30 to 200 in terms of equal units .
For convenience in interpretation these values are divided
by 1.06 , which makes the unit of the equal - unit scale 1/100
of the difference between 70 and 170 of the original scale .
By interpolation and smoothing , letting the two scales coin
cide at 100 , we obtain the equivalents of Table 73. It may
be noted that what scant data we have above 200 indicate
that the rise from 12.52 to 17.77 ( or 11.81 and 16.76 , after
division by 1.06 ) is not a matter of the sampling error . The
data give 21.60 ( or 20.38 ) as the value for 200 to 209 .
The interval from 20 to 29 has a value of 18.30 ( 17.26
after division by 1.06 ) by the sixth - grade groups , but this
is too unreliable for use without confirmation . We have
considered the facts for a fifth -grade population of 3,058
individuals and a fourth - grade population of 1,500 pupils .
We do not, of course , know that in these grades the distri
242
.
TABLE 68
.
NATIONAL A. SUMMARY OF DETERMINATIONS OF VALUES IN EQUAL UNITS
-o
By Grades By Grades Average By Grades By Grades Average Ot
6
9
8
6
9
7
7
Interval and and Interval and and
45 to 49 2.63 2.63
50 54 3.38 2.30 2.84
55 " 59 3.04 4.94 3.99 50 to 59 6.42 7.24 6.83 .41
60 64 2.35 1.78 2.07
65 " 69 4.50 3.82 4.16 60 69 6.85 5.60 6.23 .625
70 74 3.91 2.72 3.32
75 79 5.42 5.24 5.33 70 79 9.33 7.76 8.55 .78
80 84 4.82 4.02 4.42
" " " """

85 89 3.74 4.06 3.90 80 89 8.56 8.08 8.32 .24
THE MEASUREMENT
90 " 94 5.22 4.70 4.96
" " " "

888888
95 " 99 5.52 4.25 4.89 90 99 10.74 8.95 9.85 .895
100 104 4.76 4.65 4.71
105 109 4.15 5.32 4.74 100 66 109 8.91 9.97 9.44 .53
110 66 114 4.89 5.22 5.06
""
""
115 119 5.06 4.89 4.98 110 119 9.95 10.11 10.03 .08
120 124 4.73 4.63 4.68
""
125 129 4.51 4.03 4.27 120 129 9.24 8.66 8.95 .29
130134 4.86 4.81 4.84
135139 5.35 4.82 5.09 130 139 10.21 9.63 9.92 .29
140 144 5.15 5.47 5.31
OF INTELLIGENCE
145 "" 149 5.49 6.16 5.83 140 149 10.64 11.63 11.14 .505
150 154 5.85 5.73 5.79 " " " "
155159 7.37 7.78 7.58 150 159 13.22 13.51 13.37 .15
160 164 8.20 9.20 8.70
""
165 169 5.25 6.53 5.89 160 169 13.45 15.73 14.59 1.14
170 174 5.26 9.97 7.62
In Table 68 the results of the 6 , 9 and of the 7 , 8 deter-

minations are put side by side and averaged , and measures
of the unreliability of the averages are attached .
We have made a provisional extension of the transmu-
tations for National A down to original scores of 20 by
using the assumption that Grades 5 and 4 will show ap-
proximately normal distribution of perfectly measured in-
TABLE 69.
NATIONAL A: GRADE 4 ( n = 1677 ) AND GRADE 5 ( n = 2487 )
Original
Interval Distributions Values in Equal Units Aver- Average X
4 5 4 5 age .924
10- 19 10 3
20- 29 33 7 11.51 9.30 10.40 9.61
30-39 69 23 9.92 10.19 10.06 9.30
40-49 113 63 8.24 10.44 9.34 8.63
50-59 214 118 10.07 9.42 9.75 9.01
60-69 285 275 9.87 12.00 10.94 10.02
70-79 311 347 9.99 10.02 10.01 9.25
80-89 257 417 9.36 10.13 9.75 9.01
90-99 200 412 10.26 10.01 10.14 9.37
100-109 106 350 9.49 10.27 9.88 9.13
110-119 55 226 10.85 9.55 10.20 9.42
120-129 7 147 7.69 10.83 9.26 8.56
2 66 10.59 10.44 10.52 9.71
22
11
1
1
tellect . Table 69 shows the original distributions , the values

of the intervals in equal units by the assumption , the quo-
tients when these are divided by 1/50 of the difference
between 70 and 120 of the original scale , the averages for
the two grades , and these averages after multiplication by a
factor which equates the 70 to 120 difference with the 70 to
120 difference of the scale derived by the use of Grades 6
and 9.
TABLE 70 .
EQUIVALENTS FOR NATIONAL A SCORES FROM 20 TO 170 , IN A SCALE WITH

EQUAL UNITS 1 = APPROXIMATELY 1/50 OF THE DIFFERENCE BE-
TWEEN 100 AND 150 OF THE ORIGINAL SCALE.
20 30.2 50 57.8 80 82.4 110 110 140 139

1 31.2 1 58.6 1 83.2 1 111 1 140.1
2 32.2 2 59.4 2 84.1 2 112 1 141.2
3 33.2 3 60.2 3 85.0 3 113 3 142.3
4 34.1 4 61.0 4 85.8 4 114 4 143.4
5 35.1 5 61.8 5 86.6 5 115 5 144.5
6 36.1 6 62.6 6 87.4 6 116 6 145.6
7 37.0 7 63.4 7 88.2 7 117 7 146.7
8 38.0 8 64.2 8 89.0 8 118 8 147.8
9 38.9 9 65.0 9 89.8 9 119 9 148.9
30 39.8 60 65.7 90 90.7 120 120 150 150.0
1 40.8 1 66.5 1 91.6 1 120.9 1 151.3
2 41.8 2 67.3 2 92.6 2 121.8 2 152.6
3 42.7 3 68.1 3 93.6 3 122.7 3 154.0
4 43.6 4 68.9 4 94.6 4 123.6 4 155.3
44.6 5 69.7 5 95.6 5 124.5 5 156.6
6 45.5 6 70.5 6 96.5 6 125.4 6 158.0
7 46.4 7 71.3 7 97.5 7 126.7 7 159.3
8 47.3 8 72.1 8 98.5 8 127.2 8 161.6
9 48.2 9 73.0 9 99.5 9 128.1 9 162.0
40 49.1 70 73.8 100 100.5 130 129.0 160 163.4
1 50.0 1 74.6 1 101.4 1 130 1 164.8
2 50.9 2 75.5 2 102.4 2 131 3 166.2
3 51.8 3 76.3 3 103.3 3 132 3 167.6
4 52.7 4 77.2 4 104.3 4 133 4 169.0
5 53.6 5 78.0 5 105.2 5 134 5 170.5
6 54.5 6 78.9 6 106.2 6 135 6 172.0
7 55.3 7 79.8 7 107.1 7 136 7 173.5
8 56.2 8 80.6 8 108.1 8 137 8 175.0
9 57.0 9 81.5 9 109.0 9 138 176.5
170 178.0
We combine these results with that from Grades 6 , 7 , 8,

and 9 , allowing equal weight to each of the two , and so have ,
as provisional values for these low intervals , the following :
20-29 9.61
30-39 9.30
40-49 8.63
50-59 7.92
60-69 8.12
Using these values up to 70 and that of Table 68 from

70 on , and making the original scale and the scale in equal
units coincide at 120 we have Table 70 as our transmuting
,
table .
THE OTIS ADVANCED EXAMINATION
In the case of the Otis Advanced Examination we have
the distributions shown in Table 71. We obtain the σ values
in equal units for each interval for each group , as shown in
the case of Army Alpha . In Groups I , II , and III we then
divide each of these by the difference between 70 and 140
(in equal units ) ; average I and II with respective weights
of 2 and 1 ; combine this average with III
, giving equal
weight to Grade 6 and to Grade 9. In Group IV we divide

each of the σ values by the difference between 100 and 170
(in equal units ) and then multiply each by a factor which
makes the 100 to 170 difference for Group IV equal to that
I II III
for the , , weighted average . The , I II III
, weighted
average and the IV result are then averaged with weights
of 3 and 1, respectively
The essential steps in these com-
.
putations are shown in Table 72 , the last column of which
shows the combined estimate of the relative values of the
10 - point intervals from 30 to 200 in terms of equal units .
For convenience in interpretation these values are divided
by 1.06 , which makes the unit of the equal - unit scale 1/100
of the difference between 70 and 170 of the original scale .
By interpolation and smoothing , letting the two scales coin-
cide at 100 , we obtain the equivalents of Table 73. It may
be noted that what scant data we have above 200 indicate
that the rise from 12.52 to 17.77 ( or 11.81 and 16.76 , after
division by 1.06 ) is not a matter of the sampling error . The
data give 21.60 ( or 20.38 ) as the value for 200 to 209 .
The interval from 20 to 29 has a value of 18.30 ( 17.26
after division by 1.06 ) by the sixth -grade groups , but this
is too unreliable for use without confirmation . We have
considered the facts for a fifth - grade population of 3,058
individuals and a fourth - grade population of 1,500 pupils .
We do not , of course , know that in these grades the distri-
TABLE 71.
OTIS ADVANCED : DISTRIBUTIONS .
I
Grade 6
II III IV
Grade 6 Grade 9 Grade 12
Interval n = 4298 n = 1654 n = 3627 n = 1226
10 to 19 3 1
20 " 29 19 12
30 " 39 74 23
40 " 49 168 56
50 " 59 334 107 9
60 " 69 504 183 40
70 79 659 243 79
80 " 89 738 268 174 6
90 " 99 587 244 262 23
100 109 499 209 443 38
110 66 119 346 135 520 58
120129 193 93 541 95
130139 97 45 547 153
140 66 149 51 25 409 187
150 159 20 8 317 191
160 169 5 1 190 187
170 179 62 139
180 " 189 1 1 24 85
190 " " 199 10 50
200209 13
210 219 1
butions of truly measured intellect are of Form A ; but their

low end will not diverge enough from the corresponding
sections of Form A to invalidate the comparisons which we
shall make .
Assuming the low end to be of the geometrical form of
the corresponding section of Form A , and expressing the
true values of the interval 10 to 19 , and of the interval 20
to 29 in terms of the interval 30 to 39 , we find the following :
The relevant facts are that in Grade 4 we have 33 cases
( out of a total of 1,500 ) from 0 to 9 , 121 cases from 10 to 19 ,
246 cases from 20 to 29 , and 248 cases from 30 to 39 ; in
Grade 5 we have 6 , 41 , 112 and 257 ( out of the total of
3,058 ) in these same intervals .
Interval Grade 4 Grade 5 Average
10-19 1.65 1.37 1.51
20-29 1.43 1.00 1.22
We allow equal weights in averaging , because the larger

population of Grade 5 is offset by the larger proportion of
Grade 4 in the intervals studied .
TABLE 72 .
OTIS ADVANCED : EQUIVALENTS FOR EACH 10 -POINT INTERVAL OF THE ORIGINAL

SCALE IN EQUAL UNITS .
A B C D E F G H
6 6 2A + B 9 C+D 12 3E + F
Interval n = 4298 n = 1654 3 n = 3627 2 n = 1226 4 G 1.06 /
30 to 39 14.29 10.04 12.86 12.86 12.86 12.1
40 to 49 11.89 10.80 11.53 11.53 11.53 10.9
50 to 59 11.74 10.54 11.34 11.34 11.34 10.7
60 to 69 10.91 11.18 11.00 11.00 11.00 10.4
70 to 79 10.89 10.80 10.86 10.13 10.50 10.50 9.9
80 to 89 10.99 10.46 10.81 10.69 10.75 10.75 10.1
90 to 99 9.47 9.84 9.60 9.35 9.48 9.48 8.9
100 to 109 10.00 10.31 10.10 10.74 10.42 10.52 10.45 9.9
110 to 119 10.23 9.18 9.88 9.90 9.89 9.10 10.09 9.5
120 to 129 9.63 10.45 9.87 9.56 9.72 9.74 9.72 9.2
130 to 139 11.54 9.38 10.82 10.41 10.62 11.27 10.78 10.2
140 to 149 9.77 9.77 11.24 10.14 9.6
150 to 159 11.23 11.23 10.96 11.15 10.5
160 to 169 13.10 13.10 12.08 12.84 12.1
170 to 179 10.16 10.16 12.27 10.69 10.1
180 to 189 12.52 12.52 11.8
190 to 199 17.77 17.77 16.8
Multiplying the 12.86 of Column G and the 12.1 ( more

exactly 12.13 ) of Column H of Table 72 by 1.51 and 1.22 , we
have these values for the intervals :
10-19 by 1.06 ) .
19.42 ( or 18.32 when divided
20-29 15.69 ( or 14.80 when divided by 1.06 ) .
We may use these as provisional values subject to further
investigation . They are used in the extension of Table 73
by Table 73a .
THE HAGGERTY EXAMINATION , delta 2
In the case of the Haggerty Delta 2 we have the distribu-

tions shown in Table 74. After estimating the frequencies
18
TABLE 73 .
EQUIVALENTS FOR OTIS ADVANCED SCORES FROM 30 TO 200 IN A SCALE WITH
1
EQUAL UNITS . 1 = 120 OF THE DIFFERENCE BETWEEN 50
AND 170 OF THE ORIGINAL SCORES .
о с C C O C O C
30 26.4 70 70.6 110 109.5 150 147.8 190 192.3

31 27.7 71 71.6 111 110.5 151 148.9 191 193.6
32 28.9 72 72.6 112 111.4 152 149.9 192 195.1
33 30.2 73 73.6 113 112.4 153 151.0 193 196.6
34 31.4 74 74.6 114 113.3 154 152.0 194 198.1
35 32.6 75 75.6 115 114.3 155 152.9 195 199.7
36 33.8 76 76.6 116 115.2 156 154.0 196 201.3
37 35.0 77 77.6 117 116.2 157 155.0 197 203.0
38 36.2 78 78.6 118 117.1 158 156.1 198 204.7
39 37.4 79 79.6 119 118.1 159 157.2 199 206.5
40 38.6 80 80.6 120 119.0 160 158.3 200 208.3
41 39.7 81 81.6 121 120.0 161 159.4
42 40.8 82 82.5 122 120.9 162 160.5
43 41.9 83 83.5 123 121.9 163 161.6
44 43.0 84 84.5 124 122.8 164 162.7
45 44.1 85 85.4 125 123.8 165 163.8
46 45.2 86 86.4 126 124.7 166 164.9
47 46.3 87 87.4 127 125.7 167 166.0
48 47.4 88 88.3 128 126.6 168 167.1
49 48.4 89 89.3 129 127.6 169 168.2
50 49.5 90 90.3 130 128.6 170 169.3
51 50.6 91 91.3 131 129.5 171 170.4
52 51.6 92 92.2 132 130.5 172 171.5
53 52.7 93 93.2 133 131.4 173 172.6
54 53.8 94 94.2 134 132.4 174 173.7
55 54.9 95 95.1 135 133.4 175 174.8
56 56.0 96 96.1 136 134.3 176 175.9
57 57.0 97 97.1 137 135.3 177 177.0
58 58.1 98 98.0 138 136.2 178 178.1
59 59.2 99 99.0 139 137.2 179 179.3
60 60.2 100 100.0 140 138.2 180 180.5
61 61.3 101 101.0 141 139.1 181 181.6
62 62.3 102 101.9 142 140.1 182 182.8
63 63.4 103 102.9 143 141.0 183 183.9
64 64.4 104 103.9 144 142.0 184 185.1
65 65.5 105 104.8 145 143.0 185 186.3
66 66.5 106 105.7 146 143.9 186 187.5
67 67.5 107 106.7 147 144.9 187 188.7
68 68.6 108 107.6 148 145.8 188 189.9
69 69.6 109 108.6 149 146.8 189 191.1
TABLE 73a .
PROVISIONAL VALUES FOR OTIS ADVANCED SCORES FROM 10 TO 29 .
0 C O C O с O C
10 - 6.7 15 2.6 20 11.6 25 19.1

11 - 4.8 16 4.4 21 13.1 26 20.6
12 - 3.0 17 6.2 22 14.6 27 22.1
13 - 1.1 18 8.0 23 16.1 28 23.6
14 .8 19 9.8 24 17.6 29 25
in intervals of from the irregular arrangement of

10 III
and IV the values of each interval of each group in equal
,
units are computed . These values for , , and are put I II III
in terms of the difference between original 70 and original
130 ,to make them comparable . The two Grade 9 deter-
minations are then combined with weights of 1 and 3 , re-
spectively With these averages are combined the deter-
.
minations from Grade 6 , with weights of 2 for the former
and 1 for the latter . The determinations from Grade 12 are
TABLE 74 .
HAGGERTY DELTA 2 DISTRIBUTIONS
I II III IV
Grade 6 Grade 9 Grade 9 Grade 12
Interval n = 916 n = 473 n = 1995 n = 668
10 to 19 1
20 25 to 42 1
30 4
40 12 43 to 54 3
50 39 1 55 to 65 10
60 87 5 66 to 76 29
70 127 6
80 161 36 77 to 86 73 1
90 164 54 87 to 99 225 13
100 154 89 100 to 114 555 45
110 86 109 115 to 119 212 36
120 61 79 120 to 129 415 121
130 17 73 130 to 139 283 162
140 2 42 140 to 149 155 170
150 to 159 1 12 150 to 159 31 102
160 to 169 1 160 to 169 3 16
170 to 179 2
made comparable with this composite determination by

multiplying them by a factor such as makes the 100 to 150
difference the same for the Grade 12 group as for the com
posite . They are then combined with the composite deter
mination , the weights being 1 for the Grade 12 items and 3
for the composite . The essentials of these computations
appear in Table 75 , the last column of which gives the final
estimate . The units of the Haggerty Delta 2 score become
progressively " harder " ( that is , larger ) when put in equal
units , from some point in the 70's up to 160 .
TABLE 75 .
HAGGERTY DELTA 2 : VALUES IN EQUAL UNITS .
A B C D E F G
B + 3C A + 2D 3E + F
Interval n = 916 n = 473 n = 1995 4 3 n = 668 4
50 to 59 10.85 10.85 10.85

60 to 69 10.55 10.55 10.55
70 9.31 5.20 9.29 8.27 8.62 8.62
80 9.19 13.25 8.38 9.62 9.48 9.48
90 9.03 9.62 9.08 9.21 9.15 9.15
100 10.30 10.42 11.06 10.90 10.70 10.05 10.54
110 8.78 11.48 10.67 10.87 10.17 9.50 10.01
120 13.41 9.82 11.53 11.10 11.87 12.61 12.06
130 16.61 11.87 13.06 13.06 12.86 13.01
140 10.37 16.16 14.71 14.71 15.46 19.86
150 20.88 20.88
Interpolating , smoothing , expressing the values in

terms of 1/60 of the difference between original 60 and orig
inal 120 , and letting the two series coincide at 90 , we have
the equivalents of Table 76 .
THE TERMAN GROUP TEST

In the case of the Terman Group Test of Mental Ability ,
we have the scores of 5,582 pupils in Grade 7 , 9,087 in
Grade 8, 10,881 in Grade 9 , 6,730 in Grade 10 , 4,206 in
Grade 11 , and 4,886 in Grade 12. [ Terman , '23 , p . 9. ]
These are reported in the form of the point on the scale
TABLE 76 .
EQUIVALENTS FOR HAGGERTY DELTA 2 SCORES FROM 50 TO 160 , IN A SCALE

WITH EQUAL UNITS .
O с O с с
50 50.6 87 87.1 124 124.8

51 51.7 88 88.1 125 126.0
52 52.8 89 89.0 126 127.2
53 53.9 90 90.0 127 128.4
54 55.0 91 90.9 128 129.6
55 56.1 92 91.8 129 130.8
56 57.2 93 92.7 130 132.0
57 58.3 94 93.7 131 133.3
58 59.4 95 94.6 132 134.6
59 60.5 96 95.6 133 135.9
60 61.5 97 96.5 134 137.2
61 62.6 98 97.4 135 138.5
62 63.7 99 98.3 136 139.8
63 64.7 100 99.3 137 141.1
64 65.8 101 100.3 138 142.4
65 66.9 102 101.4 139 143.7
66 67.9 103 102.4 140 145.0
67 69.0 104 103.4 141 146.9
68 70.0 105 104.5 142 148.8
69 71.0 106 105.5 143 150.7
70 72.0 107 106.5 144 152.6
71 72.9 108 107.6 145 154.5
72 73.8 109 108.6 146 156.4
73 74.7 110 109.7 147 158.3
74 75.5 111 110.7 148 160.2
75 76.4 112 111.7 149 162.1
76 77.2 113 112.8 150 164.0
77 78.1 114 113.8 151 166.1
78 79.0 115 114.8 152 168.2
79 79.8 116 115.9 153 170.3
80 80.7 117 116.9 154 172.4
81 81.6 118 117.9 155 174.5
82 82.5 119 119.0 156 176.6
83 83.4 120 120.0 157 178.7
84 84.3 121 121.2 158 180.8
85 85.2 122 122.4 159 183.0
86 86.2 123 123.6 160 185.0
252
.
TABLE 77
.
TERMAN GROUP TEST
I
H K L M
A
B D E F G J
to
in
the Intervals between
to
Differences Equal Units Corresponding
Points Corresponding
B
Scale to G.
.
Entries Successive Scale Points of Columns
Permille
Grade Grade Grade Grade Grade Grade
Grade Grade Grade Grade Grade
9
Per Grade
8
12
7
10 11
8
12
7
10 11
mille
20 30 35 48 55 63
10 13.42 13.23
58 66 74 12.21 12.38 12.59 12.63
25 25 36 44
10.83 10.86 11.54 11.38
-
43 53 67 77 86 10.50 10.64
50 31 13.12
THE MEASUREMENT
100 12.11 12.27 12.49 12.52 13.31

100 38 52 63 79 90
8.17 8.28 8.42 8.45 8.98 8.85
43 58 71 86 99 109
150 6.72 7.14 7.03
92 105 112 6.49 6.58 6.69
200 47 64 76
5.57 5.65 5.74 5.76 6.12 6.03
250 51 69 81 98 112 122
5.00 5.07 5.15 5.17 5.50 5.42
54 73 86 103 118 128
300 9.93 9.78
128 138 9.03 9.16 9.31 9.34
400 61 81 .95 113
147 8.44 8.56 8.70 8.73 9.28 9.14
500 68 89 104 122 138
8.44 8.56 8.70 8.73 9.28 9.14
75 97 113 131 147 156
600 9.34 9.93 9.78
141 158 165 9.03 9.16 9.31
700 83 107 123
5.00 5.07 5.15 5.17 5.50 5.42
750 88 112 128 147 163 169
168 174 5.57 5.65 5.74 5.76 6.12 6.03
800 93 118 135 152
6.58 6.69 6.72 7.14 7.03
OF INTELLIGENCE
126 142 159 174 179 6.49

850 100 8.85
8.45 8.98
-
151 166 180 185 8.17 8.28 8.42
900 109 135
189 194 12.11 12.27 12.49 12.52 13.31 13.12
950 122 148 164 177
200 10.64 10.83 10.86 11.54 11.38
975 134 159 172 185 196 10.50
12.59 12.63 13.42 13.23
-
147 170 181 194 203 207 12.21 12.38
990
below which a certain permille of the group lies . Thus

Table 77 states that 10 permille or 1 percent of the popula
tion of Grade 7 had scores up to 20 , 25 permille had scores
up to 25 , 50 permille had scores up to 31 , and so on .
TABLE 78 .
TERMAN GROUP TEST .

Sample of the Six Sets of Values in Equal Units Whence the General
Transmutation Table Is Derived .
Gr. 7 Gr . 8 Gr . 9 Gr . 10 Gr . 11 Gr . 12 Average
100 100.00 100.00 100.00 100.00 100.00 100.00 100 .

101 100.91 100.92 100.99 101.04 101.19 100.98 101 .
102 101.82 101.84 101.98 102.08 102.38 101.97 102 .
103 102.72 102.76 102.97 103.11 103.57 102.95 103 .
104 103.63 103.67 103.97 104.04 104.76 103.93 104 .
105 104.54 104.58 104.94 104.98 105.95 104.92 105 .
106 105.45 105.50 105.90 105.91 106.82 105.90 106 .
107 106.36 106.41 106.87 106.85 107.70 106.88 107 .
108 107.26 107.42 107.83 107.78 108.57 107.87 108 .
109 108.17 108.44 108.79 108.71 109.45 108.85 109 .
110 109.10 109.45 109.76 109.65 110.32 111.19 110 .
111 110.03 110.46 110.73 110.58 111.19 113.54 111 .
112 110.96 111.48 111.70 111.52 112.07 115.88 112 .
113 111.90 112.42 112.67 112.45 112.99 116.48 113 .
114 112.83 113.36 103.60 113.42 113.90 117.09 114 .
115 113.76 114.30 104.53 114.39 114.82 117.69 115 .
: : : : :
: : : : :
: : : : :
190 195.75 198.69 199.79 198 .
191 197.15 200.34 201.25 200 .
192 198.56 201.99 202.70 201 .
193 199.96 203.64 204.16 203 .
194 201.36 205.29 205.62 204 .
195 206.94 207.52 207 .
196 208.58 209.41 209 .
197 210.50 211.31 211 .
198 212.42 213.21 213 .
199 214.34 215.10 215 .
200 216.26 217.00 217 .
We therefore find the value in equal units for the given

intervals , rather than for successive intervals of 5 or 10 .
These values are then divided by :
1/86 of the difference between 61 and 147 in the case of

Grade 7.
Grade 8 .
Grade 9 .
Grade 10 .
Grade 11 .
Grade 12 .
The results appear as the last six columns of Table 77 .

We then construct a transmutation table of values for each
grade , in each case coinciding with the original scale at 100 .
A sample piece of this table is shown as Table 78 .
A final transmutation table ( Table 79 ) is obtained by
averaging all six values for each point of the original scale ,
or all five values , if there are only five , or all four or three ,
if there are only four or three . The table does not extend
down or up beyond the points where three grade distribu
tions were available .
THE MYERS MENTAL MEASURE

Using the scores of 724 pupils in Grade 6 and 311 pupils
in Grade 9 in the Myers Mental Measure , we find the facts
of Table 80 and construct Table 81 as the transmuting table .
The two scales coincide at 46 .
THE PINTNER NON - LANGUAGE TEST

For the Pintner Non - Language Mental Test we have
the scores of 1,237 pupils in Grade 6 and 258 in Grade 9
shown in Table 82 ( columns I and II ) . From these we de
rive the values in true units for each twenty - point interval
and express them in terms of one two - hundredth of the dif
ference between 260 and 460. They appear as columns III
and IV of Table 82. When they are averaged , with weights
TABLE 79.
TERMAN GROUP TEST OF MENTAL ABILITY : VALUES IN EQUAL UNITS OF EACH

POINT ON THE ORIGINAL SCALE FROM 35 TO 193 .
о с O C O C O C 0 C
35 21 70 68 100 100 130 129 160 158

9 8 7 8885
223
69 101 129 159

6
4 3 2 1
4 3 2 1
4 3 2 1
4 3 2 1
24 71 102 130 160
25 72 103 131 161
27 73 104 132 162
40 28 75 74 105 105 135 133 165 163
30 75 106 134 165
9 8 7 6
9 8 7 6
9 8 7 6
9 8 7 6
12
31 76 107 135 166

33 77 108 136 167
3
4 34 78 109 137 168

45 36 80 80 110 110 140 138 170 170
37 81 111 139 171
6
4 3 2 1
4 3 2 1
4 3 2 1
4 3 2 1
39 82 112 140 172

40 83 113 141 174
8
41 84 114 142 175

9
8
50 43 85 85 115 115 145 143 175 176

44 86 116 144 178
4 3 2 1
9 8 7 6
7 6
8 7 6
8 7 6
46 87 117 145 179

47 88 118 146 181
8
49 89 118 147 182

9
9
9
55 50 90 90 120 119 150 148 180 183

51 91 120 149 185
9 8 7 6
4 3 2 1
4 3 2 1
4 3 2 1
4 3 2 1
52 92 121 150 186

54 93 122 151 188
55 94 123 152 189
60 56 95 95 125 124 155 153 185 191
58 96 125 154 192
3 2 1
8 7 6
8 7 6
7 6
59 97 126 155 194

7
60 98 127 156 195

8
68
61 99 128 157 197

4
9
9
65 63 190 198
64 200
6
4 3 2 1
65 201
66 203
789
67 204
9
TABLE 80 .
MYERS MENTAL MEASURE ; GRADE 6 ( n = 724 ) AND GRADE 9 ( n = 311 ) : VALUES

OF INTERVALS IN TERMS OF EQUAL UNITS , EXPRESSED AS MULTIPLES
1
OF
45 x ( DIFFERENCE BETWEEN 36 AND 81 ) .
Interval Grade 6 Grade 9 Average

21 to 25 9.69 9.69
26 " 30 6.78 6.78
31 35 7.89 11.83 9.86
36 " 40 6.98 6.10 6.54
41 "" 45 5.44 7.36 6.40
46 "
50 5.89 4.95 5.42
51 "" 55 5.01 5.23 5.12
56 " 60 6.43 5.38 5.91
6165 3.19 3.46 3.33
66 " 70 4.57 4.39 4.48
7175 3.86 5.62 4.74
76 " 80 3.61 1.62 2.62
81 " 85 5.35 5.35
TABLE 81 .
EQUIVALENTS OF SCORES FROM 21 TO 86 FOR MYERS MENTAL MEASURE :

IN EQUAL UNITS .
C C 0 с O C
21 5.7 41 39.6 61 61.0 81 78.0

22 7.7 42 40.9 62 62.0 82 79.0
23 9.7 43 42.2 63 62.9 83 80.0
24 11.6 44 43.4 64 63.9 84 81.0
25 13.5 45 44.7 65 64.8 85 82.0
26 15.4 46 46.0 66 65.7 86 82.9
27 16.7 47 47.0 67 66.7
28 18.1 48 48.1 68 67.6
29 19.4 49 49.2 69 68.5
30 20.8 50 50.3 70 69.4
31 22.2 51 51.4 71 70.3
32 24.2 52 52.4 72 71.3
33 26.2 53 53.4 73 72.2
34 28.2 54 54.4 74 73.1
35 30.2 55 55.5 75 74.0
36 32.1 56 56.5 76 74.9
37 34.4 57 57.4 77 75.5
38 35.7 58 58.3 78 76.1
39 37.0 59 59.2 79 76.7
40 38.3 60 60.1 80 77.3
of 4 and 1 , we have column V. The transmutation table ,

which coincides with the original scale at 300 , appears as
Table 83 .
THE I. E. R. TEST OF SELECTIVE AND RELATIONAL THINKING ,

GENERALIZATION AND ORGANIZATION
For the I. E. R. Test of Selective and Relational Think-
ing , Generalization and Organization , we have scores from
TABLE 82 .
PINTNER NON -LANGUAGE MENTAL TEST . ORIGINAL SCORES AND VALUES

OF INTERVALS IN EQUAL UNITS .
I II III IV V
Original Scores Values of Intervals Weighted
Interval Grade 6 Grade 9 in Equal Units Average
n = 1237 n = 258 Grade 6 Grade 9 Value
0 to 19 1
20 " 39
40 " 59 2
2
60 " 79 2
2
80 " 99 6
100 119 11 19.6 19.6
120 139 17 22.0 22.0
140 159 32 13.5 13.5
160 66 179 28 1 13.7 13.7
180" 199 37 1 12.3 12.3
200219 41 12.2 12.2
220 "" 239 48 12.2 12.2
240259 56 19.3 29.7 21.4
61565
260279 101 15.5 19.7 16.3

280 299 92 13 19.8 26.4 21.1
300 319 125 13 18.6 18.2 18.5
320329 119 11 17.7 12.7 16.7
340359 108 27 23.4 25.3 23.8
360379 126 21 19.3 16.8 18.8
380 399 85 28 23.2 21.1 22.8
"
400 419 78 26 17.4 19.4 17.8

420 439 42 25 23.9 19.8 23.1
440459 38 23 20.9 20.7 20.9
460 66 479 20 17 18.7 18.7
480 66 499 10 14 19.8 19.8
500 519 14 30.3 30.3
1 3 7
520539 23.0 23.0

6
540 559
560579
1 23
580
1
TABLE 83
A
O
.,
-
EQUIVALENTS FOR PINTNER NON LANGUAGE SCORES FROM 100 TO 380 IN SCALE WITH EQUAL UNITS AND REFER TO THE
..
A
ORIGINAL SCORES AND THE SCORES TRANSMUTED INTO SCALE WITH EQUAL UNITS
C
C
C
C
с
O O 0
100 136 140 177 180 205 220 229 260 263 300 300 340 341
258
137 178 205 230 263 301 341
12
138 179 206 230 264 302 342
1 2 3
1 2 3
12 3
139 179 206 231 265 303 343
12 3 4
140 180 207 231 266 303 344
141 181 208 232 267 304 345
142 181 208 233 268 305 346
143 182 209 233 268 306 347
144 183 209 234 269 307 348
1 213456789
5 6 7 8 9
456789
678 9
5 6 78 9
11234567 8 9
145 183 210 235 270 308 349
110 146 150 184 190 211 230 235 270 271 310 308 350 350
1
147 185 211 236 272 309 351
12
148 2. 185 212 236 272 310 352
12 3
148 186 213 237 273 311 353
1234
149 187 213 238 274 312 354

150 187 214 238 275 313 I 355
6
THE MEASUREMENT
151 188 214 239 276 313 356

GOT
7
152 189 215 239 277 314 357
3 4 5 6678
56 78
189 216 240 277 315 ∞ 357
153
3456789
1 2 3 4 5 66889
123HST89
8∞
9
2 3 56789
12 34 5 6 789
154 190 216 241 278 316 358
120 155 160 191 200 217 240 241 280 279 320 317 360 359
1
156 192 217 242 280 318 360
12
12
158 192 218 243 281 319 362
12 3
1 2 3
159 193 219 244 282 320 363
12 3 4
160 194 219 246 283 322 364
45
161 194 220 247 284 323 365

162 195 221 248 285 324 366
196 221 249 286 325 367
OF INTELLIGENCE
163
78
78
164 196 222 250 287 326 368
23456789
456789
567889
78 9
5 6 7 8 9
165 197 222 251 288 327 370
130 166 170 198 210 223 250 252 290 289 330 329 370 371
167 198 224 253 291 330 372
169 199 224 254 292 331 373
12 3
170 200 225 255 293 332 374
1234
171 200 225 256 294 333 375
3 5
172 201 226 257 295 335 376
12 45 6
5 6
1 2 3 4 5 6
202 227 258 296 336 378
LOGO
173
67
174 202 227 259 297 337 379

5 67 ∞
8
203 228 261 298

12341 678 QOLQ3HBO789
175 338 380

123456789
11020046789
89
a
78 9
9
176 204 228 262 299 339 381
3,231 pupils in Grade 9 and from two groups ( of 1,666 and

972 ) in Grade 12 , as shown in Table 84. The values of ten-
point intervals in true units derived from facts of Table 84
appear in Table 85. Table 86 is the transmutation table
TABLE 84 .
I.E.R. TESTS OF SELECTIVE AND RELATIONAL THINKING , GENERALIZATION AND

ORGANIZATION : DISTRIBUTIONS IN GRADE 9 AND GRADE 12.
Grade 9 Grade 12
n = 3231 n = 1666 n = 972
40 to 49 2
50 " 59 6
60 " 69 15 1
70 66 79 15
80 " 89 32
3 1
90 " 99 66
116
100109 82
6
110 119 127
4
"
120129 168 7
11 13
130 66 139 200 20 20
140149 271 32 20
150 159 258 39 37
"
160 66 169 295 56 31

170 179 272 79 51
180 189 303 94 65
"
190 199 246 119 76

"
200 " 209 228 119 76

210 219 195 129 95
220229 143 142 71
230239 115 145 86
240 249 74 134 66
250 259 49 145 72
260 269 26 103 59
"
270 279 20 77 32
280289 54 36
1 4 4 5 8
290 299 47 30
300 309 41
"
310 319 30
"
320 329 12
330 339
"
151
340 349 10
1
350 359
"
360 369
"
370 379 1
1
without smoothing . For rough work , it is sufficient to take

the original scores at their face value . They are very , very
close to the true values from 130 to 250 , and are nowhere
more than 3.7 points off .
TABLE 85 .
I.E.R. TEST OF SELECTIVE AND RELATIONAL THINKING , GENERALIZATION AND

ORGANIZATION . VALUES OF INTERVALS OF ORIGINAL
SCALE IN EQUAL UNITS .
Grade Grade Grade Grade Grades

Interval 9 12 12 12 9 and 12
n = 3231 n = 1666 n = 972 Average Average
90 to 99 11.9 11.9
100 109 9.7 9.7
110 119 10.6 10.6
120 129 10.4 10.4
130 " 139 9.8 10.4 13.5 11.9 10.9
140 149 11.1 10.7 8.8 9.75 10.4
150 159 9.3 9.1 11.4 10.25 9.8
160 169 10.0 9.9 7.4 8.65 9.3
170 179 9.0 10.4 9.7 10.05 9.6
180 189 10.5 9.9 10.3 10.1 10.3
190 " 199 9.3 10.5 10.3 10.4 9.9
200209 10.0 9.4 9.4 9.4 9.7
210 " 219 10.3 9.3 11.2 10.25 10.3
220229 9.7 9.9 8.3 9.1 9.4
230239 10.9 10.2 10.7 10.45 10.7
240249 10.3 10.1 8.9 9.5 9.9
250259 12.1 11.2 11.65 11.7
260 269 10.3 11.5 10.9 10.9
270 279 9.3 7.9 8.6 8.6
280 " 289 8.0 11.8 9.9 9.9
290 299 9.0 16.1 12.55 12.6
300 " 309 11.1 7.7 9.4 9.4
310 319 12.6 7.2 9.9 9.9
THE BROWN UNIVERSITY PSYCHOLOGICAL EXAMINATION
The results of similar computations for the Brown Uni-

versity Psychological Examination appear in Tables 87
and 88 .
.
TABLE 86
.
,
.
TRANSFORMATION TABLE I.E.R. TESTS OF SELECTIVE AND RELATIONAL THINKING GENERALIZATION AND ORGANIZATION
C
C
C
C
C
C
C
O 0 O O
90 87.2 120 119.4 150 151.1 180 179.8 210 209.7 240 240.1 270 272.6 300 303.7
88.4 120.5 152 180.8 210.8 241 273.5 304.6
89.6 121.6 153 181.8 211.9 242 274.3 305.5
90.8 122.6 154 182.9 213 243 275.2 306.4
90 1 2 3 4
92 123.6 155 183.9 214 244 276.1 307.4
1 2 3 4 5
5
93.1 124.7 156 184.9 215 245 277 308.3
94.3 125.7 157 186 216 246 277.8 309.3
95.5 126.7 158 187 217 247 278.6 310.3
96.7 127.8 159 188 218 248 279.4 311.3
7 8 9
1 2 3 4 5 6 7 8 9
1 2 3 4 5 6 7 8 9
1 2 3 4 5 6 7 8 9
1 2 3 4 5 6 7 8 9
7 8 9
1 2 3 4 5 6 7 8 9
1 2 3 4 5 6 7 8 9
97.9 128.8 160 189 219 249 280.3 312.3
100 99.1 130 129.8 160 160.9 190 190.1 220 220 250 250 280 281.2 310 313.3
100 130.9 161.8 191 221 251.1 282.1 314.3
THE TRANSFORMATION
101 132 162.7 192 222 252.3 283.1 315.3

1 2 3
102 133.1 163.7 193 223 253.4 284.1 316.2

103 134.2 164.6 194 224 254.6 285.1 317.2
104 135.3 165.5 195 224.9 255.7 286.1 318.2
1 2 3 4 5 6
105 136.4 166.4 196 225.8 256.9 287.1 319.2

106 137.4 167.3 197 226.7 258.1 288.1 320.2
106.9 138.5 168.3 198 227.6 259.3 289.1 321.2
1 2 3 4 5 6 7 8 9
1 2 3 4 5 6 7 8 9
1 2 3 4 5 6 7 8 9
1 2 3 4 5 6 7 8 9
8 9
1 2 3 4 5 6 7 8 9
1 2 3 4 5 6 7 8 9
5 6 7 8 9
107.9 139.6 169.2 199 228.5 260.5 290.1 322.2

110 108.8 140 140.7 170 170.2 200 200 230 229.4 260 261.7 290 291.1 320 323.2
109.9 141.8 171.1 201 230.5 262.8 292.3
OF STANDARD SCORES
111 142.9 172.1 202 231.6 263.9 293.6

112 144 173 203 232.6 265 294.8
113 145 174 204 233.7 266 296.1
114 146 175 205 234.7 267.1 297.4
115 147 176 206 235.8 268.2 298.7
261
116.1 148 177 207 236.9 269.3 300

117.2 149 178 207.9 238 270.4 301.2
1 2 3 4 5 6 7 8 9
1 2 3 4 5 6 7 8 9
1 2 3 4 5 6 7 8 9
1 2 3 4 5 6 7 8 9
1 2 3 4 5 6 7 8 9
1 2 3 4 5 6 7 8 9
1 2 3 4 5 6 7 8 9
118.3 150 178.9 208.8 239 271.5 302.4
TABLE 87 .
BROWN UNIVERSITY PSYCHOLOGICAL EXAMINATION . GRADES 12 3333 AND
N
=
,
13 , N = 2118 .
Δ
Value of interval Values in terms of
in equal units 1/35 the difference
of
Interval Grade 12 Grade 13 between 35 and 70 Average
3333 2118 Grade 12 Grade 13
=
n
=
n
20 to 24 .5304 6.25 6.25

25 " 29 .5800 6.84 6.84
30 " 34 .4764 5.62 5.62
35 " 39 .4500 .4260 5.30 5.13 5.22
40 " 44 .3780 .3741 4.46 4.50 4.48
45 " 49 .4172 .3442 4.92 4.14 4.53
50 " 54 .3976 .3851 4.69 4.63 4.66
5559 .4179 .4500 4.93 5.42 5.18
60 66 64 .4263 .4495 5.02 5.41 5.22
65 " 69 .4805 .4798 5.67 5.77 5.72
70 " 74 .4908 5.91 5.91
75 " 79 .6048 7.28 7.28
TABLE 88
.
EQUIVALENTS OF SCORES FROM 20 TO 80 FOR THE BROWN UNIVERSITY

PSYCHOLOGICAL EXAMINATION IN EQUAL UNITS
,
O O
O
C
C
20 17.0 40 41.0 60 59.8

1 18.2 41.9 60.8
4 3 2 1
4 3 2 1
2 19.5 42.8 61.9

20.7 43.7 63.0
34 22.0 44.6 64.0
25 23.4 45 45.5 65 65.1

24.6 46.4 66.2
8 7 6
9 8 7 6
26.0 47.3 67.4

1678
27.4 48.2 68.5

28.8 49.1 69.6
9
30 30.2 50 50 70 70.8
31.3 51 72.0
4 3 2 1
2 1
32.4 52 73.1
3 2
33.6 52.9 74.3

4
34
34.7 53.8 75.5
35 35.8 55 54.7 75 76.7

36.9 55.7 78.1
9 8 7 6
9 8 7 6
9 8 7 6
38.0 56.7 79.6

39.0 57.8 81.0
40.0 58.8 82.5
80 84.0
TABLE 89 .
ARMY EXAMINATION A : DISTRIBUTION OF PUPILS IN GRADES 4, 5 , 6 , 7 , 8, 9

AND 13 .
Interval 4 5 6 13
8
7
n = 463 n = 570 n = 672 685 630 701
=
=
=
n
n
0-9 4
10-19 9
20-29 16 4 1
2
30-39 21 5
40 36 10
1
50 41 21
6 7 1
60 46 26
2
2635
70 57 48
80 53 45 32 11
2 4
90 47 64 41 20
100 40 53 53 27 6
110 31 61 54 36 21
11 1
120 22 61 64 56 30
1
130 20 55 78 55 28
140 11 43 84 63 37
150 7 28 80 65 46
160 1 20 53 53 52
5 335
170 1 11 24 79 56
180 6 26 56 61
190 8
∞ 23 47 55 20
200 1 44 42 25
8
210 16 45 25
220 16 36 31
975
230 18 33 53
240 28 58
2 1 3
250 15 61
21
260 13 68
270 75
1 6 9
280 66
1
290 48
300 51
1
310 39
1
320 20
330 16
340 12
350 13
360
2
19
THE ARMY EXAMINATION A
For special reasonswe have investigated the values of

,
scores in the Army Examination a , although it is not in

present use . We have distributions from Grades 6 , 7, 8 ,
and from college freshman , nearly seven hundred in each
[ Memoirs , p . 537 ] , shown in Table 89. We shall also use to
TABLE 90.
ARMY EXAMINATION A : EQUIVALENTS FOR EACH 10-POINT INTERVAL OF THE

ORIGINAL SCALE IN EQUAL UNITS . RESULTS FROM GRADES 6 , 7 AND 8.
Original Value in Equal Units

Interval Gr . 6 Gr . 7 Gr . 8 Average
60 to 69 18.7 18.95 18.8

70 to 79 15.65 6.9 11.3
80 to 89 12.3 9.15 15.6 12.4
90 to 99 11.6 11.0 4.0 8.9
100 to 109 9.6 9.7 9.5 9.6
110 to 119 10.1 9.7 17.1 12.3
120 to 129 11.55 11.4 14.1 12.4
130 to 139 12.65 9.45 9.2 10.4
140 to 149 13.65 9.6 9.8 11.0
150 to 159 11.3 9.4 10.05 10.3
160 to 169 6.4 7.65 9.9 8.0
170 to 179 8.65 12.2 9.9 10.3
180 to 189 10.6 10.05 10.5 10.4
190 to 199 5.2 10.35 9.8 8.4
200 to 209 7.5 13.2 7.9 9.5
210 to 219 9.4 6.7 9.6 8.6
220 to 229 11.25 9.6 8.9 9.9
230 to 239 0.00 20.7 10.05 10.3
240 to 249 9.28 7.3 11.8 9.5
250 to 259 13.48 5.5 8.45 9.1
260 to 269 11.3 11.3
270 to 279 12.9 12.9
some extent the distributions of 570 pupils in Grade 5 and

463 pupils in Grade 4 , which are also shown in Table 89 .
The values of each 10 - point
interval from 60 to 270 in
equal units are computed for Grades 6 , 7 , and 8 , by the
methods previously used , and made strictly comparable by
being divided by the difference between 90 and 230 of the
original scale . They are then averaged . The true values

of the intervals from 10 to 100 are computed for Grades 4
and 5 , assuming that the low ends of these distributions
are distributed like the low end of Form A. They are made
strictly comparable and averaged . The averages are then
multiplied by a factor such that the difference 60 to 100 is
represented by the same amount in the series of true values
TABLE 91 .
ARMY EXAMINATION A : EQUIVALENTS FOR CERTAIN 10 - POINT INTERVALS OF

THE ORIGINAL SCALE IN EQUAL UNITS . RESULTS FROM GRADES 4 AND 5 .
Values in Equal Units

Interval Gr . 4 Gr . 5 Av . Av . x 1.28
10 to 19 16.4 16.4 21.0

20 " 29 13.15 13.15 16.8
30 10.1 10.3 10.2 13.1
40 11.9 10.5 11.2 14.3
50 10.1 12.2 11.65 14.3
60 9.7 9.5 9.6 12.3
70 10.8 11.9 11.35 14.5
80 10.0 8.6 9.3 11.9
90 " 99 9.7 10.4 10.05 12.9
obtained from Grades 6 , 7 , and 8 , and in the series obtained

from Grades 4 and 5. The 6 , 7 , 8 series is then extended by
the values from 10 to 100 obtained from Grades 4 and 5 ,
allowing equal weight to the two sets of values from 60
to 99 .
The values of the intervals from 200 to 360 are com-
puted from the facts for college freshmen ( Grade 13 ) . They
are then multiplied by a factor such that the difference 200
to 260 is represented by the same amount in the two series
of true values ( from Grades 6 , 7 , 8 , and from Grade 13 ) .
The 6 , 7 , 8 series is then extended by the values for 260 to
360 obtained from Grade 13 .
The essentials of these procedures and their results ap-
pear in Tables 90 , 91 , and 92 .
A transmutation table in steps of 10 is then made , let-
ting the two scales coincide at 170. This appears as Table
93 .
TABLE 92.
ARMY EXAMINATION A: EQUIVALENTS IN EQUAL UNITS . RESULTS FROM

GRADES 6 , 7 , AND 8 ; 4 AND 5 ; 13 ; AND COMPOSITE FROM ALL .
Value in Equal Units

Original By Grades By Grades By
Interval 6, 7 and 8 4 and 5 Grade 13 Composite
1- 9
10- 19 21.0 21.0
20-29 16.8 16.8
30-39 13.1 13.1
40-49 14.3 14.3
50- 59 14.3 14.3
60-69 18.8 12.3 15.6
70-79 11.3 14.5 12.9
80-89 12.4 11.9 12.2
90-99 8.9 12.9 10.9
100-109 9.6 9.6
110-119 12.3 12.3
120-129 12.4 12.4
130-139 10.4 10.4
140-149 11.0 11.0
150-159 10.3 10.3
160-169 8.0 8.0
170-179 10.3 10.3
180-189 10.4 10.4
190-199 8.4 8.4
200-209 9.5 9.5
210-219 8.6 8.6
220-229 9.9 9.9
230-239 10.3 10.3
240-249 9.5 9.5
250-259 9.1 9.1
260-269 11.3 8.7 10.0
270-279 12.9 9.2 11.1
280-289 10.3 10.3
290-299 9.7 9.7
300-309 8.1 8.1
310-319 10.3 10.3
320-329 10.8 10.8
330-339 7.7 7.7
340-349 8.6 8.6
350-359 9.7 9.7
TABLE 93 .
TRANSMUTATION TABLE FOR ARMY EXAM . A.
Original Scale in
Scale Equal Units
-
-
10 34.1
20 13.1
30 3.7
40 16.8
50 31.1
60 44.4
70 60.0
80 72.9
90 85.1
100 96.0
110 105.6
120 117.9
130 130.3
140 140.7
150 151.7
160 162.0
170 170.0
180 180.3
190 190.7
200 199.1
210 208.6
220 217.2
230 228.1
240 238.4
250 247.9
260 257.0
270 267.0
280 278.1
290 288.4
300 298.1
310 306.2
320 316.5
330 327.3
340 335.0
350 343.6
360 353.3
TABLE 94
A
IN
10
.A
80
.
EQUIVALENTS
/
FOR ARMY
OF
EXAMINATION SCORES FROM TO 360 SCALE WITH EQUAL UNITS 1-1 THE DIFFERENCE
.
BETWEEN 130 AND 210 OF THE ORIGINAL SCALE
C
C
O
C
C
O
C
C
O C
10 34 50 31 90 85
268
210 210 250 248 290 288 330 327
1
32 32 86 From 130 211 249 289 328
12
30
12
34 87 to 210 212 250 290 329
28 35 88 the values 213 251 291 330
12 3 4
26 36 89 of the two 214 252 292 330
24
12 3 45
38 90 scales are 215 253 293 331
22
.
40 91 identical 216 254 294 332
20 41 92 217 255 295 333
78
17
78
42 93 218
C423456789
256 296 334
1123456789
6 78 9
-
6889
4 5 6 7 8 9
15
1 23 4 56 78 9
44 94 219 257 297 335
20 13 60 45 100 95 220 219 260 258 300 298 340 336
11
1
46 96 220 259 299 337
-- 48 97 221 260 300 338
49 98 222 261 300 339
9 7 5
4
50 100 223 262 301 340

52 101 224 263 302 340
12 3 456
THE MEASUREMENT
53 102 225 264 303 341

55 103 226 265 304 342
1234∞
2
56 104 227 266 305 343
9
OF
8 9
5 678 9
a
45 6 78 9
2 345 6 78 9
678 9
58 105 228 12345 6 78 9 267 306 344

30 70 60 110 106 230 229 270 268 310 307 350 345
61 107 230 269 308 346
62 108 231 270 309 347
12 3
4 6 7 8
64 110 232 271 310 348
1234
10 65 111 233 272 311 349
12341
123 45
11
12 3 4 5
67 112 234 273 312 350

12 68 113
3 4 LOGO
235 274 313 350
13 70 115
INTELLIGENCE
236 275 314 351

CO∞
15 71 ∞ 116 237 276 315 352
56789
2 3 456 78 9
5 6 78 ∞
9
78 9
16
12 3 4 5 6 78 9
9
78 9
67 a
72 117 238 277 316 353

40 17 80 73 120 118 239
240 280 278 320 317 360 354
18 74 119 240 279 318
20
2
75 120 240 280 319
12 3
21 76 122 241 281 320
22 78
12 34
123 242 282 321

193410
24
12 3 4 5
79 124 243 283 322
6
25 80 125 244 284
5 TO
323
2 34 5 6 7
26 81 126 245 285 324
28
6 100
6 78
83 128
9
246 286 325
456789
1 2 3 4 5 6 78 9
30
6849
84 129
9
78 a
247 287 326

From 130 to 360 , the original scale value may be used

with little error , but from 130 down the true values of the
original scale units increase so that these 120 points of the
original scale are equal to about 164 elsewhere . Table 94
presents a detailed transmutation table made with some
smoothing .
CHAPTER VIII
THE FORM OF DISTRIBUTION OF INTELLECT IN MAN
The orthodox doctrine is that the form of distribution
of intellect in human beings of the same sex and age is
Form A, shown in Fig. 15 , representing a fact whose varia-
tions up and down from its average condition are caused by
a large number of uncorrelated factors each of which exer-
FIG . 15. Form A , The Normal Probability Surface .
cises about the same amount of influence on intellect as any

other, and being a surface enclosed by a curve approximat-
1 -x2
ing the normal probability curve y= where σ is
σ2π
e202
the mean square variation .

This doctrine was urged by Francis Galton , on the basis
partly of analogy with the facts in the case of certain bodily
dimensions , and partly of his own shrewd observations of
human abilities . Since his day it has gained very wide ac-
ceptance . This is partly because the measurements of in-
tellect and of other mental abilities in children of the same
270
THE FORM OF DISTRIBUTION OF INTELLECT IN MAN 271
age ( their units being taken at their face value ) have uni
formly shown continuity clustering around one mode , with
diminishing frequencies in proportion to remoteness from
that mode , and with no notable departure from symmetry
toward any one special form of asymmetry . It is partly
because some assumption had to be made in one investiga
tion after another for purposes of quantitative treatment ,
and this assumption was about as safe as any other one
assumption , and much easier to operate with . Hence we
gradually slid into the habit of using the doctrine . This
fashion became so strong that in recent years psychologists
have assumed symmetry , even though the units taken at
their face value produced a markedly skewed distribution .
GENERAL CONSIDERATIONS
Many of those who have made extensive use of this as
sumption have been aware of its highly hypothetical nature .
The argument from analogy is weak because so many bodily
variables are clearly skewed in distribution . Such are
weight, longevity , girth of chest , strength of arm pull . The
argument from mental measurements is weak , not only
because of the general ambiguity of the units , but still more
because the " error " has been a large proportion of the
variation in many of the investigations . The " error " being
symmetrical and " normal " tends to add a spurious sym
metry and normality to the variability . Moreover , some
times the selection is such that normality in the group mea
sured may well be an argument in favor of skewness for
man in general . So , for example , with sixteen - year olds in
high school , or twenty - five - year olds in universities .
In general the form of distribution of any variable trait
is due to the number of causes that influence variations in
its amount , their magnitudes and their interrelations.¹
Since we do not know what the causes of the variations in
1 There is a certain regrettable vagueness , not to say ignorance , concerning
the causation of variations , as when psychologists consider the amount of in
tellect to be a consequence of the presence or absence of a single Mendelian
determiner , and yet to be distributed unimodally in Form A. Either of these
beliefs really denies the other .
intellect in human beings of the same age are , we cannot as

yet count them or measure their magnitudes or determine
their correlations . We should then be very skeptical of a
priori assumptions of Form A as the form of distribution
of intellect in human beings of the same age . They are
very much stronger in the case of children in the same
school grade .
In the case of children in the same school grade the
causes are our own acts ; and we do know that school au
thorities have a rough standard of the educational ability
which belongs in a certain grade , say Grade 7 , that intellect
correlates closely with educational ability , that departures
from this standard ( that is , mistakes in grading ) are rare
in some proportion to their magnitude , that they are due
to many causes ( the different teachers ' judgments with all
the experiences upon which they are based , and the ideals
and prejudices which they exemplify, and the other causes
of error to which they are subject ), and that many of these
causes are only loosely inter - correlated . These are all fea
tures of a status productive of symmetry and normality .
In the case of children of the same age ( or age and sex
and race ) the causes are acts of nature , many of them hap
pening millenniums ago ; and we do not even know whether
the hereditary factors of variability in intellect are six big
ones or sixty small ones . We do not know whether the
words heard and acts seen in the first three years of life are
of almost zero consequence , as used to be thought when
favored children were turned over to healthy peasants dur
ing this period ; or are of enormous consequence , as is as
serted by Freud and (but for different reasons ) by Wat
son . So we may best consider the facts of the distribution
of intellect in man with little or no pre - disposition .
2 There is one special set of major causes of variation about whose action
we do know something . Certain diseases and certain accidents , either before
or during or after birth , act to prevent or reduce the development of intellect .
In some cases one of these causes may act to prevent intellect from reaching
more than a certain very lowly status regardless of what might have happened
had its action been withheld . The result may be that whatever the distribu
tion apart from these causes , there is combined with it a very small distribu
tion with a mode at a very low degree of intellect , as shown in Fig . 16 .
.
TABLE 95
,
.
:
NATIONAL INTELLIGENCE EXAMINATION DISTRIBUTION OF SCORES FOR WHITE PUPILS AGE 11
to
3
8
in
Interval Frequencies Grades Permille
3A 3B 4A 4B 5A 5B 6A 6B 7A 7B 8A 8B Total intervals of
5
10
10- 14
5
1
ст
15- 19
1
20 10
THE FORM
4500
25 10 12
13
30 14
7
35 + 13
1
40 18 11
22431323
45 26 16 27
774
50 30 18
23668 4 LURR
1 13 46
2
2
55 12 47 29 47
2113∞
6 5
1
60 15 13 33 70 43
3
65 ∞ 27 14 63 38 81
1
1
1
14 26 10 72 44
OF DISTRIBUTION
70
75 40 20 92 56 100
3 886
80 40 25 88 54
OF
85 43 16 12 99 60 114
1
90 46 16 15 99 60
5
95 56 24 18 10 122 74 135
2136626OOKER308722
1
100 44 21 24 11 110 67
1
105 35 18 20 13 93 57 124
23595ERE
110 25 13 28 12 87 53
1
115 21 17 37 13 97 59 112
9
271
120 24 33 86 52
125 15 33 12 71 43 96
8
2
130 33 14 62 38
8423
2113∞
135 30 56 34 72
76
1
140 25 39 24
INTELLECT IN MAN
1
82
145 10 19 12 36
8
8 6 2 11
1 1
150 14
1
155 11 17 10 19
273
160
4756
3 4
2 2
7
2 1 1 1
21124 6899975215223
165-169
0
23
Total 23 19 161 97 528 226 369 142 10 1638 996 1000
THE EVIDENCE
The results stated in Chapter VII permit us to free the
evidence of the past from the ambiguity and misleading of
units whose real value was unknown . If
we had the time
and facilities we could free them also from the constant
,
tendency toward symmetry and normality due to the error

of measurement , but that work must be delayed . What we
can do now is to show the form of distribution of children
of the same year - age in respect of intellect in so far as it is
measured by the Haggerty or by the Otis or by the National
A , and in so far as the children examined are a random
sampling of the children at that age . We do not separate
FIG . 16
the sexes , because the sex differences are small and the
separation would leave us with too small populations . We
do not separate races , because that cannot be done in the
records available . Negro schools are very rarely , if ever ,
included in the records ; but negro children and children of
mixed parentage doubtless are sometimes reported without
distinction , and so included in the distributions .
The ages used are 11 , 12 , 13 , and 14 , at which years cer
tain very dull children have been excluded from school at
home or in institutions .Some 14 - year olds have left school .
The measurements were taken in schools , so that there is
.
TABLE 96
.
,
12
:
NATIONAL INTELLIGENCE EXAMINATION DISTRIBUTION OF SCORES FOR WHITE PUPILS AGE
8
3
in
Interval Frequencies Grades to
:
Permille
3B 4A 4B 5A 5B 6A 6B 7A 7B 8A 8B Total intervals of
3A
5
10
10-14 0.6
2
1
15-19
1
1
20
5
25
37
THE FORM
1
30
112121
10
35
OF
3
14
40
8
11
34271
22
45
50 13
3
∞ 38 25 34
55
∞ITT
TIT
1 1 1
20
1 1 1
60 ∞ 30
52
8 3 3
49 32
124B002201
65
9
70 51 34
75 11 10 71 47 80
DISTRIBUTION
3376
1
∞TTE
771
1
80 19 18 13 62 41
24265885
33037
∞
8
85 20 15 56 37 78
1 3 1
9
90 20 30 11 78 51
1
31
95 13 28 21 10 89 59 110
221464234L1
100 11 10 19 16 20 79 52
2
105 17 34 16 13 101 66 118
3
110 11 35 14 15 92 61
1
115 11 37 12 27 106 70 131

120 22 17 31 85 56
789
1336254KLESESCOL
125 20 23 71 47 103
130 21 36 70 46
1
4 3 1 2
135 16 24 10 59 39 85
20 29 61 40
263758D02
140
222
1
145 13 22 12 59 39 79
OF INTELLECT IN MAN
5
1262213431
150 13 26 49 32
51521
155 24 41 27 59
160 15 20 13
9 1 2
165 15 10 23
275
775
1
170
63311
8
3 3
6 6
11 11
44
1
175-179
3
7
55
Total 14 55 38 266 161 359 161 337 84 33 1518 1002
too small a representation of truants and of sickly children .

City children are more fully represented than country chil-
dren . The schools are predominantly public schools ; so
that Catholic children are insufficiently represented . The
age is doubtless sometimes in error , and is a year wide .
The latter fact should spread and flatten all the distribu-
tions a little .
FIG . 17. The form of distribution of the scores of 11-year - old children in
National A , transmuted into a scale with equal units .
FIG . 18. The same as Fig . 17 , but for 12 -year - old children .
It seems unwise to tamper with the records in an effort

to allow for these various factors and make the distribu-
tions more exactly representative of " all white children of
the United States of age x . " The process of allowance
would probably make improvements , but they would be
small and uncertain and very tedious to make and to under-
stand . So we shall take the facts just as Haggerty , Otis ,
and the National Committee give them ; and do nothing to
them save transmute each scale interval into units which
are truly equal , construct the resulting distributions , and
measure certain of their properties . In our inferences from
the results we shall , of course , try to bear all the condition

ing factors in mind .
Consider first Tables 95 to 98 , which give the facts in
the case of the National A. Tables 95 , 96 , and 97 give the
original data . Table 98 gives the data for constructing the
surfaces of frequency in the shape of the true values for
each interval , which are taken as lengths along the abscissa
line , and the quotients of the permille numbers each divided
by its corresponding abscissa length . These quotients give
FIG . 19. The same as Fig . 17 , but for 13 -year - old children .
FIG . 20. An approximate composite of Figs . 17 , 18 , and 19 .
the relative magnitudes of the ordinates or heights of the

rectangles erected over the corresponding abscissa lengths .
Figs . 17 , 18 , and 19 show the resulting surfaces of fre
quency with equal units . Fig . 20 shows a rough composite
picture of the form of distribution of National A ability in
children of the same year - age . It contains the three sepa
rate distributions centered on their medians .
Tables 99 and 100 show the original data in the case of
the Otis Advanced Examination . Table 101 shows the ab
.
TABLE 97
:A
,
.
NATIONAL INTELLIGENCE EXAMINATION DISTRIBUTION OF SCORES FOR WHITE PUPILS AGE 13
to
8
3
in
Interval Frequencies Grades Permille
intervals of
3A 3B 4A 4B 5A 5B 6A 6B 7A 7B 8A 8B Total
5
10
278
1
10-14 0.8
1
1
15-19 2.8
20-24 0.8
25 0.8 1.6
1
30
121145
1
1
1
7
35
0200343
1 1 11
1
40
2
45
5
15 12 15
1
50 17 14
6
55 17 14 28
HH3LH992
11322
42
60 12 24 19
65 14 28 22 41
1
70 16 39 31
9
75 25 20 51
8 3 8
8
THE MEASUREMENT
80
4216131
47 37
1 1
1229∞8
85 10 40 31 68
34312
90 13 19 13 67 53
OF
8
95 16 10
826
49 39 92
212196706
2 3 3 3
6 5 7
100 18 16 13 69 54
105 10 15 25 73 57 111
DOUBTIL
2
8
110 19 19 13 73 57
6
115 20 16 33 71
90 128
4365995
3
120 17 28 15 83 65
311
125 24 23 10 82 65 130
8
6
130 29 14
12456∞∞∞∞∞GELDA
64 50
135 11 22
INTELLIGENCE
13 11 71 56 106
6
5
140 19 18 58 46
145 23 22 64 50 96
8 9 9
150 19 16 57 45
155 13 10 36 28
964623ILL
73
1 1
160 11 33 26
HHHHHGR45221
23127B37624
165
20 16 42
170
48421
9 5 4 1
175
52
7
180
621
0.8
0
3
Total 22 25 127 89 201 130 329 186 114 42 1268 1000.2
TABLE 98 .
NATIONAL A. DATA FOR SURFACE OF FREQUENCY IN EQUAL UNITS .
Abscissa Ordinate Heights to Make the Areas Equal to

Original
Interval
Length in
Equal
Units Age 11
-
the Corresponding Permille Entries of
Tables 95 , 96 and 97
Age 12 Age 13
10-19 10.00 (Est . ) 5 2 2.5

20-29 9.61 12.5 6.2 1.6
30 9.30 14.0 10.8 7.5
40 8.63 31.3 25.5 17.4
50 7.92 59.3 42.9 35.4
60 8.12 99.8 64.0 50.5
70 8.55 117.0 93.6 59.6
80 8.32 137.0 93.8 81.7
90 9.85 137.0 112.0 93.4
100 9.44 131.0 125.0 118.0
110 10.03 112.0 131.0 128.0
120 8.95 107.0 115.0 145.0
130 9.92 72.6 85.7 106.8
140 11.14 32.3 71.0 86.3
150 13.37 14.2 44.1 54.5
160 14.59 4.8 15.8 28.8
170 15.00 ( Est . ) 5.3 4.7
180 16.00 ( Est . ) 0.7
scissa lengths in equal units and the ordinate heights ob

tained by dividing each original permille number by the
corresponding abscissa length in equal units . Figs . 21 , 22 ,
23 , and 24 show the surfaces drawn according to Table VII .
FIG . 21. The form of distribution of the scores of 11 -year - old children in Otis
Advanced , transmuted into a scale with equal units .
20
FIG . 22. The same as Fig. 21 , but for 12 - year -olds .
FIG . 23. The same as Fig. 21 , but for 13 -year -olds .
FIG . 24. The same as Fig . 21 , but for 14 - year - olds .
FIG . 25. An approximate composite of Figs . 21 , 22 , 23 , 24 .

Fig . 25 is a composite repeating Figs .

21 to 24 , with the
four medians coinciding .
Table 102 shows the original data for the Haggerty

Delta 2 ; Table 103 shows the lengths and heights when
FIG . 26. The form of distribution of the scores of 11 -year -old children in the
Haggerty Delta Two , transmuted into a scale with equal units .
FIG . 27. The same as Fig . 26 , but for 12 - year -olds .
FIG . 28. The same as Fig. 26 , but for 13 - year -olds .

.
TABLE 99
:
OTIS ADVANCED EXAMINATION DISTRIBUTION OF SCORES AGES 11 AND 12
282
12
11
Age
10
Age
to
4
in
to
9
4
Grades
in
Interval Frequencies Grades Frequencies
Per
Per
5
mille
4
10 Total
H
5
8
9
6
mille
10
Total
1
0-9
2 2
1 1
10-19
1 4 7
1 3 6
113
23
4
20-29 10 16
2
29 34
164
30 16 20 32 22
29 12 42 49
40 27 43 70
1 3
62 72
7 9
60 77 125 31 25
50
29 41 11 86 100
60 50 15 71 115
142
1 4 2
211
29 52 12 96 112
92 150
THE MEASUREMENT
70 55 30
3
33 74 31 141 165
8 6 3 2 1
80 44 32 85 138
63 27 14 124 133
90 23 37 67 109
84 98
231T1
35 58 94 41 21 14
100 16
9 5 3
69 81
1
19
9
110 30 44 72 33 11
37 43
1 1 3 3 8 6 6 4 4
120 16 22 36 13 10
39
2
13 33
22
130 12 19
27
3
11 23
140
5 4
22
9 4 2 3
19
322
1
1222
150 10
9 H
7 4 1 1
5 6 1
6 1 1 1
160 2
OF INTELLIGENCE
12 1 3 6 5421
122T12 1
1
170
841
7 3 1
4 1 1
180
190
200
210
25 10
2
309 220 43 13 11 197 372 135 106
Total 31
equal units are used . Figs . 26 , 27 , 28 , and 29 are the result-

ing surfaces of frequency Fig . 30 is their composite .
;
Fig . 31 is a composite of the three composites , Figs . 20 ,

25 and 30 .
FIG . 29. The same as Fig . 26 , but for 14 -year -olds .
FIG . 30. An approximate composite of Figs . 26 , 27 , 28 , and 29 .
FIG . 31. A composite of three composites .

.
TABLE 100
:
EXAMINATION DISTRIBUTION OF SCORES AGES 12 AND 14
OTIS ADVANCED
14
Age
Age 13
11
to
5
in
to
Grades
in
Grades 11 Frequencies
9
Interval Frequencies
7
11 Per
6
Total
5
10
7
Per
6
11 Total
5
10
4
mille
mille
0-9
1 4
1 4
10-19
1
20 15 16
12
2 5
12
55
30
265
24
16 17
1 1 7 6
113
1 3
39 31
40 12 11
1
59 15 14 13 41 45
8
58
1 1
50 26 23 54 59
19 17
2
21 10 12 76 77
122
60 31 52
7 7 6
21 14 48
45 16 12 94 96
70 21 69 75
10 25 11
3
99 102 12
39 21 28
183
80 88 96
9 4 3 1
109 14 11 40 19
8 5
29 17 41 11 107
90
6
40 30 10 97 105
18 118 120 10
100 25 22 44
109 118
2
1
10 37 41 17
46 6
92 94
1
110 11 10 48 15
31 53 17 109 119
1 1
47 17 12 92 94
1 1 1 1
120 11 88 96
92 26 48
7 1
39 32 13 90 1
130 78
7 1 1
39 25 72
8 5 3
20 16 46 47
3
140
7 7
34 12 56 61
10 13 30 31
150 14 30 33
14
1
12 19 19
5
160
21
15 16
7 7 2 1
4 2
1 1
170
22
5 3
5 3
3 2
180
8 1 1
5 2 1
2 1
1
190
200
210
9
252 304 129

3
58 48 106 72
3
Total 117 219 125 311 144
The ability which is measured by the score of any one

of the commonly used intelligence examinations is thus
shown to be distributed in children of the same age ( from
11 to 14 in rather close approximation to Form A. There
are no demonstrable departures from unimodality or from
TABLE 101.
OTIS ADVANCED EXAMINATION : DATA BY WHICH THE SURFACES OF

FREQUENCY ARE CONSTRUCTED .
Heights of the Surface of Frequency with

Intervals Values of
Equal Units to Make the Areas Equal
by Original Intervals in
Scores Equal Units to the Corresponding Permille
Entries
11 12 13 14
0-9 20.0 ( Est . ) 1.0 0.5

10-19 18.3 1.1 2.2 .5
20-29 14.8 10.8 4.7 6.6 2.5
30 12.2 26.2 27.9 9.8 13.4
40 10.9 64.2 45.0 28.4 16.0
50 10.7 116.9 67.3 55.2 41.7
60 10.4 110.5 96.2 74.1 56.5
70 10.0 150.0 112.0 96.0 52.2
80 9.7 142.3 170.0 105.0 77.2
90 9.7 112.4 137.0 112.0 98.9
100 9.5 98.9 103.0 126.0 110.9
110 9.5 75.8 85.3 99.0 125.0
120 9.6 37.5 44.8 98.0 123.9
130 9.6 19.8 40.7 95.9 99.7
140 9.6 8.3 28.2 49.0 81.5
150 10.5 9.5 21.0 29.6 58.0
160 11.0 1.8 7.3 17.3 29.7
170 11.2 3.6 4.5 14.6
180 11.8 .9 1.7 4.7
190 16.0 .6 2.1
symmetry ; the decrease in frequency as we pass from the

mode is slow , then more rapid , and then slow again .
THE FORM OF DISTRIBUTION AT AGES UP TO FIFTEEN

It
is reasonable to infer that the form of distribution
which is found for these examination scores , when trans
formed into a scale with equal units , will be found with very
little change for any valid measures of the altitude of In-

tellect CAVD , or of Intellect GOPI ( letting G refer to geo-
metrical tasks , O to opposites , P to picture completions , I
to information ) , or of any representative sampling of intel-
lectual tasks . It is reasonable to carry the inference on to
any valid measures of the histological and physiological
basis of altitude of intellect . It is probably safe also to ex-
TABLE 102 .
HAGGERTY DELTA 2 ; DISTRIBUTION OF SCORES . DATA FROM MADSEN [ 22 ] .
Age 11 Age 12 Age 13 Age 14

Grades 3 to 8 Grades 3 to 10 Grades 3 to 10 Grades 3 to 11
n Per- n Per- n Per- n Per-
mille mille mille mille
0-9
5 2
7 6
10- 19
25
10
8 5
4
568
486
9 132
507
20-29 10 3.5
6
30-39 39 48 22 28 11 14
6
40-49 78 96 35 44 16 21 11 19
50-59 80 98 62 79 26 34 13 23
60-69 99 122 74 94 48 62 32 57
70-79 106 130 92 117 65 85 37 66
80-89 106 130 93 118 88 114 66 117
90-99 100 123 107 136 99 129 76 135
100-109 89 109 95 121 115 149 83 147
110-119 44 54 69 88 112 145 77 136
120-129 36 44 65 83 76 99 76 135
130-139 17 21 33 42 68 88 47 83
140-149 18 23 26 34 28 50
1
1
150-159
3
7
1 3 9
4
1
2 7
160-169
170-179
1
1
Total 815 787 770 565
tend the inference back ages eleven

to to
one since there

to
is
,
no evidence that mortality from one twelve selective

is
respect altitude intellect any considerable ex-

in
of
of
to
tent or that the environment acts during those years

to
,
reduce and counteract tendencies multimodality skew-

to
ness and other departures from Form A. With little less

a
,
may back the germ cells and

to
assurance we extend
it
,
assert that , to a close approximation , the original capacities

of white children in the United States to manifest given
altitudes of intellect are distributed in a surface that is ap-
proximately unimodal , symmetrical , and of Form A.
THE FORM OF DISTRIBUTION IN ADULTS

Extending the inference to later ages is a very different
matter . If
the distribution is " normal " at 14 , it may still
become skewed at 24. This would happen if the gains made
by those of different degrees of intellect at 14 differed in
certain ways and by certain amounts . For example , sup-
pose that the altitude of intellect of fourteen - year - olds is
distributed as shown in column I
of Table 104 , and that
from fourteen to twenty - four those individuals of abilities
1 , 2 , 3 , 4 , and 5 gain 0, while those of abilities 6 to 19 gain as
shown below :
Ability Gain
6 0 to .15
7 .15 to .35
8 .35 to .55
9 .55 to 1.3
10 1.3 to 2.0
11 2 to 5
12 5 to 10
13 10 to 16
14 16 to 40
15 40 to 80
16 80 to 150
17 150 to 300
18 300 to 600
19 600 to 1000
The distribution at age twenty - four would then have its

low extreme at 1 as before , its mode and median at about
11 , and an enormous skew running up to about 1,000 . To
take a much less extreme state of affairs which might be
real , suppose the condition at fourteen to be as in columns
II and III
of Table 104 , and the gains to be as shown in
column IV . Then the condition at twenty - four will be as
shown in column V with a clear skew .
We also have evidence that a positive relation of gain

to ability exists in the case of the ages above fourteen ,
though we do not know its exact nature or amount . Imbe
ciles notoriously gain very little . Thorndike has shown
[ '23 ] that the sort of pupil who attends high school gains
up to eighteen at least , in the ability measured by stock
intelligence tests , and that the white pupils gain much more
than the colored pupils .
TABLE 103 .
HAGGERTY DELTA 2. DATA FOR SURFACE OF FREQUENCY IN EQUAL UNITS .
Original
Abscissa Ordinate Heights to Make the Areas
Length in Equal to the Corresponding Per
Interval
Equal Units mille Entries of Tables
Age 11 Age 12 Age 13 Age 14
0-9 12.00 (Est .) 5 4 2

10-19 12.00 ( Est . ) 6 8 4 4
20-29 10.70 9 7 7 3
30 9.50 51 29 15 6
40 9.33 103 47 23 20
50 10.85 90 73 31 21
60 10.55 116 89 59 54
70 8.62 151 136 99 77
80 9.48 137 124 120 123
90 9.15 134 149 141 148
100 10.54 103 115 141 139
110 10.01 54 88 145 136
120 12.06 36 69 82 112
130 13.01 16 32 68 64
140 19.86 0.5 12 17 25
150 20.88 0.5 2 4 3
The differential gain could be caused by several differ

ent factors . Inner mental growth is less in amount in the
dull at all ages ; it may , and probably does , slow up and
approach zero earlier in the dull . Insofar as ability with
intellectual tasks is due to environment and training , the
expectation will be that each added acquisition will be a
stimulus to others and an aid in acquiring them . So learn
ing to read commonly leads to the acquisition of a wider
vocabulary and a better score in opposite tests and comple
tion tests than would have been attained by oral intercourse

alone . The more intellectual the individual is , also , the
more will he give his free time to intellectual pursuits .
Finally , vocational selection is such that the more intellec-
TABLE 104 .
THE EFFECT OF CORRELATION BETWEEN STATUS AND GAIN WHEN GAIN IN-
CREASES IN A GEOMETRIC RATIO .
I II III IV V
Status Frequency at 14 Gain Frequency at 24
Grouped Grouped 14 to 24
by 1's by 3's
.02 .1+ 0.2

.2
012
.19 .125
1.14 .156
3
4.85 21.5 .195 21.5

4
15.50 .244
38.76 .305
77.52 242 .381 242
9 5678
125.97 .477
167.96 .596
10 184.76 521 .745 510
11 167.96 .931
12 125.97 1.16
13 77.52 242 1.45 214
14 38.76 1.82
15 15.50 2.27
16 4.85 21.5 2.84 53
17 1.14 3.55
18 .19 4.44
19 .02 .2 5.55
6
20
21
22 0.2
23
24
25 0.02
tual individuals continue school and engage

in
clerical
in
and professional work that involves intellectual activities

,
while the dull leave school for labor which requires little
thought and sometimes does not even permit
it
,
.
There may of course be a marked increase of gain for

those of higher abilities without producing skewness . If
gas + b, no skewness will be produced , no matter how
steep the relation line may be . The variability will in
crease , but the form will still be Form A , as shown in Table
TABLE 105 .
THE EFFECT OF CORRELATION BETWEEN STATUS AND GAIN WHEN GAS + B.
Status Frequency Gain Frequency

at 14 14 to 24 at 24
8 6 4 2
1
10
45
1
12345
120
5 210 10
252 12 10
9 8 7 6
210 14
120 16
45 18 45
22220
10 10
11
1
12 120
13
14
15 210
16
17
18 252
19
20
21 210
22
23
24 120
25
26
27 45
28
29
30 10
31
32
33
1
105 , where the abilities 1 , 2 , 3, 4 , 5 , 6 , etc. , have gains of 2 ,

4, 6, 8, 10 , 12 , etc.
The causes which influence differences in gains in intel-
lect up to about fourteen do seem to produce them in a
rough proportion to the differences in ability , so that the
form does remain that of Form A.
The only data that we have found for measuring the
form of distribution of anything approximating a random
sampling for any age above fifteen are the well -known
Army records with Alpha , Beta , and Examination a . We
have no satisfactory means of determining the value of
Beta scores in a scale of equal units . So we limit our in-
quiry to Alpha and Examination a.
50 100
FIG . 32. The form of distribution of the scores of recruits in Army Alpha
transmuted into a scale with equal units .
Using the equal - unit values for Alpha and a derived in

Chapter VII , and proceeding as in the case of the National ,
Otis , and Haggerty scores for children , we obtain the re-
sults shown in Fig . 32 for of the
51,620 native - born whites
draft [ Data from the National Academy of Sciences Mem-
oirs , '21 , p . 764 ] .
The equal - unit values of the interval from 0 to 20 in
Army Alpha and from 0 to 30 in Examination a are esti-
mates from exceedingly scant data .
The lower end of Fig . 32 would be extended if the illit

erates who were exempt from Alpha had been included . It
would have been extended still further if the men rejected
for dullness by the examining boards had been included .
The upper end would be extended if the officers had been
included .
Using the equal - unit values for Examination a in the

same manner , we obtain the results shown in Fig . 33 for
100 200
FIG . 33. The form of distribution of the scores of recruits in Army Examina
tion a , transmuted into a scale with equal units .
63,647 enlisted men in four camps . [ Data from the Mem

oirs , '21 , p . 492. ] The same considerations concerning the
inclusion of men rejected for dullness and of officers apply
as applied in the case of Alpha . In these four camps , 13.9 %
had been excluded from examination as illiterate .
It
is difficult to reach any secure conclusion from the
facts of Fig . 32 and Fig. 33 , except that there is no evidence
of negative skewness . From Alpha alone in the general
draft it would appear that even after generous allowance
for the dullness of the illiterates and others who were ex
cluded from examination , the distribution was skewed posi
tively , i.e. , toward the high end . With Examination a in
the four camps , however , the skewness of the original scores
disappears when the values in equal units are used .
We cannot even estimate with surety what the distribu

tion of 51,620 of the native - born whites would have been if
they had been measured with Examination a , or what the
distribution of the 63,647 enlisted men in the four camps
would have been if they had been measured with Alpha .
That is, we cannot decide how far the difference between
Fig . 32 and Fig . 33 is caused by the tests used and how far
it is caused by the groups tested .
On the whole , we may provisionally regard the sort of
3
intellect measured by Alpha and a as distributed in the

adult native - born white population of the United States
with some positive skewness . We may provisionally as
sign, as the cause of the change from the symmetry and
normality found in children , a differential gain from the
age of fourteen to twenty and beyond , whereby some indi
viduals increase these abilities very greatly , whereas others
increase them little or not at all . This should be only pro
visional . The whole matter of adult intellect should some
time be studied with the care which it deserves .
For the ages from 14 up to 17 or 18 , we may assume
symmetry and normality without much probability of more
than a small error . Or , we may have a slightly greater
prospect of correctness if we allow for a very little positive
skewness , increasing year by year .
3 Certain facts of the distribution of men in occupations , of the distribution
of wages , of the distribution of schooling , and the like rather favor the sup
position that adult intellect is distributed with positive skewness .
CHAPTER IX
A SCALE FOR MEASURING ALTITUDE OF INTELLECT

It was not a part of our original plan to make an actual
scale for measuring intellectual difficulty , but only to find
methods whereby this could be done . We have proved that
the form of distribution of altitude of intellect in grade
populations from Grade 6 to the first year of college can be
known , so that the degree of intellectual difficulty of a com
posite task which is truly intellectual can be measured by
the percentage of successes in such a grade population .
We have also shown that the form of distribution of intel
lect of an age population 10 to 14 is approximately of Form
A, that of the normal probability surface , so that the same
procedure can be followed in one of these age groups . It
is highly probable that it can be followed in lower age
groups .
Although we did not plan for scaling the difficulty of
actual tasks and are not able to do it precisely with the time
and facilities at our disposal , it seems best to make a begin
ning , if only to illustrate the workings of the principles and
techniques involved in an actual case .
The work on this scale may best be considered in two
parts , that which evaluates the differences in difficulty of
Composites I to Q, and that which evaluates the differences
in difficulty of Composites A to I. The latter was done
primarily to put the values for , , K
I J Q in relation to
the absolute zero for which
, purpose chance errors in the
I
determinations of B - A , C - B , D- C ฀ ฀ ฀· -H are of minor
importance , since they tend to equalize one another . These
lower intervals are less precisely determined than those
I
from to Q ; and we report them and their derivations sepa
rately in the latter part of the chapter .
294
A SCALE FOR MEASURING ALTITUDE OF INTELLECT 295
THE DIFFICULTY OF COMPOSITES I , J , K, L, M, N, O, P , AND Q
We present first the measurement of differences in dif-

ficulty between tasks and , between I J
and K , between K J
and L and so on with M N O P and Q. The facts at our
, , , , ,
disposal for the measurement of differences in difficulty

amongst these composite tasks are the percentages correct
in various groups as shown in Table 106. Group 5 refers
to the 147 pupils measured at the end of Grade 5 and at the
beginning of Grade 6 with composites , , and K. Group I J
pupils of Grade 9 who are measured
J
91 refers to the 246
with composites , , I K, L , and M. Group 911 refers to the
192 pupils of Grade 9 who are measured with composites
TABLE 106 .
PERCENTS OF VARIOUS GROUPS SUCCEEDING WITH 20 OR MORE SINGLE TASKS
OF CAVD 40 -COMPOSITES I TO Q.
Composite Percents Succeeding

52 91 911 13 17
I
J
91.2 99.6
29.1 89.4
K 11.5 61.4 47.0
L 32.9 16.3
M 5.3 7.2
N 1.1 81.5 95.4
48.1 77.1
Р 27.5 56.7
3.7 22.9
K, L , M, and N. Group 13 refers to the 189 candidates for

entrance to college who were measured with composites N,
O , P , and Q. Group 17 refers to the 240 college graduates
who were measured with N , O , P , and Q.
Ifwe know the form of distribution of a group and the
percent of the group succeeding with a task , it requires
only straight - forward mensuration to find the point on the
base line corresponding to that percent , and the distance of
that point plus or minus from the median ( or mode , or
other point of reference defined by the distribution of the
21
group ) in terms of the mean square variation ( or other de-

fined measure of variability ) of the group in whatever abil-
ity is measured by that task.
The form of distribution is taken as normal for each
of the grade groups , 51 , 91 , 9II , and 13 , in consequence of
the facts outlined in Chapter II
and presented in detail in
Appendix III
. The form of distribution of Group 17 , which
was composed of first - year law - school students , all college
graduates , was determined by a special investigation which
is reported in Appendix VI . The same has been done for
TABLE 107 .
THE DIFFICULTY OF COMPOSITES I

TO Q IN VARIOUS GROUPS EXPRESSED IN
EACH CASE AS A DEVIATION FROM THE DIFFICULTY FOR THE MEDIAN OF
THAT GROUP , IN TERMS OF THE σ OF THAT GROUP IN THE ABILITY
MEASURED BY SUCCESS WITH THE COMPOSITE IN QUESTION . IS -
EASIER , IS HARDER .
Composite Difficulty
512 91 911 13 17
I - 1.35 - 2.65
JK + .55 -- 1.25
- .29
+ 1.20 + .08
L + .44 + .98
M +1.62 +1.46
N - - 1.862
-
+ 2.29 .897
-
O + .048 .714
Р + .598 .153
Q + 1.787 + .738
the form of distribution of certain groups used , later in this

chapter , namely , for the 180 adult imbeciles of mental age
from 2 to 5 years , for the 100 adult feeble - minded of men-
tal age near 61 , for the group of 50 feeble -minded at or near
mental age 8 , for the group of 101 dull pupils 13 years old
or over, in special classes in New York City , for the popu-
lation of Grade 4 ( second half year ) and for the population
of Grade 5. The evidence and argument in all these cases
appear in Appendix VI.
Table 107 gives the difficulty of various 40 - composites
in various groups , expressed in each case as a distance
from the difficulty which 50 % of that group can succeed

with and in terms of ot , ( the mean square deviation of
whatever ability is measured by that composite in that
group ) . The next procedure in constructing a scale of dif-
ficulty is to make all these different measurements of dif-
ficulty commensurate and put them all into relation to the
same point of reference . This is a complicated procedure
involving the following steps :
Each measurement in σt , for a given group is to be
turned into a measurement in o , for that group , o , being the
mean square variation of the group in altitude of intellect
perfectly measured in truly equal units .
Each measurement in the o , of a certain group must
be made commensurable with measurements in the Οι of
any other group , by finding the comparative magnitudes
of o₁ of the 240 graduates , o₁ of the 189 college entrants ,
σ of the 246 pupils in Grade 91 , o , of the 192 pupils in
Grade 911 , and so on . All of the different o , values may
then be multiplied or divided by numbers so that all will be
expressed in the same units . We shall use the mean square
deviation of pupils in Grade 9 as our unit for this purpose .
The measurements , now in units of σ19 , must be ex-
pressed , not as distances plus or minus from the CAVD
difficulty for the median now of one group now of another ,
but all from some common point of reference such as the
median for Grade 9.
ESTIMATING σ₁ FROM σt₁

We turn the measurements in σt , into terms of o , by
σ, = ot , Vrt ,t, or by using o₁ = σt , rt , t .
using o₁
The self - correlation rtt₂ is , of course , for a 40 - composite
with another 40 - composite of equal difficulty , not for an in-
finitely extensive set of tasks of a certain difficulty with
another equally extensive set . Also rtt2 is the correlation
for the specific group of restricted range which is being
used , not the correlation for a group of wide range , such
as all persons of age 20 .
For
precise determinations of Vrt , t , or of rt , we need
measurements with more extensive groups and alternate
forms of our 40 - composite tasks . With the material at our
disposal we can hope only for approximate results .
We measure or infer rtit2 separately for each composite
with each group . We may , however , wisely modify the esti
mate for each composite with each group in view of the
facts concerning rt1t2 for the same composite in other
groups , or for other neighboring composites in the same
group .
Consider , for example , the 40 - composites K , L , and
M. The correlations of each of these with a 40- composite
of different content but similar difficulty estimated by
2r20, 20
1 + 120, 20 are as shown below
according to the group and
kind of coefficient computed .

K L M
Group 246 ( Sheppard ) .74+
.86
.68
.70
.68
.86
/2
Group 246 ( Pearson ) .65+ }
}
}
Group 192 ( Sheppard ) .80 + } .77 .86

.73 + }
~ .872 .78
.75
%
Group 192 ( Pearson ) .89 }
The correlations of each them with 40 composite of
of
a
-
different content but similar difficulty may also be esti

mated by adding .03¹ their average correlations with
to
their nearest neighbor composites with some justifica

or
(
,
tion by adding .02 or .01 even Using .03 we have

or
0
,
,
).
the results shown below

.
K L M
Group 246 Sheppard .70
.68.68
)
(
.69
Group 246 .70.68

%
Pearson .69
}
)
(
Group 192 Sheppard .50

.57 ฀ .752 .75
%
.70.67
~
)
(
Group 192 Pearson .64 .75

}
)
(
Combining the two sorts estimates we have

of
K L
,
M
Group 246 .70 .77 .68
Group 192 .67 .771 .75
See Appendix IV for the derivation and justification of this allowance

1
for remoteness
.
Moreover , we may consider that chance played some

part in making the self - correlation of L higher than the
other two ; and so lower its rt₁t2 and raise theirs somewhat
to balance . Similarly we may consider that chance played
some part in making these rt , t , ' s higher in the 192 group
than in the 246 group , and allow somewhat for that . Thus
we may replace the last set of figures by
K L M
Group 246 .72 .76 .70
Group 192 .67 .741 .75
in which slight smoothing by these allowances is made .

In Table is collected all the information concerning
108
the rt , t's for each 40 - composite in each group . I and II
refer to the two methods of determining rt1t2 . In we use I
the correlation between the two halves of a 40 - composite
obtained by taking 5C + 5A + 5V + 5D at random , the sec
ond half being composed of the remaining 5C + 5A + 5V +
2r20, 20
5D ; and estimate · That is , Tt1t2 =
by
1+
140
40
20
20
,
2120
In II we
20
use the obtained correlation between

,
1+ T20
·
20
,
the 40 composite question and its nearest neighbor com

in
-
posites adding .03

,
The correlations for composites and group

in
N
O
P
Q
,
17 under By other data were obtained follows

as
A
"
"
composite almost identical with was correlated with an

N
other of very closely equal difficulty giving .72 com

=
A
r
,
posite almost identical with was correlated with com

O
posite very closely equal difficulty giving .75 The

=
of
r
,
composite almost identical with was also correlated with

N
composite almost identical with giving .73 The

=
O
r
a
composite almost identical with was correlated with

Q
O
giving From these correlations allowing for

=
.73 .03
+
r
,
.
The results by method are general higher The differences Method

in
II
2
(
.
Method are .09 .12 .172 .002 .20 .20 .00 .05 -.07 .05 .05
%
I-
0
)
-.06 averaging .063

,
.
300
.
TABLE 108
-
-
AS ESTIMATED FROM CORRELATIONS BETWEEN NUMBER OF SINGLE TASKS CORRECT IN ONE HALF OF 40 COMPOSITE AND NUMBER OF
Tt1t2
;
A
SINGLE TASKS CORRECT IN THE OTHER HALF AND ALSO AS ESTIMATED FROM CORRELATIONS BETWEEN NUMBER CORRECT IN
A
.
-
-
40 COMPOSITE AND NUMBER CORRECT IN NEIGHBORING 40 COMPOSITE
40 Composite Method II Average
2r By
= 20,20
=
other
40
.03 +140with nearest
Tt1t2
20,20 data
THE MEASUREMENT
91 911 13 17 52 91 911 13 17 17 52 91 9II 13 17

Group 5½
.71 .78 .62 .78 .662
I฀
.782 .78 .622 .78 .722
К .70 .70 .57 .70 .67
.86 .8712 .682 .672 .772 772
.682 .752 .68 .75 .68 .75
.75 .66 .692 .73 .86 .74 .72 .6912 .80
%
.81 .76 .862 .75 % .78 .81

OF INTELLIGENCE
Р .81 .76 .792 .782 .792
TUKLMNOPQ
฀
.66 .72 .72 .76 .69 .74
one step of remoteness , we have the self- correlation of N

as .72 or .76 , averaging .74 ; that of O as .75 or .76 , aver-
aging .75 ; and that of Q as .76 .
91 and 911 differ almost nil in the general magnitude of
rtit2 for the three 40 - composites used with both of these
groups , the average difference ( I-II
) being -.013 with a
TABLE 109 .
VALUES OF tit2 DERIVED FROM TABLE 108 , AND THE VALUES OF Vrt₁₂ USED
TO OBTAIN TABLE 110 FROM TABLE 107 .
Tt1t2 VI₁₂12
512 91 911 13 17 52 91 911 13 17
I
J .662
.78 .883 .815
.78 .722 .883 .851
K .682 .682 .828 .828
L .772 772 .880 .880
M .70 .73 .837 .854
N .72 .722 77 .8482 .8512 .8772
O .77 .822 .8772 .908
P .77 .81 .8772 .900
Q .69 .74 .831 .860
mean square error of .033 , three times as great as the

difference . So we shall probably be nearer the truth by
using .68 and .68 in place of the .70 and .67 , and .70 and
.73 in place of the .68 and .75 .
In general rtit2 is .04 higher in 17 than in 13 , and the
use of this fact to smooth out the irregularities in the values
for N , O , P , and Q will probably be an improvement . Thus ,
columns 3 and 4 below are probably truer than column 1
and 2. The totals for each group and for each composite
are unaltered by the amendments .
From the table Amended

1 2 3 4
N 691 80 721 77
781 81 77 821
Р 781 791 77 81
69 74 69 74
We make the amendments noted in the last two para-

graphs and so use the rt , t , ' s listed in Table 109 in estimat-
ing the difficulty in terms of o, for each composite in each
group . The results are shown in Table 110 .
TABLE 110 .
THE INTELLECTUAL DIFFICULTY OF COMPOSITES I

TO Q IN GROUPS 52 , 91 , 911 ,
13 AND 17. EXPRESSED IN TERMS OF σ15 , O DI, Ơi 911901 18 OR σ1 179
AS DERIVED BY THE USE OF TABLE 109 .
Com-
Difficulty
posite
In 01 5 In 191 In σ1 911 In 01 18 In σ1 11
17
I - 1.53 - 3.25
J + .62 - 1.47
K 1.36 .35 .10
+
+
L .50 +1.11
+ +
1.94 +1.71
MANNO
2.70 1.054 2.120
- -
+
-
-
.055 .786
+ + +
Р .681 .170
2.150 .858
FO
+
♦
Estimating the by oot by

,,
we obtain
rt
rt
o
s
,
,
It₁₁
,
'
,,
rt₁1 which the obtained correlation be-

Vri
in
rt
is
i',,
tween the 40 composite question and the summation score

in
-
in long CAVD series and

,,
the self correlation

of
is
a
r
,
this summation score In certain cases we have to estimate

.
but the error of the estimate small and its effect

is
is
,,
ri
1
,
reduced since only the square root of

,,
used The
is
ri
values of and r₁₁i used are those used for another pur-
,,
rt
pose in Appendix V. The results the computations are

of
shown in Table 111

.
Using the estimates Table 111 we obtain the

of
of
of rt
,
,
estimates the difficulty composite for each group

of
each
terms of the of that group which are presented
in
in
σ
,
For grade population the empirical values r₁₁12 vary from .91
to
.95
of
a
3
In group 5½ and group the summation score from

17
is
where we estimate
,
very long series we use .95

so
a
.
Table These differ on the average from those of Table

112.
110 , as shown below , the median difference regardless of
-
signs being .03 and the average difference .05 .
σ (by Vrtit2 ) σi (by rt₁1 )

-.15 to -.06 4
-.05 to +.04 12
+.05 to +.14 3
+.15 to +.24 1
TABLE 111 .
VALUES OF It ,
ESTIMATED FROM CORRELATIONS BETWEEN NUMBER OF SINGLE
TASKS CORRECT IN A 40 -COMPOSITE AND NUMBER CORRECT IN A LONG
CAVD SERIES .
Composite 52 91 911 13 17
I .933 .759
.882 .907
J
.854 .819
.944 .896
.849 .922
.819 .824 .872
LUKLMNOPQ
.917 .944
Р .948 .913
.790 .882
♦
Except the case of composite group does

91
in
in
it
I
not matter much whether we use the estimates of Table 110

or those of Table 112 or averages of the two We have
.
averaged each pair of determinations with the results

shown in Table 113 which are used as the values in all
σ
that follows
.
EXPRESSING THE σ₁ OF EACH GROUP IN TERMS OF COMMON

A
UNIT
We make the two groups and commensurate
of
B
s
σ,
,
'
by finding the difference difficulty between two tasks

in
in
terms of and terms of OIB provided the two groups

in
σ
overlap sufficiently Thus we find the case of the group

in
,
,
.
( 17 ) of
240 college graduates and the group ( 13 ) of high
-
school graduates , that :
P-
Composite Composite N = 1.120 , 13 and 1.3601 17.
Composite Composite O = .61σ₁ and .600 ,
Q-
13 17.
Composite Composite P = 1.55σ , 13 and 1.0201 17.
01 13 = 1.210₁ 17 or .9901 17 or .660 17, according to the

successive pair of composites used . we take the most If
remote composites which include all the data , Q and N , we
have 3.280 132.9801 17, whereby σ1 13.9101 17.
TABLE 112 .
THE INTELLECTUAL DIFFICULTY OF COMPOSITES I
TO Q IN TERMS OF σ sả
O191, ETC .; AS DERIVED BY THE USE OF TABLE 111 .
In σist In σ1 91 In σ1911 In σ1 18 In σ1 17
I
J
1.45 3.49
- -
-
.59 1.38
+
+1.29 - .34 .10

.47 +1.0912
+
+1.91 +1.58
2.80 1.09 2.132
-
+
-
KLMNOPQ
.05 .76
+ + +
.63 .17
Р
2.26 .84
+
฀
In thesame way we find the Group of

91
the case
of
in
,
,
pupils Grade and the Group 9II pupils

in
of
246 192
in
9
Grade that
9,
Composite
911
Composite ――
.840 and 1.000
= =
L K
or 91
ML
Composite Composite
911
91 and
-
1.440 .5501
,
according the pair

911
of
com
911
911.1901 .380
or
to
01
posites used we take the most remote pair which

If
in
.
clude all the data and we have 2.2801

M
1.5501 9119
91
=
,
,
whereby 91.6801
911
σ
In the same way we find with Group and Group

91
5
,
,
,
that
:
K-
Composite Composite
――――
2.10σ and
= =
1.9401
91
51
J
JI
5
,
Composite Composite .720 and 1.0901

91
5
,
.
01 51
= .920 or 1.510 , according to the pair of com-
If we use K and I
91 91
posites used . , which include all the data ,
we have 2.820 51 = 3.020 , 91, whereby σ , 51

5,
= 1.070 , 91.
For precise work in scale construction , the groups

should be large and close enough together to have a consid-
erable overlapping . The measurement of the σ , of any one
group in terms of the o , of any other group may then be de-
termined with as small an error as is desired .
Our groups are obviously not large enough , since there
are so great differences between the estimates of the com-
TABLE 113 .
THE INTELLECTUAL DIFFICULTY OF COMPOSITES TO Q. AVERAGES I OF THE

DETERMINATIONS OF TABLE 110 AND TABLE 112 .
In σ15 In σ1 91 In σ1 911 In σ 13 In σ1 17
I - 1.49 - 3.37
J + .61 - 1.43
K + 1.33 - .35 + .10
L + .49 + 1.10
M +1.93 +1.65
N +2.75 - 1.07 - 2.13
O -
-
+ .05 .77
P + .66 .17
Q + 2.21 + .85
parative variabilities according to the composites which we

use . There is particular risk in using the estimates of com-
parative variabilities in different groups which depend
upon a composite that is very easy or one that is very hard
for the group . In the case of the very easy composites
carelessness may play a part that affects the results . In
the case of the composites which are very difficult for a
group , lack of effort and persistence and interest may be a
disturbing factor ; and it is possible that , in spite of care
taken to give what seemed to be abundant time , certain in-
dividuals may not have exhausted their abilities for lack of
sufficient time . The eccentricity of the results with Com-
.306 THE MEASUREMENT OF INTELLIGENCE
posite M in Group may be due to this fact . In general ,

91
Group 91 was superior to Group 9II and the reversal to
notable inferiority with Composite M may be explainable
by the fact that this was the hardest composite taken . It
was not truly the last in point of time , since all the C's were
done in one division of the examination , all the A's in an-
other division of it , all the V's in another division of it ,
and all the D's in another division of it .
In
cases where the material is not notably richer than
this of ours and in cases where the groups are spaced so
far apart that there is little or no overlapping , valuable aid
may be derived from a general consideration of the com-
parative variability of groups similar in school grade or
other indication of intellect to the particular groups which
are used in scaling the difficulty of the composite tasks .
Moreover , facts concerning the comparative variability of
grade populations are valuable as a check on even the best
determinations made by using two or more composites with
two or more groups . Consequently we have made a rather
exhaustive study of the variability of grade populations
from 6 through 13 , using all the data that we could discover
which had sufficiently large populations to make the deter-
minations of variability reasonably precise .
In order to discover the relative variability of different
grade populations from 6 through 13 , if each individual
were measured in truly equal units , we may proceed in
either one of two ways :
We may argue after the fashion of the argument in Ap-
pendix III
that inequalities in the face - value units will neu-
tralize each other so that the general average result from
many tests , each with its own sort of inequality , will be near
the truth . In this case , we simply take the sigmas by the
original scoring for these different grades and get their
general drift . Dr. Bregman has done this for all the ma-
terial available with populations large enough to give re-
liable sampling of the grades . The results are shown in
Table 115 and in more detail in Table 114 .
The second method is to transmute the face - value mea-

sures for such tests as Army Alpha , National A , Otis Ad-
vanced , etc. , into terms of equal units before computing the
sigmas . The results of the investigations reported in
Chapter VII enable us to do this , since in that chapter we
determined the value in equal units of each interval of the
TABLE 114 .
DATA FOR COMPUTING RELATIVE VARIABILITIES OF DIFFERENT GRADES IN INTEL-

LECT ; AND FOR COMPUTING DISTANCES BETWEEN MEDIANS
OF DIFFERENT GRADES IN INTELLECT .
Original median refers to the median by the standard method of scoring ;

corrected median refers to the median by a scale in equal units ; original σ
refers to the mean square deviation by the standard method of scoring ; cor-
rected o refers to the mean square deviation by a scale in equal units .
Median σ
Grade Number Original Corrected Original Corrected
Army Alpha *
6 281 54.9 55.6 18.4 19.1
9 1721 97.94 97.94 24.0 24.2
10 1223 24.0
11 977 23.8
12 1387 125.39 125.39 24.24 24.8
12 766 128.04 128.04 24.13 24.9
Coll . 1 2545 128.50 128.50 28.20 29.2
66 1 158.5
400 157.8 19.99 23.3
Army Examination A*
6 742 139.8 139.8 36.9 38.94
7 685 158.6 158.6 39.2 40.90
8 630 186.1 186.1 43.04 43.39
9 311 204.36 204.36 45.89 45.53
12 53 276 274 36 36
Coll . 1 701 267.33 265.33 40.63 39.25
National A **
6 1668 111.9 111.9 22.8 21.8
9 494 141.75 140.85 16.8 16.5
* All computations exact .

TABLE 114 ( Continued ) .
Median σ
Otis *
5952 86.8 87.2 24.3
67 3896 96.98 97.1 24.4

8 4598 111.93 111.4 25.08
9 3627 125.04 123.8 24.62
12 1226 151.83 149.7 24.06
Haggerty *
6 916 91.4 91.3 20.4 20.7
67 737 105.07 105.2 20.2
8 689 113.9 113.7 19.46
9 473 113.7 113.5 17.5 19.54
9 1995 116.5 116.4 18.2 23.25
12 668 135.83 139.3 15.31 22.4
1. E. R. Form A **
6 379 83.9 81.0 32.41
9 3231 173.4 173.4 42.9
10 1935 191.1 191.1 40.3
11 1533 202.6 202.6 42.4
12 972 219.81 219.8 44.99
12 1666 227.79 227.5 45.85
I. E. R. Form B **
10 1656 209.0 43.55
11 1453 219.7 44.0
12 1207 229.9 44.7
Terman Group Test **

9 1438 102.16 102.16 32.0
12 4886 144.55 142.55 32.61
* All computations exact .
** The sigmas in equal units will vary inappreciably from the sigmas by the
original are not computed .
scale and
*** The effect of inequalities in the units will be almost identical for Grade
9 and for Grade 12 ; hence the relative values of the sigmas will not be influ-
enced thereby . Consequently the sigmas for values in equal units have not been
computed.
original scale for Army Alpha , National , Otis , Haggerty ,

Army a , Terman Group Test , and several others . We have
made these computations with results as shown in Table
114 .
TABLE 114 (Continued ) .
Median σ
Brown University *
12 3333 45.69 46.2 11.59
Coll . 1 2118 56.62 56.3 11.11
Myers Mental Measure **

6 724 46.3 46.3 13.1 12.6
7 696 49.61 49.8 14.65
8 950 54.15 54.55 13.72
9 311 57.1 57.5 13.05 13.75
Pintner Non - Language **

6 1237 316.7 313.7 86.7 86.5
7 755 339.0 339.0 73.18
8 530 379.6 381.6 73.24
9 258 400.6 403.0 75.0 78.5
Pressey Cross - Out ***

1057 51.18 10.30
69
7 998 56.10 10.30
8 725 63.12 10.0
9 303 72.5 10.0
Trabue Completion **
6 1454 21.8 5.5
7 1456 25.39 5.67
8 1740 27.61 6.29
9 273 30.05 5.9
* The inequalities of units in the scale are such as balance one another and
leave the relative values of the sigmas by the original units undisturbed . Con-
sequently new values are not computed .
** The sigmas according to a scale with equal units are computed by finding
½ the distance required to exclude 15.87 % at each extreme .
*** Scores in equal units have not been determined .
The data which we have used to measure comparative

variabilities are the same as those which will be used
later to measure the differences between the medians of
various grade groups in intellect . We present them in
Table 114 classified according to the examination used .
In connection with each examination we record the results
for Grades 6 , 9 , 12 , and 13 ( or first year of college ) and
occasionally for other grades or groups . We report the
number of individuals ; the median score , taking the
TABLE 114 ( Concluded ) .
Median σ
Illinois Examination *
6 588 75.52 17.01
9 380 101.4 18.5
Thorndike Intelligence Examination , Part 1 *

12 1527 91.4 18.1
Coll . 1 166 101.7 17.6
66 1 466 108.4 17.0
66 1 319 107.1 18.5
Weighted average
66
1 ( weights 1 , 2 and 106.5 17.7
2)
* Scores in equal units have not been determined .
units at their face value ; the median score in a scale

with equal units ; the mean square deviation , taking the
units at their face value ; the mean square deviation ,
using a scale with equal units . In the latter case the
sigmas have been computed exactly , where it was possible ,
but in many cases we have had to resort to approximations .
In cases where the scale with equal units was so closely
similar to the original scale that little , if any , difference
would be made in the mean square deviation , we have used
the original figures . Notes are appended to Table 114 de
scriptive of what was done in this regard in each case .
From the facts of Table 114 are computed the ratios of

Table 115 .
From the facts of Table 115 we may conclude that the
forces of selection and gradation which determine the vari-
ability of grade populations result in a slight increase from
Grade 6 to 4 percent ( giving
9 , which we may estimate as
twice as much weight to the results from equal - unit scaling
TABLE 115 .
THE RELATIVE VARIABILITY OF DIFFERENT GRADE POPULATIONS .
Examination Using the Original Using Scales with Equal

Scale Units Units
016 0112 0113 0113 016 O112 0113 0113
019 019 0112 019 019 019 0112
Army Alpha . .77 1.01 1.00 1.00 .79 1.03 1.09 1.06
Army a.. .80 .782 .882 1.13 .8512 .79 .86 1.09
National A 1.36 1.32
Otis Adv .99 .98 .99 .98
...
Haggerty 1.13 .85 .93 1.00

I.E.R. Sel Gen. .76 1.06 .76 1.06
.
Terman Group....... 1.02 1.02

Brown Univ......... .96 .96
Myers Mental...... 1.00 .92
Pintner 1.16 1.10
Th Part N .97
I
.
Trabue Comp ....... .93

Illinois .92
Pressey 1.03
Median .99 .992 .94 .9812 .92 1.01 .972 1.06

Average .99 .95 .94 1.012 .96 .98 .972 1.04
as those from the original scores From there

12
to
to
9
).
Oi12
little or no change The medians for the ratios aver-
is
019
age 1.001 The .78 and .79 Army which make the aver-
of
a
.
ages lower .95 and .98 are from very small group
53
of
a
(
which should be given very little weight This group was

.
used because enriched somewhat our very scanty ma-

it
0113
terial on the comparison We may then estimate the
0112
.
22
variabilities of Grades 6 , 9 , and 12 as 96 , 100 , and 100. Com-

paring Grade 13 with both Grade 9 and Grade 12 , we find
for the original - scale units a median

σ113
of .98 and
or σ19
0112
an average of .99 ; for the scales in equal units , there is a

median of 1.06 and an average of 1.01 . The best estimate ,
in view of the fact that the .88 and .86 by Army a deserve
less weight than the other determinations , seems to be
about 102. We then have 96 , 100 , 100 , and 102 as the rela-
tive variabilities of Grades 6 , 9 , 12 , and 13 .
These general facts may be used to correct the eccentric
and unreliable determinations from the composites them-
selves ( see page 304 f. ) . The use of the entire stretch of
overlapping gave σ15 as 1.070191 , for our particular group ,
but in general σ15 may be expected to be about .96019 . We
know of no facts which make it probable that our groups 5
and 91 differ from Grades 5 and 9 in general in such a way
as to make a variation of 0151

up from .96 , any more prob-
019
able than a variation down . The scientific procedure would
be to apply the same examinations to these two particular
groups , and compute the variabilities in units of known
value , but this was not practicable . The best thing to be
done in the circumstances is to attach some reasonable
weights to the two lines of evidence , and so obtain a work-
ing estimate . Giving the general facts about Grades 6 and
9 a weight of 4 , and the particular facts from the composites
used in both groups a weight of 1, the ratio

0151
is .98.
σi91
The next matter to be cleared up is the comparative
L-K
σ191
variability of 91 and 9II . was 1.19 by and .38 by
σ1911
M -L . We shall disregard these determinations entirely

and treat the variability of 91 as equal to that of 911 , for
the following reason . These two groups were constituted
by a division of all the pupils in Grade 9 in a certain school
at random , so far as is known . There is nothing in their
summation scores to show that one is more variable than

the other . - -
The L K and M L determinations are enor
mously at variance , and so deserve very little weight .
Between 9 and 13 there is no overlapping , so that the
general facts of grade variability are the only means of
estimate . As has been stated , our group 13 is a group of
candidates for college entrance , not of actual freshmen .
They were , however , candidates already selected by certain
tests and were of intellect comparable to the freshman
groups reported in Table 114 , differing probably toward
less variability rather than toward more , if they differed at
0113
all in this respect . 1.02 or 1.00 is then suitable as the
019
ratio , so far as is known .

The last comparison to be considered is of group 13 and
group 17. The determinations from the composites taken
in common were :
O113
- 1.21 or .99 or .66 , with a median of
0117
.99 and an average of.95 . The use of the widest stretch

between composites gave .91 . The .66 and .91 and .95 are
probably too low , inasmuch as all depend on the + 2.21 for
composite Q in the 13 group . This is the most unreliable
of the eight determinations , and is probably too high . The
difference between the general level of ability of group 17
and that of group 13 is 1.06 by composite N , .82 by O , .83 by
P , and 1.56 by Q. The median .99 is the most probable esti
mate from the composites used in both groups . The gen
eral drift of the facts for Grades 6 , 9 , 12 , and 13 gives the
expectation that the variability in Grade 17 will be some
what but not much higher than that in Grades 12 or 13 , per
haps 1.04 or 1.05 times the variability of Grade 9 , giving a
Oi13
ratio for of about .95 .
0117
The records for Examination a with 136 college students

of Grades 14 , 15 , and 16 , and with 27 graduate students
show , however , decreases in variability much below that of
0113
Grade 9, making well above 1.05 . So the general con
Oi17
siderations can hardly be used to favor change from 1.00 in
13 17
either direction . On the whole .99 for or 1.01 for
17 13
is fairly well justified by both methods .

The values recommended for turning the various o₁'s
into σ's are then :
0151
.98
0191 or σ191:
0191 σ1911
0113
==1.02
σ191 or σ1911
Oi17
= 1.03
σ191 or σ1911
Nothing in the particular comparisons from the composites

themselves is inconsistent with these estimates . What has
been done is to use general considerations to locate ratios
within the limits of those which were reasonable in view of
the particular comparisons . Using them the measures of
Table 113 become those of Table 116 .
EXPRESSING THE MEASURES OF DIFFICULTY AS DISTANCES FROM A

COMMON POINT OF REFERENCE
The differences in difficulty, of composites , , and K I J

plus and minus from the median of group 5½ may be ex
pressed as differences from the median of group 91 , by find
ing the differences between the difficulty for the median of
group 5 and the difficulty for the median of group 91. This
may be found by using the composite tasks which were used
with both groups . Thus composite I is , by Table 116 , 1.52019
easier than the task which just 50 % of group 5 can master
and 3.37019 easier than the task which just 50 % of group 91
can master . By this determination , the difficulty of the
median task for 5 is 1.85 less than the difficulty of the
median task for 91. Using the facts of Table 116 for com
posites J
and K in the same manner , gives 2.05019 and
1.71019 . The average is 1.87019 ; the median is 1.85019 .
In the same manner , K is 35019 easier than the task
which just 50 % of group 91 can master and .10σ harder
than the task which just 50 % of group 9II can master . By
this determination , the difficulty of the median task for 91
is greater than the difficulty of the median task for
.45019
9II . Using the facts for composites L and M gives .6109
greater and .28019 less . The average of the three determi
nations is .26019 ; the median is .45019 .
Composite N is 2.750 , harder than the task at which
50% of 911 succeed , and 1.05019 easier than the task at which
50 % of group 13 succeed . So the difficulty of the median
task for group 13 is 3.8019 greater than that of the median
taskfor 9II .
Using N , O , P , and Q in similar manner , the difficulty of
the median task for Group 17 is found to be 1.02019 or
or .81019 , or 1.34101 , greater than the difficulty of the
.80019 ,
median task for Group 13. The average is .99019 ; the
median is .92019 .
Relating the difficulty of the median task for each group
to the difficulty of the median task for a group half-way
between 91 and 9II , we have :
Computed Computed
-
by average by medians
-
The median for 51 Median 91 + 911 -1.74 -1.621
--
The median for 91 - Median 91 + 9II + .13 + .22
The median for 911 Median 91 + 911 .13 .22
-
The median for 13 Median 91 + 9II +3.80 + 3.80
The median for 17 Median 91+ 911 +4.80 +4.72
The reasonableness of these estimates may be checked

by the facts for the difference between the median scores
in Grade 5 and Grade 9 and Grade 13 in intelligence ex
aminations in general , expressed in terms of the variability
of Grade 9 , or in some other unit of measure .
We have collected the available facts concerning the

median scores of Grade 6 , Grade 9 , Grade 12 , and the first-
year of college , in Army Alpha , Army Examination a , Na-
tional A, Otis Advanced , Haggerty , I. E. R. Sel . Rel . Gen.
Org . , Terman Group , the Brown University Examination ,
the Myers Mental Measure , the Pintner Non - Language
Test , the Trabue Completion , the Illinois Examination , and
the Pressey Cross - Out Test . They are reported in Table
114 ( on pages 307 to 310 , inclusive ) . For all save the last
three , we have computed what the differences between the
medians in question are by a scale of equal units . The re-
sults , both by the original scale and by the scale with equal
units , are shown in Table 117 .
TABLE 116 .
THE INTELLECTUAL DIFFICULTY OF COMPOSITE TASKS I TO Q IN TERMS OF σ19.
Difficulty
Composite By 5½ By 91 By 911 By 13 By 17
I - 1.46 - 3.37
J + .60 - 1.43
K +1.30 .35 + .10
L + .49 +1.10
1.93 +1.65
+
2.75 1.09 2.19

-
+
..79
MNOP
.05
+ +
.67 .172
+2.25 .872
+
฀
The variabilities used computing Table 117 are of

of in
course the variabilities the respective groups the

in
,
ability measured by the particular instrument used such

as
,
Army Alpha National

A.
or
or
Oalpha
is
ONational
σ
;
(
mg me MAlpha9 mAlpha6
is or
09 бAlpha9
mNat.9 mNat.6
or the like and will be smaller than
;
ONat.9
----
MCAVD9 MCAVD6
or
m19 m16
since σalpha or σNat.9 will be
9
OCAVD Οι
larger than σ19
)
.
σAlpha should be treated just as we treated ot's . We have

to estimate 019 from σAlpha9 or σNat.9 or Ootis9 . This has to be
done rather crudely since neither the self - correlations of
most of these tests , nor their correlations with any such
criterion as the score of one of our long CAVD series , have
been worked out . The self - correlation of the I. E. R. for
TABLE 117 .
DIFFERENCE BETWEEN GRADES IN SCORES ATTAINED IN VARIOUS INTELLIGENCE

EXAMINATIONS .
Using the
original scores Using scores in Equal Units
m, -m. M-13 m12 m, - m. -

m12 m , -
m13 m , -
m13 M12
σ, σ,
Army Alpha 1.75 1.28 2.50

Army a 1.42 1.382
National 1.76
Otis Adv . 1.49 1.05
Haggerty 1.09 1.21
I. E. R. 2.15 1.20
Terman 1.26
Brown .87
Myers .81
Pintner 1.14
Thorndike .832
Trabue 1.40
Illinois 1.40
Pressey 2.13
Median 1.40 1.452 1.21 1.94

Average 1.64 1.45 1.20
Median + Average
2
1.52 .832 1.45 1.21 1.94 .87
two different forms of the examination taken a year apart

is .82 for 1,000 boys of Grades 9 , 10 , and 11 , and is .86 for
489 sixteen - year - old boys in these grades [ Bailor , '24 , p .
8 ] . We have computed the self - correlation of the Terman
Group Test for 209 cases of high school pupils in Grades 9 ,
10 , and 11 , finding it to be .92 . The correlation of the Hag-
gerty test against a combined score in Army Alpha , Thurs-

tone , Otis , Pressey , and other tests is .89 for a group of 60
college seniors . This would make the self - correlation about
.80 . The self - correlation in " an entire school , " the two
trials being on the same day , is .90 [ Haggerty , '23 , p . 54 ] .
From the data given in the Memoirs [ '21 , pp . 315-17 ] , we
estimate the self - correlation of Army a as about .80 for a
grade population . The Otis Self - Administering correlates
.88 with the Terman Group Test in a group covering Grades
7 to 12 [ Clark, '25 , p . 15 ] .
Allowing for the restriction of the range in our groups
as compared with those reported above , we may expect the
self - correlations of these various examinations within one
grade to vary around a central tendency of about .80 for
rt , t,. Dividing by V.80 , we have , for the data from equal-
unit scores :
mo m6 = 1.62019 .
m12 m. = 1.35019-
m13 mg == 2.17019 .
m13 m12 = .973019 .
The same divisor with the data from original scores

gives :
m, — mg = 1.70019 .
M13 m12
= .934019 .
Allowing a weight of 4 to the determinations from scores

in equal units and a weight of 1 to the determinations from
the original scores , we have :
mo –m = 1.640 .
m12 - m9 = 1.35019 .
m13 m, == 2.17019 .
=
-
m13 m12 .961019 .
We have two independent estimates of m13 m9 , 2.17019

by the direct comparison and 2.3109 by the comparison
via Grade 12. Allowing equal weight to each gives 2.24019
as the combined estimate .
-m
-
The 1.64019 for m, , agrees very well with the ob-
served results of the Av . , - 1.74019 , and the Median ,
1.62019 , for m , -ms ; and we may reasonably accept
-1.74019 or -1.62019 or the average -1.68019 . We shall
take the last , and use ___ 1.709 as the m, -ms , difference .
The observed comparison of 91 and 911 may be taken as it
-
stands , there being no relevance of the general facts to it .
So 91 is .13019 or .221019 above m, and 9II is .13019 or
.223019 below it . We use +.2019 and 2019 .
The 2.24019 is much below the observed result of 3.60019
for our Group 13- Group 9

and , since this 3.600 , depends
;
upon the single determination by Composite N , it is wise to

consider possible amendments of it in view of the general
facts .
The following additional facts will help in the decision .
The individuals of Groups 13 and 17 were tested with half
of the Composite M and with D4 , which is only a little
harder than D4 . We can infer approximately what the
percent of successes with Composite M would have been ,
if it had all been given , by allowance for the missing C
( Completion M ) and for the replacement of D4 by D41 .
In the case of the 189 individuals of Group 13 there
were four who might perhaps have failed to have 20 or
more right out of 40 in Composite M if they had been tested
with it . By our estimates two probably would have so
failed . This gives 1.06 % or 2.340t13 below the median dif-
ficulty for Group 13. This , in terms of σ118 would be 2.75 ;
in terms of σ19 it would be 2.70 . This would make the 9
median 4.5019 below the 13 median .
Among the 240 of Group 17 there was no individual

who would not have had 20 or more right if he had been
tested with all the 40 tasks . There were some who probably
would have had only 22 , 23 , 24 , or 25 right . By our esti-
mates
2 would have scored 22 .
66 66 66
1 23 .
3 66 66 66
24.
66 66 66
2
32 25.
The level of Composite M is

then probably more than
30t17 below the median of Group 17 , but not much more
than that . A reasonable placement would be -3.2017 .
This in terms of σ117 would be 3.64 ; in terms of o₁9 it would
be 3.53 . This would make the 9 median 5.32019 below the 17
median or about 4.3019 below the 13 median .
In view of these additional facts it seems best to con
sider that our Group 13 differs more from our Group 9
than the college freshmen classes of our general survey dif
fered from the ninth grades of that survey, and that the
3.60019 is approximately correct . This means that we are
treating Groups 13 and 9 as if only about one in ten of the
latter were equal or superior to the lowest tenth of the
former in altitude of Intellect CAVD ; and this would not,
in our opinion , seem too small an overlapping to anyone
who knew the two groups .
The difference ( 1.00019 av . or .92019 median ) between
Group 13 and Group 17 is determined from four different
composites and with a mean square error of only .11 . There
is no reason to alter this in one direction rather than in
another .
So we put all the measures of difficulty of Table 116 into

differences from the difficulty of the task at which 50 per
-
cent of our Group 9 would succeed by the following :
- 1.7019 for Group 51
66 66
+.2019 91
66 66
.2019 911
66 66
+3.6019 13
66 66
+4.6019 17
The results are shown in Table 118 .

The average values , allowing equal weight to each deter
mination , are
I
:
——- 3.2019 M
J=
+ 1.8019
= 1.2019 N = + 2.5019
K: - .2019 0 = +3.7019
L= .8019 P = + 4.3019
Q= + 5.7019
The differences all in terms of

J-I =
, σ19 , are :
N M = .7
K-J =
2.0*
-
0.6
1.0 * N= 1.2
- P O
-
L -K = 1.0
-
M -L = 1.0 = 1.4
Q
P
The measurement of the unreliabilities of these deter
minations is beyond our facilities both of time and skill .
They are doubtless large , perhaps as large as .15 . They
are , however , not as large by far ( relative to the differ
ences to be measured ) as are those of the best forms of the
Binet .
TABLE 118 .
THE INTELLECTUAL DIFFICULTY OF TASKS I

TO Q EXPRESSED IN EACH CASE AS A
DIFFERENCE FROM THE MEDIAN DIFFICULTY FOR GROUP 9 , IN UNITS OF σ19.
Task Difficulty
By 52 By 91 By 911 By 13 By 17
I
J
3.16 3.17
- -
- -
1.10
.40
1.23
.15 - .10
-
.69 .90
+ + +
+ +
2.13 1.45
2.55 2.51 2.41
+
+ + + +
KLMNOPQ
3.65 + 3.81
4.27 4.422
Р
5.85 +5.472
THE DIFFICULTY OF COMPOSITES AND H³

A
G
D
C
B
,
,
,
,
E,
F,
,
As was stated the beginning the chapter the mea

at
of
surements of these lower levels of difficulty are less secure

than those of tasks since investigations of the form
to
Q
I
of distribution of the various groups used and of their dif

,
ferences central tendency and variability comparable

in
to
the investigations the case of Grades 13 have not

to
in
These estimates will be amended by the results from other large groups
4
to become 1.9 for and 1.1 for K

-J
J-
I
Composite contained only 30 single tasks having no sentence com

H
5
pletions
.
been made . The results of such investigations as we have

made are reported in Appendix VI .
The basal facts for measuring the differences in diffi
culty between A and B , B and C , C and D , and so on , are
the results of experiments with 180 adult imbeciles of Stan
ford Mental Age from about 2 years to 5 years , 100 adults
of mental age 6 ( a few over 84 months ) , 50 feeble - minded
comprising all the children graded as Class III in one in
stitution for the feeble - minded , 101 pupils in ungraded
classes in a large city , ' 163 pupils in Grade 4 ( second half ) ,
311 pupils in Grade 5 , and 44 adults , recruits in the United
States Army. These groups will be referred to in order as :
im . 3, im . 6 , f ., sp . , 4 , 5 , and ad . ( The use of im . and f. in
volves no theory of classification , but is solely for con
venience . )
In groups im . 3 , im. 6 , f. , and sp . , the tasks were given
orally . In groups f. and sp . ( and in some cases in group im .
6 ) , the individual tested was allowed to look at the book
let as the questions were asked , and read it if he could . In
groups 4 and 5, the tasks were all presented in print . The
comparative difficulty for any given group of oral and
printed presentation has not been determined . In the com
putations of differences between groups in variability and
central tendency which follow , the assumption is made that
the pupils in Grades 4B and 5 would do better , but vary
about as much , if they were tested in the manner used with
the lower - level groups , as they did when tested with the
printed booklets . The amount of allowance made will be
described when the differences of groups below group 4
from groups 4 and above are computed .
The percent succeeding for each of the 40 - composite
tasks is reported for such of the groups as were measured
by that task, in Table 119. Table 119 thus corresponds to
Table 106 .
This Class III corresponds roughly to grade 3 of an ordinary school .
The chronological ages ranged from 9 to 21 , only 6 being below 12 and only
2 over 18 .
7 The distribution of ages reported was : 15 from 13-0 to 13-11 , 37 from
14-0 to 14-11 , 39 from 15-0 to 15–11 , and 10 from 16-0 to 16-11 .
Using for the respective groups the forms of distribu

tion derived and described in Appendix VI, the difficulty of
each 40 - composite is found in terms of its difference from
the difficulty of that 40 - composite which exactly half of the
group in question would have succeeded with , in terms of
the mean square deviation of the group in question in the
ability measured by that 40 - composite . These measures
appear in Table 120 , which corresponds to Table 107 .
TABLE 119
.
PERCENTS SUCCEEDING WITH VARIOUS COMPOSITES IN GROUPS IM3 IM6
,
F,
SP and ad
5,
,
4,
.
Groups Institutions Special Regular School Adult
in
for the Feeble Minded Classes Classes Recruits

-
B 5
7
im3 im6 sp ad
4
f
MA 22 MA MA
7 6
to to to 10+
5
180 100 50 = 101 163 311 44

=
=
n
=
=
n
n
=
n
n
n
88.3
48.3
12.8 98.0
с
฀ 00.6 73.0
45.0 96.0 98.0
14.0 94.0 96.0 100.0 100.0
03.0 66.0 88.1 98.8 100.0
ABCDEFGHILE
н 68.0 67.3 91.4 97.7 97.7
J
06.0 34.7 35.6 63.3 70.5
03.1 13.2 56.8
K 00.0 00.3 47.7
ESTIMATING FROM σt
σ
By means of determinations
of rtite for the various 40
composites the various groups the measures units
in
of
in
,
CA im3 Oc im69 and the like are transmuted into units of

,
ime and the like The essential facts of these deter

σi
im3
O
,
,
minations are shown below The results appear Table

in
.
122 which corresponds Table 113

to
,
In general we have measured rtit2 both by the Spear

,
man formula using two twenties and by the correlations

,
of neighboring forties To economize time only one method

,
.
is used in the case of group im . 3 and group im . 6 and group

ad .; and only 98 of the 180 individuals are used in group
im . 3 .
The self - correlation of one random half of a 40 - com-
posite with the other half for 98 of the imbeciles of men-
tal age 2 to 5 years was found to be .86 for A, .77 for
B , .86 for C , and .76 for D. The self - correlation of one
40- composite with another at the same level may then be
2r20
estimated ( by as .927 for .874 for .924
B
A
140
+ 120
=
,
,
1
for and .864 for D.

C
,
TABLE 120
.
THE DIFFICULTY OF COMPOSITES IN VARIOUS GROUPS EXPRESSED AS
K
TO
A
A
,
DEVIATION FROM THE DIFFICULTY FOR THE MEDIAN OF THAT GROUP IN
,
TERMS OF THE OF THAT GROUP IN THE ABILITY MEASURED BY
σ
SUCCESS WITH THE COMPOSITE IN QUESTION
.
Group im im sp ad
4
5
f
3
n 180 100 50 101 163 311 44
1.68
-
В .05
+
с
-
+1.13 1.90
-
+1.83 .45
.29 1.33 2.61
+
- -
- --
+1.25 1.25 2.31 <-3.10
- <-3.10
+2.08 .33 1.54 2.26
- -
ABCDEFGHILE
.41 - .44 1.37 2.00 2.00

-
- -
+1.17 .36 .37 - .34 1.00

JΙ
+
+
+1.87 +1.12 - .35

K > +3.10 +2.75 .06
+
Dividing the entries under im Tables 120 by √.927

in
3
V.874 V.924 and V.864 respectively we obtain values

in
,
terms of from the values for 1m3OB 1m3 etc. They

σ
σ
1m3
,
are - 1.74 +.05 +1.18 and +1.97 as shown Table

in
:
,
,
,
,
122
.
The inter correlations the 40 composites

of
E
D
C
-
F,
and in the case of the 100 adults of mental age were as

G
shown in Table 121. The correlations with neighboring

composites were .685 for A , .703 for B , .725 for C , .769 for
D , and .809 for E. We add .03 to obtain estimated rt₁t2's .
Dividing the entries in the im 6 column of Table 120 by
V.715 , v.733 , v.755 , v.799 , and V.839 , respectively , we
obtain values in terms of σ ime from the values for σc ime ,
OD im6 , OE ime , etc. They are : -2.25 , — .53 , + .33 , +1.40 , -
and 2.27 as shown in Table 122 .
The self- correlation of one random half of a 40 - compos-
ite with the other half for group f ( the 50 feeble - minded in
class 3 ) was found to be .638 for E , .809 for F , .638 for G ,
.876 for H , and .588 for I. The self- correlation of one 40-
TABLE 121 .
RAW INTERCORRELATIONS OF COMPOSITES C , D , E , F AND G IN THE CASE OF 100

INDIVIDUALS CHRONOLOGICALLY SIXTEEN OR OVER , AND MENTALLY SIX .
D E F G
.685 .685 .588 .426

D .721 .638 .426
E .729 .509
CARF
.809
-
composite with another the same level of difficulty thus
at
is
2120
by г40 .779 for .894 for .779 for .934 for
E
+ 120
G
,
F,
,
1
and .741 for

H
I.
,
The inter correlations of the 40 composites

H
E
G
=F
F.59
-
,
,
,
-
and for group were with with .81

E
H.88
I
f,
,
:
with and with .81 The correlations with

H
=
I
,
neighboring composites are thus .59 for .70 for .841

E
,
F,
for .841 for and .81 for Adding .03 as an allowance

H
I.
G
,
for remoteness gives .62 .73 .871 .871 and .84

,
Allowing equal weight these two determinations the

to
rt2 are respectively .70 .81 .83 .90 and .79

of
values
,
,
,
Dividing the entries column Table 120 by v.70 V.81

in
in
f
V.90 and V.79 respectively we obtain values

in
V.83
,
,
,
,
terms of from the values of and

OF
OEt Oot Oнt

σ
o
,
,
,
They are -1.59 ,

in Table 122 .
1.39 , -
. 36 , - -
.43 , and +1.32 , as shown
In group sp ( the 101 pupils in special classes ) the inter-

correlations of neighboring composites were : F with G =
.62 , G with H = .77 , H with I
= .86 . Adding .03 allowance
for remoteness , rtit2 is .65 for F , .73 for G, .82 for H , and
.89 for I.
The self - correlation of one random half of a 40 - com-
posite with the other half in group sp is .73 for F , .54 for G ,
.64 for H , and .82 for I. The correlation of a 40 - composite
with another of equal difficulty , that is , rtit2 , may by these
2120
facts be estimated ( by as .844 for .701
140
=
+ 120
F,
I. 1
for .780 for and .901 for
H
G
,
Giving equal weight these two determinations we

to
,
have values .80 and .891 for
in as
of
rtit2 .75 .711
H
G
,
F,
,
,
group sp Dividing the entries
sp
the column
and
in
I
Table 120 by v.75 v.715 v.80 and v.895 respec-

of
,
tively we have values terms of from the values of
sp
in
oi
,
etc. They are -2.67 -1.82 49 and +.38
as
OE
-
OF
Sp
sp
,
,
,
,
,
.
,
entered in Table 122

.
group the 163 cases Grade 4B the intercorre-

In
of
I.86
4
)
= G (
lations were with .831 with with =

=
H
J
I
:
.63 with .47 Adding .03 as allowance for remote-

K
J
;
ness for for for and .58 for

J.
.861 .88
H
.77½
G
is
rt1t2
I ,
,
,
,
In group the self correlations one half with the

of
4
other half each 40 composite were .69 for for

of
.79
H
G
,
,
-
.83 for and .65 for The correlation 40 composite

of
J.
a
I ,
with another equal difficulty thus .817 for for

of
.883
H
G
is
J.
,
,
.907 for and .788 for

I ,
Giving equal weight the two determinations

to
of
rt1t2
,
we have .84 .88 .84 and .68 for and respec-

H
G
J
I
,
,
,
,
,
,
tively Dividing the entries in the column Table 120 by

of
4
.
√.84 √.88 √.84 and V.681 respectively we have values

,
-
,
in terms of
o₁4 from the values of OF OG They OH 49 etc.
Он
,
4
4,
4,
are 2.47 1.47 +.40 and +2.26 as entered in Table

,
,
,
,
122
.
The intercorrelations in the case of the 311 pupils of

Grade 5 were : H with = .77 , with I
= .85 , with K = I J J
.61 . The correlations with neighboring composites , ele
vated .03 to allow for remoteness , are thus : .80 for H , .84
I
for , .76 for , and .63 for K.J
The self - correlations in this group , using 20 elements
with 20 , are : .68 , .77 , .70 , and .51 for H , , , and K in order . I J
The correlation of one 40 - composite with another of equal
difficulty would then be .81 for F , .87 for I , .82½ for , and J
.67 for K.
Allowing equal weight to the two determinations of rt1t2 ,
TABLE 122 .
THE DIFFICULTY OF COMPOSITES A TO K IN , Oime, Of , ETC.
TERMS OF O1m3
im3 im6 sp ad
5
4
f
A 1.74
-
.05
+
+1.18 2.25
с
฀ 1.97 .53
+
.33 1.59
- -
+ + +
1.40 1.39 2.67

-
- -
-
1.82
-
2.27 .36 2.47
- -
н .43 .49 1.46 2.23 2.34

- -
BCDEFGHIK
1.32 .38 .40 .37 1.09

+
+ +
2.26 +1.26 - .35

J
3.41 .06
+
we have .80 for .85 for .79 for and .65 for K. Di
H
J
,
I,
viding the entries Table by V.805

of
Column
in
120
5
V.855 V.79 and V.65 respectively we have values

in
,
,
,
terms of o₁5 from the values for σн They are OK

σJ
σ1
-—
5.
5,
5,
5,
2.23 .37 +1.26 and +3.41 as shown in Table 122

—
In the case of the 44 adults the intercorrelations of the

,
40 composites were with .75 with .65 with

=
=
H
H
G
I
,
,
-
.96 and with Allowing +.03 for remote

=
=
K
.91
J
J
,
ness we have .78 .73 .83 .95 and .94 as the probable
,
,
,
correlation of with another 40 composite

each
K
H
G
J
,
I,
,
equal difficulty Dividing the entries Column ad

of
in
of
23
Table 120 by v.73, v.835 , V.955, and V.94 , respectively ,

we have values
OH ad,
in terms of σ ad from the values of og ad,
I ad, бJ ad, and σk ad. They are .2.34 , — 1.09 , — .35 , -
and .06 , as shown in Table 122 .
EXPRESSING THE σ OF EACH GROUP IN TERMS OF σ19
In
accordance with the earlier findings , σ191 and σ1911 are
treated as equal .
The σ , of group 5 ( 311 pupils in Grade 5 ) is made com-
parable with σ19 by finding the difference in difficulty be-
tween two tasks in terms of σ15 and in terms of σ191 , which is
equal to σ19 . Thus
K-J = 2.15015 and whereby
J-I = 1.63015
1.090191 , σ15 .510191 .
and 1.940191 , whereby σ15

== 1.190191 .
It is also possible to proceed indirectly by way of σ15 which
was found to equal .98019 . Thus
K- J= =
J-I -
2.15015 and 72015 or 705019 , whereby σ15 .33019 .
1.63015 and 2.1001st or ==1.26019 .

2.06019 , whereby σ15
It is also true in general that the variability of Grade 5 in

intellect will not be much different from that of Grade 51 .
If an estimatehad to be made from general considera-
tions , would be expected to be at least .95019 . We assign
σ15
equal weight to .85 ( the median of the .51 , 1.19 , .33 , and
1.26 ) and to .95 ; and use .9009 as the value of σ15 . The
— 2.23 , — .37 , -1.26 , and +3.41 of Table 122 in terms of
σ15 thus become the
- 2.01 , -.33 , 1.13 , and +3.07 of
Table 123 in terms of σ19 .
Next , the o₁4 is put in terms of o₁9 both directly and via
J-I =
015 .
1.860,4 and 1.94019 whereby =

J-I =
, σ14 1.04019 .
1.86014 and 1.63015 , whereby σ14
= .88015 or .79019-
I-H1.86014 and 1.86015 , whereby σ14
= 1.00015 or .90019 .
From these facts is taken to be approximately equal

, σ14
to .94015 or .85019 . The -2.47

2.47 , -1.46 , +.40 , and + 2.26 of
Table
and
122 in terms of σ14 thus become
1.92 of Table 123 in terms of σ19 .
- 2.10 , — 1.24 , + .34 ,
Next the Osp ( the 101 special class pupils ) is put in

terms of Oi9 via 015 and via σ14.
H -G = 1.33019 and 1.01014 , whereby σ1Sp = .76014 or .65019-
I -H = .87σisp and 1.86014 , whereby op = 2.14014 or 1.93019 .
Nothing is known precisely of the general tendency of

pupils over 13 in such special - class populations to vary ,
though the expectation would be that the variation would
be fairly wide , from pupils who really belonged in an insti-
tution for the feeble -minded to pupils who really belonged
in a regular Grade 4. Giving equal weight to the three de-
terminations , Oisp = 1.47019 . Giving equal weight to the
I-H and the H G pairs , σisp = 1.26019 . We use the latter .
-
In a similar manner σif , σi im39 and σi ime are put in terms
of 019. The essential facts are :
I-H = 1.75σ₁ , or 1.86015
Oif
O₁t
-or or .8701sp , whereby
1.86014
1.06015 or .950199
σ₁ = 1.06014 or .900199
or σit
or of 5001sp or .63019 .
H -F = .960 or 2.1801p , whereby o = 2.2701sp or 2.86019 .
We take the median of these four observations , .921019 .
-
G F.8701 ime and 1.03σ , and .8501sp , whereby
= 1.180 or 1.09019 or
=
O ime
01i ime
Οι .9701sp or 1.22019 .
F -E 1.0701 1m6 and .200₁ , whereby σ₁ im6
= .1870 or .17019 .
Since group im6 contains only individuals of Stanford

Mental Age 6 , it may be assumed to be much less variable
than group f or group im3 , or any other group used here .
The average of the three determinations ( 1.09019 , 1.22019 ,
and .17019 ) which is is used , giving in terms of σ19
.83019 ,
1.87 , -.44 , .27 , +1.16 , and +1.88 as the entries in
Table 123 .
8 Plus two individuals of mental age 7.
D -C = .7901 1m3 or 1.7201 ime, whereby

бi im3
== 2.180₁ 1m6 or 1.81019 .
Using 1m31.81019 , the entries

01
σ1 for Table 123 are — 3.15 ,
+.09 , +2.14 , and +3.57 .
The facts for the adult group are :
K- J=
J-I =
.4101ad or 1.09019 whereby olad 2.67019-
.7401ad or 1.94019 whereby σlad = 2.62019 .
I -H = 1.2501ad or or
-
1.86015 or 1.86014 .8701sp or 1.75011 .
By these four indirect computations , σ₁ad = or 1.34619 .

1.49015
Olad 1.49014 or 1.27019 .
Olad = .7001sp or .88019 .

Olad = 1.40σ₁ , or 1.30019 .
TABLE 123 .
THE DIFFICULTY OF COMPOSITES A TO K, IN TERMS OF 19.
im3 im6 f sp . 4 5 ad . 52 91
A - 3.15
B + .09
C +2.14 -1.87
D + 3.57 - .44
E -1.47
- 1.29
+ .27
F +1.16 -3.36
G +1.88 .33 -2.29 - 2.10
H
I
.40 .62 -1.24 -2.01 -5.12
- 1.52 - 3.37
J
- -2.39
-
+ 1.22 + .48 + .34 .33
+1.92 + 1.13 .77 + .62 - 1.43
K +3.07.13 + 1.36 - .35
We weight each of the direct comparisons as equal to

the median of the four indirect comparisons and thus by
averaging, have , as the estimate used , σ1 ad = 2.19019 . This
value is not unreasonable , since the group of adults in-
cluded men of schooling all the way from Grade 3 to Grade
Table 122 , are then : — 5.12 , -

12. The entries for Table 123 , derived from Column ad in
2.39 ,
-
.77 , and +.13.
EXPRESSING THE MEASURES OF DIFFICULTY AS DISTANCES FROM A

COMMON POINT OF REFERENCE
The measures of Table 123 are in every case distances

from the median difficulty for the group in question . We
shall express each ( in Table 124 ) as a distance from the
median difficulty for Grade 91 + II
, in terms of σ₁9 as a unit ,
by estimating the distance of the median difficulty for

group 5 from the median difficulty for group 91 + , and II
similarly for the median difficulty for each of groups ad , 5 ,
4, sp , f, im6 , and im3 . The essential facts and procedures
are stated below . We use M9 + II to denote the difficulty of
the task which exactly 50 % of the Grade 9 group will suc-
ceed with , M , to denote the difficulty of the task which ex-
actly 50 % of the group 5 will succeed with , and similarly
for Mim3 , Mime , Mt , Map , M4 , Mad , M5 M.
KM
, M91 , and
=
+ 1.36019 .
-
J = M5.62019
-5 +3.07019
M5 , whence M5 M5 = 1.71019 .
= M5 +1.13019 , whence M5 ,
- M =
I
.51019 .
-
= M5 + -1.52019 .
= M, -- M
J = M5.62019
.33019 , whence M5 ; 1.19019 .
- - M4 = 1.30019
.
IM
= whence M5 ,
-
M4 +1.92019 , .
, -1.52019 .
JM
= M4 + .34019 , whence M₁ =
-
M5 1.86019 .
, +1.13019 .
= M4 +1.92019 , whence M5 , M₁
M4 =
= .79019 .
I = M 5 + .33019 .
= M₁.34019 , whence M - M₁ =
-
.67019 .
H = M₁ 2.01019 .
= M4·4 -1.24019 , whence M , M₁ = .77019 .
In view of the above , we take :

.70019 as the difference between M4 and M5,
.90019 as the difference between M , and M5 , and
1.600 , as the difference between M4, and M51 .
These figures everything well except the determina-
fit
―
,
tion of M5 M by and this we believe deserves less
K
,
;
-
weight than the others
.
M5 -1.52019
=
= Msp
I
,
.48019 whence M5 Msp 2.00019 and
-
+
,
- 2.00019-1.60019 or .40019
M4 M sp
-
=
.
M .33019
=
= Msp.48019
,
I
whence Msp and
-M
=
.81019
M ,
,
M4 or
=
.8119 .70019 .11019
MSP
,
H= M
.
-2.01019
,
Msp whence M5 M sp and

=
-
,,
.62019 1.39019
,
M4 Msp 1.39019.70019
or
=
-
- .6919
.
M₁
-
=
.34019
I
Msp + whence M4 Msp
=
.48019 .14019
--
=
.
H= M
M₁ -1.24019
H
= Msp
,
whence M4 M
-
,,
.62019 62σ19
,
.
M -2.10019
=
G
= Msp -2.29019 whence M4 -- Map.19019

sp
,
Taking these six differences their face value there

at
is
,
median difference of .15019 and an average difference of
a
.1109 We have allow for the fact that presentation was

to
to .
oral group sp Lacking experimental evidence this al-

,
.
lowance arbitrary We allow .27019 making the special

is
-
,
.
class .4009 below group

.4
M5 -1.52019
=
I
,
M₂ +1.22019 whence Ms M
M₂
- and =
=
-
2.74019
=
,
,
Msp
M
.74019
,
M -
=
-M
.33019
-
I
, ,
= whence M5 and
=
M
+1.22019 1.55019
-
f,
,
Msp M₁ .45019
=
-
M -2.01019
=
H
=M .4009 whence M¸ M 1.610 and

M
-
=
,
,
-
,
,
-
:
Msp
M
.51019
,
M4
=
.34019
=
I
+1.22019 whence M4 and

-
.88019
M
=
=
,
M M .48019
SP
,
-
.
MM =
H = M4 -1.24019
= M₂ .40019 , whence , = .84019 and
―-Me := .44019 .
MSD
-M =
G = M4 - 2.10019
= M, .33019 , whence M₁ , 1.77019 and
Msp — M, =
I = Msp
1.27019 .
+ .48019
= M ,f +1.22019 , whence Ms , - M₂ = .74019-
I == M sp
- .62019
M .4009 , whence M ,, M, =
-
Msp .22019 .
= Msp
-
2.29019
= M, .33019 , whence Msp M₁, = 1.86019
M
= Map
.
sp 3.36019
= M,f -1.29019 , whence Msp --- M₁f = 2.07019 .
The four direct comparisons with the sp group which

had oral presentation are the most important . Their aver-
age is 1.2209 ; their median , 1.30019 . The average of the
other six is .65019 ; their median , 50019. We use 1.100199
which is very close to the result obtained by weighting the
result from direct comparison 3 and the result from indi-
rect comparison 1. This puts the group as 1.500₁9 below f
group 4 , which is not unreasonable , since these feeble-
minded individuals were doing approximately the work of
a regular school grade 3 .
G=
--
Msp
Mim
-2.29019
61.88019 , whence Msp - Mim 6
M₂- Mim
4.17019
6
and
= 3.07019 .
F = Msp -3.36019
-
= Mim 6 + 1.16019 , whence Msp Mim6 4.52019 and
G = M, - .33019
Mr Mim 6 = 3.42019 .
= 2.21019 .
-
== Mim
-
6 + 1.88019 , whence M, - Mim 6
F = M, 1.29019
= Mim + M, -Mim 6
-
6 1.16019 , whence 2.45019 .
E = M, -- 1.47019
== Mim 6.27019 , whence M, - Mim 8
6 = 1.74019 .
The average of the five determinations is 2.58019 ; the

median is 2.45019 . We use 2.50019 .
D = M₁m 6 - .44019
= Mim 3 + 3.57019 , whence Mim 6
-
-Mim 3 = 4.01019-
C = Mim 6 1.87019
=
-Mim 3 + 2.14019 , whence Mim 6- Mim3 =4.01019 = .
We use 4.00019-
These determinations of differences are obviously far

less reliable than is desirable , and should some time be
made precisely . They are , however , presumably free from
constant errors , and the variable errors do not prevent
them from satisfying one main purpose of relating the mea-
I J
sures of , , K , L , M , N , O , and P to an approximate abso-
lute zero . We find the differences between M, and M , and
M,, etc. , as follows :
M, = by the data presented earlier in the
-
M5 1.70019 ,
chapter .
Mst M = and M , Ms
M5 =
- M4
.90019 2.60019 .
=
M₂
-
- M₁ .70019 , SO M5 M₁ = 1.600199
and M ,
-
M4 = 3.30019 .
M4 - - Msp = .40019 , SO M5 Msp = 2.000199

and M , --
-M -
― MSD = 3.70019-
Msp - - M₂ = 1.10019 , SO M5 , =3.100199

=
- Mim -
and M , ― M , = 4.80019 .
= 2.50019 , so
M, 8 SO Ma
M5 Mim 65.60019 ,
and M , -
-
Mim 6-7.30019-
-
Mim6 Mim 34.00019 , SO M5 Mim 39.600199
and M, Mim 311.30019
-Mim 3 .
These facts are used to put all the entries of Table 123
into differences from the median difficulty of group 9. The
result is Table 124 , which is thus a continuation of Table
118 .
The adult group is given its location by the following :

K = Mad.13019
= M91.35019 , whence M91
-— Mad = .48019 .
J= Mad
-
.77019
M91- Mad =
= , whence
I
M91 1.43019 .66019 .
= Mad -- 2.39019
Mor -
= M91 3.37019 , whence Mor
91 Mad = .98019 .
K = Mad +
= M5 +
.13019
1.36019 , whence Mad - M5-1.23019 .
JMad.77019
= M51 + .62019 -
-
whence M5 + =
I=
, Mad 1.39019-
Mad 2.39019
= M5-1.52019 , whence Mad M51
= .87019 .
TABLE 124.
THE INTELLECTUAL DIFFICULTY OF TASKS A TO K EXPRESSED AS A DIFFERENCE

FROM THE MEDIAN DIFFICULTY FOR GROUP 9 , IN UNITS OF σ19.
im3 im6 sp ad 52 91
5
4
f
.
.
Α 14.45
- - - -
В 11.21
9.16 9.17
-
7.73 -7.74
-7.03 -6.27
-6.14 -6.09 -7.06
5.42 5.13 -5.99 -5.40
-
- - -
ABCDEFGHIK
5.20 4.32 -4.54 4.61 5.65

Ι
- -
- - -
3.58 3.22 2.96 -2.93 2.92 3.32 3.17

-
-
- -
- -
1.38 -1.47 1.30 -1.08 -1.23
J
К .47 .41 - .34 .15

+
on the average .71019 below M91 or .51σ9

Thus
is
Mad
M.
,
below and 1.16σ₁ above Since M 1.7009 above

M
is
,
M5 this second determination equivalent .5609 below

to
is
-
.. ,
transmute the
--
We use the average .53019 and
-
so
M
,
,
.77 and Table into

—
—
of
5.12 2.39 .13 123 5.65

in +
,
,
2.92 1.30 and — .41 Table 124

,
.
,
The data from groups and are also repeated

in
91
5
Table 124. Table 124 then has four determinations of the

difficulty of five determinations of the difficulty
of
K
J ,
,
seven determinations the difficulty of and so on

of
I ,
.
We first consider whether there is any need to modify

our earlier estimate of the difficulty of K , , and I in view J
of the extensive new facts.
-
K was.2019 from the median of Group 9 , by the aver-
age
Groups
of three determinations
5 , and
+47019 by Group 5.
91 , 9II . It is
(
In view of the
-
.34 ,
.5109 by
and -.10 ) from
. 15 ,
the adults and

great discrepancy , and
the fact that K is perhaps too hard a task to enlist full
effort from pupils in Grade 5 , we leave the - .2 as the esti-
mate for K. J
was -1.16019 from the median of Group 9 .
It is 1.3809 by the averages of the determinations from
4 , 5 , and ad . Then 1.30,9 is more probable than 1.20₁ , as its
value. Iwas — 3.2019 by Groups 5 and 91. It is -3.12019
by the average of the five new determinations . We leave it
as So we may use -3.2019 for , -1.30₁9 for , I J
J
3.2019 .
and.2019 for K. The difference -I is then 1.901 , instead
J
-
of 2.0019 , and the difference K- is 1.10 , instead of 1.0019 .
-
The difficulty of H is 4.7019 or-- 4.6019 ( average and
-
median ) that of G is -5.409 ; that of F is -6.4019 or
;
- 6.109 ; E , D , C , B , and A are , in order ,

the difficulties of
-6.610199 -9.2019
7.7019 , , -11.2019 , and 14.41019 .
The difficulties for L , M , N , O , P , and Q were found ( in

Table 118 ) to be respectively +.8019 , +1.8019 , +2.5019 ,
+3.71019 , +4.4019 , and +5.6019 .
These measures are all deviations from the difficulty for

II
the median of Group 91+ . Expressed as deviations
from the difficulty of Task A , they are , in the A , B , C order ,
A = 0 , B = 3.21 , C = 5.21 , D = 6.71 , E = 7.8 , F = 8.2 ,
J
G = 9.0 , H = 9.8 , I = 11.3 , = 13.11 , K = 14.21 , L = 15.21 ,
M = 16.21 , N = 16.9 } , O = 18.2 , P = 18.81 , and Q = 20.01 .
THE SCALE
In Chapter X it will
be shown that the distance from an
approximate absolute zero of intellectual difficulty to the
difficulty of Composite A is about 4.35 times the difference
in difficulty between Composite A and Composite C , which

is 5.28 . So the difficulties of A , B , C , etc. , measured from
an approximate absolute zero are , in units of σ19 :
23 361 or 36.1
J
В 261 K 371 or 37.2
281 L 381 or 38.2
293 M 391 or 39.2
ABCDE
303 N 40 or 39.9
F 311 0 41 or 41.1
32 Р 412 or 41.7
G
H 323 43 or 43.1
Q
341 or 34.2
I
The unit may now be given more realistic defini-

a
σ
tion one twentieth the difference difficulty be-

of
It
in
is
-
.
tween such composites an adult approximately mental

as
of
age three can succeed with the sense of obtaining 20 or

in
(
more right the single tasks and such only the ablest
of
as
)
fifth of college graduates can succeed with one twen-

It
is
-
.
tieth of the difference between tasks which over 999 per

at
thousand adults can succeed and tasks which only about

at
,
ten per thousand can succeed one tenth of the differ-

It
is
-
.
ence between tasks which nineteen out of twenty pupils

at
in Grade can succeed and those which only fifth of

at
5
a
,
college graduates can succeed corresponds about 13

to
It
.
average years of mental age from 12

to
6
The relative magnitudes from 23 for to 43 for will

A
seem preposterous to many critics who will deem incred-

it
ible that the intellects of the top one percent of men should
be less than twice as high as the intellects of the lower
"
"
ranges of asylum inmates that child three has at-

of
or
a
;
tained two thirds of the intellectual altitude which he ever

-
will attain
.
The relative magnitudes are far from secure depending

,
much they do on the experimental determinations

as
so
The first column gives the measures the nearest quarter unit which
to
9
,
-
closer than the data justify for tasks

even to H. The second column
A
is
gives measures the nearest tenth of unit for tasks to Q.

to
I
from the im3 group and the expert ratings to be described

in the next chapter . The 23 for A may conceivably be in
truth only 13 or 10 or even lower , though we have no evi-
dence that it is likely to be lower rather than higher . Even
if it were as low as 10 or 12 , the critics would still find the
relative magnitudes nearly as preposterous . The difficulty
is probably due to a confusion of altitude of intellect with
intellect in toto , a confusion which the Binet and other tests
seem to have stimulated . Intellect in toto is proportional
to altitude only if the number of tasks is approximately
equal at each level of difficulty . The number selected to
make an examination , such as the Binet , or our CAVD
series , may be so , but we shall show that the number of
tasks that are or can be made increases with their difficulty ,
so that what we may call the " area " of an intellect of alti-
tude 40 may be , not two times that of an intellect of altitude
20 , but twenty or two hundred or perhaps two thousand
times it . Any further discussion of these matters may best
be deferred until after the treatment of the location of zero
difficulty in Chapter X , and the treatment of the measure-
ment of width and " area " of intellect in Chapter XII .
CHAPTER X
THE ABSOLUTE ZERO OF INTELLECTUAL DIFFICULTY
We may expect the same sort and amount of advantage

to the scientific study of intellect from a determination of
its absolute zero as accrues to the study of temperature or
electrical resistance from the determination of their abso
lute zeros . Just as we cannot properly add or subtract or
average numbers representing degrees of intellectual diffi
culty until we know that the units called equal are really
equal , so we cannot properly make the " times as much "
judgment , or divide one amount of intellectual difficulty by
another to form a ratio , until we can state these amounts
as differences from a true absolute zero meaning just barely
not any intellectual difficulty .
Knowledge of the location of zero intellectual difficulty
not only will put all our measures of difficulty , or altitude
of intellect , into numbers capable of treatment in ratios ,
but also will put all our measures of what may be called the
total 66 surface " or " area " of intellect , ¹ into numbers
capable of similar treatment .
Relations within the field of intellect , and relations be
tween it and other facts measurable in known units from a
known absolute zero will then be susceptible of simple ,
straightforward study and presentation as lines in the ++
quadrant of a system of coördinates . In place of our pres
ent laborious and somewhat ambiguous determinations by
correlations and regressions , that such and such an excess
over or deficiency below the central tendency for a certain
group in trait Y is related to such and such an excess over
or deficiency below the central tendency of the same group
in trait X , we shall have simple and definite statements that
y = ax + b, found by plotting the straight line of best
fit
,
These terms will be defined Chapter XI

in
1
339
just as the physicist or engineer does . Where the relation

is curvilinear and intricate , the gain will be even greater ,
since in such cases the correlation and regression technique
is especially laborious and subject to ambiguity and mis
leading .
We shall also be stimulated to study psychological re

lations over a large part of the range of intellect from near
zero to near its maximum . The intellects of dogs , cats ,
primates , and men will more easily be made commensurate
and put in relation to the same fact in the same plot .
LOCATING ZERO DIFFICULTY BY EXPERIMENT

By the methods previously described , we can measure
differences in difficulty in Intellect CAVD from tasks such
as only one adult in a thousand can do down to tasks of
level A ( shown in Chapter III ) which over nine hundred
and ninety -nine adults in a thousand can do . Tasks in giv
ing the opposites of words or in answering informational
questions can be measured in respect of their intellectual
difficulty from an equally high down to an equally low level .
Consider now such a series of tasks as AA to AAAAA
below . Adult imbeciles and idiots who cannot do twenty
out of forty tasks as hard as those of Level A, may succeed
with twenty out of forty as hard as AA . A hundred dogs or
cats of specified age and training could be measured as to
their ability in composites of tasks like AA to AAAA . Am
phibians and fishes of specified selection and training could
be measured as to their ability in composites of tasks like
AAA to AAAAA.²
AA Responds to the direction " Come here ."
AAA Can find his way to some very familiar
place , such as his own sleeping - place .
AAAA Will not try to eat a familiar nasty -tast
ing object .
AAAAA Will not try to bite off his own toes .
2 We are , of course , here concerned with the difficulty of tasks learned
by intellect , not with that of tasks provided for by original nature .
THE ABSOLUTE ZERO OF INTELLECTUAL DIFFICULTY 341
A PROGRAM OF TASKS TO USE IN MEASURING TASKS OF VERY

LITTLE INTELLECTUAL DIFFICULTY .
Composite Task . Groups in Which the Percent of Suc

cesses Is between 100 and 0 .
I Earthworms
II 66
III 66
Crabs
IV 66 66
V Fishes 66
VI 66 66
VII 66
Frogs
VIII 66 66
IX Turtles 66
X 66 66
XI 66
Rats
XII 66 66
XIII Cats 66
XIV 66 66
XV 66
Dogs
XVI 66 66
XVII Monkeys 66
XVIII 66 66
XIX 66
Chimpanzees
XX 66 66
XXI Human adults , A³ 66
XXII 66 66
XXIII 66
Human adults , B³
XXIV 66 66
XXV Human adults , C³

66
XXVI 66 66
A+ 66
Human adults , D³
B₁ 66 66
C4 Human adults , E 66
D₁ 66
3 The A, B ,
C , D , and E groups of human adults are to have means and
mean square variations in mental months of about 12 ± 4 , 20 ± 4 , 30 ± 4 ,
394 , and 48 ± 4 .
4 These tasks A , B , C and D are the CAVD composites A , B, C and D

or tasks equal to these in intellectual difficulty .
With sufficient care and ingenuity , we could doubtless

devise composite tasks I , II , III , IV , V , and so on , such that
there would be percents between 100 and 0 succeeding
within each of the groups shown above and overlapping of
groups as shown in the program above . It is probable that
the number of test composites and of animal groups neces
sary will be much less than the program shows . We could
then measure the differences in difficulty from A down to
that intellectual task which earthworms can master , by the
same general methods as we have used from high levels
down to A.
The difficulty of the intellectual task which the earth
worm's intellect can master is so near zero difficulty that a
level slightly below it may safely be accepted as an approxi
mate absolute zero of intellectual difficulty , sufficiently close
for all purposes , theoretical or practical .
It is to be hoped that such a determination of zero intel
lectual difficulty by actual experimentation will sometime be
made . The time and facilities required made it imprac
ticable to include it among our investigations . We have
had to content ourselves with cruder methods .
LOCATING ZERO DIFFICULTY BY A CONSENSUS

What we have done is to utilize a consensus of psycholo
gists , especially such as are expert in animal and infant
psychology , or the psychology of the very dull . Each of
them in entire independence of all the others ranked 56
tasks shown below ( but presented in a random order ) in
accordance with the following instructions :
Please rank the tasks or achievements described on the

enclosed slips according to their intellectual difficulty , that
is , according to the degree of intellect required for a man to
perform each, supposing the man to have lived 20 years
with the average opportunities of a person born and bred
in an average English - speaking home ( or institution , if his
intellect is so slight that he has to be brought up in an in
stitution for the feeble - minded ). Mark the task or achieve
ment that requires least intellect , 1 ; the one that requires

next to least , 2 ; the one that requires second to least , 3 ; and
so on . Use your own conception of intellect in doing so .
If any of the tasks or achievements seem equally difficult ,
assign them the same rank .
MB
FIG . 34. Drawings used with tasks 57 , 58 , 81 , and 82. Reduced to about
2/3 original size .
57. Responds correctly to the direction " Draw a line

around the cup , " Fig. 34a being shown .
58. Responds correctly to the direction " Make a cross in
the square , " Fig . 34b being shown .
69. "
Can answer correctly What is the person that you
send for when you are sick ? "
70. Can answer correctly " Tell me something that walks on
four legs . "
75. Can recognize four fingers when four are held up and
he is asked " How many ? " in three trials out of five , the
trials being interspersed with other tasks .
24
76. The experimenter shows a box is empty and shows five

pennies clearly and puts them in the box . He then takes
out four pennies from the box and shows these clearly
and asks " How many cents are in the box ? " Two cor-
rect responses out of three are required , the trials being
interspersed with other tasks .
81. Can respond correctly when told " Put your finger on
the pistol , " or " Point to the pistol , " or " Find the pis-
tol , " or " Which is the pistol ? " or other familiar expres-
sion of similar meaning , Fig. 34c being shown .
the tiger , " or " Point to the tiger , " or " Find the tiger , "
or " Which is the tiger ? " or other familiar expression
of similar meaning , Fig . 34d being shown .
55. Responds correctly to the direction " Make the other
arm on this man . " ( Pointing . )
FIG . 35. Drawing used with task 55. Reduced to about 2/3 original size .
63. Responds correctly to the direction " Stand back of

your chair . ”
65. Can answer correctly " Is this morning or afternoon or
evening ? Which is it ? "
66. Can answer correctly " Tell me something that you are
afraid of. "
74. The experimenter shows that a box is empty . He holds
up one penny and puts it in the box . He then holds up
two pennies , both clearly in view, and puts them in the

box . He then asks " How many cents are in the box ? "
(Two correct responses out of three are required , the
three trials beng interspersed with other tasks . )
the window , " or " Point to the window , " or " Find the
"
window , or " Which is the window ? " or other familiar
expression of similar meaning , Fig . 36a being shown .
OC
FIG . 36.
3
Drawings used with tasks 79 and 80. Reduced to about 2/3 original
size .
the envelope , " or " Point to the envelope , " or " Find
the envelope , " or " Which is the envelope ? " or other
familiar expression of similar meaning , Fig . 36b being
shown .
51. Responds correctly to the direction " Make a line like
this , " the experimenter showing him by drawing a line
on a sheet of paper. ( Anything approximating a
straight line is to be scored correct . )
52. Responds correctly to the direction " Make a cross like
this . " ( Two lines that cross anywhere are to be scored
correct. )
59. Can answer " What do you wear on your head when you
go out ?"
60. Can answer" Tell me something that is good to eat .
Something else . Something else . " ( 3 required . )
71. Can respond correctly to " Show me the littlest square ;
show me the littlest one of all , " showing three as here .
( Three squares with sides respectively ", 1 ″ and 11 ″

are shown . )
72. Can respond correctly to " Show me the biggest square ;
show me the biggest one of all , " showing three as here . 1
( Three squares shown as in 71. )

the apple , " or " Point to the apple , " or " Find the
apple , " or " Which is the apple ? " or other familiar ex-
pression of similar meaning , Fig. 37a being shown .
FIG . 37. Drawings used with tasks 77 and 78. Reduced to about 2/3 original
size .
"
the dog , or " Point to the dog , " or " Find the dog , "
or " Which is the dog ? " or other familiar expression of
similar meaning , Fig. 37b being shown .
45. Can give a correct response to " Tell me the name of
something you eat . "
49. Can give a correct response when someone shows him a
watch and says " Tell me what this is . "
46. Can give a correct response when someone shows him a
penny and says " Tell me what this is ."
4. Responds to the direction " Give me the pencil , " assum-
ing that a pencil is in clear view before him and that you
are near enough for him to hand it to you.
5. Responds to the direction " Take the pencil , " supposing
one to be in clear view within his reach .
42. Can give a correct answer to " What do you call this ? "
(the questioner touching the nose of the one ques
tioned) .
29. Will not try to put a large object through a hole less
than one - fourth its size , for example , will not try to put
a baseball into an inkwell or put a football into his
pocket .
50. Can get into bed and cover himself with the bedclothes .
31. Will be disturbed if, after turning away from two cher
ished objects ( such as two pieces of cake ) he turns
back to find only one left .
18. Can open a door by turning an ordinary knob .
11. Responds to the direction " Shake hands " by holding
out his hand .
3. Responds to the direction " Hold up your hand . "
17. Can put on his hat .
40. Familiar and attractive food being on his plate , he will
be able to put it in his mouth with a spoon .
1. Responds to the direction , " Stand up . "
7. Responds to the direction , " Come here . "
33. Can find his way to some very familiar place , such as
the dining - room , his own bedroom or the bathroom .
26. Will not walk off a roof or wharf or the like where the
distance to the ground or water is 20 feet or more .
10. Responds to the direction , " Come here , John " ( sup
posing " John " to be his own name and assuming also
that the person giving the direction is a familiar friend
speaking in a pleasant voice , witha smile and with open
arms , representing a very habitual situation to which
approach has been the response ) .
23. Will go toward an object six feet off in case it is a
familiar , attractive , desired object .
24. Will go around a familiar object , or push it out of his
way, if it is movable , that is an obstacle in the way of
his passage to some attractive , desired , familiar object
which he is approaching.
16. Understands the meaning of " bad boy, " spoken in a

harsh voice with disapproving looks . (Use " girl " for
" boy " in the case of a female . )
39. Familiar and attractive food being on his plate , he will
be able to get it into his mouth somehow or other.
35. Being in the presence of some familiar source of great
heat ( wood fire , coal fire , gas flame , or the like , accord
ing to his environment ) , will not put his hand in it .
28. Will not try to eat a familiar nasty - tasting object , such
as soap or ashes or a shoe .
14. Understands his own name so as at least to feel or think
differently when it is spoken than when some other
word is .
34. Being in the presence of some familiar , attractive , de
sired object , such as food when he is hungry , will go
toward it rather than away from it.
41. Will bend his head or body to avoid a blow directed
from in front straight at his nose .
32. Responds to his best and kindest friend , for example ,
mother or nurse , differently from his response to
strangers .
27. Will not try to pull off his own fingers or toes .
38. Familiar and attractive food being offered him in a
familiar way , will take it .
37. Having an object of bitter , nasty taste in his mouth ,
will spit it out more often than hold it there .
36. Having an object of sweet , pleasant taste in his mouth ,
will keep it there more often than spit it out .
A point somewhere between the last two tasks listed ( 36

and 37 ) and tasks 27 and 32 may fairly be taken to repre
sent approximately zero intellectual difficulty , since tasks
36 and 37 are comparable to acts done by animals that can
not learn or can learn only in an infinitesimal way , and are
done by human beings probably without any learning .
Tasks 27 and 32 (" Will not try to pull off his own fingers
or toes " and " Responds to his best and kindest friend
. . ." ) are such as very dull animals can learn . Tasks 51 ,

52 , 59 , 60 , 71 , 72 , 77 , and 78 are a sampling from our level A.
Tasks 55 , 63 , 65 , 66 , 74 , 79 , and 80 are a sampling from our
level B. Tasks 57 , 58 , 69 , 70 , 75 , 76 , 81 , and 82 are a sam
pling from our level C. We know from experiment the dif
ference in difficulty between A and B , and between B and
C. If we can measure the differences C - B and B -A , we
have extended our scale to an approximate absolute zero .
We make this measurement by measuring the following dis
tances :
36 , 37 to 27 , 32 , which are of nearly equal difficulty inter se .

27 , 32 to 14 , 28 , 35 , 39 , which are of nearly equal difficulty
inter se.
14 28 35 , 39 to 10 , 26 , 33 , which are of nearly equal diffi
, ,
culty inter se .
10 , 26 , 33 to 3 , 11 , 18 , 31 , 50 , which are of nearly equal diffi
culty inter se .
3 , 11 , 18 , 31 , 50 to 46 , 49 , which are of nearly equal diffi
culty inter se .
46, 49 to the level A tasks , which are of nearly equal diffi
culty inter se.
The level A to the level B tasks , which are of nearly equal
difficulty inter se .
The level B to the level C tasks , which are of nearly equal
difficulty inter se .
The measurements , in terms of the size of the minority

( that is , the number of judges placing the one task as
harder than the other , though it is in truth easier ) , are
shown in Table 125. The corresponding distances in terms
of the variability of expert opinion , assuming it to be of
Form A, and taking 1.00 as its Q or Median Variation , are
also given in Table 125.5
The difference between 36 , 37 , and A is 4.7 times the dif
ference between A and C. The difference between 27 , 32 ,
5 These values will be in approximately the same proportions by any rea
sonable assumption about the form of the distribution , since we are using only
minorities between 6 and 15 out of 40 .
and Ais 4 times the difference between A and C. We may
sionally at A 4.35 ( - ). C CA
then set the absolute zero of intellectual difficulty provi-
-
A equals 5.28019 by the
facts reported in Chapter IX , whereby zero is 23019 below
A or A has a difficulty of 23019 measured from zero or just
not any intellectual difficulty . We estimate the unreliability
of this 23 due to the small number of judges as a probable
error of about 2.
TABLE 125 .
THE ESTIMATED DIFFERENCES IN DIFFICULTY OF INTELLECTUAL TASKS FROM

ZERO DIFFICULTY TO THE DIFFICULTY OF COMPOSITE C.
Average Difference
Minority % in Difficulty
36 , 37 , to 27 , 32 , 10.25 25.6 .97
27 , 32 , " 14 , 28 , 35 , 39 , 9.75 24.4 1.03

66
14 , 28 , 35 , 39 , 10 , 26 , 33 , 10.33 25.7 .97
"" 24.9
10 , 26 , 33 , 3 , 11 , 18 , 31 , 50 9.97 1.00
3 , 11 , 18 , 31 , 50 " 46 , 49 , 8.40 21.0 1.19
46 , 49 , " A 6.72 16.8 1.42
A " B6 10.95 27.4 .89
B C6 14.66 36.6 .51
36 , 37 , " A 6.58
27, 32 , " A 5.61
A " C 1.40
We present all the facts concerning all the ratings in

Table 126. The distance from zero to A can be determined
via different tasks from those which we have used , but the
result will be substantially the same provided that enough
tasks are used to reduce the effect of chance .
6 In the case of the A to B, and B to C comparisons , it would perhaps
be better to use the median minority rather than the average minority , be-
cause of the irregularities in the separate comparisons . The two medians
which would then replace 10.95 and 14.66 are 9.0 and 15.5 . The differences
in difficulty would then be 1.02 and .42 , giving a total of 1.44 for the C A -
difference in place of the 1.40 . The end result is thus almost the same . If
the median minority is used for the 46 , 49 to A comparison , we have 6.75 in
place of 6.72 .
.
THE TABLE 126
(
NUMBER OF PSYCHOLOGISTS
A
OUT OF 40 JUDGING
NUMBER CERTAIN TASK TO BE MORE
DIFFICULT
A
JUDGING THE TWO INTELLECTUALLY
,
.
TASKS TO BE EQUALLY
DIFFICULT THAN
:"
CERTAIN OTHER AND THE
IT
THE TABLE READS
) ,7
AND EQUAL TO IN DIFFICULTY BY TASK 36 WAS JUDGED
HARDER THAN TASK 37 BY 13
2
TASK 36 WAS JUDGED HARDER
THAN TASK 38 by
3
,
AND EQUAL TO IN DIFFICULTY
36 WAS JUDGED HARDER BY TASK
"
IT "
IT
THAN TASK 27 BY 10 AND
EQUAL TO
2
,'
37 38 27 32 41
IN DIFFICULTY BY AND SO ON
.
34 14 28 35
=
39
V
16 24
=
>=
V
> = 23 10
II
>
7
26
1
> = > 33
88 ^
> >= > > =
=
> = 11
A
> =
V
> >
V
> =
7
> =
2
36
3
13 10 61
> > = > =
4
1
61
1
1
37 62
6
3 5
12 10
6 6
38 61
2 21
21
1 1
2
20 22
A 2277
2
3
15 11
1
13 1
1 2
27 11 12
12 1
22 18 15 12
32 11 11 15 12 91
2
1
19 12 11
9
A L2721
7
10
7 1
112
A 34513
41
1
23 18
1
34 16 16 17 15 15 15
2
12
43
7
14 14 14
4
13 5
32
4
2
28 21 32
1
17 20 16 13 17 11
1425
35 20 15 13
14 13
8 8
16 11
3243
17 16 23492 54332
4123
6 8
112
12
79
11
39 14 11
13 4 4
11 1 1
16
4123 2
17 15
1112 1
15
9
1
10
2
24 15
18 20 11
23 18 17
214
29 12 16 11
10
1 122
4
2 1
10 14
2 2 1
15720
3277 7784333
26 18
5 728
17
33
7
19
113 1
17 13
13
18 14
33
13
44
V8 8 6
22
1 3
V 65 7 6
12
1 1
11
6
V 5
16
> =
11280
51
1 1 I
V₁ 411
1 13 30 3 37652 870∞
∞
∞
1 4 1 8
441 3245033-
||
3 7 5
91
10
20
4
3340
.) ^ 531 42 12
74 87
1 1 6
>=
5 ♡
Continued
18
22
21
( 1 1
>=
4 2052 39
3 6
18
23
10
126-
1 1
V 29744 477
||
20
11
E
123
TABLE
|| 2 2
5
2 112
1 2
||
V8 6 249
11 1
15
10
15
13
17
11
|| 1 1 1 2222
FA 4 9
12
10
312
92
3 5
|| 21 22
2 2
15
16
20
1 1 1
VA
13
19
13
17
21
2337
10
26
1
V 5
14
= 21 2
8
2000 251000
81 >
1 T
11
va 2 56 22
1
> =
8
75
2
va
11
V889
16
1
V8 4 9
13
16
17
23
27
.)
> =
4249
57
2112
>=
Concluded
80
( 43
11
24
126-
1 TT
21122 154
>=
74
TABLE
A 33009
7
11 1
69 5734r
5 9877
11 21
V8 2222
|| 1 122
Va 12 954
11 1
Vã 5 5 9 9
17
11 1
EA 550 7 8
18
11
V 13
||
ΕΛ 342
CHAPTER XI
THE MEASUREMENT OF THE ALTITUDE OF AN INDIVIDUAL

INTELLECT
It will , of course , very rarely happen that an individual
will have exactly twenty single tasks right out of the forty
in any composite
. We have to estimate the level at which
he would have exactly fifty percent right from such a record
J
as that shown below .
Composite Difficulty I K L M N O P Q
Numbers Correct 24 18 11 8 1
We may use all or part of his record , bearing in mind

that percents near zero and 100 ( that is , numbers right out
of forty near zero and forty ) are of less value than those
nearer 50 percent . To use his record most effectively we
need to know the general form of the curve ( especially near
the 50 percent point ) and to perform the equivalent of plot
ting the particular curve of this general form which best
fits the observations . We have determined the general form
of the curve from 10 percent of successes to 90 percent of
successes by the method described below .
THE FORM OF THE CURVE OF PERCENT CORRECT IN RELATION

TO DIFFICULTY
All the cases of group 17 and 13 are grouped into five

groups according to their general altitude as shown by the
sum of their scores ( number right ) in N , O , P , and Q.
The median score at each of the four levels for each of
the five groups is computed ; and the five curves are drawn .
They appear in Fig . 38 .
These curves are subject to a very slight error due to
the fact that they were computed using a value 41.9 for the
difficulty of composite P , which was later found to be 0.2 off
354
THE ALTITUDE OF AN INDIVIDUAL INTELLECT 355
from the correct value ( 41.7 ) . The difference to the general

argument is so trifling that we have not recomputed the
measures or redrawn the curves .
Q- 2 3
P-
0—
FIG . 38.
ļ
5 10 15 20 25 30 35
Curves of percent correct in relation to difficulty ; Groups 13 and 17 .
-------
!0 1
10 15 20 25 30 35
FIG . 39. The curves of Fig . 38 shifted so that similar percents correct fall
approximately on the same points .
The four curves other than that for the middle group
are then shifted up or down until each fits the curve for the
middle group as closely as may be for such part of the range
from zero to forty correct as they have in common . The
result is shown in Fig . 39. A curve ( Fig . 40 ) representing
the central tendency of all the five in Fig . 39 is drawn . Fig .
40 represents the most probable general form of the curve
of decrease in percentage correct with increase in difficulty
so far as groups 17 and 13 reveal it .
12 10 15 20 25 30 35
FIG . 40. The probable form of the decrease in percent correct with increase
in difficulty , for Groups 13 and 17.
Fig . 40 has heights as follows ( in terms of tenths of σ19 ) :

5 correct , 301 .
10 correct, +16
correct , + 71
-
15
20 correct , 0.
25 correct , 9.
30 correct , - 18 .
35 correct , -31 .
All the cases of group 91 are grouped into five groups

according to the sum of their scores in , , I
, L , and M. J K
The median score at each of the five levels is computed for
each of the five groups ; and the five curves are drawn .
They appear in Fig. 41 , being curves 6, 7 , 8 , 9, and 10 .
‫أ‬
N-
‫ا‬
M-
‫أ‬
K- 5
J—
10 15 20 25 30 35
FIG . 41. Curves of percent correct in relation to difficulty ; Groups 9 -I and
9 -II .
The four curves for 91 other than that for the middle
group are then shifted up or down until each fits the curve
for the middle group as closely as may be for such part of
the range from zero to forty correct as they have in com-
mon . The result is shown in Fig. 42. A curve ( Fig . 44a )
representing the central tendency of all the five in Fig . 42
is drawn .
All the cases of group 9II are grouped into five groups
according to the sum of their scores in , , M , N , and O.K L
The median score at each of the five levels is computed for
each of the five groups ; and the five curves are drawn .
They appear in Fig . 41 , being curves 1 , 2 , 3 , 4 , and 5 .
The four curves for 9II other than that for the middle
group are then shifted up or down until each fits the curve
--
--
.
---
ļ
1
ļ 10
16 15
15 20 25 30 35
FIG . 42. The curves for 9 - I ( 6–10 of Fig . 41 ) shifted so that similar percents
correct fall approximately on the same points .
for the middle group as closely as may be for such part of

the range from zero to forty correct as they have in com
mon . The result is shown in Fig . 43. A curve ( Fig . 44b )
representing the central tendency of all the five in Fig. 43
is drawn . The curve of Fig . 44b is then shifted down so as
to Fig 44a over that part the range which they have
of
fit
.
in common and an average of the two is drawn . This is

Fig . 45 , which represents the most probable general form
of the curve of decrease in number correct with increase in
difficulty so far as groups 91 and 911 reveal it .
FIG . 43.
0
ó į
The curves for 9
10
16
-II ( 1-5 of
15
Fig .
20 25 30
41 ) shifted so that similar percents
This curve has heights as follows ( in terms of tenths of

019 ):
5 correct , +28 .
10 correct , +15 .
15 correct , + 8.
---
20 correct , 0.
25 correct , 8.
30 correct ,
- 18 .
35 correct , 32 .
25
15
!0 5
1ļ 10 15 20 25 30 35
FIG . 44. The central tendency of the five curves of Fig . 42 ( a ) , and the cen-
tral tendency of the five curves of Fig . 43 ( b ) .
Fig.
40 and Fig . 45 are very closely alike in form , and
each is well represented by a curve with heights ( in terms
of tenths of σ19 ) as follows :
5 correct , 30 .
10 correct , +18 .
15 correct , + 8.
20 correct, 0.
8. -
-
25 correct ,
30 correct -
, 18.
35 correct , 30 .
All the cases of Groups 4 and 5 are grouped
I , J, and
into six
groups according to the sum of their scores in H ,
K. The six curves are drawn ( Fig . 46 ) , and shifted up or

down so that similar numbers correct fall on approximately
1
0 5 10 15 20 25 30 35
FIG . 45. The curves of Fig . 43 and Fig . 45 shifted so that similar percents
the same points ( Fig . 47 ) , and combined to form a curve
( Fig . 48 )
representing the central tendency for number cor-
rect in relation to increase in difficulty in the same way that
Fig. 40 was formed from the facts of Fig . 38. This curve
has heights as follows ( in terms of tenths of σ19 ) :
5 correct, +281.
10 correct , +15 .
15 correct , + 6 .
20 correct , 0.
25 correct, 7.
30 correct , -14 .
35 correct , -241.
The cases of Group im3 are grouped into five groups

according to the sum of their scores in A , B, C , and D ; and
the same procedure followed as hitherto . The facts appear
L-
123 4 5 6
K-
ī
H-
! 10 15 20 25 30 35 374
FIG . 46. Curves of percent correct in relation to difficulty ; Groups 4 and 5 .
in Figs . 49 , 50 , and 51. Fig . 51 has heights as follows ( in

terms of tenths of σ19 ) :
5 correct , 341 .
10 correct , +21.
15 correct , 91.
20 correct, 0.
25 correct, — 91 .
30 correct , 22 .
35 correct , -36 .
The cases of Group im6 are grouped into three groups
according to the sum of their scores in C , D , E, F , and G ;
and the same procedure followed . The facts appear in
Figs . 52 , 53 , and 54. Fig . 54 has heights as follows ( in

terms of tenths of σ19 ) :
5 correct, +16.
10 correct , +11 .
+ 6.
-
15 correct ,
20 correct , 0.
25 correct , 61.
30 correct , ― 151 .
35 correct ,
- 31 ( approx . )
The facts whence Figs . 38 to 54 are derived are pre-

sented in Table 127 .
•
0 5 10 15 20 25 30 35
FIG . 47. The curves of Fig . 46 shifted so that similar percents correct fall
on approximately the same points .
From Composite Q down to Composite , or from the I

College - Graduate group down through Grade 5 , there is a
notable similarity not only in the shape but also in the
amount of slope of the curves . From 5 correct to 35 cor-
rect equals 6.2019 at or near Composites O and P , 6.00₁9 at
or near Composites L and M, and 5.3019 at or near Com-
posites J
and K. At lower levels there is considerable
5 10 15 20 25 30 35
in difficulty , for Groups 4 and 5 .
disparity , the spread around Composites B and C being

7.0019 , while that around Composites E, F , and G is only
about 4.8019 . The determinations around levels A to G
are much less reliable than those from to Q. A combina- I
tion of Figs . 51 and 54 with equal weight gives ( in terms
of tenths of σ19 ) + 25 , +16 , +8 , 0 , , -19 , and-8 34 , -
with a spread of 5.9019 .
ESTIMATING THE CAVD ALTITUDE OF AN INDIVIDUAL

From I to Q the altitude at which an individual will
have exactly twenty right out of forty can be estimated
from Table 128 , which approximately represents a curve
of the general form of Figs . 40 , 45 , and 48 , running from
+3.0019 at 5 correct to -3.0019 at 35 correct .
12 3 4
D
C-
B-
A-
! 5 10 15 20 25 30 35
FIG . 49. Curves of percent correct in relation to difficulty ; Group im 3.
Table 129 is a similar table for Composites A to H,

which may be used provisionally until the difficulties of
these tasks and the nature of the relation between difficulty
and number correct at these levels are more accurately de-
termined .
5
.
ļ
FIG . 50.
1ļ 10
10 15 20
1
25 30
The curves of Fig. 49 shifted so that similar percents
approximately the same points .
35
correct fall on
0 1ļ 10 15 20 25 30 35
in difficulty , for Group im 3.
!!
67 8
!
E—
!
! 5
10 10 15 20 25 30 35
FIG . 52. Curves of percent correct in relation to difficulty : Group im 6.
\6
1ļ ฀ 15 20 25 30 35
FIG . 53. The curves of Fig .52 shifted so that similar percents correct fall
on approximately the same points .
Any 40 - composite task at which an individual has more

than zero and less than forty of the single tasks correct
can be used to estimate the level or altitude at which he
would probably have exactly 20 right . Scores near zero or
forty deserve , however , relatively little weight , because the
former are so much affected by chance successes in the case
of the vocabulary tasks , and the latter are so much affected
by careless slips.
A simple and impartial procedure which works very well
in general is to utilize for any individual the three succes-
1
10 5 10 15 20 25 30 35
in difficulty , for Group im 6.
sive levels whose sum of rights is nearest to 60 , and to attach

equal weight to each of them . This is the procedure used
in all the measurements of altitude which are reported in
later chapters .
A possibly better procedure is to take the two successive
levels whose sum of rights is nearest 40 , and the neighbor
ing level above or below which is nearest 20. This seems to
be somewhat quicker in operation , and is less likely to in-
clude levels with extremely low or high percents correct .
TABLE 127
(
)
-
MEDIAN OF THE SCORES NUMBER RIGHT OUT OF 40 FOR EACH OF TWENTY NINE GROUPS WITH EACH OF FOUR OR MORE NEIGH-
. .
BORING COMPOSITES
Composite Task
C
G
F
I
.J
Group A B D E H K M Q
370
L N Р
5
im3a 15 2.5
b
1 2
24 13.1
c 32 21 11 4.3
d
36 26.8 16 7.3
e 38 34 21 12.2
©
im6a 27 16.2 12.2
b
0 2
31 22.9 18.3
499
c 34 30.3 27 19 9.8
+
4
5a 21.7 8.9 4.1 3.3
b
28.8 14.6 6.5 3.4
с 32.7 18.3 8.8 4.2
THE MEASUREMENT
d
34.3 22.6 11.9 5.8
e 36.6 27.4 14.4 7.5
f
37.4 32.0 19.9 10.7
6
91a 30.0 19 14.5 4.0
b
32.5 23 16.7 10 6.2
c 36.0 28 22.4 16 8.3
d
37.5 31 27.3 23 11.0
e 39.0 35 30.3 30 18.3
OF INTELLIGENCE
9IIa 15.3 7.3 5.0 3.0 1.7
b
20.0 13.5 7.3 6.3 2.7
с 23.8 17.8 12.6 7.3 3.0
d
24.8 22.8 16.5 9.3 4.3
e
30.4 28.5 23.2 14.4 7.9
+
13 17a 16.3 9.7 6.5 4.0
b 24.8 17.2 12.5 6.0
с 28.7 22.6 18.4 10.5
32.0 27.7 24.0 16.5
35.1 31.3 29.7 21.3
TABLE 128 .
ALTITUDES CORRESPONDING TO ANY NUMBER CORRECT FROM 5 TO 35 OUT OF 40
FOR TASKS I TO Q.
I J K L M N O P Q
31.2 33.1 34.2 35.2 36.2 36.9 38.1 38.7 40.1

31.4 33.3 34.4 35.4 36.4 37.1 38.3 38.9 40.3
31.7 33.6 34.7 35.7 36.7 37.4 38.6 39.2 40.6
9 5678
31.9 33.8 34.9 35.9 36.9 37.6 38.8 39.4 40.8

a 32.2 34.1 35.2 36.2 37.2 37.9 39.1 39.7 41.1
o
10 32.4 34.3 35.4 36.4 37.4 38.1 39.3 39.9 41.3
11 32.6 34.5 35.6 36.6 37.6 38.3 39.5 40.1 41.5
12 32.8 34.7 35.8 36.8 37.8 38.5 39.7 40.3 41.7
13 33.0 34.9 36.0 37.0 38.0 38.7 39.9 40.5 41.9
14 33.2 35.1 36.2 37.2 38.2 38.9 40.1 40.7 42.1
15 33.4 35.3 36.4 37.4 38.4 39.1 40.3 40.9 42.3
16 33.6 35.5 36.6 37.6 38.6 39.3 40.5 41.1 42.5
17 33.7 35.6 36.7 37.7 38.7 39.4 40.6 41.2 42.6
18 33.9 35.8 36.9 37.9 38.9 39.6 40.8 41.4 42.8
19 34.0 35.9 37.0 38.0 39.0 39.7 40.9 41.5 42.9
20 34.2 36.1 37.2 38.2 39.2 39.9 41.1 41.7 43.1
21 34.4 36.3 37.4 38.4 39.4 40.1 41.3 41.9 43.3
22 34.5 36.4 37.5 38.5 39.5 40.2 41.4 42.0 43.4
23 34.7 36.6 37.7 38.7 39.7 40.4 41.6 42.2 43.6
24 34.8 36.7 37.8 38.8 39.8 40.5 41.7 42.3 43.7
25 35.0 36.9 38.0 39.0 40.0 40.7 41.9 42.5 43.9
26 35.2 37.1 38.2 39.2 40.2 40.9 42.1 42.7 44.1
27 35.4 37.3 38.4 39.4 40.4 41.1 42.3 42.9 44.3
28 35.6 37.5 38.6 39.6 40.6 41.3 42.5 43.1 44.5
29 35.8 37.7 38.8 39.8 40.8 41.5 42.7 43.3 44.7
30 36.0 37.9 39.0 40.0 41.0 41.7 42.9 43.5 44.9
31 36.2 38.1 39.2 40.2 41.2 41.9 43.1 43.7 45.1
32 36.5 38.4 39.5 40.5 41.5 42.2 43.4 44.0 45.4
33 36.7 38.6 39.7 40.7 41.7 42.4 43.6 44.2 45.6
34 37.0 38.9 40.0 41.0 42.0 42.7 43.9 44.5 45.9
35 37.2 39.1 40.2 41.2 42.2 42.9 44.1 44.7 46.1
In the rare cases where is desirable to estimate altitude from score

it
below the following may be used For and

or
above 35 subtract
5
3
1,
2,
4,
:
4.3 3.9 3.6 and 3.3 respectively from the score for 20. For 36 37 38 and 39
,
add 3.3 3.6 3.9 and 4.3 respectively the score for
20
to
,
.
TABLE 129 .
APPROXIMATE PROVISIONAL ALTITUDES CORRESPONDING TO ANY NUMBER CORRECT

FROM 5 TO 35 OUT OF 40 FOR TASKS A TO H.
A B C D E F G H
5 20.0 23.2 25.3 26.7 27.8 28.3 29.0 29.8

6
16 20.2 23.4 25.5 26.9 28.0 28.5 29.2 30.0
7 20.5 23.7 25.8 27.2 28.3 28.8 29.5 30.3
8 20.7 23.9 26.0 27.4 28.5 29.0 29.7 30.5
9 21.0 24.2 26.3 27.7 28.8 29.3 30.0 30.8
10 21.2 24.4 26.5 27.9 29.0 29.5 30.2 31.0
11 21.4 24.6 26.7 28.1 29.2 29.7 30.4 31.2
12 21.6 24.8 26.9 28.3 29.4 29.9 30.6 31.4
13 21.8 25.0 27.1 28.5 29.6 30.1 30.8 31.6
14 22.0 25.2 27.3 28.7 29.8 30.3 31.0 31.8
15 22.2 25.4 27.5 28.9 30.0 30.5 31.2 32.0
16 22.4 25.6 27.7 29.1 30.2 30.7 31.4 32.2
17 22.5 25.7 27.8 29.2 30.3 30.8 31.5 32.3
18 22.7 25.9 28.0 29.4 30.5 31.0 31.7 32.5
19 22.8 26.0 28.1 29.5 30.6 31.1 31.8 32.6
20 23.0 26.2 28.3 29.7 30.8 31.3 32.0 32.8
21 23.2 26.4 28.5 29.9 31.0 31.5 32.2 33.0
22 23.3 26.5 28.6 30.0 31.1 31.6 32.3 33.1
23 23.5 26.7 28.8 30.2 31.3 31.8 32.5 33.3
24 23.6 26.8 28.9 30.3 31.4 31.9 32.6 33.4
25 23.8 27.0 29.1 30.5 31.6 32.1 32.8 33.6
26 24.0 27.2 29.3 30.7 31.8 32.3 33.0 33.8
27 24.2 27.4 29.5 30.9 32.0 32.5 33.2 34.0
28 24.4 27.6 29.7 31.1 32.2 32.7 33.4 34.2
29 24.6 27.8 29.9 31.3 32.4 32.9 33.6 34.4
30 24.8 28.0 30.1 31.5 32.6 33.1 33.8 34.6
31 25.0 28.2 30.3 31.7 32.8 33.3 34.0 34.8
32 25.3 28.5 30.6 32.0 33.1 33.6 34.3 35.1
33 25.5 28.7 30.8 32.2 33.3 33.8 34.5 35.3
34 25.8 29.0 31.1 32.5 33.6 34.1 34.8 35.6
35 26.0 29.2 31.3 32.7 33.8 34.3 35.0 35.8
CHAPTER XII
THE MEASUREMENT OF WIDTH AND AREA OF INTELLECT
The width or range of intellect at any altitude or level

of difficulty is measured by the number of tasks mastered
at that altitude . Thus , suppose that Intellect X is mea-
sured with ten 40 - composite tasks ( N1 , N2 , N3 , etc. ) , each
equal to Composite N in difficulty ; and has the following
score :
Number of single tasks right in N₁ = 20.
66 "" 66 66 66
N2 = 19 .
66
Ng =
66 66 "" 66 "" 66
N3 = 21 .
N₁ = 20 .
66 "" 66 66 66
66 ฀
66 66 66 66 66
N5 = 20 .
N. =
฀
-=
66 66 66 66 "" 66
. No =
18 .
66 ฀
66 "" 66 66 66
, =
N = 19 .
66 "" 66 66 66 66
Ng = 21.
66 66 "" 66 ""
N, 22 .
66 "" 66 66 66 66
N10 = 20 .
Success at one of these 40 - composite tasks means attain-
ing 20 or more single tasks correct . The width of Intellect
X at Altitude N is 7 out of 10 for Tasks N₁1 to N10 . It may
also under certain conditions be considered as 200 out of
400 for the single tasks composing N₁1 to N10 , or as a certain
number out of 40 for the same single tasks grouped in 10-
composites , or as a certain number out of 100 for the same
tasks grouped in 4- composites .
WIDTH OF INTELLECT IN THE CASE OF TRULY INTELLECTUAL

TASKS
Consider first the first and most correct meaning , that

is , the number of composite - tasks correct , here 7 out of 10
for Tasks N₁ to N10 . If
the ten are a representative sam-
pling of tasks of intellectual difficulty N , Intellect X may
373
be expected to have approximately 70 successes out of 100 ,

or 700 out of 1,000 , or in general approximately 70 percent
of successes with tasks at the intellectual altitude of N. If
there are 200 such tasks , his probable width is 140 ; if there
are 60,000 such , his probable width is 42,000 . If
, when mea
sured in respect of ten 40 - composite tasks representative

of intellectual difficulty M , his scores are 25 , 25 , 22 , 24 , 26 ,
may be expected to have 100 per
25 , 23 , 24 , 26 , and 21 , he
cent of successes with tasks of intellectual difficulty M. If
there are 150 such tasks , his probable width is 150. there If
are 40,000 , his probable width is 40,000 .
This illustration directs our attention to two meanings
of width , namely , width of intellect in the sample examined
and width of intellect in the entire series which the sample
represents ; and also to the fact that the sample examined
may have a larger representation of tasks at one altitude
than of tasks at another .
Suppose , for example , that the sample contains
40 single tasks between difficulty 30.0 and 30.99 ,
66 66 66 66
40 35.0 and 35.99 , and
66 66 66 66
40 40.0 and 40.99 , and
that there really are one million CAVD tasks between 30.0
and 30.99 , two million between 35.0 and 35.99 , and three mil
lion between 40.0 and 40.99 . Then the sample has twice as
large a representation of level 35.0 to 35.99 as it has of level
40.0 to 40.99 , and three times as large a representation of
level 30.0 to 30.99 as it has of level 40.0 to 40.99 . If
an indi
vidual can do 9 out of 10 of the sample at level 30.0 to 30.99 ,
he can probably do 500,000 tasks at that level . But if he
can do half of the tasks of the sample at level 40.0 to 40.99 ,
he can probably do 1,500,000 tasks at that level .
Ifeach of the tasks , the number of which measures
width , is perfectly intellectual , depending for success upon
all of intellect and nothing but intellect , the change from
one hundred percent of successes to zero percent of suc
cesses , as the intellect in question is tested at higher and
THE MEASUREMENT OF WIDTH AND AREA 375
higher altitudes , will be instantaneous . When a small

amount of inadequacy and error is present , as in our 40-
composites for Intellect CAVD , the change will still be very
sudden . The conditions in representative intellects , each
measured by a score or more of tasks like our 40 - composites
P
O
N
II
M
L
Π
K
‫ل‬
I I
0 10 20 30 40 50 60
I
70 80 90 100
FIG . 55.The probable percentages of successes II
of three intellects , I ,
and III ,
in a series of 360 tasks , 20 of difficulty A, 20 of difficulty B , and
so on , each task having r₁₁
it = approximately .9 . The drawings are not from
precise computations , being for illustration only , not for mensuration of the
effect .
26
at each altitude in Intellect CAVD , will be roughly as shown

in Fig . 55 .
The evidence for this is the correlations between one 40
composite and another at or near the same level , and the in
frequency of reversals from failure to success in our series
of tasks . For example , in the 240 individuals of Group 17 ,
of those failing with P ( 103 in all ) , only 4 or 3.9 percent
succeeded with Q , which is 1.1 harder .
In the 246 individuals of Group 91 , of those failing with
K ( 93 in all ) , only 9 or 9.7 percent succeeded with L , which
is 1.0 harder .
The measurement of CAVD width at any altitude , in the
rigorous sense of number of intellectual tasks mastered at
that altitude , is thus given for most altitudes by the mea
surement of altitude itself . Nearly up to that altitude the
percent is one hundred ; above it the percent very soon
drops to zero . Within the short distance of uncertainty
the widths may be determined by experiment or estimated
fairly closely from the altitude .
This will hold true of any sort of intellect defined and
treated in the same manner as Intellect CAVD . In propor
tion as each task depends for success upon all of intellect
and nothing but intellect , a smaller and smaller increase in
difficulty will cause a shift from success to failure , the alti
tude where it does so varying with the intellect that is being
measured .
WIDTH OF INTELLECT IN THE SENSE OF THE NUMBER OF SINGLE
SHORT TASKS MASTERED , ANY ONE OF THESE TASKS BEING

ONLY A VERY PARTIAL REPRESENTATION OF INTELLECT
For many purposes it is desirable to know how many

single tasks from a set which are nearly or quite alike in
difficulty and which are nearly or quite as intellectual as
any short single tasks can be , a given intellect can succeed
with . If
, for example , two intellects A and B have identical
CAVD altitudes exactly at Level N , and if A has average

scores at Levels K , L , M , N, O, P , and Q of 39 , 36 , 29 , 20 , 17 ,
11 , and 6, whereas B has scores of 30 , 28 , 27 , 20 , 18 , 6 , and 0 ,

there is a difference between A and B which may need ex-
pression . Between 40 and 20 right , and between 19 and
zero right in the case of such 40 - composites as the CAVD
series , there are ranges of difference which may be of great
importance for theory or for practice or for both .
The measurement is , of course , a simple count of suc-
cesses in the sample used in the examination , and an esti-
mated count for the entire series which is represented by
the sample . If
the single tasks in K represent a selection
of 40 out of 10,000 , while those in L represent a selection of
40 out of 15,000 , and those in M represent a selection of 40

out of 25,000 , A's scores of 39 , 36 , and 29 in the examina-
39 36
tion mean probabilities of success with 40 ×10,000,40x
40
29
15,000 , and X25,000 , or with 9,750 , 13,500 , and 18,125
40
single tasks of the sort chosen as components of Com-
posites K
, L , and M , respectively .
A series of names is needed to designate different sorts

of width , from the width of an intellect in perfectly intel-
lectual tasks , down through its width in various composite
tasks less and less representative of all of intellect and
nothing but intellect , to its width in such tasks as giving the
opposite of one word , or understanding one sentence , or
tracing a way through one maze , or repeating one series of
five digits backward . We suggest the use of a series of
W's , each followed by a notation describing the tasks , and
being in each case the percent of successes .
Thus , W ( 10C + 10A + 10V + 10D ) would refer to the
percent of successes with 40 - composite tasks made up
equally of C , A, V , and D ; W ( 1C or 1A or 1V or 1D ) N
would refer to the percent of successes with a series of tasks
made up of single C's , A's , V's , and D's . W ( 10M ) would
refer to the percent of successes with a series of composite
tasks each made up of ten mazes . The altitude at which W
is measured will require very careful description in every
case .
AREA OF INTELLECT
Area or volume seems the best term to use to mean the
total number of tasks of some specified sort at which an in-
tellect succeeds ; and area seems preferable . Area , like
width , will have two distinct meanings , namely , the number
of successes in the sample set of tasks examined , and the
number of estimated successes in the entire inventory of
tasks which have been or can be made , and of which the ex-
amination -tasks are a representative sample .
Area of intellect , like width , is , in the strictest usage ,
the number of truly intellectual tasks , each of which mea-
sures all of intellect and nothing but intellect . In this sense
the area found will be a function of the altitude ; Intellect
X , of Altitude N , will succeed with all tasks up to that alti-
tude , and with none beyond it .
As in the case of width , it will be desirable to use area of
intellect in a loose sense to mean the total number of tasks
mastered which are proper components of composites
which , as totals , are intellectual , all the way down from
composites which are nearly perfectly intellectual to short
single tasks like the single C's , A's , V's, and D's . A nota-
tion like A ( 10C + 10A + 10V + 10D ) , A ( 1C or 1A or 1V
or 1D ) N , A ( 10M ) , and the like may usefully be adopted to
describe the kind of " area " that is being measured .
We shall consder as a typical case the measurement of
A ( 1C or 1A or IV or 1D ) . Everything is simple so far as
concerns finding this area for the sample examined . But
the effort to estimate the area as a fraction of all the dif-
ferent sentence - completions that might be desired , all the
different arithmetical problems which could be collected or
invented , all the word -knowledge tasks ( Shall other than
English words be used ? ) possible , and all the sentences or
paragraphs or books that might be heard or read , and so to
estimate effective A ( 1C or 1A or 1V or 1D ) brings us up
squarely against great difficulties due to lack of knowledge
of the relative frequency of different C's , A's , V's , and D's
at different levels of difficulty .
If
we know the width of an intellect at each level in an
adequate sample of tasks , we can measure its total " area , 99
'
provided we know the number of tasks at each level . Thus ,
if the C , A, V , and D single tasks of Intellect CAVD at
levels zero to forty¹ number , in order , 100 , 100 , 100 , 100 , 100 ,
200 , 200 , 200 , 200 , 200 , 300 , 300 , 300 , 300 , 300 , 400 , 400 , 400 ,
400 , 400 , 500 , 500 , 500 , 500 , 500 , 700 , 700 , 700 , 700 , 700 , 1000 ,
1000 , 1000 , 1000 , 1000 , 2000 , 2000 , 2000 , 2,000 , and 2000 , and
if Intellect JS , when measured with a representative sam-
pling of 40 at each level , scores 40 at each level up through
level 30 , and 38 , 32 , 24 , 20 , 10 , 4 , 0 , 0 , 0 , 0 in order there-
after , we find his A ( 1C or 1A or 1V or 1D ) as 14,200 out of
a possible 26,000 . If there had been 650 tasks at each level ,
the same record in the examination would have meant
21,580 out of a possible 26,000 .
Such a computation of the area of an intellect would not

be a mere theoretical curiosity or statistical tour de force ,
but would be a systematic and accurate way of measuring
something of very great importance . Common - sense
thought and action about intellect often deal with some-
thing which this concept of area makes definite and objec-
tive . Just as terms like acuity , originality , and intellectual
genius refer to intellect with especial emphasis on its alti-
tude , so terms like breadth , scope , and intellectual power
refer to intellect with especial emphasis on its " area . " We
should not expect common sense to make clean - cut distinc-
tions or to avoid confusions , for the very good reason that
altitude and area are closely correlated , so that for most
practical purposes , we can describe a man's intelligence
adequately by simply rating him for intelligence as a unit .
But the concept of a man's general average probability of
correct response to intellectual or semi - intellectual tasks
has been real and useful ; and it will be more so now that it
can be made definite and measurable .
1 Level 0 includes all C , A , V or D tasks from 0 difficulty up to a difficulty

of 1.00 , 1 includes all from 1.00 up to 2.00 , 2 includes all from 2.00 up to 3.00 ,
and so on .
It
is possible to discover approximately the number of
single tasks at each level of Intellect CAVD or any other
defined intellect , though such estimates are beset by many
difficulties . The enumeration of the C or A or V or D tasks
harder than the average of those in Composite N and easier
than the average of those in Composite O is , indeed , prob
ably comparable in complexity to the enumeration of all
the species of animals .
The chief and most obvious difficulty is that of deciding
how much one task must differ from another in order that
they shall be counted as two rather than one . Consider , for
example , these fourteen tasks to be given orally :
1. John is now. How old will he be in 3 years ?
5 years old
2. Tom is now. How old will he be in 3 years ?
5 years old
3. John is 5 years old now . Tom is 3 years older than
John . How old is Tom ?
4. John is 5 years old . Will is 3 years older than John .
How old is Will ?
5. John has 5 cents now. How much will he have if his
father gives him 3 cents ?
6. John has 5 cents now . How much will he have if his
mother gives him 3 cents ?
7. How many dollars are five dollars and 3 dollars ?
1a . John is 6 years old now. How old will he be in 3 years ?
2a . Tom is 6 years old now . How old will he be in 3 years ?
3a . John is 6 years old now. Tom is 3 years older than
John . How old is Tom ?
4a . John is 6 years old . Will is 3 years older than John .
How old is Will ?
5a . John has 6 cents now. How much will he have if his
father gives him 3 cents ?
6a . John has 6 cents now. How much will he have if his
mother gives him 3 cents ?
7a . How many dollars are 6 dollars and 3 dollars ?
How many different tasks are there ? All competent

students of intellect will deny that there are fourteen .
By any reasonable view, we should not count 2 as a dif
ferent arithmetical task from 1. Whether the problem is
put about John or Tom or Will or Mary, does not , we

think , make any difference to it as an arithmetical or intel
lectual task . Our thinking is probably sound , and we shall
later state the facts and principle which justify it . But
note that if we think in a stiff pseudo -logical way that the
name of the boy makes no difference , we shall err . Let
Tom , well known to be of age ten , be sitting in full sight
and the task is now not quite the same , requiring for suc
cess that the intellect shall not be misled by the temptation
to think of the present Tom . Or let the problem be stated
as " Sneezer Snoop Squibb is 5 years old now. How old
will Sneezer Snoop Squibb be in 3 years ? " and the task is
not quite the same , requiring that the intellect be not dis
tracted by the seductive name into inattention to the num
bers .
If a psychologist should list all the arithmetical tasks
that ever have been set , and add to them all that a decade
of ingenious thought could devise , and then try to cull out
the duplicates , he would find some that would be indubitably
so , and some that would be as unlike as arithmetical tasks
can be ; but with many he could only say that the two tasks
were somewhat different . So he would have to set up some
standard of the amount of difference which would qualify
two tasks to be counted as two , or some scheme for frac
tional counts .
The facts which he should use for these purposes are the
facts of the differences of the tasks as tasks for intellects .
By this is meant not merely that the facts are facts in the
minds or neurones of individuals , but also that they are
facts in the action of intellects to which the tasks are pre
sented for solution . Two sentences differing in print only
by an apostrophe or comma may differ enormously in the
intellectual actions which they evoke in an intellect set upon
solving them , and two questions which have not a word in
common may arouse very similar behavior , as is the case
with " Solve 2x² + x = 21 " and " What does y equal if
2y² + y = 21 ? " for competent students of algebra . And
either may arouse very different behavior according as the

person reacts to it as a task to be solved or as , say , a mere
question to contemplate .
So the investigator seeking to measure the differences

amongst tasks ( apart from differences in difficulty ) must
be expert in the psychology of thinking , and must be skilful
in examining and cross - examining individuals who do the
tasks in question and report what they did . He will often
have to make subtle distinctions in cases where two tasks
arouse different action in two intellects , and when it is
doubtful how much of the difference lies in the tasks and
how much in the intellects .
The objective method of correlation will be helpful . Ob-
viously , if an intellect can do task 1 and cannot do task 2 ,
then the two tasks are different for that intellect ; two tasks
are not perfectly alike as tasks for intellect , unless every
intellect that can do one can do the other . Other things
being equal , the more individuals there are within any given
group who can do the one task and not the other , the greater
the difference between the two ( for that group ) will be .
More generally, the differences with which we are concerned
here are measured , other things being equal , by the lack of
perfect correlation between the ability to succeed with one
and the ability to succeed with the other , in some defined
group of intellects , difficulty being kept constant . If two
tasks are identical as tasks for intellects , rt1t2 will be 1.00 .
If they are of equal difficulty , the more unlike they are the
lower rt1t2 will be until it reaches a minimum which repre-
sents the amount of likeness which two tasks must have to
be included in the series of arithmetical tasks which is to
be enumerated .
This argument from correlation will not hold good if a

task is a composite where success is defined as obtaining a
certain percent of successes with the elementary tasks , or
attaining a certain score by some system of credits . Two
such composites may show very high correlations in respect
of success as just defined , and yet have hardly a single
detail of one like any detail of the other . The correlations

are between scores , each of which measures chiefly ability
in what is summed up in or common to all the single tasks
of the composite , not what is characteristic of any one of
them as a totality . The composites are closely alike in the
sense that what is summed up in or common to all the single
tasks of A , is closely like what is summed up in or common
to all the single tasks of B.
PROPORTIONAL COUNTS
For some purposes , the relative numbers of tasks at the

different levels of difficulty will serve in place of the abso-
lute numbers . Thus , if we wish to know what percent of
A's area B's area is , we will do as well by knowing that the
numbers of tasks are in the proportions K , 3K , 9K , 27K ,
81K , as by knowing their absolute amounts .
It may well be that such proportional counts may be
made with greater accuracy , as well as with greater ease
and speed , than absolute counts . Certain factors of error
may act alike at all levels and so do no harm to the propor-
tional counts . Certain arbitrary schemes of fractional al-
lowance for overlapping tasks may also act alike at all
levels and so do no harm to proportional counts . Even such
proportional counts , however , will require much sagacity
and industry to achieve even approximate truth for even a
small fraction of intellect . A reasonably satisfactory pro-
portional count of the number of tasks at each level of even
so small a representation of intellect as CAVD is , will in-
deed require an enormous amount of observation and ex-
periment . New tasks , like new species of animals , are
coming into existence while we count them ; tasks a and c
seem enough different to count as two , and tasks b and d
seem enough different to count as two , but when a , b , c , and
d are considered together , they do not seem to deserve a
credit of four ; it seems as if some sorts of tasks at some
levels of difficulty were innumerable ; when task X is simply
a task where both a and b must be performed successfully
to bring success in X, shall we count a and b and X as 3 or

as 2 ? These difficulties , together with those which have
been mentioned and others which might be , may make per-
fect or even approximately correct counts impossible .
The best way to find out what is possible is to begin
work at actual counts . We have begun , but have not
progressed far enough to report results , save one . This
one , which the reader's sagacity may have anticipated , is
that the number of different tasks per unit of altitude of
intellect is not equal , but increases as we go up from zero
altitude .
That this is true for sentence completions can be easily

realized if one will try to make as many different C tasks
as he can between the average difficulty of those in A and
the average difficulty of those in B ( 23 to 261 ) , and to do
the same for the stretch of difficulty from N to Q ( 40 to 43 ) .
It will be found very hard and perhaps impossible to devise
five hundred of the former , whereas there seems almost no
limit to the possible number of the latter . Apparently the
harder the task, the greater the number possible , though it
is not easy to devise extremely hard completions which are
linguistic rather than informational in their difficulty .
In the case of the arithmetical tasks the number of dif-
ferent tasks surely increases from the very easy levels up
to a certain point , after which there is some doubt . The
doubt seems , however , to be due mainly to our averseness
to fabricating problems whch are so elaborate and intricate
as to be extremely unreal , rather than to the paucity of
such . In the case of the disarranged - equation task , it is
obvious that the number of different ones possible increases
rapidly with increased difficulty and has no limit .
In the case of vocabulary, the fact is unquestioned if
other languages than English are included , and probably
holds true for English alone .
In the case of the understanding of sentences and para-
graphs , the increase is obscured by the facts that people
usually try to make their statements as easily intelligible
as may be , and that the number of persons who are con

cerned with very subtle and intricate ideas is few . Also ,
the number of different statements and questions of even a
moderate degree of difficulty which can be fabricated is so
enormous that comparisons are very difficult . Also , only
persons of very high directions ability can frame state
ments which are sensible and correct but very hard to un
derstand , and still free from any great informational dif
ficulty . Sentence comprehension cannot , however , well be
kept distinct from informational abilities ; and if informa
tional difficulties are allowed to enter freely , the number of
sentences very hard to understand is practically infinite .
Even if one abstains from these rather rigorously , the num
ber of very hard D's that can be made is enormously
greater than one would expect from the number found in
reading . Merely by combining and permuting causal , con
ditional , and concessive clauses and pronoun references ,
one can produce an enormous number of different tasks
"
like , A change in ab would cause a similar change in og if
ek did not produce its usual effect upon il , although ek did
act upon um , and ba would cause an increase in ab , provided
bi did not occur in unison with bo . What will happen to og
if ba and bi and ek happen shortly subsequent to bo , pro
vided the ek-il action is neutralized by bo , and um does not
occur ? "
We have not even begun a count for the entire series of
tasks which might reasonably be made constituents in com
posites designed to measure intellect in general . Conse
quently , we are not able to make more than a very rough
estimate of how much number increases with altitude , or of
the way in which the increase comes . We think the increase
for Intellect CAVD is so great as to make the number of
different tasks at level 40 to 40.99 at least a hundred times.
the number at level 20 to 20.99 . We also think that it comes
smoothly and with acceleration , at least up to a certain
In
level , after the pattern of Fig . 56 or Fig . 57 or Fig. 58.
tellect CAVD can hardly be said to have an appreciable
42
40
38
36
34
32
30
28
26
24
22
20
50
48
46
44
42
40
38
36
34
32
30
28
26
24
22
20
40
38
36
34
32
30
28
26
24
22
20
FIGS . 56 , 57 and 58. Samples of possible patterns of the increase in the num
ber of different intellectual tasks with increase in intellectual difficulty .
386
area below level 20 , since it probably requires an altitude

of 20 to complete ten out of any twenty sentences , no matter
how easy , or to solve ten out of any twenty arithmetical
tasks . The increase for intellect in general will be found ,
we think , to increase to a similar degree and in a similar
manner , with at least fifty times as many tasks at 40 as at
20 , and at least several hundred times as many at 40 as at 10 .
An intellect of altitude 40 may then have an area , not
twice that of an intellect of altitude 20 , but ten or twenty or
perhaps two hundred times it . The common - sense view
that the greatest intellect of a thousand men is many times
as great as the worst intellect of the thousand may be en
tirely correct , if we mean by " great " something corre
sponding to area .
Moreover , if we think of intellect as a hierarchy of unit
connectionsor bonds between ideas or between the neural
correspondents of ideas , the number of different connec
tions required to enable a person to respond correctly to 20
out of 40 of the elements of task N at level 40 may be not
twice the number required to enable one to respond cor
rectly to 20 out of 40 of a task 3 below A , but ten or twenty
or two hundred times it .
Intellectual altitude , by our definition , shows a small
relative rise from the imbecile to the average and then to
the gifted adult , by the argument followed in Chapter IX ,
so small as to arouse astonishment and incredulity concern
ing the usefulness of the definition and the validity of the
argument , at first thought . If , however , the altitudes of the
imbecile , average and gifted were in the proportions of 5 ,

15 and 20 or 1 11 and 16 instead of about 25 , 35 , and 40 ,
, , , , ,
we might find the relative areas of intellect in the three

groups much more preposterous in the reverse way . The
scale of altitude must not be criticized for the lack of at
tributes which are appropriate only for a scale of area ,
unless it can be shown that width is approximately the
same at all altitudes . It is not .
CHAPTER XIII
THE RELATIONS OF ALTITUDE TO WIDTH , AREA , AND SPEED
The number of CAVD tasks at any given level of diffi-
culty is unknown . Consequently all the relations with
width which are considered in this chapter are relations
with percents . No comparison or conclusion will appear
which involves the absolute number of tasks in two levels .
THE RELATION BETWEEN ALTITUDE AND W ( 10c + 10a + 10v +

10D ) , I.E. , NUMBER OF 40 - COMPOSITE CAVD TASKS SUC-
CEEDED WITH AT A GIVEN LEVEL OF DIFFICULTY
N individuals are measured , each with , say , a score of

CAVD composite tasks , each composite being of the same
difficulty as any other , and each consisting of so many
single tasks that the correlation between the number right
in any one composite and the number right in any other is
perfect. Then any one of the N individuals who succeeds
with any one of these composites ( in the sense of having 50
percent or more of the single tasks correct ) will succeed
with any other of them ; and the W of any individual will
be one hundred percent or zero percent . Suppose that the
same N individuals are measured perfectly in respect of
altitude of Intellect CAVD . The correlation between alti-
tude CAVD and W ( 10C + 10A + 10V + 10D ) will be per-
fect , every one of the individuals who succeeds with these
composites having a higher altitude than any one of those
who fail with them . If each task at a certain level of diffi-
-
culty is extensive enough to represent and measure all of
CAVD difficulty and no other difficulty all of CAVD intel-
-
lect as it operates with tasks at that level of difficulty and
nothing but it then everyone who succeeds with these will
have a CAVD altitude as high as , or higher than , the alti-
tude which they represent and no one who fails with them
388
ALTITUDE , WIDTH, AREA , AND SPEED 389
will have a CAVD altitude as high as the altitude which

they represent . That is , if each task measures all the
CAVD intellect which can operate at that level and nothing
but it , the percent of tasks mastered at that level will be
zero or one hundred and will correlate perfectly with alti
tude CAVD .
Stated in another way , any individual who succeeds with
any task of difficulty d which measures CAVD perfectly as
it operates at that level of difficulty , will succeed with all
tasks of less difficulty than d , if these also measure CAVD
perfectly as it operates at their respective levels of diffi
culty ; and any individual who fails with any task of diffi
culty d will fail with all tasks of greater difficulty than d ,
if these also measure CAVD perfectly as it operates at
their respective levels .
These are not axioms necessitated logically by the defi
nition of Intellect CAVD and of difficulty CAVD ; but con
clusions reached by observations of facts . The facts could
be otherwise . Some men might conceivably succeed with
tasks like O , P , and Q and fail with tasks like M , N , and O.
We do not give an absolute empirical proof of these con
clusions , because we have not any tasks which measured all
of the CAVD intellect which operates at any given level of
difficulty . All the evidence , however , goes to prove their
truth .
Evidence may be found in the correlation between the
altitude measure and the score of success or failure in
20 - composites ( 5C + 5A + 5V + 5D ) corrected for attenu
ation , so as to give the correlation between a precise mea
sure of altitude and the number of s's in an examination
with a very large number of such 20 - composites . For ex
ample , the average correlation ( bi - serial r ) of the mea
sure of altitude with success in a CAVD 20 - composite in the
case of 98 adult imbeciles was .984 for A , .916 for B , .875
for C , and .757 for D , averaging .883 . The self - correlation
of the altitude measure is .94 , the inter -correlations of the
three determinations whose average it is being .92 , .77 , and
.83 . The self - correlation ( tetrachoric r ) of a CAVD 20-

composite in this group is .96 for A , .76 for B , .79 for C ,
and .99 for D , averaging .87 . The correlation between a
precise measure of altitude and success in 50 percent or
more of a number of CAVD 20 - composites of equal diffi-
.984
culty may then be expected to be for A ,
V.96 .94
.916 .875 .757
for B, for C , and for D ,
V.76.94 V.77 X .94 V.99 X.94
.883
or, on the average , or .97 .
V.875X.94 '
,
Also the correlations between altitude and W ( 1C or 1A

or 1V or 1D ) are very near unity , as will be demonstrated
in the next section . The correlations between altitude and
W ( 10C + 10A + 10V + 10D ) a fortiori will be near unity .
In view of such evidence the conclusions stated in the
first two pages of this chapter may be accepted as true .
There is no reason to expect that the case will be different
with any fairly catholic form of intellect ( such as Picture-
Completions Opposites + Geometrical Relations + Rea-
soning Problems of the type devised by Burt + Informa-
tion ; or Analogies + Number - Completions + Arithmetical
Computation + a Common Element test of the type devised
by Otis ) from what it is with CAVD .
THE RELATION BETWEEN ALTITUDE AND W ( 1c OR 1A OR 1V OR

1D ) , I.E. , THE NUMBER OF SINGLE TASKS SUCCEEDED
WITH AT A GIVEN LEVEL
This correlation is very close . There are a certain num-
ber of individuals who are , relatively to others , much better
( or worse ) in arithmetical tasks than they are in the lin-
guistic tasks , and whose records prevent perfect correla-
tion . Also , there are probably other minor specializations
within Intellect CAVD . But on the whole , individuals
would be found to follow rather closely the general pattern
of CAVD intellect shown in Fig . 59 if each of them had
ALTITUDE , WIDTH , AREA , AND SPEED 391
been tested with several hundred tasks ( one - fourth being

C ; one - fourth , A ; one - fourth , V ; and one - fourth , D ) at
each level of difficulty from 0 to 44. In general , that is ,
if intellect A has a higher altitude than intellect B , intel-
lect A will also show a greater W ( 1C or 1A or 1V or 1D )
than B at all levels between those where both A and B have
one hundred percent right and those where both A and B
have zero percent right ; and the amount of superiority of
A to B in W will be closely similar to the amount of superi-
ority in altitude .
To prove this , we have to estimate the relation as it will
be found with a very large number of single tasks at the
level of difficulty in question , from data where this number
is only 40 or less . The evidence is as follows :
In the case of 237 individuals of group 17 , the correla-
tions between altitude CAVD and percent succeeding in
tasks N , O , P , and Q were as follows ( P means the Pearson
r ; Sh means the Sheppard r ) :
Р Sh
N .86 .76
O .93 .94
P .91 .91
Q .81 .86
Average .88 .87
The self - correlations for % s in N , O , P, and Q in this

group may be taken as approximately .76 , using the data
given in Appendix V , which show that the correlations of
neighboring 40 - composites average .73 in this group . .03
is added for the effect of the slight remoteness . The self-
correlation of the measure of altitude in this group is com-
puted as .90 from the intercorrelations of the three inde-
pendent measures of altitude of which it is the average .
They are .80 , .76 , and .71 , averaging .757 . By the well-
3 ( .757 )
known formula of Spearman , r 3 with 3 will equal
+ 2 ( .757 ) 1
By this determination , a precise measure of altitude will
27
correlate with a precise measure of W ( C or A or V or D )

.875
to the extent of or 1.06.
V.76 X.90
As a check on this determination , we have computed the
obtained correlation between the measure of altitude and
the sum of the numbers of rights in N , O , P , and Q. It is
.99 . The correlation between a precise measure of altitude
and a precise measure of W ( C or A or V or D ) should be
higher than this obtained correlation .
In the case of 189 individuals of group 13 , the correla
tions between altitude CAVD and % s in tasks N , O , P , and
Q were as follows :
P Sh
N .875 .84
O .925 .901
P .916 .89
Q .782 .83
Average .874 .866
The average self - correlation for % s in N , O , P , and Q

in this group may be taken as .74 , from the data given in
Appendix IV . The self - correlation of the measure of alti
tude in this group is found by the Spearman formula to be
.89 . The intercorrelations of the three independent mea
sures of altitude of which it is the average are .71 , .64 , and
.81 . The correlation between altitude and W ( C or A or V
or D ) , both being measured accurately , will thus be
87
or 1.07 . As a check , we have a correlation of
V.74 X.89
.95 between the obtained measure of altitude and the sum
of the numbers correct in N , O , P , and Q , and a part of M.
In the case of 246 individuals of group 91 , altitude
CAVD correlates with % s in composites , , K , L , and M I J
I J
as follows : .58 for , .82 for , .92 for K , .82 for L , and .64
for M ( all by the Sh formula ) . The self - correlations of % s
J
in I , , K , L , and M in this group are respectively .73 , .80 ,
.74, .86 , and .69 . The self - correlation of the measure of
altitude in this group is .79 , the intercorrelation of the

three measures of which it is the average being .56 , .58 ,
and .52 . It is perhaps unwise to average correlations such
as these which show wide and regular differences . So we
correct each for attenuation separately and have , as the
five resulting determinations of the correlation between
altitude and W ( C or A or V or D ) , .76 , 1.03 , 1.20 , 1.00 , and
.87 . The average of these is .97 ; the median is 1.00 . As a
check we have the correlation between the altitude measure
and the sum of the numbers correct in I , J , K , L , and M. It
is .91 .
In
the case of 192 individuals of group 9II , altitude
CAVD correlates with % s in composites K , L, M , and N as
follows : .73 for K , .90 for L , .91 for M, and .66 for N. The
average is .80 . The self- correlations of , L, M , and N are K
respectively .761 , .871 , .751 , and .75 , averaging .80 . The
self - correlation of the measure of altitude in this group may
be taken as .83 , the intercorrelations of the three measures
of which it is the altitude being .50 , .635 , and .73 . So a pre
cise measure of altitude will correlate with a precise mea
sure of W ( C or A or V or D ) to the extent of .99 ( .91 by K ,
1.07 by L, 1.15 by M, and .84 by N ) . As a check we have
a correlation of .96 between the measure of altitude and the
sum of the numbers right in K , L , M , and N.
In the case of 63 university students the correlations
between altitude CAVD and % s in tasks N , O , P , and Q
were as follows :
Sh
.77
.92
.90
NOPY
.70
Q
The intercorrelations of and are with

N
N
O
O
,
,
.58 with .70 and with .73 The self correlations

P
O
Q
;
,
;
P,
-
.
of and may be estimated and .76

as
.61 .67 .74

N
P
O
to Q
,
by adding .03 the correlation between neighboring com

posites . The self - correlation of the measure of altitude in

this group may be taken as .83 , the three measures of which
it is the average having intercorrelations of .80 , .54 , and .50 .
The correlation between a precise measure of altitude and
a precise measure of width is then computed as 1.08 for N ,
1.23 for O , 1.14 for P , and .88 for Q , averaging 1.08 .
As a check on this result , we have the correlation of .98
between the measure of altitude and the sum of the numbers
right in N , O , P , and Q.
We have thus five determinations of what the correla-
tion between altitude CAVD and W would be if both were
measured precisely , namely ,
for group 17 1.06

66
13 1.07
66
91 .97
66
911 .99
66
Univ . students 1.06
with an average of 1.03 ± a mean square error of .019 .

There is an element of insecurity in these corrections
for attenuation , especially in so far as the self - correlations
for W ( C or A or V or D ) are estimated by adding .03 to
the obtained correlations for neighboring composites .
However , the empirical correlations between the obtained
altitude measure and the obtained sum of the W's ( .99 , .95 ,
.96 , .91 , and .98 ) show that the corrected correlations should
be near unity .
The same close correlations obtain in groups at low alti-
tudes . In the case of the 100 individuals of group im6 , the
correlations of the measure of altitude with % s in C , D , E ,
F, and G , respectively were .79 , .86 , .89 , .86 , and .54 , aver-
,
aging .79 . The self- correlations of the measures of % s

were , respectively , .80 , .86 , .84 , .83 , and .81 , averaging .83 .
The self - correlation of the measure of altitude in this group
is .67 by the Spearman correction , the average intercorrela-
tion of the three determinations of which it is the average
being only .407 . The correlation between a precise measure
of altitude and a precise measure of W ( C or A or V or D )

is then 1.06 by this determination .
The correlation of the obtained measure of altitude with
the sum of the numbers right in C , D , E , F , and G was .93 .
In the case of the 50 f, the correlation between the ob-
tained altitude measure and the sum of the number right
in E , F , G , H , and I was .98 .
J
In the case of 162 individuals of group 4 , the correla-
tions between altitude CAVD and % s in tasks F , G, H , , , I
and K were .48 , .83 , .93 , .95 , .75 , and .53 , respectively . The
intercorrelations of % s in F, % s in G, and so on , are : F
with G = .67 ; G with H = .81 ; H with I = .851 ; I with = J
J with K = .51 . The self - correlations may therefore be
J
.63 ;
taken as .70 for F , .77 for G , .86 for H , .77 for I , .60 for ,
and .54 for K. The self - correlation of the measure of alti-
tude is .81 , the average intercorrelation of the three mea-
sures of which it is the average being .59 . The most prob-
able correlation between a precise measure of altitude and
a precise measure of width is then .64 for F , 1.05 for G ,
1.111 for H , 1.20 for I , 1.08 for J , and .80 for K, with an
average of .98 .
As a check on this determination we have computed the ,
correlation between the measure of altitude and the sum of

the numbers of rights in F , G , H , I , , and K. It is .96 .J
A rough calculation of the correlations for the 180 cases
of group im3 shows that with them the raw correlations of
the altitude measure with W ( 1C or 1A or 1V or 1D ) in com-
posites A , B , C , and D will be around .90 and that the cor-
rected coefficients will be near unity .
The closeness of these correlations indicates that each
individual would , if adequately measured by a large num-
ber of single tasks at each level of difficulty , show a pattern
closely of the type of Fig . 59. Individuals might be of
widely different patterns , such as those shown in Fig. 60 ,
Fig . 61 , and Fig . 62 , so that individuals of the same altitude
would differ widely in width at any level . But , in fact , such
large divergences in pattern are very scarce in Intellect
CAVD .
Fig . 59 Fig. 60 Fig. 61 Fig. 62
FIG . 59. The pattern of decrease in percent of single tasks correct with in
crease in difficulty , which corresponds to close correlations between
altitude and W ( 1C or 1A or 1V or 1D ) .
FIGS . 60 , 61 and 62. Patterns of decrease in percent of single tasks correct
with increase in difficulty such as individuals would show
if the correlations between altitude and W ( 1C or
1A or 1V or 1D ) were much below 1.00 .
How small and scarce they will be in other forms of in

tellect that is , how close a resemblance between altitude
,
and width will be found for any other form of intellect , will
depend upon the constitution of the form in question . In
CAPIma , with picture completions and information about
music and art replacing vocabulary and directions tasks ,
the correlations will probably be lower . However , so long
as the constituents of our composite tasks all concern the
ability to deal with ideas and symbols for ideas , the amount
of specialization will be small in comparison with the total
variation in ability , so that the correlations will be high .
THE RELATION BETWEEN ALTITUDE AND AREA OF INTELLECT

The facts brought forward in the first and second sec
tions of this chapter prove that the A ( 10C + 10A + 10V +
10D ) of any intellect and the altitude of that intellect are
determined almost or quite entirely by the same cause or
causes .
The facts of the third and fourth sections prove that to
a very considerable extent this is true for the A ( 1C or 1A
or 1V or 1D ) of any intellect and its altitude . A verifica
tion of this by the direct measurement of A ( 1C or 1A or 1V
or 1D) is not yet possible because the number of tasks at
each level of difficulty is unknown . Indirectly , it may be
partially verified as follows : If n single tasks are taken
from each level from zero to forty - five , one - fourth being C ,
one - fourth A , one - fourth V , and one - fourth D , and individ
uals are measured in respect of these , n being sufficiently
large , the A's so obtained will have the same rank as A's
obtained by an examination where the intellects are tested
with all tasks at all levels . The area for the selection of
n at 0 , + n at 1 , + n at 2 , + n at 3 , and so on , may be taken
to be approximately the area found by assuming that each
intellect will succeed with all or nearly all of the single ele
ments at levels below the highest level where it obtains 100
percent right and will fail with all or nearly all of the single
elements at levels above the lowest level where it obtains
zero percent right ( or only that percent which mere chance

guessing could give ) .
By permitting some estimating of scores , this procedure
may be carried out . The results appear in Fig . 63. The
cases entered in Fig . 63 are all taken at random so far as
461
45-
44-
43-
42-
41
40-
39-
38-
37
36-
35-
34-
33-
32-
31-
30-
29-
28
27-
12 15 16 T8
FIG . 63. The relation between CAVD altitude and area in a sampling of tasks
comprising N tasks for each unit of altitude .
the relation in question is concerned . Those used were all

that had 37 or more right in the easiest altitude with which
they were tested , or a random selection from all such . The
groups used were im3 , f , 4 , 91 , 17 and the group of 63 univer-
sity students . The area number was computed as follows :
I. Assume that , at each unit of altitude up to the easiest
altitude at which the person was measured , he had 40 ( i.e. ,
all ) right . II
. Count the number he had right over the
range at which he was measured ; and estimate from this

how many he would have had right had he been tested with
40 single tasks at each unit of altitude over this range . . III
Estimate the number which he would have had right at all
altitudes above the highest at which he was tested , using
arbitrarily the number which he had right at the highest
altitude at which he was tested . The area number is the
sum of the three numbers obtained by , , and I II. III
The area number thus ranges possibly from 957 for an
im3 who had 37 right in Composite A and none right in any
higher composite , to 1,800 for a person who had 40 right in
N and also in O , P , and Q. The lowest actual area number
among the cases used was 1,063 ; the highest was 1,760 .
The very close interdependence of area and altitude shown
by Fig . 63 would be little if at all reduced if more extensive
and precise measures were available.¹
There is thus a high degree of genuine unity to Intellect
CAVD , not assumed but discovered . We began with a mea-
surement in the form of an inventory , differing from a bare
enumeration of success or failure with actual tasks only in
that the tasks were graded in difficulty . We end with mea-
surements of altitude , width and area which intercorrelate
so closely that they may reasonably be treated as results of
a closely knit set of causes . Whatever makes one intellect
able to do much harder CAVD tasks than another intellect
1 It would be reduced inasmuch as some of the errors now involved act in
the same direction on the altitude measurement and on the area measurement .
It would be increased inasmuch as the purely chance fraction of the error acts
to reduce the correlation .
also makes it able to do many more tasks than that other

can do . After the necessary data have been collected , width
at any altitude , and so total area , will be predictable in the
case of Intellect CAVD ( and presumably in the case of
other forms of intellect ) rather precisely from altitude
alone .
THE RELATION OF ALTITUDE OR LEVEL OF INTELLECT TO SPEED
It is important to know the relation between level and

speed for two reasons . If the relation is very close , the
speed of performing tasks which all can perform would be
an admirable practical measure of intellect . The record
would be in time , an unimpeachable and most convenient
unit . If
on the other hand , the correlation is very low , the
,
practice of giving credit for speed in group examinations

should probably be amended .
Dr. Hunsicker [ 25 ] has made extensive individual mea
surements upon 82 adults and 81 school children , taking the
time for easy problems in arithmetic and for easy comple
tions , such as appear in our composites E , F , and G ; and
then testing the person with harder and harder tasks until
the level was reached where he could not obtain fifty per
cent right .
The correlations which she obtained between altitude
and rate (the reciprocal of the time required for tasks done
with no , or very , very few errors ) are shown in Table 130 .
They are much too low to make it advisable to use the speed
at easy tasks as a measure of the altitude or width or area
of intellect , except possibly in the case where the time avail
able for the examination is very short. They are indeed so
low that it seems unwise to attach much weight to speed in
intelligence examinations in general . A graded or ladder
test of thirty minutes containing 5 levels each consisting of
ten words and five arithmetical problems using small num
2 Except , of course , in the case of tests ( such as the substitution test )
where speed measures the speed of learning .
3 Or containing ten opposites and ten questions of arithmetical informa
tion , or containing five directions and five arithmetical problems .
bers , will in all probability show a closer correlation with

any reasonable criterion of intellect than will a thirty min-
utes ' speed test .
TABLE 130 .
CORRELATIONS , RAW AND CORRECTED FOR ATTENUATION , BETWEEN RATE AND

LEVEL . ( AFTER HUNSICKER , '25 , TABLE V. )
Arithmetic Completion
Individual No. in Raw Corrected Raw Corrected Average
Testing group r r I r
W. C. 28 .29 .35 .50 .56 .46

S. C. 54 .46 .55 .19 .23 .39
P. S. 189 32 .49 .58 .49 .64 .61
P. S. 6 49 .29 .35 .41 .50 .43
Average .46 .48 .47
We have extended Dr. Hunsicker's work by a measure-

ment of the speed of doing a collection of CAVD tasks
chosen from levels I and below in the case of 63 university
students for whom a measure of CAVD altitude was ob-
tained by the use of composites N , O , P , Q , and a still harder
composite .
There were some errors in the easy tasks , so we have

computed rsa.e , the partial correlation between speed and
altitude , for those making equal numbers of errors in the
rate test .
Isa.403 Ise
= .084 rae
- .484 ,
hence Isa.e
= .416 .
The self - correlation of the measure of altitude is .83 for

this group ; the self- correlation of the measure of speed is
not known but is almost certainly between .7 and .9 . If the
.403 were corrected for attenuation , the result for CAVD
would thus be fairly close to Dr. Hunsicker's results for A
and C.
CHAPTER XIV
THE MEANING OF SCORES OBTAINED IN STANDARD

INTELLIGENCE EXAMINATIONS
THE MEANING OF THE BINET MENTAL AGE
A Binet Mental age is a rough measure of relative alti-

tude A D Inf Ot , using Ot to mean " other tasks found or
alleged to deserve inclusion in a battery to measure intelli-
gence " ; or , more exactly , of the relative A ( la or 1d or linf
or lot ) of a sampling of a certain number of tasks at each
of certain levels . This A will correlate closely with alti-
tude . Up to about M. A. 14 , Binet scores are defined by the
probable median or average chronological age of those who
would obtain such a score , in the group by which the ex-
amination was standardized . Above M. A. 14 , the scores
are arbitrary .
Until the numbers of tasks at each level of difficulty are
known, and perhaps even after they are known , a Binet
Mental Age may best be treated as a measure of altitude—
of how hard tasks the person can succeed with . If this is
done , nothing will be lost from sound present uses and cer-
tain misapprehensions will be avoided . For example ,
everyone will understand that a very small increment of
mental age at the high ages may mean a very large incre-
ment in area of intellect or percentage of success with the
total mass of intellectual tasks which life may offer , and
that a very large increment of mental age at the low ages
may mean a relatively small increase in the total number of
tasks achievable or in the total number of connections
formed .
The great merit of the Binet Test is that it is a graded
scale for intellectual difficulty , and it is only weakened by
being interpreted loosely as a measure of some mysterious
essence called intelligence which grows in man . The weak-
402
MEANING OF SCORES IN INTELLIGENCE EXAMINATION 403
ening is not disastrous simply because , as was shown in the

previous chapter , altitude and width ( and consequently
area ) of intellect are so closely correlated .
Miss Rowell is measuring the values of Stanford Binet
M. A. 10 , M. A. 11 , M. A. 12 , etc. , in terms of the absolute
units of the CAVD scale in so far as one can be said to mea-
sure the equivalence of two series of magnitudes which may
not be measures of exactly the same fact in nature , and of
which one (the Binet ) may not measure varying amounts
of the same fact . We have found that adults of Stanford
Binet Mental Age 48 months , or 4 years , will show an alti-
tude of about 26 in Intellect CAVD ; and that adults of
Stanford Binet Mental Age 78 months , or 6 years , will
show an altitude of about 30 in Intellect CAVD . When , by
these measurements or by others , the differences in the
M. A. scores are put in equal units and referred to the abso-
lute zero of intellectual difficulty as located by us , or as
more accurately located by future workers , the Binet scale
and measurements will have a much greater value than they
now have .
What has been said of the Binet applies equally to the
Herring Examination , which is an alternative Binet .
THE MEANING OF SCORES OBTAINED IN STANDARD GROUP

EXAMINATIONS
The significance of scores in group tests such as the

Army Alpha , National , Otis , may best first be considered
with disregard of the factor of speed ; that is , on the as-
sumption that the scores of individuals represent what
they can do with time enough allowed to exhaust their
abilities .
The score does not measure either altitude or width or
area of intellect . It does not measure altitude , because the
number of tasks between levels equally far apart is not nec-
essarily the same . It does not measure width , because the
score is not divided up into a number of sub - scores , each
representing the number of successes at a certain level . It
does not measure area , because it measures neither altitude

nor width , and because the percent which the tasks are of
those that might be had at any level of difficulty is not
known .
Although one of these group tests does not in a rigorous
sense measure any one of the three , the score in it is about
as closely symptomatic of altitude as the score in any test
requiring so short a time could be . It is also closely symp
tomatic of the average width of intellect at and near the
levels of difficulty represented by its tasks . One of these
group examinations is in fact very much like what we have
when we put together five or six of our CAVD 40 - com
posites that are in a sequence for difficulty . The difference
between a set of these CAVD composites from about G to N
and Army Alpha or the National or the Otis ( no time limit
being set ) is that in the case of the CAVD composites , we
know how many single tasks there are at each level of diffi
culty , and we know how far apart the levels of difficulty are ,
and we can not only make a summation of credits , but also
can make an altitude score , and a width score at each level .
In Army Alpha or the National or the Otis , the total sum
mation score is not susceptible of such an analysis .
Except for the speed element , then , one of these stock
intelligence examinations may be regarded as a series of
composites unequal in the number of their elements , and
undefined as to the distances between levels . The addition
of the speed element complicates matters and theoretically
makes the significance of the score incapable of interpreta
tion except in terms of what people of a certain sort do in
that kind of a test when it is scored in that way.
Practically , however , the speed element does not make
the scores in these examinations , as they are administered
in the case of most of the individuals who are measured by
them , very much different in significance from the scores
which would be obtained with no time limits set . A few
persons are nervously upset by the instructions to work as
fast as they can ; a few cautious , critical workers do not have
time enough to do as many of the hard tasks as they are

really able to do ; a few persons are scored unduly high
because they utilize the time especially shrewdly , while a
few others are scored unduly low because they dally too
long over tasks at which they fail , or leave tasks undone
which the use of a little more time would have enabled them
to finish . But , in general , the scores in these speed tests
correlate very closely with the scores obtained when a
longer time allowance is given , partly because the correla-
tion between speed and altitude is positive , but more be-
cause the standard time allowance is long enough to enable
most of the candidates to do most of the tasks which they
could under any circumstances do .
The experiments of the Army psychologists on the re-
sult of doubling the time allowance for the Alpha and Beta
examinations are well known [ Memoirs , '21 , pages 415-
420 ] . The general result was that there was a slight im-
provement in the correlation with officers ' ratings for in-
telligence , and a close correlation between the score in
single time and the score in double time ( r = .967 ) , which is
probably as high as the self - correlation of the determina-
tions would permit .
Dr. J. R. Clark has investigated the influence of alti-
tude and of speed upon the abilities measured by the Stan-
ford Binet , the Otis Self - Administering Test , and the Ter-
man Group Test , in the case of school pupils from Grades
7 to 12 .
His results are not entirely clear , because his measures
of speed are afflicted by rather large variable errors , and
are perhaps also disturbed by the presence of an altitude
factor ; but on the whole they indicate that scores in these
stock examinations are determined much more by altitude
than by speed , and perhaps are determined almost entirely
by altitude and width . The average of the six speed cor-
relations ( speed in arithmetic and speed in completions
with Binet , Otis , and Terman ) each being corrected for
attentuation , is .54 . The average of the corresponding alti-
tude correlations is .70 . The average of the four correla-

tions between speed and altitude is not given , nor all the
data whence to obtain it. Ar speed with Ar altitude ( cor-
rected r ) is .76 ; Co speed with Co altitude is .40 ; the aver-
age is thus .58 . The other two r's are not given . They
would presumably be lower . If
their average is estimated ,
we can compute the partial correlations of speed with Binet ,
Otis , and Terman for persons of equal altitude in Ar or Co
and of altitude with Binet , Otis , and Terman for persons of
equal speed in Ar or Co. Estimating this average as .48 ,
the partial correlations are .28 for speed and .58 for alti-
tude . A more instructive set of measurements is of the
relations between speed in general and altitude in general
to scores in Binet in general , Otis in general , and Terman
in general . These Dr. Clark has made . He finds that dif-
ferences amongst individuals in the score in one of these
examinations are almost perfectly correlated with differ-
ences in what is common to their two altitudes , and much
less closely correlated with differences in what is common
to their two speeds . We quote his results .
" r general level and Binet
=
(r ar. level and Binet ) (r co. level and Binet )
(r ar. level and co. level ) ( r Binet Binet )

and Binet
.65 X.65 = .93 .

( .55 ) ( .90 * )
Similarly
r general level and Otis
(r ar. level and Otiss ) (r co. level and Otis .
(r ar. level and co. level ) ( r Otis and Otis )
.83 X.61 =
= .98 ,
( .55 ) ( .90 * )
and
r general level and Terman
1 "
Binet in general " means the average score in an infinite number of
tests patterned after the Stanford Binet .
* Estimated .
|( r ar. level and Terman ) (r co , level and Terman )

( .55 ) ( .90 * )
.80 X .66 -
Approx . 1 .
( .55 ) ( .90 * )
In the same way, the relationship between ' general
speed ' scores and intelligence test scores is found to be :
r .55 X.49
general speed and Binet = .59 .
( .50 ) ( .90 * )
r = .71 X.49 = .77 .

general speed and Otis
( .50 ) ( .90 * )
r .67 X .32
-
general speed and Terman .49 .
( .50 ) ( .90 * )
The mean of these correlations is .62 . " [ Clark , '25 , p .
33f. ]
If partial correlations are computed using .97 , .62 , and
any reasonable estimate for the intercorrelation of speed
in general with altitude in general , they are very high for
altitude and very low for speed . For example , let r for
speed in general with altitude in general in this group be
.65 . We then have .95 and .00 . Letting it be .60 , we have
.95 and .20 . Letting it be .70 , we have .96 and — .34 .
Even with time limits , then , the scores in standard
group examinations may properly be treated approximately
as summation - of - credit scores of the same fundamental
nature as a Stanford - Binet summation score or as a CAVD
summation score . The chief difference for practical pur-
poses is that the Stanford - Binet summation is of rights in a
series of tasks specified as to difficulty , six for the interval
from Chr . Age 3.0 to Chr . Age 4.0 ; six for the interval
from Chr. Age 4.0 to Chr . Age 5.0 ; and so on ; and the
CAVD summation score is of tasks specified as Diff . 23 ,
Diff . 264 , Diff . 281 , and so on ; whereas the Army Alpha or
National or Terman Group summations are from an unde-
fined collection of tasks .
* Estimated .
28
We may expect that , in the future , all these group ex

aminations which have proved themselves so convenient in
getting quickly and cheaply an approximate measure of
something which is reasonably called " intelligence , " will
retain these advantages and gain those of clearer interpre
tation by certain changes in the method of construction
which are recommended by the principles which we have
formulated . Instead of being a collection of small tasks of
undefined location as to difficulty , they will be made in
levels with a definite number in each level , and the levels
will be placed , at least approximately , equally far apart .
A more liberal time allowance will be given , and each indi
vidual who is examined will be instructed to take as much
time as he needs and to go as far as he can . The group test
can then be scored by a summation of credits just as now ,
but that summation will have resulted from a combination
of scores of the number right at each of various levels of
known difficulty. An altitude score can be inferred from it ,
since its correlation with altitude will be nearly as high as
its own self- correlation permits . Or an altitude score can
be computed more directly by some such procedure as was
described in the case of CAVD in Chapter II .
THE MEANING OF SCORES OBTAINED IN TESTS OF THE ABILITY TO

LEARN AND TO IMPROVE
The only test of ability to learn which has been widely

used as a measure of intellect is the substitution test.²
There has been far too little experimentation with ability
to learn as a test of intellect . The early work seemed dis
couraging, the correlations with a criterion being appar
ently much lower per unit of time in testing and labor in
scoring for tests of improvement than for tests of status .
We have elsewhere shown [ Thorndike , '24 ] that this may
be due in large measure to the peculiar action of the error
2 The form used in the Nationalis typical . The amount done correctly in
a given time depends largely on how quickly and accurately the individual
learns the key .
of measurement upon the relation between initial ability

and gain . Thomson [ '24 ] has provided the requisite cor
rection formula . While the recognition of this error and
the correction for its influence on the correlation does not
improve the actual diagnostic or predictive value of a short
learning test , it does demonstrate that a test which is long
enough to measure improvability accurately may have a
much greater diagnostic and predictive value than had
seemed possible . Another difficulty which has discouraged
experimentation with tests of improvability is the lack of
any sound general theory for comparing gains from differ
ent starting points ; and this may greatly limit the scope of
such tests . On the other hand , measures of improvability
have exceptional advantages in respect of universality of
application and freedom from improper training . They
may also be much less susceptible to environmental differ
ences than are the ordinary measures of status .
An investigation of the altitude of learning , that is , of
how hard things a person can learn to do ; and of the speed
of learning certain things which all or nearly all can learn
ifthey have time enough , is very much needed . Until such
an investigation is made , the sound procedure with scores
in substitution tests and the like is to treat the varying
scores as representing success with tasks varying in diffi
culty . A score of K + 2 thus means success with a harder
task than does a score of K + 1, and so on . The differ
ences in difficulty corresponding to the differences in the
scores can then be measured by the principles and tech
niques presented in this volume .
There are two final matters which concern all existing
tests and scales for intellect . The first is the matter of the
selection of the tasks . The second is the matter of the re
liability of the measure .
In all of them the selection of the single tasks has been
narrow and more or less arbitrary . Binet chose tasks
which older children did and younger children could not do .
In the National the choice was made in view of a criterion
consisting of grade reached in school , intelligence as esti
mated by teachers , and the like . A general fund of knowl

edge that such and such tasks are allied to various symp
toms and assumed criteria of intelligence is more or less
wisely used . In none of them has the selection been made
so as to represent or sample in any defined way any total
series of tasks which the authors of the examination re
garded as being the totality of intellectual tasks .
This is not a serious defect for the ordinary purposes
of ranking individuals according to that defined trait
ill
-
known intelligence for two reasons First most of the
as
,
,
.
examinations common use are made up of verbal and
in
numerical and factual tasks in somewhat the same propor

tions and second the intercorrelations of different sorts
;
,
,
of tasks so long all concern the ability manage ideas

as
to
,
,
are all fairly close
.
However there no loss and considerable gain espe

is
a
,
,
cially for rigorous treatment definite plan for the selec
if
a
,
tion of tasks used as in the CAVD series that the

If so
is
,
nature of the fact measured clearly defined the Army
is
Alpha for example were called Cs So Di Ac An

A
D
,
,
Inf and constructed so that there were three of each of
,
these sorts of tasks eight levels would be

of
each
at
it
a
,
more useful instrument

.
In respect the reliability of the determination our

of
examinations have been far too lax especially the high

at
,
levels serve well for scientific purposes for such prac

or
to
,
tical purposes require any considerable exactitude The

as
meaning of any score obscured by the fact that large

so
is
portion of chance error At the high levels small

is
it
a
.
error altitude may mean very large error area The

in
in
a
neglect of this matter has indeed been almost scandalous

,
since the case of many widely used intelligence examina

in
tions the amount of the chance error not even known

is
,
In connection with our inquiry Miss Woodyard con

is
,
ducting careful investigation

of
the nature and amount

a
and causation of the chance error in mental measurements

,
which will treated separate report We present here

be
in
a
only the facts concerning the reliability of determination

a
of CAVD altitude such as results from a measurement of

an individual with four 40 - composite tasks near his level
of difficulty . To obtain such a measurement in an indi-
vidual examination will require about thirty minutes of
exploratory testing and about two hours with the four com-
posites . To obtain it in a group examination will require
usually that the group be tested with five 40 - composites re-
quiring from three hours at levels H to L , up to five hours
at levels N to R.
The mean square error of an altitude determined from
the three successive 40 - composite scores whose sum is near-
est to 60 is as shown below for various groups .
MEAN SQUARE ERROR OF A CAVD ALTITUDE IN UNITS OF THE CAVD

SCALE ( 1.00 EQUALLING 019 )
The median error is reported for each group .

100 adults of mental age 2 to 4 .41
100 adults of mental age 6 .32
115 pupils in Grade 4 .40
100 pupils in Grade 9 .29
63 university students .37
The mean square error is thus about .35 , or about one

fiftieth of the difference between a low grade imbecile of
mental age 3 years and a very gifted adult , one person in a
thousand .
By improving the CAVD composites , the error doubt-

less can be reduced somewhat . Also , further investigation
may disclose a procedure more serviceable than the " 60 "
rule (described in Chapter XI , p . 369 ) , and not more com-
plicated or time - consuming . But the decreases will prob-
ably be small . In the main , decrease in the error must be
attained by increase in the time of the examination .
The existing stock intelligence examinations may be
superior to CAVD in this respect , but the probability is
that they are inferior to it . It has the advantage of build-
ing upon the results gained by them , and of choosing from
the best task -material known to date .
CHAPTER XV
THE NATURE OF INTELLECT
We have learned to think of intellect as the ability to

succeed with intellectual tasks , and to measure it by mak
ing an inventory of a fair sampling from these tasks , ar
ranging these in levels of intellectual difficulty , and observ
ing how many the intellect in question succeeds with at
each level ( and , if we wish , how long a time each success
takes ) . From this graded inventory , we may compute mea
sures of altitude or level , of width or range or extent at
each level , and of total area . For Intellect CAVD , the pat
tern for a sample with n tasks at each level , if n is large
enough , is very similar for all individuals of roughly simi
lar training . Consequently , the altitude , the total area , and
the width at any level are closely interdependent .
Any defined intellect can be treated as we have treated
Intellect CAVD .
Such a definition in terms of tasks accomplishable , and
such a measurement in terms of the contents of a graded
inventory is sound and useful , but is not entirely satisfying .
One cherishes the hope that some simpler , more unitary
fact exists as the cause of intellect and that variations in
the magnitude of this fact may provide a single funda
mental scale which will account for levels and range and
surface . one realizes the desirability of search
Moreover ,
for the physiological cause of intellect , regardless of
whether that cause be single and simple or manifold and
complex .
Our consideration of these matters has led us to a hy
pothesis concerning the nature and causation of intellect
for which we have found fairly substantial evidence . It is
the purpose of this chapter to present and discuss this
hypothesis .
412
THE NATURE OF INTELLECT 413
A WORKING DEFINITION OF INTELLECT
Since this hypothesis concerns intellect in general as

well as any defined segment of it such as Intellect CAVD ,
we may revert to vaguer and more catholic conceptions .
The fact of human life of whose nature we seek a more
exact description is the ability to deal with things or per
sons or ideas by the use of ideas . We contrast intellectual
power over things , as by ideas about length or weight or
heat , with non - intellectual power over things , as by strength
or skill or acuity of vision . We contrast intellectual power
over people , as by consideration of facts about them , with
non - intellectual power over them , as by good temper or
courage or physical charm . We contrast intellectual power
over ideas , as by using other ideas to gain success with
them , with non - intellectual power over them , as by industry
or patience .
The facts and arguments which we shall present do not ,
however , depend for relevance or value upon the acceptance
of this particular identification and demarcation of intel
lect . They will apply nearly or quite as well to any prelim
inary description which any competent psychologist would
devise for those features of life wherein the Aristotles dif
fer most widely from the inmates of asylums for idiots , and
wherein the life of a man thinking effectively about mathe
matics or medicine or manufacturing differs from the life
the same man eating , drinking , swimming or playing tennis
without , as we say , an idea in his head .
If a score of competent psychologists should list on the
marily upon intellect -
one hand all the products whose production depends pri
all the tasks for success with which
intellect is the sine qua non ; and on the other all the prod
-
-
ucts or tasks which they regarded as non - intellectual suc
cess thereat being independent of intellect they would
show very substantial agreement . Where they appeared to
differ , the differences would be unimportant for our pur
pose . Very seldom would the same task appear on opposite
sides of the ledger . When it did , the difference would re
solve itself into a difference in favor of a narrower restric

tion of intellect ( for example , to ability to deal with ab
stract ideas or to ability to deal with relations ) , or in favor
of a wider extension of it ( for example , to certain tasks
where ideas are not at work , at least not obviously ) . The
facts and arguments which we shall present will apply re
gardless of such shrinkage or swelling in the area regarded
as intellectual .
Intellect may be CAVD , or CAVD plus ability in giving
the opposites of words , making it CAVDO ; or that , plus in
sight into spatial relations , making it CAVDOS ; or that
plus ability in inductive and deductive reasonings , making
it CAVDOSR , and so on .
Beginning , then , with this loosely determined group of
products which intellect produces , tasks at which intellect
brings success , we may inquire concerning its observable or
surface nature as a fact in human behavior or its deeper
nature as a fact of fundamental processes in the mind or
brain . We may investigate the thinking and action of men
who have much intellect to discover more precisely and
fully the features in which they differ from men who have
little ; or we may try to discover more ultimate causes of
these differences . We may compare a man's obvious life ,
when he is using his intellect little or not at all , with his life
when he makes large use of it , other factors remaining the
same , to see just what the differences are ; or we may try to
discover hidden forces which produce these differences .
We may study the nature of intellectual tasks , the produc
tion of intellectual products , or the nature of the ultimate
power or powers whereby a man can succeed with such
tasks . There may be , of course , much to be revealed con
cerning facts intermediate between the description of in
tellectual tasks and the discovery of their ultimate cause .
The standard orthodox view of the surface nature of
intellect has been that it is divided rather sharply into a
lower half, mere connection - forming or the association of
ideas , which acquires information and specialized habits of
thinking ; and a higher half characterized by abstraction,

generalization , the perception and use of relations and the
selection and control of habits in inference or reasoning,
and ability to manage novel or original tasks . The ortho
dox view of its deeper nature , so far as this has received
attention , has been that the mere connection or association
of ideas depends upon the physiological mechanism whereby
a nerve stimulus is conducted to and excites action in neu
rones A, B , C , rather than any others , but that the higher
processes depend upon something quite different . There
would be little agreement as to what this something was ,
indeed little effort to think or imagine what it could be , but
there would be much confidence that it was not the mecha
nism of habit formation .
THE HYPOTHESIS THAT QUALITY OF INTELLECT DEPENDS UPON

QUANTITY OF CONNECTIONS
The hypothesis which we present and shall defend ad

mits the distinction in respect of surface behavior , but as
serts that in their deeper nature the higher forms of intel
lectual operation are identical with mere association or
connection forming , depending upon the same sort of physi
ological connections but requiring many more of them . By
the same argument the person whose intellect is greater or
higher or better than that of another person differs from
him in the last analysis in having , not a new sort of physio
logical process , but simply a larger number of connections
of the ordinary sort .
More exactly our hypothesis is as follows : Let c repre
sent whatever anatomical and physiological fact corre
sponds to the possibility of forming one connection or asso
ciation or bond between an idea or any part or aspect or
feature thereof and a sequent idea or movement or any part
or aspect or feature thereof . Then if individuals I1 , 12, 13 ,
14 , etc. , differing in the number of c's which they possess
but alike in other respects , are subjected to identical en

vironments , the amount or degree of intellect which any
one of them manifests , and the extent to which he mani

fests "higher " intellectual processes than the other indi
viduals , will be closely proportional to the number of c's
which he possesses . If we rank them by intelligence - exami
nation scores , the order will be that of the number of c's .
If we rank intellectual processes in a scale from lower , such
as mere information , to higher , such as reasoning , the indi
viduals who manifest the highest processes will have the
largest number of c's.
The view of common sense , of educational science , and
of those who have constructed tests for measuring intellect ,
has been that intellect is a power to respond correctly , that
the quality of the responses is a primary criterion of the
degree of intellect . The teacher and the test - maker would
insist that correct judgments and valid inferences required
more intellect than the wrong judgments and faulty in
ferences .
The hypothesis which we present accepts this view, but
makes a sharp distinction in this regard between the orig
inal intellectual capacity which a man has and the actual
intellectual products which he produces . It credits the
quality of the ideas that a man acquires , and the truth or
falsity of the judgments which he makes , and , to some ex
tent , even the validity of the inferences which he draws
from any given data , largely to his training . The average
man today has better ideas about lightning than Aristotle
had , can make more correct judgments about eclipses than
Moses could , and , if trained in science , may well draw more
valid inferences from observing the action of acids on
metals than either Aristotle or Moses did . With approxi
mate equality of training , the quality of intellectual re
sponses is an essential index of intellectual capacity , but it
may be deceptive if the inequalities are great.
Our hypothesis limits itself to the original capacity . If
by original nature , apart from all training , a man possesses
tendencies to be right rather than wrong in his judgments ,
to hold true rather than false ideas , to make justifiable
rather than unjustifiable inferences , more or less than other

men , in so far forth those tendencies are due to his having
more or fewer c's than other men .
The essential element of our hypothesis is that it offers
a purely quantitative fact , the number of c's, as the cause
of qualitative differences either in the kind of operation
( e.g. , association versus reasoning ) or in the quality of the
result obtained ( e.g. , truth versus error , wisdom versus
folly ) , so far as these qualitative differences are caused by
original nature .
We need to make clearer what is meant by " one connec
tion or association or bond between an idea or any part or
aspect or feature thereof, or a group of ideas and a sequent
idea or movement or any part or feature or aspect thereof . "
By " connection or association or bond between a and b , ” is
meant the probability or certainty that if a occurs in a per
son , b will occur in him shortly thereafter ( say within a
second ) unless some counteracting force prevents . For the
sake of simplicity , we may think of all cases as cases of cer
tainty . The existence of the connection a → b then means
that whenever a occurs , b will follow within a very short
time interval , unless restrained by some contrary force .
Thus in a child who has learned the multiplication table , the
idea 2 times 5 will always be followed by the idea 10 , unless
some contrary force prevents .
The b which follows a may be the suppression or pre
vention of an idea or movement as truly as its appearance .
Connections where a inhibits some event may indeed well
be as numerous as connections where a releases or produces
some event . The b may also be one step toward an event or
one partial condition of it so that , for example , a , may be
followed by b₁ with no obvious difference in the person , and
a2 may be followed by b₂, again with no observable differ
ence in the person , but if a , and a , act closely together in
time , an obvious difference c in the person may always fol
low . Similarly the b may be one step toward or one partial
condition of the inhibition of some occurrence , so that , for
example alone may not prevent c , and a , alone may not

, a,
prevent c , but a₁ + a, in close enough temporal proximity
may entirely prevent c from occurring .
By an " idea " is meant any small portion of the stream
of thought , such as a precept , or an image , or a meaning ,
or a word of inner speech.¹ It is intended to omit unorgan
ized sensory stuff such as the person does not ever isolate
or identify or name , and emotional stuff such as excitement ,
irritability , or fear . But no sharp lines need be drawn , for
the hypothesis will remain valid even if " ideas " are inter
preted loosely to include more than we have in mind ; and
the hypothesis will be fruitful even if " ideas " are inter
preted very , very narrowly , say to include only words . It
is not intended to beg the question of consciousness or
bodily action by the use of the word image . Whether
" "
ideas be envisaged as facts of conscious awareness or
as facts of bodily behavior is a matter of indifference to the
hypothesis .
There remains to be clarified the apparently innocent

and unambiguous word " one , " which really conceals a nest
of difficulties . We think readily about " one idea , ” "
“ .part
of an idea , " " many ideas , " " two images that are nearly
alike , " " ten thousand percepts , " and the like ; and the
treatment of percepts , images , and relations as separable
and capable of enumeration is doubtless useful and in a cer
tain correspondence with reality . Yet it is hard to decide
when and why the reader's percept of , say , the word
" Adam " shall be counted as one percept of a word , or four
percepts each of one letter , or ten percepts each of one line ;
or when and why his visual image of a square shall be
counted as one image of a square , or as eight images of
lines and angles , or as a fraction of a total visual image of
a square on a certain background . Naïve common - sense
calls " Two and three are five " one idea , and with a certain
1 Sensori -motor connections , such as are formed in learning to swim , dive ,
box , or wrestle , and moral or temperamental connections , such as are formed
in keeping one's temper , or being courageous , are thus considered as , at least
partially , outside of intellect .
suitability , but that sort of a " one " is obviously different

from the " one " used when a person has the idea of two ,
or the idea of three , or the idea of " and - ness . "
In strictness the entire status of the mind ( or of the
central nervous system ) at any instant connects with or
leads to its entire status at the next instant , and any seg-
regation of a part of this total status as one idea , and any
attribution of a part of the sequent status to this idea's
associative potency , is an incomplete statement of what
happens .
These difficulties and others of like nature may best be

met , not by trying to set up rigorous criteria for what shall
be one idea , what a part of an idea , and what a group of
ideas , nor , on the other hand , by a refusal to use the anal-
yses which common sense and science have found profitable ,
but rather by realizing that ideas are not like eggs in a
basket nor like eggs in an omelet , but are what they are . If
we must liken them to something , let it be tones in a sym-
phony or factions in a party or neurones in the brain , or
some other case of a very complex organization where sci-
ence can and should separate the total into parts , but where
these parts are splitting and combining from time to time
and are being influenced in their action by more or less of
the total organization .
We must , of course , be consistent , in any comparative
enumerations , not calling the same fact now one idea , now
two , and now ten , to the prejudice of the truth .
The next matter to note is that there may be associa-
tions or bonds from different ideas , from a to b , from a to
c , from a to d , and so on , according to slightly different con-
ditions in the general status of which a is a part . Thus the
idea of " 12 " may call up " 11+ 1 , " or " 10 + 2 , ” or “ 4 X
"
3, or " 62 , " or " dozen , " or " not prime , " or " a num-
ber, " or " XII , " according to relatively minor conditions
attached to " 12. " There may likewise be connections from
many different ideas to the same idea , or to the same move-
ment , as when we think " yes " or nod the head to a hundred
different questions . This does not mean that the same

cause can have different results or different causes have
the same result , any more in the mind than in physical
nature . Different total statuses of the person never do lead
to one absolutely identical status . It means simply that
when certain accessories are neglected , one same conse-
quence follows a hundred different stimuli .
We can then analyze the stream or web or panorama of
intellectual life , finding in it ideas . These occur not hap-
hazard , but always by some cause . Chief among the causes
are the bonds or links whereby one idea tends to be fol-
lowed by a certain other idea . One man may have per day
or per lifetime many more different ideas than another
man . Of two men having the same number of ideas , one
may have many more different connections than the other
man has . For example , individuals A and B may each be
capable of the ideas a , b , c , d , and the movement e . A may
have only the connections a → e , b → e , c → e , and d → e .
B may have the connections a → bc , b → ac, c → db , d → ae ,
ab → abcc , ac → bbcd ,, ad →
→ abce , bc → abcd , bd → aace , and
others .
This greater fund of ideas and connections is partly due
to larger life and more varied and stimulating life , but it
may be and certainly is partly due to original nature . It
has some anatomical or physiological cause or parallel . Our
hypothesis regards this anatomical cause or correspondent
of the original possibility of having more such connections
(call it C ) as the cause of the original differences in intel-
lect among men . It also supposes that the correspondence
is such that C can be analyzed into a number of c's so that
a C which allows the formation of many connections with
ideas has many c's , whereas a C which allows the formation
of only a few has few c's . As we stated it , C consisted of
c's , one c meaning the possibility of the formation of one
connection , two c's meaning the possibility of forming two
connections , and so on . This form of statement was chosen
primarily for clearness and brevity . It needs amendment ,
or at least explanation .
Whether there is any such one to one corespondence be-

tween facts of anatomy and physiology , and facts of mental
association , bonds or links in behavior is , however , not of
much consequence to the hypothesis . What is essential to
the hypothesis is that by original nature , men differ in re-
spect of the number of connections or associations with
ideas which they can form , so that despite identical outside
environments , some of them would have many more than
others . "
The number of c's a man has " means simply the
original constitutional basis of the number of ideational
connections which he has . It is highly probable that the
original basis of quantity of connections is itself a matter
of quantity , that a more potent C is one that has more of
something than a less potent C has . But it is not necessary
for the hypothesis that this should be so . So our hypoth-
esis may better be amended to read :
Let Ca , Carb , Ca +b+e, Ca +b+c+d, and so on , represent original
natures such that with identical outside environments , the
man having C₁a will form the man having Ca +b
ab
a connections ,
will form connections and so on . Then , with iden-

,
tical outside environments , the amount of intellect which a

man manifests , and the extent to which he manifests
" higher " powers than other men , will depend largely upon
his C.2
Negatively , the hypothesis asserts that no special quali-
tative differences are required to account for differences in
degree of intellect ; the higher processes or powers have no
other basis in original nature than that which accounts for
differences in the number of bonds of the associative type .
The reader who is impatient with these subtleties may
forget them all with no great loss . The gist of our doctrine
is that , by original nature , the intellect capable of the high-
est reasoning and adaptability differs from the intellect of
2 Certain other inner conditions , such as the strength of curiosity , the satis-
fyingness of thought for thought's sake , and the appeal of non -intellectual
activities , in so far as they are distinct from the man's C , would have to be
allowed for to make a perfect prediction .
an imbecile only in the capacity for having more connec

tions of the sort described .
The bearing of the hypothesis upon the problem of mea
surement lies in the fact that we may be able for many pur
poses to replace our measurement via a sample inventory
of tasks , by a more or less direct measurement of C. we If
can get access to C so as to measure it ( and if the hypoth
esis is valid ) we can measure intellectual capacity , and can
measure it perhaps at a very early age . If also one C does
vary from another simply by consisting of a larger number
of c's we have a single variable in the most convenient of
all units .
Any person familiar with the finer anatomy of the
brain will at once think of the number of possible contacts
(or possibly coalescences ) of the fibrils of axones with den
dritic processes in the associative neurones which act in
perception thought , and speech as a highly probable C.
,
We have had it in mind as the possible C which we should

investigate first if opportunity offered . We do not , how
ever , make the hypothesis depend upon this particular C.
EXPERIMENTAL VERIFICATION OF THE QUANTITY HYPOTHESIS
The hypothesis may be submitted to an almost crucial

test by determining the correlations within the upper half
of intellect , those within the lower half , and those between
the upper and the lower halves . Do the " higher " abilities
of selective and relational thinking , abstraction , generali
zation and organization display close interdependence
among themselves and marked independence of the
" lower " or purely associative abilities ?
At our suggestion , Mr. J. W. Tilton has made this test
in the case of 250 boys at the time of graduation from
Grade 8. As measures of the " higher " or " control " abili
ties , he used sentence completions , arithmetical problems ,
and analogies tests . As measures of the more purely " as
sociative " abilities , he used vocabulary tests , routine and
informational arithmetic , and information tests .
The following is a sample of the arithmetic :
I. E. R. ARITHMETIC , ASSOC ., II
Add :
a. b. C. d. e. f.
822 .g
91 92 3 wk . 4 da . 7 lb. 12 oz . 11
1 2 51 4 wk . 2 da . 8 lb. 5 oz . 27
52 61 wk da lb. 14 oz 11
6
1
6
.
.
Multiply
:
7
h
ft
.
i.
.
.
5 5 .
254 9.6 in 16 12
%
3
.
214 16
4
6
Divide
:
m n 0
.p
.
.
.
50 .138 31 ÷
÷
=
6
7
9
Write the answers to these questions
:
cent how many mills

is is
.q
1
pint how many gills

1
?
r.
square mile how many acres

is
?
1
t s
. . .
How much 20 of 60
% %
is is
How much of 200 51

u
Which months have only 30 days

. .v
rod how many feet

is is
2. .y 2. w
1
acre how many square rods

1
meter how many inches

is
1
What the square root of 64

is is
aa What the cube root of 649

.
bb equals how many

63
?
.
The following sample of the information tests

is
a
E. R. INFORMATION AND F1
I.
E
2
,
Write print your name and age and grade here very plainly
or
school
in
Name Age Grade

In each of the sentences below you have among four words choice
a
,
Draw line under the one of these four words which makes the truest sentence
a
E
2
The Gnome engine airplanes automobiles trac-

chiefly used
in
1.
is
tors motorcycles
.
Vinegar made from picric acid apples bark lemons

is
3. 2.
Adobe the name of building material Indian tribe Chinese offi-

is
cial flower
.
4. One of the books of the Bible is Jacob Jesse Joshua Judah

.
5. Oxo the name of meat extract automobile cigar toothpowder

is
6. Lille in Belgium England France Switzerland

is
Queen Elizabeth of England was born about 1425 1525 1625

7.
1725
.
29
8. From Leningrad to Moscow is about 500 mi . 1,000 mi . 1,500 mi .

2,000 mi .
9. Corot is the name of a musician painter pigment general .
10. The angle of incidence is equal to the angle of coincidence reflec-
tion refraction subsidence .
F 1
1. The namber of rows of kernels on an average ear of corn is about

5 15 25 35
2. The ten commandments are called the decagon decalog decament
decemvirate .
3. The ratio of the size of Africa to that of Europe is about 2 to 1 3
to 1 6 to 1 9 to 1.
4. Brahmaputra is the name of a flower goddess language river .
6 lb. 12 lb. 18 lb.
5. A pint can full of lead would weigh about
24 lb.
6. Hydrogen becomes liquid at about -300 ° C. - 150 ° C. -10 ° C.

+ 60 ° C.
7. The number of a crab's legs is four six eight ten .
8. One inch equals about 2 cm . 21 cm . 2 cm. 2 cm .
9. An irregular four -sided figure is called a scolium trapezium paral-
lelogram pentagon .
10. One of the books of the Bible is David Eleazar Leviticus Uzziah .
The correlations ( corrected for attenuation ) , which are pre-

sented in Table 131 , give a clear answer . In this group the
" higher " abilities correlate as closely with the associative
abilities as the higher do inter se , or as the associative abil-
ities do inter se . The average of these six cases is .558 ±
.029 ( P. E. ) . The average of the three cases of " higher "
with “ higher ” is .544 . The average of the three cases of
" associative " with " associative " is .571 . The average of
the six cases of " higher " with " associative " where the
content differs just as much as it does in the " higher with
higher " and in the " associative with associative , " is .577 ±
.021 ( P. E. ) . The three cases Co. with Voc . , Ar . Cont . with
Arith . Ass . , and Anal . with Inf . where the content is similar
(words , numbers , and facts ) , have an average correlation
of .71 .
If we apply the attenuation formula so as to measure
the relation between ( a ) what is common to any two of the
associative tests with words , numbers , and facts , and ( b )
what is common to any two of the " higher " tests with
words , numbers , and facts, we obtain nine correlations "
whose average is 1.07.05 ( a ) . If
we eliminate the influ-
ence of the correlations between higher and lower with simi-
lar content ( not using rcov, Tanint , or TACAR ) the correlation
between what is common to the associative abilities and
what is common to the higher abilities , has an average of
1.00 ± .03 ( o ) .
There thus seems reason to refer the higher , originat-
ing , directing abilities to much the same fundamental causes
as the associative . The higher powers are in their causa-
tion as much like the lower as like one another . This is not
because the correlations are insensitive indices . On the
contrary , similarity in the content or data thought about ,
raises the correlation from .58 to .71 . Nor is it because the
subjects of the experiments did not have and use the higher
abilities .
We have extended , and , in general , confirmed , Mr. Til-
ton's findings by experiments of the same general nature .
458 pupils in Grade 11 in city K were tested with 350
vocabulary tasks and also with two forms of the I. E. R.
Selective and Relational Thinking , Generalization and Or-
ganization examination . "
676 pupils in Grade 11 in city K , closely similar in abil-
ity to the 458 just mentioned , were tested with over a hun-
dred reading tasks and also with the two forms of the I. E.
R. Sel . Rel . Gen. Org . examination .
The raw correlations of the general intelligence score
with the total vocabulary score ( sum of rights ) and with
the total reading score ( sum of rights ) were .72 and .73 , re-
spectively . The correlation of the general intelligence score
with that from another similar pair of examinations is .92
by the Spearman - Brown formula , the correlation of one
5 1.01 , 1.06 , .82 , 1.31 , 1.26 , 1.07 , 1.11 , 1.05 , and .97 , with a median
of 1.06 .
6 1.00 , 99 , 77 , 1.09 , 1.09 , 1.04 , 1.01 , 1.06 , and .90 , with a median of 1.01 .
This is a composite of stock tests of so -called general intelligence .
form with the other being .85 . The correlation of the vocab
ulary score with that of another similar examination is .98
by the Spearman - Brown formula , the correlation between
two random halves thereof being .97 . The correlation of
the reading score with that of another similar examination
is not known , but will not be far from .9 . Using .92 , .98 , and
.90 for these self - correlations , we have .76 and .80 as the
TABLE 131 .
INTERCORRELATIONS ( CORRECTED FOR ATTENUATION ) OF SENTENCE COMPLETION

( Co) , VOCABULARY ( V ) , ARITHMETIC CONTROL ( AC ) , ARITHMETICAL
ASSOCIATION ( AA ) , ANALOGIES ( AN ) , AND INFORMATION
(INF ) , IN 250 PUPILS OF GRADE 82 . ( COMPILED
FROM TABLES OF TILTON [ '25 ] , p . −) .
Control with Control Control with Asso . Control with Asso .

(content different ) (content similar )
Co An .522 Co Inf .722 Co V .865
Co Ac .523 Co Aa .550 An Inf .604
An Ac .587 An V .563 Ac Aa .643
An Aa .486
Ac V .491
Ac Inf .575
Asso . with Asso .
V Inf . .787
V Aa .433
Inf . Aa .592
Self -Correlations
V-I II
V- .815.0143
Inf I Inf II .600.0273
Aa I Aa II .829.0133
Co I Co II .744.0190
An I An II .920.0103
Ac I Ac II .950.0041
coefficients corrected for attenuation between a stock intel

ligence score and vocabulary and reading , respectively .
The mere knowledge of single words seems almost as “ in
tellectual " as the comprehension of paragraphs .
If we use for each individual a level score representing
the degree of difficulty at which he can succeed with 50 % of
the tasks , a similar result is obtained . The coefficients of
correlation for the group in question are :
Corrected for
Raw attenuation³
General intelligence score with vocab
ulary .72 .77
General intelligence score with compre
hension of paragraphs .66 .76
We have measured 100 university students in (1) a com

posite of sentence completions and comprehension of para
graphs , (2) a composite of picture completions , pictorial
analogies , and geometrical analogies , ( 3 ) arithmetical prob
lems , ( 4 ) a vocabulary test , and ( 5 ) an extensive informa
9
tion test.10 The intercorrelations are shown in Table 132 .

The " higher " abilities show an average corrected cor
relation among themselves of .48 by P and .35 by S ; the
" lower " show a correlation of .67 by P and .59 by S. The
average for a " higher " with a " lower " is .47 by P and .38
by S. In general , the correlation is nearly as close between
a " higher ' and a " lower " as within the higher or within
the lower .
The estimated correlation between what is common to C
and A and what is common to V and Inf is .90 . The esti
mated correlation between what is common to C and Pic
and what is common to V and Inf is .78 . The estimated
correlation between what is common to A and Pic and
what is common to V and Inf is .56 . The average is
.75 ± .08 (o) .
Additional evidence is found in the correlation in the
case of 126 pupils in Grade 5 for each of whom summation
8 The self -correlation of the vocabulary level -score is .94 . The self - corre
lation of the reading level -score is approximately .80 ( .77 with a level -score
from a less extensive test ) .
9 The arithmetical tasks were not hard enough to measure the ability of
the group well , and the correlations would probably be considerably higher
with an adequate set of mathematical tasks . But they would hardly surpass
the information correlations .
10 This was not as purely a test of associative thinking as would have
been most desirable , a certain amount of organization and inference being
of assistance in some of the tasks ; but it was so to an enormously greater
extent than the other composite .
428
.
TABLE 132
).
(
,.
,.
THE INTERCORRELATION OF FOUR TESTS OF THE HIGHER Co. READ ARITH AND PIC AND TWO TESTS OF ASSOCIATIVE THINK
=
.
=
;
P
.
..)
.
(
ING VOC AND INF 100 UNIVERSITY STUDENTS BY PEARSON FORMULA SH BY SHEPPARD FORMULA
Raw Correlations Correlations Corrected for Attenuation

.
.
.
Arith Pic Voc Inf Arith Pic Voc Inf
THE MEASUREMENT
P SH P SH P SH P SH P SH P SH P SH P SH
.
Co. Read .28 .34 .44 .31 .58 .54 .51 .37 .33 .39 .52 .40 .63 .58 .58 .43
Arith .48 .40 .23 .22 .42 .40 .59 .53 .25 .242 .49 .48
%
Pic .29 .16 .43 2 .25 .33 .20 .53 .33
...
.59 .67
)1( )3( )2( )4(

Voc .61 .6512
OF INTELLIGENCE
scores in C , A , V, D , and Inf were available . The reliabili-

ties of these scores have not been determined at all exactly ,
but they are high and approximately equal , each represent-
ing about 40 minutes of work . Table 133 presents these
correlations . The average of the intercorrelations of C , A ,
and D is .66 . The correlation between V and Inf is .81.
The average of the intercorrelations of C or A or D with
V or Inf is .62 . The change from " higher " to " lower "
abilities does not reduce the correlations as much as the
change from words to numbers within the higher abilities .
C and D correlate .81 , whereas C and A correlate only .64 ;
and D and A only .52 .
TABLE 133 .
THE INTERCORRELATIONS OF THREE TESTS OF THE HIGHER AND TWO TESTS OF

THE LOWER OR ASSOCIATIVE THINKING . 126 PUPILS IN GRADE 51 .
THE CORRELATIONS ARE ALL RAW CORRELATIONS
BY THE SHEPPARD FORMULA .
A V D Inf .
C .64 .75 .81 .59
.52 .52 .41
V .80 .81
GAYA
D .64
The correlation between what common to and

A
C
is
and what common and Inf given by

to
V
is
is
V.75 X.59 .52 X .41

X
or .77
*
V.64 X.81
9
The correlation between what common to and A

D
is is
and what common and Inf given by

to
V
is
V.80 X .64 .52 X .41

X
*
or .71
V.52 X.81
.
The correlation between what common to and

D
is
and what common and given by

to
A
is
is
D
V.64 X.81 X.52 X 1.00

or 1.11 .
V.81 X.52
The average is .86.10 ( error is in terms of σ t - o ) .
SUMMARY
We may combine the results from Mr. Tilton's 250 cases

and our two groups of 100 and 126 cases roughly , giving
equal weight to the last two determinations and as much
weight to Mr. Tilton's determination as to the other two
together . " The weighted average correlations are :
Higher with higher .53

Lower with lower .641
Higher with lower .57
What is common to two higher with
what is common to two lower .94
These facts are almost crucial . They prove that mere

association and the higher abilities have in the main the
same cause . Almost all of whatever is common to the one
sort is common to the other sort . If
we are to avoid the
conclusion that associative ability is this cause , we must
either place the causation of associative ability in the higher
ability , or seek a common cause for both which is different
from either , such as a general mental energy or vitality .
The first of these assumptions is absurd , because associa-
tive ability occurs abundantly without any trace of the
higher abilities , but these never occur without it . In the
lower animals , in idiots , and low imbeciles , and in the young
infant , mental connections are formed without the appear-
ance or use of abstraction , generalization , or relational
thinking . If either is to be derived from the other , it is
surely best to derive the higher abilities from the associa-
tive abilities . The second assumption is tolerable , though it
11 His determination probably deserves even more relative weight than
this because his test material was better adapted to bring out any differences
between the higher and the lower forms of intellect . If more weight is given ,
the higher and lower become still less distinguishable .
seems defensive and evasive . It also is entirely empty and

meaningless until the " energy " or " vitality " is expressed
as some fact known to science . What shall that fact be ?
Until that fact is chosen , the doctrine that ability in con
nection - forming , and ability in the higher processes , have a
common cause which is not the former , is a mere statement
of ignorance . We can think of no fact so suitable as C , the
physiological parallel of number of mental connections .
The cause must not be a too general vigor or health or en
ergy or sensitivity or conductivity of neurones . For the
correlations between intellect and other functions of the
nervous system are very far from perfect . Between intel
lect and mental health or balance , between intellect and sen
-
sori - motor skill , between intellect and sensory acuity , be
tween intellect and morality no one of these correlations
would be as close as the correlation between the associative
ability and " higher " ability within intellect .
We do not maintain that C is the sole cause of intellect
in original nature, so that two persons with identical num
bers of C's and identical training will necessarily have iden
tical intellectual achievement . We have already noted , as
factors which play a part , strength of curiosity , satisfying
ness of thought for thought's sake , and competition from
non - intellectual activities and interests . Other thing must
be equal , such as health and energy . There is also perhaps
a capacity for having the neurones act with reference one to
another, that is , with integration , whose low or negative
extreme is pronounced dissociation as in hysteria , and
whose high or positive extreme appears as a notable good
sense or adequacy in the use of one's experiences . This
capacity may be largely irrespective of C. There is also
perhaps a capacity for resisting intellectual panic and con
fusion , whose low or negative extreme is mania or " flight
of ideas " and whose high extreme is a notable steadiness
and regulation of each individual connection by the general
set or adjustment of the mind at the time . The strength of
this capacity may be largely irrespective of C.
These and other possible qualifications do not impair the

value of the hypothesis . They amount to making distinc
tions between basal intellectual capacity and interferences
with it or handicaps to it by disassociation as in hysterical
lapses , or by irrelevance as in mania . We should have just
as much need to make these distinctions if we attributed
differences in intellectual capacity to differences in the
quality , or in the shape , or in the chemical action of neu
rones . Nor would two sorts of neural action qualitatively
different relieve us from them .
We shall not discuss general arguments pro and con in
this report , but will simply note that both the phylogeny
and the ontogeny of intellect seems to us to show selection ,
analysis , abstraction , generalization , and reasoning coming
as a direct consequence of increase in the number of connec
tions ; and that what little is known of the status of the neu
rones in very dull individuals is in harmony with the quan
titative theory.12
12 See especially Hammarberg , Studier öfver Idiotiens , Klinik och Patologi .
CHAPTER XVI
THE MEASUREMENT OF ORIGINAL INTELLECTUAL CAPACITY

AND OF ACQUIRED INTELLECTUAL ABILITY
THE PRESENT STATUS OF OPINION

Psychologists are often credited with the opinion that
the intelligence examinations which they have devised , such
as the Binet or Army Alpha or Army Beta , measure an in
dividual's original intellectual capacity , irrespective of the
opportunities which he has had , or the time he has spent in
intellectual activities , or the zeal with which he has engaged
in them .
So extreme a view is not , however , held by any of the
leaders in this field . The nearest approach to it that we
have noted in their statements is that of L. S. Holling
worth , who defines intelligence as " the capacity for learn
ing , the capacity for comprehending and making adapta
tions to the environment " which " cannot be acquired by
any course of training " [ '23 , p . 192 ] ; and says of the Stan
ford Binet " It measures intelligence " [ '23 , p . 67 ] , when
discussing means " of singling out intelligence from all the
other factors which complicate efficiency in school work "
[ '23 , p . 62 ] .
The following quotations from Colvin , Whipple and
Terman are representative :
" There is no reasonable doubt that the present intelli
gence tests do indicate to a fair degree native ability to
learn " [ Colvin , '23 , p . 336 ] .
Colvin and MacPhail ( speaking of Professor Bagley's
article , " Educational Determinism or Democracy and the
I. Q. " ) say : " He is right if he means that it is not always
easy to determine what this innate learning capacity , this
native intelligence is , and that mistakes in individual cases
may be made ; but he is wrong if he would convey the idea
433
that in general it is not possible to determine within reason

able limits what the native learning capacity of the child is ,
provided adequate intelligence tests are employed and vari
ous common sense precautions taken " [ ' 22a , p . 114 ] .
"
Terman writes : As a matter of fact, all the ' intelli
gence testers ' will readily agree with Mr. Lippmann that
their tests do not measure simon pure intelligence , but al
ways native ability , plus other things , with no final verdict
yet as to exactly how much the other things affect the
score . However , nearly all the psychologists believe that
native ability counts very heavily " [ "22 , p. 119 ] ; and else
where , " It would , of course , be going too far to deny all
possibility of environmental conditions affecting the result
of an intelligence test . Certainly no one would expect that
a child reared in a cage and denied all intercourse with
other human beings could by any system of mental mea
surement test up to the level of normal children . There is ,
however , no reason to believe that ordinary differences in
social environment ( apart from heredity ) , differences such
as those obtaining among unselected children attending ap
proximately the same general type of school in a civilized
community , affect to any great extent the validity of the
scale " [ '16 , p . 116 ] .
Whipple writes : " In presenting these results , it ought
to be made clear at the outset that no psychologist is foolish
enough to suppose that native intelligence is the sole factor
in academic success ; all that is contended is that it is one
factor , and probably the most important single factor , and
that it is measurable by wholesale rapid methods with a
reasonable degree of precision " [ '22 , p . 262 ] ; and else
where , " We know that the organism arrives at approximate
maturity of growth in stature and in many other physical
traits in early adolescence ; the fact that our test scores
indicate the maturing at about the same time of whatever
it is we are measuring , like the fact that , regardless of
chronological age , the correlation between stature and men
tal age is high , may very well indicate that our tests are
ORIGINAL AND ACQUIRED INTELLECTUAL ABILITY 435
also measuring an intrinsic capacity which matures accord

ing to laws of its own and with relatively little influence
from the environment " [ "23 , p . 597 ] .
On the other hand , there are some emphatic assertions
that a person's score in a stock intelligence examination is ,
in very large measure , the product of purely environmental
forces . So Cyril Burt writes of the Binet : " Errand boys
and paper boys will answer smartly in the money tests . . . .
The busy little housewife from an illiterate home , who there
carries out the most intricate duties , will yet be unable to
put those duties into words . The solitary child of a cul
tured family — profiting , perhaps , rather by daily inter
course with educated adults than by special inborn gifts
will respond with an information and a phraseology beyond
anything he would spontaneously invent or acquire . •
" Of these numerous intervening agencies the most
-
potent is, without doubt , educational opportunity . Many
of the tests some of them withdrawn by Binet in his final
revision —are sheer tests of school attainments . Reading ,
writing , dictation are learnt in English lessons ; counting
and addition and subtraction of money , in arithmetic les
sons ; drawing from copy and drawing from memory , in
drawing lessons ; the date is put at the head of every written
exercise on every day of the term , and with equal regularity
is never heard and never recollected on any day of the vaca
tion . Estimated by the Binet - Simon scale , therefore , a
child's apparent intelligence must depend in no small mea
sure upon his class in school " [ '21 , p . 175 ] .
GENERAL PRINCIPLES
We have hitherto defined and measured intellect without
restrictions as to its origin , and without distinction between.
sheer ability at thinking and a love of thinking which makes
one think oftener , longer , and harder . If, however , either
the altitude or the area of Intellect CAVD is entirely due
to an original capacity that is entirely independent of the
kind or amount of training received and of the intellectual
interests inherited or acquired , the fact should be known .

If analytic measures are possible whereby certain tasks or
symptoms isolate and measure the original capacity , cer
tain others measure the element of interest or zeal , and still
others measure the potency of the environments provided ,
such would be useful in many ways . It might be impossible
to measure capacity separately from interest , and still be
possible to measure inherited intellectual promise ( due to a
mixture of capacity and interest ) separately from environ
mental alterations thereof . Or , it might be possible to mea
sure the sheer capacity apart from the interest factor , but
not to separate nature's share from nurture's . Either of
these last two possibilities would be useful .
We may best begin by certain simple axioms , or , more
modestly , truisms . ( 1 ) If two men had been subjected to
identical circumstances in life , each and every difference
between them would be due to original nature ; if two were
alike originally , all their later differences would be due to
the circumstances of life .
( 2 ) In proportion as an intellectual task is one in re
spect of which all persons have had equally adequate train
ing , so that no conceivable classification by environmental
opportunities would correlate at all with success in the task ,
that task is a measure of original capacity ( plus original
interest ) .
( 3 ) In proportion as a series of intellectual tasks gives
on the whole as much advantage to any one set of environ
mental opportunities as to any other set , that series is a
measure of original capacity ( plus original interest ) .
( 4 ) Intellectual tasks , success in which requires zero
training and is uninfluenced by any kind or amount of train
ing , do not exist and cannot exist, at least not in shape to
measure appreciable amounts of intellect .
The first three axioms are self - evident and undisputed ,
but the fourth may seem to run counter to the beliefs , or at
least the hopes , of some psychologists . Indeed , one is
tempted to think that children who are set tasks in filling
form - boards or tracing mazes , never before having seen

one , start at zero and get all the relevant training in the
course of the test itself . But that is never the case above
the lowest levels . The tracing of a maze rests upon habits
of response to location and direction which we once
learned ; the filling of the simplest form - board depends
upon habits of perception of shape and size which we learn
as truly as we learn Euclid or shorthand . It is because we
have all learned them and learned them early that we tend
to forget that they are influenced by training . Because
almost everybody has learned them , these tasks are , by
axiom 2 , more suitable ( other things being equal ) to mea
sure original capacity than shorthand or Euclid would be .
But they cannot be said to require zero training .
It is conceivable that by some direct method of examin
ing the finer anatomy or physiology of the neurones , as by
some technique analogous to the X - ray technique , original
intellectual power and interest may be separated from ac
quired ability without any reliance upon axioms 2 and 3 .
-
But so long as we measure intellect by the production of
intellectual products success with intellectual tasks we
can never reduce environmental forces to zero ;
-
we must
always seek to equalize them .
THE USE OF NOVEL TASKS

One common method of obtaining some degree of equali
zation is to make the tasks novel , so that at least no person
will have been taught to do that particular task by environ
mental forces . This has played a part in the disarranged
sentence test , the number - series completions , and the mixed
relations or analogies of Army Alpha , in the maze , cube ,
rhythm , picture - completion , and card - cutting tests of Army
Beta , in a majority of the pencil and paper tests for young
children unable to read , and in many others . Each year
brings forth new selections or adaptations or inventions of
tasks where intellect operates with novel data or with old
data in new ways .
This method had the further alleged advantage that
-
novelties were supposed to measure ability to analyze and
infer and reason the higher mental processes better than
more customary tasks could . We have seen that this ad
-
vantage may be of less moment and magnitude than has
been supposed , since the " higher " and the " lower " proc
esses measure nearly the same abilities at bottom .
The equalization of environmental influence obtained by
novelty in and of itself has one notable practical disadvan
tage . Special coaching for the tests is likely to produce
very great inequalities in favor of those who receive it .
For example , the syllogism test of Rogers shown below will
be made very much easier for many persons if they are
taught to make a diagram representing the given facts by
position along a line , as shown below .
Fill in with conclusions which can be correctly drawn from the given facts
in each set as shown in the first line .
Given Facts Conclusions

therefore Y is thicker than V
Z is thicker than X therefore X is H

(1 ) H is as thick as Z
V is thicker than H therefore Y is H
HH
V is thinner than Y therefore X is V

therefore Z is Y
therefore B is A
D is greater than B therefore D is F
B is equal to E
(2 ) E is greater than F therefore E is A
C is less than F
therefore B is C
A is greater than D
therefore A is F
X ฀
thin H V Y thick
E
less C F B D A great
THE USE OF FAMILIAR TASKS

A contrasting method is to use tasks that are so familiar
that everybody has had somewhat nearly adequate environ
mental stimulation to master them , any person's success
depending consequently chiefly on his intellectual capacity
and interest . Thus , " Are you a boy or a girl ? " " Is it
morning , afternoon , or evening ? " "
What month is it ? "
" Name all the months of the year ," tying a bow knot , nam
ing the days of the week , naming six coins , and other ele
ments of the Stanford Binet are tasks which obviously lack
novelty . Their merit is not that training has little effect
on them , but that training treats all people somewhere
nearly alike in respect of them .
The equalization of the environment's influence by
choosing tasks where it is adequate for nearly all , is aided
by relying on the third rather than the second axiom and
using an extensive sampling of things which the world in
general stimulates nearly everybody to learn or do . Thus
it may be argued that the total number of such words known
will be a better index of original intellectual capacity and
interest than the knowledge of any particular score of
words . The environment may decide that A learns words
about things and mechanisms , that B learns words about
animals and plants , and that C learns words about people
and their actions , but may well have less power over the
total number learned .
One method of accentuating the original factors seeks
to equalize environmental influence in respect of the data
or content involved in the task, by familiarity ; and seeks to
reduce environmental influence in respect of the operation
with the data involved in the task, through novelty . The
comprehension of paragraphs whose words singly are all
well known , the completion of sentences about familiar
facts in familiar words , the solution of unconventional
arithmetical problems ' are typical cases . For example , as
1 Such as :
A. If 7 multiplied by some number equals 63 , what is the number ?
30
sume that a thousand individuals aged 16 have been in at-

tendance upon school in the United States eight years or
more ( of 150 school days or more ) and are measured by a
battery of tasks chosen with care ( say , from those printed
in this volume ) to contain only words from the 4,000 known
to almost all such individuals , and to require only such
arithmetical facts and techniques as are taught in grades
three and four .
Differences in the degree of success with such a battery
of tasks might reasonably be regarded as largely indepen-
dent of differences in school environment save in so far as
these differences themselves were caused by original differ-
ences in capacity and zeal .
THE USE OF A SERIES GRADED FOR SUSCEPTIBILITY TO

ENVIRONMENTAL INFLUENCES
We can then , by one or another of these methods , select
or devise a battery of intellectual tasks or tests the score
in which (in a group of individuals of the same time , coun-
try , and general manner of life ) bids fair to be determined
B. What part of 16 equals half of 24 ?
C. How many quarters of a quarter equal half of a half ?
D. In the lines below , each number is gotten in a certain way from the
numbers coming before it . Study out what this way is in each line , and then
write in the space left for it the number that should come next . The first two
lines are already filled in as they should be :
2 4 6 8 10 12
Samples
} 11 12 14 15 17 18
1. 38 34 30 26
2. 103 95 87 79
3. 1 10 100 1000
E. Write the numbers and signs in each line below in the proper order , so
that they make a true equation as shown in the two sample lines .
6 =
3 3
Sample lines } 4 7 8 20 = X
1. 2 2 3 5 15 11
2. 2 5 6 7 10 =
3. 1 4 8 15 20 11
2 How largely will depend upon the disciplinary values of school training
and upon the extent to which our novelties are really novel .
to a considerable extent by original nature , another battery

the score in which a larger fraction of the score is due to
circumstances , and so on . The application of this series of
examinations graded from " least subject to environment "
to " most subject to environment " would provide a partial
analysis of a man's intellectual ability and attribution
thereof to his original nature and to his acquisitions . Thus ,
suppose that the examinations A , B , C , D , and E , when ap
plied to native -born white citizens of the United States ,
aged 21 , depend respectively on nature and nurture in these
proportions .
Nature Nurture
A 7 3
B 6 4
C 5 5
D
E 43 67
Suppose that individuals I and II score as follows :
A B с D E Total
I 90 80 70 60 50 350
II 50 60 70 80 90 350
Then obviously , and I II

, who are of equal present status ,
have it from very different causations . had much theI

better original equipment , but has not much improved it .
From such measurements , we could infer the relative con
ditions of individuals at the limit where original nature was
10 and environment 0 in the causation . Thus in the case
above , suppose all units to be truly equal and referable to

a true zero point . Then by original nature , would have I
120 or six times the
, original possibility of 20 that would II
have . If the environment could have been made as favor
able for him as it was for II
, he would have had a total
score six times as large as II's .

The importance of such analytic measures by a graded
series , and of such inferences about the conditions at its
limits , depends on ( 1 ) the surety with which we can pick
tasks to be differently sensitive to original capacity , ( 2 )

the extent to which the differential series will extend up
and down from the modal condition of sharing of original
nature and environment , whatever that may be , ( 3 ) the
nearness of that mode to the condition where one hundred
percent is caused by nature and zero percent by environ
ment , and ( 4 ) the restrictions which we have to impose on
environmental differences in order to work the plan at all .
Little has been known with surety concerning any of
these matters ; and we have not had the time or facilities to
make more than a beginning at the investigations which are
needed . All our work concerns these problems in the case
where environmental differences are limited to such as hold
for white individuals born and bred in the United States ,
belonging to the same generation ( born say , not over 20
years apart ) , provided with opportunity to go to school for
at least 6 years ( or 900 school days ) unless they were de
monstrably so stupid as to be unable to learn at school ,
and not deaf, dumb , blind , or insane . If, for example , we
state that environmental differences cause only K percent
of the differences found in a test in completing sentences or
solving arithmetical originals , we do not mean that the per
cent would not exceed K in a group composed half of pres
ent - day Americans and half of their ancestors fifty thou
sand years ago ; or in a group composed half of present - day
Americans and half of African pigmies ."
In
accord with the principles already stated , the tasks
which would be chosen as especially indicative of original
capacity are sentence - completions , arithmetical problems ,
and comprehension of paragraphs , especially such as re
quired the use of familiar data in new ways . At the other
extreme would be the knowledge of single words and iso
lated informational items .
When we apply these two extremes , we find that they
are really very close together . Either they do not measure
3 An investigation of tests which may be freed from these limitations in
whole or in part is being made by a group of psychologists , with funds sup
plied by the Spelman Memorial .
two very different things , one the intellect a man has by

nature , and the other the intellect which he has acquired by
training , or these two are in almost perfect correspondence .
For the correlation between whatever is common to the
tasks of one extreme and whatever is common to the tasks
of the other extreme is almost perfect . The evidence is the
same as that brought forward by Mr. Tilton and by us in
Chapter XV to prove that associative thinking and analytic ,
inferential , original thinking have almost identical roots .
As a check , we may contrast , within each form of test ,
certain elements which home and school advantages would
benefit less with certain other elements which they would
benefit more . Thus , within the field of sentence - completion ,
we may divide our elements 65 to 130 into two halves
as shown below , as a result of a consensus of expert opinion .
BENEFITED MORE BY HOME AND SCHOOL ADVANTAGES

66. The of the World were
kinder than the kings and nobles
Old .
68. The of five and ten is fifteen .
70. At time was progress rapid
during the last half of the nineteenth
74. The source wealth in Den-

mark agriculture .
75. The laws inheritance are for the most
unknown .
76. In to maintain health , one
should have nourishing
82. At ancient banquets the of the day seems
have the chief
of conversation .
84. The Declaration affirms that
the Creator all men with certain inalien-
able
85. This was done a view caus-
ing the of carfares
three cents .
93. One of the most difficult problems of representative
is that of getting large assemblies to
the work of legislation
and efficiently .
98. The word virtue is derived from a

strength .
99. The struggle for among the lower
has a commonplace of modern scientific
thought .
100. The World wished to the
New .
101. Saddles and bridles no unimportant
in the art .
102. India is rich in of scenery and climate ,
the mountains to vast
deltas raised a few
above sea
104. Undue consciousness often the flow of expres-
sion diffuseness is detrimental to a clear
and exposition of our ideas .
106. Throughout the river plains of northern India , two harvests ,
and , some provinces , are
..each
116. Few historians would the fact that Marx had
a larger and more thorough on the social
of his time any
living man .
119. Let us very briefly examine the social forces
at work concentrating or
the ownership wealth .
123. Modern inherits
innate pugnacity and all the love of glory of
124. Let the class that itself to transportation , for
example , working and the disastrous
to the rest of the can
scarcely imagined .
125. The monuments of Persepolis
the use incense as
in ancient Persia as Baby-
lonia .
127. The orderly peaceful of our
industrial mechanism is of public
and be secured in
way or
128. Ever since the hearing before him the gov-
ernor giving
spare moment a of the case .
129. So far the displeasure of
the people by the will of their represen-
tatives , a President generally gains by the
bold use of his veto power . It conveys the
firmness ; it
shows
has a view and does
to give effect to it .
BENEFITED LESS BY HOME AND SCHOOL ADVANTAGES

65. Itmay effort and a long but
the result is sure .
71. He will come to the meeting
the fact he
rather stay quietly at home .
73. No what happens wrong is
right .
77. His friends , wished to dissuade him from this
undertaking , asserted that he followed
their advice would withdraw their sup-
port .
78. It
would several pages to
contain the list .
79. Standing beside the grave
great Englishman enough
for us to know lived and
died , and made the his heirs .
80. You may safely conclude that you in yourself
the means of at the truth .
81 . the fact that you disagree with me , shall I
continue to aid you .
83. As the treasure he had come to seek , prob-
ably it existed in his own
86. The of a man is to be useful to his
87. They who are miserable have medicine other

hope .
88. The best is too
him .
89. The sublime the ridiculous
often so nearly it is
to class them separately . One step above
the makes the ridiculous , and one step
the makes the
again .
90. spite many severe
he is still alive the of ninety-
one .
91. It appears whether his debts will be paid .
94. If the of the year were holi-
days playing be tedious
working .
95. Gratitude the fairest that

sheds perfume
heart .
97. Impatience , it
while us to surpass
generations , disposes to
overrate their happiness .
103. The American boy wishes go
college and to go , has
only his own weaknesses to thank for it .
104a . injury nor retaliation nor warding off
by evil is ever
105. Knighthood and Chivalry are words
are nearly not
synonymous .
107 . a man time sufficient for all
laudable pursuits , and sufficient for all
generous purposes , he is free
shadow of blame or reproach .
108. Maize contains small a proportion of nutri-
tious matter it not
for horses which fast
work is
109. The drafting a measure depends
the pains and skill exerted by its
110 . is natural that being dissatisfied with the

we should form a too
estimate of the past .
115. Virtue knowledge , and is the
fruit of ignorance .
117. He will do as you request
his own feelings .
117a . Where in nature is grandeur
displayed as in the Grand Canyon ?
126. He would assign no reason his action
to his
to enemies .
Within the field of arithmetical problems the same con-

sensus lists ten problems as those benefited most by home
and school advantages , and nine as benefited least by them .
We put these "
Most " and " Least " tasks into four
groups : Ma , Mb , La , and Lb , Ma and Mb being random
halves of the M's , while La and Lb are random halves of
the L's .
In a group of two hundred normal - school students , the

correlations are as follows :
Ma with Mb , .74
La with Lb , .81
Ma with La , .69
Ma with Lb , .66
Mb with La, .67
Mb with Lb , .72
The correlation between the " Most " and the " Least "
is thus nearly as high as the correlation within the " Most "
or within the " Least , " and the correlation between what is
common to the " Most " and what is common to the " Least "
is .88 .
On the whole , the difficulties in the way of analyzing a
man's intellect into the contribution of nature and that of
nurture by the use of tasks much subject to environmental
influences , and tasks little subject to them , are very great .
The method is sound , but hardly practicable .
THE TEST AND RESULTS OF BURT
Burt has sought to measure the relative shares of intel

ligence , school environment , and age . His work is so im
portant that we quote the report of it in full . In estimating
the meaning of his results , we must bear in mind that his
" intelligence" is in reality the score in the test ( Test 29.—
Graded Reasoning Test ) quoted below . His final conclu
sion , that B ( the Binet Mental Age ) = .54S + .331 + .11A ,
should then be modified to read as follows :
" Of the gross result , then , one - ninth is attributable to
age alone , one - third to the ability measured by the Burt
Reasoning Tests alone , and over one -half to the ability mea
sured by school attainment alone . "
Test 29. - GRADED REASONING TESTS
(Short List )
1. Tom runs faster than Jim :
Jack runs slower than Jim .
Which is the slowest of the three ?
7 Years .
2. Kate is cleverer than May .
-
May is cleverer than Jane .
Who is the cleverest Jane , Kate , or May ?
3. Ihave bought the following Christmas presents : a pipe , a

blouse , some music , a box of cigarettes , a bracelet , a toy
engine , a bat , a book , a doll , a walking - stick , and an
umbrella .
My brother is eighteen : he does not smoke , nor play cricket ,
nor play the piano .
Iwant to give the walking - stick to my father , and the umbrella
to my mother .
Which of the above shall I
give my brother ?
8 Years .
4. I don't like sea voyages

And I don't like the seaside
:
.
I must spend Easter either in France , or among the
Scottish Hills , or on the South Coast .
Which shall it be ?
5. The person who stole Brown's purse was neither dark , nor tall ,
nor clean - shaven .
The only persons in the room at the time were
1. Jones , who is short , dark , and clean - shaven :
2. Smith , who is fair , short , and bearded :
3. Grant , who is dark , tall , but not clean - shaven .
Who stole Brown's purse ?
9 Years .
6. Three boys are sitting in a row :
Harry is to the left of Willie :
George is to the left of Harry .
Which boy is in the middle ?
7. In cold , damp climates , root crops , like potatoes and turnips ,

grow best :
In temperate climates , there are abundant pastures , and oats
and barley flourish :
In sub -tropical climates , wheat , olives , and vines flourish :
In tropical climates , date - palms and rice flourish .
The ancient Greeks lived largely on bread , with oil instead
of butter : they had wine to drink and raisins for fruit .
Which climate do you think they had ?
10 Years .
8. There are four roads here :
I have come from the south and want to go to Melton .
The road to the right leads somewhere else :
-
Straight ahead it leads only to a farm .
In which direction is Melton North , South , East or West ?
9. The doctor thinks Violet has caught some illness .

If she has a rash , it is probably chicken - pox , measles , or
scarlet fever :
If she has been ailing with a cold or cough , she may develop
whooping - cough , measles , or mumps .
She has been sneezing and coughing for some days : and now
spots are appearing on her face and arms .
What do you think is the matter with Violet ?
11 Years .
10. Where the climate is hot , gum - trees and rubber will grow :
Heather and grass will grow only where it is cold :
Heather and rubber require plenty of moisture :
Grass and gum - trees will grow only in fairly dry regions :
Near the river Amazon it is very hot and very damp .
Which of the above grows there ?
11. Father has just come home in a brand new overcoat : there is
clay on his boots and flour on his hat .
The only places he can have been to are Northgate , Southgate ,
Westgate , or the City ; and he has not had time to go to
more than one of these .
There is no clay anywhere in the streets except where the
pavement is up for repair .
There are tailors ' shops only in Southgate , Westgate , and the
City .
There are flour mills only in Northgate , Westgate , and the City .
I know the roads are not being repaired in the City , though
they may be in the other places .
Where has father been ?
12 Years .
12. Field - mice devour the honey stored by the humble - bees : the
honey which they store is the chief food of the humble - bees .
Near towns , there are far more cats than in the open country .
Cats kill all kinds of mice .
Where , then , do you think there are most humble - bees
in the neighbourhood of towns or in the open country ?
13. I started from the church and walked 100 yards :

I turned to the right and walked 50 yards :
I turned to the right again and walked 100 yards .
I
How far am from the church ?
13 Years .
14. A pound of meat should roast for half an hour :

Two pounds of meat should roast for three - quarters of an hour :
Three pounds of meat should roast for one hour :
Eight pounds of meat should roast for two hours and a quarter :
Nine pounds of meat should roast for two hours and a half .
From this can you discover a simple rule by which you
can tell from the weight of a joint for how long it
should roast ?
15 . What conclusion can you draw from the following facts ?

Iron nails will not float in a pool :
A cup of pure gold dust weighs nearly twenty times as much
as a cup of water of the same size :
If you drop a silver sixpence or a copper coin into a puddle ,
it will sink to the bottom :
A cubic inch (about a tablespoonful ) of water weighs less than
half an ounce ; a cubic inch of brass weighs over two
ounces :
A leaden weight will drop to the bottom of the ocean .
Sum up all these observations in one short statement of the
following form : " Most are
""
14 Years .
16. John said : " I

heard my clock strike yesterday , ten minutes
before the first gun fired . I
did not count the strokes , but
I
am sure it struck more than once , and I
think it struck
an odd number . "
John was out all the morning from the earliest hours : and his
clock stopped at five to five the same afternoon .
When do you think the first gun fired ?
17. Captain Watts and his son James have been found shot the -
father in the chest and the son in the back . Both clearly
A
died instantaneously
gun fired
-
.
close to the person -
as , for example , when a man
shoots himself will blacken and even burn the skin or
clothes fired from a greater distance , it will leave no such
mark.
The two bodies were found near the middle of a large hall used
as a rifle range . Its floor is covered with damp sand ,
which shows every footprint distinctly . Inside the room
there are two pairs of footprints only . A third man stand-
ing just outside the door or window could aim at any part
of the room : but the pavement outside would show no foot-
marks .
Under Captain Watts ' body was found a gun : no such weapon
was found near James .
In each case the coat , where the bullet entered , was blackened
with gunpowder , and the cloth a little singed .
Captain Watts was devoted to his son , and would have died
sooner than harm him purposely : hence it is impossible to
suppose that he killed him deliberately , even in self-
defence . But some think that James secretly disliked his
father, and hoped to inherit his fortune at his death .
( 1 ) Was Captain Watts ' death due to murder , accident , or
suicide ?
( 2 ) Was James ' death due to murder , accident , or suicide ?
[ Burt , '21 , pp . 239-242 . ]
" For every child in an entire school , comprising just over three
hundred pupils aged between seven and fourteen , have secured I
the following measurements : first , the child's age ; next , his school
attainments , measured by an educational examination , the results
being revised by the teachers ; thirdly , his intelligence measured by
special tests of reasoning , the results , again , being checked by the
teachers ; and , lastly , his mental age , given directly by the present
version of the Binet - Simon scale , unchecked and unrevised .
" The first column of figures in Table XX . ( our Table 134 )
shows the six correlations subsisting between these four measure-
ments coupled with one another in every one of the six ways
possible .
" From the six ' total ' coefficients , taken each in turn , have I
first of all eliminated one or other of the four factors operative .
From the gross figures I
have , by discount , found the net . The
resulting ' partial ' coefficients are given by the second column of
figures in the table . A comparison of these values at once invites
several inferences . The resemblance between the Binet - Simon
results and the child's school standing seems due more to the com-
mon influence of age than to the common influence of intelligence .
The resemblance between the Binet - Simon results and the child's
intellectual maturity , estimated independently , seems due more to
4 See Appendix IV
. , pp . 239–242 .
5 With a group of nearly 300 children , the probable error for correlations
less than .12 ranges between +.038 and +.039 . A under .07 , there-
coefficient
fore , has little or no significance ; one over .11 may be received as trustworthy .
-
the common influence of school standing than to the common influ
ence of age . The estimates for intellectual maturity owe their
-
correlation with school standing a correlation by no means high
even at the outset chiefly , but not entirely , to the common influ
ence of age . When the influence of intelligence is excluded , there
that is unexpectedly -
still remains a correspondence between age and position in school
indeed , I
apprehend , unwarrantably
promotion goes suspiciously with seniority . The negative correla
close : -
tion between school standing and intelligence , obtained when dif
ferences in Binet age are eliminated , may seem odd ; but even were
it larger than it is , it would not be at all inexplicable . In a group
homogeneous in regard to mental age , children who are older
chronologically would , in a test measuring inborn intelligence rather
than mere mental growth , appear duller ; yet , because they are
older , the school system elevates them to a somewhat higher class .
Hence the paradox of a group whose mental age is uniform : the
higher the class , the duller the child .
TABLE 134. ( Table XX of Burt ) .

OBSERVED PARTIAL CORRELATIONS BETWEEN AGE , INTELLIGENCE , SCHOOL
AND
ATTAINMENTS , AND THE RESULTS OF THE BINET -SIMON TESTS .
Partial Partial
Factors Observed Factor Coefficient Factors Coefficient
Correlated Coefficients Eliminated (First Eliminated ( Second
Order ) . Order ) .
B Tests and Intelligence .78 Intelligence

School Work .91 Age .68 and Age .61
B Tests and School Work .58 School Work
Intelligence .84 Age .65 and Age .56
B Tests and Age .83 School Work .19 School Work
Intelligence .62 and
Intelligence .13
School Work and Tests -.06 Tests and
Intelligence .75 Age .40 Age -.07
School Work and Tests .49 Tests and
Age .87 Intelligence .73 Intelligence .49
Intelligence and Tests .01 Tests and
Age .70 School Work .15 School Work .05
" Let us now examine the partial coefficients of the second order ,
coefficients , that is , obtained where two factors have been cancelled
in succession ( last column of Table XX ) .
"Intelligence , it may be remembered , was observed to correlate
with the Binet tests by .84 and with school attainments by .75 .
Mediated solely by intelligence , therefore , a correlation between the
Binet estimates and school attainments could be predicted amount
6 The coefficient in question is barely twice its probable error .
ing at least to .75 X .84 , that is , .63 . The total correlation found ,
however , was as much as .91 . The excess is due , in part at least ,
to the second common factor of age . But , on eliminating also the
effect of age , there is still left a substantial surplus . With both age
and intelligence constant , the ' partial ' correlation between school.
attainments and Binet results remains at .61 . Of all the partial
coefficients of the second order this is the largest . There can , there
fore , be little doubt that with the Binet - Simon scale a child's mental
age is a measure not only of the amount of intelligence with which
he is congenitally endowed , not only of the plane of intelligence at
which in the course of life and growth he has eventually arrived ;
it is also an index , largely if not mainly , of the mass of scholastic
information and skill which , in virtue of attendance more or less
regular , by dint of instruction more or less effective , he has progres
sively accumulated in school .
" The correlation of .49 between age and educational attainment ,
left after the elimination of ability both tested and observed , con
firms our previous suspicion of the undue influence of age upon
school classification . The only other correlations surviving after
the double elimination are those between the Binet tests , on the one
hand , and intelligence and age respectively upon the other .
" From the three final correlations thus furnished by the tests ,
and from the relevant standard deviations , can be calculated the
several so - called ' regressions . '
-
relative proportions in which the three factors
and school attainments
-
The regressions will indicate the
age , intelligence ,
together determine a child's achievements
in the Binet - Simon tests . The complete equation is as follows :
B.54 S.33 I.11 A,
where B = mental age according to the Binet - Simon scale ,
S = school attainments expressed in terms of educational
I = intellectual
age ,
development also measured in terms of
,
years and
A= the chronological age .
" Of the gross result , then , one - ninth is attributable to age , one
third to intellectual development , and over one - half to school attain
ment . School attainment is thus the preponderant contributor to
the Binet - Simon tests . To school the weight assigned is nearly
double that of intelligence alone , and distinctly more than that of
intelligence and age combined . In determining the child's per
formance in the Binet - Simon scale , intelligence can bestow but little
more than half the share of school , and age but one - third the share
of intelligence . " [ Burt , '21 , pp . 181 to 183. ]
We are greatly indebted to Burt for this study , espe

cially for the application of the partial - correlation tech
nique to the problem , but we should be very cautious in our

use of the facts . Small differences in the observed correla-
tions may make large differences in the partials . These
small differences may be in whole or in part due to sam-
pling error , or to differences in the amount of chance error
in the original measures producing differences in the atten-
uation . Suppose , for example , that Burt's Reasoning Tests
really measured exactly the same ability as the Binet , but
with a greater chance error ( the two reliability coefficients
being .95 and .80 ) and that the true correlations for a very
large population , each person being perfectly measured ,
were as shown in Table 135 .
Ifthe facts were so , I
and B would measure identical
abilities and the regression equations , B = aS + bI + cA
I
and = dS + eB + fA would be identical save that B and I
would change places .
Yet all that is required to produce Burt's six coefficients
from those of Table 135 is that the self - correlation of Burt's
I
determinations of be .15 points lower than that of his de-
termination of B (somewhat lower it almost certainly is ) ,
and that , by the sampling error of his 300 cases , his corre-
I
lation of B with be .03 too low, his correlation of S with I
I
be .09 too low , and his correlation of with A be .06 too
low . There is one chance in fifty for the first of the three ,
one chance in 5,000 for the second , and one chance in sev-
enty for the third . The chance of all three occurring to-
gether is very small ( though by no means so small as
1/50 X 1/5000 X 1/70 , since the same sort of cases that
would lower FBI would tend somehat to lower rs and r₁ also ) ,
and there is an equal chance that 30,000 cases in place of 300
would strengthen Burt's argument . Indeed , we agree with
Burt that score in the Binet series is influenced by school
training , and probably to a greater extent than score in
Burt's reasoning test series will be . We very much doubt ,
however , that the amount of influence is so large in the one
case or so small in the other as the second - order partials of
.61 and -.07 indicate .
It is perhaps worth while to compute what these partials

would be if we assume that all of Burt's coefficients are
valid , but assume in addition that the self - correlations of
his four measures with a second set of independent mea-
sures of the same qualities or abilities ' are .95 for B , .80
TABLE 135 .
THE INTERCORRELATIONS OF ONE BINET TEST ( B ) , ONE BURT REASONING TEST

( I ) , ONE MEASURE OF SCHOOL WORK ( S ) , AND AGE ( A ) BY CERTAIN
ASSUMPTIONS CONCERNING THE INTERCORRELATIONS IF AN INFI-
NITE NUMBER OF SUCH TESTS HAD BEEN USED .
1 2 3 4
10000 T₁ by Burt Difference ( 2–3 )
฀
BS .96 .91 .91 0
BI 1.00 .87 .84 .03
BA .85 .83 .83
SI .96 .84 .75 .09
SA .89 .87 .87
ΙΑ .85 .761 .70 .06
for I , 1.00 for A , and

TBS.IA
-
.95 for S. They
.91 and IIS.BA
will be
- .79 .
:
Finally , as further evidence of the need of caution in

arguments from small differences between correlations , let
us apply the well - known attenuation formulae to Burt's six
correlations , and obtain their answers to certain questions .
The first is : What is the correlation between ( a ) whatever
is common to the ability measured by the Binet and the
ability measured by the Burt Test , and (b ) whatever is
common to School work and Age ? The answer is :
* V.83 X .91 X .87 X .75
or .98 ; that is , almost everything .
V.84 X.87
The second is : What is the correlation between ( a ) what-
7 To prevent possible ambiguity , we add that we mean , not repetitions of
the particular Binet and Burt tasks , but a set drawn at random from the same
general collection of tasks from which the Binet Tests may be considered
a random draft , and a set drawn from the same general collection of tasks from
which the Burt Tests may be considered a random draft .
31
ever is common to the ability measured by the Binet and

School work and (b ) whatever is common to the ability
measured by the Burt Test and to that measured by Age ?
The answer is :
*V.84 X.83 X.75 X.87

or 1.02 ; that is , everything .
V.91 X.70
We obtain a similar answer ( 1.03 ) to the third question
concerning the correlation between whatever is common to
Binet and Age and whatever is common to Burt and School
work ?
These answers are absurd . Not the most passionate
adherent of one general ability as the cause of all possible
excellences would assert that whatever is common to the
ability to score well in Binet and Burt is also common to
Age and School Attainment . Neither Burt nor any other
competent psychologist would entertain the notion that
what is common to success in the Binet and success in
School is the same thing that is common to success in the
Burt and Age .
Such fantastic perfect correlations between common
factors sometimes are due to a statistical fallacy , there
being really nothing common in either pair of traits ; but
Dr. Burt will be too sagacious to plead this in the present
case .
The fact is that when a number of traits are somewhat

nearly equally intercorrelated , as is the case with B , , S , I
and A in Burt's 300 pupils , partial correlations and in
ferred correlations between common factors will often show
queer , not to say absurd , results . Both procedures are of
very great value , but they are very sensitive to the influ
ence of the " errors " due to measuring traits in too few per
sons and with too few tests per person .
The partial correlations computed by Burt for the in
fluence of the separate features of school work seem to us
in better harmony with the view that B and I are two tests
differing moderately in their susceptibility to school train
ing than with his view that B ( for persons alike in ) is I

very susceptible and I ( for persons alike in B ) not at all so .
They are :
TABLE 136 ( XXI of Burt ) .

OBSERVED AND PARTIAL CORRELATIONS BETWEEN THE BINET - SIMON TESTS AND
ATTAINMENTS IN THE SEVERAL SCHOOL SUBJECTS .
Partial Coefficients
Observed ( Age and Intelligence
Coefficients Eliminated )
Composition .63 .32

Reading .54 .26
Dictation .52 .21
Arithmetic ( Problems ) .55 .07

Arithmetic ( Mechanical ) .41 .15
Writing .21 .01
Drawing .15 -.08
Handwork .18 -.06
It is easy to see how Composition might be a better

symptom than Dictation and Arithmetic of those parts of
intellect which B measured but I
failed to measure , but
hard to see how training in it could improve B more than
training in Dictation and Arithmetic did .
Burt's study, then , though it is by one of the most ex-
perienced and able workers in this field and is the most ex-
tensive and the most searching study of the problem that
we have found , does not convince us either that his Rea-
soning Tests , when given to a group identical in age and in
ability measured by the Binet , will measure original capac-
ity uninfluenced by school advantages , or that the Binet ,
when given to a group identical in age and in the ability
measured by his Reasoning Test score , will measure a com-
posite made up , half or more , of school advantages . we If
knew of any test that did the former , we should have re-
peated his experiments testing out both the Binet and the
Burt with the aid of the test of sheer original capacity
instead of making this long counter - explanation of his
results . But our general study of the problem has not re

vealed any such test . To this general study we may now
return .
THE USE OF ALTITUDE AND WIDTH OF INTELLECT

According to the orthodox views of what original nature
is likely to contribute and what the environment is likely to
contribute , it would be reasonable to choose the altitude of
intellect and the width W ( 1C or 1A or 1V or 1D , etc. )
as the two extremes , the area A ( 1C or 1A or 1V or 1D, etc. )
being intermediate in its causation . It seems , at least , much
easier for a good home or school to increase the number of
easy things which a child can do than to enable him to do
harder things than he has ever done . Dull men who could
never learn to use indirect discourse correctly in Latin can
learn the easy features of a score of languages . A favor
able opportunity and assiduity seem to be all that are
needed to teach anybody twice as many thousand easy ac
complishments as he has acquired with meagre opportunity
and less study .
This seems almost axiomatic . We were almost con
vinced of it until we investigated the actual relations be
tween altitude , width , and area of Intellect CAVD and be
tween the higher selective and organizing abilities and
the lower or associative . The correlations are such as to
cast doubt upon the doctrine that the number of easy in
tellectual accomplishments which a person learns depends
chiefly , or even largely , on the stimulus of the environment .
On the contrary , the number which a person can learn
seems to be limited by his nature almost as much as is the
degree of difficulty which he can master . If, by a miracle ,
intellectual accomplishments were all of exactly the same
difficultywe have reason to believe that the number which
,
a person had learned at a given age would show the same

hereditary relations as are shown now by the altitude
which he reaches . As things are , the competent intellects
learn approximately all the easy things which the incom
petent intellects learn , plus a large balance of harder things .

The imbecile probably could not learn twice as many things
as he now does , no matter what opportunities were pro
vided for him .
The explanation of the difference between the expecta
tion from general psychology and the results shown in our
correlations is to be found in the fact that there are rela
tively so few easy intellectual accomplishments . In order
to increase the number of important things an intellect can
do , you soon have to put it to doing harder things , because
the easy ones are so soon used up . If there were ten mil
lion different single tasks at each level from 25 to 45 , it
would perhaps be possible to take two intellects of equal
original capacity , and by certain deprivations hold one
down to three million at level 25 , and by certain advantages
stimulate the other to reach four or five or six million at
that level . And doubtless any wise psychologist would con
sider this attempt under these conditions much more prom
ising than the attempt by an equal difference in depriva
tions and advantages , to hold one down to inability to do
anything harder than 25 , while the other was made able to
do tasks at 35 or 40. But the conditions are unreal . To
push the second intellect's score up by two or three million
would , as things are , mean to have him master tasks above
level 25 .
Another part of the explanation is found in the fact that
altitude as it is used in our correlations is not the same as
the altitude meant in the statement that it is much easier
for environment to increase the number of easy things an
intellect can do than to enable it to do harder things than
it has ever done . To do a harder thing means for us to do
a certain percent of the tasks at a higher altitude , one in
forty, or two in forty . Now , if there are twice as many
tasks at level 31 as at level 30 , the intellect which is stimu
lated to do 1 in 40 at level 31 has to learn twice as many
tasks as the intellect which is stimulated to do 16 instead of
15 out of 40 at level 30. This is naturally harder for the
environment or for any other force .
A third part of the explanation is to be found in the fact

that our correlations represent the status as it is created by
such differences in environment as do now act , not the
status as it might conceivably be. If
the action of the en-
vironment is positively and closely correlated with that of
original nature , so that the best born are the best bred , and
the worst born the worst bred , approximately , the high cor-
relations between altitude and W ( 1C or 1A or 1V , etc. )
which we find , may not be necessary as results of the nature
of intellect, and may be alterable , as by giving equally ad-
vantageous training to all . Or , if the effect of difference in
environment is now very small in comparison to the effect
of original nature , so that the high correlations found are
chiefly due to original nature , these may be alterable by in-
creasing the differences in environment .
We may then retain in a somewhat tempered form the
expectation that environment can extend the area of intel-
lect by adding width somewhat more easily than by adding
altitude .
So far as concerns the special question of a differential
diagnosis between nature and nurture in the world as it
now is , however , we shall receive almost no assistance by
using altitude for the symptom of the former , width W ( 1C
or 1A or 1V or 1D , etc. ) for the symptom of the latter .
Where the width is not 100 % or 0% , its correlation with
altitude is too near perfection . In the world as it now is ,
they are due to almost the same causes .
OTHER METHODS OF SEPARATING ORIGINAL CAPACITY FROM

ACQUIRED ABILITY
There are two other ways of approaching the problem ,

besides that of differential tests . The first is to approxi-
mate original nature by measurements early in life before
environmental forces have had much opportunity to act on
intellect . The second is to measure a man's intellect as we
find it and to make the best allowance we can for the favor-
able or unfavorable action which the environment has had .
The former is not so fantastic as it would have been a

few years ago . It would then have been even more absurd
to claim to measure original capacity for intellect at three
or four or five than to claim to measure at that age the orig
inal tendency to adult stature . We now have far better
tests of intellect at low levels and in early stages ; and the
correlations between intellect at three and intellect at
thirty , environment being equalized , may be closer than we
think , much closer for example than the correlations be
tween stature at three and stature at thirty with equalized
environment . The contribution of original nature is all
there in the individual at three years , or at three days .
How much of it is revealed in external behavior , and how
much is hidden in the constitution of the neurones , is a
question for investigation .
On the other hand , we have to reckon with the evidence
and arguments now being brought forward to prove that
the environment of the first three years is very potent .
Freud and others contend that the trends of character are
much influenced by the environmental forces acting in these
years ; and we may expect some of them soon to make the
same contention with regard to intellect .
The method of measuring original capacity by measur
ing attainment and making an allowance for the benefits
and handicaps of environment , is what has been and is used
in many scientific investigations and in practice by wise
educators and advisers of youth . For example , of two boys
making equal scores in Army Alpha , one from an English
speaking family with four years in an excellent city school ,
the other from an immigrant family and an inferior school ,
the second will be rated as the better in natural intellect
and future intellectual promise . Its success depends , of
course , on the adequacy of one's knowledge of what the
environment has been in each individual case , and on the
wisdom of one's theories concerning the action of environ
mental forces on intellect .
In far as our work improves the measurement of in

so
tellect , it will improve this sort of measurement of , or in
ference about , original capacity for intellect . The balance
of the program is to improve measurements of the environ
mental forces which help or hinder the attainment of in
tellect .
SUMMARY
On the whole , the problem of analyzing a person's in

tellectual ability into an amount due to nature and an
amount due to nurture , is unsolved . No task or test has
been proved to be a measure of the former alone . The wis
est procedure at present is to equalize environmental forces
by using a wide variety of data with which all individuals
have had adequate experience , and to make as correct al
lowances as we can for what we cannot equalize . With the
progress of science , we may hope for accurate measure
ments of intellect at earlier and earlier ages , and for truer
rules for making allowances for environmental differences .
CHAPTER XVII
CHANGES IN THE ALTITUDE AND AREA OF INTELLECT WITH AGE
It is not at present possible to distinguish at all accu

rately within the general gain in intellect with age from 0 to
15 years or later , the share of mere maturity , mere inner
growth , from the share of the experience and training which
age implies . The reasons for this failure are much the
same as for the failure to distinguish the shares of original
capacity and environmental circumstances in the determi
nation of intellect. We lack tests which measure maturity
by itself , and tests which measure training by itself . We
also lack extensive investigations using the partial - correla
tion technique with such imperfect symptoms of maturity
and training as are available . In this chapter , therefore ,
age means chronological age , and whatever it involves
under the conditions of present - day life in America .
ALTITUDE
The curve of altitude of Intellect CAVD with age is of

the general parabolic form shown in Fig . 64. There is a
rise from 0 to about 30 at 61 , to about 341 at 101 , and to
about 361 for adults 21 years old.¹
1 These estimates subject to a thorough -going investigation which is
are
being made by Miss Rowell , using our data and additional data obtained by
her . They are made from the following facts : The average altitude CAVD
of imbeciles of mental age 6-0 to 6-11 is 30.1 . Taylor [ '23 ] has shown that
ordinary children of chronological age 6-0 to 6-11 do not differ much in such
tasks as these , from imbeciles of mental age 6.0 to 7.0 . The average altitude
CAVD of pupils 10 yr . 0 mo . to 10 yr . 11 mo . in grade 4B of School X , is
33.85 . The average altitude CAVD of pupils of the same age in grade 5A
of School X is 33.9 . The average for pupils of the same age in 5B of School
X , is 35.15 . ( A is the lower half ; B is the upper half .) In School X the
ordinary pupil 10.0 to 11.0 in age is in grade 5A or 5B . The numbers of
ages 9.0 to 10.0 grades 4B , 5A and 5B are 86 , 24 and 27 ; those of age 10.0
to 11.0 are 39 , 47 and 88 ; and those of age 11.0 to 12.0 are 18 , 29 and 48 .
The ordinary pupil of age 10 may then be expected to score somewhat above
463
It has
been shown that the score in stock intelligence ex-
aminations such as the National , Otis , and Haggerty , is
substantially a measure of altitude . Consequently , we may
take the curves of these scores in relation to age as approxi-
mate curves for the altitude of intellect in relation to age ,
first transposing the scores into terms of equal units .
This has been done for the National A , Otis Advanced ,
and Haggerty Delta 2 , with the results shown in Table 137 .
The three examinations do not agree at all closely , the Hag-
gerty official norms being especially divergent . Scores in
equal units are available for all for the interval from age
10 to age 15. The gain from age 10 to age 11 is 28 percent
of the gain from 10 to 15 in the case of the National ; 25 per-
cent of it in the Otis , and 20 percent of it in H ( I ) , and 33
II
percent in H ( ) . The gain from 11 to 12 is 27 , 22 , 17½ , and
31 percent of the gain from 10 to 15 in the National , Otis ,
I II
H ( ) , and H ( ) , respectively . Corresponding percents
for the gain from 12 to 13 are 25 , 18 , 15 , and 35 ; for the
gain from 13 to 14 they are 15 , 18 , 21 , and 6 ; for the gain
from 14 to 15 they are 5 , 17 , 261 , and — 5 .
The discrepancies are about as great if the original
scales are used , the percents of the gain from 10 to 15 then
being :
Haggerty Haggerty
Otis Delta 2 Delta 2
Age National A Advanced Official Madsen
10-11 28.3 23.6 18.3 33.4

11-12 26.7 21.8 18.3 30.9
12-13 25.0 18.2 16.7 34.3
13-14 15.0 18.2 21.7 5.6
14-15 5.0 18.2 25.0 - 4.2
the average of 33.9 and 35.15 , which is 34.53 . As a check on the estimate of
34 or more for age 10 , we have the fact that the 86 nine year olds in grade
4B have a median score of 34.0 . In this school the ordinary 9 year old has
reached grade 4B or , less often , 4A . The average altitude CAVD of the 44
adult recruits in the United States Army is 36.5 . These are enlisted men
chosen for training in the Signal Corps , all but one between 18 and 25 years
old , whose median amount of schooling is grade 8 , and who may therefore be
taken to represent the median of the white population , 18 to 24 years old , or
a point a little above it in intellect CAVD .
CHANGE IN INTELLECT WITH AGE 465
There is evidently need for a careful critical study of

these facts . If the
three examinations measure the same
thing , and if the averages or norms for the ages 10 to 15
are correctly determined , the curves in relation to age
should be the same.
TABLE 137 .
THE RELATION OF SCORES IN NATIONAL A , OTIS ADVANCED , AND HAGGERTY
DELTA 2 , TO AGE . H ( I ) REFERS TO THE OFFICIAL AGE -NORMS ;
II
H ( ) REFERS TO THE MADSEN RESULTS .
Age Scores by the original scale Scores by a scale of equal units

Nat . Otis H ( I ) H ( ) II Nat . Otis I
H ( ) H ( ) II
8 46 25 25 38.8 51.2
9 61 40 43 56.6 64.5 38.6 57.8
10 76 55 55 66.8 78.0 54.9 56.1 68.8
11 93 68 66 78.7 94.0 68.6 67.9 79.6
12 109 80 77 89.7 109.5 80.6 78.1 89.7
13 124 90 87 101.9 124.0 90.3 87.1 101.3
14 133 100 100 103.9 132.4 100.0 99.3 103.3
15 136 110 115 101.8 135.4 109.5 114.8 101.8
16 120 106.7 119.0 106.9
17 127 125.7
18 130 128.6
19 130 128.6
Gains By the original scale By the scale with equal units

Nat . Otis H ( I ) H ( II ) Nat . Otis H ( I ) H ( II )
8-9 15 15 18 17.8 13.3
9-10 15 15 12 10.2 13.5 16.3 11.0
10-11 17 13 11 11.9 16.0 13.7 11.8 10.8
11-12 16 12 11 11.0 15.5 12.0 10.2 10.1
12-13 15 10 10 12.2 14.5 9.7 9.0 11.6
13-14 9
-
10 13 2.0 8.4 9.7 12.2 2.0
14-15 3 10 15 2.1 3.0 9.5 15.5 - 1.5
15-16 10 4.9 9.5 4.9
16-17 7 6.7
17-18 3 2.9
18-19 0 0.0
The disagreement between the Haggerty official age

norms and Madsen's results is especially noteworthy be
cause Madsen's selection of 14- year - olds and 15 - year - olds is
presumably of the superior , the duller ones being more
likely to leave school . The average gains from 13 to 14 and

from 14 to 15 for all children would consequently tend to
be even less than those for his groups .
The general drift of these determinations is toward a
parabolic curve of the form of Fig . 64. If
the two sets of
determinations ( by equal - unit scores ) of the Haggerty
Delta 2 are combined with equal weight , and if the three
curves for the three examinations are then combined with
equal weight, the rise in the ordinates from 10 to 11 , 11 to
12 , 12 to 13 , 13 to 14 , and 14 to 15 are in the proportions :
26.5 , 24.4 , 22.7 , 15.5 , and 10.75 .
FIG . 64. The general nature of the relation of altitude of intellect to age
in years , 0 to 20 .
These results for altitude of intellect in relation to age

may be compared with those attained by Brooks [ '21 ] , in
his careful and extensive experiments . In general our work
corroborates his . The gains in what he terms the " higher "
functions from 10 to 15 by years according to his final com-
bined table [ '21 , p . 68 ] , allowing equal weight to the boys
and to the girls are in the proportion 24.8 , 19.7 , 19.4 , 18.8 ,
and 17.4 . These gains may all be unduly large and the later
gains at later ages unduly large in comparison with the
gains at earlier ages , because of the practice effect which
retesting involves . Data are not available to correct for it .
CHANGE IN INTELLECT WITH AGE 467
Probably an allowance of one - twelfth of the gains from age

10 to age 15 would be considered enough by all students of
the matter , and too much by a majority of them . this If
very large allowance is made , the gains are in the propor-
tions : 28.2 , 19.5 , 18.9 , 17.9 , and 15.5 . Neither in our CAVD
results , nor in the National - Otis - Haggerty estimate , nor in
Brooks ' results is there any justification for the doctrine
that the gain in altitude of intellect of the sort measured
by existing intelligence tests is zero after 14 , or after 15 , or
even after 16. It decreases , but it should not become inap-
FIG . 65. The relation of area of intellect to age in years 0 to 20 , assuming

Fig . 64 as correct , with a slight increase in the number of tasks with
increase in difficulty .
According to our results the de-

preciable until 18 or later .
crease from 14 to 18 is not an abrupt slowing up of a gain
that has been steady hitherto , but is part of a general nega-
tive acceleration which began long before the age of 61 .
AREA
The form of the curve for area of intellect in relation to
age is not known even approximately for CAVD or any
other specified intellect , since the number of tasks at each
altitude is not known . The arguments presented in Chap-
ter XII make it highly improbable that the curve for area
is like Fig . 64 with a rapid rise at the lowest ages and
decreasing annual increments thereafter . A very moderate
increase of the number of tasks with increasing difficulty is
sufficient to make the area curve from 0 to 16 , one with in

creasing annual increments as in Fig. 65 .
GENERAL CONSIDERATIONS
The limitations of the special tasks used in CAVD , Na

tional , Otis , and the like , should be kept in mind in all
thought concerning the relation of either altitude or area
of intellect to age . The verbal and mathematical tasks
which bulk so largely in these examinations may be more
like those which occupy the intellects of children from five
to fifteen than those which occupy the intellects of young
people from fifteen to twenty - five , or those which occupy
the intellects of men and women from twenty - five to thirty
five . It is conceivable and probable that the person who
ceases to improve in altitude CAVD may continue to im
prove in altitude Bu , Ch , Ho , So. ( Business , Child Manage
ment , Household Management , and Social Arrangement ) .
It
is also the case that after a person acquires a certain
amount of general linguistic and mathematical ability , and
of general information about the sort of things which every
body is ashamed not to know, he usually devotes his mental
abilities to the specialized abilities useful in his trade , busi
ness , or profession , hobby , and social circle . The correla
tion between CAVD ( or any similar ability ) and such spe
cialized abilities is doubtless high , but it is probably not
perfect ; and these specialized abilities may begin their
rapid rise in altitude at an age when CAVD altitude has
almost ceased to gain .
CHAPTER XVIII
SUMMARY OF RESULTS AND APPLICATIONS TO THE MEASURE-

MENT OF HUMAN ABILITIES IN GENERAL
SUMMARY OF RESULTS
All the measurements of intellect which have been made
hitherto and which psychologists may expect to make in
the future , unless means are found of defining and counting
units of connection in the neurones , are inventories . They
are records of the degree of success in accomplishing intel-
lectual tasks . If all intellectual tasks are listed and ar-
ranged in levels of difficulty , the inventory may be syste-
matized into a record of how many the intellect in question
can do at each level and how quickly he can do them . From
the record of how many it can do at each level , three useful
measures may be abstracted . One is altitude , that is , the
degree of difficulty at which a given percentage of success
is attained . The second is width , that is , the percent of suc-
cesses at any given altitude or the average percent of suc-
cesses at any given series of altitudes . The third , which
may be called area , is the total number of tasks done cor-
rectly , or the percentage which this total is of the number
of tasks in the entire list .
An intellectual task is one , success in which depends
upon all of intellect and nothing but intellect . Intellect is
definable by a series of tasks , and we have so defined one
variety of it , Intellect CAVD , and could so define any other
variety of it . A CAVD intellectual task is , then , one suc-
cess at which depends upon all of Intellect CAVD and
nothing but Intellect CAVD . Tasks can be devised which
do substantially meet this requirement , success at one of
them correlating perfectly ( or as closely as its own self-
correlation permits ) with success in the entire series .
CAVD intellect is nearly or quite homogeneous in the sense
469
"
that the ability which determines success at any one level

of difficulty is , to a close approximation , simply a larger or
smaller amount of the same ability that determines success
at any other level of difficulty . The evidence for this is the
very small and uncertain reduction of correlation as more
and more remote levels are taken , and the substantially per-
fect correlation between score at any one level and the sum
of the scores at all levels , or any other score representing
in any reasonable way ability at the entire series .
No short single task, however , can be measured in re-
spect of its intellectual difficulty, for no short single task
can be devised which depends for its success upon all of in-
tellect and nothing but intellect . Even composite tasks
made up of forty or more single tasks well selected to repre-
sent Intellect CAVD ( or any other specified sort of intel-
lect ) are not perfectly intellectual and must be treated as
measuring intellect plus an error for which allowance must
be made in all inferences from measures of their difficulty
to measures of their intellectual difficulty . With short
single tasks , this error becomes so large and so variable
amongst different tasks that no trustworthy allowance can
be made for it .
The difficulty of a task and the difference in difficulty
between one task and another may be measured by the per-
centages of certain groups which succeed with it . For it has
been demonstrated that the form of distribution of a school
grade population from 6 to 13 in respect of altitude of in-
tellect is to a very close approximation that of the normal
1 X2
probability surface defined by y = -- e σ2

The diffi-
σ2π
,
culty of an intellectual task may also be measured by the

percentage of successes among the various trials of the
same individual , or by the average of such percentages
from any given number of individuals of the same general
degree of intellect , since it has been demonstrated that the
varying conditions of an individual from time to time
( omitting such extreme conditions of sleep , illness , intoxi-
SUMMARY AND APPLICATIONS 471
cation , and the like , as would obviously unfit him for being
tested ) are distributed in close approximation to the nor-
mal probability surface . We have not made use of this
method because it demands , in practical use , a large num-
ber of composite intellectual tasks of equal difficulty , and
these are not yet available .
The 40 - composite tasks A , B , C , etc. , were constructed
on the basis of measurements of the difficulty of some thou-
sands of single tasks and some hundreds of 10 - composites ;
and were measured in respect of their intellectual difficulty .
The result is a series of tasks ranging from A , at which 88
percent of adult imbeciles of mental age 2 to 5 years suc-
23
В 261
с 281
฀ 291
301
311
32
321
341
Ι
361
J ABCDEFGHIKLMNOP
371
381
391
40
411
411
& 43
฀
which only 23 percent college graduates

at
ceed
to
of
Q
,
,
succeed The differences difficulty are determined ap-

in
.
proximately They can be determined any given degree

to
.
of precision by the methods outlined

.
The distance from the difficulty of Task to the diffi-

A
culty of Tasks 36 and 37 which are or near the absolute

at
,
¹
zero of intellect has been determined by consensus of ex-

a
,
perts as about 4.35 times the difference between Task and

A
Task which 5.2809 So we have an approximate scale

is
C
,
See Chapter page 339

X
1
32
of intellectual difficulty from an absolute zero in equal units

as shown below . This scale is at all points more accurate
than the best scales previously available ; and
is accurate enough for many scientific and
I
practical uses from to Q, covering the inter
val from the upper extreme of the feeble
minded to the 98 or 99 percentile adult intel
lect . It should , however , be improved by
more extensive experimentation .
If alternative CAVD composites are con
structed to be like A , B , C , etc. , respectively ,
in difficulty , but different from them in con
tent , until all possible CAVD tasks are used ,
it will be found that the number is not the
same at each level . In general , the easier the
level , the smaller the number of CAVD tasks
that can be made without using the same
single task twice .
Ifthe number of tasks possible at level 1
is a , the number possible at levels 6 to 10
ab
to 5
is , the number possible at levels 11 to
15 is a + b + c , and so on . If
the tasks of our
series are represented by the column of Fig .
66 , the total number which could be con
structed and which are exemplified or sam
pled by A , B , C , etc. , would be represented by
a figure with a top very much wider than its
bottom , as in Fig . 67 , or Fig . 68 , or Fig . 69 .
When the exact shape of this surface of fre
quency of intellectual tasks acording to diffi
culty determined the width and area of in
66.
FIG . The
is
pattern of area
of intellect in tellect can be measured The measurement
.
Na
sampling of of an intellect should then be measurement

a
tasks at each
of its success or failure with each of series
,
level of diffi
a
culty
of composite tasks each which depends for
of
.
success substantially on all intellect and nothing but in

of
tellect If the number of these tasks in the scale used to test

.
the intellect in question is proportional at each level to the

number of such tasks that the world offers , the width of the
intellect in question at each level and its area are given in
the record . If the
number of tasks in the scale is not pro-
portional to the total existing number in this way, the num-
bers of the record at each level must be multiplied by suit-
able factors to obtain widths and area . In either case the
total record for a person whence altitude , widths and area
FIG . 67 .
are derived may be with high probability inferred from

the scores in three 40 - composite tasks where the percent of
single tasks right is near fifty .
Such a measurement is fundamentally right , and im-
proved varieties of it are all that can be expected on the
level of external behavior with voice , paper , and pencil , and
the like . The physiological facts in the neurones which pro-
duce and , in the deeper sense , are intellect , are not known .
When they become known , they may or may not be amen-
able to observation and measurement .
As a result of the high correlations found between mere
associations or connections and the so - called " higher "
FIG . 68 .
FIG . 69 .
FIGS . 67 , 68 and 69. Samples of probable relations of number of tasks to
level of difficulty .
processes of abstraction , generalization , organization and

control , inference and reasoning , we have advanced the
hypothesis that the original basis for altitude , width , and
area of intellect is the mere number of possible connections
in the neurones whose connections correspond to having
and using ideas . If this hypothesis is verified , it may be
possible sometime to discover means of counting the num-
ber of possible connections in distinction from the number
which actually function , and so of distinguishing original
intellectual capacity from acquired intellectual ability . So
far as tasks for external behavior are concerned , there are
none which measure original capacity for intellect uninflu-
enced by training . One can only measure intellect , and then
make such allowances for advantageous and for disadvan-
tageous training as are shown to be reasonable .
Attaining any specified score in a stock intelligence ex-
amination such as the Stanford Binet or Army Alpha or
National may best be considered as succeeding with a cer-
tain more or less intellectual task . That is , to attain 190 in
Army Alpha under standard conditions is to succeed with a
certain task ; to attain 185 is to succeed with a certain
easier task , and so on . With this interpretation of the
scores , a scale in equal units can be worked out for any
such test over most of its range by suitable experimenta-
tion ; and this was done for many of these examinations .
Using equal - unit scales , the form of distribution of in-
tellect in a group of the same age was determined for ages
10 , 11 , 12 , 13 , and 14 , and by inference for younger ages .
The results of the determinations for adults were incon-
clusive , those from Army Alpha and those from Army Ex-
amination a being very different .
The change in altitude of intellect with age is obviously
characterized by negative acceleration , the curve being
roughly as shown in Fig . 64. The change in area of intel-
lect , however , will show a very different course . Even with
an extremely conservative estimate of the increase in the
number of intellectual tasks as difficulty is increased , the
additions to area of Intellect CAVD are larger from 4 to 8

than from 0 to 4 , and larger from 8 to 12 than from 4 to 8,
in the case of children with ordinary school facilities .
APPLICATIONS TO THE MEASUREMENT OF HUMAN ABILITIES IN

GENERAL
The principles and techniques which we have developed

for the measurement of intellect are , with minor modifica
tions , suited to the measurement of a very large proportion
of human abilities .
Consider the following list , which is representative of
traits which either have been measured in the last twenty
years or have been suggested as traits which, for theoreti
cal or practical reasons , it would be desirable to measure ,
and includes sensory , sensori -motor , academic , moral ,
social , and economic abilities :
Hearing .
Discrimination of pitch .
General motor skill .
Ability in assembling mechanisms .
Ability in drawing .
Ability in the written use of the vernacular .
Ability in spelling .
Honesty about money .
Honesty about cheating .
Popularity .
Ability in carpentry .
Ability in salesmanship .
A consideration of each case will show that with occa

sional exceptions the following principles and techniques
are applicable :
1. What is
measured is a product produced , a task
achieved . This may seem somewhat far - fetched in the case
of hearing , honesty , and popularity . In hearing , the pri
mary product produced is within the nervous system , being
evidenced by the person's awareness of sound , but from
the point of view of measurement this is evidenced by the

" "
correctness of his Yes and " No " products as the trials
are made with the sound and with no sound . In honesty ,
the products produced include acts of not doing as well as
acts of doing certain things . In popularity the primary
products are in the nervous systems of other people , but
these are measured by smiles , votes , loans , companionship ,
and favors of various sorts .
2. The measurement of any of the products produced
involves valuation . In the case of hearing , discrimination
of pitch, and honesty about money and about cheating , the
process of valuation is included in the definition of the abil
ity . For hearing , to hear is obviously better than not to
hear . For discrimination of pitch, knowing which of two
tones is higher is obviously better than not knowing . For
honesty about money , not stealing is obviously better than
stealing The question of valuation does exist , but it has
.
been settled by the statement of the kind of product to be

produced .
3. In measuring
intellect , we favored the arrangement
of tasks so that the score could be success or failure , though
we carefully left room for a scale of credits for various de
"
grees of goodness " in the accomplishment of an intellec
tual task . In the cases of drawing , written composition ,
ability in carpentry , and to a less extent in some of the
other abilities , the arrangement of many tasks each for a
two - compartment score may not be so effective as the ar
rangement of fewer tasks each for a score graduated in
perhaps fifty or more compartments . How to make full
use of such graduated scores and still obtain intelligible
measures of difficulty , range and speed then becomes a
problem . We may best defer our answer until we have
considered the next two principles .
4. The measurement of the ability is in essence an in
-
ventory . We can satisfactorily define the ability only by a
list of the products which it produces the tasks which it
achieves . We measure it only by measuring a sample
which represents this total series . Even the case of hear-

ing is a case of an inventory , though a relatively simple one .
A person's hearing is not equally good at all pitches , and
there may be other complexities . Discrimination of pitch
may vary with the intensity and timbre of the tones .
5. The tasks in such an inventory ( a ) vary in difficulty²
and (b) may be usefully considered as varying in difficulty

for hearing (or for discrimination of pitch, or for whatever
the ability is ) as a totality , or for such parts of the ability
as the task requires , with more or less intermixture of diffi-
culty due to other factors than hearing ( or whatever the
ability is ) . That is , we may usefully entertain the abstract
conceptions of difficulty for all of hearing ability and noth-
ing but hearing ability , difficulty for all of ability to dis-
criminate pitches and nothing but that ability , difficulty for
all of motor skill and nothing but motor skill , etc.
This should be unquestioned for hearing , discrimina-
tion , motor skill , assembling , and spelling . But in the case
of the other traits in our list , it is not so clear that the tasks
form a graded series in difficulty , and it is much less clear
that it will be useful to apply the conception of difficulty
for all of a certain ability and nothing but it , and of alti-
tude as a feature of the ability which is measured by the
degree of its kind of difficulty at which it can succeed .
For example , we do not often think of increases in abil-
ity in English composition as the achievement of harder
and harder tasks , but rather as the production of better
and better products . This procedure can , however , be put
into conformity with the general plan which we have
adopted for intellect . All that is needed is to define a
certain degree of difficulty as the difficulty of producing a
product of a certain excellence . Thus when a person
writes a composition on " Fishing " which is scored as 64
on the Hillegas scale , we may regard him as having at-
tempted the following tasks with the stated results .
2 Difficulty being defined for some specified group by the percentage of
failures at the task in question .
To write a composition on " Fishing " of quality

40 or better S
66
45 " 66
S
66
50 " 66
S
66
55 " 66
S
66
60 " 66
S
66
65 " 66
F
66
70 " 66
F
66
75
66 66
F
This is a part of the answer to the problem raised in

connection with the third principle , providing a method of
utilizing graded - credit scores and still retaining the ad-
vantages of the " success - failure " scoring . It could have
been utilized in our completion tasks , scored as 3 , 2 , 1 , or 0 ,
by calling a 3 a success at one level and a 2 a success at a
lower level . In using it we should bear in mind that
graded - credit scores for one task usually have a consider-
able subjective element and may need some special pro-
visions to eliminate the errors thereof . We should also
bear in mind that nothing is added to the reliability or
weight of a determination by re - stating it as a series of
successes and failures . To replace " scored 64 by a com-
position , " by " scored 40S , 50S , 60S , 70F , 80F , 90F by a
composition , " or to replace the latter by " scored 30S , 35S ,
40S , 50S , 55S , 60S , 65F , 70F , 75F , 80F , 85F , 90F by a com-
position , " gives no added reliability or weight , so long as
it is the same composition scored by the same persons .
We do not have to abandon measurement by the qual-
ity of the product measured by a gradation of credits , if
we institute measurement by degree of difficulty mastered .
The latter may be added without displacing the former .
In the case of drawing , the latter is already in good use .
We not only rate products on scales of general and special
3 This is so unless they are made up by putting together credit points for
specified objective features in the product . In that case they are better treated
as summations of S's in separate tasks , to wit , the production of those objec-
tive features .
merit , but also set series of tasks graded in difficulty . In

the case of honesty about money and about cheating , the
concept of difficulty is applicable . It is more difficult to
be honest in paying a railroad than to be honest in paying
a newsboy . Many more men will cheat the government
than will cheat their partners . In the case of popularity ,
the concept is applicable . Indeed , when we say that John
is much more popular than James , we usually mean pre-
cisely a combination of " altitude " and " width , " that , for
example , John is more liked than James by the people who
like both , and that John is liked by more people than
James is at any degree of liking that we may choose to
take . In such a case the degrees of difficulty are furnished
by the differences in persons and likings , it being " harder "
to win much liking than little from the same person , and
to win the liking of the least friendly people than that of
the most friendly . Carpentry and salesmanship are much
like drawing and the written use of the vernacular in re-
spect of the applicability of the concept of difficulty . We
can measure variations in the ability either by a graded
scale of credits for the quality of the product , or by a series
of tasks graded in difficulty ; and the second method can
be operated with no loss to the first .

Whether in any case it is desirable to operate measure-
ment in terms of the point on a scale consisting of tasks
graded for difficulty , where 50 % S ( or some other assigned
percentage of successes ) is reached , depends in large mea-
sure upon the usefulness of the attempt to abstract out
difficulty for, say , ability in carpentry in its entirety and
untainted by anything other than ability in carpentry , and
measure in terms of it .
In the past, we have not tried to distinguish at all pre-
cisely the ability in carpentry from other abilities . a If
thousand boys made chairs , we have used the quality of
the chairs which they made as a measure of their several
abilities in carpentry , despite the fact that John may have
4 Or that certain people like John better than any people like James .
succeeded chiefly by general carefulness , and James may

have failed by inability to read the instructions , and Tom
may have won his standing largely by the artistic excel-
lence of some carving and painting which adorned his
chair , and a dozen others may have won their standings ,
each by a different compound of abilities . In the case of
intellect , it is well worth while to seek rigorous measures
of intellectual difficulty , because intellect is so important
an ability and because altitude of intellect turns out to be
a fairly unified , coherent variable properly represented by
cardinal numbers . But will this be so with the altitude of
ability in carpentry ? How many thousands of abilities
are we to assign specific difficulties and altitudes to ? A
girl makes a dress , and we easily thereby have some mea-
surement of her ability to make that sort of a dress . How
far shall we go in our use of this task of making a dress as
a measure of difficulty for ability in sewing , difficulty for
ability in design , difficulty for motor skill , difficulty for
executive ability ; difficulty for honesty in not obtaining
illegitimate help , and so on and on ?
The general answer is that it is desirable to settle on
the most fundamental and mutually exclusive abilities and
learn to measure them first ; that many of the thousands of
abilities with which the sciences and arts of man deal are
doubtless compounds or derivatives of more elemental
abilities that most of them are far from being mutually
;
exclusive ; that new professions and trades and topics for
study add to the number of these abilities without any alter-
ations in the fundamentals of human nature ; and that con-
sequently science may often refuse to measure the altitude
of, say , ability in carpentry in its entirety and uninfluenced
by anything save ability in carpentry . Ability in car-
pentry may be left to a looser definition and cruder mea-
surement . If, however , science does undertake to make a
true measure of ability in carpentry , it is possible to apply
the concept of specific difficulty for ability in carpentry .
For the most satisfactory measurement it is necessary to
do so . The same will hold of other mental abilities , with

few exceptions .
6. There are usually more tasks than one at each level
of difficulty , so that the range or width of the ability at any
level may be measured by the percentage correct at that

level, and , if desired , a measure of surface may be made by
summing the widths at all levels .
This is in a sense a corollary of principle 4. It needs
no comment .
7. Measurements of speed should be of the speed of
successful performances. Nothing useful concerning an
ability is measured by the time required to fail at a task.5
8. Width being measured by number and speed by time ,
the technical problem with any of these abilities is to pro
vide a scale for altitude , that is , for its specific difficulty .
Knowledge of the differences in difficulty ( for the ability
in question ) of a series of tasks , and of their differences
from an absolute zero of that difficulty , are the two desid
erata in such a scale .
9. Knowledge of the differences in difficulty (for the
ability in question ) may be had if (a ) the form of distribu
tion of the varying conditions of the ability in an individual
is known ; or if
( b ) the form of distribution of the varying
abilities of the individuals in a group is known . These
5 The relative amounts of importance of altitude , width , and speed in the
measurement of an ability vary . Consider , for example , drawing , spelling , and
typewriting , asking the three questions , " How difficult a task ? " , "
How many
tasks at easy levels ? " , and " How quickly ? " . In drawing , altitude is of
prime importance , since if a person can do the hard tasks , that is , can draw
well , he can usually do , or very quickly learn to do , nearly all of the easy
tasks , that is , draw identifiable cows , chickens , houses , trees , and the like . In
spelling , width is of especial importance , since the ability to spell very difficult
words is less valuable for the world's welfare or as a symptom of excellence ,
whereas failure with an easy word is an annoyance to many ; and since the
ability to do the hard task does not so often as in drawing presuppose ability
with ( or quick acquisition of ) the easy tasks . In typewriting , speed is ob
viously relatively more important than it is in drawing or spelling . This is
partly because the typist early learns to do all or nearly all of the tasks , and
thereafter improves his ability by learning to do them without lapses into
error , and more quickly .
may be ascertained with a high degree of probability by

submitting the individual or the group to many graded
series of tasks , each series being made with the intent to
have tasks spaced at equal intervals of difficulty and to
have as many tasks at any one level of difficulty as at any
other, by individuals or committees uninfluenced by any
preconceptions about the form of distribution . The score
to be given is a level score , like those derived and used in
Chapters IX , XI , and XIII . Whatever distribution is ap-
proximated by the average of these distributions , and more
closely when the scores for two or more are averaged than
when they are used singly , has a strong probability of being
near to the form of distribution which the altitude of that
ability would show if measured in truly equal units .
There may be other useful ways of estimating these
amounts of difference . In the case of the beauty of draw-
ings, a consensus of opinion may be preferable , since
amount of beauty may be precisely the amount of beauty
as felt . In the case of ability to discriminate pitch , it may
be permissible to forego equality in units of the ability
itself and be content with the mere definition of the levels
in terms of the physical differences ; or the progress of
research may justify some method of inference such as that
amounts of the ability are inversely proportional to the
time required to perceive the difference .
10. Altitude of intellect is distributed approximately in
Form A in the case of grade populations from Grade 6 to
Grade 13 , and of age populations from 11 to 14. There
is a substantial likelihood that any one of these grade popu-
lations will show a similar form of distribution in the case
of the altitude of academic abilities ( like ability in reading ,
or in the use of language , or in academic information )
which are bases for promotion and are rather closely corre-
lated with intellect . There is some likelihood that age popu-
lations will show a similar form of distribution for altitude
of general sensory acuity , or for general sensori - motor
capacity , or for general motor skill , or for general mental
health , since variations in these amongst children of like

age are probably in large measure due to the same sorts of
causes as are variations in intellect in children of like age .
11. An ability is defined by making a series of tasks
such that the score in this total series depends on the abil
ity. Thus an ability is defined by a total series of tasks .
This series may be constituted by assuming that such and
such tasks are as a whole a measure of the ability , or by as
suming that the ranking of individuals in an order for their
amounts of the ability by some defined consensus of experts
is valid , and selecting a series of tasks success in which
correlate perfectly with that order . This is in so far ar
bitrary . If the ability does correspond to some unified ,
coherent fundamental fact in the world , later work with
,
the series of tasks will , to some extent at least , reveal that

such is the case. The correlations of parts of the series
with other parts will be high ; their correlations with abili
ties outside the series will be lower ; the relations to hered
ity , to age , and to sex will be fairly simple . Some reason
able hypothesis concerning the physiological parallel of the
ability is likely to suggest itself . If
the ability does not so
correspond , but is just a name for some concatenation of
aspects which various fundamental abilities take under
rather artificial conditions , later work with the series of
tasks will probably show it .
12. The ability having been defined by a total series of
tasks ,it may be arranged in subseries or composite tasks
in an approximate order of difficulty by experiment or con
sensus or both .
13. Each composite task should be representative of the
total series qualitatively and should be large enough so that
rat ( letting a stand for the total ability ) is 1.00 , or as high
as the self- correlation of the composite permits .
14. Composite tasks will be efficient instruments for
measurement in proportion as their single elements are
equal in difficulty ( not necessarily in specific difficulty for
the ability in question ) and give a high multiple correlation
with ability a.
15. These composite tasks should be measured in re

spect of their differences in difficulty ( for ability a ) one
from another. Even a very inadequate measurement is
better than none , so long as its inadequacy is nowise con
cealed . For we have everything that we had before the
measurement was made , undisturbed by it . For example ,
suppose that in motor skill , composite tasks a , b , c , d , e
· h are devised and tested with populations at age 2 ,
4 6, 8 , 10 , 12 , 14 , 16 , and 18. Even if we do not know the

,
forms of distribution of altitude of motor skill in these age

groups at all accurately , it is better to make a reasonable
hypothesis about them and act on it than to remain con
tent with the mere rank order for motor difficulty , that
a < b < c < d , etc. Whenever anyone does anything with that
rank order in the way of using it in the comparison of any
difference or the expression of any relation , he makes some
assumption about the amounts of difference b-a , c-b , d - c ,
etc. There is no workable arithmetic of pure ranks , and
cannot be . So long as it is unknown whether c b is equal to -
b-a, or ten times b-a or one - tenth of b-a , and the like , we
cannot add , subtract , multiply , or divide with rank differ
ences as such .
16. These composite tasks should also be measured in
respect of their differences from an approximate absolute
zero of ability a. Even though the location of zero is hypo
thetical and has a large margin of possible error , the esti
mate can do no harm if its unreliability is not concealed .
Nothing is distorted by it . It is simply an added feature
to be used or neglected as seems wise . Anybody who
6 Thus the Spearman correlation formula
1 62g
T
r=2 coss
3 (1 - R ) - 1 where R= -n² 1
assumes that the form of distribution of the two traits is Form A , that of the
normal probability surface .
Spearman's formula , r = 2 sin )89 (
where =
3 1 n (n² - 1)
62D2
assumes that the form of distribution of the two traits is Form A. If the
form is rectangular , r is taken as equal to ? .
wishes to utilize the " times as much " comparison with

the facts will have to make some assumption about the loca
tion of zero ; and the assumption made by the author of the
scale is likely to be better than the average of the assump
tions of those who use it . He has presumably more knowl
edge and better judgment about it than the average of them .
No one of them is compelled to accept the location which
he assigns .
Aqualification of " usually " might have been inserted .
in the paragraph above , because there may be important
abilities where any location of the absolute zero is so diffi
cult that any attempt to do so with present knowledge is
pretentious and absurd . We doubt this , however . On the
contrary , when an ability has been defined by a series of
tasks , the extension down from the easiest composite task
to a task of the same general sort which requires just a
bare trifle of the ability in question is almost a necessary
consequence . If a reasonable location of the absolute zero
cannot be attained , it is a sign that the ability itself is not
an important , unified variable in nature .
There are two important special cases of locating abso
lute zero which may suitably be considered here . The first
is that exemplified by discrimination of pitch , where it is
customary to measure down from an upper limit of perfec
tion , using the amount of error made , or the smallest
amount of difference discernible , and thinking of its dif
ference down from zero error rather than up from zero
ability . This has its advantages , and it need not be alto
gether discarded ; but it produces numbers which are very
adapted for quantitative thinking and may well be sup
ill
plemented by measures up from zero meaning just barely

a
not any ability

.
The second the case where the standard point of refer

is
the ordinary modal man Thus people com

or
ence
is
,
.
monly measure popularity up and unpopularity down from

condition average popularity This procedure has
of
a
been widely recommended recent psychological and edu

in
cational science , and has the very great merits that its
point of reference is one near where the measurements are
made , is easily definable in reality , and is convenient for
+ and -
studies by correlation and partial correlations . Measures
from such a central tendency in units of the varia
bility of the group are also honest , making no pretence at
being more than they are . It has the disadvantage of in
stability , shifting with the group taken . Again science
a dog
FIG . 70.
FIG . 72 .
W глу
FIG . 71 FIG . 73 .
FIG . 74 FIG . 75 .
may retain all these advantages , and still add the different
advantages of a true zero .
Finally , attention may be invited to the difference be
tween natural and conventional zero points . In the former ,
there is a genuine beginning of an important natural phe
nomenon . In the latter , there is a beginning only from the
point of view of some human institution or custom . Thus
motor skill in the case of the hand may be said to have its
task of zero difficulty somewhere below the point of reach
ing toward an object and touching it , or of grasping an
object touched and bringing it somewhere near the mouth .
This is near a natural zero . The tasks of zero difficulty
33
for ability in penmanship and of zero difficulty for ability

in drawing would by any reasonable view be tasks which
would be much above this zero of manual motor skill . The
zero task for penmanship would be to make something
which could be identified as writing , though no letter or
word could be identified . For example , to copy Fig . 70
with the result shown in Figure 71 may be regarded as
near the zero of handwriting . To use pencil or crayon to
make a line as " good " as Figure 73 when Fig . 72 is
shown as the model may be regarded as near the zero of
drawing . To use pencil or crayon to make any kind of a
scribble such as Fig . 74 or Fig . 75 may be regarded as zero
for the use of a mark- making tool . All three of these zeros
are well above the zero of manual motor skill . They come
later in life ' and are impossible for idiots for whom the
reaching and grasping are possible .
The number of conventional zeros is legion . We can
have tasks of zero difficulty for ability in typewriting , in
stringing beads , in shuffling cards , in playing the piano , in
playing the violin , in tapping , and scores of other varieties
of manual skills ; and for certain educational and economic
problems each of these zeros may be of value and the refer
ence of measures of ability to it may be desirable .
They will be even more useful if each of them is mea
sured off from the natural zero for manual motor skill , so
that reference can be made to it also . This holds as a gen
eral principle . Any conventional zero will be made more
useful by being itself measured off from some natural zero .
There are two important fields of mental measurement
where the principles and techniques which we have devel
oped for intellect do not apply , at least not without radical
changes .
The first is the general case of the description and mea
surement of the connections or bonds whereby any given
7 The dates may be set roughly asyr . for reaching and grasping , 1 year
for making a mark , and 2 years for copying a line . The date for making
anything enough distinguishable from general scribbling and mark - making to
be called a zero specific to handwriting ability is not known , but is later than 2
years .
condition or state of affairs in a mind leads to another con

dition or state of affairs in that mind . In such cases we
need not and do not use valuation or the concept of diffi
culty , but simply record that such and such connections exist
in such and such degrees of strength from zero up . By
the strength of a connection between condition A and con
dition B , we mean the probability that A will be followed
by B. The strength of the connection is measured crudely
by the percent which A →B is of A→ B + A → C , + A →D ,
• + A → N , and by the length of time that the connec
tion will last without exercise ; and somewhat more exactly

by this percent and this time in relation to the number and
strength of competing tendencies . This is by far the com
monest sort of mental measurement .
The second is the measurement of likes and dislikes , in
terests , desires , " drives , " motives , or whatever one chooses
to call the facts whereby certain states of affairs are satis
fying and others are annoying to the animal . The whole
subject of satisfiers and annoyers is in more or less doubt
and dispute , but by any view the measurement of how
much a man likes the taste of olives or a brisk run up a
hill is very different from the measurement of either his
intellect or his tendency to contract the pupil in a bright
light .
The most striking difference is that the measurement
is now up and down from a zero of indifference to A , where
the animal neither seems moved to do anything which will
change the state of affairs in respect of A , nor seems moved
to do anything which will cause A to continue or recur .
Another notable fact is that whereas it is easy to dis
cover whether A or B is more satisfying to any given in
dividual X , it is hard to discover whether A is more satis
fying to individual X than to individual Y.
According to our notions , the satisfyingness or annoy
ingness of any state of affairs to any mind is measurable
8 The cases where A occurs and apparently leads to nothing are to be in
Icluded in the count .
by the products produced , including the internal neural

products which lead to honest reports of satisfaction and
annoyance , welcome and rejection , " for " and " against . "
We also think that different varieties of satisfyingness
such as that of food when hungry and hearing music and
being looked at with respectful glances , are intercommen-
surable ; and also susceptible of algebraic computation
against amounts of annoyingness . The discussion of any
theories of such an hedonic calculus is , however , beyond
the scope of this chapter or this book .
APPENDIX I.
THE FORM OF DISTRIBUTION OF AN INDIVIDUAL'S

VARIATIONS IN INTELLECT
We have elsewhere¹ reported evidence showing that the

variations of an individual in separate tests of an ability ,
each a half hour or so in length , were distributed almost or
quite symmetrically about his average or mode , extreme
deviations upward being almost or quite as common as ex-
treme deviations downward . It is the purpose of this ap-
pendix to present , in a very brief and summary manner ,
additional evidence to the same effect .
We have confirmed the results in the case of the Army
Alpha by using the records of 81 high school students in
Milan , Michigan , each with three trials of Alpha , a year
apart , given by Superintendent Tape . ( Fall of 1919 , 1920 ,
1921. ) Fifteen students were at those dates eighth grade ,
freshmen , and sophomores , respectively ; thirty - two were
freshmen , sophomores , and juniors ; and thirty - four were
sophomores , juniors and seniors . The allowances for prac-
tice effect and growth were : first group , 15 and 13 points ;
second group , 16 and 12 points ; third group , 13 and 13
points . ( The average scores were : first group , 91 , 105 ,
and 121 ; second group , 87 , 103 , and 115 ; third group , 104 ,
117 , and 130. ) The median of the scores thus corrected
was found for each individual , and the two deviations from
this median . The distribution of these plus -minus devia-
tions is shown below . The 81 medians themselves are not
included .
The average deviation upward is 8.60 points . The aver-

age deviation downward is 8.54 points . There are 23 plus
deviations and 24 minus deviations which are 10 points or
greater .
1 Jour . of Exp . Psy ., vol . 6 , pp . 161–167 .
491
Frequencies Total Grouping

Deviations 1st Group 2nd Group 3rd Group Total by 9 Points
- 32 , 33 , 34 1 1 1
-29 , 30 , 31 1 1
- 26 , 27 , 28 2 2 3
- 23 , 24 , 25
-20 21 22 3
3
,
,
121
17 18 19 3
1
- - -
,
,
14 15 16
4 2
6
,
,
11 12 13
7
,
,
5 5 5 7 6 9 24377
10 16 40
2
3, 0, 3, 6, 9,
5, 2, 1, 2, 5, 8,
― 5 17
4 7
-
18
25615
16 53
4 1753979
+1
19
7 4
+ + + +
17
1 1 1 3 3
6,
10 12 35
9,
8,
11 12 13
1 2 1 1 1 4 8 6
,
,
1 10
+14 15 16
5
,
,
17 18 19 13
+ + + + + +
221
,
,
20 21 22
,
,
23 24 25
1
T
,
,
26 27 28
4
1
,
,
29 30 31
1
1
,
,
32 33 34
1
1
,
,
39 pupils Grades and took the Courtis tests

in
in
6
5
computation four times during the year The effect of

.
practice was allowed for by adding the first score the

to
average gain of trial over trial and subtracting from

2
1,
the third score the average gain of trial over trial

3
2
.
Trial was treated similarly The deviations each in-

of
4
dividual's scores from his median score were then com-

puted As the result we have
,
:
.
Deviations Frequencies
T
to 10.5
-
-
2 1
-2 -5 -8
66 7.5
110 4.5 19
-1.5 1.5 107
+ + +
4.5 20
+
2
+5 7.5
0 4
+8 +10.5
' "
INDIVIDUAL VARIABILITY IN INTELLECT 493
37 pupils in Grades 5 and 6 took the Woody Test , Series

A, three times in Oct. ( or Nov. ) , Jan. , and May of a
school year . The scores were corrected for the practice
effect by adding the average gain of trial 2 over trial 1 to
the score for trial 1 , and subtracting the average gain of
trial 3 over trial 2 from the score for trial 3. The devia-
tions of each individual's scores from his median score
were then computed . As the result , we have , including
the 37 deviations of 0 belonging to the median measures :
- 23 to - 31 1
-14-22 3
5 " -13 16
- 4 + 4 83
+ 5 + 13 8
+14 " +22 2
+23 " +31 0
In records in tests of spelling treated in a similar man-

ner ,we have , as the distribution of an individual's devia-
tions from his median , the following :
- 32 to - 40
-23 " - 31
0
2
- 26 " -22
-
5
" -13
-
5 24
466 + 4 53
+5 " +13 30
+ 19 + 22 4
+ 23 66 +31 1
+ 32 + 40 1
Pupils in Grades 4, 5 , and 6 were tested with Stanford

Binet , National A , National B , Otis Advanced , 2 Myers Men-
tal Measure , Haggerty Delta 2 , Illinois , and certain parts
of Dearborn . Each score was first turned into a devia-
tion from the median for the group in that test , in terms
of the variability of the group in the test in question . Then
2 Some had the Otis Primary instead . For these , estimated scores in the
Otis Advanced were computed .
it was expressed as a deviation from the average of the

eight such deviation - scores for the individual in question .
These last deviations represent the variability of an in-
dividual around his average ability in intelligence tests .
Their distribution is as follows :
- 165 to -- 194 0
- 135-164 2
-105-134 1
- " -104
-
9
- 45 "
75
74 40
15 "- 124
-
44
14 + 14 137
+ 15 "" + 44 86
+ 45+ 74 41
+ 75 " +104 10
+ 105 "" + 134 3
+ 135 + 164 2
"" 1
+165 + 194
65 pupils in Grades 8 to 12 were tested with Alpha Form

5, Alpha Form 8 , Terman Group Test Form A , Terman
Group Test Form B , and half of Part I of the Thorndike
Examination for High School Graduates . These five scores
for each pupil were treated just as the eight scores de-
scribed in the previous paragraph , except that the final
deviations are deviations from the individual's median in-
stead of from his average .
The resulting distribution was as follows ( including the
65 zero deviations of the medians themselves ):
- 110 to 129 1
- 90 66 - 109 2
- 70 " - 89 10
- 50 ""
-
69 10
- 30 " 49 34
- 10 " 29 47
9 66 + 9 117
+ 10 "" + 29 55
"" + 49 27
+ 30
+ 50 "
+ 69 15
+ 70 66 + 89 5
+ 90 +109 2
+110 " +129 0
INDIVIDUAL VARIABILITY IN INTELLECT 495
The variations in I. Q.'s from different Stanford Binet

examinations at long intervals are not exactly the sort of
evidence that we wish , since variability is complicated with
growth and the measure is an indirect ratio , but it is of
interest to note the facts . So we have tabulated the devia
tions of an individual's separate I. Q.'s from his median
I. Q. in all of the cases of Baldwin and Stecher ( '22 , p . 24 )
who had at least three separate examinations . The results
are as follows :
-26 to - 291 1
- 23 " -26
- 19 ( -221 13
-16 " " - 19 5
- 12 66 - 15 6
- " -12
- 81
9 19
- 5 28
2 "" 5 53
1 11 " + 11 149
"" +
+ 2 5 60
+5 " + 81 27
""
+ 9 +12 14
+ 12 + 15 5
+16 ""
+19 1
""
+191 +22 2
""
+ 23 +26
+ 26 + 29 10
In no one of these six cases is there more than a very

slight excess of extreme downward deviations . On the
whole , the balance is about even as shown by the two
rough summaries below .
We may expect , in a thousand deviations , about 12 that
are 2340 or more minus to about 7 that are 2340 or more
plus ,if we assume that the scoring units in the cases studied
are on the average of about the same real value at the lower
as at the upper extreme of the ranges studied . About the
real value of the units we know very little , but it seems
likely that the tests have more tasks at the lower levels of
difficulty than at the upper , rather than the reverse . So
getting equal units would probably reduce the excess of
Courtis Woody Spelling Total

low 0 0 0 0
1 1 2 4
2 3 5 10
19 16 24 59
central 107 83 53 243
20 8 30 58
4 + 10
1 4
0
1 1
200
high 0 ++
1
test test repeated I.Q. Total
+
+
8
low
3 0
1
2 0
9 1
++
14
50 54 53 157
central
+++
347 219 262 828

54 47 46 147
high
0 9
0 3
0 2
0 4
low deviations rather than increase Also the excess of
it
,
.
low deviations will be reduced we omit the somewhat

if
ambiguous data
Q.
I.
These facts with those of the previous report certainly

justify the conclusion that the real variations an in-
dividual in 30 minute test of intellect or school achieve- of
a
ment will found be approximately symmetrical This

be
to
justifies the use of the average practice and clears the

in
way for important advances the theory mental mea-

in
of
surements and the scaling of tasks

.
APPENDIX II .
THE RELATION OF AN INDIVIDUAL'S VARIABILITY TO HIS

ABILITY IN TESTS OF INTELLIGENCE
An individual who takes a number of different trials in
a test , using alternative forms arranged to be of equal diffi
culty , attains varying scores . It is of importance for many
purposes to know how the variability of an individual in
any ability is related to the amount of the ability which he
has , that is , to his average or modal achievement . We
have investigated the matter somewhat thoroughly in the
case of seventeen tests of intelligence .
It is necessary to distinguish between ( a ) the apparent ,
or face - value , relation observed between the variability of
an individual's separate scores and his average score , and
( b ) the real relation that would be observed if these scores
were transmuted into measures such that 1 , 2 , 3 , 4 , 5 , 6 , 7 ,
etc. , represented a real arithmetical progression of amounts
of the ability . For example , we find that twenty individ
uals each of whom took ( after two preliminary trials , to
eliminate the practice effect ) from eleven to thirteen forms
of Part I of the Thorndike Intelligence Examination for
High School Graduates , showed the results of Table 138 .
If the scores are taken at their face value , it appears that
the variability of an individual whose median score is about
105 ( from 100 to 113 ) is very nearly the same as the varia
bility of an individual whose median score is about 128
( 125 to 132 ) . If, however , the units of the scoring from
90 to 120 really represent smaller increments of ability than
the units from 120 to 145 , the real variability of an in
dividual of ability 105 is less than the real variability of
an individual of ability 128. The converse is true , if the
units of the scoring scale from 90 to 120 really represent
larger increments of ability than the units from 120 to 145 .
497
498
.
TABLE 138
(
)
I
VARIATIONS OF THE SCORES OF THIRTEEN OR FEWER 30 MINUTE TRIALS WITH PART OF THE THORNDIKE INTELLIGENCE EXAMINA-
TION FOR HIGH SCHOOL GRADUATES FROM THE MEDIAN SCORE FOR THE INDIVIDUAL IN ALL THIRTEEN
.
,A
- ,
C
.
, D
,E
,
TRIALS 20 GIFTED PUPILS U. 13 DIFFERENT DAYS
i
d
j
1
Individual
r
.r
.o
u
to
S
,a ,kh
j
f,e, ,nq,
,g, ,t.
P
u m to
Median 87 95 96 99 99 87 to 99 100 to 113 125 125 128 128 132 125 to 132
11 -12 13
18
-
1
10
10
155
-
2 45
10
THE MEASUREMENT
7 4
11 20 11
1 2 00 5
22 12
18 32 13
2 2 5 2
9
6
8
Frequency
+ ++
1 7
3 1 1
of 10 20 11
1 4 304 1
1 2
1 0 4 3 1 1 1
++
14 1
2 21 3
2 3 2 1
Variations +10 12
212
+
52
++
21
82
3 1 3 1 1
11 +12 13
34
+14 +15 +16
+
17 +18 +19
+
+20 21 22
++
-, ,-9 ,6 ,3 ,0 4,3 ,6 ,9 , , , , ,
+
, ,8 ,5 ,2- ,5 ,8 , , , , ,
0 1
23 24 +25
n 63 120 64
OF INTELLIGENCE
Average
Variation 4.2 4.8 4.8
THE RELATION OF VARIABILITY TO ABILITY 499
The facts concerning the scores taken at their face value

are worth knowing and recording , but it is the relation of
the " real " variability to the " real " amount of ability that
is the essential problem . Our procedure and reasoning are
as follows :
We record the face - value score results for many differ
ent sorts of tests of the ability in question , and note in each
case any facts about the construction of the tests which
concern the probability that its units progressively swell or
shrink in " real " value over any considerable fraction of
the range we are concerned with . We note especially the
score results in those cases where there is no reason to
expect swelling more than shrinking . The average rela
tion between variability and ability found in these cases
may be taken to represent approximately the real relation ,
until someone produces evidence that , in all or nearly all
tests for the ability in question , there are forces leading
psychologists , quite without intention , to devise scoring
plans which make for progressive swelling or shrinking of
units at the same places along the scale of real ability .
The most desirable material for our purpose would be
the records of individuals representing a very wide range
of ability , there being many of them , and each being mea
sured with many alternative forms of the test in question .
Range is desirable to accentuate the relation and measure
it throughout . Large populations are desirable to reduce
the disturbing effect of individual differences . A large
number of trials of the test is desirable to locate exactly
the ability to which the variability is related . Unfortu
nately, there is no such material , and we have to do the
best we can by putting together several small ranges , and
by using individuals who have been measured by only two
trials of the test . Indeed , we use individuals who have
been measured by only one trial, by a method which will
be described shortly . Whatever the method , however , the
result is an estimate of what the relation would be between
( a ) the variability of an individual in an infinite number
of trials or tests or measures of the ability and (b ) the

amount of his ability as measured by the mode or other
central tendency of these separate trials or measures .
In general , any material is useful for our purpose by
which we can measure the variability of a number of trials
or individuals whose central tendency in the ability is low ,
and also the variability of a number of trials or individuals
whose central tendency in the ability is high . The essen-
tial difference is that in certain material the relation is
very much more liable to blurring and distortion by the
inaccuracy of the determinations of average ability itself
than it is in other material .
The effect of the inaccuracy of the averages upon our
determinations of the relation between average intellect
and variation in intellect is somewhat subtle and compli-
cated . We need not discuss it here . After we have gath-
ered and organized the best data we can bearing on the
relation , and found , as we shall , that they are best ex-
plained by the hypothesis that the variability of an in-
dividual neither increases nor decreases according as his
average is low or high , we shall have to ask whether the
case would be otherwise if the inaccuracies of these aver-
ages were reduced .
When we have only two separate trials with a test , it is

best to measure an individual's variability by the difference
between the two scores , after suitable correction for the
practice effect and for differences in the difficulty of the
forms used . ( Call this difference D. ) The average varia-
tion of an individual from his average or median ( call this
A.D. ) may be taken as this difference divided by √2. The
average of the A.D.'s of a group of individuals , all of the
same or nearly the same average ability , is the average of
their D's divided by V2 . Table 139 shows the results in
I
the case of a set of scores in the Thorndike Part , series
of 1919 to 1922 , so treated .
If we obtained the actual average for each individual
and computed his A.D. directly by the formula , Σ ( devia-
TABLE 139
..
IS
).
(
THE RELATION BETWEEN AN INDIVIDUAL'S ABILITY AND HIS VARIABILITY THE VARIABILITY THAT OF ONE TRIAL 30 MIN OF
2
,
,I
.
THE THORNDIKE TEST PART FROM THE AVERAGE OF AN INFINITE NUMBER OF SUCH TRIALS THE TRIALS
.
WERE TAKEN ON THE SAME DAY
THE RELATION
S.
in
Average score Th H.
,I
55
Part forms and M 45 50 55 60 65 70 75 105 110 115 120 125 130 135
OF
M
. I -I(
, )
Average differences
after allowance for prac
tice and for the difficulty
.
of the two tests 10.5 12.0 9.0 8.3 9.3 8.0 10.1 8.1 7.8 6.7 4.5 6.7 6.0 8.3
Average difference of an
individual's score in one
1
VARIABILITY
trial of form of the

test from his average
TO
in
. (
score the test Row
2
÷
).
V2 7.4 8.4 6.3 5.8 6.5 5.6 7.1 ... 5.7 5.5 4.7 3.2 4.7 4.2 5.8
6
3
4
Number of individuals 34 48 61 84 102 65
127
50 20 13
ABILITY
501
501
tions ) / 2 , we should have A.D.'s throughout seven -tenths as

large as those of Table 139. This would be because in each
case the central tendency had been automatically chosen so
as to vary from the true central tendency in the direction
of making the A. D. a minimum .
A still more convenient way at times is to estimate the
variability of an individual's separate trials around his
average indirectly from the variability of n individuals
( each measured by a single trial ) around the average of
the n individuals , using individuals all equal in ability to
the individual in question . If n individuals in one trial
with Army Alpha all score 165 , and in another trial score
168 , 160 , 171 , 159 , etc. , etc. , averaging 165 , it may be argued
that the deviations of the 168 , 160 , 171 , 159 , etc. from 165
are comparable and proportional to the deviations of one
individual of ability 165 who took Alpha n times .
3
More
exactly , the facts are that any one of n individuals who
are all really of ability K in a test does resemble any other
of them in variability ; and that if there is taken from each
of the n distributions representing their separate variabili-
ties , one variation at random , the composite of these se-
lected variations will resemble the distribution of the varia-
tions of any one of the n individuals . In the particular
matter with which we are concerned , the average of these
selected variations will vary only by chance from the aver-
age of the averages of the separate individuals ' variations ,
and this will vary only by chance from any one of these
averages . The matter of concern to us is that , when n in-
dividuals are measured each by only one trial and all score
K , they really are not all of ability K , but vary around it .
Consequently , when n individuals measured each by one
trial are used in place of one individual measured by n
trials , the “ attenuation " or blurring of the relation is much
greater . Using one individual measured by n trials relates
the variability to one precisely located amount of ability .
3 Assuming of
course that in both cases the effects of practice , interest ,
and the like are properly allowed for .
Using one trial of each of n similar individuals relates it to

an amount which is located with a considerable error . If
we keep this attenuation in mind and allow for it , we may
use material of this sort when it is desirable .
Even if we have only one trial with a test , it is still
possible to estimate the approximate average of the aver-
age variations of a group of individuals each from his
average , provided the individuals are also measured by
some test which correlates fairly closely with the test in
question . For example , suppose that ten individuals all
score 75 in the score for one trial of the entire Thorndike
Intelligence Examination for High School graduates , and
score 30 , 32 , 34 , 35 , 36 , 37 , 38 , 40 , 42 , and 48 in Part III
of it ; and that the general correlation between a one trial
score for the entire examination and average ability in a
large number of trials of Part III is .90 ; and that the corre-
lation between the average score in two trials of Part III
and the average ability in a large number of trials of Part
III is .90 for a similar group . Then we do as well to take
the variations in Part III of the ten individuals who all
scored 75 in the entire examination as we would do to take
the variations of ten individuals who all scored alike in the
average of two trials with Part IIIitself .
We shall make very large use of this indirect method
because by its use we can cover wide ranges and have large
populations , and also in certain instances because the ma-
terial available can be treated by no better method .
It should be noted that when the two trials of a test by
an individual occur in a single session , the variability mea-
sured is that due to the difference of one hour from an ad-
jacent hour , plus , of course , the differences of the content
of the two forms of the test that were used . Cases of
single - session variability will be so designated wherever
they appear in what follows . Except for them , the varia-
bility studied is that of a random sampling of different days
of an individual (but omitting days when by sickness or
other causes he would not be submitted to a test ) , with a
34
random sampling of alternative forms of the test in ques-

tion .
Let us first illustrate certain further matters concern-
ing method . Table already shown the method
138 has
where we have several trials with the test by an individual
on different days . Table 139 has shown the method where
we have two trials by an individual . Table 140 below
shows the method where we use the array of scores in trial
2 of n individuals scoring alike in trial 1 to represent ap-
proximately the n different trials of one individual . The
test is still Part I
of the Thorndike Examination , as in
Tables 138 and 139 .
It will be noted that in computing the variability of an
array we use as à central tendency , not the actually ob-
served average or median of that array , but the central ten-
dency which seems most probably the true one for it in
view of the facts of the entire table , that is , the central
tendency given by the total relation line or regression . If
the actually observed central tendencies were used , there
would result a fallacious diminution of the variability in
all arrays , and this would sometimes be a very large error
when the number of cases in the array was small . The re-
lation line is determined by plotting the observed medians
and drawing a smooth line such as approximately ( 1 )
makes the sum of the plus deviations equal to the sum of
the minus deviations from it , and ( 2 ) distributes these
-
deviations about equally + and along the course of the
line , and ( 3 ) has a geometrical form not much variant from
the form found in general for the test in question . ( That
form is , for most of our material , rectilinear ) . There is
thus an element of personal judgment in the decision as to
the relation line and consequent placement of the central
4 These were on the same day in immediate sequence giving single -session
variability .
5 For in Table 140 arrayed under 45 would show
example , the three cases
an average deviation of 4.5 from the average obtained by considering them
alone , whereas they show an average deviation of 8.35 from the average that
the whole table indicates as the probably true average for that array .
tendencies from which the variabilities were computed .

This , in general , has almost no effect save upon the arrays
toward either end , and its effect is greatly reduced so far
as our general problem is concerned , by reason of the over
lapping of the groups which we have used . Finally , the
personal judgment did not , so far as we know, favor any
one sort of relation of variability to magnitude more than
another ; it acted as a chance error .
The midpoint of a class interval of the array which was
nearest to the relation line was used as the central tendency
from which to compute the variability of the array . The
reason for adopting this crude procedure instead of com
puting the lines of best and computing the variability
fit
of
,
each array from the exact point on that belonged the
to
it
array in question was that we had to choose between
,
spending our time on few relations computed exactly and

a
on many relations estimated this rough way The latter

in
.
procedure seemed much more instructive per unit of time
spent
.
Table 141 shows the method where we use the array of

scores in one trial Test individuals scoring alike
of
in
in some other test closely enough correlated with Test X

to serve instead of trial with Test itself The central
X
a
tendency an array from which compute its variability

of
to
was determined just the case Table 140

as
of
in
We have four tables of the type Table 139 and six

of
teen of the type Tables 140 and 141 for Part of the
of
I
,
Thorndike Examination Their summarized results ap

.
pear as Tables 142 and 143. Our next task to combine

is
them give general estimate of the relation of the varia

to
bility of an individual his ability the case of this test

in
to
There are many suitable ways of doing this each having

,
certain merits The method which we have adopted

to
is
.
express each entry of line Table 142 percent of

of
as
a
to a
that line's average entry under abilities 70 89 and

to
average the percents under each ability after weighting

them as follows
:
This roughly weighted average

is
6
.
N Weight
1 or 2 1
3 - 5
-
2
12
6
·
8 7 6 5 234
13 19
20 29
30 39
-- - -
40 59
60 79
80 99
9
100 over 10
The use of the percents makes convenient compare
to
it
the relation found for this test with that found for others
.
The particular weighting used gives some weight each
to
group studied apart from the number contains
of
cases
it
,
and adds no weight size of population over 100 any
in
to
array any one group and economizes time The

in
one
,
.
same procedure was followed with the facts of Table 143
.
The relationship shown by the final line of Table 142
and the final line Table 143 may be seen more clearly
of
we group the results more coarsely In combining for

if
the coarse grouping the average percentages already ob-

,
tained are weighted each by the sum of the weights at-

tached Thus 133 with weight of and 50 with
to
it
a
5
.
weight of become 109 with weight of 94 with weight

a a
2
7
;
of and 79 with weight of become 88 with weight

a
3
8; 5
of 116 with weight of and 90 with weight

of
a
a
9
become 104 with weight of 17. The relations of Tables

a
142B and 143 with coarser grouping are shown Table

in
should be observed that the grouping carried out

It
144.
this manner after the variability groups or arrays
in
of
has been obtained from the original fine grouping does

,
not add any new attenuation All groupings what

in
"
"
follows were carried out in this manner

.
We have in the case of many tests material of the sort

,
,
shown by Tables 142 143 and 144 for the Thorndike

,
Part In two instances results of different tests have

I.
been combined abbreviate the presentation The first

to
is
.
where results of the regular Army Beta and the weighted

.
TABLE 140
1.
.
,
,2
S.
,
.
,I
.
.
THORNDIKE INT EXAM FOR H. GRADUATES '19 -'22 SERIES PART TRIAL ARRAYED UNDER TRIAL TEST OF FEB 22. 30-30 TO 34
,
35 35 TO 39 ETC.
1
Score in Trial 30 35 40 45 50 55 60 65 70 75 80 85 90 95 100 105 110 115 120 125 130 135
Score in
2
1
Trial 40
45
1
50
1
1
1
2
1 1
60
948 585
65
1
70
1
1
75
4
2
80
3 ∞
85
24 +32
1
90
1686
23 1231
3144
2
95
1222
1145
5
1
122 12221
100 1 3 1 1
2
3
1
3
2
2
105 3
110 10
1504
6 5
8
1 1
115
1
120
100030301
712
1214
25464
125
1
1
1
1
1232 1
130
12
1
2
12
135
140
1
1 1
145
150
1
4
2
1
3
3
1
n 13 15 18 21 27 20 32 27 16 23 10 11
27
Central Tendency
Used .50 60 65 65 70 75 80 85 90 90 95 100 105 110 110 115 120 125 125 130 135
Average
Deviation .10.0 5.0 8.35 6.65 5.00 9.40 9.60 10.35 10.0 7.85 8.50 10.5 8.30 5.95 7.5 6.75 6.5 6.80 11.25 6.25 7.5
TABLE 141
.
S.
.
.
THORNDIKE
,I
INT
. ,2
EX FOR H. GRADUATES '19 -'22 SERIES PART
.
TRIAL ARRAYED UNDER THORNDIKE EXAM TOTAL SCORE WOMEN STU-
,,
,
.
;
=
,. ,
DENTS IN HIGH SCHOOL NORMAL SCHOOL COLLEGE AND UNIVERSITY 30 30-34 35 35-39 ETC.
Score in
Thorndike Total 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95 100 105 110 N
Score in
I,
Part
2
Trial 40
1
2
ROROAA
41
1
01197
1
13
12
11
27
1
29
19 10 11
2VERTE
55
122TAT
44202
124 46521
9 7 5
THE
BEJ
28
19 19 10 87
5
95 + 17 28 22 14 97
100
1
19 36 19 10 106
IN
6 42
1
105
166
20 32 30 21 125
2200 797 1
110 13 18 13 19
30
HERE
104
-
HOODR
1
200 200006
115 1 1 2 BOL32
6
H 24 20 23 17 108
4 1
120
2 15 23 14 74
125
45
809
531 10 37
130
3
" 19
2577 4232
2215
3421
135
137 000+
312
8 84
2
16
140
1
1
2
2
6
8
n 26
28
38
2
64
35
5
0
68 124
11
136
1T
130 95 90 52 41 18 10 917
Central
Tendency
Used 50 55 60 65 70 75 80 85
115
90 95 100 105 110 115 120 125 130 135 140 150
Average
Deviation
in steps of
(5
0
).
units each 2.00 3.00 50 1.00 1.50 1.54 1.16 1.25 1.25 1.54 1.15 1.39 1.58 1.14 1.31 1.24 2.17 1.10 2.00 2.00
TABLE 142 ( A ) .
THORNDIKE INT . EX . FOR H. S. GRADUATES . PART I AVERAGE DIFFERENCE BETWEEN TWO TRIALS
SUM OF TRIAL 1 AND TRIAL 2
( SINGLE SESSION ) IN RELATION TO THE AVERAGE SCORE 2 TRIAL 2 )
A Normal School students . C = Night school men .
B = Candidates for college entrance . D = S . A. T. C. candidates .
( The average here is of 3 trials . The difference is Trial 3 - Trial 2. )
Average score 15 20 25 30 35 40 45 50 55 60 65 70 75 80
Average
Difference A 10.5 12.0 9.0 8.3 9.3 8.0 10.1 7.9
n 12 34 48 61 84 103
9
Average
Difference B 9.0 10.0 13.0 6.8 10.0 9.0
n 12 12 10
3
2
3
Average
Difference C 9.75 10.5 9.0 10.0 6.0 10.2 8.1 3.8 6.0 9.8 8.1 8.0 10.1 7.0
127
n 4 2 4 6 10 14 18 22 26 22 16 24
7
Average
Difference D 11.5 4.9 9.4
n 6
22 22
Weighted Average
Percent of the Va
riability at 70-89 117 127 108 120 72 123 107 84 90 108 117 84 118 100
Sum of weights 2 1 2 13 17 21 16 23
3
7
3
Average score 85 90 95 100 105 110 115 120 125 130 135 140 145
Average
Difference A 9.2 8.1 8.4 7.8 8.1 7.8 6.7 4.5 6.7 6.0 8.3
n 145 145 115 122 102 65 50 20 13
3
Average
Difference B 7.6 7.5 8.5 7.9 7.6 6.2 7.1 6.6 6.8 5.4 7.4 4.2 15.0
n 32 30 32 42 49 52 65 43 37 19 15
5
Average
Difference C 8.8 5.0 7.1 5.7 7.5 6.3 7.5 10.5 10.5 4.5
n 18 20 17 10
9
2
2
3
9
Average
Difference D 8.4 7.6 6.1 6.0 4.9 6.5 5.8 4.4
n 48 63 113 120 128 114 70 23
Weighted Average
Percent of the Va
riability at 70-80 103
88888
82 92 185
22223
96 87 75 78 76 62
88888
8885
8883
225
Sum of Weights 27 29 20 30 30 28 16 23 20 12
2
1
6
TABLE 142 ( B ) .
SAME AS TABLE 142 ( A ) , EXCEPT THAT THE DIFFERENCE IS BETWEEN TRIALS ON DIFFERENT DAYS ,
AND THAT THE AVERAGE SCORE IS FROM FOUR TRIALS . NORMAL SCHOOL STUDENTS .
Average
Score 15 20
20 25 30 35 40 45 50 55 60 65 70 75 80
Average
Difference 8.0 5.2 10.8
n 2 10 5
Weighted Average
Percent of the Va-
riability at 70-89 102 66 137
Weight 1 3 2
Average
Score 85 90 95
35 100 105 110 115 120 125 130 135 140 145
Average
Difference 8.5 6.3 8.0 7.3 8.4 7.4 6.8 6.1 1.0 9.0
n 19 21 17 24 28
86 18 16 7 1 3
Weighted Average
Percent of the Va-
riability at 70-80 108 80 102 93 107 94 87 78 13 115
Weight 3 5 4 5 5 4 4 3 1 2
Army Beta and a certain picture test ' have been combined ;
the second is where the results for the Haggerty Delta 2 ,

the Myers Test , the Kelley Trabue Completion , and the
Thorndike Visual Vocabulary are combined . In these com-
binations each is treated separately up to the point of the
Weighted Average Percents of the Variability at Ability X,
and care is taken to chose Ability X to represent closely the
same percentile ability in each of the tests to be combined .
Consequently , two variabilities are combined , only if they
belong to approximately the same ability . We use the
same method of weighting that is used for any one test.
given to different groups .
7 Consisting of tests 3 , 4 and 5 of the Thorndike , Part II .
TABLE 143 .
THORNDIKE INT . EXAM . FOR H. S. GRADUATES . PART I. VARIABILITY OF SCORE IN ONE TRIAL
ARRAYED UNDER SCORE IN ANOTHER TRIAL OR UNDER TOTAL SCORE IN THE ENTIRE
EXAMINATION . 10 10 TO 14 ; 15 15 тo 19 , ETC.
Night College candidates College freshmen

A
school men entrance
H
G
=
B
D
C
,
,
,
E,
F,
High School Normal
.
-
plus some others . Men of

= Women School and College
K
, in
L
=
J
I,
,
.
.
college grade and below School students Normal School
at
Summer
M
N
.
. Men in High School and College
O
P
=
,
.
CENTRAL TENDENCY OF THE ARRAY
.
10 15 20 25 30 35 40 45 50 55 60 65 70 75
Group 12.5 30.0 22.0 19.0 15.2 16.6 15.6

A
n 16 27 35 50
4
Group 5.0 5.0 10.0 15.0 13.8 8.4 21.6 16.8 20.6 16.0 20.0 11.2 15.2 14.0
B.
n 13 19 14 15 18 17 19 10
2
8
4
Group 38.4 10.0 10.0 20.0 12.0 12.6 10.0 16.4 22.5 17.0 14.4 17.2 9.4
1 0
C
n 11 12 13 18 18 16
4
7
6
Group 5.0
D
2
Group 15.0 11.8
E
.2
11
2
Group 30.0 13.0 10.0
F
1 0
n 2
10 10
Group 10.0 17.5 14.1

G
TIMES
n 27
2
Group 20.0 7.8

H
n
DEVIATION
Group 40.0 20.0 11.7 25.0 30.0 19.6 12.5 12.5

I
n 26 40 64
8
2
1
Group 20.0 30.0 5.0 10.0 15.0 15.4

AVERAGE
n 26
8
1
2
2
Group 20.0 10.0 15.0 10.0 18.8

K
n
8
1
1
6
Group 10.0 15.0 10.0 18.0 10.0 12.9 10.8 15.8

L
n 12 31
2
5
1
Group 20.0 10.0 20.0 17.1 17.5 12.8 15.1

M
1 0
n 14 20 18 71
3
1
Group 10.0 25.0 24.0 17.5 13.8 18.2 15.6 16.5 17.9
N
n 12 16 22 89 49 67
5
2
1
Group 40.0 5.0 11.4 14.0 19.5 15.0 15.9 14.7

O
n 10 21 36 51 71
2
7
2
Group 30.0 10.0 20.0 18.6 18.2 19.0 14.4

P
n 11 21 36
7
1
Weighted Aver
age Percents of
the Variability
888888
at 70 to 89
Sum of Weights
.133 50 94 79 116
฀
90 125
15
106
16
101
28
120
22
114
45
99
38
96
64
96
66
9
8
5
3
2
5
TABLE 143- ( Continued )
80 85 90 95 100 105 110 115 120 125 130 135 140 145
Group A 12.7 18.3 14.0 13.3 15.0 10.0 20.0

D 37 41 30 18 6 4 2
Group B 12.9 11.8 13.0 15.6 5.0 16.6 16.6 10.0

n 7 11 10 9 4 3 6 2
Group C 16.7 18.8 16.4 7.2 12.6 10.0 15.6 10.0 10.0
n 15 16 11 7 8 6 9 2 2
Group D 11.4 50.0 11.7 16.2 13.7 16.5 11.4 12.5 9.4 11.2 13.6 12.9 10.0
4
671
n 7 2 12 13 41 92 44 112 50 85 25 14
Group 12.0 9.2 15.2 13.7 12.9 14.8 11.9 12.5 14.5 16.0 10.0 2.5 10.0
E
n 10 12 31 38 43 61 59 61 68 16 13
2
4
.2 Group 18.3 16.8 13.0 14.4 13.3 16.3 13.2 11.9 14.4 13.8 10.0 15.0
F
n 12 31 37 45 61 60 106 21 16 13 4
2
Group 17.5 15.6 15.5 14.6 17.7 11.1 5.0
TIMES
n 71 115 163 118 44
2
9
.
Group 10.4 16.5 12.7 12.7 13.7 10.5 16.0

H
DEVIATION
02
26 68 118 165 110 43 10
Group 15.7 15.2 16.6 15.5 15.3 11.9 11.5 15.5 11.1 13.0 16.0 10.0
J I
n 70 127 134 139 102 93 53 42 18 10
1
Group 11.6 12.5 12.5 15.4 11.5 13.9 15.8 11.4 13.1 12.4 21.7 11.0 20.0
AVERAGE
n 38 64 68 124 130 136 95 90 52 41 18 10
5
Group 19.2 20.7 17.7 17.0 21.0 16.6 13.0 13.5 13.0 16.0 12.5 15.0
K
n 13 15 39 27 20 32 43 23 10 15 4 2
Group 17.9 17.3 16.1 22.3 14.6 14.5 13.8 20.0 15.7 20.0 5.0
L
n 19 22 23 22 35 31 32 7
5
2
1
Group 15.6 14.4 16.1 14.8 16.3 12.0 18.8 6.7 11.3 10.0
M
100
10
n 55 107 38 44 41 10
8
3
8
Group 14.8 13.9 16.4 16.1 21.8 8.9 20.8 15.0 10.0 20.0
N
n 48 33 42 23 11 12
2
9
1
1
Group 16.7 15.4 14.9 17.3 17.6 16.6 14.4 15.2 11.5 15.6 10.8 11.4
O
n
10
73 90 89 109 107 87 103 48 39 25 12

7
Group 14.3 13.4 14.9 14.0 15.8 14.6 13.5 14.4 12.6 12.5 13.2 12.8 7.1
P
1 0
n 53 70 73 91 90 108 108 85 102 48 41 36

7
Weighted Aver-
age Percents of
the Variability
at 70 to 89 ...........
103 102 112 103 103 99 91 95 91 92 85 72
8000
1 0
535
995
395
398
23
Sum of Weights 86 84 92 89 106 90 100 77 75 56 34 26 11

35
.
TABLE 144
,
.
THE SUMMARIES OF TABLE 142 AND TABLE 143 WITH COARSER GROUPING
Ability 10 20 30 40 50 60 70 80 90 100 110 120 130 140
;
Relative Variability
Table 142 75 118 90 100 90 62 115
2
6
9
4
00 8
4
Weight 10
;
Relative Variability
Table 143 109 88 104 115 112 107 96 102 108 99 95 93 89 62
฀฀฀
THE RELATION OF VARIABILITY
7
8
Weight 17 31 50 83 130 170 181 196 177 131 60 *14
TO
ABILITY
513
Our results for all tests are summarized in Table 145 .

Each pair of lines of entries represents the results from
one or more such tables as Table 142 or 143 giving the
relative variabilities with increasing amounts of the ability
and the weights appertaining thereto . In Table 145 all
the entries in one column are for the same amount of abil
ity , to a rough approximation . That is , we arrange the
results for any test so that it will the general scale from
fit
about 10 percentile fourth grade intellect about 95
to
a
a
ap
-
percentile college intellect We have done this only
.
proximately because absolute precision unattainable
is
with present knowledge and because only rough approxi
a
mation needed for our purpose discover any general
is
to
relation between the amount ability and the variability
of
attached thereto after allowance for swelling shrinking
or
,
of the real values of the units of the scores
.
Although all the variabilities any column pertain
in
to
approximately the same amount of ability they cannot ,
properly be added by columns show the general drift
of
to
Table 145. For the numbers in one row may be percen

tages on base very different from that used in some other
a
row For example the Army Exam numbers are on

a
.
approximately median 9th Grade ability as base the

a
Haggerty Myers Trabue combination numbers are on ap

-
proximately median 6A Grade ability as base the last

a
seven rows are on approximately median Columbia College

Freshmen ability as base
a
We shall later mass the results of Table 145 in form

a
more suitable reveal the general relation Table 145

to
is to show the relation in each of the different tests or test

.
amalgamations shows no evidence any tendency for

of
It
.
variability increase with ability On the contrary

to
if
,
.
one had to choose between law of increase with amount

a
ability and law of decrease with amount ability he

of
of
a
would have choose the latter Table 145 also shows in

to
,
.
all but three of its lines tendency for the numbers to

a
,
increase from the lowest extreme for certain distance

a
,
or to decrease toward the highest extreme or to do both .

The increase comes in the low range for a test , regardless
of whether the ability there is that of a low 4th grade
pupil , or a low ninth grade pupil . The decrease also comes
in the high range for a test ; regardless of what real ability
that range covers . This is what would follow from the
presumably common tendency of a person or committee
devising tests , to provide rather fully tasks along the range
of abilities which the test was designed to measure , and
to have fewer tasks over an equally long range below or
above that . If each task is given an equal credit in the
score , the same real variability will of course be repre-
sented by a smaller number in the low or in the high than
in the middle range , supposing that abundant time is given
to exhaust the subjects ' abilities . Table 146 gives an illus-
tration of such a case . When the time is limited so that
only some , or even none of the individuals do all that they
could do , the same effect may be found .
We have studied the facts of Table 145 in connection
with the tests themselves in considerable detail and from
several points of view . On the whole it seems to us that
the most satisfactory explanation is to regard the relation
as y = k , that is , to consider the variability of an individual
as independent of his ability , accounting for the rises and
falls in the curves of Table 145 by irregularities in the
amount of real ability corresponding to one unit of score
in a given test at different ranges of the score . We shall
not rehearse the evidence which has impelled us to that con-
clusion . Some of it is personal . The reader who doubts
it should try any alternative relation , using the same cri-
teria of merit for the two .
The general drift of the facts of Table 145 can be stud-
ied more easily if they are more coarsely grouped and if
the percentages , where possible , are computed on the base
of the variability attached to an ability of approximately
the median ninth - grade pupil . Table 147 gives such a pre-
sentation . In the case of the Modified Thorndike II , the
.
TABLE 145
.
THE RELATION OF THE VARIABILITY OF AN INDIVIDUAL TO HIS AMOUNT OF ABILITY IN FIFTEEN TESTS OR AMALGAMATIONS OF TESTS THE UPPER
IS
;
;(
)
NUMBER THE MEASURE OF VARIABILITY THE LOWER NUMBER IN ITALICS IS THE WEIGHT ATTACHED TO IT
4th Grade 6th Grade

Median Median
1.
Alpha 61 73 97 104 102 132 110 109 128 84 107
8
8
3
8
12 17 10 12 10
A
2. Exam 68 132 88 91 107 94 90 103 102 93 87 104 99 107 93 109
5788
9
4
888
89853
36
18 20 16 34 17 16 22 34 45 10 29 41 37
3.
I.
.
. ,
Th Part 133 50 94 79 116 90 125 106 104
5
9
5
3
8
2 15 16 28
.
3T
4. Otis Adv 82 89 54 93 73 99 83 62 97 124 96 99 110 80 85 111 92
3
3
4
898
366
9
7
8
3
3 67 10 15 10 13 13 13
5.
Terman Group 56 96 117
3
5
4
6.
Stanford M. A. 202 83 91 106 100 109 92
3
766
5
1288
7. National 94 92 107 108 104 107 93 78 ฀
7
8
20 14 37 14 23
8.
.
฀฀฀
Hag Myers Composite ... 21 71 74 98 106 80 102 92 92 116 111 105 92 108 102 90 91 96 100 86 82 104 80
6
9
10 10 16 18 14 16 27 18 16 11 25 28 10 22 23 15 19 21 13 ฀15 10
9.
+
Beta Picture Test 65 73 95 87 89 83 102 106 97 99
175
8888
355
16 10 22 16 26 13 20 18 14
10. Toops Clerical 89 121 76 107 5 112 87 92 118 131 161 115 118
5
27
5
6
7
6
77 3
11. Th New Part 63 42 59 59 49 89
28828
271
4
4
7
5392
12. Th Part II 131 64 118 101

3
5
6
5
,AI .
, II ,
13. Th Part Mod
14. Th Part III 95

66
15. Th Part II New
, I.
,
.
,
16. Th New Pt 45 min
,. ,. ,. ,. ,. ,. ,.
17. Th Total 72 119
26
18
(
).
TABLE 145— Continued
College
9th Grade Freshman
Median Median
125 117 108 95 69 75 97 95
1.
86 114 121 89 129 71
Alpha
9
8
38
8889
10 11 11 11 18 10 12
93 84
.
2. Exam A. 102 113 95 111 103 120 91 90
5
14 18 12 20 12 15 18 14 10
102 112 103 103 95 99 91 95 91 92 85 72
I.
96 96 102
,.
97
3.
Th Part .120 114
84 92 89 106 90 100 77 75 56 34 26 14
22 45 38 64 66 86
87 54 67 67 67
.
4. Otis Adv 65 127 116 92 80 78 86 99 89 66
2254
1
9
20180
9
11 11 13 12 11 10 11 10 4
98 100 84 110 85 83 85
5.
90 92 99 135
Terman Group
4
8284
6
6
6
10 11 10
335
11
6.
Stanford M. A. 99 74 100
3
7. National 65 59 79 60
8
2809
7
13
8.
.
Hag Myers Composite 80 64
6
15
9.
Beta Picture Test 78 81 94 90 88 46
4
co
12 12 11 12
10. Toops Clerical 98 79
3
3
92 122 89 104 42 98 63 184
,I
฀฀
,
100 110
,.
11. Th New Part A. 83
1
1
2
5
4
8
6
6
฀
85 104 115 93 93 85 111 45 50 100 116 65 103 30 69
.II
Part 86 98 83 100 106 106 109 109 93 100
฀฀ 95 107 110 88 97
836
12. Th
4
5
3
2
4
5
8
4
88888
8
7
4
9
3
3
6
14 16 16 19 10 16 10 17 10 18 17 14 14 >
8285
53 89
฀฀
90 79 69
.
123 92
,
Part Mod 56 50 100 96 103
13. Th
5
4
5
6
7
II
4
20
2
2
฀ ฀
88 102 109 100 93 102 118 85 66 67
.
Part 60 86 81 93 82
14. Th III
2
4
13
8888
13 25 23 26 30 27 27 26 19 18 93
83 100 109 91 77 77 65
,
15. Th Part New 45
© 91 70 80 ➢87
3
6
6
II
7
5
5
7
3 ฀฀฀
11000
97 117 141 119 115 135

° 113 93 125 100
.
102 67
,I
.,
New Pt 45 min 122
16. Th
6
4
8487
7
7
5
6
}
» 3 ฀ ฀
100000
105 104 95 90 81 75 162
,. ,. ,. ,. ,. ,.
Total 117 113
5
17. Th
1888
55 69 88 99 78 46 26 12
123302
.
TABLE 145
THE RELATION OF THE VARIABILITY
.
OF AN INDIVIDUAL TO HIS AMOUNT OF ABILITY IN FIFTEEN TESTS OR AMALGAMATIONS OF TESTS THE UPPER
IS
;
IS
;(
)
NUMBER
IT
THE MEASURE OF VARIABILITY THE LOWER NUMBER IN ITALICS THE WEIGHT ATTACHED TO
4th Grade 6th Grade

Median Median
1.
Alpha 61 73 97 104 102 132 110 109 128 84 107 83
7
8
8
3
3
8
6
12 17 10 12 10
2.
.
Exam A 68 132 88 91 107 94 90 103 102 93 87 104 99 107 93 109
9
18
8
4
20 16 34 17 16 22 34 45 10 29 41 37
,.
I.
3.
Th Part 133 50 94 79 116 90 125 106 104
5
5
3
9
2 15 16 28
4.
.
Otis Adv 82
8 89 54 93 73 99 83 62 97 124 96 99
] 110 80 85 111 92
3
5
3
4
6
8
9
฀ 67 7 10 15 10 13 13 13
5.
3
5
4
6.
Stanford M. A. 202 83 91 106 100 109 92
7
6
7
5
8
3
4
7. National 94 92 107 108 104 107 93 78 73
5
8
20 14 37 14 23 13
8.
.
.
Hag Myers Composite 21 71 74 98 106 80 102 92 92 116 111 105 92 108 102 90 91 96 100 86 82 104 80
6
9
10 10 16 18 14 16 27 18 16 11 25 28 10 22 23 15 19 21 13 15 10
9.
+
7
16 10 22 16 26 13 20 18 14
10. Toops Clerical 89 121 76 107 112 87 92 118 131 161
5 115 118
5
6
7
5
7
6
6
5
77
,
11. Th New Part 63 28 42 59 59 49 89
1
4
4
7
2
12. Th Part II 131 64 118 101
5
6
5
,
,AI .
13. Th Part II Mod
14. Th Part III

Petra
95
6
,
15. Th Part II New
.
.,
,I
,. ,. ,. ,. ,. ,. ,.
17. Th Total
72 119
6
18
(
.)
TABLE 145- Continued
College
9th Grade Freshman
Median Median
1.
Alpha 86 114 121 89 129 71 125 117 108 95 69 75 97 95
9
9
6
4
3
10 11 11 11 18 10 12 3
2. Exam A. 102 113 95 111 103 120 91 90 93 84
14 18 12 20 12 15 18 14 10
3.
I.
.
. ,
Th Part 120 114 97 96 96 102 102 112 103 103 95 99 91 95 91 92 85 72
22 45 38 64 66 86 84 92 89 106 90 100 77 75 56 34 26 14
.
4. Otis Adv 65 127 116 92 80 78 86 99 89 66 87 54 67 67 67
9
9
9
4
4
1
11 11 13 12 11 10 11 10
5.
Terman Group 90 92 99 135 98 100 84 110 85 83 85
7
6
4
6
6
11 10 11 10
6.
3
3
3
7. National 65 59 79 60
7
8
2
13
8.
.
Hag Myers Composite 80 6+
6
15
9.
Beta Picture Test 78 81 94 90 88 46
4
3
36
12 12 11 12
3
3
11. Th New Part A. 83 100 110 92 122 89 104 42 98 63 184
5
5
2
1
1
8
6
4
1
,.
12. Th Part 98 83 100 106 106 109 109 93 100 95 107 110 88 97 85 104 115 93 93 85 111 45 50 100 116 65 103 30 69
8686
2439
II
3
4
6
7
8
8
4
5
8
4
5
4
3
2
14 16 16 19 10 16 10 17 10 18 17 14 14
,
,I .
13. Th Part II Mod 56 50 100 96 73 103 90 123 79 92 69 53 89
5
5
6
4
4
2
4
7
4
3828
92
.
80
14. Th Part III 60 86 81 93 82 88 102 109 100 93 102 118 66 67
9
$
2
8888
886
13 25
J 88
23
Ş 26 30 27 27 26 19 18 13
11895
,
15. Th Part II New 45 91 70 80 83 87 100 109 91 77 77 ฀ 35
3
5
5
8
7
6
6
4
.,
.
,I
16. Th New Pt 45 min 122 102 67 113 93 125 100 97 117 141 119 115 135
88888
3
8
6
3
7
6
7
4
฀฀
ao ฀ 3
฀
฀฀฀
,. ,. ,. ,. ,. ,. ,.
17. Th Total 120 117 113 105 104 95 90 81 162
5
3886
38
37 55 69 88 99 78 46 26 ฀
.
TABLE 145
.
IS
;
;(
)
NUMBER THE
IT
MEASURE OF VARIABILITY THE LOWER NUMBER IN ITALICS IS THE WEIGHT ATTACHED TO
4th Grade 6th Grade

Median Median
1.
Alpha 61 73 97 104 102 132 110 109 128 84 107 83
7
8
3
3
8
6
12 17 10 12 10
2.
A
Exam 68 132 91 107 94 90 103 102 93 87 104 99 107 93 109
88888
9
18
8
4
20 16 34 17 16 22 34 45 10 29 41 37
3.
I.
.
. ,
Th Part 133 50 94 79 116 90 125 106 104
5
9
5
3
8
15 16 28
.
4. Otis Adv 82
฀ 89 54 93 73 99 83 62 97 124 96 99 110 80 85 111 92
3
3
5
3
4
6
8
7
9
67 10 15 10 13 13 13
5.
5
4
6.
Stanford M. A. 202 83 91 106 100 109 92
7
6
5
8
4
7. National 94 92 107 108 104 107 93 78 73
7
5
8
20 14 37 14 23 13
8.
.
.
Hag Myers Composite ... 21 71 74 98 106 80 102 92 92 116 111 105 92 108 102 90 91 96 100 86 82 104 80
6
9
10 10 16 18 14 16 27 18 16 11 25 28 10 22 23 15 19 21 13 15 10
9.
+
7
16 10 22 16 26 13 20 18 14
10. Toops Clerical 89 121 76 107
5 112 87 92 118 131 161 115 118
5
6
7
7
6
5
77
,
11. Th New Part 63 28 42 59 59 49 89
2
1
4
4
7
12. Th Part II 131 64 118 101

5
,
,AI .
13. Th Part II Mod
14. Th Part III

Petra 95
6
15. Th Part II New
,
, ,I.
.
,. ,. ,. ,. ,. ,. ,.
17. Th Total 72 119
6
18
(
).
College
9th Grade Freshman
Median Median
1.
Alpha 86 114 121 89 129 71 125 117 108 95 69 75 97 95
9
3
8
9
6
4
38
10 11 11 11 18 10 12
.
2. Exam A. 102 113 95 111 103 120 91 90 93 84
5
14 18 12 20 12 15 18 14 10
I.
,.
3.
Th Part 120 114 97 96 96 102 102 112 103 103 95 99 91 91 92 85 72
22 45 38 64 66 86 84 92 89 106 90 100 77 75 56 34 26 14
.
4. Otis Adv 65 127 116 92 80 78 86 99 89 66 87 54 67 67 67
2859
9
9
4
9999654
1
9
11 11 13 12 11 10 11 10
5.
Terman Group 90 92 99 135 98 100 84 110 85
2883
85
6
4
4
6
6
6
11 10 11 10
6.
3 3
7. National 65 59 79 60
7
8
2080
13
8.
.
6
15
9.
+
Beta Picture Test 78 81 94 90 46
888
4
co
12 12 11 12
3
3
,
11. Th New Part A. 83 100 110 92 122 89 104 42 98 63 184
1
5
5
2
6
1
1
6
4
2
1
.
12. Th Part II 86 98 83 100 106 106 109 109 93 100 95 107 110 88 97 85 104
฀฀ 115 93 93 85 111 45 50 100 116 65 103 30 69
6
3
9
8
4
g28
6
8
5
4
5
8
4
3
2
14 3 16 19 10 16 10 17 10 18 17 14 14
,I .
13. Th Part II Mod 56 100 96 73 103 90 123 79 92 69 53 89
5
2
4
5
09890
7 7
4
2828
4
4 33932
2828
14. Th Part III 60 86 81 93 82 88 102 109 100 93 102 118 85 66 67
8
2
1886
13
88888
25 23 26 30 27 27 26 19 18 13
15. Th Part II New 45 91 70 83 87 100 109 91 77 77 65 35
฀80
5
3
5
7
3
7
6
6
4
,
, . , ,I.
16. Th New Pt 45 min 122 102 67 113 93 125 100 97 117 141 119 115 135
88787
5
฀฀
7
6
3
8
6
202023
4
,. ,. ,. ,. ,. ,. ,.
17. Th Total 120 117 113 105 104 95 90 81 75 162
5
37 55 69 88 99 78 46 26 12
.
TABLE 145
THE RELATION OF THE VARIABILITY OF AN INDIVIDUAL
.
TO HIS AMOUNT OF ABILITY IN FIFTEEN TESTS OR AMALGAMATIONS OF TESTS THE UPPER
IS
;
;(
) IS
NUMBER
IT
THE MEASURE OF VARIABILITY THE LOWER NUMBER IN ITALICS THE WEIGHT ATTACHED TO
4th Grade 6th Grade

Median Median
1.
Alpha 61 73 97 104 102 132 110 109 128 84 107 83
3
6
8
12 3 17 10 12 10
.
2. Exam A 68 132 91 107 94 90 103 102 93 87 104 99 107 93 109
8828
8888
18
8
20
4
16 34 17 16 22 34 45 10 29 41 37
I.
3.
,.
Th Part 133 50 94 79 116 90 125 106 104
5
9
5
3
2 15 16 28
4.
.
Otis Adv 82 89 54 93 73 99 83 62 97 124 96 99 110 80 111 92
3
3
18989
4
6
7
3
8
฀
9
13
1895
7 10 15 10 13 13 13
5.
5
4
6.
Stanford M. A. 202 83 91 106 100 109 92
7
6
7
5
8
3
4
7.
National 94 92 107 108 104 107 93 78 73
7
8
20 14 37 14 23 13
8.
.
Hag Myers Composite ... 21 71 74 98 106 80 102 92 92 116 111 105 92 108 102 90 91 96 100 86 104
82 ฀ 80
6
9
10 10 16 18 14 16 27 18 16 11 25 28 10 22 23 15 19 21 13 ฀ 15 10
9.
7
16 10 22 16 26 13 20 18 14
10. Toops Clerical 89 121 76 107 112 87 92 118 131 161 115 118
5
5
6
7
5
7
5
6
6
S
77
A
11. Th New Part ...... 63 28 42 59 59 49 89
2
1
4
4
7
12. Th Part II 131 64 118 101

3
5
5
Svettek
,
,I .
, II
13. Th Part Mod
14. Th Part III 95

6
15. Th Part II New
,
, ,I.
,. ,. ,. ,. ,. ,. ,.
17. Th Total 72 119
6
18
(
.)
College
Freshman
9th Grade
Median Median
117 108 95 75 97 95
121 89 129 71 125
1.
86 114
9
Alpha
38
699
10 12 3
10 11 11 11 18
120 91 90 93 84
.
A. .102 113 95 111 103
2. Exam
18 12 20 12 15 18 14 10
14 85 72
112 103 103 95 99 91 95 91 92
96 96 102 102
I.
120 114 97
,.
3.
Th Part 106 90 100 77 75 56 34 26 14
45 38 64 66 86 84 92 89
22 67
78 86 99 89 66 87 54 67
.
80
4.
127 116 92
1
65 4
A988
Adv
9
Otis
9
12 11 10 11 10
11 11 13
99 135 98 100 84 110 85 83 85
5.
90 92
4
8784
Terman
6
Group
7
11 10
6
11 10
6.
3
65 59 79 60
7. National
18000100
8
7
13
80 6+
8.
.
Hag Myers Composite
6
15
94 90 88 46
9.
Beta Picture Test 78 81
4
12 12 11 12 3
Clerical 98 79
10. Toops
3
3
100 110 92 122 89 104 42 98 63 184
83 8
1
1
New Part A.
1
11. Th
8
6
6
* 2
93 100 95 107 110 88 97 85 104 115 93 93 85 111 45 50 100 116 65 103 30 69
,.
86 98 83 100
8
2
4
5
88
5
4
฀ 106 106 109 109

7
8
A
Part
8888
12. Th 43
2439
6
II 14
4
14
3
16 19 10 16 10 17 10 18 ฀ 17
14 3 16
73 103 90 123 79 92 69 53 89
56 50 100 96
,I .
,
3
8787
4
5
6
Mod
5
Part
88987
13. Th
4
4
4
20
2
II
102 109 100 93 102 118 85 66 67
86 81 93 82 88
.
60
9
2
14. Th Part III 27 26 19 18 13
4
13 25 23 26 30 27
388338
83 100 109 91 77 77 ฀฀65 35
45 91 80
,
฀
New
7
15. Th Part
5
3
฀฀฀
1092
97 117 141 119
1
115 135
122 102 113 93 100
.
,I
II ,
.
88787
3
New 45 min
5
16. Th Pt
2
° ฀฀฀
105 95 90 81 le75 162
120 117 113
5
50013848
,. ,. ,. ,. ,. ,. ,.
17. Th Total
geaiad
55 69 99 78 46 26 ?
37
888888
1388888
.
TABLE 145
.
IS
;
;(
IS
)
NUMBER THE
IT
MEASURE OF VARIABILITY THE LOWER NUMBER IN ITALICS THE WEIGHT ATTACHED TO
4th Grade 6th Grade

Median Median
1.
Alpha 61 73 97 104 102 132 110 109 128 83
84 107
7
8
3
6
8
12 3 17 10 12 10
2.
Exam A 68 132 91 107 94 90 103 102 93 87 104 99 107 93 109
80086
9
18
3886
8
4
20 16 34 17 16 22 34 45 10 29 41 37
.
. ,
I.
3.
Th Part 133 50 94 79 116 90 125 106 104
5
5
3
9
8
2 15 16 28
.
4. Otis Adv 82 89 54 93 73 99 83 62 97 124 96 99 110 85 92
111
4
3
6
8
6
13
7
9
22
1280
3 3 7 10 15 13 13 18
5.
5
4
6.
Stanford M. A. 202 83 91 106 100 109 92
7
6
7
8
5
Co
3
4
7.
National 94 92 107 108 104 107 93 78 73
7
5
8
20 14 37 14 23 13
8.
.
...
Hag Myers Composite 21 71 74 98 106 80 102 92 92 116 111 105 92 108 102 90 91 96 100 86 82 104 80
CATERIN
6
9
10 10 16 18 14 16 27 18 16 11 25 28 10 22 23 15 19 21 13 ฀15 10
9.
7
2928
16 10 22 16 26 13 18
20 14
฀฀฀฀
10. Toops Clerical 89 121 76 107 112 87 92 118 131 161 115 118
5
5
6
7
7
6
5
5
27
฀฀฀ 3
11. Th New Part 63 28 42 59 59 49
฀฀ 89
221
*
4
4
8828
7
Part
Ampl
12. Th II 131 64 118 101
6
5
5
, II ,
,AI .
13. Th Part Mod
14. Th Part III 95

6
,
15. Th Part II New
,I
.
.,
,. ,. ,. ,. ,. ,. ,.
17. Th Total 72 119
6
18
(
).
College
9th Grade Freshman
Median Median
1.
Alpha 86 114 121 89 129 71 125 117 108 95 75 97 95
6988
9
8
37 38
10 11 11 11 18 10 12 3
2. Exam A. .102 113 95 111 103 120 91 90 93 84
4 5
14 18 12 20 12 15 18 14 10
I.
3.
.
. ,
Th Part 120 114 97 96 96 102 102 112 103 103 95 99 91 95 91 92 85 72
61
22 45 38 64 66 86 84 92 89 106 90 100 77 75 56 34 26 14
REGU865
633888
.
4. Otis Adv 65 127 116 92 80 78 86 99 89 66 87 54 67 67 67
AAYOGBAN
20180
3889
4
9
4
1
11 13 12 ฀ 11 10 11 10
5.
Terman Group 90 92 99 135 98 100 84 110 85 83 85
6
7
NO
87844
6
6
4
6
11 10 11 10
6.
3
7. National 65 59 79 60
1801353
7
8
2
8.
.
15
+
9. Beta Picture Test 78 81 94 90 88 46
4
12 12 11 12 3
3
4682893
,
11. Th New Part A. 83 100 110 92 122 89 104 42 98 63 184
6
5
8
2
6
4
1
1
1
2 ฀
.
12. Th Part 86 98 83 100
≡ 106 106 109 109 93 100 95 107 ฀฀฀
110 88 97 85 104 115 93 93 85 111 45 50 100 116 65 103 30 69
88888
II
3
3
4
6
7
2439
8
8
5
4
5
4
3
2
14 16 16 19 10 16 ฀ 10 17 10 18 17 14 14 ฀
,I .
56 50 100 96 73 103 90 123 79 92
4103
13. Th Part II Mod 69 53 89

5
5
2
4
4
1727
4
6
88987
4
02
฀ 92
14. Th Part III 60 86 81 93 82 102 109 100 93 102 118 85 66 67
4
9
888
8888
2
13 25 23 26 30 27 27 26 19 18 13
15. Th Part New 45 91 70 80 83 87 100 109 91 77 77 65
฀฀ 35
3
5
8
5
3
7
7
6
6
4 ฀฀฀ 3
,
, . ,II I.
,
.
16. Th New Pt 45 min 122 102 67 113 93 125 100 97 117 141 119 115 135
189 190
5
8
8787
6
7
7
6
3
2 3 3
195
,. ,. ,. ,. ,. ,. ,.
17. Th Total 120 117 113 105 104 95 90 81 75 162
12
5
37 55 69 88 99 78 46 26 ?12
1000085
Thorndike II
, new style , and the Thorndike I , new style ,
with 45 minutes time , the data at or near median ninth

grade ability are very scant , so we have combined the re-
sults for these three tests giving equal weight to each and
report them as an addition to Table 147 .
TABLE 146. THE EFFECT OF SELECTION OF TASKS
Individual's Credit
Real Given Individual's
Real Average for the Average Devia-
Diffi- Individual's Deviation Task in Individual's tion in Terms
culty Real Average from his the test Average of the test
Task . of Task . Ability . Average . Score . Test Score . Scores .
a 101 1
b 106 106 5
10 1
1 1 1
c 111 111 10
5 1
6∞ 5 234
d 116 116 5
10 1
e 121 121 10
5 1
1
126 126 10
5 1 1+
g 130
1
131
h
132
1 1 1 1 1 1 1 1 1
j i
133
5-
10 510
134 134 11
k
135 135 12 5
10 5
1
m 136 136 13 5
H234
51 5
n 137 137 14 5-
O 138
139
r q Р
*
140
1+
10
141 141 18
5 5
1
10 10
8 146 146 19
1
151 151 20
10 510 5
1
1
t
u 156 156 21
1
*
V 161 161 22
22
510 5
1 1 1
W 166 166 23
1
X 171
There this coarser table as in Table 145 no evi-

in
is
,
dence that we can detect that the variability either in-

creases or decreases general with the magnitude
It
in
Concerning the effect of the error due inaccuracy averages men-

of
to
8
tioned on page 500 we now find that no allowance need be made The error
.
might have reduced some real tendency but there little evidence any
of
so
is
,
tendency that we cannot tell whether the reduction has been from tendency
a
for the variability increase from tendency for Not

or
to
to
decrease
it
a
,
knowing in which direction make the correction we need not make any
to
.
.
TABLE 147
IN
FIFTEEN TESTS OR AMALGAMATIONS OF TESTS USING
OF THE VARIABILITY OF AN INDIVIDUAL TO HIS AMOUNT OF ABILITY
THE RELATION
.
EIGHT LEVELS OF ABILITY
Seventh Upper College Upper

Sixth Ninth
Low Level Fourth High Freshmen College
Eighth Grade
Grade Grade School Alpha
Fourth or Grades Alpha
+
Alpha Alpha Alpha 150 or

Below Alpha Alpha Alpha 130-149
90-109
฀฀
35
15-29 30-49 50-69 70-89 110-129
48 101 35 88 10
42 107
5
101
8
109 35 108 36 79
67 15 97 104 63 83 24
Alpha 88 98 98 82 100 81 266
88 89 84
.
A. 90 54 94 399 92 382
Exam 15 102 48 100 144
91 86 40 62 11
Thorndike 100 48 85 41
4
98 51
2895
6
77 31 34 43 89
I.
Otis Adv 8512 1062 21 116 17 94
85 100
8 9
116
8
Terman Group 117 100
115 13 116
THE RELATION
138 10 33 109 10
St. Mental Age 56 119 31 100
139 41 153
National
,
,
Haggerty Myers 141 106 113 100 46
฀
100 99 120 23 81
etc. 118 38 100 38 106 1
103 32 103 51
+
Beta Picture 812 23 31 100 14
98 30 110
107 15 129
2
Toops 53 14 100 31
66 96 99 87 49
9
Thorndike New Ia 100 52 105 90
87
111 104 123 114½ 42
Thorndike 100 31
288806
II 83 223 75 43
157
6
Thorndike III 100 110 92
61
Thorndike Total 100 1062 852
105 102
Median 852 100
104 97
6 85
Average equal 97 100
92 103 106
, , , ()
weights
Average half
to
weight entries
sum of 104 96 85
whose 99 100
93 103 108
( )
10
<<
OF VARIABILITY TO ABILITY
weights
Combination of
Thorndike modi
123 41
II
8
fied New 121 57 143 78

II 100
and New
,I (
Median including
Thorndike com
519
.
bination with wt 100 107 95 8712
105 102
100
2
of 852
)(
Average including
the Thorndike
combination with 100 107 1042 922
103 106 97
92
)2
weight of
could be argued that it increases a trifle over low levels

of intellect and decreases a trifle over high levels of intel-
lect , but the facts presented in Chapter VII make it much
more probable that the tasks of these tests were too scant ,
or the credits too low , or both , at the easy and hard ex-
tremes of the tests .
We thus have grounds for replacing cumbrous and am-
biguous measures of unreliability of a test in the shape of
its self correlations , by the simple probable error of the
measurement . An adequate determination of this will
suffice anywhere along the range from a dull 10 - year - old to
a gifted 20 - year - old . We also have validated a new method
of measuring the real differences signified by differences
in arbitrary scores . That transmutation of arbitrary
scores is best ( other things being equal ) which makes the
variability of an individual the same for all levels from
that of a dull 10 - year - old to that of a gifted 20 - year - old .
We have further made a step toward using the arrays of
a distribution separately in measuring either the real dif-
ferences of total scores or the real differences of single
tasks in difficulty. The form of distribution of an array
is determined by three causes , the form of distribution of
the average abilities of individuals in the group in ques-
tion , the form of distribution of the variations of an in-
dividual around his own average , and the relation of the
variability of an individual to his average ability . We
have previously shown that the variability of an individual
is symmetrical and roughly of the form of the normal
probability surface . If it is also of equal magnitude irre-
spective of the amount of the ability , scaling tasks by what
we may call the array method has many advantages over
the methods hitherto used .
APPENDIX III
ON THE FORM OF THE DISTRIBUTION OF INTELLECT IN THE
SIXTH GRADE , THE TWELFTH GRADE , AND
AMONG COLLEGE FRESHMEN
We have shown elsewhere¹ that the form of distribution
of intellect in the ninth grade conforms closely to the so
called normal curve of error . This is the report of a series
of similar investigations into the manner in which intellect
varies in the sixth grade , the twelfth grade , and among col
lege freshmen .
It has been commonly assumed , whenever tests or tasks

have been scaled for difficulty in a group of given educa
tional status , that the form of the distribution of intellect
within such a group was truly represented by the normal
1 -x2
or Gaussian curve , whose equation is y = e202
σν2π
It
is of practical importance to test the validity of this
assumption if the scaling of tasks for difficulty is to be upon
a sound basis .
THE FORM OF DISTRIBUTION OF INTELLECT IN THE SIXTH GRADE

Our data for this grade consist of a series of frequency
tables of the scores made by sixth - grade pupils in many
different cities of the United States on eleven well - known
intelligence examinations . They are listed below with the
size of the population for each examination . The sources
of the data appear on page 522 .
From these data we have plotted the percentage dis
tributions for each test , from the mean as central tendency ,
1 E. L. Thorndike and E. O. Bregman , Journal of Educational Research ,
November , 1924 ," On the form of Distribution of Intellect in the Ninth
Grade . "
521
Test n
1. Otis Advanced Examination 5952¹
2. National Intelligence Examination , A 16682
3. Trabue Completion Exercise A 14543
4. Pintner Non -Language Examination 13774
5. Pressey Mental Survey ( Cross - out ) E 10575
6. Hagggerty Delta 2 9166
7. Army Examination A 7427
8. Myers Mental Measure 7248
9. Illinois Examination 588⁹
10 . I. E. R. Tests of Selective and Relational Thinking ,
Generalization and Organization 37910
11. Army Alpha 28111
1. Otis , A. S. , Manual of Directions .

Coxe , W. W. , Variation in General Intelligence , Jour . Ed . Re-
search , 1921 , vol . 4 , pp . 188-189 .
2. Data of the N. I. T. Committee supplied by Professor Guy M.
Whipple .
3. Trabue , M. R. , Completion Test Language Scales , p . 8 .
4. Unpublished data supplied by Professor Rudolph Pintner .
5. Pressey , S. L. , A Brief Group Scale of Intelligence for Use in
School Surveys , Jour . Ed . Psy . , vol . 11 , p . 96 , 1920 .
6. Madsen , J. N. , Intelligence as a Factor in School Progress ,
7. Memoirs -
School and Society , 1922 , vol . 15 , p . 285 .
The National Academy of Sciences , vol . 15 , Table
187 , p . 537 and Table 20 , p . 334 .
8. Layton , L. H. , Myers , G. C. , and Myers , C. E. , Group Testing
in Altoona , Pa . , School and Society , vol. 13 , p . 624 .
9. Data supplied by Professor W. S. Monroe , Univ . of Illinois .
10. Unpublished data of the Institute of Educational Research ,
Division of Psychology , Teachers College .
11. Data of the Kansas State Teachers College , Emporia , Kansas .
and in units of one - tenth of the standard deviation of each

distribution . The plots appear in Figures 76 to 86.
These individual curves are more or less irregular in
outline and of no very constant form . On the whole the
curves for the larger populations are the smoothest and
most regular .
We are not concerned , however , with the form of dis-
tribution based upon any single test or examination . The
form of any such single distribution , granting that the
sample was both representative and numerically adequate ,
might not reflect the true form of distribution of intellect
3D -20 -10 10 20 30
FIG . 76. Percentage distribution of sixth -grade scores in the Otis Advanced
Examination . n = 5952 .
SD -30 -20 10 10 20 30 40
FIG . 77. Percentage distribution of sixth -grade scores in the National Intel
ligence Examination , Form A. n = 1668 .
SD-> -30 -20 -10 10 20 30 40

FIG . 78. Percentage distribution of sixth -grade scores in the Trabue Comple
tion Exercise A. n = 1454 .
-20 10 10 20 40.
FIG . 79. Percentage distribution of sixth -grade scores in the Pintner Non-
Language Examination . n = 1377 .
SD- -30 -20 10 10 20 30 40

FIG . 80. Percentage distribution of sixth -grade scores in the Pressey Mental
Survey ( Cross Out ) E. n = 1057 .
-20 -10 10 20 30
FIG 81. Percentage distribution of sixth - grade scores in the Haggerty
Delta 2 Examination . n = 916 .
3.D 30 -20 -10 0 ΤΟ 20 30 40
FIG . 82. Percentage distribution of sixth -grade scores in the Army Examina-
tion A. n = 742 .
SD-- -30 -20 10 10 20 30 40

FIG . 83. Percentage distribution of sixth -grade scores in the Myers Mental
Measure . n = 724 .
S.D -30 -20 10 10 20 30 40

FIG . 84. Percentage distribution of sixth -grade scores in the Illinois Examina-
tion . n = 588 .
in this grade , either as the result of the error of measure-

ment in the individual scores , or through the effect of in-
equalities in the units of the tests .
In so far as inequalities in the units of the tests occur
purely by chance , however , inequalities in one direction in
one test will tend to be balanced by like inequalities in the
opposite direction in some other test .
SD -30 -20 10 10 20 30 40
FIG . 85. Percentage distribution of sixth -grade scores in the I.E.R. Tests of
Selective and Relational Thinking , Generalization and
Organization . n = 379 .
5D4 -30 20 10 10 20 30 40
FIG . 86. Percentage distribution of sixth -grade scores in the Army Alpha .
n = 281 .
THE DISTRIBUTION OF INTELLECT BY GRADES 527
We have therefore combined the eleven separate distri

butions , equal weight being attached to each , into a single
composite distribution , by averaging the frequencies for
each successive one - tenth sigma , and plotting the resulting
curve .
It is shown in Figure 87. For purposes of comparison ,

the theoretical Gaussian curve has been dotted in .
SD + 30 -20 10 10 20 30 50
FIG . 87. Composite curve for the sixth grade ,
based upon eleven single curves .
The broken line indicates the theoretical normal curve .
The two conform closely . The fit of the observed curve
son's Goodness of Fit Method , 2 P.999999 . -

to the theoretical has been numerically determined by Pear
In this grade , however , it is possible that inequalities in

the units of the tests , although present , do not occur by
chance . The sixth grade approximates the middle region .
of the range of ability for which these tests have been de
vised . The normal curve bears an excellent reputation
in psychological literature . One might conjecture with
some show of reason , therefore , that in the construction of
these tests there has been a more or less general and con
scious effort to adjust the units of the tests so as to dis
tribute pupils according to the normal curve , and that since
the sixth grade approximates the mean of the range of
ability for which the tests have been generally devised , such
2 For the method see Tables for Statisticians and Biometricians , edited by
Karl Pearson , Intro . pp . XXXI - XXXIII and Table XII .
deliberate inequalities would probably be most effective in

and near the sixth grade .
Figure 87 would not then bear evidence as to the in
trinsic nature of the distribution of intellect in grade six ,
but possibly in the psychologists who are responsible for
the tests only to the skill in producing a given form of the
tests . We should expect such skill to have its limits , how
ever . It seems highly improbable that it would extend to
having the same inequalities produce a spurious symmetry
not only in the neighborhood of the sixth grade , but through
out the whole range of ability for which the tests are in
tended . In Grade 9 , for instance , it seems reasonable to
expect that the potency of any such hypothetical inequali
ties to produce symmetry would have largely , if not com
pletely , vanished . If
we find , then , that the same tests that
display a normal curve in Grade 6 display the same form
of distribution in Grade 9 , we may justly conclude that
there are no concerted inequalities in the units of the tests
to which the symmetry observed may be attributed .
We have therefore plotted ninth- grade distributions³
3 The ninth - grade population and source for each test are as follows :
Test n Source
Otis Advanced Examination 3627 Otis , A.S. , Manual of Directions , '21
National Intel . Ex . A 494 Unpublished data of the I.E.R.
Trabue Completion Ex . A. 273 Trabue , M. R. , Completion test Language
Scales , p . 10
Pintner Non - Language Ex . 258 Unpublished data supplied by Prof. Ru
dolph Pintner
Pressey Mental Survey Pressey , S. L. , Jour . Ed . Psy . , vol . XI , p . 96
(Cross Out ) E 303
Haggerty Delta 2 2648 Compiled from unpublished data of the
I.E.R. and from records supplied by Dr.
1 W. J. Osburn , Wisconsin Dept. of Ed .
Army Examination a 805 Memoirs National Academy of Sciences ,
vol . 15 , Table 25 , p . 344 , Table , 187 , p .
537
Myers Mental Measure 311 Myers , C. E. and G. C. , Measuring Minds ,
P. 24
Illinois Examination 380 Unpublished data of the I.E.R.
I.E.R. Tests of Selective and Unpublished data of the I.E.R.
Relational Thinking , Gen
eralization and Organization 3214
Army Alpha 1721 Cobb , M. V. , J. Ed . Psy . , Nov. , '22 , Table
IV
for the same series of tests which constitute the sixth -grade
data , and have combined these into a composite curve in
the same manner as already described . This curve is
shown in Figure 88. It is as close a fit to the theoretical
curve as Figure 87 , the sixth - grade curve . P for Figure
88 is unity .
SD 30 -20 -10 20 30 40
FIG . 88. Composite curve for the ninth grade , based upon single curves for
the same eleven tests from which the sixth grade composite ( Fig . 87 ) was
derived . The broken line indicates the theoretical normal curve .
The symmetry and goodness of observed the dis

fit
in
tribution of intellect the sixth grade cannot therefore

in
be due consistent inequalities the units of the indi

to
in
vidual tests
.
One factor remains other than the real nature of ability

Grade which the symmetrical form of Figure
87
in
to
6
,
may still be due This the effect of the error of mea

is
.
surement the individual scores The effect of this may

in
be to produce spurious appearance symmetry and

of
a
normality For example all the pupils this grade

of in
if
"
"
,
.
were of absolutely equal ability measurement them with

,
fallible instruments would result in distribution of scores

a
resembling the normal curve of error and their real dif

if
;
ferences were such produce right triangle the

as
as
to
If
a
distribution the error might largely mask this these

,
children actually vary ability according the normal

to
in
curve of error , diminishing the errors of measurement

should not lessen the symmetry of the group . As a series
of measurements of a group grew less fallible , we should
expect that the original symmetry , if spurious , would be
accordingly diminished , but if intrinsic , would be main
tained .
We have attempted to discover the effect of reducing

the error of measurement upon the form of distribution
of 216 sixth - grade children . Each child had been tested
with six different intelligence examinations as follows :*
Test
1. Haggerty Delta 2
2. Kelley -Trabue Completion
3. Myers Mental Measure
4. Otis Advanced Examination
5. National Examination B
6. National Examination A
The sum of an individual's scores in two of these tests

should give a better measure of that individual's ability
than either test alone , and successive summation of the re
maining test scores should result in progressive improve
ment in the measures . The error of measurement pre
sumably grows less and less . If
such symmetry as plots
from the distribution of scores on the single tests show is
spurious and exists solely by virtue of the error of mea
surement in these single scores , then plots from the sum
mated scores , in which the error is progressively dimin
ished , should grow less and less symmetrical .
Such is not the case . Figures 89 to 94 are plots of the
216 scores in each of the six tests . The fit of each to the
theoretical normal distribution is given in Table 148 .
Figures 95 to 99 are plots of the distributions obtained
by successive addition of the test scores in the order in
which the tests are listed in Table 148 .
4 We are indebted to Dr. J. L. Stenquist for these unpublished data .
TABLE 148 .
THE CLOSENESS OF FIT OF SIX TEST SCORES , TAKEN SINGLY .
Test Р
Haggerty Delta 2 .570
Myers Mental Measure .272
Kelley Trabue Completion .738
Otis Advanced Examination .496
National B .543
National A .067
Score Fig
Fig 89 90
Score
Fig 91 Fig92
.
Score
Fig93 Fig 94
FIGS Distributions of sixth grade pupils various
in
to
89 94. 216 scores

-
'
.
examinations
.
FIG 89. Haggerty Delta

2
.
FIG 90 Myers Mental Measure

.
FIG 91. Kelley Trabue Completion

-
.
FIG 92. Otis Advanced Examination

.
FIG 93 National Int Ex B.

.
FIG 94 National Int Ex A.

.
.
.
The fit of each of these to the theoretical is given in

Table 149 .
There is in
these graphs and in these figures no such
consistent change , paralleling the change in the error of
measurement , as would justify us in attributing such sym-
metry as the group shows to the error of measurement .
Score-
Fig. 95 Fig 96
Score
Fie 97 Fig 98
,
Fig 99
FIGS . 95-99 . Distributions of 216 sixth - grade pupils ' combined scores .
FIG . 95. Haggerty Delta 2 and Myers Mental Measure combined .
FIG . 96. Haggerty Delta 2 , Myers Mental Measure and Kelley -Trabue Com-
pletion , combined .
FIG . 97. Haggerty Delta 2 , Myers Mental Measure , Kelley -Trabue Completion
and Otis Advanced Examination , combined .
FIG . 98. Haggerty Delta 2 , Myers Mental Measure , Kelley - Trabue Completion ,
Otis Advanced Examination , and National Intelligence
Examination B , combined .
FIG . 99. Haggerty Delta 2 , Myers Mental Measure , Kelley - Trabue Comple-
tion , Otis Advanced Examination , National Intelligence Examination
B , and National Intelligence Examination A , combined .
When Haggerty and Myers scores are combined the fit

is .376 as compared with an average of .421 for them taken
separately . When Haggerty , Myers and Kelley - Trabue
are combined the fit is .706 as compared with an average of
.526 for them taken separately . When Haggerty , Myers ,
Kelley -Trabue and Otis are combined the fit is .547 as com-
pared with an average of .519 for them taken separately .
When Haggerty , Myers , Kelley - Trabue , Otis and National
B are combined the compared with an average
fit
as
.748
is
of .524 for them taken separately When all six are com-
.
bined the fit .353 as compared with an average of .471 for
is
them taken separately

.
TABLE 149
.
THE CLOSENESS OF FIT OF SIX TEST SCORES TAKEN TWO OR MORE AT
A
TIME AND ARRANGED
.
Tests P
Haggerty and Myers .376
Haggerty Myers and Kelley Trabue .706
,
Haggerty Myers Kelley Trabue and Otis .547

,
,
Haggerty Myers Kelley Trabue Otis and National .748

B
,
,
,
Haggerty Myers Kelley Trabue Otis National and

B
,
,
-
National .353
A
We may conclude therefore that intellect Grade

in
if
6,
measured truly equal units varies general accordance

in
in
,
with the normal probability curve Intellectual tasks may

.
therefore be scaled for difficulty in this grade by this

hypothesis with close approximation the truth
to
,
THE FORM OF THE DISTRIBUTION OF INTELLECT IN THE

TWELFTH GRADE
For the twelfth grade we have series ten frequency
of
a
distributions representing the scores made by twelfth-

grade pupils in as many different intelligence examina-
tions
.
The examinations are listed below with the number of

,
pupils who took each examination and the sources from

which the data were obtained
.
Examination n Source
1. Terman Group Test of 4886 Manual of Directions , page 9. The
Mental Ability World Book Co. ( The table of per-
centile scores was converted into a
frequency table )
2. Brown University Psycho- 3333 The Intelligence of Seniors in the
logical Examination High Schools of Massachusetts .
Stephen S. Colvin and Andrew H.
MacPhail . Bulletin , 1924 , No. 9 .
Department of the Interior , Bureau
of Education . Page 14
3. I.E.R. Tests of Selective 26385 Unpublished data of the Institute of
and Relational Think- Educational Research
ing , Generalization and
Organization , Forms A
and B
4. Thorndike Intelligence 1527 A Mental Educational Survey . G. M.
Examination for High Ruch . University of Iowa Studies
School Graduates , Part in Education , Volume 2 , No. 5 ,
1, Forms D and N page 22
5. Army Alpha Examination 1387 The Limits Set to Educational
Achievement by Limited Intelli-
gence . Table VII . M. V. Cobb ,
Journal of Educational Psychology ,
November , 1922
6. Otis Group Intelligence 1226 Manual of Directions , page 60 .
Scale Advanced Exami- World Book Co.
nation
7. Strickland Test 1020 This is a two - hour examination . The
frequency distribution was supplied
by the kindness of Professor V. L.
Strickland , Kansas State Agricul
tural College , Manhattan , Kansas
8. Mentimeter Scale 874 Monthly Bulletin , Bureau of Educa-
tional Research, University of North
Carolina , July 21 , 1923
9. Miller Mental Ability 739 Manual of Directions , page 17.
Test World Book Co. The percentile
distribution was converted into a
frequency distribution
10. Haggerty Intelligence Ex- 668 From data supplied through the
amination , Delta 2 courtesy of Dr. W. J. Osburn ,
State Department of Education ,
Wisconsin
5 1666 for Form A , 972 for Form B. Plots were made for each form sepa-
rately and were then combined into a single distribution by averaging .
The frequency distributions of these tests have been

converted into percentage distributions , and combined into
one composite distribution in the manner already described
in Section I. The individual percentage distributions are
shown in Figures 100 to 109 .
3D -30 -20 -10 10 20 30 40

FIG . 100. Percentage distribution of twelfth - grade scores in the Terman Group
Test of Mental Ability . n =4886 .
SD- -30 -20 -10 10 20 30 40
FIG . 101. Percentage distribution of twelfth -grade scores in the Brown Uni-
versity Psychological Examination . n = 333 .
The composite twelfth - grade distribution appears in

Fig. 110. P for this equals .999911 .
Since it is hardly likely , in view of the evidence which
has already been presented in connection with the sixth
36
grade and the ninth grade and which will be presented in

the next section for college freshmen , that the normality
and symmetry of Figure 110 result from the error of mea-
surement , we may conclude that intellect in the twelfth
grade is also distributed in conformity to the Gaussian
curve .
฀SD- -30 -20 -10 10 20 30 40
FIG . 102. Percentage distribution of twelfth -grade scores in the I.E.R.

Tests of Selective and Relational Thinking , Generalization and Organization .
Total n = 2638 . ( This curve is an average of the separate curves for Forms
A and B. )
-30 -20 -10 10 20

FIG . 103. Percentage distribution of twelfth -grade scores in the Thorndike
I
Examination for High School Graduates , Part , Forms
D and N. N = 1527 .
O
SD- -30 -20 -10 10 20 30 40
FIG . 104. Percentage distribution of twelfth - grade scores in the Army Alpha
Examination . n = 1387 .
LSD → -30 -20 -10 O 10 20 30 40

FIG . 105. Percentage distribution of twelfth - grade scores in the Otis Group
Intelligence Scale , Advanced Examination . n = 1226 .
SD 30 20 -10 10 20 30 46
FIG . 106. Percentage distribution of twelfth - grade scores in the Strickland
Test . n = 1020 .
-30 20 -10 20 30 40
FIG . 107. Percentage distribution of twelfth -grade scores in the Mentimeter
Scale . n = 874 .
SD- -30 -20 -10 10 20 30 40

FIG . 108. Percentage distribution of twelfth -grade scores in the Miller Mental
Ability Test . n = 739 .
THE FORM OF THE DISTRIBUTION OF INTELLECT AMONG

COLLEGE FRESHMEN
The composite distribution for college freshmen is

based upon percentage distributions plotted from fre-
quency tables for the following examinations :"
6 Whenever two or more groups , whether for the same or different forms
of the examination , are listed under any examination , separate percentage dis-
tributions have been computed for each group so listed . These have then
1. Army Alpha 2545 University Students Intelligence Rat
ings According to the Army Alpha
Test . E. L. Noble and George F.
Arps . School and Society , Volume
11 , page 234 .
400 Intelligence Tests of Yale Freshmen .
J. E. Anderson . School and So
ciety , Volume 11 , page 419
SD 30 20 10 10 20 30 40
FIG . 109. Percentage distribution of twelfth - grade scores in the Haggerty
Intelligence Examination , Delta 2. n = 668 . The composite twelfth -grade
distribution appears in Figure 110. P for this curve equals .999911 .
S.D -30 -20 10 10 20 30 40

FIG . 110. Composite curve for the twelfth grade , derived from ten single
curves . The broken line indicates the theoretical normal curve .
been combined by averaging , equal weight being attached to each group , into
a single distribution for each examination . Thus no examination receives
greater weight than any other in the final composite distribution .
2. Brown University Psycho- 2118 Distribution of scores made by six

logical Examination freshmen classes at Brown Univer-
sity . Data supplied by the courtesy
of Dr. A. H. MacPhail
3. Army Examination A 701 Memoirs of the National Academy of
Sciences , Volume 15 , page 537 ,
Table 187 .
4. Iowa Comprehension Tests Data supplied by the courtesy of Dr.
B- 1 1046 G. M. Ruch
D- 1 1085
5. Minnesota Recognition Vo- 1208 Data supplied by the courtesy of Dr.

cabulary , A - 2 M. E. Haggerty and Dr. D. G.
Paterson
6. Morgan Mental Test 1250 Data supplied by the courtesy of Dr.
J. J.
B. Morgan
7. Princeton Examination , 623 Data supplied by the courtesy of Dr.
Series II C. C. Brigham
8. Smith College Entrance Distribution of scores made by four
Examination No. 1 371 freshmen classes at Smith College .
Examination No. 2 486 Data supplied by the courtesy of
Examination No. 3 604 Dr. D. C. Rogers
Examination No. 4 596
9. Thorndike Intelligence Data supplied by the courtesy of Dr.
Examination for High G. M. Ruch
School Graduates
I , Form
J and K
Part B 1085
Part I, Forms 1046

Part I
, Forms E and F 834
and Part II
, Form C
Total score Smith 525 Data supplied by the courtesy of Dr.
College freshmen Agnes L. Rogers
Total score , Columbia 356 Unpublished data of the I.E.R.
freshmen
Part VI , Form C 272 Data supplied by the courtesy of Dr.
R. M. Smith
Thorndike Intelligence
Examination for High
I , Forms E and J,
School Graduates
Part
Part II , Form C and 241 Unpublished data of the I.E.R.
Part III , Form AA
10. Thurstone Psychological 5495 A Cycle Omnibus Test for College
Examination , Test IV Freshmen . L. L. Thurstone . Jour-
Liberal Arts freshmen nal of Educational Research , 1921 ,
Volume 4 , Table 2
11. Yale Examination Data supplied by the courtesy of Dr.

Preliminary Form 647 J. E. Anderson
Forms 1 and 2 815
Forms 1 and 2 and 3a 829
or 3b
Forms 1, 2 and 4 820
The plots for each examination are shown in Figures

111 to 121 .
The composite obtained by averaging Figures 111 to
121 is presented in Figure 122. Its fit to the normal curve
is expressed by P = .999988 .
SD -20 -10 10 20 40
FIG . 111. Percentage distribution of college - freshmen scores in Army Alpha .
Composite of separate curves for two groups . Total n ==2945 .
10 10 20 30
FIG . 112. Percentage distribution of college -freshmen scores in Brown Uni-
versity Psychological Examination . n = 2118 .
SD 30 20 10 10 20 30 40
FIG . 113. Percentage distribution of college - freshmen scores in Army Exami-
nation A. n = 701 .
SD 30 20 10 10 20 30 40 50
FIG . 114. Percentage distribution of college - freshmen scores in Iowa Com-
prehension Tests . Composite of separate curves for Forms
B - 1 and D -1. Total n = 2131 .
3D- 30 -20 10 10 20 30 40
FIG . 115. Percentage distribution of college -freshmen scores in Minnesota
Recognition Vocabulary , A-2 . n = 1208 .
SD 30 20 -10 O 10 20 3.0 40
FIG . 116. Percentage distribution of college - freshmen scores in Morgan Men
tal Test . n = 1250 .
SD 30 20 -10 0 10 20
FIG . 117. Percentage distribution of college - freshmen scores in Princeton
Examination , Series II
. n = 623 .
SD 30 20 FO 10 20 30 40
FIG . 118. Percentage distribution of college - freshmen scores in Smith College
Entrance Examination . Composite of separate curves for Forms
No. 1 , 2 , 3 and 4. Total n = 2057 .
SD 30 -20 -10 10 20 30 40
FIG . 119. Percentage distribution of college - freshmen scores in Thorn
dike Examination for High School Graduates . Composite of separate curves
for seven groups , many different forms . ( See tabulation , page 540. Total
n = 4359.
SD 30 -20 -10 10 20 40
FIG . 120. Percentage distribution of college - freshmen scores in Thurstone
Psychological Examination , Test IV . n = 5495 .
SD 30 -20 10 10 20 30 40
FIG . 121. Percentage distribution of college -freshmen scores in Yale Exami
nation . Composite of separate curves for four groups , several forms .
(See tabulation , page 541. Total n = 3111 .
THE EFFECT OF THE ERROR OF MEASUREMENT
We have , for each of six groups of college freshmen ,

several series of test scores .
For four of these groups , which we have called Groups
1 , 2 , 3 and 4 , we have a series of five scores for each student .
Three of the five scores represent each student's record on

three major parts of one long examination , or on two parts
of one examination , and the total score on a second ex-
20 20 30
FIG . 122. Composite curve for college - freshmen , derived from eleven single
examination curves . The broken line indicates the
theoretical normal curve .
amination . The fourth and fifth scores represent succes-

sive summation of the first two , and all three single scores ,
respectively .
For the remaining two groups , which we have called .
Groups 5 and 6 , we have , for each student , three sets of
scores . The first two represent performance on two major
parts of a single examination , or on two separate examina-
tions . The third score is the sum of these two single
scores .
The examinations performed by each group , the size
of each group and the sources of the data are tabled below .
These data enable us to observe the effect , upon the
form of the distribution , of progressively diminishing the
error of measurement . If
the symmetry and normality
Examination Series Source
A
J
1. Thorndike Int . Exam . for H.S.G. , 1046
Part I, Form Data supplied by the cour
B Thorndike Int . Exam . for H.S.G. , tesy of Dr. G. M. Ruch
I
Part , Form K
C Iowa Comprehension Exam . B- 1
2. A Thorndike Int . Exam . for H.S.G. , 834
Part I, Form E
B Thorndike Int . Exam . for H.S.G. , Data supplied by the cour
Part I, Form F tesy of Dr. G. M. Ruch
C Thorndike Int . Exam . for H.S.G. ,
Part II , Form C
3. A Smith College Ent . Exam . Form 633

3 , Part 1
B Smith College Ent . Exam . Form Data supplied by the cour
3 , Part 2 tesy of Dr. D. C. Rogers
C Smith College Ent . Exam . Form
3, Part 3
4. A Princeton Examination Series V 629 Data supplied by the cour

B Princeton Examination Series VI tesy of Dr. C. C. Brigham
C Princeton Examination Series VII
5. A ThorndikeInt . Exam . for H.S.G. , 1085 Data supplied by the cour
I
Part , Form B tesy of Dr. G. M. Ruch
B Iowa Comprehension Exam . D – 1
6. A Minnesota Recognition Vocabulary 1208 Data supplied by the cour

A-2 tesy of Dr. M. E. Hag
B Minnesota Opposites and Comple- 1203 gerty and Dr. D. G. Pat
tion of Definitions erson
decrease with each successive combination of scores , they

are in so far due to the chance error . If they do not de
crease with more and more combination , the symmetry and
normality are in so far really characteristic of the group
itself .
We have examined the effect of squeezing out the error
in individual measurements , by successive combination of
test scores , upon the form of the distribution of each one.
of the six groups , and also upon the composite curves ob
tained by combining the similar series of distributions for
each group .
Plots of the series of single and summated scores for

each of the six groups are given in Figures 123 to 128 .
In each figure the observed frequency distribution is
given in solid outline . The broken line gives the frequen-
cies to be expected by the theory of the normal curve wher-
ever they differ from the frequencies of fact .
Seare A
Score- D
Score B
Storan
FIG . 123A . Group 1 : Distribution of 1046 Freshmen scores in Thorndike

I
Examination for High - School Graduates , Part , Form J.
FIG . 123B . Group 1 : Distribution of 1046 Freshmen scores in Thorndike
I
Examination for High - School Graduates , Part , Form K.
FIG . 123C . Group 1 : Distribution of 1046 Freshmen scores in Iowa Compre-
hension Examination B -1.
FIG . 123D . Group 1: of 1046 Freshmen scores in Thorndike

Distribution
for High -School Graduates , Part , Form I
J
Examination
plus Part , Form K.I
FIG . 123E . Group 1 : Distribution of 1046 Freshmen scores in Thorndike
Examination for High - School Graduates , Part , Form I J
plus Part I ,
Form K plus Iowa Comprehension Examination B - 1.
The fit of each observed distribution of scores , single

and summated , to each series of theoretical normal fre-
quencies has been numerically determined by the Pearson
Goodness of Fit Method . The P's are shown in Table 150 .
Examination of the figures and the table reveals no con-

sistent and progressive tendency in the individual groups
to lessened symmetry and normality with progressive re-
duction of the error .
In the composite curve the the normal becomes
fit
to
,
practically progressively better
,
.
12616-
SLOVE
→
E
C
FIG 124AGroup Distribution of 834 Freshmen scores Thorndike Ex-

in
2
:
.
amination for High School Graduates Part Form

E.
,
I,
-
FIG 124B Group Distribution 834 Freshmen scores Thorndike Ex-

of
in
2
:
.
amination for High School Graduates Part Form F.

,
I,
-
FIG 124C Group Distribution of 834 Freshmen scores Thorndike Ex-

in
II
2
:
.

C.
,
,
-
FIG 124D Group Distribution of 834 Freshmen scores Thorndike Ex-

in
2
:
.

E
,
I,
-
plus Part Form F.

I,
FIG 124E Group Distribution of 834 Freshmen scores in Thorndike Ex-

2
:
.
amination for High School Graduates Part Form plus

E
II
,
I,
-
Part Form plus Part Form

C.
F
I,
Our data give us two sets of composite distributions

,
one which all six groups are represented and which con-
in
sists of two curves for the two sets of single scores and
a a
,
third distribution of these single scores summated and

,
second set of distributions which only the four groups

in
are represented for which we have three sets of single

,
scores and the two sets of successively summated scores .

These distributions have been obtained in the same man
ner as all of our composite distributions , namely , by plot
ting percentage distributions for each frequency table , and
deriving the composite by averaging the percentage fre
12074
Score
→
Scovere
с
FIG . 125A . Group 3 : Distribution of 633 Freshmen scores in Smith Entrance

Examination , Form 3 , Part 1.
FIG . 125B . Group 3 : Distribution of 633 Freshmen scores in Smith Entrance
Examination , Form 3 , Part 2 .
FIG . 125C . Group 3 : Distribution of 633 Freshmen scores in Smith Entrance
Examination , Form 3 , Part 3 .
FIG . 125D . Group 3 : Distribution of 633 Freshmen Scores in Smith Entrance
Examination , Form 3 , Part 1 plus Form 3 , Part 2 .
FIG . 125E . Group 3 : Distribution of 633 Freshmen scores in Smith Entrance
Examination , Form 3 , Part 1 plus Part 2 plus Part 3.
quencies for each successive 1/100 from the mean , equal

weight being attached to each distribution .
Figures 129 , 130 and 131 present the composite distri
butions of the single and summated scores of all six groups ,
Figures 132 to 136 the composite curves of Groups 1 to 4 .
The broken lines in Figs . 129 to 136 indicate the " normal "
curve .
The fit of each composite curve to the theoretical nor-
mal curve is given in Table 151 .
SCOTC➡ D
Scove
+ B
Score
FIG . 126A . Group 4 : Distribution of 629 Freshmen scores in Princeton Ex-

amination , Series V.
FIG . 126B . Group 4 : Distribution of 629 Freshmen scores in Princeton Ex-
amination , Series VI .
FIG . 126C . Group 4 : Distribution of 629 Freshmen scores in Princeton Ex-
amination , Series VII .
FIG . 126D . Group 4 : Distribution of 629 Freshmen scores in Princeton Ex-
amination , Series V plus VI .
FIG . 126E . Group 4 : Distribution of 629 Freshmen scores in Princeton Ex-
amination , Series V plus VI plus VII .
The error of measurement is therefore not the potent

factor in determining the normal symmetrical form of dis-
tribution of intellect among college freshmen . We may
then conclude that this form of distribution describes the
actual variation of intellect in this group .
13
FIG . 127A . Group 5 : Distribution of 1,085 Freshmen scores in Thorndike
Examination for High -School Graduates , Part I , Form B.
FIG . 127B . Group 5 : Distribution of 1,085 Freshmen scores in Iowa Com-
prehensive -
Examination D 1.
FIG . 127C . Group 5 : Distribution of 1,085 Freshmen scores in Thorndike
Examination for High - School Graduates , Part I , Form B plus
Iowa Comprehensive Examination D 1. -
FIG . 128A . Group 6 : Distribution of 1,208 Freshmen scores in Minnesota
Vocabulary Examination .
FIG . 128B . Group 5 ; Distribution of 1,203 Freshmen scores in Minnesota
Examinations , Opposites and Definitions .
FIG . 128C . Group 6 : Distribution of 1,203 Freshmen scores in Minnesota
Examinations , Vocabulary plus Opposites and Definitions .
37
-
TABLE 150 .
GOODNESS OF FIT OF OBSErved DistribUTIONS GROUPS 1 TO 6—

TO NORMAL CURVE .
P. for dis . of P. for dis . of sums P. for dis . of sums

single scores of A and B scores of A , B and C scores
Series
Group 1 A .863808 .979015 .681535
B .340511
с .000000
Group 2 Α .705301 .108533 .965324

B .001299
0 .523111
Group 3 Α .019390 .598101 .279157

B .411783
C .000717
Group 4 A .028408 .000038 .003595

B .000000
C .006401
Group 5 Α .444183 .486321

B .521361
Group 6 A .974138 .261391

B .269385
TABLE 151 .
GOODNESS OF FIT OF COMPOSITE DISTRIBUTIONS TO NORMAL CURVE .
P. for composite P. for composite P. for composite

dis . of single dis . of sums of dis . of sums of A,
scores A and B scores B and C scores
Series
Groups 1 to 6 A .999981 1.
B .999765
Groups 1 to 4 .986097 .999818

1
.
В .999713
ABC
с .949824
SD -2 :0 -10 10 20 30 40
FIG . 129. Composite distribution of college - freshmen scores based on six
curves .. Series of A scores .
S.D.→ -3.0 -2.0 -10 10 20 30 40

FIG . 130. Composite distribution of college -freshmen scores based on six
curves . Series of B scores .
S.D. 30 -20 -10 10 20 3.0 40

FIG . 131. Composite distribution of college -freshmen scores based on six
curves . Sum of the A and B scores .
SD -30 -20 -10 10 20 3-0 40
FIG . 132. Composite distribution of college -freshmen scores based on four
curves . Series of A scores .
30 -20 -10 10 20 30 40
curves . Series of B scores .
SD -30 -20 10 20 30 40
curves . Series of C scores .
SD . -20 -10 10 20 3.0 40

FIG . 135. Composite distribution of college - freshmen scores based on four
curves . Sum of the A and B scores .
SD -30 20 -10 10 20 3.0 40

FIG . 136 . Composite distribution of college - freshmen scores based on four
curves . Sum of the A , B , and C scores .
APPENDIX IV .
THE HOMOGENEITY OF INTELLECT CAVD AT ALL LEVELS

OF DIFFICULTY
Our question is how far the ability required to succeed
with hard CAVD tasks differs from that required to suc-
ceed with easy CAVD tasks by being a greater amount of
the same kind of thing , and how far, on the contrary , it
is qualitatively different . If
we had composites at each
level of difficulty , made up of, say , a thousand single tasks
or enough to measure the ability at each level perfectly ,
and tested a random million of age twenty with them , how
nearly would the correlations between different levels of
difficulty approach 1.00 ; and how nearly would the remote-
ness of one level from another approach zero in its in-
fluence upon the correlations ?
We have to estimate the correlations for composites of
1,000 or more from the correlations for composites of 40
or less . This is done by the well - known attenuation for-
mula of Spearman . We have to estimate the correlations
in such an age population from the correlations in various
groups of more restricted range . This may be done by
the Pearson formulae for correction for range . We have
to estimate the effect of remoteness over the whole range
of difficulty from tasks like those of the 40 - Composite A
to tasks like those of the 40 - Composite Q by the effect of
remoteness of two or three steps , since none of our groups
was tested over the whole range of difficulty .
We present the facts from four groups . The first is
98 adult imbeciles ; the second is 121 candidates for college
entrance ; the third is 246 pupils of Grade 9 ; the fourth is
192 pupils of Grade 9 .
The 40 - Composites A , B, C , and D were divided each
into two 20 - composites , by taking elements 1 , 2 , 3 , 9 , and
556
THE HOMOGENEITY OF INTELLECT CAVD 557
10 from C , and D to make AI , BI , etc. , and by taking

A , V,
elements 4 , and 8 from C , A, V , and D to make AII ,
5 , 6, 7 ,
BII etc. The correlations between one and another of
,
these 20 - composites AI , AII , BI , BII , CI , CII , DI , and

DII in the case of 98 adults of mental age 30 months
to 60 months , are as shown in Table 152. The self - corre-
lations of AI with AII , BI with BII , CI with CII, and DI
with DII , are also shown there . The correlations between
two composites of infinite length , at various differences of
difficulty , may then by this determination be expected to
be as shown in Table 153 .
The average of the correlations of neighboring com-
posites is .94 ; that of the correlations of composites one
step removed , is .86 ( .84 by Pearson , .885 by Sheppard ) ;
that of the A to D correlations is .78 ( .71 by Pearson , and
.84 by Sheppard ) .
TABLE 152 .
SELF AND INTER -CORRELATIONS OF FOUR 40 -COMPOSITES OF CAVD , EACH

DIVIDED INTO TWO RANDOM HALVES ( AND I II
) . 98 IMBECILES ( P MEANS
PEARSON COEFFICIENT , SH MEANS SHEPPARD COEFFICIENT ) .
A II BI BII CI CII DI D II
P. Sh Р Sh P Sh P Sh P Sh P Sh P Sh
A I .86 .87 .78 .80 .78 .78 .67 .71 .73 .78 .62 .76 .52 .70
A II
I
.78 .76 .78 .81 .68 .71 .70 .74 .58 .71 .49 .67
B .81 .74 .74 .83 .80 .81 .63 .76 .59 .72
B II
CI
.76 .78 .80 .80 .70 .70 .62 .77
.81 .91 .65 .67 .64 .79
C II .73 .70 .74 .80
D I .69 .83
D II
In
these and in similar correlations between levels at
different degrees of remoteness , it should be kept in mind
that the range is very restricted , and that if all Americans
of the same chronological ages as these 98 imbeciles had
been measured by A , B , C and D , the correlations would
have been very much higher . The o of the group of 98
adults was about 8 mental months . That of the group of
"all Americans of comparable chronological ages " would

be at least four times as great.¹
In a group with a variability four times that of the
group of 98 imbeciles , the correlations corresponding to
the .94 , .86 , and .78 would be approximately .992 , .99 ,
and .98 .
TABLE 153 .
THE INTER -CORRELATIONS OF FOUR CAVD COMPOSITE TASKS LIKE A, B , C,

AND D IN CONSTITUTION AND DIFFICULTY , BUT EACH CONSISTING OF
AN INFINITELY LARGE NUMBER OF SINGLE TASKS . THE INTER-
CORRELATIONS OF TABLE 152 CORRECTED FOR ATTENUATION .
B с D
Р Sh P Sh Р Sh
.93 .982 .83 .83 .71 .84
.96 .98 .85 .94
ABC
с .92 .85
TABLE 154
.
SELF- AND INTER CORRELATIONS OF FOUR 40 COMPOSITES OF CAVD EACH
,
-
DIVIDED INTO TWO RANDOM 121 HIGH SCHOOL

II
HALVES AND
I
(
GRADUATES MEANS PEARSON SH MEANS SHEPPARD

P
.
)
(
NI NII ΟΙ OII PI PII QI QII

P Sh P Sh P Sh P Sh P Sh P Sh P Sh
NI .57 .46 .64 .57 .68 .62 .55 .57 .61 .65 .54 .59 .47 .37
NII .64 .62 .61 .62 .58 .62 .63 .65 .48 .37 .52 .41
ΟΙ .66 .71 .65 .68 .60 .63 .60 .65 .43 .19
OII .60 .68 .70 .75 .55 .62 .53 .39
ΡΙ .62 .75 .54 .48 .50 .48
PII .46 .39 .58 .53
QI .49 .47
QII
group of 121 candidates for entrance

Similar facts for
a
college
measured with composites formed by divid-
20
to
ing and two 15 composites formed by dividing

N
P
O
,
are shown in Tables 154 and 155. Here the average

Q
,
The of the random sample the draft from camps

of
in
653 men was

σ
1
34 mental months Memoirs 391.

.p
,
[
.
]
correlations are 1.022 ( .995 for P and 1.05½ for Sh ) ,

1.002 ( .96 for P and 1.05 for Sh ) , and 1.14 ( .95 for P and
1.34 for Sh ) in order of remoteness .
TABLE 155 .
THE INTER -CORRELATIONS OF FOUR CAVD COMPOSITE TASKS LIKE N, O, P

AND Q , IN CONSTRUCTION AND DIFFICULTY , BUT EACH CONSISTING
OF AN INFINITELY LARGE NUMBER OF SINGLE TASKS .
121 HIGH SCHOOL GRADUATES .
Q
P Sh Р Sh Р Sh
1.05 1.062 .99 1.05 .95 1.34
.992 .94 .92 1.05

NOA
Р .94 1.15
The 40 composites and showed the inter-

M
K
J
I
,
L,
,
,
,
-
correlations of Table 156 in the case of 246 pupils in

Grade The self correlations when we divide each 40-
9.
composite into two 20 composites made 5C 5A 5V and

of
,
,
-
5D taken at random were

:
P Sh
.53 .56
JI
.63 .66
K .4812 .59
L .75 .75
M .59 .53
The self correlations of the respective 40 composites

-
,
-
each with another 40 composite the same difficulty are

= of
-
2r₁
then by Spearman's formula as shown below
+ ri
r₂
.
1
((
P Sh
.70 .73
JI
.77 .80
K .65 .74
L .86 .86
28588
M .68 .69
The correlations between two composites of infinite

length , at various differences of difficulty , derived from
Table 156 by using TAB =
TA1B1
are as shown in
VIA1A2 TB1B2
Table 157.
TABLE 156 .
THE ' RAW ' INTER -CORRELATIONS OF FIVE CAVD 40 -COMPOSITES IN 246
PUPILS OF GRADE 9.
Sh
J P Sh
K
P Sh
L
P Sh
M
P
J
.59 .53 .62 .49
I
.70 .66 .72 .66 .72

K .67 .66 .69
TUELN
.63 .67
M
TABLE 157
.
THE INTER CORRELATIONS OF FIVE CAVD COMPOSITE TASKS LIKE THOSE OF

,
-
TABLE 156 IN CONSTITUTION AND DIFFICULTY BUT EACH CONSIST-

,
ING OF AN INFINITELY LARGE NUMBER OF SINGLE TASKS

.
Joo K∞ Loo Moo

Sh Sh P Sh Sh
P
P
P
I∞ .77 .72 .78 .69

J∞ .91 .93 .87 .81 .97
Koo
.84 .88 .962
Loo .82 .87
The average of the correlations neighboring compo-

of
sites this group .85 the average composites one

in
of
is
step removed that composites two steps removed

of
.84
is
.872 the correlation between the two which are three

is
steps removed .69

is
The composites and showed the inter corre-

M
K
N
in L
-
,
,
,
lations of Table 158 the case of 192 pupils Grade

in
9
.
The self correlations using 20 with 20 are order .67

in
,
,
-
and .58 using Sheppard's formula and .58 .80

57
.76
,
.64 and .62 using Pearson's formula

,
.
The correlation between a 40 - composite and another 40

2r1
composite of the same difficulty are then ( by r₂ = 1
+r
as shown below :
Sh P
K .80 .73
L .86 .89
M .73 .78
N .73 .77
TABLE 158 .
THE " RAW ' INTER -CORRELATIONS OF FOUR CAVD COMPOSITES IN 192 PUPILS
OF GRADE 9.
L M N
Sh P Sh P Sh P
K .47 .61 .59 .69 .59 .61
L .77 .73 .50 .57
M .68 .71
TABLE 159 .
THE INTER -CORRELATIONS OF FOUR CAVD COMPOSITES LIKE THOSE OF TABLE

158 IN CONSTITUTION AND DIFFICULTY , BUT EACH CONSISTING OF AN
INFINITELY LARGE NUMBER OF SINGLE TASKS .
Loo Moo Noo

Sh P Sh P Sh P
K∞ .56 .75 .77 .91 .73 .81

Loo .97 .88 .63 .69
M∞ .93 .92
The correlations for composites of infinite length , at

the various differences of difficulty , derived from Table
in Table
TA1B1
by using are shown 159 .
158 TAB =
VTA1A2 TB1B2
For this group of of the correlations of

192 , the average
neighboring composites is .832 ; that of the composites one
step removed is .75 ; that of the two which are two steps
removed is .77.
Table 160 summarizes the facts for all four groups .

We correct for restricted range in the groups of high
school graduates and pupils in Grade 9 , by taking the varia
bility of age 20 to be twice that of high school graduates
or of pupils in Grade 9. The variability of the literates of
the draft who took Alpha was 13 times that of pupils in
Grade 9 or of pupils in Grade 12 tested with Alpha . If
the illiterates at one extreme and the officers at the other
had been included , the variability would have been greater .
TABLE 160 .
SUMMARY OF INTER -CORRELATIONS CORRECTED FOR ATTENUATION .
Neighboring Composites Composites Composites

Composites One Step Two Steps Three Steps
Removed Removed Removed
98 imbeciles .94 .86 .78

121 H. S. graduates 1.0212 1.002 1.14
246 9th grade. .85 .84 .872 .69
192 9th grade.. .852 .75 .77
Average .91 .862 .89 .852
CORRECTED FOR RESTRICTED RANGE

98 imbeciles .992 .99 .98
Average of the other
groups .972 .96 .983
Weighted average of
all four groups .98 .964 .984
The variability of 63,647 enlisted men in four camps in

Army Examination a was almost 14 times that of pupils
in Grade 9 , and almost 2 times that of college freshmen .
Here also the inclusion of the illiterates and officers would
have raised the variability of the adult group .
We have in these four determinations , taken together ,
a proof that the effect of remoteness in difficulty upon the
inter - correlations of various CAVD composites is small ,
and a rough measure of how small it is .
2 When .69 is averaged with the figures in preceding column .
3.96 ifthe .69 is used in the average .
4.96 % if the .69 is used .
The change as we pass from neighboring composites to
-
composites one step removed in difficulty is — .08 , — .01 ,
-.08 , and -.02 .02 ,, with an average of -.05 which has a
σt - o of ± .034 . The change as we pass from neighboring
composites to composites two steps removed is - - .16 ,
+.02 , — .06 , and +.111 , with an average of -.02 ,
which has a ot - o of .102 . If
we count the three - step
case where the change is.16 ( .85 to .69 ) in with the two
step cases , we have an average of -.05 with a σt - o of
± .107 . Ifthe effect of remoteness were large , twice the
amount of remoteness would have a greater relative effect ,
and the effects would all be larger relative to their unre
- -
by -.08 , -.01 .01 , -
liabilities . One step of remoteness changes the correlation
.08 , — .02 , — .08 , +.131 , + .031 , + .02 ,
and ― .18 . The average of these , -.03 , has a σt- ฀ of
± .086 . So there is a probability of 36 in a hundred that
remoteness raises the correlations , and a probability of 64
in 100 that it lowers them .
When the correlations are corrected for the restriction
of range , the correlations are around .97 and the average
drop for one step of remoteness is .001 .
It is hard to state in any concise fashion how much of
a difference in difficulty this obtained drop of .03 corre
sponds to . We may best simply list the changes in the
percent of successes to which the facts which it averages
correspond. They are :
In the imbeciles , from 48.3 to 12.8
66
12.8 to 6.0
In the college graduates , from 48.1 to 27.5
66
27.5 to 3.7
In the 246 of Grade 9, from 89.4 to 61.4
66
61.4 to 32.9
66
32.9 to 5.3
In the 192 of Grade 9, from 16.3 to 7.2
66
7.2 to 1.1
It is somewhat strange that since the drop is so small ,

the correlation between neighboring composites should not
be around 97 and all the obtained correlations nearer to

unity than they are . This may be just a matter of chance
due to the small populations . Or there may be factors of
carelessness and lack of zeal on the part of some of the
subjects , and even some undetected cheating , which the
correction for attenuation may not properly allow for .
This and other imperfections in the determinations do
not , however ,impair the essential result that the correla
tions between composites far apart in difficulty are little ,
and possibly not at all , lower than the correlations between
neighboring composites . The series of CAVD composites
I
from one so easy as the to one so hard as the Q do mea
sure much the same sort of thing . We do have a right to
call it by one name and to measure increases in it by the
series of cardinal numbers .
APPENDIX V.
THE ADEQUACY OF TASKS OF ANY ONE LEVEL OF DIFFICULTY

AS A MEASURE OF ALL OF INTELLECT CAVD
We wish to know how nearly certain CAVD composite

tasks would measure all of intellect CAVD and nothing but
intellect CAVD , if the number of elements in the tasks were
extensive enough to make its own self - correlation 1.00 .
The answer is given by the correlation corrected for atten
uation between any CAVD composite task and any random
sampling of all of CAVD . It would , however , be extremely
difficult to obtain facilities to test any large group with
samples from CAVD at all levels of difficulty . College stu
dents , for example , would properly rebel at being given a
long list of absurdly easy tasks . It would discourage chil
dren in Grade 5 to be asked hundreds of questions none of
which they could answer . It is not necessary to sample
all of CAVD in the case of any one individual . If our
sampling goes down far enough to get near his level of one
hundred percent of successes and up far enough to get
near his level of zero percent of successes save by chance ,
our purpose will be served . The correlation between num
ber right in such a sampling and number right in a com
plete sampling will be almost perfect .
Our data are for such samplings . For example , a group
of 246 pupils in Grade 9 were measured in some 200 CAVD
tasks , sampling CAVD from level I to level K. This sam
pling was cut into two halves at random, save that each
half had the same number of tasks at each level of diffi
culty and the same number of C , A, V , and D. The corre
lation of these two halves was .826 by Sheppard's and .855
by Pearson's formula . The correlation of the score in the
200 with the score in another 200 may then be taken as .90
or .92 according to the formula used .
565
1506
.
TABLE 161
THE CORRELATIONS BETWEEN THE
-
NUMBER OF SINGLE TASKS RESPONDED TO CORRECTLY IN VARIOUS 40 COMPOSITES AND THE
A
NUMBER OF TASKS RESPONDED TO CORRECTLY IN LONG SERIES OF CAVD TASKS RANGING FROM TASKS VERY EASY
FOR THE
P
(
GROUP IN QUESTION
.
TO TASKS
=
VERY HARD FOR THE GROUP IN QUESTION GROUPS 91 AND 9II
=
;
, : ).
PEARSON COEFFICIENT SH SHEPPARD COEFFICIENT
Raw Self- Self- Corrected

Correlation Correlation Correlation Correlation
with Long of of with Long
-
Series 40 Composite Series
Long Series
THE MEASUREMENT
-
Group 40 Composite Sh Р Sh Р Sh Р Sh P
246 of Grade .72 .73 .70 .90 .92 .89
IJ
246 of Grade .89 .84 .80 .77 1.05 1.00
246 of Grade K .82 .81 .74 .65 1.00 1.05
246 of Grade L .93 .87 .86 .86 1.05 .98
9 9 9 9 9
246 of Grade M .83 .79 .69 .68 1.05 1.00
192 of Grade К .77 .82 .80 .73 .95 .93 .88 .99
192 of Grade .88 .86 .86 .89 1.03 .95
OF INTELLIGENCE
192 of Grade .88 .91 .73 .78 1.06 1.07
KLMN
9 9 9 9
192 of Grade .77 .82 .73 .77 .92 .97
THE ADEQUACY OF A SINGLE LEVEL 567
The raw correlation between the number right in the

40 - composite I
and the number right in the entire 200 is .72.
The self - correlation of the former is .73 . The correlation
corrected for attenuation is .89 .
The raw correlation between the number right in the
40 - composite J
and the number right in the entire 200 is .89
by the Sheppard formula and .84 by the Pearson . The self
J
correlation of is .80 by the Sheppard and .77 by the Pear
son . The correlation corrected for attenuation is 1.05 by
the Sheppard and 1.00 by the Pearson .
These and similar facts for tasks of various degrees of
:
difficulty are shown in Tables 161 and 162 , which report the
results from this group of 246 and from three other groups
as follows
192 other students in Grade 9 were tested with a some
what similar long series , including the tasks of the 40 - Com
posites K , L , M and N. 121 candidates for college entrance
were measured by a series of 240 completions , 56 arithmetic
tasks , 50 vocabulary tasks and 41 paragraph - reading tasks ,
ranging from such as almost all could do to such as hardly
any could do . A summated total score for the number
right was computed . The tasks included the 40 - Composites
N , O , P and Q. 240 first -year students in a Law School ,
all college graduates , were measured by a series of 53 com
pletion tasks , 56 arithmetic tasks , 100 vocabulary tasks and
41 - paragraph - reading tasks , ranging from such as almost
all could do to such as hardly any could do . A summated
score for CAVD giving equal weight to C , A , V and D was
computed . The tasks included the 40 - Composites N , O , P
and Q.
The self - correlations of the 40 - composites for the group
of 121high - school graduates are the averages from deter
2r20,
20
minations , first by 1+ and second by .03 + 140 , using
T20, 20
only nearest neighbors among the composites . Those for
the group of 240 college graduates are determined by the
second method .
38
The average of the seventeen corrected coefficients of

Table 161 is 1.00 with a mean square error of ± .058 .
The average of the eight corrected coefficients of Table
162 is 1.00 with a mean square error of ± .034 .
These results are corroborated by results from a group
of 100 university students , from the 240 college graduates
of Table 162 but using different 40 - composites , and from a
group of 147 pupils in Grade 51 .
TABLE 162 .
THE CORRELATIONS BETWEEN THE NUMBER OF SINGLE TASKS RESPONDED TO

CORRECTLY IN VARIOUS 40 -COMPOSITES AND THE NUMBER OF TASKS RE
SPONDED TO CORRECTLY IN A LONG SERIES OF CAVD TASKS RANGING
FROM TASKS VERY EASY FOR THE GROUP IN QUESTION TO TASKS
VERY HARD FOR THE GROUP IN QUESTION : GROUPS 13 AND 17 .
Raw Cor Self Cor Self Cor Corrected

Group 40 relation relation relation Correlation
Compo with Long of 40 of Long with Long
site Series Composite Series Series
121 high -school graduates N .80 .6912 .90 1.01
121 high -school graduates O .89 .782 .90 1.00
121 high -school graduates P .92 .782 .90 1.03
121 high -school graduates Q .77 .69 .90 .98
240 college graduates .85 .86 .95 .94
240 college graduates .92 .862 .95 1.02

NOP

O
฀
One hundred students of education graduates of col

,
leges or normal schools were measured by 40 sentence

,
completion tasks 46 arithmetic tasks 150 vocabulary tasks

,
24 paragraph reading tasks 170 information tasks 180

,
,
-
tasks involving information plus more or less general

sagacity picture completion tasks pictorial analogy
40 20
32
,
,
-
tasks and geometrical relations tasks general sum

A
.
mated score was compiled by combining the results

of
S₂
,
the 40 150 and 24 tasks so as give reasonable

to
V
D
C
a
,
weight each sort task Two composites and

to
of
75
C O
(
.
Q961 each consisting single tasks

26
of
10 10
+
V
(
),
+6 the scores being multiplied by approxi

of
D
1
,
)
mately equal difficulty within each composite and differing

between composites so that 77 percent of the group had 15
or more right in the O 75 composite while 21 percent had
15 or more right in the Q 963 composite , showed correla-
tions of .91 and .78 with the total summation score and a
correlation of .66 one with the other .
The self - correlation of the total summation score is
almost certainly not over .96 and the average self - correla-
tion of the 40 - composites is almost certainly not over
.66.10 or .76 . Using .96 and .75 , the corrected correla-
tion between one level and the total score in all levels in
.91.781
this group is 2 or a little under 1.00 .
V.96 X .76
The 240 college graduates were measured by three com-
posite tasks slightly different from N, O and Q , which we
shall denote by 1 , 2 and 3. Calling the long total score S1 ,
the correlations are :
=.73
= .63
T112₁
T1131
= .73
= .85
T2131
r1181
T2181.91
T3181
==.86
The self - correlation of 1 , 2 and 3 may be set as about
.03 higher than the correlations between 1 and 2 and be-
tween 2 and 3 , or at .76.¹
We do not know directly what the self - correlation of
the s , score is , but it can hardly be higher than .95 . Using
.76 and .95 , the corrected correlations between score for a
composite of tasks at one degree of difficulty and score in
the total CAVD series are 1.00 , 1.07 and 1.01 , averaging
1.03 with a mean square error of ± .031 .
One hundred and forty - four pupils at the very begin-
ning of Grade 6 were tested with a fairly long series of
1 In the case of the 240 college graduates the correlation between two
40 composites of about equal difficulty was .732 .
CAVD , with time to do all that they could do . A summa

tion score was given with approximately equal weight to
C , A, V and D. Included in the series were the 40 tasks
of the composite I
and the 40 of the composite J. The
raw correlation between the number right in level I and
J
the summation score was .91 . The raw correlation between
the number right in level and the summation score was
.86 . The raw correlation between the number right in com
posite I
and the number right in composite was .75 . J
We do not know the self - correlations in this group di
rectly , but that for a 40 - composite will be near .78 and that
for the total score will be near .95 . The corrected coeffi
cients will thus here also average close to unity ( using .78
and .95 , they are 1.00 and 1.06 ) .
We may approach the question of whether one of our
40 - compositeCAVD tasks measures ( except for the chance
error due to its having only 40 tasks ) all of intellect CAVD
and nothing but intellect CAVD by another method . If it
does , the average raw correlation r₁₁ should in a group of
wide range in intellect be little , if any ,
less than Vrtit2 .
We have found r₁₁₁₁ in of the range of a school
groups
grade to be , according to the group and the composite , .72 ,
.77 , .781 , .791 , .79½ , .80 , .81 , .81½ , .85 , .85 , .86 , .86 , .86 , .861 ,
.87 , .89 , .89 , .891 , .90 , .91 , .91 , .91 , .92 and .92 . The median
is .86 ; the average is .853 .
The self - correlation rt2 , estimated by the Spearman
formula from the two halves , or by taking the correlation
with a neighboring composite +.03 , is , for the same com
posites in the same groups , 3 .68½ , .69 , .69 , .69½ , .69½ , .71½ , .72 ,
.75, .75 , .75 , 76 , .76 , .76 , .76½ , .78 , .78 , .781 , .781 , .791 , .791 ,
.86 , .86 , .86 and .87 . The median is .76 ; the average is
.765 .
2 Where the correlation is determined by both Sheppard and Pearson for

mulae we use the average of the two results .
3 As before , when both Sheppard and Pearson coefficients were obtained ,
the average of the two is used .
Within a grade range then rt is .857 and rt1t2 is

.857
.761 . It is .873
of Vrtit2 or .98 Vrtit2 . If the range is
increased rt will approach still nearer to Vrt1t2 .
,
We may then safely conclude that the ability measured

by a sufficiently extensive composite of CAVD tasks at a
level of difficulty which is appropriate for the group in the
sense that the percent of successes in the group will be be-
tween 10 and 90 , is substantially identical with the ability
measured in that group by the total CAVD series . Our
CAVD 40 - composite tasks measure intellect CAVD as far
as their self - correlations permit .
The facts presented here concerning the correlations
between 40 - composites and long series which approximate
to total CAVD the conclusion in Ap-
series corroborate
pendix IV from the correlations between composites that
intellect CAVD is to a high degree homogeneous , the
higher levels requiring much the same ability as the lower ,
but more of it .
There is good reason to believe that , if we had taken
intellect GOPI , composed of tasks in observing geometrical
relations , giving opposites , completing pictures and an-
swering questions requiring information such as intellec-
tual people acquire , and carried out the same sorts of in-
vestigations as we have carried out with intellect CAVD ,
we should have found that intellect GOPI also was nearly
or quite homogeneous at all levels of difficulty , and that a
task composed of equal parts of G , O , P and all of ap- I
proximately the same difficulty, measured all of intellect
GOPI and nothing but intellect GOPI , so far as its own
self - correlation permitted . The same would also probably
hold for any other selection of tasks in thinking correctly
with ideas and symbols . Nothing was done in our selec-
tion of CAVD to favor homogeneity . There is no reason
to believe that CAVD is any more homogeneous at differ-
ent levels of difficulty than any other selection of tasks
for intellect .
APPENDIX VI
THE ESTIMATED FORM OF DISTRIBUTION OF VARIOUS GROUPS
We do not know exactly what the form of distribution

of intellect in the 180 adult imbeciles is . They were se-
lected to include the individuals 16 years old or over in two
institutions who were from 2 to 5 years of mental age by
the Stanford Binet ; and on this basis we should expect the
5.0 -40 -3.6

FIG . 137. A section of the probability surface from -3.6 to - 5.0 .
-6.0 -5.0
FIG . 138. A section of the probability surface from -5.0 to -6.0 .
572
THE FORM OF DISTRIBUTION OF INTELLECT 573
distribution to be a small segment of the low end of the

distribution of all persons 16 years old or over , but modi-
fied by the error of measurement . Empirically the distri-
bution of the scores in the Stanford Binet is as shown in
FIG . 139. The result of the application of an error of measurement to the

group represented in Fig . 138 .
Low High
FIG . 140. The distribution of 180 adult imbeciles in Stanford Binet Mental
Age .
Fig. 140. By the summation of credits in the 240 CAVDIO

tasks , the distribution was that of Fig . 141 .
In view of these facts , we have chosen as the probable
form of distribution of altitude of intellect in these 180
individuals , if they were measured in equal units , the form
shown in Fig . 142.¹ According to it the σ values of the

difficulty of the CAVD composites A, B , C , and D are re-
spectively , 1.68σA , + .05σв , + 1.130c and + 1.83σD .
If
the low end of the distribution were a fraction of the
probability surface undisturbed by the operation of any
large factors , such as diseases , accidents at birth , and the
High
Low
FIG . 141. The distribution of 180 adult imbeciles in a summation of credits
I
in CAVDIO ( = Inf ., O = Opp .)
FIG . 142. The probable form of distribution of altitude of intellect in the

group of 180 adult imbeciles .
like , a section of it would be like that shown ( from -3.6σ

to -5.00 ) in Fig . 137 , or from — 5.00 to -6.00 in Fig .
138.2 The application of an error to such a surface would
alter it to something like Fig . 139 , by blurring its distinc-
tions and providing it with tails at both ends .
1 We make no claim that this is the best guess at the form of distribution
that could be made . On the contrary , we could ourselves improve it by ( a )
giving other tests to this group , ( b ) by making a census of records of the
mental ages of the inmates ( 16 years old or older ) in institutions for the
feeble -minded , and in other ways . We simply have not the time to do so . It
is not a matter of much importance except in the case of the estimate of the
difficulty of Composite D. In that case the exact determination of the upper
tail of the distribution would be very desirable . However , the estimate for
Composite D would be hazardous in any case because of the unreliability of the
00.6 as the percent of successes .
2 The ordinate scale of Fig . 138 is magnified 100 times that of Fig . 137.
We know very little , however , about the forces operat-

ing to produce individuals so dull as these ; or the form
which the extreme low end of the distribution of human
intellect takes .
-4.2 -3.7
FIG . 143. A section of the normal surface of frequency from -4.2 to 3.7 .
We do not know exactly the form of distribution of the

100 feeble - minded of Stanford mental age 6 yr . 0 mo . to
6 yr . 11 mo . If they were a random sampling of all per-
sons sixteen years old or older of mental age 6 , the dis-
tribution would be approximately that of a segment of the
surface
37
-
of frequency for adult intellect from an I.Q. of
to an I.Q. of 44.4 , or from -6.250 to - 5.560 , ifwe
use Terman's estimate of the variability of intelligence
quotients [ '16 p . 78 ] . However , the variability of the in-
tellects of persons chronologically 16 or over is probably
much greater than that given by Terman's figures for chil-
-
dren, at least from the mode toward the low end ; and from
4.20 to -3.7o seems a more probable status by random
selection .
As has been stated , we do not know the exact form of the

surface of frequency for adult intellect, at its low extreme .
It almost certainly is continuous from M.A. 6-11 , to M.A.
6-0 , and diminishing somewhat in area per unit of abscissa .
shown in
in Fig . 138.
Fig.
137. That from -
The low extreme of Form A from -5.00 to -3.60 was
6.0 to 5.0σ was shown
Not much weight should be attached to any
estimate from theory of the amount of the diminution in
frequency as we go to very dull levels of the total adult
which are acting to create these levels . Using -

population , since we know very little about the causes
4.20 to
-3.70 of Form A we should have the distribution shown
,
in Fig . 143 .
However , the selection for commitment is not random ,
the duller ones being more often committed than the
brighter . So the pitch of the curve would be expected to
be less sharp than that of the general adult population ,
whatever that may be .
Turning to the actual measurements , the form of dis-
tribution of the 100 imbeciles of mental age 6 , if we as-
sume that one month of mental age from 6 to 7 by that scale
equals any other month , and that the Stanford Binet mea-
sures intellect perfectly , is that shown in Fig . 144 .
We do not know what the values of these mental months
of the Stanford Binet are in truly equal units , as there have
never been any measurements of grade or age groups by
the Stanford Binet which are large enough to enable us
to apply the procedure which we have used with the Na-
tional , Otis , Terman group and other examinations . In

all probability , the differences in the true values are not
great .
We know that the Stanford Binet does not measure in
tellect perfectly ; and we know roughly the amount of the
error of a single determination of the Binet as a measure
ment of the abilities that would be measured by a long
series of examinations of the very same sort as the Binet .
52 56 60 64 68 72 76 80 84 88 92 96 100 104
FIG . 144. The form of distribution of Group im . 6 in Stanford -Binet Mental
Age .
This is a mean ot - o of about 6 mental months , at or near

mental age 10. It probably is less around mental age 6 ,
but it will still be large in comparison with the range of
12 months in the measurements themselves . Taking it
as 4 mental months , the true intellects of the 100 would
range from about 5 yr . 0 mo . of mental age to about 8 years
of mental age . However , a long series of examinations of
the very same sort as the Binet would probably not mea
sure all of intellect . So that an allowance for the error of
the Binet sort of examination , no matter how extensive ,
must also be made .

It is thus very difficult to make anything like a valid
estimate of the probable form of distribution of altitude
of intellect measured without error and in equal units .
What we have done is to apply an error of the magni
tude shown in Table 163 to the empirically obtained scores
grouped as :
25 cases M.A. 6 yr . 0 mo . to 6 yr . 3 mo . inclusive ,

33 66 66
6 yr. 4 mo . to 6 yr . 7 mo . 66
40 66 66
6 yr . 8 mo . to 6 yr . 11 mo . 66
2 66 66
7 yr. 0 mo . to 7 yr . 3 mo . 66
72 76 80 84 88
FIG . 145 . The form of distribution resulting by the application of an error
of measurement to the Stanford -Binet scores of Group im . 6.
This gives us Fig . 145 as the form of distribution of this

group .
Using the form of distribution of Fig . 145 , the sigma
respectively -1.90σc , -
values for the difficulty of Composites C , D , E , F and G are
.45σD , + .290E , + 1.250F , and
+ 2.08σc .
TABLE 163 .
DISTRIBUTION OF THE ASSUMED ERROR WHEREBY A STANFORD BINET MENTAL

AGE DIFFERS FROM THE MENTAL AGE WHICH WOULD BE FOUND BY
A PERFECT MEASUREMENT OF ALTITUDE OF INTELLECT .
- 18 to - 21
- 14
mo . mo . 1
" "" - 17 66 2
-10 " 66
-13 " 3
- 6 " "" - 9 66 6
" "" - 5 "" 8
-
2
2 + 1 10
+ 2 " ""
+ 5 " 8
+6 66 66
+ 9 66 6
+10 " 66
+13 " 3
+14 " ""
+17 66 2
+18 " ""
+21 " 1
The 50 cases of feeble - minded of Class 3 may be as

sumed to be of approximately the normal form of distribu
tion in respect of intellect , since they represent those se
lected by the educational authorities of the institution as
belonging in Grade 3 rather than in Grade 2 or Grade 4 .
We have seen that the process of educational selection for
a grade tends in general to produce symmetry and an ap
proximation to Form A. However , there would probably
be a curtailment at the upper end and an extension at the
lower , since there would be a scarcity of children who were
much too bright really to belong to Grade 3 , but an abun
dance of children much too dull to belong there . Whereas ,
in an ordinary school the forces acting to produce grada
tion in Grade 3 select from a rectangle , in an institution
for feeble -minded they select from a surface which is pre
sumably much higher at the low than at the high end .
In a total CAVD summation score , the distribution was
that shown below .
Quantity Frequency
60 to 79 1
80 " 99 0
100 " 119 1
120 " 139 1
140159 0
160179 3
180 199 4
200219 7
220239 13
240 259 7
260 " 279 7
280299 4
300 319 2
In view of these facts we have assumed the form of dis

tribution of this group to be that shown in Table 164 .
-
-
Using this , the difficulty of Composite E is -1.35σ ; that
of Composite F is ―- 1.250F ; that of G is — .33σ ; that of
I
His - .41σ ; that of is +1.1701 .
It has not been practicable to secure sufficient informa
tion outside of our own tests for even the roughest em
pirical determination of the form of distribution of the

Special - Class group . Nor is enough known about the
policies of the different schools , principals and teachers ,
nor about the accuracy of the diagnoses to justify an
a priori estimate of the selective forces which relegated
these pupils to the special classes . There should theoreti-
cally be much of negative skewness , since the selection is
surely from the dull half and probably from the dullest
TABLE 164 .
FORM OF DISTRIBUTION ASSUMED IN OBTAINING MEASURES OF THE DIFFICULTY

OF VARIOUS COMPOSITES FOR THE GROUP OF 50 FEEBLE -MINDED .
Interval Frequency
a to a + k
a+ k " a+ 2k
112
66
a + 2k + 3k
a a a a a a a a a a a a
a + 3k " + 4k
+ 4k 5k
3 23
" "
+ +
a a a a a a a a a a
5k 6k
+ + +
6k + 7k 10
7k 8k 12
" "
+ + +
+ 8k 9k 20
9k 66 10k 20
+ + + + +
10k + 11k 12
"
11k 12k 10
+ + +
12k 13k
" "
2 2
13k 14k
quarter or eighth rather than from the brighter which

results if the effort is to select the fraction of the
normal surface between -1.5 S.D. and -2.7 S.D. but this
disturbed by an error distributed as follows -- .2
,
effort
is
S.D. —.1 S.D. S.D. +.1 S.D. S.D.

+
.2
4
4
1
1
;
;
,
Fig 146
.
This gives surface of the form shown

in
a
Using this form of distribution the differences difficulty

in
,
of and from the median difficulty for the

H
E
-
,
I,
,
F,
,
Special Class group are respectively units of the

of
-
in
o
,
the group 2.610E -2.310F 1.5400 .44σ and

—
,
,
,
+
.36σ1
.
-28 -1.3
FIG . 146. The form of distribution assumed in the case of the special class
group .
FIG . 147. The form of distribution of Grade 5 in Army Alpha , in equal units .
FIG . 148. The form of distribution of Grade 5 in Examination A, in equal

units .
582
THE MEASUREMENT
.
in
FIG
5
149. The form of distribution
in
Grade
,
of the National
.
Intelligence Examination equal units

OF INTELLIGENCE
39
THE FORM OF DISTRIBUTION
5
in
,
.
.
FIG 150. The form of distribution of Grade in the Otis Examination equal units
OF INTELLECT
583
To ascertain the form of distribution in Grade 5 and

the form of distribution in Grade 4, we have collected the
distributions of scores made by these grades in any of the
examinations whose scores we have transmuted into scales
FIG . 151. The form of distribution of Grade 4 in Examination A , in equal

units .
FIG . 152. The form of distribution of Grade 4 in National A , in equal units .
in equal units . The surfaces of frequency are drawn , using

the equal - unit scales . Figures 147 , 148 , 149 and 150 show
the essentials in the case of Grade 5 ; Figures 151 , 152 , and
153 show the essentials in the case of Grade 4.
FIG . 153. The form of distribution of Grade 4 in Otis Examination , in

equal units .
In view of these curves and the general probabilities of

the it does not appear that there is justification for
case ,
assuming any considerable flattening or any considerable
skewness . So the form of distribution for Grade 5 and
for Grade 4 is taken as that of the normal probability sur-
face .
In order to perfect our measures of the differences in
difficulty of levels N , O , P and Q , we need knowledge of
100 150 200 250

FIG . 154. The form of distribution of first -year law students , '24 , in Army
Alpha in equal units .
the form of distribution of altitude of intellect in the group

of 240 college graduates whose percentage of success at
each of these levels we have computed . This group of 240
comprised the first - year class entering the Columbia Law
School in 1924. We have their scores in Army Alpha taken
on the same day that they took the CAVD examination ;
and we have learned ( in Chapter VII ) the approximate
values of Army Alpha scores in truly equal units . The
distribution of the 240 in Alpha in equal units is shown in
Fig. 154. The Law School class entering in 1924 may be
regarded as differing from the classes entering in 1921 ,
1922 , and 1923 , only by chance in respect of the form of
distribution . In the case of each of these classes we have
the scores in several intelligence examinations , namely :
CLASS OF 1921
A completion test of about 1 hour .

A paragraph reading test of 1 hour.
A paragraph reading test of 1 hour ( selective type ) .
Thorndike Exam . , series 1925–1930 , Part , Form Q. I
Thorndike Exam . , series 1925-1930 , Part I , Form R or S.
An arithmetical composite .
CLASS OF 1922
A paragraph reading test of 1 hour.
A paragraph reading test of 1 hour ( selective type ) .
An examination containing opposites , Briggs's gram
matical analogies and a vocabulary test
Thorndike Exam . , series 1925-1930 , Part I , Form S or V
CLASS OF 1923
An extensive test in sentence completion and arithmeti
cal problems .
An examination containing opposites , Briggs's analo
gies , and a vocabulary test .
A paragraph reading test of 1 hour .
These records permit the application to " a first - year

class in the Columbia Law School" of the same processes
of reasoning used in Appendix III in the case of " a grade

population . " The surfaces of frequency for each sepa-
rate score are shown in Figs . 155 to 167. They show in
-3 -2 O +1 +2
FIG . 155. Form of distribution of first -year law students , '21 , Completions .
-3 -2 +1 +2
FIG . 156. Form of distribution of first -year law students , '21 , Reading I.
-3 -2 +2
FIG . 157. Form of distribution of first -year law students, '21 , Reading II .
-2 -0 +1 +2
+2
FIG . 158. Form of distribution of first -year law students , '21 , Thorndike ,
Part I- Q.
-3 -2 O +1 +2
FIG . 159. Form of distribution of first -year law students , '21 , Thorndike ,
I
Part , R or S.
-3 -2 + -0 +1 +2
FIG . 160. Form of distribution of first -year law students , '21 , Arithmetic .
general a departure from " normality " in the shape of

a longer tail at the low end . This is still clearer when we
rid the measurements of chance variations by combining
FIG . 161.
-3 -2 J +1
Form of distribution of first -year law students, '22 , Reading
+2
I.
-3 -2 -0 +1 +2
FIG . 162. Form of distribution of first -year law students , '22 , Reading II .
-3 -2 + -0 +1 +2
FIG . 163. Form of distribution of first -year law students , '22 , Verbal
Relations .
-3 -2 +1 0 +1 +2
FIG . 164. Form of distribution of first -year law students , '22 , Thorndike I,
S or U.
the scores for each individual in the case of the 1921 and
groups , as shown in Figs . 168 and 169. When the
3
1922
surfaces of frequency are combined (with equal weight for
each examination ) for each class , we have Figs . 170 , 171 ,
and 172. When these are combined ( with equal weight to
1921 , 1922 , and 1923 ) we have Fig . 173 .
3 2 +1 0 +2
FIG . 165. Form of distribution of first -year law students , '23 , Co. and Ar .
Both the direct evidence from the 1924 group itself,

when measured by Alpha transposed to a scale of equal
units , and the evidence from the groups of 1921 , 1922 , and
1923 in which the inequalities of units may be assumed to
have approximately counterbalanced one another , show a
negative skewness . So also does the distribution of the
1923 group when measured by Army Alpha with a scale of
equal units . This is shown in Fig. 175. So we have taken
3 This would be true also with the 1923 group , but we have not had time
to make the computations exactly .
MMI-3
‫ھچا‬฀
FIG . 166. Form
-2
of distribution of first -year law
Relations .
students ,
+2
'23 , Verbal
-3 -2 +1 0 +1 +2
FIG . 167. Form of distribution of first -year law students, '23 , Reading I.
-3 -2 -0 +1 +2
FIG . 168. Form of distribution of first -year law students , '21 , Total Score .
-3 -2 +1 O +1 +2
FIG . 169. Form of distribution of first -year law students , '22 , Total Score .
1
-3 -2 +1 +1 +2
FIG . 170. Composite of the distributions of Figs . 155 to 160 .
-3 -2 +1 +1 +2
FIG . 171. Composite of the distributions of Figs . 161 to 164 .
฀฀
฀฀฀
฀฀
฀฀
-2 +1 +1 +2
3
-
FIG 172 Composite the distributions Figs

of
of
to
165 167
.
-2
+
FIG 173. Composite of the curves of Figs 170 171 and 172 with equal
,
.
,
to .
weight allowed each

.
the distribution of Table 165 , which is that of Fig . 173 with

some smoothing , as representing the probable form of dis
tribution of altitude of intellect in the group of 240 .
TABLE 165 .
THE PROBABLE FORM OF DISTRIBUTION OF ALTITUDE OF INTELLECT IN GROUP

17 ( LAW STUDENTS ) .
Frequency Frequency
Interval permille Interval permille
L to L+ 1 + L + 33 " L +34 37
L+ 1 " L+ 2 } L + 34 " L +35 38
L + 2 " L+ 3 + L + 35 " L + 36 39
L+ 3 " L + 4 } L + 36 " L +37 391
L+ 4 " L+ 5 1 L + 37 L + 38 391
L+ 5 " L+ 6 1 L + 38 " L +39 39
L+ 6 " L+ 7 11 L + 39 " L +40 391
L+ 7" L+ 8 11 L + 40 " L +41 39
L+ 8 " L+ 9 2 L + 41 " L +42 38
L+ 9 " L + 10 2 L + 42 " L +43 37
L + 10 " L + 11 21 L + 43 " L +44 36
L + 11 " L + 12 21 L +44 "" L +45
341
L + 12 " L + 13 21 L + 45 " L +46 33
L + 13 " L + 14 3 L + 46 " L +47 31
L + 14 " L +15 31 L + 47 L +48 29
L + 15 " L + 16 4 L + 48 " L +49 251
L + 16 " L + 17 41 L + 49 " L + 50 22
L + 17 L + 18 51 L + 50 66 L + 51 19
L + 18 " L + 19 6 L + 51 ´´ L + 52 16
L + 19 " L + 20 7 L + 52 L + 53 131
L + 20 L + 21 8 L + 53 "" L + 54 11
L + 21 " L + 22 9 L + 54 " L + 55 81
L + 22 " L +23 10 L + 55 66 L + 56 6
L + 23 " L + 24 12 L + 56 " L +57 4
L + 24 "" L + 25
14 L + 57 " L + 58 31
L + 25 " L + 26 16 L + 58 L +59 3
L + 26 " L +27 181 L + 59 " L + 60 21
L + 27 L + 28 21 L + 60 L + 61 11
L + 28 L +29 24 L + 61 L +62 11
L + 29 " L + 30 27 L + 62 " L +63 1
L +30 " " L +31 30 L + 63 " L + 64 1
L + 31 L +32 33 L + 64 L + 65 1
L + 32 L +33 351 L + 65 "" L + 66 }
-
The percents correct for the four levels in question
-
being 95.4 , 77.1 , 56.7 , and 22.9 , the values in terms of o dis
tances from the median , are 1.862 , .714 , — .153 , and
+.738 in terms of σN , σo , σp and σ , respectively .
FIG . 174.
Л
The distribution of Fig. 173 , with some smoothing .
50 100 150 200

FIG . 175. Form of distribution of first - year law students , '23 , in Army Alpha ,
in equal units .
In the case of the 44 recruits , a normal form of distri

bution is assumed because nothing demonstrably better is
suggested by the facts available . These facts are : the re
ported amount of schooling

the distributions of scores for
,
the group in Otis Intermediate , G.E. Form I ( a General

Electrical Information test ) and the sum of the scores in
J
Composites I , , and K ( but with the arithmetic of K esti-
C + V+ D
They appear in Table
mated as ). 166 .
3
TABLE 166 .
DATA FOR ESTIMATING THE FORM OF DISTRIBUTION OF ALTITUDE OF INTELLECT

IN THE GROUP AD . (44 RECRUITS ) .
Grade Reached Otis Int . G. E. I+J+ K

Grade Freq . Score Freq . Score Freq . Score Freq .
3 or 4 1 15-19 4 3- 5 3 10-29 2
5 or 6 10 20-34 10 6-9 6 30-49 9
7 or 8 18 35-49 13 10-13 26 50-69 9
9 or 10 11 50-64 12 14-17 8 70-89 11
11 or 12 3 65-79 4 18-21 2 90-109 10
22-25 0 110-129 3
26-29 1
30-33 0
34-37 0
38-41 1
UNIVERSITY OF MICHIGAN
Original from
Digitized by Google
9)/6006-pd¥#asn ssad0e/bs0'1sn4u Ty ey MMM//:dI121yU / paztiT6Tp-a16009 ‘utTewog 2T1qGNd
8ZOSITOOOSTOGE Apu/ZZ7OZ/JOU* a pueYy*1pYy//:sdz1y / IW9 SS'Z0 OZ-vO-€Z0Z UO pazesauag
LIST OF REFERENCES
Anderson, J. E. '20 Intelligence Tests of Yale Freshmen . School and
Society 11 : 417-420 .
Bailor , E. M. '24. Content and Form in Tests of Intelligence . Teach-
ers College , Columbia University , Contributions
to Education , No. 162 , 1924 .
Baldwin , B. T. , and
Stecher , L. E. '22. Mental Growth Curve of Normal and Superior
Children . University of Iowa Studies 2 : No. 1 .
Binet , A. '16. The Development of Intelligence . ( English trans-
lation by Kite . )
Boas , F. , and
Wissler , C. '06 . Statistics of Growth . Report of the U. S. Com-
missioner of Education for Year ending June
30 , 1904 , 1 : 25-132 .
Brooks , F. D. '21 . Changes in Mental Traits with Age , determined by
Annual Re - tests . Teachers College , Columbia
University , Contributions to Education , No. 116 ,
1921 .
Bulletin , July 21 , 1923 ... Bureau ofEducational Research , University of

North Carolina .
Burt , C. '21 . Mental and Scholastic Tests .
Carey , G. L. , and
Kline , L. W. '23 . The Kline - Carey Measuring Scale for Free - hand
Drawing . The Johns Hopkins University
Studies in Education , No. 5a , 1923 .
Clark , J. R. '25 . The Relation of Speed , Range , and Level to
Scores on Intelligence Tests .
Cobb , M. V. 22 .. The Limits Set to Educational Achievement by
Limited Intelligence . Jour . Educational Psy .
13 : 449-464 and 546-555 .
Colvin , S. S. '22 . The Present Status of Mental Testing . Educa-
tional Review 64 : 196-206 and 320–327 .
Colvin , S. S. , and
MacPhail , A. H. '22a ....TheValue of Psychological Tests at Brown Uni-
versity. School and Society 16 : 113-122 .
Colvin , S. S. , and
MacPhail , A. H. '24….……..... The Intelligence of Seniors in the High Schools
of Massachusetts . Dept. of the Interior , Bureau
of Education , Bulletin No. 9 , 1924 .
Coxe , W. W. '21 .. Variation in General Intelligence . Jour . of Edu-
cational Research 4 : 187-194 .
601
40
602 LIST OF REFERENCES
Haggerty, M. E. '23 Haggerty Intelligence Examination ; Manual of

Directions .
Hammarberg , Carl '93 . Studier öfver Idiotiens Klinik och Patologi .
Hillegas , M. B. '12 Scale for the Measurement of Quality in English
Composition by Young People . Teachers Col
lege Record 13 : 331-384 .
Hollingworth , L. S. '23 .The Psychology of Subnormal Children .
Kelley , T. L. '19 . The Measurement of Overlapping . Jour . Educa
tional Psy . 10 : 458–461 .
Kelley , T. L. '21………… The Reliability of Test Scores . Jour . Educational
....
Research 3 : 370–379 .
Kelley 23 Statistical Method .
T.
L. L.
,
Kelley T. 23a The Principles and Technique Mental Measure
of
..
,
'
ment American Jour of Psy 34 408-432
.
:
.
.
.
Madsen N. '22 Intelligence Factor School Progress

J.
as
in
a
,
.
School and Society 15 285ff
.
Memoirs '21 National Academy of Sciences Psychological Ex
.
.
amining Army edited by Robert M.
U.
the
in
S.
,
Yerkes 15 1921
.
:
Miller W. '22 Miller Mental Ability Test Manual of Directions

S.
,
,
First Revision for Forms A and B.

Myers E. and
C.
,
,
Myers 21 Measuring Minds

S. G. C. G.
C.
,
Myers
E.
,
Myers and
C.
,
,
Layton H. '21...... Group Mental Testing Altoona Pa School and

in
,
Society 13 624-628
:
Noble and
G. E.
F. L.
,
Arps University Students Intelligence Ratings Accord

20
,
'
ing the Army Alpha Test School and So

to
ciety 11 233–237
:
Otis A. '21 Otis Group Intelligence Scale Manual Direc

S.
of
,
tions 1921 Revision

,
Pearson K. '14……………… Tables for Statisticians and Biometricians

..
,
Pressey '20 Brief Group Scale of Intelligence for Use

L.
in
S.
A
,
School Surveys Jour Educational Psy

11
:
.
89-100
.
Ruch G. M. '23....... Mental Educational Survey Univ of Iowa

A
,
Studies in Education No.

2
5
:
Spearman '04 The Proof and Measurement of Association be

C.
,
tween Two Things American Jour of Psy 15

:
.
.
.
72-101
.
LIST OF REFERENCES 603
Spearman , C. '07 .Demonstration of Formulae for True Measure

ment of Correlation . American Jour . of Psy .
18 : 161-169 .
Spearman , C. '10. Correlation Calculated from Faulty Data . British

Jour . of Psy . 3 : 271–295 .
Spearman , C. '13 . Correlations of Sums and Differences . British
Jour . of Psy . 5 : 417–426 .
Spearman , C. '23 The Nature of Intelligence and the Principles of
Cognition .
Taylor , G. A. '23.. An Inventory of the Minds of Individuals of Six
and Seven Years Mental Age . Teachers College ,
Columbia University , Contributions to Educa
tion , No. 134 , 1923 .
Terman , L. M. '16 . The Measurement of Intelligence .
Terman , L. M. '22 . The Great Conspiracy or The Impulse Imperious

of Intelligence Tests , Psychoanalyzed and Ex
posed by Mr. Lippmann . The New Republic
33 : 116-120 .
Terman , L. M. '23 . Terman Group Test of Mental Ability for Grades

7 to 12 ; Manual of Directions .
Thomson , Godfrey H.
'24 A Formula to Correct for the Effect of Errors of
Measurement on the Correlation of Initial
Values with Gains Jour of Experimental Psy
.
.
321-324
3
... Handwriting . II
:
Thorndike , E. L. '10 Teachers College Record , , No. 2 .
Thorndike , E. L. '13..... The Measurement of Achievement in Drawing .

Teachers College Record 14 : 1–39 .
Thorndike , E. L. An Introduction the Theory of Mental and
to
'13a
...
Social Measurements
...
.
Thorndike Aesthetic Appreciation Jour Educa

E.
Tests
L.
of
'16
,
tional Psy 509–522

7
:
.
Thorndike On the Improvement Intelligence

E.
Scores from
L.
'23
in
,
Fourteen to Eighteen Jour Educational Psy

.
.
.
14 513-516
... The
:
Thorndike Influence of the Chance Imperfections

to of
L.
E.
'24
,
Measures upon the Relation of Initial Score

Gain or Loss Jour Experimental Psy
3
.
:
.
225-232
.
Thorndike , E. L. ,
Bregman , E. O. , and
Cobb , M. V. '24 . The Selection of Tasks Equal Difficulty by
of
Consensus of Opinion Jour Educational Re

.
search 133-139
9
:
.
604 LIST OF REFERENCES
Thorndike , E. L. , and
Bregman , E. O. 24. On the Form of Distribution of Intellect in the
Ninth Grade . Jour . Educational Research 10 :
271-278 .
Thurstone , L. L. '21 . Cycle omnibus Intelligence Test for College
-
Students Jour Educational Research 265
4
:
.
.
278
.
Trabue , M. R. '16. Completion - Test Language Scales . Teachers Col
lege , Columbia University , Contributions to Edu
cation , No. 77 .
Vincent , Leona M. '24... A Study of Intelligence Test Elements . Teachers
College , Columbia University , Contributions to
Education , No. 152 , 1924 .
Whipple G. M. 22 Intelligence Tests in Colleges and Universities .
,
The 21st Yearbook of the National Society for

the Study of Education , Chap . X : 262 .
Whipple , G. M.
-
The Intelligence Testing Program and its Objec
23.
tors Conscientious and Otherwise School and
.
Society 17 561-568 and 596–604
:
.
Yerkes R. M. '20 Examining the United States
in
Psychological
..
,
Army Memoirs National Academy Sciences
of
,
,
.
15
.
INDEX
Abilities measurable by I. E. R. tech Arbitrariness in units , 1 , 3ff .; in
nique , 476 choice of tasks as intellectual , 61
Ability , defined , 484 ; to deal with Arbitrary scores transmuted to make
persons , things , ideas , by ideas , variabilities equal , 520
413 ; to learn as criterion of intel Area of intellect , 24 , 339 , 378ff ., 469 ;
ligence , 17 ; to learn as intelligence and nature vs. nurture , 458 ; and
test , 408ff .; vs. variability , 43ff ., stock examination score , 404 ; as
56 , 497ff . related to altitude and width , 378
Absolute vs. proportional counts of Arithmetic problems , as tests , 9 ;
tasks , 383f . composites , construction of , 193 ff .
Absolute zero , 294 , 336f ., 339ff ., 471f ., Army Alpha , 1, 2 , 3 , 6ff . , 10 , 15 , 22f .,
482 , 485f .; and Binet Scale , 403 42 , 44 , 46 , 49 , 52 , 55 , 99 , 227 , 307 ,
Acceleration of level with age , nega 403 , 404 , 410 , 433 , 437 , 461 , 475 ,
tive , 467 , 475 491 , 494 , 522 , 528 , 534 , 539 , 562 ,
Accident , effect on form of distribu 586 , 592 ; form of distribution ,
tion of intelligence , 272 291f .; letter ratings , 7 ; scores and
Acquired vs. original ability , 95f ., officers ' ratings , 405 ; scores in
433 ff . equal units , 228ff ., 309 , 316
Adequacy one CAVD
of level for Army Beta , 15 , 18 , 222 , 433 , 437 , 506 ;
measurement , 565 ff . form of distribution , 291f .
Advantages from location of zero Army Examination a , 228 , 475 , 514 ,
point , 339 522 , 528 , 562 ; formof distribution ,
Age changes in altitude and area , 291f .; scores in equal units , 264ff .,
463ff .; in specialized abilities , 468 ; 309 , 313 , 316
grouping for intellectual difficulty , ARPS , G. F. , 539
28 Array, form of distribution of, to
Allowance for environmental effect on measure difficulty , 40 , 54ff .; in cor
intelligence , 460ff . relation table to measure ability vs.
Altitude , as related to width , 376 ; variability , 505 , 520
and nature vs. nurture , 458 ; and Arrays used to measure difficulty , 40
speed of learning , 409 ; and stock Associative vs. selective and generaliz
examination score , 403f ., 405f .; ing thought , 414f .
curve with age , 463 ; measured Assumptions in measuring intellectual
without error , 577 ; of intellect , 24 , difficulty , 38 , 59
33 , 104 , 469 Attenuation , formula , 60 , 111 , 177 ,
Ambiguity in content of tests , 1ff . 560f .; formulae used to find com
Analogies test , 9 mon part of tests , 406f ., 429f .;
Analysis of measurement into level , corrected for in correlations , 424f .;
width , speed , 35ff . in composites with few tasks , 556 ;
Anatomical cause of intellect , 420 in variability vs. ability , 500 , 502 ,
ANDERSON , J. E. , 539 , 541 506 ; used on correlations of Burt ,
Annoyers , 489f . 455f .
Applications of I. E. R. technique to Average ability of individual , 493f .
measurement of human abilities , Axioms of mental measurement re
476ff . original and acquired ability , 436
605
606 INDEX
BAGLEY , W. C. , 433 Central tendency as reference point ,

BAILOR , E. M. , 135 , 317 42 , 43 , 57 , 109 , 148 , 209ff ., 224ff .,
BALDWIN , B. T. , 495 297ff ., 314ff ., 339 , 491ff ., 497ff .
Batteries of tests , 1 , 9 CHAMBERLAIN , A. F. , 9
Beauty of design , scales for , 139 Chance error , in original measures
Beta , Army , 15 , 18 , 222 , 433 , 437 , affecting partial correlations , 454f .;
506 ; form of distribution , 291f . in stock examination scores , 411 ;
BINET , A. , 16 , 409 smoothing for , 299 ; in variations ,
Binet , mental age , 3 , 6 , 16 , 17 , 97 , eliminated , 590ff .
402 , 447 , 575 ; scale , zero on , 403 ; Change from 100 percent to 0 percent
tests , 9 , 18 , 23 , 338 , 433 , 435 , 451 , correct , 374f .
454 , 455 , 457 , 577 Changes in altitude and width with
Biserial r , 118 , 120ff ., 131 , 389 age, 463ff .
Blurring of variability vs. ability by Cheating , 476ff ., 564
errors of measurement , 500 , 502 CLARK , J. R. , 97f ., 106 , 318 , 405ff .
BOAS , FRANZ , 112 Coaching for tests , 437
BREGMAN , E. O. , 135 , 138 , 306 , 521 COBB , M. V. , 135 , 528 , 534
Briggs Analogies Test , 586 COLVIN , S. S. , 9 , 433 , 534
BRIGHAM, C. C. , 540 , 546 Combining novelty and familiarity in
BROOKS , F. D. , 466f . test tasks , 439f .
Brown University , 540 ; Psychological Commensurable Οι 's in various groups ,
Examination , 228 , 534 , 540 ; scores 303ff .
in equal units , 260ff ., 316 Common causation of associative and
BuChHoSo , intellect and age , 468 higher thought processes , 425 , 430
Buckingham Spelling Scale, 119 Common measure for intellect of ani-
BURT , C. , 9 , 435 , 447ff . mals and men , 340

Burt Reasoning Tests , 447ff . Common reference point for difficulty
of composites , 314ff ., 331f .
Capacities independent of intellect , Common sense , criteria for test items ,
431 64 ; test , 9
CAREY , G. L. , 134 Comparative , enumeration of ideas ,
Carpentry , 476 , 480ff . 419 ; variabilities in grade groups ,
CATTELL , J. McK ., 1 311f .; in face -value and equal
Causation of , ability , 95 ; area and units , 318ff .
altitude of intellect , 397ff ., 402f .; Comparison of , consensus with experi-
associative and higher thought proc- mental results , 143ff .; variabilities ,
esses , 425 , 430 ; form of distribution 303 ff .
of intellect , 271 ; intellect , 420 ; Completion , composites , 193ff.; tests ,
variation , 271 9 ( See also Trabue Completion
CAVD , altitude , 197ff ., 102 ; and other Tests )
tests , 97ff .; composite tasks , 66ff . , Composite tasks , as tests , 118f ., 128 ,
160 ff .; composites , construction of , 133 ; vs. single tasks as measures of
176ff .; correlation with other tests , intellect , 470
96ff .; intellect , 65 ff ., 96ff ., 101ff ., Composites , I
A to , 294 ; difficulty of,
104ff . , 114 ; and original capacity , 295 ff .;I to Q , 294 ; of specified
435ff .; and training , 435ff .; level , difficulty , 128f .; made by consensus
form of distribution , 285 ff .; vs. ratings , value of , 156ff .
width , 388f .; selection of tasks , Composition , of word knowledge lev-
410 ; width and altitude , 102 els , 184f .; scales , 134 , 476ff .
INDEX 607
Concept of altitude and area , 379 389ff . , 394ff .; of officers ' ratings
Connection between ideas and sequent and Army Alpha scores , 405 ; of
ideas , 417 one composite and summation score ,
Connection - forming vs. higher thought 302 ; of one level and sampling of
processes , 414ff . entire series , 565ff .; of single tasks
Connections to varying ideas from one and intellect , 118ff ., 129ff .; of sub
idea , 419 series in CAVD with total CAVD ,
Consensus of expert opinion as mea 102ff .; of tasks of one sort and
sure of intellectual difficulty , 134ff ., composites , 154 ; tetrachoric , 390 ;
156ff .; of experts for location of technique vs. line best fit , 339f .;
of
zero point , 342ff .; ratings , under technique vs. probable error of
and over estimation in , 157 ; valid measurement , 520
ity of , 141 Correlations corrected , by Spearman
Consequences from given stimuli , 420 Brown formula , 299ff ., 324 , 426 ,
Considerations limiting test items , 559 , 561 ; for attenuation , 406f .,
63f . 426ff ., 455 , 560 , 561 , 562 , 570f .;
Constant error , in judging , affecting for range , 155 , 557ff . , 562 , 570f .;
results , 134 ; in intellectual diffi for remoteness of level , 563
culty of tests , 141ff . , 145 , 148ff ., Correlation , self- , of ratings , 138 ; of
155f .; in scoring affecting results , sub -series in CAVD , 107 ; of task ,
132 ; tending to symmetry , 271f . 50 ; of various tests , 99ff .
Constitution of intellect , 50 Correspondence of degrees of CAVD
Construction , of CAD composites , and other variables , 103
193ff .; of V composites , 179ff . Courtis tests in computation , 492
Content of intelligence tests , 20f . COXE , W. W. , 522
Conventional zeros , 487f . Credits , graded , 29f ., 477 , 480
Correction for , attenuation , 392f ., Criteria for , choice of test items , 64 ;
424ff ., 567 ; chance , 299 , 301 ; composites , 129f .; estimating na
range , formula for , 155 ; remoteness ture vs. nurture , 441f .; intellectual
of level , 298f .; restricted range , difficulty , 63f .; intellectual tasks ,
155 , 556 , 562 ; sampling error , 112f ., 60f .; intelligence , 10 , 15ff .; used in
118 making N. I. T. , 409f .; value of
Correlation , as tool for discovering mental product , 14
differences in tasks at same levels , Curve , Gaussian , 6 , 53 , 270 , 521 , 527 ,
382 ; of associative and higher 536 ; of altitude and age , 463ff .; of
abilities , 42f .; of CAVD level and area and age , 467ff .; of percent
CAVD score , 105f .; of CAVD level correct , 351ff .
and other test scores , 105ff .; of Data for intelligence testing , 20f .
CAVD score and other test scores , Dearborn Intelligence Examination
97 ff.; of CAVD score and Stanford 41 , 493
M.A. , 97 ; of CAVD width and alti Definition of , ability , 484 ; idea , 418 ;
tude , 102 ; of CAVD with other intellect , 25f ., 27 , 412 ; intellectual
tests , 96ff .; of C and A tasks , difficulty , 26f .; product , 25ff .;
115ff .; of composites , 376 ; of com range , 26f .; task , 26f .; one idea ,
posites more and less remote , 557 ; 418f .
of consensus and experimental re Democracy and the I. Q. , 433
sults , 143ff .; ofmeasure of altitude Derivation of units of measure , 134
and success - failure in composites , Design , scale for beauty of , 134
608 INDEX
Determinism , educational , 433 stock examination scores 411 re
;
Development with age as criterion of duced by successive summation of
intelligence , 16 tending spuri
to
scores 530 545
.;
ff
,
,
Differences in CAVD composites N to ous normality 271 529 545 550
,
,
,
,
Q , 304ff . , 320f .; consensus and ex 572ff 577ff 590ff vs. initial
.;
.,
.,
perimental results , 142ff .; difficulty ability and gain 408f
.
of composites , 485 ; from common Errors of sampling 109f 112 114
.,
,
reference point , 415f ., 420f .; grade 117f affecting partial correla
as
.;
in of tion coefficients 454ff
,,,
medians terms o 218f
.;
.
individuals connection systems Estimated effect on correlation of
'
,
meaning of scores in
331ff small composites
small range of
.;
314ff
.,
,
stock examination and CAVD level age and levels 556
,
score 407 original capacity 421f Estimating CAVD altitude individ
of
.;
,
;
,
by two methods of calculation ual 369
,
o,
303 tasks at the same level 380ff .; , Examinations school tests
to as
;
1
,
,
diagnosis of nature and nurture Experiment program locate zero
,
,
460 339ff
.
Differential gains in intellect with Expert opinion as measure of intel
ability 288 age 287ff training lectual difficulty 134ff 156ff
.;
,
,
;
,
.,
.
288f vocational selection 289 Expressing CAVD altitude of
.;
an
in
,
,
Difficulty CAVD 103 general vs. dividual various sigmas in
.;
365
in
ff
;
,
intellectual difficulty 28 56f 119 common unit 303ff

,
,
.,
.
126f 129 133 intellectual 22f Extent of intellect See Width
,
.,
;
.,
)
(
26 28 62ff 109ff criteria for
.;
,
.,
,
63f of task 109 of CAVD Face value vs. true value of scores
.;
,
a
;
,
composites 210ff to 321ff 3f 16 43 53 224ff 497ff

.; H
.;
.;
A
.
.,
.,
,
,
affecting form of N.
T.
295ff related Factors
I.
to
471
to
A
Q
Q
,
I
,
;
,
distribution intellectual
in
correct 351ff
percent from zero 276
as to
.;
;
,
,
reference 349f of CAVD lev achievement 431

.;
,
,
els 179 of tasks measured by per Failure success marking 477 479
,
,
-
;
,
cent successful 39 62 109 121 Familiar tasks as tests 439f

,
.
,
Fit goodness 527ff

of
123 144f 148ff 185f 190ff

,
. ff
.
,
.,
.; .,
.,
.,
.
195 202ff 205ff of word knowl Form of curve of altitude and age
,
,
ff
.,
,
edge tasks 182f 463 area and age 467ff per

.;
.;
ff
,
,
Directions composites construction cent correct 351ff

,
.
,
ff ,
of 193 test Form of distribution by year age

.;
,
9
-
,
value of school 285ff adult im

in
Disciplinary studies 274f 294

;
.,
.,
,
group 147 572ff adult recruits

in
440
.;
3 in ,
,
.; ,
Discrimination pitch 476f 486 598f in Army Alpha 291f

.;
of
,
,
.,
Disease effect on form of distribution

Army Beta 291f Class
in
.;
,
,
group Grade group

in
of intellect 272 579

4
in ,
;
.; ,
,
Dissociation vs. integration 584f Grade group 584f

in
.;
431
5
,
,
form of See Form of im group 572ff im group

in
Distribution
.;
3
6
,
.; ,
,
Symmetry of 270f 575ff in Law School group 586ff

.;
Distribution
,
,
;
)
Drawing scales 134 476ff special class group of

in
580
;
,
,
,
adults 287ff of altitude Grades

in
.;
to ,
Errors of measurement 58 132 due VI XIII 224 of an array

294
,
,
;
;
,
to carelessness cheating 564 in of college freshmen 538ff of

or
520
.;
,
;
,
;
INDEX 609
GOPI intellect , 286 ; of Grade VI , GOPI intellect , form of distribution ,

521ff . 286
Form of distribution of, Grade IX , Gradation of responses to tasks , 477
521 ,528ff .; Grade XII ,
533ff .; Grade 6 , 9 , 12 , variability in , 104
group given composites A to D , Grade 9 as unit , sigma of variability
142 ; intellect , 270ff ., 294 , 295 ; level in, 112
in CAVD intellect , 104f .; school Graded inventory to measure intellect ,
grade population , 470 , 483f .; varia- 412
tions as check on estimates of diffi- Greatness of intellect in terms of area
culty , 41 ; variations in intelligence and altitude , 387
test scores , 104 Growth and practice effects on Alpha
Form -boards as tests , 436f . scores , 491
Formula , for attenuation , 60 , 116 , 560 ,
561 ; for coefficient of correlation , HAGGERTY , M. E. , 540 , 546
Haggerty Intelligence Examination
485 ; for correction of error in mea-
surement of gains , 409 ; for correc- Delta 2 , 41 , 228 , 274 , 291 , 309 , 316 ,
tion for restricted range , 155 ; for 493 , 510 , 522 , 528 , 530 , 533 , 534 ;
unreliability , 298f ., 324f ., 559 , 561 , increase in score with age , 464ff .;
567 ; for rti , 177 ; for It₁19 302 ; for scores in equal units , 247ff ., 276 ,
rt , 110 ; for rho , 147 , 485 ; for

316 , 464ff .
, 115f .; for o₁ , 110 , 222 , HAMMARBERG , C. , 432

0200 + 20A Handwriting scales , 134
297 ; for σt ,, 110 ; for ot , 110 , 111 ,
Hearing , tests for , 476ff .
117 ; for true σ, 122 ; Spearman-
Hereditary , ability vs. acquired abil
Brown , 425f ., 556 , 567 , 570
ity , 95f .; factors producing varia-
FREUD , S. , 272 , 461
bility , 272
Herring Revision of Simon -Binet
Gain , as related to , errors of measure-
Tests , 403
ment , 409 ; initial ability , 409 ; in
Higher thought processes vs. connec-
intellect with maturity , 463 ff .; in
intelligence test scores , 287ff .; of tion - forming , 414ff ., 473
HILLEGAS , M. B. , 134
whites vs. negroes , 288
HOLLINGWORTH ,H. L. , 134
GALTON , FRANCIS , 1 , 270
HOLLINGWORTH ,L. S. , 135 , 433
GATES , A. I. , 135
Homogeneity of CAVD , difficulty ,
Gaussian curve , 6 , 53 , 270 , 521 , 527 ,
101ff ., 469 ; intellect , 556ff ., 571
536
Honesty , tests in , 476ff .
General Electrical Information Test ,
HUNSICKER , L. M. , 400f .
599
Hypotheses on increases in intellect ,
General principles and technique of 48 ff .
measurement of human abilities ,
476ff .
Hypothesis on nature and cause of
intellect , 412 , 415
Generalizing vs. associative thought ,
414f . Idea defined , 418
Goodness of fit of , college freshmen I.E.R. , Tests of Se-
technique , 476 ;
curve , 541 ; Grade VI
curve , 529 ; lective and
Relational Thinking ,
Grade IX curve , 527 ; Grade XII Generalization and Organization , 96 ,
curve , 535 ; summated scores curve , 101 , 120 , 228 , 522 , 528 , 534 ; score
547
in equal units , 257ff ., 316
610 INDEX
Illinois Intelligence Examination , 41 , CAVDOSR , 414 ; CRPF , 222 ; GOPI ,

493 , 522 , 528 ; scores in equal units , 286 , 571 ; NIL , 222 ; OGAnS , 222 ;
316 defined , 25f . , 412 , 433 , 469 ; synonym
Imperfections in testing instruments , for intelligence , 1
10 , 109 Intellectual difficulty , 22f . , 28 , 62ff .,
Importance of altitude of intellect , 35 109ff., 119 , 126f . , 133 ; as measured
Improvement in specialized abilities by consensus , 156ff .; assumptions in
with age , 468 measuring , 38 , 59 ; criteria for ,
Inaccuracy of determination affecting 63 ff.; in Binet tests , 402 ; of CAVD
ability vs. variability , 500 levels , 179ff .; of composites A, B ,
Increase in number of tests with level , C , D vs. ratings , 142 ; of composites
384ff . A to K , 335f .; of composites toI
Increases in intellect , 48ff . Q , 302ff .; of tasks , 53 ; of task , zero ,
Individual differences , affecting corre- 348f .
lation of altitude and width , 115 , Intellectual , level , 24 , 33 , 104 ; power
395ff .; in connections , 415f ., 420f .; over persons , things , ideas , 413 ;
in original nature , 421ff . product defined , 25f .; task defined ,
Inequalities in units , 306 , 526 59f ., 469 ; tasks , 26 , 59ff ., 413f . ,
Inferences as to nature vs. nurture , 469 ; vs. non -intellectual tasks , 413f .
441ff . Intelligence ( See Intellect )
Influence , of disease and accident upon Intelligence quotient , 433 , 495f .
form of distribution of intellect , Intelligence tests ( See Analogies ,
272 ; of hereditary factors upon Army Alpha , Army a, Army Beta ,
variability , 272 ; of selection to University , Burt ,
Binet , Brown
simulate normality , 272 CAVD , Dearborn , form -boards , Hag-
Inherited ability vs. acquired ability , gerty , Herring , I.E.R. , Illinois ,
95f . , 433 ff . Iowa , Kelley - Trabue , mazes , Menti-
Inhibitions as connections , 417
meter , Miller , Morgan , Myers ,
Initial ability vs. errors of measure-
National , Otis , Pintner , Pressey ,
ment, 409
Princeton , Smith College , Stanford
Instantaneous change from success to
Binet , Strickland , Terman , Thur-
failure at higher level , 374f .
stone , Trabue , Thorndike , Yale
Instructions for rating difficulty of
Tests )
test elements , 135f .
Integration vs. dissociation , 431
Intercorrelation of associative and
higher abilities , 426ff .
Intel as unit , 6
Intercorrelations , for determination of
Intellect , and single tasks , 109ff .,
altitude , 389 , 392ff .; of 40 com-
131ff .; as ability to get truth , 15 ;
posites C to G, 325ff .
as ability to learn , 17f .; as develop-
ing with age , 16f .; as organizing Interdependence of altitude , area,
ability , 19f .; as relational thinking , width , 412
19f .; as response to novelty , 18f .; Interest and capacity , 436
CAPIma , 397 ; CAVD , 65 , 96f ., 99 , Interviews as tests , 1 , 15
101ff ., 114 , 118f . , 143 , 159 , 178 , Inventory , as tool for measurement ,
222f ., 286 , 302 , 338 , 374ff ., 383 , 115 , 399 , 469 , 477f .; of intellectual
388ff ., 403ff . , 410f ., 411 , 458 , 463 , tasks , 27 , 159 , 412
469 ff . , 476 , 556 , 564 , 565ff ., 574 ; Investigation needed of altitude and
CAVDI , 97 , 99 , 101 ; CAVDIO , speed of learning , 409
154f .; CAVDO , 414 ; CAVDOS , 414 ; Iowa Comprehension Tests , 540 , 546
INDEX 611
JASTROW , J. , 1 intellect by inventories , 469 ; intel

Judgments of intellectual difficulty , lect by measurement of connections ,
134ff . 422 ; intellect by percent of success ,
37 ; intellect by product produced ,
Kansas State Teachers ' College , 522 14 ; intellect by single tasks , 109ff .,
KELLEY , T. L. , 110 , 112 , 117 , 155 , 225 120 , 127 ; intellect , theorems for ,
Kelley - Trabue Completion , 510 , 530 ,
30ff .; intellect via form of distri
533
bution of arrays , 54ff .; intellect via
KLINE , L. W. , 134 form of distribution of intellect ,
51ff .; intellect via form of distribu
LAYTON , L. H. , 522
tion of variations , 40 ; intellectual
Level , CAVD , composites , A to D ,
difficulty , 37ff ., 470 ; intellectual
66ff .; Ito M, 160ff .; N to Q , 76ff.
difficulty of a task , 109ff .; level ,
Level , importance of, 35 ; intellectual ,
width , speed , 35ff .; likes and dis
24 , 33 , 104
likes , 489f .; mechanical ability ,
Levels of intellect , 159ff .
476ff .; motor skill , 476ff .; original
Limitation of test items , consideration
capacity , 95 ; popularity , 476ff .;
of, 63f.
products of intellect , 476ff .; sales
Limitations of growth vs. CAVD intel
manship , 476ff .; speed , 32f ., 482 ;
lect , 468
spelling , 476ff .; width of intellect ,
Lincoln school , intercorrelations in
373 ff .
tests , 97f ., 106
Measurement , technique of , 27
Location of zero point , 485 ; by con
Measuring , individual CAVD intellect ,
sensus , 342ff .; by experiment , 340ff .
351ff .; higher processes by novel
Lower vs. higher thought processes ,
tests , 438
414f .; 473
Mechanical intelligence tests items , 64
Mechanical skill , 476ff .
MACPHAIL , A. H. , 433 , 534 , 540
MADSEN , J. N. , 465f ., 522 Median error of sum of ratings , 1 ,
MAY , M. A. , 99 38f . , 141
Maturity vs. training , 463ff . Memoirs , National Academy of Sci
Mazes as tests , 21 , 437f . ences , 97 , 291f . , 405 , 522 , 528 , 540 ,
558
McCall - Thorndike Reading Test , 7
Meaning , of Binet's mental age , 402 ; Mendelian determiner , 271
of 337f . Mentaces , 6
σ199
Measures , of difficulty from common Mental age , Stanford - Binet , 3 , 6 , 16 ,
reference point , 314ff . , 331ff .; of 17 , 97 , 402 , 447 , 575f .
improvability as measures of intel Mental products , 12ff .
lect , 409 Mentimeter Scale , 534
Measurement , errors of ( See Errors ) Method , of deriving units of measure ,
Measurement of , carpentry , 476ff .; 134 ; of obtaining composite distri
compositions , 476ff .; difficulty of butions , 521 , 527 , 529 , 533 , 535 ,
composite task , 119 ; discrimination 538f ., 545ff . , 549 ; proposed for con
of pitch , 476ff .; drawing , 476ff .; structing composites , 176ff .; used in
hearing , 476ff .; honesty , 476ff .; in constructing CAVD composites , 178 ;
clusive vs. exclusive traits , 480 ff .; used in constructing 10 -composite V
individualdifferences , 376f .; intel tasks , 179
lect by competent judgments , 39 ; Milan , Michigan , Alpha Tests , 491
612 INDEX
Miller Mental Ability Test , 534 Normal probability equation , 6 , 270 ,

MINER , Z. F. , 135 , 138 470, 521
Minnesota , Opposites and Completion Normal probability surface , 470 , 527 ,
of Definitions Test , 546 ; Recogni- 585
tion Vocabulary Test , 540 , 546 Normality of distribution , in adults ,
Mixed Relations Tests , 9 274ff .; in age groups , 287ff ., 293 ;
MONROE , W. S. , 522 in group , arguing skewness in gen-
MORGAN , J. J. B. , 540 eral , 224 , 294 , 296 ; in grade groups ,
Morgan Mental Test , 540 224 , 294 , 296
Motor Skill , 476ff . NORSWORTHY , N. , 9
MYERS , C. E. , 522 Notation for width of intellect , 337
MYERS , G. C. , 522 Novel tasks as tests , 18f ., 457f .
Myers Mental Measure , 41 , 228 , 493 , Number of connections as related to
510 , 522 , 528 , 530 , 533 ; scores in ontogeny and phylogeny of intellect ,
equal units , 254 , 316 432
National Academy of Sciences , Mem- Number of tasks at various levels , 457 ,

oirs , 97 , 291f ., 405 , 522 , 528 , 540 , 472
558 Numerical data as tests , 20f .
National Intelligence Test , 1 , 2 , 15 , 18 ,
41 , 52 , 99 , 107 , 222 , 227 , 274 , 291 , Officers ' ratings and Army Alpha
403 , 404 , 408 , 464 , 475 , 493 , 522 , 528 , Scores , 405
530 , 533 , 576f .; distributions of , in One idea defined , 418f .
equal units , 276f .; selection of tasks Ontogeny of intellect cf. number of
for , 409 ; scores in equal units , connections , 432
239ff . , 307 , 309 , 316 Order , of intellectual difficulty , 27 , 62 ,
National Research Council , 9 65
Natural zeros , 487f . Original , capacity and CAVD intellect ,
Nature of intelligence , 412ff . 435 ; capacity to be correct as due
Nature vs. nurture , 95f ., 436 ; as mea- to number of connections , 416 ;
sured by tasks , 441f .; as related to in ,
nature , individual differences
altitude and width , 458
421 ; vs. acquired ability , 433
Need, for knowledge of number of
OSBURN , W. J. , 528 , 534
tasks at each level , 379 ; for wide
OTIS , A. S. , 522 , 528
range and large population to find
variability at levels of ability , 499 ; Otis , Advanced Examination , 14f ., 41 ,
52 , 228 , 274 , 291 , 403 , 404 , 464 , 493 ,
for zero as point of reference , 339
Negative , acceleration , of level with 522 , 528 , 530 , 533 , 534 , 577 ; scores
age , 467 , 475 ; and positive factors in equal units , 245ff ., 276ff . , 307 ,
in the constitution of intellect , 50 ; 309 , 316 ; Intermediate Test , 599 ;
skewness in special class group , 580
Primary Examination , 41 ; Self-
Neural mechanisms vs. thought proc- Administering Group Test , 96f ., 99 ,
106 , 405ff .
esses , 415
NOBLE , E. L. , 539 Over -estimation in consensus ratings ,
Normal distribution , 55 , 115 143 , 157
Normal probability curve , 6f ., 53 , 270 ; Over -lapping as measure of correspon-

as distribution of Grade IX intel- dence between single task and intel-
lect , 521 lect , 131f .
INDEX 613
Parabola as curve of altitude and age , periments to locate zero intellect ,

466 339ff .
Partial correlation , between speed , Proportional vs. absolute counts of
altitude , 401 ; affected by
errors , tasks , 383f .
error in measures , 454f .; in Burt Psychological theory criteria for test
tests , 451ff .; with intellect , of ele items , 63
ments in a composite , 128f .
Qualitative differences in intellects ,
PATERSON , D. G. , 540 , 546
421
PEARSON , K. , 527
Pearson , correlation cf. Sheppard , 97 ,
Quality of intellect vs. quantity of
connections , 415ff .
99 , 101 , 298 , 323f ., 391f . , 557 , 559 ,
Quality of response , as criterion of de
565ff . , 570 ; formula for correction
gree of intellect , 416 ; vs. success
for range , 556
Percent successful as related to diffi failure marking , 477 , 479
Quantity hypothesis of intellect , 415ff .
culty , 37 , 351ff .
Phylogeny of intellect cf. number of Quickness of intellect ( See Speed )
connections , 432
Range of intellect ( See Width )
Physiological , cause of intellect , 412 ,
Range , restricted , correlations found
420f .; facts of intellect , 473
in , 297
Pictorial data as tests , 20ff .
Rank order of difficulty for test ele
PINTNER , R. , 135 , 522 , 528
ments , 38
Pintner Non -Language Examination ,
Ratings for difficulty of test ele
222 , 228 , 522 , 528 ; scores in equal
ments , 138ff .
units , 254ff . , 316 Reading composites , construction of ,
Positive and negative factors in the 193ff .
constitution of intellect , 50
Reasoning tests , Burt's , 447ff .
Practice and growth effects in Alpha Reference point in CAVD , 295 , 314ff . ,
scores
, 491
331ff .
PRESSEY , S. L. , 522 , 528 Regression , equation in Burt results ,
Pressey Mental Survey , 522 , 528 ;
447 , 453 ; line in ability vs. varia
scores in equal units , 316 bility , 504
Princeton Intelligence Examination , Relation , between altitude , width ,
540 , 546 area, speed , 388ff . , 400f
.; between
Principles , for separating original ability and variability , 43 , 104 ,
from acquired abilities , 436 ; of in 497 ff .; between Binet M.A. and
tellectual difficulty , 22 , 24 , 30f .; of CAVD level , 402 ; between tasks and
measurement of human abilities , heredity , age , sex , 484f .; of CAVD
476ff . to other intelligence examinations ,
Probability curve , 6f ., 53 , 270 , 470 , 96ff . , 105 ; of error of measurement
521 , 527 , 585 to initial ability and gain , 408f .
Probable error , of determination vs. Relational thinking as intelligence , 19
unreliability , 520 ; of medians in Relative , magnitudes of levels A and
ratings , 144f .; of sum of ratings , Q , 337f .; variability of grade popu
138f ., 141 lations , 306 ; vs. absolute counts of
Product of intellect defined , 26 tasks , 383f .
Program , for measuring tasks of small Reliability , of judgments , 140 ; of
intellectual difficulty , 341 ; of ex ratings , 138f .; of score determina
614 INDEX
tion , 140 , 410 ; of test material , Sensori -motor connections and intelli
129f ., 324ff . , 567 gence , 418 , 476ff .
Remoteness , effect of , on correlations , Sheppard's correlation coefficient cf.
557, 562 Pearson's , 97 , 99 , 101 , 298 , 391 ,
Repeated , Alpha tests , 491 ; Courtis 392 , 557 , 559 , 565ff ., 570
tests , 492 ; Woody tests , 493 Shifting curves of percent correct ,
Restricted range , correlations found 353ff .
in , 297 Shrinking of units , 44
ROBINSON , E. E. , 135 , 138 Sigma , as unit of measure , 43 , 50 , 53 ,
ROGERS , A. L. , 438 , 540 54 , 57 , 109ff . , 225 ; of army in Stan
ROGERS , D. C. , 540 , 546 ford M.A. , 97 ; of CAVD altitude
Rogers Syllogism Test , 438 score , 411 ; of errors , 126 ; of group
ROWELL , D. C. , 403 , 463 as unit , 295ff ., 302 , 323f .; of imbe
RUCH , G. M. , 534 , 540 , 546 cile group , 154f .; of sum of ratings ,
RUGER , G. J. , 135 138f .
Sigmas , of groups in common unit ,
Rule for estimating CAVD altitude of
303 ff.; of various tests in common
individual , 369
unit , 317
Salesmanship , 467ff . Significance of scores in stock exami
Sampling , error of , 109f ., 112 , 114 , nations , 403f .
SIMPSON, B. R. , 9
117f .; sufficient , for correlation of
Single tasks and intellect , 109ff ., 131ff .
one level with entire series , 565
Satisfyers and annoyers , 489f . Skewness , 271 ; cf. differential gain in
score , 290 ; in adult distribution ,
Scale , CAVD , 294ff .; CAVD Levels
A- D , CAVD Levels
66ff .; - M, I 287ff ., 293 ; in special
580
class group ,
160ff .; CAVD Levels N -Q , 76ff .

Smith College Examination , 540 , 546
Scales for composition , designs , draw
SMITH , R. M. , 540
ing , handwriting , 134 Smoothing for chance error , 299
Scores in tests in equal units , 43ff ., Social intelligence items as data for
105 , 224ff . , 306ff ., 403 , 475 , 497ff .,
tests, 64
576 , 584 , 586 , 592 Space - forms as data for tests , 20f .
Scoring , by success - failure vs. graded
SPEARMAN , C. , 9 , 19 , 110ff . , 391 , 394 ,
credit , 477 , 479 ; products of intel 559
lect , 22f ., 29 Spearman -Brown correction for self
Selection of word -knowledge tasks , 183 correlation , 298 , 323 , 394 , 425f ., 556 ,
Selective vs. associative thought , 414f . 559 , 569 , 570
Self -correlation , of altitude measure , Specialization , in CAVD levels , 397 ;
389ff .; of Army Alpha , 318 ; of in intellect within levels , 390f .
composites , 297 , 325 ; of Haggerty Speed , as measure of intellect , 24f . ,
test , 318 ; of I.E.R. tests , 101 ; of 32 , 104 ; factor in stock intelligence
Otis tests , 99 ; of percent s at levels , examination , 404f .
391ff .; of ratings , 138 ; of sub Spelman Memorial Fund , 442
series in CAVD , 107 ; of task , 60 ; Spelling , tests in, 476ff .
of Terman Group Test , 317f .; of Stanford - Binet , 1 , 3 , 8 , 15 , 22 , 38 , 41 ,
Thorndike Intelligence Examination 55 , 96 , 97 , 99 , 106f ., 433 , 439 , 475 ,
H.S.G. , 97 ; of various combined 493 , 572f . , 576 ; speed and altitude ,
tests , 318 405ff .
INDEX 615
Status of neurones cf. number of con gan , Myers , National , Otis , Pintner ,
nections , 432 Pressey , Princeton , Rogers , Smith
STECHER , L. M. , 495 College , Stanford -Binet , Stenquist ,
STENQUIST , J.
L. , 9 , 530 Strickland , Substitution , Terman ,
Strength of connections , 489 Thorndike , Thorndike - McCall , Thurs
Strickland Test , 534 tone , Trabue , VanWagenen , Word
STRICKLAND , V. L. , 534 Knowledge , Woody , Yale )
STRONG , E. K. , 134 Tests , indicative of original capacity
Sub -series , of CAVD tasks , 66ff ., vs. nurture , 442ff .; more or less sus
160ff .; of tasks in levels , 484 ceptible to training , 440f .; types of ,
Substitution Test , 222 ; as measure of 9, 20f.
ability to learn , 408f . Tetrachoric r , 390 .
Success - failure vs. graded credit Theorems on measurement of intellect ,
scoring , 477 , 479 30ff .
Sufficient sampling for correlation of THOMSON , G. H. , 135 , 138 , 409
one CAVD level with entire series , THORNDIKE, E. L. , 112 , 134f ., 138 , 288 ,
565 408 , 521
Summary of conclusions , 469ff . Thorndike Intelligence Examination
Swelling and shrinking of units , 44 for High School Graduates , 42f . ,
Symmetry in distribution of intellect , 96f . , 107 , 130 , 494 , 497 , 500 , 503ff . ,
270f . 515 , 518 , 534 , 540 , 546 , 586
Thorndike - McCall Test of Reading , 7
TAPE , H. A. , 491 Thorndike Visual Vocabulary Test , 510
Tasks , intellectual , 26 , 59ff ., 413f ., Thought processes vs. neural mechan
469 ; for location of zero by con isms , 415
sensus , 343ff .; to measure altitude , THURSTONE , L. L., 79 , 540 .
183
Thurstone Psychological Examination ,
Technique of measurement of intellect , 540
TILTON , J. W. , 35 , 422f ., 430 , 443

27
Techniques of measurement of human Trabue Completion Tests , 119 , 522 ,
abilities , 476ff . 528 ; scores in equal units , 316
Terman Group Test , 42 , 52 , 96f ., 99 , Trabue Mentimeter , 228
106f . , 228 , 494 , 534 , 577 ; scores in TRABUE , M. R. , 522 , 528
equal units , 250ff ., 309 , 316 ; speed Training , and CAVD intellect , 435f .;
and altitude , 405ff . on test items , 437 ; vs. maturity ,
TERMAN , L. M. , 8 , 9 , 16 , 183 , 433f ., 463 ff .
576 Transformation of scores into equal
Terms expressive of altitude and area units , 224ff ., 306ff .
of intellect concept , 379 Transmutation of arbitrary scores to
Tests ( See Analogies , Army Alpha , make variabilities equal , 520
Army a, Army Beta , Binet , Briggs , True value vs. face -value scores , 3f . ,
Brown University , Buckingham ,
16 , 43 , 53 , 224ff ., 497ff .
Burt , CAVD , Courtis , Dearborn , Truth as criterion for value , 14f .
Drawing , Ebbinghaus , Form -boards ,
General Electrical , Haggerty , Hand Under -estimation in consensus ratings ,
writing, Herring , I.E.R. , Illinois , 157
Iowa , Kelley - Trabue , Mazes , Menti Unequal units , 306ff.

meter , Miller , Mixed Relations , Mor Unitary cause of intellect , 412
616 INDEX
Units , equal, in intelligence tests , cause as measure of intellect , 412

43 ff ., 105 , 224ff . , 403 , 475 , 497ff ., Variety in test forms , 21f.
576 , 584 , 586 , 592 ; inequalities in , Various sigmas in a common unit ,
306 , 526 ; of measure , method of de- 303f . , 328ff .
riving , 134 ; swelling and shrinking Verbal data as intelligence tests , 20f .
of, 484 Verification of quantity hypothesis of
Unity of intellect , 63 ; CAVD , 399 intellect , 422ff .
Unreliability , of judgments , 138f .; of VINCENT , L. M. , 122 , 129ff . , 135 , 138
ratings , 140 ; of score determination , Vocabulary levels , 180ff .
122 , 410 ; of test material , 129f . Volume or area of intellect , 378ff .
Use , of arrays to measure difficulty ,
40 ; of tests more and less sus-
WATSON, J. B. , 272
ceptible to training , 440f .
WHIPPLE , G. M. , 9 , 433ff ., 522
Validity of Width of intellect , 25 , 31f ., 33 , 104 ,
consensus of judges in
ratings , 141 469 ; and nature vs. nurture , 458 ;
and stock examination score , 403f .;
Valuation , of mental products , 12ff .;
as related to altitude , 473 ; as re-
of responses in measurement , 477
lated to number of tasks at a level ,
Van Wagenen History Scales , 119
Variable errors in consensus ratings , 379 ; in sample vs. in entire series ,
142 374 ; notation for , 377 ; vs. altitude ,
Variability , for various grade popula- 387ff .
tions , 311 ; for various tests , 308ff .; WOOD , B. D. , 135 , 225
in an array , 56 ; in correlation for WOODWORTH , R. S. , 9

single tasks and intellect , 122 ; in Woody Tests , 493
experts ' opinion of easy tasks , 349 ; WOODYARD , E. , 135 , 138 , 410
in face -value vs. equal unit scores , Word knowledge , composites , 186ff .;
308 ff .; in grade 6 vs. college stu- levels , 180ff .
dents , 132 ; in grade 9 vs. age 20 ,
562 ; in Grades VI to XIII , 306 ; in Yale Examination , 541
Grades VI , IX, XII , 104 ; in low YERKES , R. M. , 9

level groups , 325ff .; in sigmas , 136 ;
of group as affecting correlations , Zero , deviations from median , 494 ;
126f .; of group as unit of measure , for Binet scale , 403 ; in discrimina-
109 , 493 ; of individual , 40f ., 43ff ., tion of pitch , 486 ; influence of re-
56 , 61 , 104 , 470f .; of individual moteness on correlation , 556 ; intel-
measured from correlation array , lectual difficulty task , 348f .; per-
502 ; of literates in draft , 562 ; vs. cent correct , 378 ; point , 4 , 27 , 36 ,
ability , 497ff . 294 , 336f ., 339ff ., 441 , 471f ., 482 ,
Variation , chance , eliminated , 590ff .; 485f .; points , natural and conven-
in the I.Q. , 495f .; in judgments , tional , 487f .; training upon test
136 ; in magnitude of a unitary items , 437

1927 Thorndike

Uploaded by

Copyright:

Available Formats

1927 Thorndike

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

1927 Thorndike

Uploaded by

Copyright:

Available Formats

THE MEASUREMENT OF INTELLIGENCE

EDWARD L. THORNDIKE , E. O. BREGMAN , M. V. COBB ,

The investigations and results reported

We are indebted , for most courteous and efficient cooperation ,

Other attempted simplifications of the process of valuation

The form of tests of intellect 21

Scoring the products intellect 22

Further facts concerning difficulty

Width or extent or range ..... 31

The relative altitude extent

CHAPTER The Measurement of Difficulty

The present status 37

Measurement differences knowledge

of the form of distribution of the variations of an individual

The relation of the variability of an individual his amount

Measurement by way of the form of distribution intellect

some defined group 51

Measurement by way of the form distribution an array

The defects of the Measurements so far described 56

CHAPTER III The Measurement the Intellectual Difficulty

of Tasks and of Level of Intellect More Rigorous and

The homogeneity of difficulty CAVD ... 101

CHAPTER IV. The Measurement of the Intellectual Difficulty of

The correlations of single tasks with measures of intellec 122

CHAPTER The Measurement of the Intellectual Difficulty

The Experiments . 135

The Ratings 138

CHAPTER VI Levels of Intellect

Composite tasks 160

The construction of composite tasks 175

10 composites word knowledge or 179

The construction composite tasks sentence completion

arithmetical problems and the understanding of sentences

and paragraphs 193

The difficulty of the 10 composites 208

The combination of 10 composites into 40 composites 211

CHAPTER VII The Transformation of the Scores Standard

Intelligence Examinations into Terms of Scales with

The method of transformation illustrated by the Thorndike

Examination and Army Alpha .......... 224

The Otis Advanced Examination ............ 245

The Terman Group Test 250

The evidence 274

CHAPTER IX for Measuring Altitude Intellect

Expressing the of each group terms of common unit

The difficulty of composites and ... 321

Estimating from 323

Expressing the each group 328

Expressing the measures difficulty distances from com-

mon point of reference..... 331

CHAPTER X. The Absolute Zero Intellectual Difficulty

Locating zero difficulty by experiment 340

program of tasks use measuring tasks of very little in-

tellectual difficulty 341

CHAPTER XI The Measurement the Altitude an Individ-

The form of the curve of percent correct relation difficulty 351

Estimating the CAVD altitude of an individual 365

CHAPTER XII The Measurement of Width Area of

Width of intellect the case of truly intellectual task- 373

CHAPTER XIII The Relations Altitude Width Area and