THE INTERNAL STRUCTURE OF THE MMPI*

WILLIAM MARSHALL WHEELER

VETERANS ADMINISTRATION NEUROPSYCHIATRIC HOSPITAL, LOS ANGELES, CALIFORNIA

KENNETH B. LITTLE AND GEORGE F. J. LEHNER

UNIVERSITY OF CALIFORNIA, LOS ANGELES

T HE MMPI is a widely used diagnostic

tool. Many studies have been made con-
cerning its use in various situations and
with various types of cases, its relationship to
hypochondriasis (Hs), depression (D), hys-
teria ( H y ) , psychopathic personality ( P d ) ,
paranoia ( P a ) , psychasthenia (Pt), mascu-
linity-femininity of interest (Mf), schiz-
other tests, the reliability and validity of the ophrenia ( S c ) , and hypomanic trends (Ma).
various scales, and the relationships among the In addition to these, three other scales, the
scales as reflected in "patterns" or "profiles" of question score (?), the lie score (L), and a
scores. Little, however, seems to have been done validity score (F), are used, respectively, to
to determine the nature and extent of the assist in assessing for a subject whether he
interrelationships among the various scales of understands the items, whether he tries to
the total test. Information concerning the in- place himself in too favorable a light, or
tercorrelations of the various diagnostic scales whether he is cooperative. Recent forms of
is needed. The present study attempts to the test include an additional scale (K) de-
provide information about scale interrelation- scribed as a "correction factor" for sharpen-
ships by utilizing a factor analysis approach. ing the discriminatory value of several of the
An examination of the literature fails to show other scales. The total of 504 items utilized
any factorial study of the scales of the MMPI, in these various scales represents 60 (arbi-
although in a study by Cottle [2] a factorial trarily) defined areas of human activity as
analysis was made of the MMPI in relation to described by Hathaway and McKinley [5].
the Strong, the Kuder, and Bell Inventories. In interpretation, items responded to as
Another aim of the present study is to ap- either "True" or "False" are scored accord-
praise statistically the possibility of making dy- ing to empirically derived keys and translated
namic interpretations in terms of different into a standard scale system with an average
scale scores. If, for example, one factor were value of 50 and a borderline value of 70. The
found, it would imply that the test measures number of items that a subject does not
degree of disturbance, not kind of disturbance, answer "True" or "False" determines his f
indicating that the test served mainly as a score. In general, the higher the score is above
screening device. If, on the other hand, several 50 the more significant is the deviation. The
different factors should emerge, each fairly meaning of scores below 50 is at present not
closely identified with a scale, the use of the clear.
test in its present form for differential or dy- As is pointed out by Schiele, Baker, and
namic interpretations would be supported. Hathaway [9], the scales were compiled by
comparing the responses of clinically diag-
DESCRIPTION OF MMPI
nosed patients with those of persons not
The scales available in the MMPI are under psychiatric care. The particular items
characterizing a symptom complex were
*Reviewed in the Veterans Administration and identified by the contrasting tendency for
published with the approval of the Chief Medical
Director. The statements and conclusions published normal and abnormal patients to respond
by the authors are the result of their own study and "True" or "False," regardless of the verbal
do not necessarily reflect the opinion or policy of
the Veterans Administration. content of the item.
134
 THE INTERNAL STRUCTURE OF THE MMPI 135

Each scale was designed to provide a mea- UCLA.

sure, in terms of a score, of the strength of Although a short 373-item group form was
a certain trend or component in personality used, all K scale items were included in the
makeup. High scores on any of the scales test, thus including all scored items.
purportedly indicate deviations in terms of To measure the interscale relationships of
which differential diagnoses may be made. the MMPI, or, more accurately, to investi-
Although the MMPI has been described as "the gate the extent to which the scales measure
first inventory measuring common specific the same personality factor or factors, Thur-
clinical syndromes, in contrast to the earlier stone's centroid method of multiple factor an-
schedules designed for either the more general alysis was employed. All computations of the
concept of 'neuroticism' or the special states Pearsonian correlation coefficients for the two
like 'inferiority,' " [9, p. 293] this assertion groups were made on the raw scores, rather
has not been tested. In the light of the data to than the T scores of the scales.
be reported from the present factor analysis Preliminary to the use of Thurstone's tech-
study, this statement is questionable—partic- nique in the present investigation, a pilot
ularly if it implies much emphasis on "specific" study was made on two groups of 30 and 53
clinical syndromes for differential diagnosis. cases utilizing a statistical technique developed
As a matter of fact, work such as that of by Gengerelli [3]. His method suggested
Meehl [8] and Gough [4], who discuss the that most of the total variance of the MMPI
clustering of scores of certain scales, and who could be explained in terms of the variance of
speak of the "neurotic scales" (Hs, D, and two or three scales. For example, these prelimi-
Hy), and the "psychotic scales" (Pa, Pt, and nary studies showed that the correlation be-
Sc), also suggests considerable overlap of the tween the Sc and Pt scales was above .80, sug-
various "specific" scales frequently employed gesting that the two tests are measuring the
for differential diagnoses. The discovery of same personality variable. The correlational
the "neurotic triad" and the "psychotic triad" matrix obtained by Cottle [2] shows an r of .84
in clinical work reinforces the need to make a for these two scales, as well as other correla-
mare detailed study of individual scale load- tions reflecting the same type of relationship
ings and their contribution to total test score among various scales.
variance. Although the authors of the test Since many of the scales contain items also
suggest that interrelationships among scales appearing in another scale or scales, the pos-
indicate "the dynamic interrelationships of sible effect of overlapping items on the pre-
different clinical syndromes" [6, p. 3], a more sent investigation was considered. Table 1
parsimonious interpretation suggests that these indicates, above the principal diagonal, the r's
relationships indicate the degree of psycholog- among scales based on the amount of item
ical identity of the scales. overlap. Below the diagonal are the number of
PROCEDURE items in common minus those items that are
scored in the opposite fashion for the respective
The MMPI, group form, was administered scales. The two halves are, of course, merely
to 110 neuropsychiatric male patients in a different representations of the same rela-
large Veterans hospital in Los Angeles, Cali- tionship. Along the principal diagonal (ital-
fornia. These patients ranged in age from 20 icized figures) is given the number of items
to 63, with a mean age of 33. The cases re- in each scale. This table is presented to
presented a random selection of neurotic and indicate the a m o u n t of communality any
psychotic patients (and a few cases in which two scales have as a result of identical items.
organic factors were also present) tested at At first glance a correction of the empir-
the hospital during the year 1948. A detailed ical correlations would seem to be indicated.
nosological breakdown of the cases was not However, the consideration of a correlation
attempted in view of the usual unreliability coefficient as an index of percentage of com-
of such categorizing. The comparison group mon elements, of which physical identity is
consisted of 112 male college students at only one kind, leads to the conclusion that
136 W. M. WHEELER, K. B. LITTLE, AND 0. F. J. LEHNER

TABLE 1
AMOUNT OF OVBLAP OF SCALES DUE TO COMMON ITEMS

L K F Hi D Hy W Mf Pa Pt Sc Mil
L 15 .05 .03 .00 .08 .03 .00 .03 —.04 —.04 —.03 .04
K l 30 .02 .00 .09 ,24 .16 .08 .06 —.05 .02 .08
F 1 1 64- .02 .03 .01 .09 .03 .18 .02 .21 .01
Hi 0 0 1 33 .18 .46 .02 .00 .03 .05 .08 .00
D 2 4 2 8 60 .20 .14 .03 .04 .26 .10 .05
Hy 1 10 1 20 12 <50 .18 .06 .08 .13 .12 .00
Pd 0 6 5 1 7 10 50 .02 .18 .12 .16 .08
Mf 1 3 2 0 2 4 1 60 .04 .02 .06 .02
Pa —1 2 9 1 2 4 8 2 40 .09 .23 .05
Pt —1 —2 1 2 13 7 6 1 4 48 .25 .06
Sc —1 1 15 4 8 8 10 4 13 15 78 .18
Ma 2 3 1 0 —3 0 4 1 2 3 11 46

extraction of these would be based on spurious ratio for this difference. All differences are
reasoning. In other words, positive item overlap significant beyond the 1 per cent level of con-
is one way in which the scales are related and fidence.
should not be interpreted as detracting from As can be seen from these data, the NP
empirical relationships based on scores obtained group scores significantly above the college
on the scales. group on all scales expcept K and Mf, where
they score significantly below, confirming the
RESULTS value of the test as a general screening device.
The mean values obtained for each group The intercorrelations obtained for the 12
and the combined groups for each of the scales used (the question scale was omitted
MMPI scales are given in Table 2. Cottle's since no records with a question score greater
means are included for comparison. Also pre- than 25 are included in the present data) are
sented are the differences between the college presented in the correlation matrices in
and NP means on each scale and the critical Tables 3, 4, and 5.

TABLE 2
MEAN VALUES FOR COLLEGE AND NEUROPSYCHIATRIC GROUPS AND THEIR DIFFERENCES FOR EACH
SCALE OF THE MMPI (THE COMBINED MEAN VALUES AND THE MEAN VALUES FROM
COTTLE'S DATA ARE PRESENTED FOR COMPARISON)

College Neuropsychiatric
N == 112 Ns= :110 N = 222
Mean SD Mean SD Diff SDdlff CR* Combined Cottle's
Scale Mean Mean
L 3.3 2.3 5.1 2.6 1.8 .33 5.45 4.2 3.8
K 16.4 4.7 14.5 5.3 1.9 .67 2.83 15.5
F 4.7 3.1 8.6 7.1 3.9 1.15 3.39 6.6 —
4.0
Hi 3.9 3.2 11.3 7.0 7.4 1.18 6.27 7.6 5.5
D 19.8 5.2 26.5 7.6 6.7 .88 7.61 23.1 19.5
Hy 20.7 4.3 25.7 7.8 5.0 .85 5.88 23.2 21.7
Pd 15.6 4.3 21.4 5.7 5.8 .63 9.21 18.4 15.3
Mf 27.7 4.3 24.7 4.8 3.0 .62 4.84 26.2 25.3
Pa 8.9 2.7 11.7 5.0 2.8 .54 5.18 10.3 9.2
Pt 10.0 7.4 16.9 10.5 6.9 1.18 5.85 13.4 12.4
Sc 10.0 6.9 18.5 12.3 8.5 1.31 6.49 14.2 10.8
Ma 16.9 4.4 18.8 5.6 1.9 .67 2.83 17.8 17.9
"All differences are significant at or beyond the .01 level.
 THE INTERNAL STRUCTURE OF THE MMPI 137

TABLE 3
CORRELATIONS AMONG THE TWELVE SCALES OF THE MMPI FOR 112 MALE COLLEGE STUDENTS

L K F Hs D Hy Pd Mf Pa Pt Sc Ma
I .366 .046 —.026 .076 .243 —.102 —.029 .127 —.299 —.334 —.151
K .366 —.105 —.305 —.068 .499 —.090 —.162 .193 —.587 —.510 —.297
F .046 —.105 .489 .561 .310 .554 .178 .301 .457 .674 .194
Hs —.026 —.305 .489 .560 .372 .417 .191 .212 .577 .554 .162
D .076 —.068 .561 .560 .458 .414 .260 .383 .570 .484 —.272
Hy .243 .499 .310 .372 .458 .348 .183 .399 .049 .077 —.075
Pd —.102 —.090 .554 .417 .414 .348 .153 .303 .420 .590 .308
Mf —.029 —.162 .178 .191 .260 .183 .153 .332 .417 .391 .124
Pa .127 .193 .301 .212 .383 .399 .303 .332 .234 .304 .014
Pt —.299 —.587 .457 .577 .570 .049 .420 .417 .234 .821 .181
Sc —.334 —.510 .674 .554 _--4*K .077 .391 .304 .821 .348
Ma —.151 —.297 .194 .162:•''—.272 )--.075 , - .124 .014 .181 .348

TABLE 4
CORRELATIONS AMONG THE TWELVE SCALES OF THE MMPI FOR 110 MALE
NEUROPSYCHIATRIC PATIENTS

L K F Hs D Hy Pd Mf Pa Pt Sc Ma
L .302 —.097 .138 .027 .134 —.121 —.085 .028 —.285 —.194 —.162
K .302 —.404 —.201 —.185 .138 —.383 —.190 —.261 —.655 —.550 —.451
F —.097 —.404 .334 .154 .021 .507 .300 .690 .536 .764 .411
Hs .138 —.201 .334 .680 .759 .405 .258 .477 .523 .516 .271
D .027 —.185 .154 .680 .690 .453 .346 .346 .635 .469 .038
Hy .134 .138 .021 .759 .690 .298 .257 .299 .344 .222 .075
Pd —.121 —.383 .507 .405 .453 .298 .297 .615 .590 .604 .515
Mf —.085 —.190 .300 .258 .346 .257 .297 .457 .344 .325 .184
Pa .028 —.261 .690 .477 .346 .299 .615 .457 .538 .703 .436
Pt —.285 —.655 .536 .523 .635 .344 .590 .344 .538 .857 .461
Sc —.194 —.550 .764 .516 .469 .222 .604 - .325 .703 .857 .539
Ma —.162 —.451 .411 .271 /:o3T\ .075 /" .515 \ .184 .436 .461 ,539
i / J

TABLE 5
CORRELATIONS AMONG THE TWELVE SCALES OF THE MMPI FOR MALE COLLEGE STUDENTS AND
MALE NEUROPSYCHIATRIC PATIENTS COMBINED

L K F Hs D Hy Pd Mf Pa Pt Sc Ma
L .241 .070 .262 .226 .274 .081 —.162 .169 —.130 —.066 —.078
K .241 —.343 —.282 —.205 .162 —.315 —.109 —.163 —.639 —.546 —.408
F .070 —.343 .471 .360 .195 .580 .118 .657 .566 .775 .383
Hs .262 —.282 .471 .728 .730 .566 .002 .519 .607 .618 .299
D .226 —.205 .360 .728 .683 .565 .119 .451 .672 .565 .025
Hy .274 .162 .195 .730 .683 .435 .082 .408 .357 .307 .099
Pd .081 —.315 .580 .566 .565 .435 .041 .591 .606 .670 .468
Mf —.162 —.109 .118 .002 .119 .082 .041 .254 .218 .173 .088
Pa .169 —.163 .657 .519 .451 .408 .591 .254 .516 .660 .347
Pt —.130 —.639 .566 .607 .672 .357 .606 .218 .516 .864 .398
Sc —.066 —.546 .775 .618 .565 .307 x:#£ .173 .660 .864 .504
>
Ma —.078 —.408 .383 .299 Xi025~ 1 .099 ( .088 .347 .398 .504
138 W. M. WHEELER^ K. B. LITTLE, AND G. F. J. LEHNER

A comparison of the two groups reveals published for the scales [7] reveals that vari-
significant differences for 17 correlations as ous pairs of scales show correlations that ap-
shown in Table 6. Relevant correlations from proximate the reliability coefficients obtained
Cottle's matrix are also presented. for either of the members of the respective
pair. This would seem to indicate that the
TABLE 6 particular scales in question do not measure
SIGNIFICANT* CHANGES IN SCALE RELATIONSHIPS different things in spite of the different labels
FROM COLLEGE TO NP SUBJECTS applied to them and in spite of the differenti-
r College NP Diff. Cottle's Data ating functions imputed to them. For ex-
__ ample, the correlation between Sc and Pt of
r
XF —.105 —.404 —.299
r
KHy .499 .138 —.361 .86 from the table of data on the combined
r
KPd —.090 —.383 —.293 — groups, if corrected for attenuation on the
"KPa .193 —.261 —.454 — basis of Holzberg's reliabilities, which are the
r
FD .561 .154 —.407 —
.344 highest published to date, ris.es to the improb-
'FHy .310 .021 —.289 .159
r
FPa .301 .690 .389 .206 able figure of 1.08, while the correlation be-
r
HsHy .372 .759 .387 .588 tween Hs and D of .73 rises to the value of
"HsPa .212 .477 .265 .183 .99.
r
DHy .458 .690 .232 .415 In Tables 7, 8, and 9 are presented the
'DMa —.272 .038 .310 —.083
factor loadings after rotation for the college
'HyPt .049 .344 .295 .182
r
PdPa .303 .615 .312 .329 group, the NP group, and for the two groups
'PaPt
f
.234 .538 .304 .363 combined.
PaSc .304 .703 .399 .415
"PaMa .014 .436 .422 .112 TABLE 7
r
PtMa .181 .461 .280 .357 FACTOR LOADINGS AFTER ROTATION,
•To compute significance of differences between cor- COLLEGE GROUP (N = 112)
relations, they were transformed to z values. All differ- h*
ences given in the table were significant at or boyond I II Ill
_ IV
_
the .05 level. L .351 .242
K —.630
— .578 .740
F .590 .450 — — .592
Two general conclusions can be drawn: — —
Hs .627 .345 .538
(1) Cottle's group tends to fall between our D .503 .530 —
— —
.595 .894
college group and our NP group not only in Hy .780 .615
mean scores but also in terms of the direction Pd —
.556 .425 — — .581
Mf .358 —
.538 —
— .441
of change of relationships among scales; (2) —
Pa .510 .339 — .413
the trend of change of relationships between —
Pt .908 — .926
scales is a consistent one and deserves further Sc .943 — — .914
investigation to determine its implications. For Ma — —
— —
—.595 .452
example, one speculation can be made about the — —
changes in relationships of the Pa scale. Pa TABLE 8
has its maximal relationship with Hy in the FACTOR LOADINGS AFTER ROTATION,
NP GROUP (N = 110)
normal group and with Sc in the NP group.
With Cottle's group, which included indi- I II III IV h2
viduals seeking psychological aid, the relation- L . .234
— —
ship of Pa to Sc is between that obtained for K —.702 .643
F .668 — —
.458 .700
the two groups in the present study. These —
Hs .529 .740 .873
relationships would seem to substantiate the D .481 .648 — .741
interpretation that paranoid reactions as mea- Hy .845 — .771
— —
sured by the items in this scale can be part of Pd .700 — .551
two different syndromes, perhaps a neurotic Mf .381 — .338
Pa .632 — —
.591 .780
and a psychotic. Pt .936 — .944
A comparison of the correlation coefficients Sc .925 — — .878
obtained here with the reliability coefficients Ma .620 — — .494
— —
 THE INTERNAL STRUCTURE OF THE MMPI 139

TABLE 9 ego-defenses are intact. Perhaps one of these

FACTOR LOADINGS AFTER ROTATION, ego-defensive mechanisms is indicated by the
COMBINED COLLEGE AND NP GROUPS (# = 222) positive loading of .510 on the Pa scale. This
I II h*
suggests that the paranoid projections serve
Ill IV
„.
more as a neurotic defense in the normal
L .469 .307
group than as a component in the schizoid
K —.670
— .633
F .688 — —
.467 — .707 pattern indicated in Factor I. We shall see in
Hi .600 —
.640 — .786 a moment that the Pa scale disappears from
D .554 .680 — — .820 Factor II in the NP group and has a loading
Hy .789 — — .747
— — in Factor I.
Pd —
.682 .416 .718
Mf — — .119 The third factor found in the normal
Pa —
.602 —
__ —
.560 —
_ .728 group has its only significant loadings on Mf
Pt .936 — — .940
— and Pa, reflecting the masculinity-femininity
Sc .931 .907
variable and its possible relationships to para-
Ma .550 — — — .392
— — — noia— a relationship often stressed in psycho-
Normal Group. An examination of the analytic theory.
data for the normal group indicates that there The fourth factor has significant loading
are two major group factors clearly defined. on D (.595) and Ma (—.595), indicating a
The first factor has its maximal loadings on bipolar relationship between these two scales.
the Sc and Pt scales, showing, respectively, This apparently indicates a dimension of
values of .943 and ,908. Other positive load- mood independent of the schizoid and neu-
ings on this factor are found in Hs, F, Pd, D, rotic patterns reflected in Factors I and II.
and to a slight extent, Mf. In contrast with Since scales D, F, and Pd have significant
these positive loadings, we find that the K scale loadings on both Factors I and II, we may
shows a high negative loading of —.630. infer that the kind of reactions typified by
An interpretation of this factor in the light these scales are parts of both the schizoid and
of these loadings would suggest that it indi- neurotic pattern.
cates primarily concern with one's self. The NP Group. The analysis of the data for
scales which are most heavily loaded, Sc and the NP group also indicates two major fac-
Pt, are two of the three commonly referred to tors. Factor I again has its maximal loadings
as the psychotic triad, and seem to reflect the on Pt (.936) and Sc (.925). Other signifi-
encapsulating withdrawal of a schizoid type cant positive loadings are found on Pd, F, Pa,
including excessive concern with compulsive Ma, Hs, D, and Mf, with a substantial nega-
needs. The extremely high negative loading tive loadings for K of —.702.
on the K scale would seem to indicate that A comparison of these loadings with those
when this factor is present to a marked degree found in Factor I for the normal group shows
in an individual, the usual ego-defensive considerable similarity. For example, Pt, Sc,
mechanisms are held in abeyance and the per- and K show very similar loadings. The dif-
son now tends to show himself in the worst ferences which appear are primarily on two
possible light. scales: (1) Hs scale has a lower loading on
The second major factor has its maximal the "psychotic" factor for the NP group than
loadings on Hy (with a value of .780) and, it does for the group of college males, reflect-
interestingly enough, on K (with a value of ing evidently the differences in defenses be-
.578). Other significant loadings on this fac- tween these two groups; (2) the other dif-
tor occur on scales D, Pa, F, Pd, L, and Hs. ference concerns the loadings of Pa. In the
This factor seems to reflect, on the basis of NP group the Pa scale has its heaviest loading
these loadings, the neurotic picture of adjust- on the same factor as the "psychotic" scales,
ment. The high positive loading of K for this whereas in the college group its heaviest load-
factor, as compared with the high negative ing is on the same factor as the "neurotic"
loading of K for Factor I, implies that the scales.
140 W. M. WHEELER, K. B. LITTLE, AND G. F. J. LEHNER

An examination of Factor II—the "neu- meaning of this drop is not clear. It may be
rotic" factor—for the NP group indicates that that this reflects a curvilinear relationship be-
it has its maximal loadings on only three scales: tween Mf and some of the other scales in the
Hy (.845), Hs (.740), and D (.648). Com- more heterogeneous population.
paring these loadings with those obtained for
the college group, we see that the Hs scale has SUMMARY AND CONCLUSION
a minimal loading (.345) for the normals, The results from the present factor analysis
whereas in the NP group it has a loading of of the MM PI seem to indicate that the am-
.740—the second highest loading for this fac- bitious goal of measuring specific clinical
tor. It is interesting to note further that in syndromes has not been completely achieved.
the NP group the K scale has no loading on The test permits diagnosis mainly in terms of
the neurotic factor, while in the normal group "neurotic" or "psychotic," but not in terms of
it has the second highest loading, indicating type of neurosis or psychosis or other more spe-
perhaps that in hospital patients the usual cific category. The results do, however, substan-
neurotic manifestations are accompanied by tiate the use of the MM PI for distinguish-
less of the type of "defensiveness" that K re- ing between neurotic and psychotic syn-
presents. It is well to remember, also, that the dromes. These two syndromes are here defined
loadings of the "neurotic" triad for the hospi- by their maximal loadings on the Hy and
tal group may have been determined by dif- Sc scales, respectively.
ferent items of the same scales than the load- Present findings would indicate that refined
ings of the neurotic triad in the normal differential diagnosis or the formulation of
group. A future item analysis may indicate dynamic personality descriptions on the basis
that the loadings of the "neurotic" triad for of MM PI profiles is a questionable pro-
the hospital group represent something quite cedure. Present results are in accord with
different than in the normal group, perhaps a such as those reported by Schmidt [10],
second "psychotic" factor. One must recog- Gough [4], and Benton and Probst [1], who
nize the danger of accepting the present scale found that specific score profiles on the vari-
titles for interpretation of factors without care- ous scales do not permit differentiation among
ful examination of the scale items contributing the patients in various psychiatric categories,
to the present loadings. though differentiation can be made between
The third factor obtained for the NP normal and abnormal persons.
group has loadings on only two scales, Pa
(.591) and F (.458). Pa is common to both Received July 17, 1950.
the normal and the NP group, with F replac-
ing Mf in the hospital group.
REFERENCES
A fourth factor was obtained for the NP
group, but it disappeared in the process of 1. BENTON, A. L., AND PROBST, KATHRYN A. A
comparison of psychiatric ratings with Minne-
rotation. sota Multiphasic Personality Inventory scores.
Combined Group, The factors presented by J. abnorm soc. Psychol., 1946, 41, 75-78.
combining the two groups are much as would 2. COTTLE, W. C. A factorial study of the Multi-
be expected from the data obtained in the phasic, Strong, Kuder, and Bell Inventories
using a population of adult males. Psychomet-
analysis of the two separate groups. The first rika, 1950, 15, 25-47.
factor again has its major scale loadings on 3. GENGERELLI, J. A. A factorial method whose
Pt (.936) and Sc (.931), with the other factors are empirical tests. Amer. Psychologist,
values in general falling between those ob- 1949, 4, 245-246.
tained for the separate groups. 4. GOUGH, G. Diagnostic patterns on the Minne-
The Mf scale, however, shows some pe- sota Multiphasic Personality Inventory. J. din.
Psychol., 1946, 2, 23-37.
culiar behavior. It will be noticed that in the
5. HATHAWAY, S. R., AND McKiNLEY, J. C. A Mul-
combined group its communality for the first tiphasic Personality Schedule: I. Construction
four factors drops to a value of .119. The of the schedule. /. Psychol., 1940, 10, 249-254.
 THE INTERNAL STRUCTURE OF THE MMPI 141

6. HATHAWAY, 8. R., AND MC&NLEY, J. C. Man- Multiphasic Personality Inventory in differenti-

ual for the Minnesota Multiphatic Personality al diagnosis. /. appl. Psychol., 1946, 30, 517-
Inventory. (Rev. Ed.) New York: Psychological 524.
Corporation, 1943. 9. SCHIELE, B. C., BAKER, A. B., AND HATHAWAY,
7. HOLZBBRC, J. D., AND AtESSi, S. Reliability of 8. R. The Minnesota Multiphasic Personality
the shortened Minnesota Multiphasic Personal- Inventory. Lancet, 1943, 63, 292-297.
ity Inventory. J. consult. Psychol., 1949, 13, 10. SCHMIDT, H. O. Test profiles as a diagnostic
288-292. aid: The Minnesota Multiphasic Inventory. J.
8. MEEHI,, P. E. Profile analysis of the Minnesota appl. Psychol., 1945, 29,115-131.