THEORETICAL RESPONSES TO NAVAJO QUESTIONS

by

ELLEN SCHAUBER KAUFMAN

B.A. Brandeis University

1968

SUBMITTED IN PARTIAL FULFILLMENT

OF THE REQUIREMENTS FOR THE

DEGREE OF DOCTOR OF

PHILOSOPHY

at the

MASSACHUSETTS INSTITUTE OF

TECHNOLOGY

June, 1975

Signature of Author...........................-.....---
Department of Foreign Literature & Linguistics
April 23, 1975

Certified b.............--...--.----..-----.
Thesis Supervisor

Accepted by................. .. ,....... .... ..-----

Chairman, Departmental Committee
on Graduate Students

ARCHIVEr
gs INST.

JUN 30 1975
&.I8RARI e
To Jay,

always the most confident,

the most optimistic.

 3

THEORETICAL RESPONSES TO NAVAJO QUESTIONS

by

ELLEN SCHAUBER KAUFMAN

Submitted to the Department of Foreign Literature and Linguistics on

April 23, 1975 in partial fulfillment of the requirements for the
degree of Doctor of Philosophy.

ABSTRACT

This dissertation surveys a range of questions in Navajo, con-

cluding on the basis of embedded questions that generally accepted
theoretical claims about constraints on movement are incorrect. I
argue that indirect question formation necessitates an unbounded right-
ward movement transformation, vitiating the Right Roof Constraint. Direct
questions formed from sentences involving only direct discourse verbs
demonstrate that unbounded leftward movement also occurs in Navajo,
although it is a SOV COMP-final language, challenging claims that move-
ment must be to COMP position. I claim that the different syntactic
forms associated with the two types of movement exist because of the
semantic interpretation imposed by the element (verb or complementizer)
governing the question word. I contrast indirect questions to embedded
questions which require no movement to establish the semantic require-
ments of the complementizers. I conclude that any theory of movement
must consider two separate factors, the extent of the movement (i.e.,
how far upward an element moves) and the direction of the movement
(rightward or leftward), arguing that semantic as well as syntactic
conditions account for movement rules.

Thesis supervisor: Kenneth Hale

Title: Professor of Linguistics
 4

ACKNOWLEDGMENTS

This dissertation owes its existence to those who offered assistance.

This author, in turn, benefited not only by completing a thesis, but also

by sharing the relationships on which it rests. Many of the intellectual

debts are deeply felt personal ones as well, none more so than to Ken Hale.

As advisor, scholar of both linguistics and Navajo, teacher and friend, his

dedication, even at the expense of his own research and tranquility, was as

inspiring as it was helpful. It is with deep respect and lasting affection

that I again say thanks, Ken.

For the Navajo consultants who made this project possible I offer

the sincere hope that its findings will in some small measure repay their

cooperation. Special thanks and credit are owed to Lorraine Honie who

helped with the earliest investigations, and to Ellavina Perkins, Lorene

Begay and Linda Platero for their long hours, friendship and insightfulness

which truly made this thesis a combined effort. I am also very grateful to

the students of the 1973 Dine'Bi'olta' Association Linguistics Workshop,

particularly to Rose Fasthorse and Louise Benally, for a summer of challenge

and assistance.

Noam Chomsky who gave most generously of his time and expertise

provided invaluable insights and challenges. He, Ken Hale and Dave Perl-

mutter were supportive mentors, even in disagreement. To them goes my

appreciation for a most helpful dissertation committee. My appreciation

goes also to our chairman Morris Halle whose dedication touches all

students in the department.

 5

To fellow linguists, friends and instructors, Avery Andrews, Nomi

Erteschik, Jorge Hankamer, Jim Kari, Jan Koster, Susumo Kuno, Paul Platero,

Marga Reis, and Haj Ross go sincerest thanks for comments, critiques and

consultations which greatly enhanced this undertaking. To Donna Furrow

and Yvonne Bordelois are owed not only credit for significant enhancement

but also my thanks for their warm friendship and bravery in the face of my

battles with graduate work.

I am indebted to the American Philosophical Society Phillips Fund

for a summer research grant, and to the faculty at M.I.T. for its finan-

cial support under N.I.H. grant 5T01 HDO0111 and N.I.M.H. grant 5P01

MH13390.

But most of all this dissertation is the product of my family. To

my parents whose love and stimulation began it all goes an equal measure

of love and appreciation for their continued presence. To Ian whose

smiles made these last eight months so very special goes a promise of

the long summer walks that the busy winter precluded. To Waffle I offer

thanks for cuddling so very closely when the going got rough. And to Jay

goes the greatest measure of love and thanks for sustaining me throughout.

His loving, encouraging, cheering, prodding and editing surely made this

dissertation possible.
 6

TABLE OF CONTENTS

Page

INTRODUCTION ....................................................... 11

0.1 Prologue .................................................. 11

0.2 Basic Navajo Structure .................................... 15
0.3 Word Order: Scrambling ................................... . 16
0.4 Types of Embeddings .............. .......................... 17
0.4.1 The /-i/ Complementizer ........................... 18

0.4.2 Direct Discourse .................................... 19

0.4.3 Structural Configuration ..................... 27
0.4.4 The /-go/ Complementizer ........................... 31
0.5 Conclusion ............................ . ............ 32

FOOTNOTES TO INTRODUCTION ..................................... . 34

CHAPTER 1: INDIRECT QUESTIONS AND UNBOUNDED RIGHTWARD MOVEMENT .... 37

1.0 Introduction ............................................... 37

1.1 Enclitic Phrases and Enclitic Raising ..................... 40
1.1.1 Identification of Spatial Enclitics and
Enclitic Phrases ................................. 40
1.1.2 EP's as NP's ....................................... 42

1.1.3 Enclitics On Verbs ................................. 44

1.1.4 Missing NP's ....................................... 46
1.1.5 Enclitic Placement................................ 48
1.1.6 Enclitic Movement .................................. .. 49
1.1.7 Unbounded Raising ..........................-.... 50
l.l.8 The Raising Transformation ...........-........ ... 52
1.1.9 Phonological Evidence Against Chomsky-Adjoining
the Enclitic ....................... ....... o. 53
1.1.10 Chorus ..................................... 56
1.2 Enclitic Phrases and Relative Clauses .......------------- . 57
1.2.1 /-1/ As Relative Clause and Complementizer Marker 58
 7

Page
1.2.2 Insufficiencies of the Relative Clause
Explanation ...................................... 60
1.2.3 EP's As Relative Clause Heads ........................63
1.2.4 The Inability of the Relative Clause Explanation
to Replace Movement .............................. 66
1.2.5 Recapitulation ..................................... 68

1.3 The EP and Its Constituents ................................ .69

1.3.1 EP's and PP's: Similarities ........................ .70
1.3.2 EP's and PP's: Differences ......................... 70
1.3.3 The Classes of NP's Which Can Occur With
EP's and PP's .................................... .72
1.3.4 Postpositions and Relative Clauses ...... o............ 75
1.3.5 Phrases With Both Postpositions and Enclitics ....... 78

1.3.6 PP's, EP's and Other Types of Relative Clauses ...... 82

1.3.7 Meaning Problems For Indirect Questions As

Relative Clauses ............................... 85
1.4 Movement and Indirect Questions .......................... 86

1.4.1 Identification ofn................................ 87

1.4.2 z Is Not a Dummy Node ............................. 87
1.4.3 Evidence That _ Is /haa/ ......................... 89

1.4.4 Further Considerations .............................. 95

1.4.5 The Rightward Movement Is In Indirect Questions 95
1.4.6 The Unbounded Movement Is Not Unconstrained ......... 97
1.4.7 Conclusion .......................................... 100

FOOTNOTES TO CHAPTER 1.............................................. 101

CHAPTER 2: DIRECT DISCOURSE QUESTIONS AND UNBOUNDED LEFTWARD

MOVEMENT .......................................... 108

2.1 Simple Questions ........................................... 108

2.1.1 Identifying the Interrogative Enclitics ............. 108
2.1.2 /-sh/ As a Distinct Enclitic .......................... 109
 Page
2.1.3 /-sht'/ As a Topic-Question Enclitic ................ 111
2.1.4 The Interrogative Particle /1i/ ..................... 116
2.1.5 Word Order .......................................... 118
2.1.6 Where Question Particles Occur Underlyingly ......... 120
2.1.7 How To Account For WH Placement ..................... 122
2.1.8 A Comparison To Baker's Proposals ................... 124
2.1.9 Relationship Between Yes/No Questions and WH
Movement ......................................... 125
2.1.10 More Problems With /-shg'/ .......................... 127

2.2 Complex Sentences .......................................... 129

2.2.1 Unbounded WH Movement To Sentence Initial Position 131
2.2.2 Contrast To Theoretical Claims ....................... 136
2.2.3 Dialect Variation ................................... 138
2.2.4 Strict Constructionists ............................. 140
2.2.5 Loose Constructionists .............................. 142
2.2.6 A Rationale For the Unbounded Movement .............. 144

2.3 The More Problematic Cases ................................. .149

2.3.1 The WH Word Is Non-Initial .......................... 149
2.3.2 Chorus .............................................. 153
2.3.3 Second Position and Case (5) ...................... 154
2.3.4 Case (6) ............................................ 156
2.3.5 Generating Case (6) ................................. 158
2.3.6 Case (6) or Case (4)? ............................ 159
2.3.7 Case (4) ............................................ 162
2.3.8 Interpreting Case (4) ............................... 166

2.4 Summary, Conclusions and Speculations ...................... 170

2.4.1 Summary ............................................. 170
2.4.2 Is There Really Any Unbounded Movement? ............. 172
2.4.3 An Argument For Separate S's: Double Questions ...... 176
2.4.4 Arguing From Emphatic Particles ..................... 177
 9

Page
2.4.5 Arguing From Negatives ...................... 179
2.4.6 Arguing From Second Position ................ 181
2.4.7 Topicalization and Unbounded Movement ....... 183
2.4.8 Speculations: The Genesis of the Movement .. 184
2.4.9 Speculations: Why So Confusing? ............ 185
2.4.10 Speculations About Case (4) ................. 190
2.4.11 Question Particle Placement ................. 192

2.5 Formulating the Transformation ............................. 193

2.5.1 Are Particle Placement and Second Position the
Same Rule? ............................. 194
..........
2.5.2 An Independent Rule ....................... 196
2.5.3 Are Particle Placement and WH Movement the
Same Rule? ........................ .... 0 ... 0 197
2.5.4 Chomsky's Proposal ................. .... 0 .... 199
2.5.5 Chorus ............................... 203
2.5.6 Meaning Changes and a Q Node ......... 203
.. 0
2.5.7 Advantages of a Pre-S Q .............. 204
2.5.8 Problems With an Underlying Pre-S Q .. 208
2.5.9 Comparing Two Theories ............... 210
2.5.10 Conclusion ........................... 212

FOOTNOTES TO CHAPTER 2 .............................. 213

CHAPTER 3: THE SEMANTICS OF THE COMPLEMENTIZERS .................... 222

3.0 Introduction ............................... ...0 222

3.1 /-go/ and Adverbs .......................... ...0 223

3.1.1 Overview ........................... ... 0 223
3.1.2 Possible Semantic Content for /-go/ . ... 0 227

3.1.3 Possible Syntactic Content for /-go/ .. *.. 228

3.2 Are /-go/ and /-i/ Clauses Always Structural ly

Differentiated? ...................... 230
3.2.1 Seemingly Anomolous /-go/ Sentences..... 230
 10

Page

3.2.2 The Structure of the /-go/ Clauses .................. 232

3.2.3 What Is the Subject? ................................ 233
3.2.4 Morphological Evidence For /-go/ Clauses As
Subjects ......................................... 234
3.2.5 /-go/ Clauses As Objects ............................ 237

3.3 Are /-go/ and /-i/ Clauses Semantically Identical? ......... 240
3.3.1 Differentiating /-go/ and /-I/ ...................... 240
3.3.2 The Semantics of /-/ ......................... 243
3.3.3 /-1/ and Semantic Subordination ..................... 248

3.3.4 An Explanation of Earlier Translations .............. 251

3.3.5 Refining the Referential Hypothesis ................. 254
3.3.6 /-go/ Is Semantically Empty ......................... 258

3.4 Syntactic Evidence For the Referential Hypothesis .......... 261

3.4.1 Clitics Determine Focus and Presupposition .......... 262

3.4.2 Complex Sentences With /-go/ ........................ 265

3.4.3 Complex Sentences With /- / ......................... 267

3.5 Embedded Questions With /-i/ and /-go/ ..................... 272

3.5.1 WH Questions In /-go/ Adverbials .................... 273

3.5.2 WH Questions With /-go/ As Contrasted To /-(/ ....... 275
3.5.3 Difference In Meaning ............................... 278

3.5.4 /-1/ and Deleted WH Words ........................... 279

3.5.5 WH Words With Presuppositions ....................... 283
3.5.6 /-go/ and Indirect Questions ........................ 287
3.5.7 /-go/ and Unbounded Movement ........................ 291

FOOTNOTES TO CHAPTER 3 .............................................. 294

CONCLUSION .......................................................... 299

FOOTNOTES TO CONCLUSION ..................................... 308

REFERENCES .......................................................... 309

BIOGRAPHICAL NOTE ................................................... 314

 11

INTRODUCTION

0.1. Prologue

Although the Navajo language has been studied extensively, it is

only very recently that transformational grammarians have begun investi-

gating how our syntactic and semantic conceptions of generative grammar

apply to this language. As has already been demonstrated by the little

work done on the language (cf. especially Hale, Platero, Creamer), signi-

ficant factors emerge in the study of Navajo that are less easily perceived

in languages heretofore more intensively investigated.

One of the factors making Navajo particularly interesting is that

it is a characteristic SOV language, and, more particularly, a verb final

language. The natural word order is SOV, relative clauses precede their

heads, complementizers follow their clauses, and postpositions, rather

than prepositions, are used. It is well recognized that there are signi-

ficant syntactic differences between verb final and non-verb final

languages. Thus, Japanese and Turkish, the verb final languages studied

in greatest detail, are often contrasted to other languages, and putative

language universals have been created to account for the differences.

However, as I intend to show, there are crucial differences between

Navajo and existing descriptions of other verb final languages, dif-

ferences requiring a reexamination of some fairly well-accepted universal

constraints on language.
 12

I will begin in Chapter 1 by looking at one type of question in

Navajo, embedded indirect questions. These are immediately interesting

because they are an example of unbounded rightward movement, a process

thought not to exist. Since there is some reason to believe that the

alleged universals under consideration are well-motivated, I believe it

is necessary to look for extenuating circumstances which would both ex-

plain the use of the unbounded rightward movement and help account for

why such movement went unnoticed.1 Thus, in Chapters 2 and 3 I turn to

other types of questions in Navajo as I believe the contrasts between the

types of questions are instructive. After a description of simple ques-

tions, incomplete but sufficient for understanding the basic Navajo yes/no

and WH questions, the remainder of Chapter 2 is devoted to a second type

of embedded question also thought not to exist in verb final (or COMP

final) languages: WH questions involving unbounded leftward movement.

I suggest that the point of view interpretation found in direct discourse

sentences is what accounts for the use of unbounded movement: a WH word

moves from one S to another to change its point of view interpretation.

Although the movement rules discussed in Chapters 1 and 2 are

completely different, I will suggest that the rationale for both rules

is the same: elements move unboundedly to escape contradictory or un-

desired interpretations. In Chapter 2 I discuss point of view interpre-

tation as specified by the governing direct-discourse verb. In Chapter 3

I discuss the semantic content associated with the complementizers. This

chapter then completes the survey of different types of embedded questions,

 13

examining the case considered typical of a verb final language, i.e., the

case in which no movement occurs. I demonstrate that each type of ques-

tion can be associated with a different complementizer, concluding that

the contrast in semantic requirements of the complementizers accounts for

the diversity of question types. Unbounded rightward movement occurs only

when it is necessary for semantic reasons to remove the WH word from the

scope of the complementizer. Since there are no semantic problems with

the question type discussed in Chapter 3, no movement is necessary.

Although I attempt to go into the fullest detail on issues related

to my major claims, I am of course describing only a minute segment of the

Navajo system. As a result, many of the conclusions I reach must be con-

sidered tentative and somewhat speculative. However, since the Conclusion

is an attempt to describe and explain the Navajo results in terms of

general theory, I hope that work on other languages as well as further

research on Navajo will both justify and help refine my claims.

The problem is exascerbated by the fact that there is little pre-

vious work on Navajo which relates directly to my inquiry. I am also

working under the further handicap of not being a native speaker of the

language. Having been particularly fortunate in having consultants who

are not only patient relaters of facts about their speech, but interested

linguists doing significant work of their own, I feel I have been able to

describe the facts relatively accurately. I have attempted to vary the

techniques used in collecting the data to get some feeling not only about

what is acceptable Navajo, but how natural a sentence type is. It is

 14

quite clear that many of the crucial grammatical examples are considered

less than natural by at least some of the Navajos consulted. What is not

so clear is whether this unnaturalness is due to the linguistic complexity

of the examples or to cultural factors, or whether it reflects uncertainty

due to contemporary changes or influence from English. None of these fac-

tors will be discussed here. However, although much of the information

has been collected by asking for Navajo translations of English sentences,

I have tried to avoid the English influence on decisions about grammati-

cality by also giving the sentences in Navajo to see if they are acceptable

and then asking for English translations. Since I created many of the

examples myself, this technique does not differentiate very well those

examples which are preferable. Nor does this technique guard against

slight changes in meaning. I have tried to cope with this latter problem

in two ways. First, I have used dialogues instead of isolated examples

since context often refines the meaning or possible meanings of a sentence.

Second, after having collected relevant data, my Navajo consultants and I

have discussed the theoretical questions involved. Understanding why a

sentence is crucial, my consultants often provided further extremely rele-

vant information and made suggestions about other sentences and directions

to investigate.

One further weakness is that, although I mention dialect variation

throughout, this is a topic needing considerably more attention. I have

found that certain types of sentences vary so markedly among the relatively

few people I have consulted that the description should really be reinforced
 15

by a dialect study and/or diachronic study. Unfortunately, I have not yet

been able to do either.

Since I am unable to motivate all aspects of the Navajo examples I

will be considering, I will limit as much as possible the claims made about

the structure of the Navajo sentence. As a result, I will make no attempt

to describe a complete underlying structure, but will only describe what a

tree would look like at the time a transformation applies. I will also

take the liberty of creating forms for expository purposes without making

any claims about their exact correlation in underlying structure. And when

I do attempt to argue that certain forms must exist underlyingly, I will do

so on the basis of arguments internal to Navajo rather than theories of

underlying forms.

Although all aspects are not well motivated, it is of course neces-

sary to work within some framework. Thus, before proceeding to Navajo

questions, I will describe some facts about Navajo that will be relevant

to the argumentation and mention some claims and assumptions I will be

making about the structure of the language.

0.2. Basic Navajo Structure

As has already been mentioned, Navajo is an SOV language, shown by

the natural word order in the following:

1. ashkii lecha'i yiyiilts

boy dog 3.3.P.see
The boy saw the dog.
 16

In the Navajo sentence, the verb is the main source of information. Notice

that the gloss on the nouns are only English nouns. This is because Navajo

nouns are not normally inflected for number2.

2
Nor are they inflected for

definiteness. Although Navajo has demonstrative pronouns, there are no

determiners corresponding to the English a and the so the example is ambi-

guous with respect to definiteness. I will in general ignore these am-

biguities, noting them only when relevant to the argumentation.

In contrast to the noun, the verb has numerous prefix positions.

I have made no effort to gloss the verb fully since doing so would not aid

the reader. 3 I have noted only the subject and object prefixes and, where

useful, the verb tense or aspect. The direct object prefix is ordered be-

fore the subject marker in the Navajo verb. Thus, in the above gloss, the

first 3 refers to the third person object, and the second to the subject.

P stands for perfective, and F for future. I will discuss specific pre-

fixes as relevant to the argumentation.

0.3. Word Order: Scrambling

Since Navajo is an SOV language, complex objects create extensive

subject-verb separation, making these sentences more difficult to process.

Navajos compensate with various scrambling rules. Subjects, especially

sentential subjects, can be postposed. Subjects and objects can be inter-

changed so long as the inversion does not lead to ambiguity. Thus, the

following are all good outputs:

 17

2a. ashkii ;le'chga' yiyii;Ltsanig i Jfan bil b 6hozin

boy dog 3.3.P.see.COMP John 3.with 3.be known
John knows that the boy saw the dog.

b. Jaan bil b ho'zin ashkii ;L4e'chgg'1 yiyii;Ltsinagii

John 3.with 3.be known boy dog 3.3.P.see.COMP
John knows that the boy saw the dog.

c. Jaan ashkii le/e'haa' yiyiisktsanigii bil beeho zin

John boy dog 3.3.P.see.COMP 3.with 3.be known
John knows that the boy saw the dog.

Sentence (2a) exhibits the natural word order, a sentential subject follow-

ed by the object and then the verb.4 In sentence (2b), the sentential

subject is postposed. Sentence (2c) interchanges the subject and the

object.

It is not evident from the English sentence that sentence (2) has

a surface sentential subject. However, literally translated, sentence (2a)

reads "That the boy saw the dog is known with John". Thus translated, the

sentential subject is readily apparent. Its verb, /-; beehozin/, most

closely resembles an intransitive passive, meaning is known about with.

5
The /-I/ is a postposition meaning with, and /Jaan/ is its object.

0.4. Types of Embeddings

In the preceding examples, the embedded S is marked by the clause

final complementizer /- gi/. This is one of the three types of embedded

S's found in Navajo. Since this dissertation deals with significant

differences in behavior among the three, I will briefly mention the three,

noting their most common syntactic uses at this juncture. In trying to

account for the variety of embedded question types, I intend to demonstrate

that there is in fact some overlap in the types of S's each complementizer

may subordinate. While complementizers were originally viewed only as

markers of syntactic subordination (cf. Elgin (1973)), I take the position

(along with Bresnan (1972), Erteschik (1973), etc.) that complementizers

may have semantic content which influences syntactic processes.

Although complementizers are central to the entire dissertation, I

cannot argue whether they occur in underlying structure, nor whether there

is a COMP node and if so how it is adjoined to the tree. For the most part,

these matters are irrelevant to the discussion. Whenever they would make a

difference in the argumentation, I will note their significance.

0.4.1. The /-i/ Complementizer

The /-i/ complementizer is normally considered a general nominalizer

appearing on any S dominated by an NP node (cf. Elgin (1973), Platero

(1974)). Although I will demonstrate in a forthcoming paper that this is

not the only use for the /- / complementizer, the sentences in (2) are in
6
fact typical of this complementizer.

I will refer to /-egll/ as one form of the /- / complementizer. The

-gii/ appears only in nominals and only if no other suffix appears on

/i/. In addition, the /- gii/ is suppressed when the past tense variant

of this complementizer is suffixed to the verb (/-ye/, /-ee/, or /-a/,

 19

depending on purely phonological rules). The sentences in (3) show /-1/

and /-yde/ alternating with tense. 7

3a. adadd' Bil b egashii yizlohe Mary bil beeh6zin

yesterday Bill cow 3.3.P.rope.COMP Mary 3.with 3.be known
Mary knows that Bill roped a cow yesterday.

b. Bil be'gashii yidoolohigii Mary bil beehozin

Bill cow 3.3.F.ropeCOMP Mary 3.with 3.be known
Mary knows that Bill will rope a cow.

0.4.2. Direct Discourse

Although /-I/ is an extremely general complementizer, there are

certain verbs whose complements never take it.

4a. *Bil b4 gashii deesh;lohigii nazin

Bill cow 3.1.F.ropeCOMP 3.think
Bill1 thinks he. will rope a cow.

b. Bil beegashii deeshloh nazin

Bill cow 3.1.F.rope 3.think
Bill. thinks he. will rope a cow.
1 1

5a. *Jaan chidf naha;lnii'igii na

John car 3.1.P.buyCOMP 3.say
John says he. bought a car.

b. Jaan chidi nahailni' ni

John car 3.1.P.buy 3.say
John says he. bought a car.
 20

Notice the gloss of the embedded verb in both (4b) and (5b). The Navajo

verb is in the first person, not in the third, as it is in English. The

embedded sentence is structured as if it were spoken by the subject of the

matrix S. A better translation of (5b), for example, is "John says, 'I

bought a car'". While the /- / embedded sentences were instances of in-

direct discourse, this second type of embedded sentence is characterized

by direct discourse. Contrast the following indirect discourse sentence

with (5b), a sample of the direct discourse sentences above:

6. Jdan chid nahaInii'ig yaa ntsikees

John car 3.1.P.buyCOMP 3.about 3.think
John is thinking about the fact that I bought a car.

(6)

NP NP V

Jaan yaa ntsikees

John 3.about 3.think

NP NP V COMP

(sh ) chidi nah knii' L

I car 3.1.P.buy
 21

(5b)

NP NP V

Jaan nai
John 3.say
Si

NP NP V

(shi) chidi naha-nii'

I car 3.1.P.buy

Sentence (6) is a typical indirect discourse sentence with an /-,/ comple-

mentizer. In indirect discourse, the entire embedded S is understood from

the point of view of the speaker. In direct discourse sentences, each

embedded S is interpreted from the point of view of the subject of the

immediately dominating S. In sentences (5b) and (6) the embedded S's

both contain the same first person verb /naha;rnii'/. The difference be-

tween direct and indirect discourse is demonstrated by how first person

is interpreted. In indirect discourse, first person is understood from

the speaker's point of view. In sentence (6), therefore, it is the speaker

who bought the car. However, in direct discourse, first person is under-

stood from the point of view of the subject of the immediately dominating

S. In (5b) the first person in S1 is understood from John's point of view

(the subject of SO). I therefore refers to John, i.e., it is John who

bought the car.

 22

Since (5b) involves only one embedded S, it is impossible to

differentiate the hypothesis that an S embedded under a direct discourse

verb is interpreted according to the immediately dominating S from the

hypothesis that such an S is always interpreted from the point of view

of the matrix S. Example (7) demonstrates that the former hypothesis is

correct.

7. Ja'an Mary chid nahideeshnih nizin ni

John Mary car 3.1.F.buy 3.want 3.say
John says Mary wants to buy a car.

(7)

NP SV

Jdan ni
John 3.say
NP S V

Mary nizin
Mary 3.want
NP NP V

(sh ) chidi nahideeshnih

I car 3.1.F.buy

The verb in S2 is in the first person. If the latter hypothesis were

correct, the first person would refer to John, the subject of the matrix

S, i.e., (7) would be talking about John's buying a car. However, this
 23

is not a possible meaning for (7). In this sentence, it is Mary who will

do the buying. In other words, the first person in S2 is understood as

Mary, the subject of S1 , the immediately dominating S,and not the subject

of So, the matrix S. (7) demonstrates that the point of view of each

direct discourse S must be interpreted separately. Or, to put it another

way, the scope of a direct discourse verb is only one S, and not every-

thing it dominates. Since it is the verb which determines whether an

embedded S will be interpreted as a direct or indirect discourse S, I

consider the verb the governing element.

Notice that it would be insufficient to say that in a direct

discourse S the point of view is interpreted solely according to the

immediately dominating subject. Thus in (8) and (9), the matrix subjects

are identical, the embedded verbs are identical, but the subjects of the

embedded verbs are different.

8. Jaan Mary chid nahid iinih yilni

John Mary car 3.2.F.buy 3.3.say
John told Mary to buy a car.

9. Jaan chid nahid i nih shilni

John car 3.2.F.buy 1.3.say
John told me to buy a car.

Notice that the second person verb form does not refer to the hearer as

it would in an indirect discourse S. In these cases, the subject of the

embedded verb is determined according to who John is speaking to, i.e.,

by the object of the matrix verb. (8) could be paraphrased, "John told
 24

Mary, you work", and (9), "John told me, you work". (8) and (9) can be

thought of as reports of what John said when talking to someone.

The second person form of the verb will not always be used when an

embedded subject is identical to a matrix object. Consider the following

example which contains a third person direct discourse verb:

10. Jaan Mary chidi neidiyoolnih yo'ni

John Mary car 3.3.F.buy 3.of.3.say
John expects Mary to buy a car.

In this case, John is not talking to Mary, but about Mary, so that we find

a third person rather than a second person verb.

Having discussed verbs in the first, second and third person we

are still left with a question as to how the speaker refers to himself.

A first person verb will be interpreted as an action by the subject of

the immediately dominating S. A third person verb can only refer to a

third party. Sentence (9) above demonstrates that a second person embedded

verb will refer to the speaker if the matrix verb contains an object pro-

noun which denotes the speaker. Another way to indicate the speaker in a

direct discourse S is to use the fourth person form of the verb.

11. Ja'an chid nahizhdoolnih sho'nI

John car 3.4.F.buy 1.of.3.say
John expects me to buy a car.

The fourth person is normally used as a third person form, either to

differentiate two different third persons, or as an impersonal form

 25

equivalent to the English one (cf. A1majian and Anderson (1968)). In

sentence (11), the embedded verb gets interpreted from the point of view

of the immediately dominating S. It is because of the first person ob-

ject in /sh 'n / (expect me) that /nahizhdoolnih/ (buy) refers to the

speaker. Thus, if the object pronoun changes, so does the reference for

the embedded verb:

12. Ja'an chid nahizhdoolinih n6 'ni

John car 3.4.F.buy 2.of.3.say
John expects you to buy a car.

So far as I know, a direct discourse S can refer to the speaker

only if the governing verb refers to the speaker, either with /sh 'n /

(expect me) and an embedded fourth person verb, or with /shilni/ (tell me)

and an embedded second person verb (cf. example (9)). There is no way to

use the verbs which do not contain such a reference and still refer to the

speaker. For example, "John wants me to buy a car" can only be paraphrased

by (9) and (11), not translated directly by /nfzin/.

Since in sentence (10) the embedded verb is in the third person,

it is impossible to tell from the verb agreement whether this verb is

being interpreted as an example of direct or indirect discourse. However,

the direct discourse style is normally obligatory in the complements of

/ni/, /n 'zin/, and /y 'n/.9 Thus, a third person verb embedded in a

direct discourse S cannot refer to the subject of the matrix. Sentence

(13), for example, cannot be identical to (5b).

 26

13. Jaan chidi nayiisnii' ni

John car 3.3.P.buy 3.say
*Johni says he. bought a car.
Johni says hej/she bought a car.
He./she says John bought a car.

Sentence (13) means only that one person said of somebody else that he/she

bought a car. Since pronouns generally delete in Navajo, there is no

reason to assume that the subject of the embedded S deleted under identity

to that of the matrix S. In fact, /Jaan/ can be the subject of either S,

although it cannot be the subject of both. Sentence (14) clearly shows

that the subjects are non-identical.

14. Ja'an Mary chidi nayiisnii' ni

John Mary car 3.3.P.buy 3.say
John says Mary bought a car.

Although the verb agreement in (13) does not indicate that this

sentence is an example of direct discourse, the absence of a complementizer

does. Recall that in indirect discourse sentences the embedded S is marked

with a complementizer (cf. sentence (6)). To see that the presence of a

complementizer is obligatory in normal indirect discourse, consider the

following:

15a. Mary Bil dib4 yiyiisxanigi yoodla

Mary Bill sheep 3.3.P.killCOMP 3.3.believes
Mary believes that Bill killed the sheep.
 27

15b. *Mary Bil dibe yiyiisx4 yoodla/

Mary Bill sheep 3.3.kill 3.3.believe

It thus seems clear that direct discourse correlates with the absence of

any overt complementizer.

So far as I know, there are three direct discourse verbs in Navajo,

/n /, meaning say or tell, /nizin/, meaning think, want, wonder, and

/yo'n /, meaning expect (him). Although all three are verbs of thinking

or saying, it is difficult to define direct discourse verbs in terms of a

semantic class. Notice that believe in example (15) is not a direct dis-

course verb. There will be examples of other indirect discourse verbs

which would fit into the same semantic class. I will therefore make no

attempt to characterize the direct discourse verbs, but will only consider

the structures in which they occur.

0.4.3. Structural Configuration

The difference between direct and indirect discourse is reflected

in their structural configurations. Two alternative explanations suggest

themselves: either only indirect discourse involves embedded S's, or both

direct and indirect discourse involve embedded S's where only the latter

are also NP's.

In the first case, one could try to treat the Navajo direct dis-

course as one would an English quotation. In the English sentence

16. John said: "Who do you think killed the sheep?"

one would not want to embed the quotation under John said because the

quotation contains its own root sentence as demonstrated by the position

of who. From example (5b), however, it is apparent that the direct

discourse S occurs internally and therefore must be considered an embedded

complement, at least at this stage of the derivation. In Chapter 1, it

will become clear that, for the purposes of movement rules, direct dis-

course clauses must also be embedded.

In addition, it is necessary to differentiate quotations from direct

discourse within Navajo. Although the verbs /ni/ (say) and /nizin/ (in
the sense of think) can be used in quotations, the direct discourse verb

/yo'n / does not take quotations. One way to differentiate direct dis-

course sentences from quotations is to use the particle /ya'/, the Navajo

equivalent to a tag question. Consider the following:

17. Jaan Bi1 chidi neidiyoolnih ni ya'

John Bill car 3.3.F.buy 3.say right
(i) John said that Bill will buy a car, didn't he?
(ii) John said, "Bill will buy a car," didn't he?

18. Jaan Bil chidi neidiyoolnih ya' ni

John Bill car 3.3.F.buy right 3.say
(i) *John said that Bill will buy a car, won't he?
(ii) John said, "Bill will-buy a car, won't he?"

19. Jaan Bil chidi neidiyoolnih yo'ni ya'

John Bill car 3.3.F.buy 3.of.3.say right
John expects Bill to buy a car, doesn't he?
 29

y / I/
20. *Jaan Bil chidi neidiyoolnih ya' yo'ni
John Bill car 3.3.F.buy right 3.of.3.say

/ya'/, like an English tag, can only appear in non-embedded S's. When

occurring with a seemingly subordinate S, the sentence is good only if

interpreted as a question, i.e., a non-embedded S. Although /y 'ni/ takes

direct discourse, it does not take embedded quotations (just like the

English expect). Sentence (20) is ungrammatical because the /ya'/ requires

an embedded quotation but /yo'n$/ allows only direct discourse, not quota-

tions. Thus, for Navajo we must differentiate indirect discourse, direct

discourse, and quotations. Only the last will not involve embedded S's.

Since it is necessary to adopt the second alternative, i.e., that

direct discourse involves embeddings, the above difference between direct

and indirect discourse can be explained by claiming that only indirect

discourse S's are also NP's (as implied in the trees for (6) and (5b)

above). The Navajo verb morphology supports this claim. Consider the

following:

21a. ashkii mosi yiyiiIts

boy cat 3.3.P.see
The boy saw the cat.

b. *ashkii mos yiilts/

boy cat 3.P.see
The boy saw the cat.
ill
c. *ashkii yiyiilts/
boy 3.3.P.see
The boy saw.
 30

The /yi-/ prefix is a direct object agreement marker. It occurs when and

only when there is a third person subject followed by a third person direct

object. Thus (21c) is ungrammatical on the reading where /ashkii/ is the

subject because there is no stated object, and (21b) is ungrammatical be-

cause there is a direct object but no agreement.

We can use this fact to determine whether sentential complements

are direct objects. Since direct objects, by definition, are NP's, we can

differentiate those sentential complements which are not direct objects

by claiming that they are not dominated by NP, at least at the stage of

the derivation at which this /yi-/ agreement occurs. Consider the following:

22a. Jaan mosi yinizin

John cat 3.3.want
John wants a cat.

b. *Jaan mcsi nizin

John cat 3.want
John wants a cat.

23a. *Ja'an mos diyeeshxeel yinizin

John cat 3.1.F.kill 3.3.want
John wants to kill the cat.

b. Jaan mcsi diyeeshxll n/zin

John cat 3.1.F.kill 3.want
John wants to kill the cat.

The sentences in (22) affirm the fact that the /yi-/ prefix must appear on

/nizin/ when it has a direct object. The impossibility of a /yi-/ prefix

in (23) proves that these direct discourse sentential complements are not
 31

direct objects. In contrast, the sentences in (24) prove that indirect

discourse complements are direct objects, since the matrix verb has a

/yi-/ prefix.

24a. Kii naakaii tl'o'ol yizhbizhigii yiyiinii'

Kee Mexican rope 3.3.P.braidCOMP 3.3.P.hear
Kee heard that the Mexican braided the rope.

b. *Kii naakaii tl'6lc yizhbizhig i y nii'

Kee Mexican rope 3.3.P.braidCOMP 3.P.hear
Kee heard that the Mexican braided the rope.

I will therefore assume that these direct discourse sentential complements

are not dominated by NP. This fact can be used to distinguish these com-

plements from those with /-I/.

Sentence (24) is another example demonstrating that is a

possible structure for /- gi/ complements. Recall that this is the normal

condition. I will thus assume that /- g / may occur on complements only

when there is such a structural configuration.12 In Chapter 1, it will be

important to characterize when an /-i/ should appear instead of direct

discourse to see how these two different types of embedded S's interact

with enclitics.

0.4.4. The /-go/ Complementizer

Although ignored in Chapters 1 and 2, Navajo has a third important

method of subordinating S's.13 The /-go/ complementizer has typically been

referred to as a subordinator for adverbial clauses. "In English, therefore,

 32

one may render /-go/ with participial ... j.ng, or conjunctive 'if, when, in

case that' or, possibly, ignore the enclitic" (Haile (1942), 11.32). While

I will be discussing the more unusual uses of /-go/ in Chapter 3, the fol-

lowing are typical examples of the use of the /-go/ complementizer, in-

cluding approximate English translationsl0:

25. shizh4'e' niyaago da'diidi'

1.father 3.P.comeCOMP lpl.F.eat
When my father comes, we'll eat.

26. 1i' sinilohgo shizh'4 neidiyoolnih

horse 2.P.ropeCOMP 1.father 3.3.F.buy
If you rope the horse, my father will buy it.

27. Mary shaani{ydago Jdan bil hozho

Mary 1.3.P.comeCOMP John 3.with 3.be happy
Because Mary came to see me, John is happy.

/-go/ is like /- / in that the clause it subordinates is interpreted as an

indirect discourse rather than a direct discourse complement, i.e., it is

interpreted from the point of view of the speaker and not the subject of

the immediately dominating S. Thus, in (27), the first person pronoun

/shaa/ (to me) is interpreted as referring to the speaker and not to John.

The uses of /-go/ will be discussed in Chapter 3.

0.5. Conclusion

In summary, the basic Navajo facts described thus far which will

be crucial to the following argumentation include the following:

 33

1) Navajo is an SOV, COMP final language.

2) Navajo verb morphology provides crucial information about the

entire sentence. For example, person is marked in the verb. So are

direct objects.

3) The two most common sentential complements are

a) sentential NP's with /- (gif)/ complements

b) embedded direct discourse S's, which are not NP's.

4) /-go/ is primarily an adverbial subordinator.

Further detailed background material will be given as applicable. Since

this dissertation deals primarily with the formation of embedded questions,

I felt that an introduction to this topic required a view of basic word

order and various types of embeddings. As we examine the different types

of questions, I will be trying to demonstrate that it is the word order

plus the requirements of the various types of subordination that provide

the rationale for the complex system of questions found in Navajo and the

somewhat unexpected transformations required for these outputs.

 34

FOOTNOTES TO INTRODUCTION

1. It has been brought to my attention that there is also evidence from

Hindi and Telugu that there are rightward movement rules which are

not upward bounded. Cf. Satyanarayana and Subbarao (1973).

2. However, /ashkii/ is one that happens to be inflected for number, it

being the singular form. Those that are inflected for number normally

refer to people. Cf. Young and Morgan (1971).

3. See Kari (1973), Stanley (1963), etc. on verb morphology and phonology.

4. For some people, (2c), rather than (2a), is considered the most

natural word order. This variation seems to be related to the fact

that it is not clear what the underlying word order is. It may be

that the object of the postposition can be the underlying subject

for some people.

5. Postpositions have object agreement markers. Thus /bi-/ in /bil/ is

the third person personal pronoun.

6. There are also adverbial clauses with this /-1/ complementizer.

i. shizhI'eniyah da, da'diidil

1.father 3.P.comeCOMPpst-time 1pl. F.eat
Since my father has come, we will eat.

ii. t 'a' shizhe' niyahi, da'diid ;l

just 1.father 3.P.comeCOMP 1pl.F.eat
As soon as my father comes, we will eat.
 35

Since /-gi / never occurs on adverbial clauses, I conclude that /g/,

not /-i/ is in fact the nominal marker.

7. The most convincing argument is based on the semantics of the /-I/

complementizer and the fact that the same semantic interpretation

would be present for both (3a) and (3b). If /yje/ were a simple past

tense marker, it would not require such an interpretation. See

Chapter 3 for a discussion of the semantics associated with /-I/.

8. As we will see in Chapter 3, this contrasts with the scope of the

/-I/ complementizer.

9. For some people, an indirect discourse interpretation is possible in

addition to the direct discourse one. See 2.2.3.-2.2.5. for a

discussion of Strict Constructionists and Loose Constructionists.

Except when specifically mentioned, I will be discussing the dialect

of Strict Constructionists only, i.e., of those allowing only a direct

discourse interpretation.

For convenience, I am using the third person imperfective form

of the verb to refer to the verb since stems can be used in a variety

of verbs. Note that I gloss /yo 'n/ (expect) as including the stem

/nf/ of say. I chose this gloss because there is no exact English

equivalent and because phonologically the stem of expect is identical

to that of say. However, I have no other direct evidence that the

stems are identical.

 36

10. I am using the term root sentence as used by Emonds (1970). Emonds

distinguishes root, or matrix, S's from all others because there is

a group of transformations (root transformations) which apply only to

this highest S. Thus, English question formation applies only to

root sentences, ruling out such ungrammatical (non-echo) questions as

(i) *You think who killed the cat?

11. (21c) is of course good, meaning He saw the boy. Also although con-

sidered ungrammatical out of context, (21c) is grammatical if it

means the boy saw it where it refers to something in the discourse.

For a discussion of /yi-/ and reference, see section 3.2.5.

12. I am using the term complement to describe sentential subjects and ob-

jects, as contrasted to the other subordinate clauses, e.g., adverbial

clauses. Recall that /-i/ but not /- g i/ occurs in adverbials only.

13. When I claim three types of subordination, I am referring primarily

to subjects and objects. Although /-go/ is normally an adverbial sub-

ordinator, I argue in Chapter 3 that it can also subordinate senten-

tial NP's. There may also be a fourth type of subordinator (/-dddl/),

found only in adverbials. See my forthcoming article for discussion

of this point. In addition, there are types of embedded clauses I

will not be describing at all.

14. Note, for example that there is no Navajo word in these sentences for

if, when, or because. See my forthcoming article for discussion of

adverbial clauses.
 37

CHAPTER 1

INDIRECT QUESTIONS AND UNBOUNDED RIGHTWARD MOVEMENT

1.0. Introduction

In this chapter, I will describe the behavior of spatial enclitics

in embedded indirect questions in Navajo, demonstrating that some movement

rule involving unbounded rightward movement is necessary to account for

surface configurations. These questions are particularly interesting be-

cause they require reconsidering some of the most widely accepted work in

transformational theory, Ross's Right Roof Constraint and alleged universal

constraints on question formation, formulated by Bach and Baker.

John Ross (1967) divided transformations according to whether or

not they made crucial use of a variable. Rules that did so were con-

sidered unbounded. Beyond this important distinction, Ross also first

proposed the Right Roof Constraint, which claimed that all transformations

which move elements rightward are upward bounded. That is, no element will

ever move rightward out of its own S. The Right Roof Constraint thus

claims that there will be neither rightward movement involving crucial

variables nor rightward movement across a single S, such as English sub-

ject or object raising, relative clause formation through movement, etc.

Although linguists have recently begun to question the Right Roof Con-

straint (Satyanarayana and Subbarao (1973) and Grosu (1973)), this

putative universal has been widely accepted and employed.

 In particular, the Right Roof Constraint has been used to account

for the small variety of question types found in languages studied. It

has been widely accepted (based especially on the data from Greenberg

(1962)) that verb final languages have no special rule for moving ques-

tioned constituents and that in those other languages where such a rule

does exist, the questioned constituent will be moved to clause-initial

position. Baker (1969), for example, concludes that:

(i) "only languages which position their particles for yes-no

questions in clause-initial position permit a movement rule
for questioned constituents" (207); and

(ii) "there is only one possible movement rule for questions, which
differs in different languages only in the particular forma-
tive mentioned in place of the English wh:

Q X NP Y
1 2 3 4 -- 1 3+2 4

Condition: 3 dominates WIH" (ibid.).

He further explains that "those SOV languages which had particles posi-

tioned with reference to the sentence as a whole put them at the end of

the sentence. Correspondingly none of the SOV languages studied regularly

moved other question words to sentence-initial position" (ibid.).

Bach (1971) takes Baker's work one step further, using Ross's

concept of unbounded rules and the Right Roof Constraint to derive Baker's

conclusions. Bach presents four assumptions, and I repeat the three rele-

vant ones (the first being irrelevant to the conclusions here under

discussion):
 39

b) Question movement is unbounded

c) Movement is toward a governing verb

d) Only leftward movement rules may be unbounded (the RRC).

From assumptions (b) and (d) he derives the obvious conclusion that:

(i) the WH phrase will always be to the left, never the right.

And from all three assumptions, Bach further concludes:

(ii) Question movement will never occur in languages that have

the deep and surface order SOV. For if movement happens
by attraction to a governing verb (c), then in an SOV
language we would have to move to the right, a result
which we have just shown inconsistent with our assump-
tions (161).

Joan Bresnan (1972) refines even more the explanation for known

question types. Her claim that WH is a complementizer, plus The Comple-

mentizer Attraction Universal is sufficient to predict both of Bach's

conclusions.

The Complementizer Attraction Universal'

Only languages with a clause-initial COMP permit a COMP-

attraction transformation.

Here "the term COMP-attraction transformation may be understood informally

to apply to any transformation moving a constituent (perhaps over an es-

sential variable) into COMP position--for example,.. .Question Formation"

(42).
 40

While the hypotheses vary in each case, all three sets are designed

to account for the observation that there is no unbounded rightward move-

ment, and specifically none in questions. The description of embedded

indirect questions in this chapter is designed to demonstrate only that

Navajo requires an unbounded rightward movement rule. However, seeing

how the particular hypotheses compare to Navajo facts is also revealing.

I believe that such a comparison will help to explain why unbounded right-

ward movement had gone unnoticed for so long. Thus, as I look at other

types of Navajo questions in the following chapters, I will return to

these assumptions when applicable.

Let us now turn to spatial enclitics, embedded questions, and the

required movement rule.

1.1. Enclitic Phrases and Enclitic Raising

Since the raising transformation that I will propose moves spatial

enclitics from enclitic phrases, I will first identify and define the

terms under discussion.

1.1.1. Identification of Spatial Enclitics and Enclitic Phrases

Consider first examples of simple enclitic phrases (EP's):

1. Kii kinlkanidn' oolbas

Kee Flagstaff.from 3.drive
Kee is driving from Flagstaff.
 41

2. Kii kinlanigoo na'asbaaz

Kee Flagstaff.to 3.P.drive
Kee drove to Flagstaff (and back).

3. Kii kinla'niji'
C.
ni'nilbdz
LA.

Kee Flagstaff.as far as 3.P.drive

Kee drove as far as Flagstaff.

4. Kii kinlanidi na'albaas

Kee Flagstaff.at around.3.drive
Kee is driving around in Flagstaff.

5. Kii hooghangone' yah iiya

Kee house in into 3.P.go
Kee went into the house.

6. Kii bikoohgoyaa oolbas

Kee canyon.down in 3.drive
Kee is driving down the canyon.

In sentences (1)-(4), Kee was driving somewhere in relation to Flagstaff.

A comparison thus provides an approximate meaning for these. four enclitics:

"in the direction away from" (/-dle'/), "in the direction toward" (/-g6'd/),
2
"motion up to or as far as" (/-j{'/), "located in or at" (/-di/). /-go'ne'

in sentence (5) means "located in", but refers only to things that have an

entrance. In sentence (6), /-go'yaa/ indicates location down in something.

In these sentences, the enclitic phrase (EP) is defined as the enclitic

and the word it modifies. The enclitics themselves divide into two cate-

gories, the directional enclitics which are necessarily complements of

motion verbs (/-dde'/, /-goo/, and /-ji'/) and the locative enclitics
 42

which, as later examples demonstrate, may also be complements of stance

verbs. As we will see, both directional and locative enclitics may modify

phrases as well as single words.

1.1.2. EP's As NP's

In all the above examples the enclitic is attached to a noun. It

is thus reasonable to assume that an enclitic phrase is an NP. In fact,

some enclitic phrases do act like NP's: locative EP's can be subjects

or objects. Consider first sentences with EP subjects:

7. kindi shil yaa'hoot'e'h

house.at 1.with area.3.be good
I like it at the house.

8. bikoohgoyaa hotsaa
canyon.down in area.3.be big
The canyon is big.

9. a'a'a'ngone' hozaad
hole.in area.3.be deep
The hole is deep.

In sentence (7), the verb /-I ylist'e'h/ is an intransitive verb with a

postposition. More literally it means "it is good with me". Thus, the

subject is the EP in the house.

By considering sentences (l)-(6) above, we see that it is equally

apparent that EP's can be spatial complements.

More specifically, we can establish that locative EP's are direct

 43

objects. However, final confirmation of this claim must await the dis-

cussion of issues covered in later sections. For the moment, suffice it

to show the correlation between EP objects and direct objects. Consider

the following:

10. Jaan bikooh yiyiiltsa

John canyon 3.3.P.see
John saw the canyon.

11. Ja'an bikooh hooltsa'

L

John canyon area.3.P.see

John saw the canyon.

12. Jaan bikoohgoyaa hooltsa$

L

John canyon.down in area.3.P.see

John saw the canyon.

Notice that in sentences (7)-(9) and (1l)-(12), there is an agreement

marker /ho-/ in the verb, just as for any subject or object. This agree-

ment marker is obligatory whenever a subject or object is locative.

13. kin shil y 't'dh

house 1.with 3.be good
I like the house.

14. *kindi shil ya/Iat 'e'h

house.at 1.with 3.be good
I like it at the house.

Sentence (14) is ungrammatical because of the absence of a subject agreement

marker. Notice, however, that this same sentence with /ho-/ is acceptable
 44

(cf. (7)).
The presence of the prefix /yi-/ in sentence (10) shows that /bikooh/

is a direct object. In (11), however, /yi-/ is replaced by /ho-/. Yet in

both (10) and (11) the same logical relationship holds between the action

and the two nouns: John is seeing the canyon. It therefore seems that

/ho-/, as well as /yi-/, is a direct object marker. The difference is

that in (10) the canyon is simply named, while in (11) the area occupied

by the canyon is designated.

Returning to sentence (12), we see that the /ho-/ in this sentence

shows that EP's act like NP's in triggering locative object agreement.

Thus, if /ho-/ is considered a direct object marker, then EP's are not only
3
NP's, but also direct objects.

Note that /ho-/ is a locative marker only. It does not mark

directional phrases. Assuming that /ho-/ is a direct object marker ac-

counts for the fact that it does not mark directional phrases. Directional

phrases never function as direct objects because they only occur as com-

plements of motion verbs and motion verbs do not take such objects. Notice

that there is also no /ho-/ when locatives are complements of motion verbs

(cf. (4)-(6)).

1.1.3. Enclitics On Verbs

I have so far said little about the constituents of the EP, except

to point out that there is an enclitic and that in the examples thus far

considered the enclitic has attached to a noun. The exact characterization

 45

of possible EP constituents is irrelevant to this discussion. Notice, how-

ever, that a noun is not the only constituent an enclitic will attach to:

15. Ja'an di/ ashkii naaghahid "I yaa ahonizin

John this boy 3.comeCOMPfrom 3.about 3.be aware
John is aware of where this boy comes from.

16. di bilagaanaa dine/ bizaad yfhool'aa"'ladi

LL_ -
doo shil b 6hdzinda
this whiteman Navajo language 3.3.P.learn.at neg 1.with 3.be known.neg
I don't know where this Anglo learned Navajo.

17. shizhe4 oolbas/g / Bil bil be4 hozin

1.father 3.driveCOMPto Bill 3.with 3.be known
Bill knows where my father is driving to.

18. Mary Ja'an niniyahi4'i bee bil hlne '

Mary John 3.walkCOMPas far as 3.about 3.with 2.tell
Tell Mary how far John walked.

19. Jaan yah iiyahigone' Mary bil beeh zin

John into 3.P.goCOMPin Mary 3.with 3.be known
Mary knows where John went in.

In all the above sentences, the enclitic appears in the same word as the

embedded verb. In each case, the /- / which precedes the enclitic is the

complementizer. Notice also that in each case there is no Navajo word

which corresponds to the English "(the place) where". In other words, the

English spatial particles (from, in, etc.) have NP complements (where in

these examples) just as they did in all the previous examples. Yet this

is no longer true of the Navajo sentences.

 46

1.1.4. Missing NP's

It is certainly not characteristic of enclitics occurring in embedded

clauses that they have no NP complement. Sentences (20) and (21) show that

embedded enclitic complements do appear.

20. Kii kinlanigoo' na'asbtzigii Jdan bil beehdzin

Kee Flagstaff.to 3.P.driveCOMP John 3.with 3.be known
John knows that Kee drove to Flagstaff (and back).

21. Bil hooghang'ne' yah iyahigil Mary yoodla

Bill house in into 3.P.goCOMP Mary 3.3.believe
Mary believes that Bill went into the house.

However, it is a fact of the Navajo language that words like who,

what, and where generally do not have a surface form in an embedded S in

an indirect question.4 This fact is independent of enclitic phrases. Thus,

we find sentences like the following:

22. shizhe?' b'gh yiy yj' gi Ja/an bil b6e'h/zin

1.father bread 3.3.P.eatCOMP John 3.with 3.be known
John knows that my father ate bread.

23. shizhe'4 yiy ya' gi Jaan bil be'h 6 zin

1.father 3.3.P.eatCOMP John 3.with 3.be known
John knows what my father ate.

24. shichidi yineez''g bee shil hIlne'

1.car 3.3.P.stealCOMP 3.about 1.with 2.tell
Tell me who stole my car.
 47

There is nothing in the surface form of sentences (15)-(19) which

elucidates the underlying structure. In sentence (23), however, the miss-

ing word is a direct object.5 We know this because, as already discussed,

the /yi-/ prefix marks direct objects and can only occur when there is a

specified object in the surface structure. Contrast the following sen-

tences:

25a. shizhe'4' liya/'l

1.father 3.3.P.eat
My father ate.

b. *shizhe'e yiy{ yaa'

L-L

1.father 3.3.P.eat
My father ate.

26a. Bil aneez'I/'

LL

Bill 3.3.P.steal
Bill stole (something).

b. *Bil yineez' LL- '

Bill 3.3.P.steal
Bill stole.

As in English, the Navajo verbs for "eat" and "steal" are transitive. Thus,

when we say my father ate, we mean he ate something, without identifying

what. Navajo has an indefinite object marker /'i-/ to designate such a PRO

direct object.6 (In (26a) the underlying /'i-/ has become /'a-/.) Since

/'i-/ occurs only when no object appears, one could argue that /'i-/ is a

nonreferential marker, indicating that although the verb is transitive in

meaning, it never had an object. However, since /'i-/ does not occur on

truly intransitive verbs, it is reasonable to assume that it is an agree-

ment marker. In any case, the ungrammaticality of (25b) and (26b) demon-

strates that the /yi-/ prefix does not function as a nonreferential marker.

Return now to sentence (23). Since /yi-/ requires a direct object,

there must have been an underlying object in these embedded clauses.

Notice that in sentence (24), the missing WH word is a subject, not

an object. We know from the /yi-/ on the verb that there must have been

two NP's in underlying structure, an object as well as a subject. It is

therefore a general fact that WH words do not have a surface form in the

indirect discourse embedded S's treated here.

Let us now resume our discussion of the underlying structure of the

enclitic phrase. It seemed that sentences (15)-(19) semantically contained

a noun which does not appear on the surface in the Navajo. We have shown

that there is an underlying direct object in sentence (23) and an under-

lying subject in (24). Just as (20) and (21) are analogous to (22),

sentences (15)-(19) should be analogous to (23)-(24). There should be

an underlying object EP even when it has no surface realization. I will

postpone discussion of the nature of the underlying noun until section

1.4.1. For expository purposes, I will designate these "missing" nouns by .

1.1.5. Enclitic Placement

Note also that enclitic placement follows a prescribed pattern.

The enclitic may not appear on the verb if there is a noun for it to
 49

attach to. Contrast (27) with (20).

27. *Kii kin]/n! na'asb /zigo' Jaan bia: bee'ho'zin

Kee Flagstaff 3.P.driveCOMPto John 3.with 3.be known
John knows that Kee drove to Flagstaff.

The only difference between (20) and (27) is that the enclitic /-gdd/ (to)

was attached to its noun in the former and to its verb in the latter.

1.1.6. Enclitic Movement

Sentence (27) provides an argument against generating all enclitics

underlyingly where they occur on the surface. If enclitics with verbal

complements were generated independently of those with NP complements,

there would be no way to account for (27). However, we can account for

the facts so far discussed if we assume that enclitics occur underlyingly

only in phrases with an NP. Since enclitics by their very nature cannot

be independent words, we must account for their cliticization anyway.

Under the hypothesis that all enclitics originate in phrases with NP's,

an enclitic will cliticize to its NP whenever possible. However, if an

enclitic is stranded in its own phrase through deletion of its NP, a trans-

formation will move it so that it attaches to the end of the embedded verb.

Since the complementizer precedes the enclitic, the cliticization could be

either to the clause final morpheme or to the COMP. As we will see, the

two are not necessarily equivalent. What is important to note is that,

by taking advantage of a characteristic inherent to enclitics we can postu-

late a single source for EP's and predict when there will be movement.
 50

1.1.7. Unbounded Raising

We have thus far considered cases with only one embedded verb. The

truly interesting cases involve several embeddings:

28. Ja'an adeesbps n'ziniL& Mary bil beeh6zin

John 1.F.drive 3.wantCOMPto Mary 3.with 3.be known
Mary knows where John wants to drive to.

29. Jaan d ind' bi' doo 'niidigo/0'sh nil be'e4ho'zin

John 2.F.go 3.3.pl.tellCOMPtoQ 2.with 3.be known
Do you know where John was told to go?

30. shinaa ndeeshnish nazinifish nil beehozin

1.brother 1.F.work 3.wantCOMPatQ 2.with 3.be known
Do you know where my older brother wants to work?

In none of these sentences is the enclitic part of the verb it complements.

Notice that the verbs to which the enclitics are attached are incapable

of taking enclitic phrases as their complements (see examples (31)-(32)).

On the other hand, enclitic phrases may occur as complements to the lower

verbs, and such complements are semantically accurate for the above sen-

tences.

31a. *Jaan kinla'nigoo nizin

John Flagstaff.to 3.want
John wants to Flagstaff.

b. Jaan kinla'nigo' adoolbas

John Flagstaff.to 3.F.drive
John will drive to Flagstaff.
 51

32a. *Jaan MITg6oobi'doo'niid

John MIT.to 3.3.pl.tell
John was told to MIT.

b. Jaan MITgo' dooga'l

John MIT.to 3.F.go
John is on his way to MIT.

These facts demonstrate that even if some enclitics originate in

an EP and some on the verb, a transformation is still necessary. The

enclitics in (28)-(30) cannot originate where they appear in the surface

string. Given that a transformation is needed anyway, if we allow only

one underlying source for spatial enclitics (an EP with either a specified

NP or a 4), then we can express the generalization that the logical re-

lationship is the same in all the examples considered. Notice also that

these facts further demonstrate that, whatever the proper transformation

may be, its scope is more than a simplex S.

To see that the scope of the transformation is in fact unbounded,

that the enclitic may appear on a verb any number of S's away from its

original, consider the following:

33. shinaai deeshad nisin ninig6 Jdan bil be'h6zin

1.brother 1.F.go 1.want 3.sayCOMPto John 3.with 3.be known
John knows where my brother says he wants to go.

It is obvious from the meaning of (33) that /-go/ (to) must have originated

in the lowest S, as a complement of /deeshia$l/ (go). In addition, it is

 52

clear that neither want nor say takes EP complements. There is no other

possible source for the enclitic. We must therefore postulate an unbounded

enclitic raising transformation.

1.1.8. The Raising Transformation

Since enclitic movement is unbounded, it is necessary to determine

which verb an enclitic will move to. Let us first rule out the possibility

of fixed position, i.e., movement to the highest embedded verb:

34. Bil Mary naaghahidQd' yishniih ninigi shi]= b e'hozin

Bill Mary 3.comeCOMPfrom 3.1.hear 3.sayCOMP 1.with 3.be known
I know that Bill said he heard where Mary comes from.

35. dniyah-" adeesbas nisin

2.goCOMPto 1.F.drive 1.want
I want to drive to where you are going.

While the enclitic does not necessarily attach to the highest em-

bedded verb, it will in fact attach only to an embedded verb. Thus we

never find sentences like the following:

/ /J //

36. *Ja'an naaghahigii shi]: beehozind9Q'

John 3.come.COMP 1.with 3.be known.from
I know where John comes from.

To determine where an enclitic will attach, it is therefore neces-

sary to compare embedded S's. We already know that one characteristic

distinction among verbs is whether they take /- /. In all the above

 53

examples, if a verb has an enclitic attached to it, it also has an /-I/

complementizer. No verb that an enclitic has passed over has an /-/
complementizer. Thus, it seems that a stranded enclitic cliticizes not

to the V, not to clause final position, but to the /-i/ complementizer.

We are now in a position to state a tentative raising rule.

Enclitic Raising

W - ME]EP - X - Y
1 2 3 4 5 6
1 2 0 4 5+3 6

Condition: X does not contain any //.

1.1.9. Phonological Evidence Against Chomsky-Adjoining the Enclitic

The Enclitic Raising Transformation as now stated moves the enclitic

to the right of the complementizer. However, I have said nothing about how

it is adjoined. I have no evidence that the enclitic is dominated by COMP,

only evidence that the /-I/ complementizer must be present in an S for the

E to attach to that S. The E may still be directly dominated by S. There

is, however, some evidence that the enclitic is not Chomsky-adjoined to the

S. This evidence is based on sentences containing two enclitics.

Since an enclitic is associated with a particular S and a complete

sentence can be composed of more than one S, we should expect to find

sentences with more than one enclitic. In fact we find two enclitics

attached to one verb. Consider the following:

 54

37. Jdan diniyahigoodef' naagha

John 2.goCOMPto.from 3.come
John comes from the place where you are going.

38. Kii i'ilbazig6/6di nahaltin

Kee 3.P.driveCOMPto.at 3.be raining
It's raining where Kee drove to.

39. Bil Kii naaghahid4e'ji' ninay&

Bill Kee 3.comeCOMPfrom.as far as 3.P.walk
Bill walked as far as where Kee comes from.

40. Jaan Kii yo'&' anaalwodide'goo' deeshaaI nizin

John Kee away 3.P.runCOMPfrom.to 1.F.go 3.want
John wants to go to where Kee ran away from.

41. t 'aa Jaan nidoog 4liji'goK

LL
Mary oolbas
just John 3.F.walkCOMPas far as.to Mary 3.drive
Mary is driving to where John will walk as far as.

(37)
NP EP V

Jaan NP E naagha
John 1 3.come
defe
from

a dinya -
to 2.go
 55

In S we find an enclitic phrase containing a null NP. Thus, the enclitic

/-g& / must move. By the raising transformation, it moves up to the nearest

(37')
N15- EP

Jaan NP naagha
John T 3.come
dff
from

NP EP C

(n|} b diniy$ 1/g

(you) 2.go to

Now consider S . There is an EP there also. But the NP dominated by the

EP is not null; it contains an S. The enclitic therefore will not move.

We have said that an enclitic normally attaches to the end of its NP. But

the NP in question is actually a whole sentence, S The enclitic will

therefore attach to the end of S1, which is in fact the same thing as

attaching onto the COMP of S1, a lower S. Thus, we find the surface word

/diniyahigood'/.1 2
The pronunciation of the word /dinya'higo'de6'/ involves a curious

phenomenon. The /-go'o'/ is longer than normal. In fact, in some speech,

 56

there is almost a pause between /-goo/ and /-d'/. I would describe this

fact as a phonological indication of the phrase structure. Thus, according

to the above tree, /-de"6'/ is outside of S1 . As is generally known, S's

tend to be the most independent of all constituents, a fact described in

our theory by an extra set of word boundaries.13 The pause, then, is a

result of # within the syntactic word.

If we postulate that the S word boundary is not erased in order to

account for the pause in /diniyahigo#de'/, and if the enclitic movement

rule Chomsky-adjoins the enclitic onto the S, then there should also be a

word boundary and therefore a lengthening or pause between the COMP /-I/
and the following enclitic. However, no such phenomenon occurs. It is

therefore reasonable to assume that the enclitic is not Chomsky-adjoined

to the S. If movement involves such adjunction, we may postulate that the

enclitic cliticizes directly to the /-i/ complementizer.1 5

1.1.10. Chorus

To recapitulate, I have shown that Navajo spatial enclitic placement

is predictable. All the examples are semantically consistent with a single

source for the enclitics, an EP. The noun in the EP may or may not be

phonologically constituted at the stage at which enclitic cliticization

occurs. If it is, the enclitic will attach to it. If it is not, and the

enclitic is stranded, the enclitic will move rightward to the nearest /-I/,
passing over as many S's as necessary. In other words, this is a case of

unbounded rightward movement, a phenomenon believed not to exist in natural

 57

language. It has been said that there is a dichotomy between languages

with leftward movement (e.g., English) and those without (e.g., Japanese).

While unbounded movement exists in the former, the latter were said to have

movement within a clause only. The facts about Navajo enclitics, however,

demonstrate that this asymmetry does not exist.

1.2. Enclitic Phrases and Relative Clauses

Since the enclitic raising transformation as postulated is evidence

against a putative universal, the obvious questions to ask are: (1) Is

this in fact a movement phenomenon at all? Could the facts explained by

my raising transformation alternately be accounted for by deletion rules?

and (2) If we are in fact dealing with a movement phenomenon, does it have

some special properties to explain it?.

I will consider question (1) first, since it is plausible to argue

that we are dealing with relative clauses rather than headless NP's. As

I discuss in more detail below, one could argue that the above facts are

accounted for by postulating underlying relative clauses with EP heads.

The "missing NP" would then be an EP deleted through relative clause forma-

tion. I will try to demonstrate, however, that at least some of the

aforementioned examples cannot be relative clauses. Moreover, whether or

not they are relative clauses, an unbounded movement rule is necessary.

Once I have demonstrated that there must in fact be movement, I will re-

turn to factors related to the second question.

1.2.1. /-1/ As Relative Clause and Complementizer Marker

The Navajo relative marker is identical to the complementizer, a

clause final /-I(gil)/. (See Platero (1974) for a detailed discussion of

the relative clause in Navajo.) Navajo relative clauses are formed

through deletion of either the head NP or the embedded NP:

42a. kinlanidd' yoo' ana'alwodigii hastiin baa 'honisin

Flagstaff.from away 3.P.run.REL man 3.about 1.know
I am aware of the man who ran away from Flagstaff.

b. hastiin kin]a'nidee' yo'o anaalwodigii baa ahonisin

man Flagstaff.from away 3.P.run.REL 3.about 1.know
I am aware of the man who ran away from Flagstaff.

When the head NP is deleted (forward deletion), as in (42b), the sentence

is identical to an embedded complement; (42b) also means "I am aware that

the man ran away from Flagstaff".

Because of this identity, if one were to argue that EP's can be

relative clause heads, the indirect questions (28)-(32) could be formed

from relative clauses where the head is an EP. For example:

28. Jaan adeesbas nazinigoo Mary bil beeh6zin

John 1.F.drive 3.wantCOMPto Mary 3.with 3.be known
Mary knows (the place) where John wants to drive to.
 59

(28)

EP

NP Mary bi]- beehozin

Mary 3.with 3.know
NP V L, goo
to
Jaan nizin I
John 3.want
NP E V

NP E

(shf) A g adeesbqs
(I) to l.F.drive

If the embedded EP is deleted under coreference to the head EP (backwards

deletion), sentence (28) is produced. If the head EP is deleted, however,

there is no good output. Because the noun in the EP is null, the string

that is formed from deletion strands an enclitic:

43. *Jaan d e' yo'd'' adeeshwo- nizinigi Mary bi- bi hdzin

The relative clause alternative is designed to replace the movement rule.

Since under this theory there is no movement, one could say that deletion

of the head is unconstrained and that the sentence is thrown out by virtue

of the stranded enclitic.

 6o

1.2.2. Insufficiencies of the Relative Clause Explanation

Even though hypothesizing a relative clause structure explains the

sentences so far discussed, there are many reasons why these sentences

should not be considered to be relative clauses. To begin with, recall

the phonological argument in section 1.1.9 which concludes that there is

no S node between /-i/ and the spatial enclitic. Since an enclitic which

is from the head of a relative clause would be separated from its relativiz-

er /-i/ by an S node, the phonological evidence argues against relative

17
clauses.

Syntactic selections also argue against this solution. Wherever a

relative clause can appear, a simple NP should also be able to appear.

Thus, the existence of (44) should predict a good sentence in (45). (45),

however, is bad.

44. Jaan naalnishdi shi]= beehozin

John 3.workCOMPat 1.with 3.be known
I know where John works.

45. *kinli' shil be'6ho/zin

Flagstaff 1.with 3.be known
I know Flagstaff.

As a corollary to the above, sentence (44) should also predict that an EP

can occur as the subject of /-1 b64hozin/ (know). We would expect that

(44) would be formed from a relative clause with an EP head. Any head

should be able to occur without the relative clause. However, sentence

(46) is also bad.

 61

46. *MITdi shil beeh6'zin

MIT.at 1.with 3.be known
I know at MIT.

Pat Brogan (n.d.) has shown that there are "strange NP's" in English

which cannot occur without a relative clause. He gives the following ex-

amples (his (5)-(8)):

47. He told me the chair he'd put it under.

48. He told me the man to arrest.

49. *He told me the chair.

50. *He told me the man.

One could argue that (45)-(46) are ungrammatical because (44) is really a

strange NP. The Navajo would therefore be parallel to (51)-(52).

51. She told me the place where he lives.

52. *She told me the place.

The difference between the English and the Navajo is that tell

allows some simple NP's while /-1 bee'h'zin/ requires sentential subjects.

Contrast the following:

53. He told me the story.

54. *hane' shil beehozin

story 1.with 3.be known
I know the story.
 62

Navajo has a transitive version of "to know", used in the sense of

"to be acquainted with". Contrast (45) and (54) with the following:

55. kinla'n be' honisin

Flagstaff 1.know
I know Flagstaff.

56. hane' beehonisin

story 1.know
I know the story.

The contrast between /bee'honisin/ and /shil beehdzin/ provides yet another

argument against calling (44) a relative clause. Contrast the meanings of

(57) and (58), two superficially identical embedded sentences.

57. kinlanigd6 deeya'hgifi beehonisin

Flagstaff.to 3.go REL 1.know
I know the person who is going to Flagstaff.

58. kinlanigo' deeyahigIf shil be'h6zin 1

Flagstaff.to 3.go.COMP 1.with 3.be known
I know that he is going to Flagstaff.
*I know the person who is going to Flagstaff.

Sentence (58) does not have the relative clause reading of (57). If the

verb /-- bedho'zin/ can have relative clause complements, as in (44), then

(58) should be ambiguous. Since the relative clause marker is identical

to the /-i/ complementizer, the embedded S in (58) should be interpreted

as either a sentential subject or a relative clause with a deleted head,

identical to (57). Since (58) cannot be a relative clause, (44) should

 63

not be either.

The above arguments strongly suggest that the indirect questions

under consideration are not formed from relative clauses.

1.2.3. EP's As Relative Clause Heads

However, there is at least one set of facts which argues for a

relative clause structure. Consider the following:

4
59a. ga;lbah a'aIngoneI yah eelwodig hatl'' '
rabbit hole in into 3.P.run.REL area.be dark,
It is dark in the hole that the rabbit ran into.

b. *galb/h a'a'angdne' yah eelwodig tl'e'e''

rabbit hole in into 3.P.run.REL be dark
The hole the rabbit ran into is dark.

Sentence (59a) appears to contain a relative clause where the head has

been deleted. It is unclear, however, whether the head is just an NP

(/a'l'n/) or an EP (/a'angKne'/). If (59a) could only be derived from

a relative clause with an EP head, then one would have to postulate that

an EP can in fact be the head of a relative clause, the position I have

been arguing against.

Sentence (59a) is derived from the following structure through

deletion of some head:

 64

(59a)

area.be dark

NP EP V REL

NP E

//I
I1 I
ga ;Lbahi a'aan gone' yah eelwod i
ra bbit hole in into 3.P.run

We know from the meaning that the head includes /a'an/, but do not know

whether it is the NP /a''n/ or the EP /a'aa'ngone'/. Since the head should

be able to occur without the relative clause, one would normally decide

between the two by seeing which conforms to the requirements of the main

verb (/hatl'ee'/). In this case, however, both can occur with /hatl'te/.

The /ha-/ in /hatlie'?'/ is the /ho-/ locative marker discussed above. Re-

call that it is obligatory whenever a locative meaning is specifically

mentioned. It is therefore obligatory with locative NP's. Thus:

60. a'ld'ngsne' hatl'e''

hole.in area.be dark
It is dark in the hole.

61. *at a'angone' ti'e'e

hole.in be dark
The hole is dark.
 65

However, the locative marker /ho-/ may also be used with simple nouns

which can be interpreted to mean "area":

62. a Iadn hatl Iee''

hole area.be dark
It is dark in the hole.

63. a'Iaan tl'e''

hole be dark
The hole is dark.

The unacceptability of (59b), however, seems to argue that the simple NP

/afa'n/ is not the head. If the head of the relative clause were the

simple NP, then (59b) should be good since the main verb in (59b) is not

locative and therefore takes a simple NP but not an EP (cf. (61) and (63)).

The ungrammaticality of (59b) might be considered irrelevant on

the grounds that a noun cannot be used both neutrally and spatially in

the same sentence: once a noun is specifically made spatial, the inter-

pretation must be used throughout a sentence. However, to see that this

is not a valid explanation, consider the following:

64. Ja'an bikooh yiyii;Ltsanigi hozaad

John canyon 3.3.P.see.REL area.be deep
The canyon which John saw is deep.

The /yi-/ prefix in the embedded clause shows that /bikooh/ as the object

of "saw" is being used in its denotative sense. As the subject of /h/zaad/,

however, /bikooh/ is being used in its locative sense (note the locative
 66

marker /ho-/). Thus it is possible for a noun to be used in both senses

at once.

The ungrammaticality of (59b) therefore suggests that a fully

specified EP can in fact be the head of a relative clause. If this is

the case, then there could be an EP head with an unspecified EP (a).

Such a head would produce the indirect questions under consideration.

1.2.4. The Inability of the Relative Clause Explanation to Replace Movement

However, notice that if we do in fact have a relative clause struc-

ture in (59a), then deletion of the embedded EP should also be possible.

Sentence (65) shows that this is not the case.

65. *ga;bahi yah eelwodigi a'ang6ne I hatl'ee

rabbit into 3.P.runREL hole.in area.be dark
It is dark in the hole the rabbit ran into.

Therefore, if we maintain that (59a) is a relative clause structure, back-

wards deletion must be constrained. Notice further that it is precisely

by backwards deletion that we proposed to derive sentences (28)-(30) under

the relative clause hypothesis. Since there is no significant difference

between (65) and (28)-(30), blocking backwards deletion in (65) would also

block it in (28)-(30). Thus, in the relative clause theory, indirect

questions would be derived through deletion of the head noun, but the

unbounded movement rule is still required. Therefore, even if we do have

a relative clause structure, it does not replace the movement rule.

 67

There is no way to explain (59a) except to assume a relative clause

structure. Even if it should be a relative clause, not all spatial in-

direct questions can derive from relative clauses. Relative clauses assume

coreference. While in all of the above examples there may be coreference,

it is clear that there is none in the following sentences:

66. galba'h a'adnglne' yah eelwodigi hatl'e'

rabbit hole.in into 3.P.runCOMPat area.be dark
It's dark in the area around where the rabbit ran into the hole.
*It's dark in the hole which the rabbit ran into.

67. Jaan da'k'eh kti'doolyahigi naalnish

John field 3.3.plantCOMPat 3.work
John works where there is a field planted.
*John works at the field which is planted.

68. Jaan da'akleh yinaalnishidi olta'idi ndeeshnish

John field 3.3.workCOMPat school.at 1.F.work
I'm going to work at the school (in the area) where John works the field.
*I'm going to work at the school at the field where John works.

In sentence (66), it is not the hole which is dark. Ellavina Perkins

(personal communication) explained that if the rabbit hole were under a

bush, one could use (66) to describe the area under the bush as dark.

Sentence (67) does not mean that John works on the field. That is made

clearer by (68), which means that I will be working near John. It does

not mean that the school is located on the field.

If (66)-(68) were to come from relative clauses, they would have

instead the starred translations. It is only in the ungrammatical

translations that the object of the main verb is coreferential with some-

thing in the embedded S. The ungrammaticality of interpretations involving

coreferentiality therefore substantiates the claim that (66)-(68) are not

relative clauses.

Although there is at least one example which argues for a relative

clause structure, it is now clear that using a structure requiring co-

reference is incorrect for at least some of the indirect questions. It

is also clear that a relative structure cannot be used to argue against

an enclitic movement rule, a rule involving unbounded rightward movement.

1.2.5. Recapitulation

In the preceding section I have argued that postulating an under-

lying relative clause structure for these indirect questions does not

vitiate the need for an enclitic raising transformation. At the same

time, I tried to argue that the indirect questions were not relative

clauses at all, but could only conclude that at least some of the cases

considered were not relative clauses. However, if it is necessary to

postulate two different types of underlying structures, then we must con-

sider the possibility that I have been describing two different types of

sentences under the rubric of "indirect question".

It has been argued for English (Baker, 1970; Bresnan, 1972; Brogan,

n.d. ; Jespersen, 1927) that there are three phenomena which are super-

ficially identical, relative clauses, indirect questions, and strange NP's.

Although I have discussed the Navajo sentences under consideration with

 69

respect to strange NP's and relative clauses, I have not actually given

any arguments that these sentences are in fact indirect questions. In

this section I will investigate further the relationship between these

Navajo examples and relative clauses and then turn to the problem of

positive proof that there is unbounded movement from indirect questions.

I will approach the problem by discussing the constituent parts of the

EP, for although I have been discussing EP's throughout, I have yet to

investigate the nature of the constituent parts. I have also said that

EP's act like NP's and yet if they were normal NP's, they should be able

to be the head of a relative clause. Investigating the constituent parts

may shed some light on this problem.

Investigating the characteristics of the constituent parts is also

the first step to determining the underlying form of 4. I will argue

that n is /haa'/, the Navajo WH PRO form, helping to justify my claim

that there is in fact an indirect question in these examples.

1.3. The EP and Its Constituents

Having noted both that EP's act like NP's (can be subjects and

objects; require a locative agreement marker in the verb which is mutually

exclusive with other subject and object markers, etc.) and that EP's are

different from NP's with respect to relative clauses, I will now compare

EP's to postpositional phrases (PP's), another special type of NP, to try

to understand the reason for the abnormal behavior.

 70

1.3.1. EP's and PPI's: Similarities

Since EP's are so similar to PP's, a comparison of the two should

provide significant information about the character of the EP. To see that

they are in fact quite similar, note the following correspondences:

6
9a. hooghang/ne' sida
house in 3.sit
He is sitting in the house.

b. hooghan yii' sida'

house 3.in 3.sit
He is sitting in the house.

70a. kingo'' adilbaas

LL.

house.to 2.drive
Drive to the house.

b. kin bich'i' ad ibgs

house 3.toward 2.drive
Drive toward the house.

Thus in the above examples, the enclitic and the postposition seem to be

similar both by virtue of their meaning and in that they combine with a

noun to form a phrase.

1.3.2. EP's and PP's: Differences

The correspondence, however, is not as neat as it appears at first

glance. Unlike enclitics, postpositions are independent words. To see

that the enclitic but not the postposition forms a phonological unit with
 71

its noun, contrast the tone on the following:

ch'inl4 ch'inal idi ch'inil bii'

Chinle at Chinle In Chinle

Navajo has a rule which changes a stressed high tone to a falling tone

when followed by a low tone within the same word. 9 This phenomenon is

apparent when an enclitic was added above to the town name /ch'ini4/.

However, when a postposition was used, no tone change occurred. It thus

seems that postpositions do not form a single phonological word with their

nouns.

Syntactically, it is also clear that postpositions are independent

of their heads. Thus, the postposition may be stranded either by movement

or deletion. Recall that when discussing scrambling (0.2) an example was

given in which the object of the postposition was interchanged with a

sentential subject, the postposition remaining with the verb:

71. Ja'an ashkii Ie'e'chaa'i yiyiiltsanigi bi be4h6zin

John boy dog 3.3.P.see.COMP 3.with 3.be known
John knows that the boy saw the dog.

In addition, Paul Platero pointed out the following fact to me:

72a. di tsekooh bii'

this canyon 3.in
In this canyon

b. dii bii'
this 3.in
In this
 72

(72b) may be used to refer to the area in which one is standing. That is,

a noun within a postpositional phrase may be deleted as redundant. Thus,

a postposition, unlike an enclitic, may be stranded through deletion of

its noun.

Recall also that the postposition includes an object agreement

marker. /bii'/ in fact means in it, while the corresponding enclitic

/gone'/ means only in. The postposition /-ii'/ takes the full complement

of person markers, /sh-ii'/ (in me), /n-ii'/ (in you (sg. and pl.)),

/b-ii'/ (in him, her, it, them), /hw-ii'/ (in one), /nih-ii'/ (in us).

In contrast, /gone'/ does not take any person markers: */shigone?/,

*/nigone'/, etc. The fact that postpositions include agreement markers

helps explain why they can be independent of their heads: the agreement

helps to "recover" the object.

It is thus clear that the relationship between a postposition and

its head is quite different from that between an enclitic and its head,

the latter being a suffix, while the former is a separate word.

1.3.3. The Classes of NP's Which Can Occur With EP's and PP's

The contrasting grammaticality in the above examples makes it per-

fectly clear that a postposition can occur in places where a spatial

enclitic cannot. In fact, postpositions occur rather freely, while the

use of spatial enclitics is very restricted. For example, enclitics un-

like postpositions can attach to only a very small class of nouns.

 73

73a. *Jdande6' y&6 ' anaishwod

John.from away 1.P.run
I ran away from John.

b. Jaan bits'' yo' andshwod

John 3.from away 1.P.run
I ran away from John.

74a. *asaa'gonet hozaad

pot in area.be deep
It's deep in the pot.

b. assa' bii' ho'zaad

pot 3.in area.be deep
It's deep in the pot.

It thus seems correct to generalize that an EP can contain only

those lexical items which by themselves can refer to place.20 That is,

an enclitic does not make a word locative, but attaches to a locative to

provide directional or spatial information. A postposition, in contrast,

can create a locative interpretation for almost any noun. Thus, unlike

the above grammatical examples with PP's, in all the good examples I have

of EP's, the enclitics attach to words which in themselves refer to place.

For example, they attach to place adverbs:

tl'do'di
at outside

from there

kodi
at here
 74

Recall that there is a verb agreement marker /ho-/ which marks

locative subjects and objects. Since the above definition requires that

a lexical item refer to a place, we should be able to define those items

21
which can take enclitics in terms of those which can take /ho-/ agreement.

Consider the following:

75a. hooghangone' hotsaa

house in area.be big.
The house is big.

b. hooghan hotsaa
house area.be big
The house is big.

76a. bikoohgoyaa hozaad

canyon.down in area.be deep
The canyon is deep.

b. bikooh hozaad
canyon area.be deep
The canyon is deep.

77a. a'la'angone' hatl'e'e

hole.in area.be dark
The hole is dark.

b. a'a'an hatl'e'e
hole area.be dark

In contrast, the sentences in (73a) and (74a) are bad because (78) and

(79) are bad.

 75

78. *Jaan hotsaa

John area.be big
John is big.

79. *asaa hozaad

pot area.be deep
The pot is deep.

1 3.4. Postpositions and Relative Clauses

Related to the independence of a postposition and particularly to

its ability to stand alone is the fact that it is possible to form a rela-

tive clause on the NP within the PP:

6
80a. yii' sidihigi kin shil yal't'le' h
3.in 3.sit.REL house 1.with 3.be good
I like the house he is sitting in.

b. kin yii' sidahigii shil ya a't'e'h

house 3.in 3.sit.REL 1.with 3.be good
I like the house he is sitting in.

The relative clause in sentence (80a) results from the deletion of the

embedded NP. (80b) is the result of deleting the head NP. Note that it

is clear that the head of the relative clause in both (80a) and (80b) is

the simple NP /kin/ and not the PP /kin yii'/. Thus, if a PP head were

possible, (81a) should be good. The fact that (81a) is bad and (81b) is

good proves that /kin/ is the head.

 76

81a. *kin bii' shil ya'at'eeh

house 3.in 1.with 3.be good
I like it in the house.

b. kin shil y'aIt 'h

house 1.with 3.be good
I like the house.

In contrast, the comparable structure with an enclitic instead of

a postposition has no good output.

82. *go'ne' sidahigif kin shil ya'Kt'44h

in 3.sit.REL house 1.with 3.be good
I like the house he is sitting in.

83. *sidahigone' kin shil y/It'teeh

3.sit.RELin house 1.with 3.be good
I like the house he is sitting in.

Whatever a PP is, it must be composed of an NP and a postposition. Thus

sentential clitics which go into second position after phrases will come

between an NP and its postposition but not between a noun and a spatial

enclitic.

84a. kin ga' yii' sida'

house emph 3.in 3.sit
He is sitting in the house.

b. *kin ga' gone' sida/

house emph in 3. sit
 77

Given the other evidence differentiating PP's and EP's, it is

possible to account for the varying relative clause possibilities by

postulating different structural configurations for EP's and PP's. As-

suming that both EP's and PP's are NP's, it is still necessary to deter-

mine whether that NP is formed from another phrase or from a simple noun,

from (a) or (b):

(a) NP (b) NP

NP N

Since the head of a relative clause is an NP, its antecedent must be also.

We know that a relative clause can be formed on the head of a PP ((80))

and therefore must postulate that the head is an NP, not an N (structure

(a)). On the other hand, since a comparable relative clause cannot be

formed from an EP ((82)), we could account for that by postulating struc-

ture (b) for EP's.

Note that this choice of constituent structure is consistent with

all the differences already noted between EP's and PP's. A phrase is

generally considered more independent than a lexical category. Transforma-

tions refer to phrases more than categories, phrases are more likely to be

moved than lexical categories, etc. To say that a postposition is inde-

pendent of its head is also to say that the head is independent of a

postposition.

A final strong argument that the head of a PP, but not an EP, is

an NP comes from Platero (1974). He notes that PP's but not NP's may
 78

appear with a demonstrative determiner: (his (103) and (106))

85. (sh ) dil tsekooh bii? se'da

I this canyon 3.in 1.sit
I am sitting in this canyon.

86. *(sh() di tsskoohdi seda

I this canyon.at 1.sit
I am sitting at this canyon.

The fact that the noun in an EP cannot be modified is evidence that it

is not an NP.

1.3.5. Phrases With Both Postpositions and Enclitics

One problem with the above postulated constituent structures arises

from cases in which an enclitic has attached to a postposition. Consider

the following:

87a. kin yiigo'ne' sidashigii shi4 yaIt I e/h

house 3.in.into 3.sitCOMP 1.with 3.be good
I like the house he is sitting in.

b. yiigone' sidahigifi kin shil ya'it'eeh

3.in.into 3.sitCOMP house 1.with 3.be good
I like the house he is sitting in.

88a. *sidahiyii'gonel shil beehozin

3.sitCOMP.3.in.into 1.with 3.be known

b. *yii? sidahigonet shil beehozin

3.in 3.sitCOMPinto 1.with 3.be known
 79

Sentences (87a)-(87b) demonstrate that a postposition enclitic combination

can be stranded by relativization. As we have seen, this process occurs

to PP's but not to EP's. Sentences (88a)-(88b) show that movement by the

enclitic raising transformation is not possible with a postposition en-

clitic combination, although this process is associated with enclitics.

We should thus assume that a postposition plus enclitic occurs in a post-

positional phrase not in an enclitic phrase. The lack of movement can be

accounted for either by saying that there is no EP or by saying that the

enclitic has cliticized to the postposition and is therefore not stranded.

Recall that only stranded enclitics move.

The problem occurs in trying to determine the structural configura-

tion specific to these phrases containing both a postposition and an

enclitic. The following trees depict the possible alternative structures

for the phrase in sentence (87):

(87A) NP (87B)

P E
NP P E

kin yii' gone' kin yii' gone'

(87C) PP (87D) P

NP P NP

N E
k
I I/I 'ki g
kin gone' yiiI kin yii' gone,
The tree in (87A) is an example of the type of constituent structure I

argued against above, one in which an enclitic is a sister to a phrase

instead of to a lexical category. In addition, the structure claims that

the phrase should behave like an EP rather than a PP. Further, if a PP

is an NP, then (87A) predicts that /kin yii'/ is a possible head of a

relative clause. By backwards deletion, the enclitic would be stranded

and would then move to the /-l/, creating a sentence such as (89).

89. *sidahigone? kin yii' shil ya'at'ee'h

3.sitCOMPinto house 3.in 1.with 3.be good

The fact that (89) is ungrammatical provides another argument against

(87A).

The structure in (87B) on the other hand tells us nothing about

whether these are to be considered enclitic phrases or postpositional

phrases. The choice is arbitrary and therefore behavior should be random.

However, the behavior is quite consistent.

Structure (87C) predicts the correct behavior, but provides the

wrong surface order. Besides the fact that enclitic movement is not

motivated from a fully specified noun, such a transformation would be

unnatural because clitics do occur between a noun and its postposition.

Our last structure makes the correct claim about postposition

precedence and about surface order. It also makes the correct claim that

/yii'go'ne'/ but not /kin yii'/ is a phonological unit. While I believe

(87D) to be the best structure, accepting this structure means resting

our answer on the question of what it means to be an enclitic. How can

a postposition be composed of a postposition and an enclitic? Even more

important, I argued above that there should be only one source for en-

clitics. I have suggested that that source is an EP having the constituent

structure of (b) (p. 77). Now I am suggesting a second constituent

structure for enclitics.

I would like to suggest that even (87D) is suspect and that these

enclitics have no status of their own. Instead, I would suggest that some

kind of contraction has occurred forming a new postposition. While I have

no direct evidence for this hypothesis, there are other contraction rules

in Navajo which change constituent structure. As Kenneth Hale pointed out

to me, /to'biih/ (water in) may contract to /taah/. Consider the follow-

ing:

90a. to biih yi'a

water 3.in 3.1.P.handle
I put it in the water.

b. taah y3'a
water.in 3.1.P.handle
I put it in the water.

(90a) forms relatives like a normal PP:

91. awe'e to biih ya4tinigii sik'az

baby water 3.in 3.3.handle.REL 3.be cold
The water I put the baby into is cold.
In (91), /to/ was the head of a relative clause and the head was deleted.

Once contraction occurs, it is impossible for /to/ to be identical to the

embedded noun.22 Thus:

92. *awe'e taah yiltingii sik'az2 3

baby water.in 3.3.handle.REL 3.be cold

Therefore, whatever /taah/ is, it is no longer composed of an NP plus P.

Analogously, I would suggest that the postposition plus enclitic

is no longer a separate P followed by an E, but has contracted to a single

lexical item. Since these phrases act like PP's, that single item should

be considered a new postposition.

If this argument holds, then there will be one constituent structure

for PP's, that of (a) on page 77 above. More importantly, we will be able

to retain the claim that there is only one source for enclitics, (b) above.

In either case, however, notice that the generalization still holds that

an enclitic will only attach to an item which is inherently locative. We

will return to this claim in trying to define 6.

1.3.6. PP's, EP's and Other Types of Relative Clauses

Before turning to the characterization of A, let us consider the

last case of possible relative clauses involving PP's and EP's. We have

considered forming relative clauses on the objects of the postposition

and enclitic and saw that the first was possible but the second was not.

The difference in behavior was accounted for by different constituent

structures. Nothing has been said so far about forming a relative clause

on the entire PP and EP. Since both are NP's, such relative clauses

should be possible. To see that the entire PP can be the head of a rela-

tive clause, consider the following:

93a. kin biyii' yah iiyahigii shil ya'ahoot ee4h

house 3.in into 1.P.go.RE 1.with 3.area.be good
I like it in the house I went into.
// / I /

b. yah iiyahigii kin biyii' shi4 ya'ahoot'eeh

into 1.P.go.REL house 3.in 1.with 3.area.be good
I like it in the house I went into.

94a. kin biyii' shil y/la/hootte'eh

house 3.in 1.with 3.area.be good
I like it in the house.

b. *kin shil ya'a hoot '&h

house 1.with 3.area.be good

The fact that (94b) is ungrammatical proves that the head had to be the

entire postpositional phrase, /kin biyii'/.

The PP head also shows up in (93b) in which backwards deletion has

occurred. (93a), formed through forward deletion, proves that the ante-

cedent is also the full PP, as we would expect given the PP head in (93b).

We have already seen one example which needs an EP head of a rela-

tive clause. I repeat (59a) below:

 //4/ /1/
59a. ga;lbahl a'aa'ngone' yah eelwodigii hatl'eel
rabbit hole in into 3.P.run.REL area.be dark
It is dark in the hole that the rabbit ran into.

While (59a) requires us to postulate that an EP can be the head of a

relative clause, it seems there must be other restrictions on the forma-

tion of relative clauses. Consider the following:

(95)

8P hodiiltlah
3.P.fire breaks out

kin di
house at

l\P P REL

NE

man house at 3:sit

95a. *hastiin sidahigii kindi hodiiltlah

man 3.sit.REL house.at area.3.P.fire breaks out
Fire broke out at the house where the man is sitting.

b. *hastiin kindi sidahigii hodiiltlah

man house.at 3.sit.REL area.3.P.fire breaks out
Fire broke out at the house the man is sitting at.
 96a. kindi hodiiltlah
house.at area.3.P.fire breaks out
Fire broke out at the house.

b. hastiin kindi sida

man house.at 3.sit
The man is sitting at the house.

Sentences (96a) and (96b) prove that the tree should be well formed. Yet

there is no good output. While I have no explanation for this phenomenon,

I point it out to emphasize the need to further illumine the behavior of

enclitic phrases and the implication of that behavior for other syntactic

processes. In particular, an explanation for (95) might help explain

whether the sentences I have argued are indirect questions are also am-

biguous, and if not, why not.

1.3.7. Meaning Problems For Indirect Questions As Relative Clauses

I do feel, however, that the constituent structure of EP's already

provides an explanation for why indirect questions do not come from rela-

tive clauses. Consider the following:

97. Jaan naaghahide' baa ahonisin

John 3.comeCOMPfrom 3.about 1.be aware
I am aware of where John comes from.

To make sense as a relative clause, the head would have to be a plain NP

rather than an EP: "I am aware of the place" makes sense; "I am aware of
from the place" does not. However, the embedded clause contains an EP:

"John comes from A," not "John comes /". Thus, to be a relative clause,

(97) would have to be formed from a relative clause within the EP, not on

the entire EP: 24

(97)

NP

11E REL A baa ahonisin

3.about 1. aware

NP E

Jaan d ) naagha'
John from 3.come

This, however, is precisely the class of possible relative clauses that has

definitely been eliminated. The impossibility of forming a relative clause

on the object of an EP (as opposed to the object of a PP) has been shown to

be inherent to the character of EP's.

1.4. Movement and Indirect Questions

Having completed the investigation of the relationship between the

examples involving rightward movement and a relative clause structure, I

would now like to try to provide positive proof that these sentences are

in fact indirect questions. Since I want to discuss the implications for

question formation universals as well as for the Right Roof Constraint,

the proof by process of elimination which I have used so far may be con-

sidered inadequate.

1.4.1. Identification of

Since questions require a question morpheme-either a WH word or

a signification for a yes/no question--I will begin by arguing that A

is underlyingly the Navajo WH morpheme /ha /.2 5 I have been asserting

that enclitic raising exists because the enclitic has been left stranded;

the N in the EP is phonologically empty. There are, however, several ways

to account for a null N: (1) it was a dummy in underlying structure;

(2) the noun deletes; or (3) the noun has been moved away. We can

eliminate (1) on the basis of the character of the EP, verb morphology,

and meaning.

1.4.2. 4\ Is Not a Dummy Node

Recall from the investigation of the constituents of an EP that

enclitics were found to attach only to lexical items which are inherently

spatial. If we allow a dummy node, we will have to abandon that generali-

zation. Notice that generating EP constituents randomly and using a surface

filter would be insufficient, precisely because of the enclitic raising

transformation. Thus, in a case like (98), the enclitic /-di/ is attached

to the COMP /- / which does not refer to place.

 6
98. shinaai ndeeshnish nfzinidi shilk beh 'zin
1.brother 1.F.work 3.wantCOMPat 1.with 3.be known
I know where my brother wants to work.

Saying that the enclitic is really attached to the verb does not help

either, since there is nothing spatial about want (/nizin/). On the

other hand, while the generalization is inconsistent with postulating a

dummy node, it is consistent with the translations of all the examples.

They all translate as if the missing noun did indicate place. It is

therefore reasonable to assume that the A in EP's is not a dummy node,

but a lexical item indicating "place".

A second argument that A is not a dummy node is provided by the

verb morphology in (99).

99. shizhe'e' yiyifya"igii shil b4e'ho'zin

1.father 3.3.P.eatCOMP 1.with 3.be known
I know what my father ate.

In this case the L is not part of an EP, but is the direct object. In

section 1.1.4. I argued that the presence of /yi-/ proved that there had

to be an underlying object and, more particularly, a specified reference.

Recall that /'i-/, not /yi-/, agrees with PRO direct objects. Thus, un-

less there were some way to interpret a dummy node before direct object

agreement occurred, a dummy node would be incompatible with /yi-/. There

is, however, nothing in (99). We therefore must conclude that 4\, has a

distinct underlying form and is not a dummy node.

 To state the argument another way, a dummy node implies non-

distinctness of meaning. However, in every example mentioned, & always

gets translated as a WH word.26 The distinct meaning therefore argues

against a dummy node and for a specified underlying form.

1.4.3. Evidence That / Is /ha4/

Having argued against an empty node, I will now try to motivate

the specific underlying form, a WH word. So far, the only suggestion

that g is /ha'a// comes from the English translation: Whenever there

has been a gap in the Navajo, there is a WH word in the English. While

translations provide a reasonable starting point, since "translations"

are really more like paraphrases, more direct justification is called

for.

Using the English as a guide, let us turn to a consideration of

Navajo WH question words, the words which appear in the English transla-

tions. Consider the following:

lOa. Jaan deeyah!go' shil b46ho'zin

John 3.goCOMPto 1.with 3.be known
I know where John is going.

b. haaoosh Jaan deeya'

where.to.Q John 3.go
Where is John going?

lOla. Jaan yig;lid ' shi;L bihzin

John 3.walkCOMPfrom 1.with 3.be known
I know where John is coming from.
 90

b. hld</'sh Jdan yigaa

where.from.Q John 3.walk
Where is John coming from?

102a. Jaan naalnishidi shiI beeho/zin

John 3.workCOMPat 1.with 3.be known
I know where John is working.

b. haadish Jdan naalnish

where.at.Q John 3.work
Where is John working?

The (b) sentences are typical direct questions. The question word is

made up of three parts. /h -/ is the indefinite form for where which

also shows up in the indefinite pronouns for somewhere. /-sh/ is an in-

terrogative particle. Whereas English makes questions through subject-aux

inversion, Navajo uses interrogative particles. /-sh/, /la//, and /-sha'/

are used in WH questions; /-ish/ and /-sht'/ in yes/no questions. The

system for forming simple questions is quite complicated because there

is a variety of interrogative particles as well as variation in where in

the sentence they are attached. I will discuss the system in more detail

in Chapter 2 in order to discuss embedded direct questions, but I do not

believe that any other characteristics of the interrogative particles are

relevant to the present discussion. However, the third part of the "where"

words is crucial in the comparison between these simple direct questions

and the indirect questions. Compare each "where" word with the correspond-

ing enclitic in the (a) sentence. Phonologically each "where" word

 91

contains its corresponding enclitic. Notice also that it is impossible

to replace the enclitic in (a) with the full WH word (with or without the

interrogative particle).

103a. *Jaan ha'go' deeyahigii shil beehozin

b. *Jaan haa'go'o'sh deeyhigi shil beghozin

4 shil beehozin
104a. *ha'd'41 Jaan yig i/gi

b. *haad44" la'Ja'an yigaligii shil beehozin

The semantic and morphological similarities as well as the complimentary

distribution could all be accounted for by postulating that 4 is a WH

word and the (a) and (b) sentences had the same underlying form for that

WH word.

Another way to determine what a sentence means is to examine how

it would be used. Consider the following questions and responses:

105. Q: shinaaf yiskahig ish niI be'hozin

1.brother 3.3.P.shootCOMP.Q 2.with 3.be known

R: nda, ha'at' ish ninaai yiskah

no what Q 2.brother 3.3.P.shoot with an arrow
No, what did your brother shoot (with an arrow)?

106. Q: Jaan deeyahigoosh nil beehozin

John 3.goCOMPto.Q 2.with 3.be known

R: auut, na'n izhoozhig deeshfal nii ni'

yes Gallup to 1.F.go 3.say pst. assertive
Yes, he said he's going to Gallup.
 92

The above questions really contain two questions, the direct question "do

you know?" and an indirect question, asking in (105) what my brother shot,

and in (106) where John is going. That there are two questions is clear

from the responses. One would reply "yes" or "no" only to a direct ques-

tion and only to a yes/no question. On the other hand, one adds more

information to the response only if there is also an indirect question,

if more information is requested. Contrast (105) to (107).

107. Q: Jd/an shinaa yiskahigiish nil beehozin

John 1.brother 3.3.P.shootCOMP.Q 2.with 3.be known
Do you know that John shot my brother (with an arrow)?

R: auu', (Jaan ninaa yiskahig i nil bee/hozin)

yes John 2.brother 3.3.P.shootCOMP 2.with 3.be known
Yes, (I know that John shot your brother (with an arrow)).

Since (107) asks only whether you knew about something, i.e., were aware

of a fact, a simple "yes" or "no" is an appropriate answer. Notice that

if any more is said, it is merely a repetition of the question, adding

no new information.

In contrast, if a direct question containing an indirect question

is answered with a simple "yes", the questioner will feel the answer is

only partial and therefore inappropriate and will request more information.

Thus, if (105) had been answered as in (107), the questioner would come

back with
 93

108. ako ha'at' sh ninaa1 yiskah

thus what Q 2.brother 3.3.P.shoot with an arrow
So what did your brother shoot (with an arrow)?

In sum, the way these sentences are used also argues that they are

indirect questions.

Lest it cause any confusion, it should be noted that the interroga-

tive particle in the above questions is the yes/no interrogative particle

and is not responsible for the indirect question. Recall that the indirect

questions discussed earlier did not contain the particle. To see that my

assertion is correct, consider the following:

109. Jaan deeyahigo'6 nikish be hozin

John 3.goCOMPto 2.with.Q 3.be known
Do you know where John is going?

110. *Jaan deeyahig o'la'nil behozin

John 3.goCOMPto.Q 2.with 3.be known

In the examples I have been using, the interrogative particle was attached

to the end of the subordinate clause, leaving it unclear whether it was

marking a question in the higher or lower S. (109) demonstrates that the

interrogative particle is not actually part of the embedded indirect

question, i.e., is not a leftover of the deleted WH word. The fact that

the WH interrogative particle /la/ may not be used ((110)) also proves

that the particle in (105)-(107) is the yes/no particle.

One final argument that the sentences under consideration are in

fact indirect questions and that 4 is a WH word is based on the following

 94

idiom, pointed out to me by Kenneth Hale:

111. Q: hatat'ish baa nanina

what Q 3.about around.2.go
What are you doing?

R 1: *naanish baa naasha

work 3.about around. 1.go

R2 : naashnish
1.work
I am working.

The expression /baa na-aal/ requires a question word. It cannot be used

with a regular object. Thus, this expression cannot be repeated in an-

swering (111). The fact that this expression can be used in the following

argues that they are in fact questions:

112. baa naashihigish nil beehozin

3.about around.1.doCOMP.Q 2.with 3.be known
Do you know what I am doing?

113. shizh'le yaa naaghahigii doo shil be'hozinda

1.father 3.about around.3.doCOMP neg 1.with 3.be known.
I don't know what my father is doing.

Since the embedded S's contain an expression requiring a question word,

there must have been a question word underlyingly although there is none

in the surface.

In sum, all the examples discussed argue that we have in fact been

considering Navajo indirect questions which contain an underlying WH word.

 95

I in fact know of no arguments against postulating that A is underlyingly

a WH word. However, it should be noted that postulating a question forma-

tion rule which deletes the WH is unusual. But then so is the rightward

movement of the spatial enclitic. Perhaps one unusual fact can be used

to explain the other.

1.4.4. Further Considerations

To account for why there is any movement at all requires investi-

gating other factors in detail, such as the other Navajo question types

and the semantics of the complementizer system. However, having argued

against accounting for the spatial enclitic placement through deletion

instead of movement, and having described in as much detail as possible

the underlying forms involved in the movement, I will now return briefly

to the question of whether this movement phenomenon has any special pro-

perties to explain it. There are two particularly interesting characteris-

tics deserving further mention.

1 4.5. The Rightward Movement Is In Indirect Questions

One is that the movement occurs in indirect questions. It thus

reflects not only on the Right Roof Constraint but on the putative question

formation universals as formulated by Baker, Bach, and Bresnan, and dis-

cussed in the introduction to this chapter. According to them, there is

only one possible movement rule for questions, and it moves the questioned

word to clause-initial position. If this movement is to be considered a

 96

question-movement transformation, it clearly violates that claim. On the

other hand, Bach claims that question movement is toward the "governing

verb" and Bresnan (1970) and Chomsky (1971) claim that all movement out

of an S is through complementizer position. This movement is clearly

consistent with either one of these formulations.27 Recall, however,

that my formulation of the movement transformation was dependent on the

previous deletion of the noun--i.e., whenever a noun is phonologically

constituted, the enclitic does not move. If that formulation is correct,

then one could postulate that the actual question formation is the WH

deletion and that the movement rule is not in fact a question formation

rule. Thus, while WH deletion remains unusual, it does not violate the

question formation universal. In this particular case, Bresnan's COMP

Substitution Rule is not violated, although her Complementizer Attraction

Universal is.

In some current research, Hale and Platero28 are investigating the

possibility that Navajo relativization is a movement rather than a dele-

tion rule. They have suggested that by modifying my enclitic raising

transformation, relativization might be collapsed with indirect question

formation. The structural description for the enclitic raising transforma-

tion now includes a deletion site. If Hale and Platero are correct, then

the environment for movement can no longer be based upon deletion because

relativization would involve no deletion. One should therefore reconsider

whether WH-deletion occurs before or after movement and whether the raising

transformation is in fact a question formation rule. Such reconsideration

must await further refinement of their proposal.

 97

1.4.6. The Unbounded Movement Is Not Unconstrained

The second interesting characteristic of this rightward movement

phenomenon has to do with its unboundedness. While unbounded in the

sense that the enclitic moves across a crucial variable, it is not un-

constrained. The enclitic never passes over an /-i/ complementizer. Of

course, it has often been noted that question formation in English also

must be constrained. For example, as Langacker (1969) points out, WH

words from indirect questions move to the front of their clause, not to

sentence-initial position.

114. John asked when Mary had arrived.

115. *When did John ask Mary had arrived.

Nor can two WH words ever be moved to initial position.

116. *Who where did John say is going?

Bach uses the concept of movement toward but not past the governing verb

to handle these restrictions; Chomsky (1971) has a Tensed S constraint

and Specified Subject constraint. In the Navajo case, however, since

the movement is only across the direct discourse segment of the sentence,

the absence of an overt complementizer could lead one to question whether

the movement is truly across an indefinite number of S's. One might argue

that some kind of clause reduction has occurred, perhaps through pruning,

and the enclitic therefore moved to the main verb of its clause. Under

Ross's hypothesis (Ross, 1966), the S node would be pruned if unbranching.

Since there is no complementizer, it would prune in the above cases for

lack of a subject. However, although not specifically discussed here,

as we saw in the Introduction, Navajo direct discourse S's can have

surface subjects. In those cases, the enclitic will still move.

117. Mary Ja'an doogaal nizinigoo Bil bil beehozin

Mary John 3.F.go 3.wantCOMPto Bill 3.with 3.be known
Bill knows where Mary wants John to go.

As we have already seen, go, but not want, takes directional complements.

(118) is the usual form:

118. Mary Ja'an kinlanigKo' dooga' nazinigil Bil bi;L beehozin

Mary John Flagstaff.to 3.F.go 3.wantCOMP Bill 3.with 3.be known
Bill knows that Mary wants John to go to Flagstaff.

Example (117) is the result of Enclitic Raising from a A (a deleted WH

word).

Another possibility is that direct discourse verbs raise, which

would also trigger pruning in the above examples. There is, however,

neither phonological nor morphological change in the embedded verb to

suggest any raising. The most convincing way to show that the lower verb

has not been raised is to produce sentences with material from the lower

S intervening between the two verbs. Consider the following:

119. ?Jaan kina'nigoo' adeesbas yisk4go nizin

John Flagstaff.to 1.F.drive tomorrow 3.want
John wants to drive to Flagstaff tomorrow.
 99

Since Navajo is a relatively strict verb final language, especially in

embedded S's, it is difficult to find natural sentences with material to

the right of the verb. Thus, (119) is considered awkward but acceptable.

It is clear, however, that the awkwardness is not related to the problem

of clause reduction in direct discourse. I thus think it fair to con-

clude that (119) is an argument against any verb raising.

Note, however, that arguing against clause reduction is not the

same as arguing that it is irrelevant that the enclitic moves only as

far as the /-I/ complementizer. The fact that there is no clause reduc-

tion simply means that it may be some other property of the /-I/ which

causes the enclitic to attach to it, a subject to be discussed in

Chapter 3.

Chomsky has suggested that the crucial factor is that movement is

never across a COMP node. Thus he follows Bresnan (1970) in distinguish-

ing S and S':

S >COMP S' (unordered)

S' ->NP Aux VP

Since movement is never across an overt complementizer, it could be

described as passing an indefinite number of (S')'s, but would be move-

ment to the first S. One must then question whether the Right Roof

Constraint should be reformulated to differentiate S' from S.

Although I have refrained from postulating a COMP node on direct

discourse S's, I must reiterate that it was an unmotivated decision,

 100

based solely on the fact that there is no surface complementizer. It

should be emphasized, however, that I can conceive of no sense in which

direct discourse S's are "reduced S's", as infinitival S's are in English.

In Chapter 2, I introduce more evidence which demonstrates that direct

discourse S's act like full S's.

1.4.7. Conclusion

In this chapter I have argued that Navajo requires some unbounded

rightward movement rule to account for enclitic placement. I consider the

version of the rule I have introduced to be tentative because that version

is extremely narrow in two respects. Written as a language particular rule,

it does not consider whether there is any relationship to clitic movement in

other languages. Secondly, further investigation is required to determine

whether the rule can even be generalized for Navajo itself. I have just

argued that this movement occurs in indirect questions and that these ques-

tions contain an underlying WH word rather than an empty node. Since there

is an underlying WH word, it is possible that the movement is triggered by

the WH rather than the deletion site. If triggered by the WH, then it may

be possible to generalize the rule to include relative clauses as Hale

suggests. However, since a more general rule cannot be properly formulated

at this juncture, I will continue to refer to movement from a deletion site

because the explanation for the movement that I will propose in Chapter 3

is more obvious under this formulation.

The explanation for this movement becomes clear when indirect

questions are compared and contrasted to other Navajo questions. Chapter 2

does just this.

 101

FOOTNOTES TO CHAPTER 1

1. Note that this is further generalized to include relative clauses

(cf. The Expansion Universal). However, the /-\ is irrelevant here.

2. There are two forms of the locative enclitic, /-di/ and /-gi/. I

propose to treat them as variations of the same enclitic because they

act alike: they have much the same meaning, occur in the same places,

and can often occur with the same word.

3. Although I assume that an enclitic phrase is some kind of noun phrase,

I will continue to refer to it throughout the discussion as an EP, in

order to distinguish it from NP's.

4. Chapter 3 deals with cases where embedded S's do contain WH words in

their surface string. However, most, if not all, are cases of direct

not indirect questions.

5. It is interesting to note that although there is no Navajo correspond-

ing to who in this sentence, it is not ambiguous. As (23)-(24) demon-

strate, sentences with missing NP's can mean who or what, but not why,

where, etc. As I show later ((28)-(30)), where sentences require an

enclitic. There is in fact no way that I know of to get why, when

referring to the future (from Navajo /hahgo/) in exactly this con-

struction. When referring to the past (from Navajo /had#'/ is pos-

sible with the past tense form of the /- / complementizer and an

enclitic.
 102

i. Mary niya'hdada' shil b eh6zin

Mary 3.P.comeCOMP 1.with 3.be known
I know when Mary came.

6. Cf. Young and Morgan (1971).

7. Note that (24) also means "Tell me about his stealing my car". Since

there is no trace of who, the subject may also be interpreted as a

deleted third person pronoun.

8. Although in principle the scope of the movement is indefinitely large,

it is in fact difficult to construct examples where movement involves

more than three S's. One reason is that the complex sentences are

extremely difficult to process, and judgments vary randomly. Another

reason is that I know of very few verbs which take no /- /, and the

sentences become even more difficult if the subjects of the different

verbs are not identical.

9. Hale and Platero (1973) discuss an example of an enclitic attached

directly to the verb, i.e., without an /-Q/. However, the exact status

of such sentences is now in doubt. I suggest that they are actually

cases of deletion of the /-l/, resulting from fast speech.

10. Note that if direct discourse S's have no underlying COMP, it may not

be necessary to mention /-I/ specifically, but only the COMP node.

Such a system is possible only if the transformation which deletes

WH words is constrained to /-1/ clauses, as contrasted with /-go/

clauses. Cf. Chapter 3.

 103

11. Although I do not argue for the existence of a COMP node, I have

included such a node in the tree. While not necessary to my argument,

I believe it is useful for expository purposes. I have also chosen

to attach the raised enclitic to the COMP node rather than directly

to S to emphasize the fact that it forms a phonological unit with

12. This argument claims that every possible combination of enclitics

should occur. In fact, this is not the case.

ia. *Jaan naalnishidig6 deesha'a

John 3.workCOMPat.to 1.F.go
I am going to where John works.

iia. Jaan naalnishigoo deeshaaI

John 3.workCOMPto 1.F.go
I am going to where John works.

iiia. *Ja/an deeyahigoogoo adeesbas nisin

John 3.goCOMPto.to 1.F.drive 1.want
I want to drive to where John is going.

iva. Jaan deeyahigso adeesbps nisin

John 3.goCOMPto 1.F.drive 1.want
I want to drive to where John is going.

From the meaning of these examples and what we know about EP be-

havior in simple sentences, we would expect two enclitics to occur

underlyingly in both (i) and (ii), although they cannot appear on

the surface.
 104

(iii) a. Jaan kindi naalnish

John store.at 3.work
John works at the store.

b. kingoo deeshaka
store.to 1.F.go
I will go to the store.

(iv) a. Jaan da'aktehgoo deeya

John field. to 3.go
John is going to the field.

b. da'akt ehgoo adeesbas

field. to 1.F.drive
I will drive to the field.

I would argue that there are two enclitics underlyingly and propose

two environments for enclitic deletion:

(1) When there are two identical adjacent enclitics, delete one.

(2) When /-di/ (and possibly /-ji'/) occurs before another spatial

enclitic, delete it.

13. Cf. Chomsky and Halle (1968) for a discussion of word boundaries.

14. No special pause is perceptible in (39)-(41) because the first

enclitic in each case ends in a glottal stop which provides a

natural pause.

15. If this argument is valid, then one should consider whether it helps

to determine whether the COMP /-i/ is Chomsky-adjoined to the S.

There is no lengthening or pause between the verb and the /-i/ as

 105

predicted for such an adjunction. On the contrary, long high tone

stem vowels are usually shortened. I conclude therefore that if

there was a word boundary, it was eliminated. However, I have no

way of determining whether there is a rule which incorporates com-

plementizers into the verb. That is, one must account for the fact

that a complementizer will cliticize to verbs only. It is impossible

to get another word in the same S to the right of the verb when there

is an overt complementizer, although it is possible when there is no

complementizer (i.e., in matrix S's and embedded direct discourse

S's). Cf. example (119) and Chapter 2.

16. I am assuming that Navajo relative clauses have a fully specified

head. Chomsky has postulated that relative clauses should have

dummy heads and there should be a transformation to fill the head.

If I were to adopt Chomsky's theory, my case would already be proven.

17. This argument does not hold in a theory in which the -/ and the

enclitic are generated together as a Specifier and then moved to

REL or COMP position. See also section 1.4.5., section 2.5.7., and

Hale (forthcoming B) for further discussion.

18. The sentence should in fact be ambiguous, but with a third reading:

"I know who is going to Flagstaff". Thus, as I argue, there should

be three possible trees, (1) a relative clause, (2) a sentential

clause containing a non-empty NP, and (3) a sentential clause with

 1o6

a /. This third is the indirect question. No one has yet investi-

gated why no sentence seems to be three ways ambiguous.

19. So far as I know, this rule applies only to stem vowels. However,

I know of no argument which determines whether this rule applies when

there is a word boundary between the stem and the suffix. This

phenomenon, therefore, cannot at this juncture be used to decide

whether there is a word boundary between the verb and the comple-

mentizer. Phonological evidence of a word boundary would argue for

Chomsky-adjoining the COMP node. This issue will come up again in

Chapter 3.

20. This generalization provides another argument for enclitic movement.

If an enclitic were to originate on a verb, it would attach to verbs

which are not spatial and thus the generalization would be lost.

21. Obviously, the /ho-/ is relevant only insofar as the lexical items

containing this /ho-/ are locative and not directional.

22. I am unable to determine whether this is a pre-underlying structure

morphophonemic rule or a transformation. Obviously, my case is much

stronger if it is /taah/ that occurs in the underlying structure.

23. This sentence is grammatical with the meaning, "The baby I put into

the water is cold."

24. And then, of course, the enclitic raising transformation would have
 107

to apply to move the enclitic. Thus, even if possible, this would

enhance not eliminate the need for the transformation.

25. There are two WH morphemes, /ha'/ and /haa/. I have only discussed

the former in this chapter. The latter occurs incorporated into the

verb, i.e., never as an independent word (cf. 2.3.7). Perhaps this

difference could be used to account for the deletion. In any case,

I have no way of determining which of the two should be postulated

as the underlying form.

26. I am assuming that the sentential subject readings (e.g., "I know

that he ate it") do not come from A's but from anaphoric pronouns

which delete in Navajo in general.

27. Although as we will see in Chapter 2, not all Navajo questions are

consistent with these formulations.

28. This idea was first discussed in Hale and Platero (1973). Also see

Hale (forthcoming B).

 CHAPTER 2

DIRECT DISCOURSE QUESTIONS AND UNBOUNDED

LEFTWARD MOVEMENT

2.1. Simple Questions

Before looking at direct questions in embedded clauses, it is

necessary to have some idea about the general method of forming questions

in Navajo. Let us therefore take a brief look at simple questions first,

i.e., questions in sentences composed of only one S, for even here we find

that the system is rather complicated. And, unfortunately, simple ques-

tions have never been studied in detail.

2.1.1. Identifying the Interrogative Enclitics

The usual conception about Navajo question formation is well ex-

pressed by Young and Morgan (1971,16-17):

Sha', -sh, - sh...are the interrogative enclitics. A question is

not indicated in Navajo by raising the tone of a word or phrase as
we do in English, but is indicated by the suffixation of one of the
interrogative enclitics to the first word in the clause, or by use
of the particle da', whose function it is to introduce a question,
or by use of both the enclitic and the particle. Thus:

1. haisha' orI- h5ish anit'iL

Who are you?

2. ha''t' isha' or ha'at' sh ninizin

What do you want?
 109

3. haasha' or haash yinflghe

How are you called? What is your name?

4. bich'ahash dif ch'ah bila'a'h an itso

Is his hat larger than this hat? (-a'sh = - sh)

5. bee nin 4''ish

Did you steal it from him?

6. da' dichinish nil

Are you hungry?

Since Young and Morgan list the three forms separately, they

obviously consider them distinct forms. In their explanation they do not

differentiate their uses, but it is clear from the examples that they

consider /-sh/ and /-shg'/ WH question markers, and /-ish/ a yes/no

question marker.

2.1.2. /-sh/ As a Distinct Enclitic

It is not universally agreed, however, that there are three distinct

enclitics. Elgin (1973, IV.2.2), for example, says that the interrogative
1
suffix /- sh/ appears on both yes/no questions and WH questions. Phono-

logical evidence, however, argues for two distinct particles. The crucial

examples are those in which the enclitic attaches to a word ending in a

low tone vowel. Since vowels assimilate in Navajo (cf. sentence (5)), the

vowel quality is often irrelevant. However, when a high tone vowel cliti-

cizes to a low tone vowel, a rising tone results. Consider example (3).

Since the tone has remained low, we know that /-sh/ and not /-ish/ was
 110

used. The ungrammaticality of (7) proves that /- sh/ is impossible:

7. *haash yinflghe

In contrast, the yes/no questions in (4)-(6) all have the /-Ish/

suffix. This is apparent because the words the enclitics are attached to

do not end in vowels. Once again, substituting /-sh/ for /-Ish/ is im-

possible:

8. *bee nin ' 'sh

*bee nindt'4'ish

9a. ashkilsh 14/ ' nabiilgo'

b. *ashkiish 11'' nabiifgo'

Lest it be assumed that /-sh/ can only attach to vowels, and that

the previous examples of yes/no questions were ungrammatical because they

ended in consonants, (9) shows that /-sh/ also cannot attach to vowels in

yes/no questions.

Haile (1942,11.112) also differentiates /-Ish/ and /-sh/ because of

phonological evidence. He notes that "we cannot say */ha'adi/sh/ or

*/hia'j sh/ but must say /haadish/ or /haadisha /, 'whereat'; /hij'sh/,

2
/haajishV'/, 'which way'." In other words, there is no change in the

final vowel of the WH word when the interrogative enclitic is attached.

Since an enclitic beginning with a high tone vowel would affect the vowel,

the interrogative enclitic in these cases cannot begin with such a vowel.

Haile concludes from this evidence that "the reduced form /-sh/

goes back to the low-toned /-sha'/, rather than to high-toned /-Ish/."

 111

However, he gives no arguments for relating /-sh/ and /-sha' .

2.1.3. /-shz'/ As a Topic-Question Enclitic

Since the differences between the three enclitics will be important

as we proceed with more complex sentences, I will suggest briefly, without

any real argument, why I disagree with Haile's position that /-sh/ is a

reduced form of /-shl'/. Unfortunately, nobody has yet investigated this

area in sufficient detail.

To begin with, some speakers of Navajo use /-sha.'/ for yes/no ques-

tions as well as WH questions, as in (10).

10. ashkiisha' 1 ' nab'Ifgo'

boy Q horse 3.3. P. throw
Was the boy thrown by the horse?

If, as I argue, /-sh/ and /- sh/ are distinct, then how can they both be

reduced forms of /-sha'/? On the other hand, if /-sh/ and /- sh/ are not

distinct, then how come some people use /-sha'/ only for WH questions?

The behavior of /-sht'/ is sufficiently different from /-sh/ and

one.
/- sh/ that I see no reasoned way of collapsing two enclitics into

As Elgin notes (112), only /-sht'/ typically occurs on single words in-

stead of whole sentences. Her two examples are (her (23)-(24)):

11. X: Ja'an naalnish

John is working.

Y: Merisha? 3
L
How about Mary?
 112

12. Jaansha?
Where is John?

According to Elgin, as part of sequence (11) /-sha/ (sic) means approxi-

mately "and does the preceding statement apply to Mary as well?" (ibid.).

On the basis of this, she suggests that /-sha/ (sic) may be a compound,

consisting of /-sh/ and an affix /-a/ (sic) meaning "aforementioned".

Although I see no motivation for her morphological analysis, 4 I

would suggest that /-shg'/ appears to break down into the above components

because it is a topic marker as well as an interrogative enclitic. As I

will discuss in Chapter 3, /-ash/ may also perform another function in

addition to indicating a question. Of these three enclitics, however,

only /-sht'/ is inherently a topic marker. 5 Thus, we find examples like

the following:

13. Jaansha', ha'dg6dsh dooga4 ;l

John.Q where.to.Q 3.F.go
What about John, where is he going?

Note the pause between /Jdanshg'/ and the rest of the sentence. Note also

that the sentence is a question even without the /Jaansha'/. In other

words, while /-sha'/ serves to indicate both topic and question, either

may be its primary function in a particular sentence. In other words,

both aspects should be considered basic to /-sha'/. I will assume, there-

fore, that in at least some sentences, /-sha'/ is generated as a topic

marker, a process different from that of generating other question particles.

 113

The fact that /-sha'/ can be used to indicate that a question is

being asked about the topic of the sentence, as opposed to indicating a

question about the whole sentence, can be observed more clearly by com-

paring (14) to (15) and (16):

14. Maryshg' Jdan bich'i' yad.ootih

Mary.Q John 3.to 3.F.speak
(i) Is John going to speak to Mary?
(ii) Where is Mary? John is going to speak to her.

15. Mary sh Jaan bich'i' yadooltih

Mary.Q John 3.to 3.F.speak
Is John going to speak to Mary?

16. Marysha', Ja/anish bich'i' yadooltih

Mary.Q John.Q 3.to 3.F.speak
How about Mary, is John going to speak to her?

For people who use /-shg'/ as a yes/no question particle, (14) can be used

instead of (15), i.e., (i) is a correct translation for (14). However,

some people normally consider /-sht'/ ungrammatical in yes/no questions,

while accepting /-sh '/ in WH questions. Linda Platero, one such speaker,

accepted (14) but did not accept the yes/no question reading ((14i)). She

pointed out that (14) was not identical to (15). Requiring a pause after

/Maryshq'/ in (14), she gave (14ii) as the correct paraphrase. In order

to question whether John is going to speak to Mary, it is necessary for

her to use /- sh/ in addition to /-shat/ ((16)). Generating /-sha'/ as a

topic marker will account for this meaning of (14). It will also account

for the fact that there may be two interrogative particles in one sentence.
 114

(17) and (18) demonstrate that if there are two such particles, the first

one can only be /-sha'/.

17. *Mary sh, Ja'anish bich'i' yadooltih

Mary.Q John.Q 3.to 3.F.speak

18. *Marysh, Ja'anish bich'i' yadooltih

Mary.Q John.Q 3.to 3.F.speak

Assuming that /-sha'/ is generated with topics leaves open the

question of whether it is also generated as a yes/no particle for some

people and with WH words for everyone. That is, we must account for the

reading of (14) which is identical to (15) and for the WH questions in

which there is no pause after the first word and /-sha'/ is the only

interrogative particle (cf. (1), (10), etc.).

One possibility is that there are three possible categories for

/-sha'/, as topic-question marker, WH question marker, and yes/no question

marker. People simply vary as to which categories they have put /-sha.'/

into. Given this option, /-sha'/ will be generated the same way the other

question particles are, in addition to being generated with topics.

Another possibility is that /-shg'/ is only generated with topics,

but Navajo has two types of topics, just like English.

19. As for John, I like him.

20. John I like.

The left dislocation in (19) is analogous to the /-sha'/ cases with a

pause. Since there is a pause, it is easy to keep the question aspect

 115

of /-shgf/ isolated to the topic.

If the cases in which /-sha'/ is the only interrogative particle

are analogous to (20), then all words marked by /-sha'/ must be topics.

Such a claim is consistent with the fact that it is impossible to get two

/-sh '/'s in one sentence. While (13) and (16) are fine, (21) and (22)

are not:

21. *Jansha', h dgdo'sha' doogaal-

John.Q where.to.Q 3.F.go

22. *Maryshal Jaanshp' bich'l' yadoo;Ltih

Mary.Q John.Q 3.to 3.F.speak

It is also consistent with the fact that even for those who use /-sha'/ as

a yes/no particle, /-sha'/ cannot be substituted for /- sh/ when the latter
6
is not in second position.

23a. Mary Ja'anish bich'i' yadoo;ltih

Mary John.Q 3.to 3.F.speak
Is it John who is going to speak to Mary?

23b. *Mary Jaansht' bich'i1 yadooltih

Mary John.Q 3.to 3.F.speak

The use of word order to designate topics seems to be a very common pro-

cess in Navajo. If we assume that topics must always occur in initial

position and that /-sha'/ is always a topic marker, then this difference

between /-ish/ and /-sha'/ follows automatically.

 116

However, assuming that /-sha'/ is always a topic marker also causes

problems. If /-sht'/ is generated with the topic, then a second interroga-

tive particle will be generated with the WH word. It should therefore be

possible to get the two interrogative particles in one sentence without

a pause. Such a subtle difference is difficult to detect, especially

since /-shq'/ ends in a glottal stop which creates a pause of its own.

However, so far as I can tell, the pause is necessary. If my data are

correct, then a deletion rule for the extra question particle is required.

An even bigger problem is the fact that /-sht'/ occurs attached to

WH words. It has always been assumed that question words may not be

topics, but it is also unclear exactly what the import of question word

behavior is. I therefore leave unanswered the question of how many ways

-sht/ is generated.

In sum, I conclude that /-sha'/ is different from the other question

particles both because it is a topic marker as well as a question marker

and because its behavior is not precisely identical to that of any of the

other particles. On the basis of the above facts, I will thus assume that

the three enclitics are all independent, and will refer to /-shg'/ as the

Topic-Question enclitic, /- sh/ as the Normal yes/no Question Enclitic,

and /-sh/ as the Normal WH-Question Enclitic.

2.1.4. The Interrogative Particle /la/

Young and Morgan do not include the particle /la/ in their list of

interrogative particles, but it will be prevalent in the examples I use.

 117

Consider the following:

24a. hila' l11'

L %_
nabiilgo'
who.Q horse 3.3.P.throw
Who got thrown off the horse?

b. haish Iii' nabifIgo'

who.Q horse 3.3.P.throw
Who got thrown off the horse?

25a. ha'at'fifll ashkii yiyiiltsa

what Q boy 3.3.P.see
What did the boy see?

b. ha'lt' ish ashkii yiyiilts L

what Q boy 3.3.P.see
What did the boy see?

26a. haago'la Jaan deeya

where.to.Q John 3.go
Where is John going?

b. haagcdsh Jaan deey'

where.to.Q John 3.go
Where is John going?

27. *ashkiil lif' nabiilgo'

boy Q horse 3.3.P.throw

It thus seems that /l/ and /-sh/ are interchangeable. Both occur with

WH questions but not yes/no questions. However, more work needs to be

done to find out whether there is any substantive difference in the cur-

rent use of /-sh/ and /la/ as questions markers. 9

 118

One difference between /1a/ and the other interrogative particles

seems to be that only /1/ is not an enclitic. If it were an enclitic,

/la/ should cause a lengthening of a high tone word final vowel. None

of the WH words end in a short high tone vowel. However, /la/ does not

cause a lengthening of word final vowels: /ashiike/ (boys) becomes

/ashiikel /, not */ashiike la/. For our purposes, however, it makes no

difference whether or not /ia/ is part of the class of particles which

are enclitics. Because it does not seem that /la/ is an enclitic, I will

refer to /la/ as a Normal WH-Question Particle, since particles may or

may not be enclitics.

2.1.5. Word Order

From my preliminary work, it appears that, if there is a signifi-

cant difference, that difference will be found in types of sentences not

included at all in Young and Morgan's description of interrogative

particles. According to them, the particle must be placed on the first

word of the clause, i.e., in what has been referred to as second position.

While this is the most common position, it is not the only possible one.

Compare (28) with (25).10

28a. Jaan haata''il yiyiiltsa

John what Q 3.3.P.see
What did John see?

28b. ashkii ha'a't ' sh yiyii-tsd

boy what Q 3.3.P.see
What did the boy see?
 119

In addition to showing that the question particle need not appear

in second position, (28) also demonstrates that WH words need not be moved

to the front of the sentences (at least in these cases) as, for example,

they must in English. That these are two separate facts is clear from

(29)-(30) where we have a WH question and yet the question particle is

attached, not to the WH word, but to the first word:

29. ashiik4la' ha 'a'tI i yiyiilts

boys Q what 3.3.P.see
What did the boys see?

30. ashkiish ha 't if yiyiiltsa

boy Q what 3.3.P.see
What did the boy see?

To round out the logical possibilities (i.e., having discussed

(25), (28), (29)) note that the following is impossible:

31a. *ha'at'111 ashkiilal yiyiilItsa"

what boy Q 3.3.P.see

31b. *ha'lt'ii ashkiish yiyiilts6

what boy Q 3.3.P.see

In other words, a WH question particle may end up either in second posi-

tion or with the WH word, no matter where it is placed in the sentence.

 120

2.1.6. Where Question Particles Occur Underlyingly

Although I do not argue in depth for any hypothesis about where the

question particle occurs in underlying structure, I would like to specu-

late a bit about some aspects of possible explanations because they involve

some conceptions about other principles in Navajo. With Young and Morgan,

I assume that Navajo has a rule moving clitics into second position.

This Second Position rule is probably necessary for other clitics besides

the question clitics (cf. Perkins, 1973). However, it was previously be-

lieved that the Second Position rule was obligatory. Although it is not

clear to me exactly how second position should be defined, the sentences

in (25) as well as the work done by Perkins demonstrate that clitics need

not end up in second position. I therefore assume that this rule is op-

tional rather than obligatory.

Under the assumption that the Second Position rule is optional,

the simplest explanation for the above data would involve postulating

that the question particle occurs underlyingly as part of the WH word

rather than as a separate entity in a fixed position (either an initial

Q or in COMP, i.e., final, position). Generating the question particles

that occur in WH questions with the WH words accounts for the fact that

/la/ and /-sh/ cannot occur in a sentence unless there is a WH word

associated with it somewhere in that sentence. Also, if the WH question

particle is not generated with the WH word, then an additional rule is

required which optionally attracts the particle to the WH word. Generating

the two together, the Second Position rule which is needed anyway derives
 121

the other possible output. Throughout the following discussion, I will

therefore generate the question particle with the WH word, but will not

try to justify any particular constituent structure. However, in section

2.5.7, I will reconsider the possibility that the question particle is

generated separately as an initial Q, since such an assumption makes it

easier to account for certain factors of question formation in complex

sentences.

We know from the above data that, in addition to the optional

Second Position rule for particles, there is some rule which optionally

moves WH words into initial position. For expository purposes, I will

temporarily refer to it as the WH-movement rule. Notice that, since both

rules are optional, it is not necessary to apply either rule. Sentence

(28) is the result of not applying either rule. (31) is generated by

applying the Second Position rule but not WH-movement. If (31) were

acceptable, it could be generated by applying the WH-movement rule but

not the Second Position rule. Notice, however, that (31) cannot be

blocked merely by ordering these two rules since, as described, they do

not apply to the same constituent. However, the assumption that the

question marker is part of the same constituent as the WH word does allow

us to block (31). If the WH-movement rule is written to apply to the

entire constituent, then when the WH word is moved, the Q particle will

be dragged along. In other words, (31) could be generated by inter-

changing the subject and object, but leaving the Q marker behind. The

WH-movement rule must therefore be written to ensure that the Q marker

 122

cannot be stranded. Notice, however, that it is also necessary to ensure

that WH-movement applies before the Second Position rule. Consider the

following logically possible derivations:

A. ashkii ha'at'iilayiyiil-tsa/
Second Position: ashkiila ha'at'll yiyiilts
L

WH-Movement: *ha'I at ' ashkiila yiyiilts

B. ashkii ha 'at ' ila yiyiilt sd

WH-Movement: ha't ' ila ashkii yiyiiltsa'L
L

Second Position: --------

It is perfectly reasonable to assume that any rule which is designed

to account for surface linear order will not apply until all rearranging

of constituents has occurred. One should expect (B) but not (A).

It should also be noted that sentences like (25) do not involve

the Second Position rule even though the Q marker ends up in second

position. Since WHn-movement is applied before the Second Position rule,

and since WH-movement is written so that the Q marker will always be

dragged along, whenever a WH word is in initial position, the Q marker

will always be in second position before the Second Position rule has had

a chance to apply. In other words, for simple WH questions, the Second

Position rule is operative only for sentences like (29).

2.1.7. How To Account For WH Placement

Let me now mention briefly the possible rules that could account

for initial WH placement. It is clear that this is not done by Subject

 123

Object Inversion (cf. Hale, 1973; Creamer, 1974). Thus, when Subject

Object Inversion interchanges the subject and the object, it also changes

the direct object prefix on the verb from /yi-/ to /bi-/. However, in

(25) the object occurs before the subject and yet the /yi-/ has not changed

to /bi-/. In addition, it is generally agreed that word order is freer

when the subject or object is a WH word. Thus, the order of the subject

and object is generally constrained by a semantic hierarchy (cf. Creamer,

1974). That hierarchy, among other things, requires that. when there are

two NP's the animate must precede the inanimate. Thus

32a. Ja/an t se yiyiiltsa"

L

John rock 3.3.P.see

John saw the rock.

b. *ts 4 Ja'an biilts

rock John 3.3.P.see
The rock was seen by John.

33a. *ts4 Ja'an yishjizh

rock John 3.3.P.crush
The rock crushed John.

b. Jdan tse bishjlzh

John rock 3.3.P.crush
John was crushed by the rock.

However, in either case, /ha'lt' sh/ what may precede /Jaan/.

 124

34. ha'a't' ish Ja'an yiyiilkts4

what Q John 3.3.P.see
What did John see?

35. ha'lat' sh J'an yishjizh

what Q John 3.3.P.crush
What crushed John?

These facts suggest the possibility that there is a special question

formation rule.

2.1.8. A Comparison to Baker's Proposals

While such rules are not expected in SOV languages, the formulation

required for these sentences would not be a violation of Baker's universal

constraints on question formation. First, the rule could be the same as

the one Baker formulates for languages in general. The fact that there is

a variable in it is irrelevant for these simplex sentences since the

variable would be null.12 Second, Baker does not say that question move-

ment is impossible in verb final languages, only that it cannot be obliga-

tory. As is apparent from (17)-(18), the Navajo rule would have to be an

optional one. Third, Baker's statements about verb final languages are

limited to those which place particles "positioned with reference to the

sentence as a whole...at the end of the sentence" (207). As we have just

seen, Navajo interrogative particles do not normally go to the end of the

sentence.
 125

2.1.9. Relationship Between Yes/No Questions and WH-Movement

Baker also notes that it is "only languages which position their

particles for yes-no questions in clause-initial position [that] permit

a rule for questioned constituents" (207). Except for differentiating

/-sh/ and /-Ish/, I have so far said nothing about Young and Morgan's

description of Navajo yes/no questions. Recall that, in addition to the

three enclitics, they mention the particle /da'/ which introduces a

question. This particle does in fact occur in clause-initial position.

As can be seen from their one example with /da'/ ((6)), it introduces

yes/no questions. Young and Morgan do not mention it, but /da'/ cannot

occur with WH questions.

- 3 6 a. *da' Ja'an ha'at'ish yiyiiltsa/

Q John what Q 3.3.P.see
What did John see?

b. *da' ha'at' ish Jaan yiyiiltsa

1.-

Q what Q John 3.3.P.see

What did John see?

c. *da' Jaan l' ha'at' yiyiiltsa

Q John Q what 3.3.P.see
What did John see?

There is, however, one problem with trying to claim that all Navajo yes/no

questions are introduced by a clause-initial particle. /da'/ does not

always occur in a yes/no question. However, since /da'/ and /-Ish/ do

occur in the same sentence, one could postulate that they both appear
 126

underlyingly and that there is an optional rule (or rules) which can

delete either one (but not both) of the segments.

In a very interesting class conducted by Paul Platero at Rough

Rock Demonstration School, 1 3 the Navajo students listed the following

possibilities and proposed this solution to account for the evidence:

37a. *da'ish hastiin nihaanya

Q Q man lpl.3.P.come

b. dat hastiin nihaan/yK

Q man lpl.3.P.see
Did the man come to (see) us?

c. * sh hastiin nihaanya
Q man 1pl.3.P.come

38a. dal hastiinish nihaanya

Q man lpl-3-Pcome

b. da' hastiin nihaanay

Q man 1pl.3.P.come

c. hastiinish nihaanya
man Q 1pl.3.P.come

/ //' 14
39a. da' hastiin nihaanayaash
Q man 1pl.3.P.come.Q

b. da' hastiin nihaaniya

Q man 1pl.3.P.come

c. hastiin nihaanyaash
man pl.3.P.come
1ma
 127

Although there are some differences in meaning among (37)-(39),15 the

sentences within each of the numbered sets were felt to be identical in

meaning. Semantically, there is therefore no reason to postulate dif-

ferent underlying forms. On the contrary, one could account for the

identity by having a single underlying source. In addition, note that

-Ish/ cannot attach to /da'/. If /da'/ is a pro-clitic, there is nothing

more to be said. However, if /da'/ is a separate word, one could account

for this fact by having a question formation rule which either moves /-ish/
or deletes one of the elements. Since I have no evidence to argue whether

or not /da'/ is a separate word, I will make no conclusions about this

hypothesis. I mention it because it would make the Navajo facts more

consistent with Baker's observations.

2.1.10. More Problems With /-shat/

Perhaps it should also be noted that /da'/ cannot occur with /-shat/
when the latter is being used as a yes/no question particle. On the basis

of the earlier discussion about /-sha'/, this fact is not a true problem

for Baker's hypothesis. First, /-sha'/ is not a yes/no question particle

for all speakers. Second, even when designating yes/no questions, it is

probably still a topic marker. It thus seems that the normal yes/no

question is formed with /-ish/ and not /-sha'/. One could therefore claim

that Baker's claim about the correlation between initial particles and

question movement holds for Navajo, since the normal yes/no question is

derived from /da'.. ish/.

 128

In any case, this particular movement of questioned constituents

cannot at this juncture be a counterexample to Baker's claim if we assume

he is only talking about movement rules that specifically apply to ques-

tioned words. This is because, while discussing a possible question

formation rule, I have not argued that the movement must be the result

of a question formation rule. Considering some explanatory remarks made

when I asked people to contrast (25) and (28) (i.e., WH objects in first

or second position), one must wonder whether the movement is in fact a

result of a question formation rule. In general, my consultants feel

that the first word is "what you're talking about," or "what you're most

concerned about." Thus, if the discussion has been about John or the boy,

and the interest is on his movements, his name rather than the question

word is likely to appear in first position (i.e., (28) or (29)). However,

if the concern is with a particular incident, say an accident, and you

want more information about it, in asking what John or the boy saw, it

would be more natural to put the WH word first (i.e., (25)).

This description suggests that using the first position for topics

is a major factor in deciding word order for questions with /-sh/ and /l'/

as well as /-sha'/. Although I eliminated Subject Object Inversion as

the WH-movement rule, the fact that Navajo is a language which permits a

certain amount of scrambling (except that the verb is usually final) and

the fact that topics are frequently reflected in the Navajo word order

suggest that the possibility of a topicalization rule should not be

eliminated lightly. The fact that /-shg'/ can be used instead of the
 129

other question particles when they are in second position may also suggest

a close connection between topicalization and WH-movement.

One major problem with using a topicalization rule to explain the

appearance of WH words in initial position is that these question words

are considered indefinite pronouns. As Bach has pointed out, the fact

that they are indefinite means that question words will never occur as

themes (1971,115). Bach does not attempt to give a definition to the term

theme, using examples instead. It is not clear to me exactly what the

relationship is between themes and topics. One point that Bach makes is

that themes are always presupposed and one cannot presuppose the truth of

question words (162-3). One must, therefore, investigate the relationship

between presupposition, topics, and Navajo WH words. This is a subject I

will discuss in depth in Chapter 3.

At this juncture, it is impossible to present any conclusive evi-

dence about the formation of simple questions involving WH-movement,

universal question rules, the use of topics,and the principles behind these

possible transformations. Some understanding of simple questions is neces-

sary to investigate the more complex questions. But it is only after an

investigation of the complex questions that some claims can be made.

2.2. Complex Sentences

When WH words are embedded in complex sentences containing only

direct discourse verbs, all variations found in WH questions in simple

sentences are still possible. As we will see, unbounded leftward movement

 130

of the WH word into any higher S is also possible, which creates even more

possibilities. For some people, it is even possible to get the WH question

particle and WH word in separate S's, so long as the question particle is

in the higher S. The possibilities I will be discussing are summarized in

the following table, as defined for the given underlying tree.1 6

SO

Si

-WH-

LOGICAL POSSIBILITIES FOR DIRECT DISCOURSE QUESTION FORMATION

S2 1 o Status Point of Viewl7

WH
1. 1 OK subject S

2. l WH *

3. l *

4. WH la OK subject SO
WH
5. OK subject S

6. l WH *

WH
7. la OK speaker

8. WH 1 OK speaker

9. WH 1 *
 131

In the following sections I will discuss the above variations

separately, including the fact that the variations differ in meaning.

I will motivate an unbounded leftward movement transformation and link

its existence to the differences in meaning.

2.2.1. Unbounded WH-Movement to Sentence-Initial Position

Having discussed the simple sentences in which we find movement

of WH words into sentence-initial position, let us now turn to the cases

where a WH word has moved from an embedded S into sentence-initial posi-

tion. Consider the following:

40. hag6dla' Ja'an deeshaa;k na

where.to.Q John 1.F.go 3.say
Where does John say he's going?

41. ha'adila[ Jaan Mary dinilnish yibni

where.at.Q John Mary 2.P.work 3.3.say
Where did John tell Mary to work?

42. ha'at'ish Jaan nahideeshnih nizin

what Q John 1.F.buy 3.want
What does John want to buy?

43. haagold Jaan jidoogal;L sho''ni

where.to.Q John 4.F.go 1.3.say
Where does John expect me to go?

The first thing to establish is that the WH word must have originated

in an embedded S. This can be proved by noting that the WH word in each

 132

example is a possible complement to the embedded verb, but not to the

matrix verb.

44a. *hajgod/l Jaan ni

where.to.Q John 3.say
To where does John say?

b. haago'l Jaan doogaal

where.to.Q John 3.F.go
(To) where is John going?

45a. *haadild Jaan Mary yi;ln

where.at.Q John Mary 3.3.say
At where did John say to Mary?

b. haadila Mary deeshnish

where.at.Q Mary 3.work
Where will Mary work?

46 a. *ha'lt'ifsh Jaian nizin

what Q John 3.want
What does John want? 1 8

b. ha'at' ish Jaan neidiyoolnih

what Q John 3.3.F.buy
What will John buy?

47a. *hla'gddla Jdan sho''n

where.to.Q John 1.of.3.say
To where does John expect?

b. hago6l deesh/a/l
where.to.Q 1.F.go
Where will I go?
 133

Notice that in all the (b) sentences it was necessary to change the form

of the verb in order to retain the same reference as that of the embedded

verbs in (40)-(43). Since (40)-(43) are direct discourse sentences, the

embedded verbs are from the point of view of the subject of the matrix S

rather than from that of the speaker. Thus, in (40), although John is

doing the "going", the verb is in the first person rather than the third.

A direct discourse segment is interpreted from the point of view of the

subject of the immediately dominating S rather than from the point of view

of the speaker. Since the subjects of the matrix and embedded S are core-

ferent in (40), the first person is used. When the speaker is reporting

what the subject said to a third party, the second person is used instead

of the third (cf. (41)). The third person is used in a direct discourse

segment only when the subject of the higher S says or thinks something

about a third party. Notice that that leaves none of the usual person

markings for the subject to designate the speaker. As (43) demonstrates,

the Navajo fourth person is used to handle this situation.

The verb changes from (40)-(43) to (44)-(47) respectively are, of

course, irrelevant to this discussion about where the complements originate.

Adjustments for number agreement obviously do not affect the underlying

grammatical relations a verb may enter into. That is, the first person

form of a verb can take any complement the third person form takes. (48)

is as good as (44b).
 134

48. haago61a deeshaa"b

where.to.Q 1.F.go
Where shall I go?

I used (44b) instead of (48) merely to keep the meaning consistent.

The fact that the embedded verb in (40) is in the first person

proves another point: the noun is the subject of the matrix verb, not

the lower verb. The subjects of the lower verbs in these cases must have

been pronouns which deleted as pronouns generally do in Navajo. To show

that the glosses are correct, note the following:

49a. *Jian deeshail

John 1.F.go

b. Jaan doog I4
John 3.F.go
John is going.

c. deeshaA
1.F.go
I am going.

While the embedded form of the verb can stand alone as a separate sen-

tence, it cannot have /Jdan/ as its subject. 1 9 This proves that the WH

words have moved across the subject of the matrix verb as well as that of

the embedded verb. In other words, there is no question that the WH word

has ended up in the matrix S. Given my claim that direct discourse clauses

are embedded S's, we now have examples of unbounded leftward movement in

a verb final, COMP final language.

 135

To see that this movement is not only unbounded, in the sense of

moving out of one S, but can involve movement across an indefinite number

of S's, consider the following:20

50. haagoo'1 Jean Mary doogaa nisin na

where.to.Q John Mary 3.F.go 1.want 3.say
Where did John say he wants Mary to go?

Sentence (50) would evolve from a tree something like the following:

(50)

J san nI
John 3.say

nisin
1.want

Mary had goo doogaaL

Mary where to 3.F.go

The above tree is incomplete because it does not include the question

particle /la/, since where /la/ originates is irrelevant to the argument

that the WH word itself may move unboundedly leftward. However, the
 136

underlying position of /la/ will effect both the form of the WH-movement

transformation and any explanation for the existence of unbounded leftward

movement. I discuss these problems below. If direct discourse S's in-

clude a COMP node, that also has been omitted. The crucial points depicted

by the tree are that the EP originates in the lowest S, S2 , and is moved

by some transformation across several S's to sentence-initial position.

For these particular examples, it seems that Navajo has the same unbounded

leftward movement found in languages like English.

2.2.2. Contrast To Theoretical Claims

The problem is, of course, that such rules are not supposed to

exist in verb final languages. More specifically, given a theory such

as Bresnan's or Chomsky's, there is no way to account for this movement.

They say both that the only way a constituent can move out of an S is

through COMP position and that the question formation transformation moves

the WH word into the COMP node.

Notice that assuming that direct discourse S's do contain a COMP

node (as well as matrix S's) does not solve the problem because the COMP's

would be on the right, at the end of each S, rather than to the left. 21

In the most extreme case of trying to save the claim that all movement

out of an S must be through COMP position, one could postulate that each

S contains a COMP node and that the movement occurred rightward rather

than leftward. After reaching S, another rule would then move the WH

word to sentence-initial position. Such an hypothesis would still violate

 137

the following putative universals:

(i) that WH words never move rightward. Recall that one

possible explanation for the deletion of the WH word

in the indirect questions in Chapter 1 was to avoid

rightward movement of that word.

(ii) Bresnan's Complementizer Attraction Universal which says

that movement into COMP position is possible only if the

COMP is clause-initial. However, the facts in Chapter 1

probably show that this is inaccurate anyway.

While this Navajo unbounded leftward movement contradicts Bresnan's

refinement and explanation for Baker's observations, the phenomenon pro-

vides little problem for Baker himself. The transformation he suggests

does in fact work for these Navajo WH words. Such a transformation is

not prohibited because, although Navajo is an SOV language, it does not

generally position particles at the ends of S's. And Navajo does have at

least one yes/no question particle which is clause-initial, a prerequisite

for having a WH-movement rule.

The one factor that might prove a problem for Baker is his claim

that SOV languages will never have an obligatory movement rule. However,

it depends in what sense he means obligatory. If the unbounded WH-movement

rule is thought of as a general question formation rule, then the Navajo

rule is not invariant in the sense that it is not used to form all WH

questions. As we have already seen, indirect questions are not formed by

 138

leftward movement, but by deletion of the WH word, possibly before the

rightward movement rule applies. In addition, as we will see in the next

chapter, there is a type of direct question which involves no movement at

all (i.e., WH words embedded in /-go/ subordinate clauses).

Even closer to home, this unbounded movement rule is not obligatory

in that, if not applied, the string will still be well-formed. For some

people, however, the meaning must be different when unbounded movement is

not applied.22 If one includes the meaning difference in the tree before

movement takes place, the rule may still be obligatory. I will be dis-

cussing these facts below.

2.2.3. Dialect Variation

There are also people for whom the question movement is optional

and the meaning remains the same, that of a direct question from the

speaker's point of view. That is, while examples (40)-(43) seem to be

acceptable to everyone, preferred by some, and the only possible forms

for some, there are others who allow a question reading even when there

is no movement. Although some unbounded leftward movement rule seems to

exist in all dialects, it is difficult to characterize when it can apply

and when it must apply. I will now turn to cases in which there is no

unbounded WH-movement, describing what different sentences mean to dif-

ferent people. After that, I will look at cases of movement where people

disagree about grammaticality and/or meaning, and will later speculate

about the development of these various dialects.

 139

Consider the following sentence and a tree representing it:

51. Ja'an Bil ha'a't'lila yizhbizh ni

John Bill what Q 3.3.P.braid 3.say

(51)

Ja'an
John 3.say

NPNP V

Bil ha'at'ii yizhbizh

Bill what 3.3.P.braid

We know that there are two NP's underlyingly in the S with the embedded

verb /yizhbizh/ (Sl) since /yi-/ is the direct object marker only for verbs

containing a third person subject followed by a third person object. Thus,

S by itself means, "what did Bill braid?". Recall, however, that S's

change their point of view when embedded under direct discourse verbs:

As we saw previously, the point of view is no longer that of the speaker,

but that of the subject of the matrix sentence. The problem in (51) thus

is who said "what did Bill braid", John, or the speaker?. If, for the

purposes of (51) John uttered those words, then when the speaker says

them he is not asking a question but merely reporting what John had said.
 140

Thus, following the definition of direct discourse given above, (51) may

be a declarative sentence, not a question. In this case, the correct

paraphrase for (51) would be (52).

52. John asked what Bill braided.

2.2.4. Strict Constructionists

Notice that to say that (52) is the proper translation of (51) is

to claim that direct discourse is defined over S's.23 However, recall

that in all previous examples, the only word that changed from indirect

to direct discourse was the verb. It is thus possible that the term

direct discourse applies only to verb interpretation, and not to full

S's, and therefore not to object NP's for example. However, I have

tacitly assumed all along that direct discourse applies to full S's by

using as an argument the fact that subjects must agree with their verbs.

Thus, I claimed that in sentences like (53), /Jaan/ could not be the sub-

ject of the lower verb /adessbas/ because the latter is in the first

person while /Jaan/ is a third person noun.

53. Jaan kinlanigo' adeesbas n zin

John Flagstaff.to 1.F.drive 3.want
John wants to drive to Flagstaff.

One aspect of Navajo grammar that is not in doubt is the fact that the

subject and verb agree in person and number. Note, however, that since

there must be some rule anyway to make the subject and verb consistent
 141

with one another, the fact that both take a direct discourse point of view

is not definitive proof that direct discourse is defined over S's. What-

ever rule makes them agree in person and number could also be sensitive to

point of view interpretation, so that only one would have to be interpreted

independently for such an interpretation.

To see that direct discourse must be defined on S's, not V's, con-

sider the gloss of the following:

54. Jaan shizh4'6e' ha'tt' ild yizhbizh ni

John 1.father what Q 3.3.P.braid 3.say
Johni asked what hisi father braided.

/shizhe'e/ is composed of the noun /-zhe"6'/, father, plus the first person

pronoun /shi/. And-yet, when asked whose father John is talking about in

(54), the answer is John's father, not the speaker's father. In other

words, the first person pronoun is being interpreted from the point of

view of the subject of the sentence, i.e., as part of the direct discourse,

and not from the point of view of the speaker. The interpretation of the

possessive pronoun therefore demonstrates that the direct discourse point

of view cannot be defined for verbs only. There is nothing in the verb

which could indicate whose father is being referred to.

In parallel fashion, if the word /ha'at'fi l/ is to be interpreted

as part of the direct discourse point of view interpretation, then it must

be the subject of the sentence (i.e., the subject of the immediately

dominating S) and not the speaker who is asking the question. In fact,

for some people, (51) cannot be a question. Thus, for these people, it
 142

would be completely incorrect to answer the question. In other words, the

following is not a coherent dialogue:

55. A: Jaan shizh4'4 ha'a't'ila yizhbizh ni

John 1.father what Q 3.3.P.braid 3.say

B: #t-'6/' yizhbizh ni
rope 3.3.P.braid 3.say
He said he braided a rope.

I will refer to those people who do not allow (51) to be interpreted as a

direct question from the speaker's point of view, to those adhering to a

strict direct discourse interpretation, as strict constructionists.

2.2.5. Loose Constructionists

There are, however, some people who will accept the discourse in

(55) as appropriate, people who will allow (51) to be a question even

though there has been no movement. However, these people would render

(51) into English as (56 a), not (56b).

56a. What did John say my father braided?

b. *What did Johni say his1 father braided?

In other words, when the speaker is asked whose father is being talked

about, the response has changed from John's father to the speaker's father.

The point of view is now that of the speaker, not the subject of the sen-

tence. Although there is no overt complementizer (and none is possible),

there are people who will allow both a direct discourse and an indirect
 143

discourse interpretation with these verbs. The interpretation in which

(51) is a question rather than a report is thus perfectly consistent:

the entire S is being treated as an embedded indirect discourse clause

rather than a direct discourse one. Notice, however, that when a logical

possibility requires one NP in the embedded S to be interpreted from a

direct discourse point of view and another part from an indirect discourse

point of view, that possibility is considered ungrammatical. In other

words, the fact that (56b) is not a translation of (54) can be explained

by claiming that the entire S must be interpreted from the same point of

view.

Since the embedded verb in (51) is in the third person, it is

impossible to tell directly that the verb is also being interpreted as

an indirect discourse verb. To see that this is the case, consider the

following:

57. Jaan kinlnigoo adeesbas na

John Flagstaff.to 1.F.drive 3.say

58a. Johni says hei is driving to Flagstaff.

b. John says I am driving to Flagstaff.

For the strict constructionists, only (58a) is a proper translation of

(57). However, for those who accept (56a) as a possible rendering of (54),

(57) is ambiguous, allowing both (58a) and (58b) as possible translations.

In (57), the embedded verb /adeesbas/ is in the first person. It thus

 144

becomes a question of the point of view for the verb: (58a) is the direct

discourse point of view; (58b), like (56a) and (52), is the indirect dis-

course point of view. (56b) which requires mixing the two is not acceptable.

I will refer to those people who allow ambiguity between direct and in-

direct discourse as loose constructionists.

In summary, the term direct discourse refers to the point of view

of an entire S and not just to the verb or to the process of person-number

agreement. For strict constructionists, an S embedded under a small set

of verbs must be interpreted as a direct discourse S. For loose construc-

tionists, it is possible to use an indirect discourse interpretation.

Given the sentences described so far, the decision is made for the entire

S, not for the individual constituents, so that it is impossible to get

a mixed interpretation when material is embedded.

2.2.6. A Rationale for the Unbounded Movement

The fact that it is impossible for a strict constructionist to have

any indirect discourse references within an S embedded under direct dis-

course verbs provides a rationale for the WH-movement transformation. One

way to explain this is to consider the choices when faced with a desire to

utter (59) in Navajo.

59. Where does John want Bill to go?

Consider (59) in terms of the following tree which depicts the proper

grammatical relations and the required forms of the verbs.

 145

(59)

Jaan nizin
John 3.want

N E

Bil haa gol doogall

Bill where to Q 3.F.go

Under the assumption that there is no question movement rule, (59) could

remain as is. If (59) remains as is, however, the WH word and question

particle remain within S1 , the direct discourse segment of the sentence.

As such, it is the subject of the immediately dominating S (John), not the

speaker, who is asking the question.

59a. Ja'an Bil haagoola doog6ll nizin

John Bill where.to.Q 3.F.go 3.want
John is wondering where Bill will go.
*Where does John want Bill to go?

Faced with this wrong interpretation, a strict constructionist could

i. decide there is no way to say this sentence directly

ii. become a loose constructionist

24
iii. try another alternative.
 146

The first alternative is in fact a very attractive one. That is, these

sentences which are very common in English are in fact unusual in Navajo.

However, assuming a strict constructionist rejects the first two alterna-

tives, another alternative would be to delete the WH word, a process we

found in Chapter 1. This alternative, however, does not provide an

immediate solution. Recall that the WH words in Chapter 1 were all in

indirect questions and that once the WH word deleted, the spatial en-

clitic moved to the nearest /-:i/. Using the same rule for the questions

under discussion here thus creates two problems: first, indirect questions

do not contain question particles. WH deletion would not automatically

delete the direct question particle. If the particle were not deleted,

the question would still be from the point of view of the subject, i.e.,

John. If the particle were deleted, the sentence would not automatically

denote a direct question. Second, some new solution would have to be

found for the spatial enclitic because there is no /-:/. (Deletion of the

enclitic will not suffice because then the specific WH word would be lost,

creating an unacceptable ambiguity.)

Another alternative, of course, is to move the WH word so that it

is no longer within an S requiring a direct discourse interpretation.

This, of course, is precisely what we find. Thus, using Baker's uni-

versal WH-movement rule, the questioned word in (59) is moved to the

sentence-initial position:
 147

(59')

E EP

NE

I aa gooal Jaan NP V n zin

wV
here to Q John 3. want

Bil doogla/
Bill 3.F.go

While I do not intend to make any claim about the type of adjunction used,

it is clear that the EP can no longer be a part of S1 . In (59'), it is

only S1 which receives a direct discourse interpretation, so that the EP

will not receive that interpretation. This explanation assumes that the

point of view of So is determined after the movement rule. In other words,

the point of view designation cannot be pre-cyclic; it could be post-

cyclic. If cyclic, either t he determination is made some time on the

same cycle as the movement and is ordered after the movement, or the de-

termination is made at any time, but the reference of the moved constituent

is reinterpreted when the reference of So is determined.

It should now be apparent why this movement rule must be unbounded:

the point of the rule is not so much to get the questioned word to the

front of the sentence, but to get it out of the scope of the direct
discourse point of view. In (59), moving /ha 4gd&l/ to the front of S1

would be insufficient because it would still be interpreted as part of

that S, and therefore as a report of a Question rather than a direct

question.

Now consider (60).

6o. h4a'go'6la' Jaan Mary doogal'b nisin ni

where.to.Q John Mary 3.F.go 1.want 3.say
Where does John say he wants Mary to go?

(60)

Mary ha' goo doogal

Mary where to 3.F.go

In (60), not only is movement to the front of S2 insufficient, movement

across one S boundary is also insufficient. To be removed from the scope

of the direct discourse, the EP in S2 must be moved to S . In other words,

since there can be an indefinite number of S's within the scope of the
 149

direct discourse, movement must be allowed across an indefinite number

of S's in order to fulfill its purpose of removing a constituent from

within any direct discourse scope.

2.3. The More Problematic Cases

2.3.1. The WH Word Is Non-Initial

The above exposition describes the facts as if they were completely

known and completely consistent. Unfortunately, however, the data are not

that straightforward. Thus, some loose constructionists who accept (56a)

but not (56b) as a possible meaning for (51) do accept the following with

the translation as given:

61. Ja'an ha'IatI'sh nahideeshnih ni

John what Q 3.1.F.buy 3.say
What did John say he'll buy?

This is a case in which the WH word is not in initial position, the verb

is given a direct discourse interpretation, and yet the sentence can be

considered a direct question from the speaker's point of view. In other

words, the WH word is given an indirect discourse designation but the verb

is not. (61) thus seems to be a counterexample to the claim that the em-

bedded S is interpreted as a whole with respect to point of view.

Let me first demonstrate that (61) can in fact be a direct question

from the speaker's point of view. This is clear when put into a dialogue.

Thus, when I said (61), the response was (62).

 150

62. chid nahideeshnih n

car 3.1.F.buy 3.say
He says he'll buy a car.

Notice, however, that given the linear order in (61), the phrase

structure bracketing is not obvious. (61) would be bracketed as either

(63) or (64).

63. Ja'an S, [ha'at' ish nahideeshnihi 1 i

64. Jaan ha'at'iish 1 nahideeshnih] S1 n

If (63) is the correct bracketing, it is a counterexample as suggested

above. If (64) is the correct bracketing, it is not a counterexample,

but then I must explain why the WH word is not in first position where

the hypothesized transformation placed it. I want to claim that (63) is

not the correct bracketing when the sentence is interpreted as a direct

question from the speaker's point of view. I believe that (64) is the

correct bracketing and that the sentence is the result of having a topi-

calization rule apply to the matrix S after WH-movement has applied.2 5

Positive evidence that (64) is correct would be a sentence of the

form of (65), interpreted as a direct question with the embedded verb

having a direct discourse point of view.

65. Jaan 8 ,[NP ha'a't' sh V]S1 fl

Consider the following:

 151

66. Ja'an ashiike ha'at' ish ba nahidii;nih shian

John boys what Q 3.for 3.2.F.buy 1.3.say
John asked me what I will buy for the boys.
*What did John tell me I should buy for the boys?

(66) is exactly the form of (65). However, it is impossible for (66) to

be interpreted as a direct question. In other words, in any example in

which both the WH word and the question particle must be bracketed as

part of the lower S, the sentence will not be a direct question. Thus,
when a sentence is a direct question, we should expect that the WH word

and question particle will not be in the lower S. Therefore, although

the word order in (61) is insufficient to determine the bracketing, the

fact that it is a direct question from the speaker's point of view argues

for the bracketing in (65).

It should be noted that (61) is in fact ambiguous, having also the

report of a question reading, a direct question from the subject's point

of view. That is, (61) can also be translated as "John asked what he

should buy." This reading corresponds to (63). When used this way in

a dialogue, (67) is an appropriate reply:

67. ak a'a'n la' nah aniih bidin

flour some 3.2.buy 3.2.say
Tell him to buy some flour.

Note that (67) is not an appropriate answer to the direct question reading

of (61): it is incongruous to follow the question "What did John say..."

 152

with "You tell him...". Although (67) is an appropriate continuation of

a conversation, it is not a response to a direct question, but a reply to

a comment. The fact that one responds to a sentence does not mean that

that sentence was understood as a question. This distinction is not so

obvious with a response like (67) which is in fact a reaction to John's

request for information. However, (67) should be considered a gratuitous

response rather than a reply to a direct request for information.

I have been referring to all the questions discussed in this

chapter as direct questions, as contrasted with the indirect questions

in Chapter 1. That is, when (61) is interpreted as a declarative sentence,

it is not an indirect question, but a report of a direct question asked by

someone else, a direct question from the point of view of the subject of

the S immediately dominating the question. It is necessary to distinguish

the two for two reasons. First, the direct discourse point of view is

inconsistent with an indirect question interpretation. Second, all the

other cases of question particles are instances of direct questions.

Since the indirect questions in Chapter 1 do not contain question parti-

cles, if these were also indirect questions, it would be necessary to

explain why some indirect questions contain question particles and others

do not. If these direct discourse question-reports are direct questions,

we can maintain the claim that question particles always signal direct

questions.

In summary, all cases of a WH word plus question particle are

instances of direct questions. What differs is who is asking the question.

 153

When the WH word plus question particle remain in an embedded direct

discourse S, the question is asked by the subject of the S immediately

dominating the WH word and question particle, instead of the speaker.

It is only reported by the latter.

2.3.2. Chorus

So far we have looked at three kinds of cases involving direct

discourse, as summarized below.

1. The WH word remains within its S and the Q particle is

attached to it.

2. The WH word plus Q particle occur sentence initially, having

moved leftward across an indefinite number of S's.

3. The WH word plus Q particle occur as the second word of the

sentence. I claim that these examples result from having a topicaliza-

tion rule apply after question formation.

In all these cases the interrogative particle has remained with the WH

word. Recall, however, that in simple questions the interrogative par-

ticle could be attached to a word other than the WH word, without changing

what was being questioned. We therefore must consider such a possibility

in the case of direct discourse sentences. Parallel to (1)-(3), I will

consider the following cases:

4. The WH word remains within its S but the question particle

does not.
 154

5. The WH word occurs sentence-initially, but the question

particle is not attached to it.

6. The WH word occurs as the second word of the sentence, but

the question particle is not attached to it.

2.3.3. Second Position and Case (5)

Once the interrogative particle is not attached to the WH word,

it is logically possible for it to attach to any word in the sentence.

In the case of simple questions, recall that if there is a WH word in

the sentence, the only acceptable positions for the question particle

were with that word or in second position. To see that the same is true

for sentences containing direct discourse S's, consider case (5) first.

68. *hahgo Jaan la Mary dooga(l n/zin

when.F John Q Mary 3.F.go 3.want

69. *ha'ati'if Jaansh nahideeshnih nizin

what John.Q 3.1.F.buy 3.want

70. *hada' Ja'an Mary la naagha' ni

when John Mary Q 3.come 3.say

To say that the question particle may occur in second position is not

quite sufficient. Second position is a term which refers to embedded

S's as well as matrix S's. In (71)-(72), for example, /la'/ is in second

position in the lower S.

 155

71. Jaan Mary la' ha'de naagha ni

John Mary Q from 3.come 3.say
John asked where Mary comes from.

72. Jaan yiskago l Bil haa'go6 deey n

John tomorrow Q Bill where.to 3.go 3.say
John asked where Bill is going tomorrow.

73. *Jaan yisk'ago Bil l " h /g&d deeya' ni

John tomorrow Bill Q where.to 3.go 3.say

The ungrammaticality of (73) confirms the fact that the question particle

placement is no more arbitrary in lower S's than it is in matrix S's. The

Second Position rule thus applies within any S.

Since the Second Position rule applies to embedded S's, it is in-

sufficient to say that question particles can occur in second position in

general. The structure of (70) is as follows:

ha'd a' Ja'an [Mary 1 naagha] ni

S, Si

In other words, /1a/ is in second position in the embedded S. The fact

is that the question particle cannot remain in the lower S when the WH

word is moved out. Recall that when discussing simple sentences, I sug-

gested that WH-movement moves the question particle with the WH word and

that the Second Position rule does not apply until all rearranging has

occurred. These two stipulations are sufficient to rule out (70). More

generally, these two stipulations rule out any grammatical sentences in

case (5). If the WH word occurs sentence-initially, the question particle

 156

must be in the matrix S. Second position in the matrix S, however, is

precisely the end of the WH word. Thus, if the WH word is sentence-

initial, the question particle must be attached to it.

In summary, when there is a WH question, the question particle can

occur only on the WH word or in second position, with the qualification

that the question particle cannot remain in an S lower than the one the

WH word ends up in. In proceeding to cases (4) and (6), I will there-

fore limit myself to second position particles. Since the qualification

about second position follows naturally from the rules, there is no need

to stipulate it.

2.3.4. Case (6)

Since the facts are even murkier for case (4) than they are for

case (6), let us begin with case (6). Consider the following:

74. Q: Jaan la' ha'ago' Mary deeya' ni

John Q where.to Mary 3.go 3.say

Rl: *auu'
yes

R2 : kingdo' deeya ni
store.to 3.go 3.say
He said she's going to the store.

75. Q: Ja'an la hahgo Mary doogia/ rfzin

John Q when.F Mary 3.F.go 3.want
When does John want Mary to go?
 157

R: Ja'an agbn 'ni'a'ago Mary dooga l4 ndzin

John noon Mary 3.F.go 3.want
John wants Mary to go at noon.

76. Q: Jaansh Mary h ade'' naagha ni

John.Q Mary where.from 3.come 3.say
Where did John say Mary comes from?

R: Ja'an an Mary kinIanidg' naagh' n

John thus.3.say Mary Flagstaff.from 3.come 3.say
John said (thus) that Mary comes from Flagstaff.

The first thing to demonstrate is that the question particle is semantical-

ly part of the WH word, i.e., that these are direct WH questions from the

speaker's point of view rather than a WH question from the subject's point

of view or a yes/no question containing a direct or indirect WH question.

In other words, (77a), not (77b)-(77d) is the correct rendering of (74).

77a. Where did John say Mary is going?

b. John asked where Mary is going.
c. Did John ask where Mary is going?
d. Did John mention where Mary is going?

Assuming that /la/ and /-sh/ (as opposed to /-/sh/) are used only for WH

questions, this fact is obvious. In other words, the use of /la/ in (74)

automatically marks the sentence as a WH question. Since there is no

second question particle (an /- sh/ or /-sha'/), the sentence could not

also be a yes/no question. However, since the work on the question

particles is far from completed and we already know that people vary in
 158

the way they use them, I further substantiate this claim through use of

dialogues. The fact that (74) cannot be answered by a yes, but can and

must be answered by naming a place proves that in (74) the speaker is

asking a question and it is a WH question.

2.3.5. Generating Case (6)

The next thing to consider is how forms like (74) would be produced.

We have in fact already discussed the mechanisms required. They include:

1. leftward WH-movement

2. topicalization

3. second position particle placement

Using (74) for demonstration purposes, consider the following derivation:

(74A) /

Jaan 1[Mary ha'sgd6d l' deeya] 8 1 n

rule (1): hago' 1a Jaan 3 1 [Mary deeya]3 1 n

rule (2): Ja'an haldgo'V 1 5 1 [Mary deeya/]S n

rule (3): Ja'an la' ha'a'gdcd 1 Mary deeygI8 n

Although there is nothing in the derivation which requires the question

particle to originate with the WH word, I have put it there to emphasize

that semantically it is associated with the WH word.

What is important to notice is that although the WH word does not

end up in sentence-initial position and does end up adjacent to the direct

 159

discourse S, it has in fact been moved out of S, and therefore out of the

scope of the direct discourse point of view by a rule we need anyway, the

unbounded leftward movement. It is the fact that the WH word has been

moved out of the scope of the direct discourse which makes these direct

questions; it is not its occurrence in first position which does so. How-

ever, needing a rule to move WH words, it is logical to assume that Navajo

used the universal question formation rule postulated by Baker, especially

since that rule is necessary in Navajo anyway.

2.3.6. Case (6) Or Case (4)?

While the above derivation of (74) is the one most consistent with

the theory that WH-movement out of the direct discourse segment of the

sentence is necessary to create a direct question, it is not the only

derivation that will generate (74). The topicalization rule, unlike in

English, seems to apply within lower S's. Thus, (78b) is an acceptable

variant of (78a).

78a. Jdan Mary kinlanigo' adoolbas ni

John Mary Flagstaff.to 3.F.drive 3.say
John says Mary is driving to Flagstaff.

b. Ja'an kinla'nigoo' Mary adoolbas ni

John Flagstaff.to Mary 3.F.drive 3.say
John says to Flagstaff Mary is driving.

Under the assumption that (78a) is composed of two S's,

 16o

(78) SO(J an S1 [Mary kinlanig' / adoobs]/ 0

topicalization can apply to S1 , generating (78b) with the following

structure:

(78b) 0 [Jaan 3 1 [kin;ldnigo' Mary adoolbqs]S1 ]3

In a parallel fashion, (74) would be generated by applying topicalization

to S1 instead of SO'

(74B) 0 Jan 1(Mary hia/gl deeyalJ1 n ]0

rule (2): Jaan 1[haagoolal Mary deeya] S i

rule (3): Ja'an la' 1haago'o Mary deeyalq1 n1

Notice that, while the same linear string is produced by (74A) and (74B),

the bracketing is different. WH-movement applied in (74A), moving the WH

word out of S1 , the direct discourse segment of the sentence. In (74B),

however, the WH word is still within S1, although the question particle

is not. In all my explanations of how the point of view of direct WH

questions is determined, I have assumed that both the WH word and the

interrogative particle had to be removed from the scope of the direct

discourse. The arguments have concentrated on the WH word because that

is the part one least expected to find moved. However, in all previous

cases the question particle and the WH word have always ended up in the

same S, even if not together. If (74B) is a permissible derivation, then

 161

note that it is necessary to have an unbounded movement rule which moves

the question particle without the WH word.26 If (74B) is a permissible

derivation, it is necessary to decide what it means, i.e., who is asking

the question, and how it gets its interpretation. If both (74A) and

(74B) are permissible derivations, then one must question whether the

different bracketings result in different interpretations. (74B) is an

example of case (4), the next case to be discussed. However, since the

existence of (74B) creates more doubts about (74A), I will further justi-

fy the rules which give rise to (74A) before continuing to case (4).

Recall that the use of topicalization after WH-movement was

originally postulated for example (61), a direct question from the

speaker's point of view with the WH word and question particle second

in the sentence. If that bracketing is necessary, then (74A) is in fact

a derivation of Navajo.

The argument that (61) requires bracketing the WH word plus question

particle in the higher S is based on the fact that it can be a direct ques-

tion. Let me turn to another example which does not rely on the interpre-

tation of the question word. Consider the following:

79. Ja/an 1d ha'at11/ shizh'e neidiyoolnih yo'ni

John Q what 1.father 3.3.F.buy 3.of.3.say
What does John expect my father to buy?

/Iyoni/ say of him requires an object as well as a subject. Therefore

/shizhe'e'/ is in the higher S, leaving no doubt that the WH word is in

 162

the higher S although not in initial position.

Notice that although father (/shizhe'e'/) is the subject of buy as

well as the object of expect, /shizhe'e'/ cannot be in the lower S. Be-

cause of the direct discourse point of view, a first person pronoun in

the lower S would refer to John, not to the speaker, and therefore first

person possessive refers to John's father, not the speaker's father. This

is the case in (80).

80. Ja'n 1 ha'a't1 Mary shizhe'e ya neidiyoo;Lih yo'n

John Q what Mary 1.father 3.for 3.3.F.buy 3.of.3.say
What does Johni expect Mary to buy for hisi father?

In this example, linear order alone is sufficient to prove that /shizhe'e'/

is in the lower S. That is, /yo'n/ takes only two NP's, /Jaan/ and /Mary/.

(The moved WH word does not create confusion since /yo''n:/ requires an

animate subject and object.) The third animate NP must therefore be in

the lower S, a fact which is confirmed by the translation. That is, only

when in a direct discourse S can a first person pronoun refer to someone

other than the speaker.

In summary, the contrast between (79) and (80) reinforces the claim

that WH words will occur in a higher S without being first in that S.

(74A) is therefore a possible derivation of (74).

2.3.7. Case (4)

Let us now return to (74B) as a possible derivation of (74), i.e.,

to case (4). We have already seen examples like (61) with two possible
 163

derivations. If (74B) is a permissible derivation, (74) would be parallel

to (61) in that, according to the possible derivations, the WH word can

end up in either the matrix S or the embedded S. The difference between

the two is that no matter how (61) is bracketed, the WH word and questionpar-

ticle willalwaysbe in the same S. This is not true of (74). Although

the WH word can be in the lower S, the question particle cannot be. It

is easy to see how (61) is ambiguous, but what happens when the question

word is divided up? That is, I have claimed that anything within a direct

discourse S is from the point of view of the subject of the matrix S, while

everything else is from the point of view of the speaker. If the question

particle in the higher S denotes that a direct WH question is being asked

by the speaker, but there is no WH word that the speaker is questioning

because that word remains within the domain of the subject of the matrix,

is a consistent interpretation possible?

Since (74) has another possible derivation, it is more difficult

to determine the acceptability of (74B) and sort out its meaning if good.

There are, however, WH elements which can never be moved away from their

verb and therefore could not undergo the WH-movement transformation. If

there exist grammatical examples with these WH elements in a lower S and

a WH question particle in a higher S, it is clear that the WH question

particle can move unboundedly without the WH word. Such cases do exist.

Consider the following:

 164

81a. at'6e'd haash wolye

girl how.Q 3.name
What is the girl's name?

b. *haash at'e'ed woly

how.Q girl 3.name

82a. 1I' l' haandel/a/'

horse Q how.3.there be
How many horses are there?

b. *haala' 1i' LL
neel'/4
how.Q horse 3.there be

C. 11' haalaneela'
horse how.Q.3.there be

83. Jaan la haa yinishye'n

John Q what 1.name 3.say
What did Johni say hisi name is?

84. Jaan la' shill ' haaneel'a ni

John Q 1.horse how.3.there be 3.say
How many (of hiss) horses did John say there are?

Examples (8l)-(82) demonstrate that the WH element /haa/ can never move

away from the verb. 27 Therefore, in (83)-(84) there is no doubt that the

WH element has remained in the lower S. In (83)-(84) there is also no

doubt that the WH question particle is in the higher S, since /Jaan/ can-

not be the subject of either a first person verb ((83)) or a neuter con-

struction ((84)). There also seems to be no doubt about the grammaticality

of (83)-(84). It therefore seems that case (6) does exist in Navajo for
 165

at least some speakers.

One need not resort to such special cases to demonstrate the

acceptability of case (6). Word order alone is sufficient to prove that

the WH question particle has moved but the WH word has not. Consider

(85):

85. Jaan la' Mary haade/' naagha ni

John Q Mary where.from 3.come 3.say

We know that /ni/ does not take NP objects, so that /Mary/ cannot be the

object of /ni/.

86. *Jan Mary n...

John Mary 3.say

87. Jaan Mary yi;Lni...

John Mary 3.3.say
John said to Mary...

Since /naagha / is a third person verb, /Mary/ is obviously its subject.

In addition, we know that particles are attached to the end of the first

word, so that /l/ is in the same S as /Jaan/. /Ja'an/ is clearly the

subject of /ni/. Thus (85) must be bracketed as follows:

[Mary haa'dd' naaghdg] n3]

SO(J/an la 5
 166

2.3.8. Interpreting Case (4)

(85) is therefore a case in which the WH word is in the embedded

S, a direct discourse S, while the question particle is in the matrix S

and therefore not under the scope of the direct discourse. How then

should (85) be interpreted? Is it a question or a report of a question?

If a question, is it a direct or indirect question? The following methods

of reasoning lead to different conclusions:

i) It is the position of the question particle which determines

who is asking the question (the speaker or the subject of the S immedi-

ately dominating the question particle). Since the question particle is

not within the direct discourse S, (85) is a direct question. Since /1/

is not used for yes/no questions, (85) must be a WH question. Although

the WH word is within the scope of the direct discourse, the /la/ forces

a change in its point of view. (85) is therefore a direct WH question

from the speaker's point of view, just like (74). If this theory is

correct, then my rationale for WH-movement must be reconsidered. If it

is possible to change the point of view of an element without movement,

why move it at all? Note, however, that the interpretation of the question

is still based on an unbounded leftward movement rule, necessary to create

the correct point of view for the question. The difference is that this

rule moves only the particle, not the WH word also.

ii) It is the WH word which determines the point of view.

Question particles are necessary to distinguish direct questions from

WH pronouns or indirect questions. However, since any question particle

 167

can be placed in second position, those particles are not themselves

marked for point of view. Sentence (85) therefore is not a question

for strict constructionists, but a report of a question's having been

asked.

iii) Both the question particle and the WH word get marked for

point of view. Since the question particle is outside the scope of the

direct discourse, (85) is a direct question from the speaker's point of

view. Since /la/ refers to WH words, the question is about a WH word.

WH words, however, serve several functions in Navajo, usually differen-

tiated by enclitics. One function is to stand for a specific name,

either known to the speaker but not specifically mentioned (/hag 6da/),

or not known to the speaker but the speaker assumes a specific reference

and usually assumes that reference is known by someone else (/ha'a'gooshi/).

/hag6osh/ may be translated by both what and something. In other words,

there is nothing about the Navajo WH word which makes it a question ele-

ment. To leave the WH word in the direct discourse segment of the sentence

does not therefore indicate that the speaker is asking a question. (85)

is therefore interpreted as the speaker asking a question about a WH word,

using that WH word in his report of what the subject of the sentence said.

One could consider the WH word a PRO substitute for what was actually

said, similar to the English "Bill told me what to buy". Under this

theory, (85) would be asserting that John named the place Mary is from

and requesting the hearer to repeat that name. There is then, no real

difference between (85) and any other WH question asked by the speaker.
 168

Although theories (i) and (iii) make different claims about the

point of view designations of the WH word and the question particle, they

end up with the same result, i.e., interpreting (85) as a direct question

from the speaker's point of view. I therefore see no way of distinguish-

ing the two through the data. However, it should be possible to decide

whether (85) is a direct question from the speaker's point of view or

a direct question from the subject's point of view, and, therefore, from

the speaker's point of view, a report of a question. Even this determina-

tion is difficult to make, since speakers will say that these sentences

are grammatical, but hesitate when it comes to translating them. The

best test I know is the use of dialogues. But these also present diffi-

culties, as the most natural answer to (85) is

88. kin;la'nide'' naagha

Flagstaff. from 3.come
She comes from Flagstaff.

While it is true that a report of a question from the subject's point of

view is not to be answered, (88) is still an appropriate follow-up in a

dialogue. Consider the English.

89. A: John asked where Mary comes from.

B: Oh, she comes from Flagstaff.

The most natural responses are therefore not the most useful ones. Yet,

as far as I can judge, and this with some trepidation, the following data
 169

are accurate for some speakers of Navajo:

90. Q: Jaan la'Mary hiaad' naagha ni

John Q Mary where.from 3.come 3.say

R: Ja'an Mary kinlanfdd'' naagha' ni

John Mary Flagstaff.from 3.come 3.say
John said Mary is from Flagstaff.

91. Q: Ja/an llaii' haaneeld' ni

John Q horses what.3.there be 3.say

R: k'ad neezna naakai na

now ten around.pl.go 3.say
He says that now there are ten (walking around).

92. Q: Ja'an 1 d 1 at'eed haa wolye ni

John Q this girl what 3.name 3.say

R: Jaan anaigo Mary wolye' na

John thus.3.say Mary 3.name 3.say
John said (thus) that her name is Mary.

I therefore conclude that these are direct questions, i.e., that the point

of view of a direct question is determined by the discourse segment the

question particle is in. I will leave open the factual question of how

the point of view of the WH word is interpreted,since I know of no way

of making that determination. However, I will return to these issues

in the next section when speculating about how this dialect of Navajo

may have developed.

 170

2.4. Summary, Conclusions, and Speculations

2.4.1. Summary

Since I have been describing a myriad of logical possibilities

related to confusing data, I will summarize my claims and classify the

various types of sentences as an aid in accounting for the data.

I have shown that WH questions contain both a WH word and a question

particle. The particle may remain with the WH word or move to second posi-

tion in any S, with the caveat that the question particle may not end up

dominated by the WH word.28 In simple sentences, the WH word may optional-

ly move to initial position. Since it is not the purpose of such movement

to form a question (i.e., the sentence will be a question whether or not

there is movement), it is not clear whether the movement is due to a

special WH-movement rule. In complex sentences containing an indirect

discourse segment and a direct discourse segment with an underlying em-

bedded WH question, there are four logical possibilities.

1. Both the WH word and the question particle end up in the indirect

discourse segment of the sentence, i.e., in an S interpreted from the

speaker's point of view. Type (1) corresponds to alternative (7) in the

Table of Logical Possibilities for Direct Discourse Question Formation

(p. 130). These sentences vary greatly in acceptability, both from speaker

to speaker and from example to example. If accepted, however, they are

always considered direct questions from the speaker's point of view.

 171

2. Both the WH word and the question particle end up in the same

direct discourse segment of the sentence, i.e., alternatives (1) and (5)

in the table. Strict constructionists interpret all parts of this segment

which are under consideration from the point of view of the subject of

the immediately dominating S. These therefore are reports of questions,

not direct questions. Loose constructionists find these examples ambigu-

ous, adding a second interpretation from the point of view of the speaker.

Given this indirect discourse interpretation, these examples may also be

direct questions.29 People also have difficulty with these sentences.

3. The WH word moves up into a higher S and the question particle

remains within the direct discourse segment, i.e., alternatives (2), (3),

(6), and (9) in the table. Everyone agrees that these examples are un-

grammatical.

4. The question particle moves up into a higher S and the WH word

remains behind, i.e., alternatives (4) and (8). These sentences create

the greatest problem for most people. When accepted, however, they are

almost always interpreted according to where the WH question particle

occurs rather than according to the placement of the WH word itself. If

the question particle is in the matrix S, these sentences are considered

direct questions from the speaker's point of view; if the question particle

is in an intermediate S, such a sentence is taken as a report of a question

which had been asked by the subject of the S immediately dominating the

question particle.
 172

To account for type (1), I have postulated an unbounded WH-movement

transformation, moving both the WH word and the question particle to ini-

tial position. This rule applies before Topicalization and the Particle

Placement rule, accounting for all the acceptable word orderings. Type

(2) is generated when unbounded WH-movement does not apply. Since the

movement includes the question particle, type (3) will never be generated.

Type (4), presumably is generated not by WH-movement but by an unbounded

application of Particle Placement.

Assuming my account of the data is accurate, the first three types

follow nicely from a theory which says that point of view is determined

solely by the nature of the S an element ends up in, and therefore un-

bounded movement is used here in order to create direct WH questions. If

type (4) is in fact acceptable and is a direct question type, then it

brings into question both how point of view is determined and why there

is any unbounded WH-movement at all. While this type allows an element

(the WH word) to remain within the direct discourse S and yet be inter-

preted as if it had been moved, if the WH question particle is generated

with the WH word as I have been assuming, this type still requires an

unbounded particle placement transformation to signal that interpretation.

2.4.2. Is There Really Any Unbounded Movement?

I cannot provide any definitive answer to why strict construction-

ists allow type (4) questions because the data are too murky to support

any substantive claims. Rather, I would like to speculate a bit on why

 173

the data are so unusually difficult to compile. These speculations suggest

ways in which these rules and sentence interpretations may have developed.

To begin with, in this instance it seems relevant to take into

consideration the fact that these sentences are not used in normal con-

versation. While one commonly hears questions in English like "Where did

John say he's going?", "What do you want to buy for Mary?", etc., these

questions are not heard in Navajo. The reason why such questions are not

common in Navajo could, of course, be extra-linguistic. If linguistic,

there are two most likely explanations: either the unbounded leftward

movement does not really exist in Navajo or the factors involved in these

sentences are so complex that people avoid them. My work is obviously

based on the latter assumption. However, the more I investigate, the

clearer it becomes that these sentences create real confusion. Neverthe-

less, there are some examples which are pretty universally accepted. These

are examples of type (1) (not type (2)), usually with a single NP, for

example:

93. haagoold Jaan deeshaIL n-azin

where.to.Q John 1.F.go 3.want
Where does John want to go?

It is certainly possible that even these sentences are ungrammatical,

since when first presented with such sentences, most Navajo speakers

hesitate. It could be that they are then accepted by analogy to simplex

sentences. Thus, simple sentences optionally move WH words to initial

position. Like a simple sentence, (93) has only one subject. In fact,
 174

while admittedly ungrammatical, people will even say (94) instead of (93):

94. *ha'g66-1a" Jaan nizin

where.to.Q John 3.want

I reject the hypothesis that all these examples are in fact un-

grammatical and that unbounded WH-movement does not now exist for any

speakers, primarily because I think that acceptance by analogy involves

a more clear-cut distinction between what is acceptable and what is not.

Thus, if analogy were the only criterion, any sentence with two subjects

(and therefore obviously two S's) would be equally unacceptable and every-

one would agree. Everyone should agree that (95) is both unacceptable and

as bad as (96).

4
95. haa~'g/6la' Jaan Mary doog l nzin
where.to.Q John Mary 3.F.go 3.want
Where does John want Mary to go?

96. haa'go5la Jaan shizhe'6' dooga l nfzin

where.to.Q John 1.father 3.F.go 3.want
Where does John want his father to go?

While some people hesitate over (95), others find it as acceptable as (93).

And having no trouble with (95) does not mean equal comfort with (96). The

difference between (93) and (95) is that in the given form of the verb,

/nizin/ takes only one NP. Since (95) has two NP's, it looks less like a

simplex sentence than does (93) with only one NP. (96) has the same number

of NP's as (95), but has the added complication that one NP has a possessive
 175

pronoun which must be assigned a point of view, requiring a decision on

the part of the speaker which takes into consideration the fact that the

sentence contains an embedded direct discourse S.

Notice that the embedded verb in (93) is obviously being inter-

preted as within a direct discourse segment, while the verb in (95) could

be either direct or indirect since it would receive the same meaning. Is

(93) then really more like a simplex S than (95)?

There is another argument against claiming that the problems with

the data derive from the fact that there really is no unbounded movement.

People have at least as much trouble with type (2), the case with no

movement at all, as with type (1) which involves the application of un-

bounded WH-movement. This fact seems to suggest that some factor other

than (or in addition to) unbounded movement makes these sentences con-

fusing.

If one is to assume that there is no unbounded WH-movement in

Navajo, one must still account for why some examples are considered

acceptable (if ungrammatical), others unacceptable, with a large group

of examples generating ambivalent responses. In going over the data that

I now have, I see no principled way of accounting for the reactions under

the assumption of no unbounded movement.

Another way of looking at these facts is that there is no unbounded

WH-movement but that these examples are grammatical because there is no

separate S at the time the movement rule applies. The problem of proving

that direct discourse segments, having no overt complementizer, are in

 176

fact separate S's has already been discussed in Chapter 1. While I have

little proof that the S's are considered separate after WH-movement, there

is syntactic evidence that the surface strings without movement are in-

terpreted as separate S's.

2.4.3. An Argument For Separate S's: Double Questions

One argument that the direct discourse segment is a separate S is

based on the fact that it is possible to get two questions, a direct

question and a report of a question. Normally, of course, it is impossible

to have two such questions within one S. Consider the following:

97. da' Jaanish ha'at'ifl Mary ba nahideeshnih ni

Q John Q what Q Mary 3.for 1.F.buy 3.say
Did John ask what he should buy for Mary?

Notice that the WH word and the WH question particle are positioned so that

they could belong to either S. Without the yes/no question, as in (98),

this string would be ambiguous.

98. Q: Ja'an ha'Iat'iila Mary ba' nahideeshnih ra

John what Q Mary 3.for 1.F.buy 3.say
(i) John asked what he should buy for Mary.
(ii) What did John say he'll buy for Mary?

99. Jaan aiigo ak''an Ia' nahideeshnih rI

John thus.say flour some 1.F.buy 3.say
John said he'll buy some flour.
 177

The first translation of (98) is the same as that in (97).31 The fact

that (99) is a possible response to (98) demonstrates that (98ii) is also

a possible translation of (98). Since it is impossible to have a yes/no

and WH question in the same S, adding the yes/no question as in (97) dis-

ambiguates the string in (98).

The same result occurs with the particle /ya'/, the Navajo equi-

valent to an English tag question. Since the addition of /ya'/ makes an

S a question, there cannot be another question in that same S. Adding,

/ya'/ to (98) therefore disambiguates the string, forcing the interpreta-

tion in which the WH word and question particle have remained in the

lower S.

100. Ja'an ha'at'ill& Mary ba nahideeshnih ni ya'

John what Q Mary 3.for 1.F.buy 3.say right
John asked what he should buy for Mary, didn't he?
*What did John say he'll buy for Mary, right?

2.4.4. Arguing From Emphatic Particles

Another and perhaps stronger argument that there are separate S's

is based on the use of the emphatic particles /ga'/ and /yee'/. These

two cannot co-occur in an S with a question word.

lOla. ashkii ga' I' nablIgo'

boy emph horse 3.3.P.throw
It's the boy who was thrown by the horse.
 b. ashkii 11' gal nab lgo'
boy horse emph 3.3.P.throw
It's the horse that threw the boy.

c. *haila A
1' gal nablilgo'
who.Q horse emph 3.3.P.throw

d. *ashkii ga' ha'at ' ila nabil go'

boy emph what Q 3.3.P.throw

102a. Jaan Bil yee' neidii ts'in

John Bill emph 3.3.hit
It's Bill who John is hitting.

b. *hail Bil yee' neidi lts'in

who.Q Bill emph 3.3.hit

Perhaps one could finesse the double question argument by saying it is

not a matter of grammaticality, but of finding an acceptable interpreta-

tion. However, in the emphatic particle case, we have good English

sentences with no direct Navajo translation.

103. Where did John tell Mary he ran away from?

104. John asked Mary where she ran away from.

The direct translation into Navajo of (103) and (104) without emphasis

are both grammatical.

lO3a. halde' ll Jdan Mary y0d' eeshwod yin

where.from.Q John Mary away 1.P.run 3.3.say
Where did John say to Mary that he ran away from?
 179

104a. Jaan Mary haa/dy"la yo'' inilwod yiln

John Mary where.from.Q away 2.P.run 3.3.say

Adding emphasis in the normal fashion, by placing /ga'/ after the word to

be emphasized, produces only one good sentence.

103b. *ha/a'dq"'tld Jdan Mary ga' yo" eeshwod yilni

where.from.Q John Mary emph away 1.P.run 3.3.say

104b. Ja'an Mary ga' hadde'e'l yoo' iinlwod yili

John Mary emph where.from.Q away 2.P.run 3.3.say
John asked Ma where she ran away from.

(1O3b) is ungrammatical because the question word cannot be in the same S

as the /ga'/. The fact that (104b) is grammatical means that the /ga'/ and

the question word are in separate S's.

2.4.5. Arguing From Negatives

The use of the Navajo negative /doo. ...da/ provides another argument

that the direct discourse segment of the sentence is a separate S, and

probably the best argument that the S's are still separate at the time

of the unbounded movement. /doo/ can generally precede any segment within
32
the same S, although any differences still need to be investigated.

105a. Ja'an Bostongoo' doo adooIbpsda

John Boston.to neg 3.F.drive.neg
John will not drive to Boston.
 b. Ja'an doo Bostongo' adoolbasda
John neg.Boston.to 3.F.drive.neg
John won't drive to Boston.

c. doo Jaan Bostongoo adoolbasda

neg John Boston.to 3.F.drive
John won't drive to Boston.

Now consider the following:

106a. Ja'an Bostong6' jidooga/'l doo sho'raida

John Boston.to 4.F.go neg 1.of.3.say.neg
John doesn't expect me to go to Boston.

b. Jaan Bostongoo doo jidoogalda sh6"n

John Boston.to neg 4.F.go neg 1.of.3.say
John expects me not to go to Boston

c.?*Jaan Bostongdo' doo jidoogaal sh Infida

John Boston.to neg 4.F.go 1.of.3.say

(106a) and (106b) prove that the two verbs have not become one, one of

the ways the lower S node could be eliminated (cf. Chapter 1). The un-

grammaticality of (106c) follows from the hypothesis that there are two

S's. Since /shon / is in a different S from /Bostongoo/ and /jidoog'L-/,

the /doo/ and /da/ cannot be interpreted as part of the same S. If ac-

ceptable at all, it will be interpreted as two separate instances of

negation. Thus, consider the translation of the following:

107a. Mary Jaan Bostongoo' doo adoolbts niida

Mary John Boston.to neg 3.F.drive 3.say.neg
Mary didn't say whether or not John couldn't drive to Boston.
(107a) is being interpreted as deletions from (107b).

107b. Mary Jaan Bostongo'o' doo adoolbasda doo niida

Mary John Boston.to neg 1.F.drive.neg neg 3.say.neg
Mary didn't say that John isn't going to Boston.

Each of the two S's thus contains a separate instance of negation.

Having seen that a negative on the lower verb provides evidence

that there are two separate S's, consider the following:

108. haa'g61 4 Jaan doo jidoogaalda sho'ra

where.to.Q John neg 4.F.go neg 1.of.3.say
Where is it that John expects me not to go?

Sentence (108) is an example of unbounded movement since John is the sub-

ject of expect. The fact that movement is possible when the lower verb

is negated proves that the lower S is still there when question movement

occurs. Sentence (108) is positive evidence that question formation is

truly an unbounded transformation.

2.4.6. Arguing From Second Position

The last argument that the direct discourse segment of the sentence

is a separate S is based on the positioning of question particles. Recall

that I demonstrated for simple sentences that if a question particle is

not attached to its WH word, it can appear in second position and only in

second position. While it is not completely clear how second position

should be defined, some of the type (2) examples make sense only if second
 182

position is being marked in an embedded S. Consider the following:

109. Jaan Mary ada/a/d a? la hiaagdM nisiniya'yilbn

John Mary yesterday Q where.to 2.P.go 3.3.say
John asked Mary where she went yesterday.

110. Jaan Bilshat had/' naaghd n

John Bill.Q where.from 3.come 3.say
John said, how about Bill, where is he from?

Assuming a single S, there is no way /la/ in (109) could be taken to be

in second position. However, /yi;ln / requires an object as well as a

subject so that /Mary/ as well as /Ja'an/ can belong to the higher S. Since

the embedded verb is in the second person, not the third, /Mary/ cannot be

in the lower S. Thus, under the assumption that there are two S's, (109)

would be bracketed as follows:

S[Jan Mary S [ad 'd'a" la ha'go'o nisinfyal]5 yil ]5

s1 1 0o

In this case, /la/ is clearly in second position within

S .
The same reasoning applies to (110). In addition, notice that

/-sha'/ is being used not only as a question particle, but as a topic

marker. Navajos often use word order to mark topics. In general, topics

go into initial position. The use of /-sha'/ as a topic marker in (110)

therefore also argues that there is an embedded S. That is, /Bil/ is the

topic of the lower S and thus in first position in that S:

S[Jaan 5 [Bilshg' hade(' naagha]

SO S1L 5 1 n ]00
 The fact that there is a lower S with an initial topic position is

also shown by (111).

111. Jaan kinlanigoo' Mary adoolbas ni

John Flagstaff.to Mary 3.F.drive 3.say
John said, to Flagstaff Mary will drive.

112. Ja'an Mary kinlanagoo adoolbas na

John Mary Flagstaff.to 3.F.drive 3.say
John says Mary will drive to Flagstaff.

(112) is the neutral, and more normal, word order, but (111) is acceptable

especially when /kinLan / is being used as a topic.

2.4.7. Topicalization and Unbounded Movement

While the topicalization in (111) seems to argue that the direct

discourse segment is a separate S, topicalization facts are no less con-

fusing than the questions. Sentence (113) is generally accepted, suggest-

ing that topicalization, like WH-movement, is an unbounded movement rule.

113. kin;aLnig0' Jan adeesbas ni

Flagstaff.to John 1.F.drive 3.say
It's Flagstaff John says he'll drive to.

However, if (113) is generated by unbounded movement, then (112) should

have an analogous variant. (114), however, seems to create confusion for

at least some people who accept unbounded WH-movement.

 114. ?kinlangoo Ja'an Mary adoo;Lbas ni
Flagstaff.to John Mary 3.F.drive 3.say
It's Flagstaff John says Mary will drive to.

Sentences (113) and (114) thus revive the possibility that there is no

unbounded leftward movement.

2.4.8. Speculations: The Genesis of the Movement

I believe it is significant that the unbounded topicalization is

less acceptable than the unbounded WH-movement. While there is insuffi-

cient evidence to argue for any theory of change, I would like to speculate

on the genesis of these cumbersome rules. On the basis of the facts pre-

sented, it would not be surprising to find that unbounded leftward move-

ment is a relatively new phenomenon in Navajo. One possibility is that

people first started moving WH words to initial position by analogy to

movement in simple sentences. In that case, movement would be most

acceptable when the embedded subject is identical to the matrix subject.

As discussed above, (93) but not (95) would then be acceptable. Since

accepting (93) creates the ability to ask a whole new set of direct

questions, it is not surprising that the movement would be generalized

and reanalyzed as an unbounded WH-movement rule. The addition of an

unbounded question formation rule significantly increases the types of

questions that can be asked. Unbounded topicalization, on the other hand,

adds little to the expressive power of the language. There is thus more

motivation for accepting unbounded question formation than there is for

 185

unbounded topicalization. Since unbounded question formation rules create

confusion, any unbounded topicalization should create as much if not more

confusion.

2.4.9. Speculations: Why So Confusing?

Let me now return to the central claims of this chapter and try

to account for Navajo speakers' unusual hesitation with these types of

sentences. Under my hypothesis that there is unbounded leftward WH-

movement, all sentences of type (1) should be equally grammatical. The

hesitations must therefore be due to other factors. I believe that the

use of direct discourse is the root of the problem: the difficulty with

a sentence is proportional to the points of conflict between direct and

indirect discourse. Sentences may initially be given an inappropriate

interpretation because that interpretation presents fewer conflicts. It

is not unusual for me to be told "I'm sure the sentence is grammatical,

but it's difficult to translate." This remark might mean that the sen-

tence is well-formed but very opaque, i.e., difficult to understand and

not just difficult to translate. There are several ways direct and in-

direct discourse can come into conflict:

1. The points of view of the verbs will be different. Interest-

ingly, this seems to create the least confusion. Thus, there is no

hesitation in explaining that a first person verb refers to the subject

of the matrix.
 186

2. The point of view of pronouns will be different. Since subject

and object pronouns are marked in the verb, they seem to be no more diffi-

cult to interpret than the verb is. However, possessive pronouns do

present more difficulty.

3. Placing particles into second position creates conflicts. Thus,

the most natural place in a sentence for particles is after the first word

of the entire sentence, i.e., linear second position. People vary in how

strongly they prefer linear second position to other places in a sentence.

For example, there is no doubt that a question particle can always occur

with its WH word. Yet, some people find (115) much less desirable that

(116)-(117), explaining that /l// should be in second position.

115. Jan haat I i'la nayiisnii

John what Q 3.3.P.buy
What did John buy?

116. Jian 1 ha't 'i' nayiisnii'

John Q what 3.3.P.buy
What did John buy?

117. ha t i'il Jaan nayiisnii'

what Q John 3.3.P.buy
What did John buy?

For sentences which are reports of questions, the question particle re-

mains within the direct discourse S (type (2)). If the question particle

is to remain within the direct discourse segment, the use of linear second
 187

position is impossible given a full matrix subject and normal word order.

That is, second position in the embedded S is not linear second position.

This problem can be avoided by extraposing the matrix subject34:

118. ha'de''la Mary naagh/ n Jan

where.from.Q Mary 3.come 3.say John
Where is John, said Mary.

119. havat' ila' Bil yizhbizh nizin Jaan

what Q Bill 3.3.braid 3.think John
John is wondering what Bill braided.

As far as I can judge, sentences like (118)-(119) do seem easier to

interpret.

4. The question particles /la/ and /-sh/ normally indicate direct

questions. There is a conflict between the recognition that there is a

direct question and the fact that the entire sentence is really a declara-

tive sentence. In fact, some people will answer sentences like (120), at

the same time giving the translation I have given.

120. Jaan Mary l' h lde/' naagha ni

John Mary Q where.from 3.come 3.say
John asked where Mary comes from.

The reasoning seems to be that, although (120) is not actually a question,

the intent can be to get some information. It seems to me that there is

a desire to answer (120) because of the occurrence of /l// which normally

signifies that an answer is requested. It is significant, however, that

(120) would be answered only by (121a), not (121b).

121a. Mary kinbanidq' naagha'

Mary Flagstaff.from 3.come

b. #Jaan Mary kinlkanidn' naagha ni

John Mary Flagstaff.from 3.come 3.say
John says Mary comes from Flagstaff.

In other words, the question being answered is "Where is Mary from?" and

not "Where does John say Mary is from?". The WH word and question particle

are not being interpreted as a question about the entire sentence as they

are when moved out of the direct discourse S. Rather, (120) is being in-

terpreted as a request to answer a question put by John. Sentence (121a)

is thus an attempt to resolve the conflict between the question particle

request for an answer and the direct discourse indication that John is

the one who asked the question.

5. Hale and Platero (1973) demonstrate in their investigation of

Navajo relative clauses and pronominalization that there seems to be a

surface constraint having to do with possible interpretations of a sentence.

They propose that the following parsing principle is employed with varying

degrees of strictness by speakers of Navajo: "In a string : of the form

(NP) NP V, the subsequence (NP) NP is a clausemate with V if the string

exists as a well-formed simple sentence of Navajo" (12). Since unbounded

movement of NP's rearranges the linear order of the NP's in a sentence,

there is a potential conflict between movement and the parsing principle.

Getting the question out from under the scope of direct discourse requires

movement. After unbounded WH-movement has applied, a sentence may become

unacceptable or marginal because of this parsing constraint. Consider the

following:

122. ?ha'at'ill Jaan Bil yishxash ni

what Q John Bill 3.3.P.bite 3.say

(122) was rejected because the string /Jaan Bil yishxash/ ("John bit Bill")

gets interpreted as a separate S. (122) thus gets interpreted as "What

said 'John bit Bill'?". People vary in how strictly they adhere to the

parsing constraint. The constraint can be overridden if the sentence does

not make sense. Thus, since inanimates don't speak, some people will look

for another subject for /n / (say), reinterpreting the sentence to its

intended meaning, "What did John say bit Bill?", and thus accepting it.

However, nobody would interpret (123) as "Who did John say bit Bill?",

the intended meaning if unbounded WH-movement is applied to the following

well-formed underlying string:

so[Jaan 1[hil Bil

yishxash] s1 so
Oni
John who.Q Bill 3.3.P.bite 3.say

123. hail Jdan Bil yishxash ni

who.Q John Bill 3.3.P.bite 3.say
Who said John bit Bill?
 190

Sentence (123) is only given the one translation. There are thus factors

independent of the actual movement which create confusions.

It is clear that it is not the WH-movement but the parsing principle

which accounts for these facts. Thus, WH-movement applies to directional

phrases which cannot be subjects or direct objects, while the parsing

principle applies only to these latter categories. Since the parsing

principle accounts for the possible ungrammaticality of (122), a similar

example with a directional phrase should be acceptable. That is exactly

what we have found.

2.4.10. Speculations About Case (4)

Notice that there are alternatives to (122), namely (124)-(125):

124. Jaan l' ha'a't' l Bil yishxash ni

John Q what Bill 3.3.P.bite 3.say
What did John say bit Bill?

125. Jaan ha'at' i la'Bil yishxash ni

John what Q Bill 3.3.P.bite 3.say
What did John say bit Bill?
John asked what bit Bill.

Sentence (125) is less desirable than (124), both because it is ambiguous

and because the question particle is not in second position when bracketed

for the intended meaning. (As a direct question, /ha'st'ila / is in the

matrix S.) In (124), however, it is not so obvious that the WH word is in

the matrix S. The fact that /la'/ must be in the matrix already indicates
 191

that the sentence is a direct WH question. Since /yishxash/ requires two

NP's, and since Navajo has the surface parsing constraint, I would expect

a tendency to parse (124) as follows:

SO[Jaan laS 1[hatatti' Bil yishxash]1 n] 0

Such a parsing interpretation creates a conflict between the point of view

interpretations of the /la/ and the WH word, which accounts for the con-

sternation caused by this set of type (4) sentences.

It may be that thinking about sentences like (124) is what causes

people to come to accept type (4) sentences, sentences in which the question

particle can only be in the matrix S and the WH word can only be in the

embedded direct discourse S. Sentence (124) is first accepted as a direct

WH question because /la/ is obviously in the direct discourse segment of

the sentence. But then when trying to figure out the meaning, /ha'a't' f/

is parsed as part of the lower, direct discourse S. Since the position

of /la/ already signifies a direct WH question and the WH word must always

be associated with the embedded S,to figure out the meaning of the sentence,

some people might begin to ignore the actual position of the WH word. It

may be that type (4) examples seem grammatical because, ignoring the

direct-indirect discourse conflict, the sentence is a well-formed string

and using (124) as a model for finding the meaning, the type (4) examples

are easily interpretable. That is, the only conflict for WH words posed

by the difference between direct and indirect discourse is who is asking

 192

the question. There is no problem about what the word means within the

sentence. In contrast, with pronouns or verbs, there is a problem in

figuring out who is being referred to. Since the only conflict can be

resolved by the placement of the question particle, type (4) sentences

become interpretable.

This line of reasoning which suggests that type (4) sentences are

acceptable by analogy says nothing about whether type (4) sentences are

in fact grammatical. Notice, however, that if type (4) sentences become

grammatical by analogy for a given speaker, then that person no longer

requires an unbounded WH-movement rule. In other words, if Question

Particle Placement is sufficient to designate a direct WH question, why

ever move the WH word?

2.4.11. Question Particle Placement

There are two things to notice about Question Particle Placement.

First, given a theory in which the question particle is generated with

the WH word, this rule also involves unbounded leftward movement. Since

using Particle Placement instead of WH-movement to designate direct WH

questions does not eliminate unbounded leftward movement, there is no

theoretical reason having to do with possible types of movement which

would lead to the elimination of one rule in favor of the other. In

fact, people who accept both types of movement differ as to which they

prefer. Second, although Question Particle Placement involves unbounded

movement, it makes a sentence look syntactically normal. That is,

 193

ignoring the meaning problems created by the direct-indirect discourse

conflict, a sentence looks normal when the particle is in linear second

position and the WH word remains in the position bearing the underlying

grammatical relation. It is only when figuring out what these sentences

mean that there is confusion. Unbounded Particle Placement therefore is

syntactically easy to accept but semantically confusing. In contrast,

sentences involving unbounded WH-movement are syntactically difficult to

accept but semantically desirable, i.e., easy to interpret. It is there-

fore reasonable that people differ in which they prefer, depending upon

whether they are more concerned with syntactic form or an easily acces-

sible interpretation.

2.5. Formulating the Transformations

Having described, explained, and justified the data and the types

of rules needed to account for it, I will conclude with a discussion of

the exact formulation of those rules. Thus, under the assumption that

type (4) is acceptable, it would be generated by an unbounded application

of Particle Placement. I will continue to consider the theory in which

the WH question particle is generated with the WH word, and then will turn

to a theory in which the WH question particle is generated separately.

In considering actual transformations, it is not clear whether the

unbounded particle placement under discussion is a special application of

the Second Position rule, one part of a general question formation rule,

or a separate rule. I will now consider several options that would handle
 194

the facts, taking into consideration the purpose of each rule, whether or

not a rule is cyclic, and how the rules interact with point of view de-

signation.

2.5.1. Are Particle Placement and Second Position the Same Rule?

Let me first consider whether the unbounded particle movement

should be considered part of the Second Position rule. Since in both

cases the WH question particle ends up in second position, it seems like

they are one rule. However, I believe the two cases should be separate.

The unbounded movement of the question particle differs from the Second

Position rule in two important respects. First, Question Particle Place-

ment affects the meaning of a sentence, while second position rules are

normally housekeeping rules with no effect on meaning. Second position

in Navajo is a neutral position. For example, as I discuss elsewhere

(cf. Chapter 3), the yes/no particle /-Ish/ generally goes into second

position, denoting a straightforward yes/no question. If, however, the

particle is placed on another word in a simple sentence, that word becomes

the focus and the rest of the S is presupposed. More investigation is

necessary, but it may be that the Navajo negative /doo... da/ is most

neutral when /doo/ occurs in second position. /doo/ can precede any

word in the sentence; /da/ is always suffixed to the verb. There is a

clear contrast between the following, as expressed by the translations:

 195

126. Ja'an doo Bostongo'6' adoolblsda

John neg Boston.to 1.F.drive
John won't be driving to Boston.

127. doo Ja'an Bostongoo adoolbpsda

neg John Boston.to 1.F.drive
It's not John who will be driving to Boston.

While the occurrence of a WH question particle in second position may be

stylistically more correct, there is nothing "neutral" about moving that

particle from one S to another, i.e., from one point of view segment to

another. Therefore, while it might be correct to connect the occurrence

of the WH particle in second position to the fact that that position is

considered a neutral position for particles, it does not seem proper to

use the Second Position rule to move a particle from one point of view

segment to another.

The second reason for not using the Second Position rule to move

the WH particle from one point of view segment to another is that the

Second Position rule is normally a bounded rule. That is, other elements

that move into second position normally go into second position in their

own clause. The /doo/ part of the Navajo negative, for example, will not

be placed in linear second position if that requires movement up out of

its own clause. And, as we have already seen, even the WH question

particle can go into second position of a lower S. Therefore, if WH

question particle movement is incorporated into the Second Position rule,

that rule will need an extra case involving a crucial variable. That case
 196

will have to apply only to WH question particles and not to any of the

other particles which can go into second position in their own clause.

In other words, since the Second Position rule is bounded and the WH

Question Particle Placement is unbounded, the latter is not automatically

handled by the former even though they both apply to the WH question

particles and in both cases the particles end up in second position.

2.5.2. An Independent Rule

Since WH Question Particle Placement is not to be incorporated

into the Second Position rule, how then should the Question Particle

Placement rule be written? One option is to have a rule which does not

interact at all with the Second Position rule. Such a rule would take

a /la/, /-sh/, or /-shg'/ from anywhere in a tree and place it after the

first word of the sentence.

A second option would be to write an unbounded movement rule which

feeds the Second Position rule. This option claims that WH Question

Particle Movement involves two separate phenomena, moving the particle

unboundedly from one point of view segment to another, and then placing

it into a stylistically correct position. The unbounded movement has a

semantic purpose; the actual occurrence in second position does not. I

favor this approach because it emphasizes that the occurrence in second

position does not affect the meaning of the sentence. Recall that in

both simple and complex sentences the WH question particle need not end

up in second position; it may remain with the WH word which need not occur
 197

initially. So far as I can tell, when the question particle and WH word

occur in the same S, the movement of the question particle off the WH

word does not affect the meaning of a sentence.35 That is, (128) and

(129) are identical in meaning, but not necessarily in acceptability.

People differ only as to whether they prefer leaving the question particle

with its WH word or putting it into second position.

128. Jaan lhai yiyiiltsg

John Q who 3.3.P.see
Whom did John see?

129. Jaan h la'yiyiilts

John who.Q 3.3.P.see
Whom did John see?

2.5.3. Are Particle Placement and WH-Movement the Same Rule?

Given an approach in which the unbounded movement is not linked to

occurrence in second position, it is possible to consider whether WH

question particle placement and leftward WH-movement should be the same

rule. It would be nice to collapse them because they both have the same

purpose, moving a question element from one point of view segment to an-

other. Creating one rule captures the generalization that one or the

other, but not both, will be used in a given sentence. In addition, they

both need to refer to the WH question particle. Recall that the WH word

may not move and leave the question particle behind, i.e., type (3) sen-

tences are ungrammatical. To rule out these sentences, I suggested that

 198

the WH-movement rule be written so that the question particle obligatorily

moves with the WH word. If the two types of movement are collapsed, one

could say instead that the WH word optionally moves with the question

particle. That is, the rule would collapse the following two cases:

i) X -WH Q- Y
1 2 3 >2-1-3

ii) X -Q - Y
1 2 3 42-1-3

As written, this looks like a straightforward case for parenthesis notation

iii) X - (WH) Q - Y
1 2 3 >2 - 1 - 3

Notice, however, that when case (ii) should apply, there is still a WH

word present in the string. If used here, parentheses would be signifying

that one could choose whether to consider the WH word part of the Q or part

of the X rather than signifying that the presence of the WH word in the

string is optional. Rule (iii) is therefore an incorrect use of the

parenthesis notation.

Although there is another rule which moves the Q particle, it is

impossible to take advantage of it to separate the Q from the WH word

before the question formation rule has applied. That is, one could opt

to order the Second Position rule before question movement in order to

separate out the question particle. But then the question particle would

be stranded sentence-initially.
 199

Jaan [Mary h 'ad"'1 naash] n

2nd Position: Jaan [Mary i1/hsa'de(' naasha] ni
Q movement: *l Jaan [Mary haadee't naash ] ni

It is only when the WH word moves with the Q particle that the question

formation rule places that particle in second position. Question forma-

tion must therefore follow the Second Position rule in order to insure

that the question particle ends up in second position.

Notice that the notation problem is analogous to the problem found

in English when a WH word is part of a prepositional phrase, i.e., how to

allow the preposition to be optionally dragged along in order to generate

both (130)-(131).

130. Where does Mary come from?

131. From where does Mary come?

The Navajo WH word is parallel to the English preposition, not to the

English WH word. In both cases it is the leftmost element which is op-

tionally dragged along. So far as I can tell, whatever solution works

for the English will work for the Navajo.

2.5-4. Chomsky's Proposals

Chomsky's solution for the English case is worth mentioning because

it brings forward several interesting facts about the Navajo case (Chomsky,

1971). He proposes a cyclic movement rule which includes a WH feature.

 200

He then says that the WH feature may be placed on either the PP or the NP.

The optionality therefore is not in the rule, but in the tree.

Looking back at the Navajo, I have so far said nothing about any

feature to trigger the question formation rule, describing the rule in

terms of the question particle only. Such a description, however, is

inadequate because it does not distinguish the WH question particles from

the yes/no ones. Such a distinction is necessary when the WH word is not

part of the structural description, i.e., for case (ii). There is no

reason therefore why the feature used to distinguish WH and yes/no ques-

tions could not also be used instead of the WH word in case (i).

Chomsky's use of a cyclic transformation is based on his theory

of movement rather than on facts about question formation since the WH

word in English can never remain in any intermediate tree. Such cases,

however, do exist in Navajo. Consider the following:

132a. Ja'an Bil la' ha'g6'6 Mary difnas yilIni ni

John Bill Q where.to Mary 2.F.go 3.3.say 3.say
John said, where did Bill tell Mary to go?

b. Jaan Bil la' Mary haLgo' diinal yilni ni

John Bill Q Mary where.to 2.F.go 3.3.say 3.say
John said, where did Bill tell Mary to go?

Such sentences are exceptionally difficult to process because of the long

strings of verbs and of nouns. They are of course made even more difficult

by the changing points of view, and probably by the repetition of the same

verb. While nobody would ever choose to use such sentences, so far as I
 201

can tell they are grammatical. Sentences (132a)-(132b) would be derived

from the following tree:

(132)

iP 1 V

Ja'an ni
John 3.say

NP 2

Bil Mary yiln

Bill Mary 3.3.say

NP

N E Q
I I .Ile I
(ni) h goo la diinda l
(you) where.to.Q 2.F.go

Sentence (132a) is derived by moving the WH word plus Q particle

from S2
to S 1 ; sentence (132b) by moving only the Q particle to S . Since the

movement is optional, it is of course possible to get a good output by

leaving the WH word and Q particle in

S2'

133. Jaan Bil Mary ha'agoo'la' diinaal yiln' ni

John Bill Mary where.to.Q 2.F.go 3.3.say 3.say
John said Bill asked Mary where she i s going.

Since the movement is cyclic, it is also possible to get a good output

by moving the question particle with or without the WH word

to SO'
 202

134a. hadg6dll Ja'an Bil Mary d fndaa yijrn ni

where.to.Q John Bill Mary 2.F.go 3.3.say 3.say
Where did John say Bill told Mary to go?

b. Jdan la" Bil Mary h'ag6 d ina/a' yilni ni

John Q Bill Mary where.to 2.F.go 3.3.say 3.say
Where did John say Bill told Mary to go?

What is not acceptable is to strand the WH word in S1 , moving the Q

particle to S . This is analogous to stranding the preposition in English.

135. *Ja'an la' Bil ha'gd' Mary d na/A yiln ni

John Q Bill where.to Mary 2.F.go 3.3.say 3.say

136. *Where did John say to Bill told Mary to go?

Whatever mechanism is used to throw out (136) should also throw out

(135). Whatever the proper solution for English, it should not be based

on the properties of COMP position or principles for moving elements out

of the COMP node because such a solution is not available for Navajo.

There are other mechanisms available, however.

One option is to use the feature marking, which is a way of saying

that once a choice is made, it is made for the entire derivation. Another

option is to create some kind of meta-constraint to the effect that there

cannot be connections from more than one S back to a trace. I have no

theoretical motivation for choosing any particular option.

 203

2.5.5. Chorus

In sum, collapsing WH-movement and Question Particle Placement is

desirable because the rules are disjunctive, apply in part to the same

constituent, and have the same semantic intent. Having one rule makes

it easier to discuss the different preferences people have. That is, it

is no longer a preference for one rule over another, but a choice as to

whether the WH word should be considered part of the Q. Given one rule,

it is easier to see how type (4) sentences do not require the addition

of a new rule (WH Question Particle Placement), but the elimination of

an element in the Q. As separate rules, the two would share the problems

of how to trigger the rule and whether it should be allowed to apply

cyclically, generating different meanings. Writing one rule, however,

creates some difficulty because the parenthesis notation cannot be used.

Having one cyclic rule also creates the problem of how to avoid stranding

the WH word in an intermediate tree. Neither of these problems is unique

to Navajo, as was seen when looking at English prepositional phrases.

2.5.6. Meaning Changes and a Q Node

There is one other factor about the Navajo that should be considered.

Each of the sentences has a different meaning because the point of view of

the question changes. Is it thus possible to use the point of view desig-

nation as a controller?

I can find no reasonable way of using the point of view designation.

Since an element is from the point of view of the immediately dominating

 204

subject, if point of view is marked before movement, the question element

will never change its point of view. There is nothing in the tree which

stipulates who should be asking the question unless one adds a Q node, or

perhaps some kind of semantic feature. The decision whether or not to

use point of view to control the movement of the question element becomes

a decision about whether a given tree should generate sentences with more

than one meaning. Let me just note that once the point of view is intro-

duced underlyingly, the question of whether the rule should be cyclic or

post-cyclic becomes academic since movement will apply only once.

2.5.7. Advantages of a Pre-S Q

A desire to introduce point of view information in the underlying

form creates one type of motivation for a Q node. However, it is not

necessary to use point of view interpretation to motivate an underlying

Q node. There are also purely formal considerations, as well as syntactic

ones, which could lead one to postulate an S-initial Q node.

I have been assuming throughout that the WH question particle is

generated underlyingly with the WH word. A second possibility is that

the WH question particle is generated separately as a pre-S Specifier.36

That is, one could postulate that in addition to an S-final COMP, Navajo

also has an S-initial Specifier. If filled by a Q, it could be spelled

out either by a yes/no question particle or a WH question particle. Re-

call that the initial interrogative particle /da'/ is found in yes/no

questions. Having an S-initial Q thus accounts for the placement of that

 205

particle. WH questions never have an initial question particle. However,

the Second Position rule which is needed anyway would insure that the WH

question particle does not end up sentence-initial.

Since, under this hypothesis, the Q node is generated separately

from the WH word, it is possible to have an underlying form with the Q

node spelled out as a WH question particle (e.g.,/la'/) in one S and the

WH word in another S. That is, given such a theory, the following would

be generated as underlying forms:

(137) (138)

9S V ~
la

S
N V

la'

-WH- -WH-

Under this hypothesis, no unbounded Question Particle Movement transforma-

tion is necessary. Type (4) sentences would be generated without any

unbounded movement, requiring only the Second Position rules to move the

/la/ from initial to second position. That is, (137) generates (137a),

and (138) generates (138a), the two acceptable cases in which the WH word

and /la/ are not in the same S.

 206

137a. Jaan la Mary yiskiggo haa'gdd adeesbas nizin ni

John Q Mary tomorrow where.to 1.F.drive 3.want 3.say
Where does John say Mary wants to drive to tomorrow?

138a. Ja'an Mary 1a yisk ago ha'go'6 adeesbas nizin ni

John Mary Q tomorrow where.to 1.F.drive 3.want 3.say
John asked where Mary wants to drive to tomorrow.

Notice that this hypothesis still requires an unbounded leftward

WH-movement transformation. Since the underlying position of the WH

word is determined by grammatical relations, (137b) must be derived from

(137) by unbounded WH-movement, and (138b) from (138) by the same rule.

137b. hddg661d Ja'an Mary yiskqggo adeesbps nizin ni

where.to.Q John Mary tomorrow 1.F.drive 3.want 3.say
Where did John say Mary wants to drive to tomorrow?

138b. Jaan ha'goo'll Mary yiskjago adeesbas nizin ni

John where.to.Q Mary tomorrow 1.F.drive 3.want 3.say
John asked where Mary wants to drive to tomorrow.

Under this hypothesis, the movement is defined by the position of

the Q. That is, WH-movement would now be written as a rule which optional-

ly attracts the WH word to the /1a/, to an S-initial Specifier. While

Navajo still requires an S-final COMP for /-go/ and /- (gI )/, such an

S-initial Specifier would be similar to a COMP node both in position and

some functions. Bresnan has suggested for example that for English the

COMP is marked L+WH], [+WH] being identical to Q (Bresnan, 1972). She

uses one feature to combine relative clauses, questions, and complementizers.

 207

In Navajo, [+WH] would signify an initial COMP, and [-WH], a final COMP.

I know of no real definition of COMP. It has been suggested, however,

that complementizers are to S's what determiners are to NP's. Both are

subsumed under the term Specifier (cf. Pope, 1972).37 One could hypo-

thesize that some of the functions of the English COMP, including indi-

cating direct questions, are performed by other Specifiers in other

languages. Since Navajo determiners, i.e., NP Specifiers, precede NP's,

it is already necessary for the term Specifier to refer to initial as

well as final elements in this one language. 3 8 Thus, while it seems

implausible to hypothesize both initial and final COMP nodes of the same

type, it may be possible to divide up the functions of Specifiers in such

a way as to make an initial S-Specifier plausible even with an S-final

complementizer. Under such a theory, one could then say that all unbound-

ed movement must be to a Specifier, a result which is more consistent with

Bresnan's and Chomsky's proposals that all unbounded movement be either

to or through COMP position.

Notice that once a Q node is introduced, the point of view inter-

pretation can be done either on the underlying or the surface form. If

done on the underlying form, all WH words will be treated as exceptions

to the usual point of view interpretation rule. That is, the point of

view of the WH word will not be determined by the S immediately dominating

the WH word, but by the Q, whose point of view interpretation in turn will

be determined by its immediately dominating S. If the point of view in-

terpretation is done on the surface, then only WH words which do not occur
 208

in the same simplex S as their governing Q's (i.e., type (4) cases) will

be exceptions to the general point of view interpretation rule. In other

words, doing the point of view interpretation on the surface necessarily

distinguishes type (4) cases from the other cases, a distinction which

seems to be necessary.

2.5.8. Problems With an Underlying Pre-S Q

Besides the theoretical issues of whether to use both initial and

final S-Specifiers in the same language and whether to allow Q nodes,

generating an S-initial Q presents other problems. For one, since such

a theory does not generate /l/ and the WH word together, a filter is

needed to eliminate the ill-formed structures: those in which there is

no WH word for the /l/,39 and those in which the WH word is in a higher

S than the /la/.

The existence of an initial Q creates problems in explaining

sentences like the following:

139. hastiin alk'esdisi haild yeini'6

man candy who.Q 3.3.P.give
Who did the man give the candy to?

Under a theory in which the /1 / is generated with the WH word, (139) is

a result of applying neither the WH-movement rule nor the Second Position

rule. If, however, the /1/ is generated initially, a new rule is needed

to attract the /1 / rightward to the WH word. One must consider whether

 209

such a rule is natural, given the existence of the optional rule which

attracts a WH word leftward to a /la/.

One way of arguing for such a rule is to hypothesize that /1a/

is a type of focus marker. Since other particles create foci in Navajo,

including the yes/no particle /-ish/, one could hypothesize that whatever

rule positions the other focus particles also positions the /1/. Such

a theory requires assuming that WH words may be foci and that in WH

questions, the WH word is the only possible focus (since /l// cannot

attach to any other word, except when in neutral second position). To

say that a WH word is a focus is to use the term focus in a new, but not

unfamiliar, fashion. Just as nobody has been able to define for English

the relationship between WH words and foci, I can find no positive evi-

dence in Navajo that WH words are foci. However, such a conclusion is

suggested by the fact that the emphatic particle /ga'/ which creates foci

cannot occur in the same S as a WH word (cf. section 2.4.4).0

Having both a rule which attracts a WH word to a /l/ and a rule

which attracts a /la/ to a WH word still would not explain how the fol-

lowing sentence is generated:

140. Jaan shizhe' ha'lt'i ll neidiyoolnih y"'ni

John 1.father what Q 3.3.F.buy 3.of.3.say
What does John expect my father to buy?

Sentences such as (140) have been insufficiently investigated. From my

preliminary work, it seems that they are grammatical. Since /yo'nr/

 210

(expect) allows direct questions from the speaker's point of view only

(i.e., *"John expects what his father will buy" vs. "John asked what his

father will buy"), (140) could have only one possible interpretation.

Such an interpretation requires (140) to be bracketed with the /l/, and

therefore the WH word, in the matrix S:

so[Ja'an shizh'e' ha' t'ill 1[neidiyoolnih]5 1 yoi]SO

In other words, the /1,/ in (140) would be generated in So, i.e., sentence-

initial position. The WH-movement rule would then attract the WH word to

the /1/,i.e., to sentence-initial position. How then does the WH word

end up as the third NP in the sentence?

If such sentences are in fact grammatical, they present problems

for any WH-movement rule which moves the WH word into initial position.

Recall that I have said that the purpose of the unbounded WH-movement is

not to get the WH word into initial position, but to move it from an S

requiring one point of view interpretation to an S requiring a different

point of view interpretation. Sentences like (140) suggest entertaining

the possibility that WH-movement is not to sentence-initial position at

all. However, I will not pursue this problem here.

2.5.9. Comparing Two Theories

So far as I can tell, the Navajo facts can be handled equally well

whether one postulates an underlying pre-S Q node, generating WH particles

separately from any WH word, or whether one generates the WH word and Q
 211

particle together, postulating an unbounded Question Particle Placement

rule. Both hypotheses require some unbounded leftward movement. The

former hypothesis has the advantage of easily specifying the direction

and extent (i.e., to which S) of the WH-movement for any given sentence.

However, such a theory does not explain why these direct questions are

formed through leftward movement rather than rightward movement such as

exists in the indirect questions discussed in Chapter 1. Such a theory

also cannot explain why type (4) sentences create so much more confusion

than the other types. If anything, this hypothesis suggests that type

(4) sentences should be the most acceptable since they require no movement

transformations. The second theory, on the other hand, makes it easier

to discuss the different preferences people have. Assuming that WH-

movement and Question Particle Placement are one rule, the choice between

types becomes a choice as to whether or not the WH word should be con-

sidered part of the Q in the structural description of the movement rule.

I have claimed that the unbounded WH-movement exists in these cases in

order to move the question word from one point of view segment to another.

Since this theory provides no other way of specifying how far the WH word

and question particle will move, it leaves open the question of whether

the rationale for the movement should be incorporated somehow into the

transformation. While the second theory does not require postulating

either underlying initial, as well as final, Specifiers or Q nodes, it

will require introducing other theoretical devices.

 212

2.5.10 Conclusion

Ignoring all the questionable cases, it is still clear that Navajo

requires some unbounded leftward movement transformation, despite the fact

that it is a verb-final, complementizer-final language. This rule applies

in sentences containing direct discourse verbs, verbs which govern point

of view interpretation. I claim that the unbounded movement originated

in order to move a question word from one point of view segment to another.

The WH question particle signifies that a direct question is being asked.

Who is asking the question depends upon the point of view interpretation

of the S that the particle and (usually) the WH word end up in. Since a

given embedded S is interpreted from the point of view of the immediately

dominating subject, only the matrix S must be from the speaker's point of

view. For the speaker to ask a question, movement to the matrix S is re-

quired, as, therefore,. is unbounded movement.

Point of view interpretation requires unbounded movement, but this

says nothing about the direction of the movement. Whether or not a Pre-S

Specifier is introduced, it is still necessary to explain why this move-

ment is leftward rather than rightward. I will return to this problem in

the Conclusion, using a concept developed in Chapter 3 to provide some

criteria for leftward versus rightward movement.

 213

FOOTNOTES TO CHAPTER 2

1. Reichard (1951) also says that /-sh/ and /- sh/ are the same interro-

gative particle, analyzing the latter as composed of /-sh/ plus the

nominalizer /-I/.

2. Haile uses his own Navajo orthography which I have transcribed.

3. Elgin spells this enclictic without a final glottal stop.

4. Among other things, the vowel in the enclitic is /g'/ while her

"aforementioned" morpheme is /aa/ although she spells it /a/.

5. Normally elements that are focused on do not reduce. Since topicaliz-

ing creates added emphasis, if /-sh/ were a reduced form, one could

account for the ungrammaticality by saying that reduced forms cannot

be used on focused elements. However, I believe that other factors

are at work, and that /-sh/ is not a reduced form anyway.

6. Cf. Chapter 3 for a discussion of /-ish/, focus, and presupposition.

7. I avoid the term neutral because questions with /-ish/ need not be

neutral. Cf. Chapter 3.

8. Note that the semantic hierarchy does not seem to apply to WH words.

Cf. section 2.1.7.

9. In section 2.5.8, I discuss the possibility that /l/ may be a focus

marker when not placed in neutral second position.

 214

10. Although there is no question that these are grammatical, people

disagree about how preferable they are. That is, some have a strong

preference for putting the question particle in second position,

while others prefer to leave it with the WH word. I discuss this

further in my conclusions and speculations, section 2.4.

11. While this sentence is grammatical, it is preferable to use /la/ in-

stead of /-sh/. The difficulty seems to be with the phonology rather

than the syntax. That is, if the noun ends in a vowel, it is diffi-

cult to differentiate the yes/no enclitic /-ish/ from this /-sh/, i.e.,

/ashkiish/ from /ashkilsh/. If the noun ends in a consonant, un-

natural consonant clusters arise, e.g., /Jaansh/. /l// creates

neither of these difficulties because it sounds very different from

the other question particles and because it is a separate syllable.

As a result, some people prefer /la/ to /-sh/, especially when de-

tached from the WH word.

12. I discuss the cases with a crucial variable in section 2.2.

13. This class was part of a summer workshop on the Navajo Reservation,

summer, 1973, sponsored by Dine Bi'olto'Association.

14. Notice that it is possible for the question particle to be placed at

the end of the sentence. It is not clear whether Baker meant to be

describing languages which can or which must place their particles

sentence-finally.
 215

15. Cf. Chapter 3 on /-Ish/, /hanii/ and presupposition.

16. The table ignores the problem of describing where within an S the

WH word and the /1/ end up. I also discuss this below.

17. The point of view given in the table is that permitted by a Strict

Constructionist. Loose Constructionists allow in addition an inter-

pretation from the speaker's point of view in every grammatical

example. Cf. section 2.2.5.

18. The translation is misleading. It is not that it is impossible to

say "What does John want?", but that when want has a direct object,

a prefix appears on the verb,

i. ha'at' ish Ja'an yinizin

what Q John 3.3.want

As explained in the Introduction, direct discourse complements are

not direct objects.

19. I am assuming that these sentences appear in a discourse so that there

is an antecedent for the pronoun. Strictly speaking, no sentence in

Navajo is complete without a subject. One must distinguish, however,

between normal pronoun deletion and missing subjects.

20. Although in principle the scope of the movement is indefinitely large,

it is in fact difficult to construct examples where movement involves

more than three S's. One reason is that the complex sentences are
 216

extremely difficult to process, and judgments vary randomly. Another

reason is that I know of very few direct discourse verbs, and the

sentences become even more difficult if the subjects of the different

verbs are not identical.

21. In section 2.5.7, I discuss the possibility that there may be an

initial COMP-like node--a Specifier--in addition to the S-final COMP.

I am assuming here that COMP refers to the position in which the com-

plementizers appear (/- (gi1)/ and /-go/ in Navajo).

22. As I discuss below, some people can use unbounded WH question particle

movement instead of WH-movement to generate the same meaning. For

those who consider these to be separate rules, WH-movement is obvious-

ly optional.

23. As Reinhart (1973) points out, adverbs are often interpreted different-

ly from the rest of the sentence. A cursory investigation suggests

that,when in the direct discourse segment, an adverb may be interpreted

either from the subject's or the speaker's point of view. Therefore,

when I talk of whole S's, I am referring only to NP's and verbs. Fur-

ther investigation of adverbs is required before any claims can be

made.

24. Another alternative is to reorder the sentences so that the lower S

is not center-embedded. For example

 217

i. yiskkgo kinlanigo deesh a n/ Jaan

tomorrow Flagstaff.to 1.F.go 3.say John
I am going to Flagstaff tomorrow, said John.

I am not certain, however, whether this alternative exists for the

other direct discourse verbs. Thus, (ii) presents problems, both

in deciding whether or not it is grammatical and deciding what it

means if grammatical.

ii. ?*Mary kin;6anigo6 doogaa; yo'n Jaan

Mary Flagstaff.to 3.F.go 3.of.3.say John

Normally it is subjects which extrapose. But, normally the NP

closest to /yo-/ is considered its object. Thus, it is not clear

whether (ii) should mean, "John expects Mary to go to Flagstaff" or

"Mary expects John to go to Flagstaff". See also footnote (34).

25. It is also possible that, although /ha'ati' sh/ has moved from one S

to another, the movement was not to initial position in the higher S

as I am assuming. In sections 2.3.1 and 2.5.8, I discuss examples

which create problems for movement to initial position.

26. Such a rule is necessary only under the assumption that the WH ques-

tion particle is generated with the WH word.

27. These sentences provide another argument that the Q particle is

generated with the WH word, since the Q particle may occur between

the WH word and the verb.

 218

28. In all the examples so far given, the Q particle has in fact been in

the highest S or the one it originated in. I will give examples be-

low which show that it can be in any S.

29. I do not have the facts to describe the loose constructionists'

dialect in greater detail. In particular, I do not know whether

sentences in which the question particle is placed on a word in the

direct discourse segment other than the WH word also receive a direct

discourse interpretation.

30. Some people will allow more than one yes/no question in a simple

sentence. I have not investigated this dialect, but at least in

the cases I know about, using more than one /-ish/ is linked to the

fact that /-ish/ can be used as a focus marker. For discussion of

/-0sh/ and focus and presupposition, cf. Chapter 3.

31. As I discuss below (pp. 187f. ) one can show that (i) is a possible

translation of (98) by the following possible response:

i. ak'aan la' nahTniih bidin

flour some 2.buy 3.2.say
Tell him to buy some flour.

(i) shows that the question can be what to buy and not what John

said he would buy.

32. But cf. Perkins (1973) for some discussion of /doo... da/ as contrasted

with another negative, /hanii/.

 219

33. Another possibility is that WH-movement to initial position originated

by analogy to English. The differences in grammaticality could still

be related to a comparison with simple sentences. Such a theory could

be tested by talking to monolingual speakers of Navajo.

34. It is not clear to me whether (118)-(119) are instances of embedded

direct discourse, equivalent to English indirect discourse, or cases

of quotations. Since Navajo does allow extraposition of subjects, it

is reasonable to assume the former is at least one possible interpre-

tation. However, if these are direct discourse examples, I have no

explanation for why examples with /yo'n / are ungrammatical, as in

the following (b) sentences:

ia. Jaan kinlnigo'' Mary adoolbas yo"n

John Flagstaff.to Mary 3.F.drive 3.3.say
John expects Mary to drive to Flagstaff.

b. *kinlanigo'o' Mary adoo-lbas yo'n Jaan

Flagstaff.to Mary 3.F.drive 3.3.say John

iia. Jaan hail Bil yidiyoolhe(a y 'n

John who.Q Bill 3.3.F.kill 3.3.say
Who does John expect will kill Bill?

b. *h la Bil yidiyoo-lhe' yo'n Jaan

who.Q Bill 3.3.F.kill 3.3.say John

More work needs to be done on extraposition of subjects before one

can decide whether these sentences are ungrammatical because /yo'n /

 220

does not allow quotations or are ungrammatical for some other reason.

35. However, the order of the WH word may affect the meaning. That is,

while (128) and (129) are identical, (i) may be slightly different.

i. halila' Ja'an biia-ts

who.Q John 3.3.P.see
Who was it that John saw?

It is not clear whether who should be considered some kind of focus,

or topic, or what. This is an area needing more investigation.

36. This possibility was first suggested to me by Kenneth Hale.

37. The term Specifier has been used in an even broader sense. Chomsky

(1968) refers to "the phrase associated with N, K, V in the base

structure as the 'specifier, of these elements" (27). Thus, [Spec, 5]

will be analyzed as the determiner, similar to my suggestion. In ad-

dition, [Spec, V] would be analyzed as the auxiliary and [Spec, K]

"perhaps as the system of qualifying elements associated with adjec-

tive phrases" (ibid.). It may be that this broader notion is the

more useful one.

38. Hale (forthcoming B) suggests that the final /-I/ complementizer

originates as an initial determiner. Under that assumption, the use

of a Specifier node would characterize the connection between Navajo

determiners and complementizers.

 221

39. So far as I can tell, multiple WH questions are impossible in Navajo.

Therefore, the filter would have to throw out any sentence with more

than one WH question particle. However, any theory would have to

eliminate these cases.

40. I am assuming here that emphatic particles are analogous to intona-

tion centers in English. Chomsky (1969) defines focus as follows:

"The focus is a phrase containing the intonation center; the presup-

position, an expression derived by replacing the focus by a variable"

(26). While the intent of this definition seems correct for most

uses of Navajo emphatic particles (as well as the yes/no particle

/-ish/), there are cases where a focus in an embedded clause may

actually be part of the presupposition (cf. Chapter 3).

 222

CHAPTER 3

THE SEMANTICS OF THE COMPLEMENTIZERS

3.0. Introduction

We have now examined two types of Navajo subordinate clauses and the

questions that are formed when there is a WH word embedded in each type.

With the /-1/ complementizer we found indirect questions, formed through

deletion of the WH word and unbounded rightward movement of the spatial

enclitic. In direct discourse S's, we found that the interrogative particle,

and optionally the WH word, is moved by an unbounded leftward movement

transformation. I postulated that this is the universal question formation

transformation and that it has been applied in this particular case to re-

move the question word from within the scope of the direct discourse. It

is due to the unbounded leftward movement that direct questions are pos-

sible with direct discourse verbs.

Let us now turn to WH questions embedded under the third type of

Navajo subordinator, the /-go/ complementizer. In this case, neither

deletion nor movement is required to produce direct questions.1 This

then represents the form of question expected in verb-final, COMP-final

languages. Using this type for contrast, I will generate a rationale for

the form of question found embedded under the /- / complementizer. Al-

though the form of the /- / questions and direct discourse questions is

totally different, I will claim that the rationale behind both atypical
 223

forms is the same: some rule is needed to get the question word out of

the scope of an element which gives the question word an undesired or con-

tradictory semantic interpretation. Speculations about why the different

types of transformations will be postponed until the Conclusion. This is

because I believe the question formation facts are related to the nature

of the complementizers. Therefore, before turning to questions embedded

under /-go/, I will compare /-go/ and /-I/, demonstrating that although

their most common uses are very different, there are cases where they

appear in superficially identical clauses. I will argue that, at least

in some cases, their structures are in fact identical. However, although

the clauses may be structurally identical, their meanings are not. I

attribute the meaning difference to the complementizers themselves, and

will then use that difference to explain the different forms of embedded

questions. In this chapter, the semantic differences are based primarily

on an examination of sentential subjects and objects, although the most

natural use for /-go/ is as an adverbial subordinator. Therefore, be-

fore turning to a contrast between /-go/ and /- /, I will briefly describe

adverbial clauses to help explain my claims about 7-go/.

3.1. /-go/ and Adverbs

3.1.1. Overview

In the Introduction I stated that the /-go/ complementizer has

always been described as an adverbial subordinator (0.4.4). Although I

will not detail how adverbial clauses get their meanings, it will be useful
 224

to examine somewhat superficially this most common use of the /-go/ com-

plementizer.2 Such an examination will both help to explain why people

have always made the assumption that /-go/ is an adverbial marker, as well

as help justify my claim that the same morpheme may be found as a nominal

subordinator. Consider the following sentences, examples of the most

natural uses of /-go/:

1. shizhe'e niyaago daIdiidjfl

1.father 3.P.comeCOMP pl.1.F.eat
When my father comes, we'll eat.

2. Mary shaan/yd/ago Jaan bil h6zh6

Mary 1.3.P.comeCOMP John 3.with 3.be well
Because Mary came to see me, John is happy.

3. Aff' sinflohgo shizhe'6 neidiyoolnih

horse 2.P.ropeCOMP 1.father 3.3.F.buy
If you rope a horse, my father will buy it.

4. Jaan chidi nayiisnii'go bil hozhpogo shaanay

John car 3.3.P.buyCOMP 3.with 3.be wellCOMP 1.3.P.came
Because John had bought a car, he was happy when he came to see me.

Navajo does not always distinguish temporal, conditional, and

resultative subordinate clauses. Thus, sentence (3) can have another

meaning. It also translates as "When you rope a horse, my father will

buy it". Notice that there is no word in any of these examples that

designates the type of condition which the subordinate clause is intended

to communicate. Sometimes a sentence does not seem ambiguous because one

reading is so much more natural than the other. Thus, example (1) could
 225

be translated, "If my father comes, we'll eat". Since such a dependency

is somewhat odd, one tends to give the sentence a unique translation.

In a similar fashion, tense and aspect relationships tend to disambiguate

a sentence. If sentence (2) were to have a "when" reading, it would have

to mean

2. *John is in a good mood when Mary came to see me.

If, however, the matrix clause is put in the perfective, the sentence does

in fact have a "when" reading:

5. Mary shaaniyaago Jdan bil h o'zhY 2 d

Mary 1.3.P.comeCOMP John 3.with 3.P.be well
When Mary came to see me, John was in a good mood.

Since the only difference between (2) and (5) is the aspect relationship

between the two clauses, it is clear that aspect is one of the factors

which determines the relationship between the two clauses. Since /-go/
appears in both sentences, it is also clear that /-go/ itself does not

differentiate among "if", "when", and "because" clauses. It is now clear

that there is no single word in the above examples which performs this

function.

As further proof, consider the following example in which such a

word does exist. This example is no longer ambiguous.

6. Mary shaaniya'ago biniinaa Jian bil h6'zhg 2 d

Mary 1.3.P.comeCOMP because John 3.with 3.P.be well
Because Mary came to see me, John was happy.
*When Mary came to see me, John was happy.
 226

While I will not discuss /biniinaa/ in detail, it is sufficient to note

that example (6) makes it clear that Navajo possesses more specific ways

to designate the specific semantic relationship between a dependent ad-

verbial clause and its independent clause than are found in examples(1)-

(4). Morphemes such as /biniinaa/ have definite semantic content. That

content provides a unique meaning for the sentence, one that requires such

clauses to be adverbial. It is thus not clear what function /-go/ performs.

In such clauses, /-go/ seems merely to be a syntactic subordinator. How-

ever, when there is no other word to designate the adverbial relationship,

/-go/ is the only morpheme available to perform that function. It is

clear, however, that /-go/ does not mark any specific relationship since

it is used to replace when, if and because. Therefore, if /-go/ has any

semantic content, it must be some abstract concept consistent with all

three meanings, a concept I will refer to as adverbial since it occurs

in adverbial clauses and define simply as the semantic common denominator

of such words as if, when, and because. Adverbial is being used here as

a heuristic device, one needing no further clarification since I will

argue that /-go/ in fact has no semantic content at all. What is important

to note is that there are two ways in which /-go/ could be an adverbial

marker. One is to contain some semantic content which requires it to be

adverbial. The second is to be a syntactic marker that appears in ad-

verbial clauses. I will consider both possibilities and show that neither

is in fact possible.
 227

3.1.2. Possible Semantic Content For /-go!

The morpheme /-go/, in addition to being a complementizer, is an

affix added to adjectives and verbs to form adverbs, similar to -lz in

English. Consideration of some simple adverbs shows how impossible it is

to find any semantic content for /-go/. Consider the following pairs with

respect to the function of /-go/:

7a. nizhon 7b. nizhonigo

good,nice well

8a. tlf 'ed' I8b. tl'leigo

dark, night when it's dark, at night

9a. yisk 9b. yisklggo

it has dawned when it has dawned, tomorrow

Given these examples, /-go/ could still be a semantic adverbial marker.

However, it is now even less clear what the content of adverbial would

be. In examples (8)-(9), /-go/ could be said to be doing more than just

changing the part of speech. The English dark, darkly or night, nightly

is not analogous to /tl'e''/, /tl'"elgo/. In examples (8)-(9), /-go/

could perhaps be said to contain the notion when, at least as much as it

did in example (1). However, it is difficult to see how anything from

the notions if, when, or because has been added with /-go/ to (7b). In

this case, /-go/ merely signifies a change in the part of speech. Thus,

the English good, well in (7) is parallel to the Navajo /nizhon /,

/nizhonigo/.
 228

It thus seems that if /-go/ has any semantic content at all, the

notion adverbial must be limited to whatever semantic content, if any,

there is in switching a part of speech. Even to claim that adverbial could

be defined as the common denominator of if, when, and because is too speci-

fic. So far as I am aware, there is no evidence that changing a part of

speech involves a semantic change. I therefore will assume that if there

is anything about /-go/ which is inherently adverbial, it must be syntactic.

3.1.3. Possible Syntactic Content For /-go!

The second way /-go/ could be an adverbial marker is to be a syn-

tactic marker. If the /-go/ marker appeared in only one syntactic con-

figuration, it could be considered a syntactic marker of that configuration.

As such it need have no semantic content to be an adverbial marker.

While I have stated that examples (1)-(4) contained adverbial

clauses, the only evidence given was the translations. To see that they

really are adverbial clauses, notice that in the above examples, the

matrix clause is independent of the /-go/ clause. The /-go/ clause is

not acting as a subject or object of the matrix S. Thus, if the /-go/

clause were deleted, there would still be a full, i.e., non-elliptical,

sentence. Compare the following with examples (1)-(4):

1' da'Idiidil
We will eat.

2'. Jaan bil hozho

John is happy.
 229

3'. shizhe'e neidiyoolnih

My father will buy it.

4'. shaanlya
He came to see me.

The two clauses in examples (1)-(4) interact to the extent of pro-

nominalizing identical nouns. Pronouns generally delete in Navajo. Thus,

(3) has a deletion site where horse would be, and (4) one where John would

be. It is clear from the meaning, however, that it is the noun and not the

adverbial clause which has pronominalized.

Recall that the /- / complementizer can only mark NP's.3 It cannot,

for example, replace /-go/ in these adverbial clauses. Contrast the fol-

lowing with (1) and (3) above:

10. *shizhe'le niyahigii da'diid 1

1.father 3.P.comeCOMP pl.1.F.eat

11.e'? /
11. *:1'L-L sinlohigi shizh(e'6 neidiyoolnih 4
horse 2.P.ropeCOMP 1.father 3.3.F.buy

Given the information so far presented, we can hypothesize that

sentences with /-go/ clauses have the following structural configuration:

A V S

rS
 230

I am using the node ADV to depict the claim that these clauses are adverbi-

als and that they are subordinate clauses, i.e., that the /-go/ comple-

mentizer is subordinating its clause with respect to the main clause.

While I have no intention of motivating this particular node, I believe

that the claim that /-go/ clauses are syntactically subordinate is suffi-

ciently important to be depicted graphically. However, since I will

include the notion "subordinate" in my eventual definition of /-go/,

there is no a priori reason why an underlying form must contain any sub-

ordinating node. 5

Among the questions to be considered in this chapter are whether

/-go/ clauses are found only in such a configuration and how its syntax

bears on its semantic content. Further discussion of the semantic content

of /-go/ must await further discussion of its syntactic behavior, especially

in comparison to the /-I/ complementizer.

3.2. Are /-go/ and /-i/ Clauses Always Structurally Differentiated?

3.2.1. Seemingly Anomolous /-go/ Sentences

So far all the examples given have shown that /-go/ and /- / are

both structurally and semantically distinct: the former marks subordinate

adverbial S's; the latter marks sentential NP's. The following, however,

are sentence pairs which have identical translations:

12a. ch'e''Itiing'6 naanlego shil beehdzin

entrance.toward 3.playCOMP 1.with 3.be known
I know that he is playing in the yard.
 231

12b. ch'e''e'tiing 6 naanehigi shil b e'h6zin

entrance.toward 3.playCOMP 1.with 3.be known
I know that he is playing in the yard.

13a. Mary kinanigo'6' iyaago yishniih

Mary Flagstaff.to 3.P.goCOMP 1.hear
I hear that Mary went to Flagstaff.

b. Mary kinaniago6 iiyahigii yishniih

Mary Flagstaff.to 3.P.goCOMP 1.hear
I hear that Mary went to Flagstaff.

14a. Mary kinla'nigo' na'asbfazgo Bil bil ya'at' eh

Mary Flagstaff.to back.3.P.driveCOMP Bill 3.with 3.be good
Bill is happy that Mary drove to Flagstaff.

b. Mary kinlanig na'asbfzigi Bil bil y 'at

'eeh
Mary Flagstaff.to back.3.P.driveCOMP Bill 3.with 3.be good
Bill is happy that Mary drove to Flagstaff.

15a. Mary Bil yich'i' ylti'go baa akonisin

Mary Bill 3.to 3.3.talkCOMP 3.about 1.be aware
I am aware that Mary is talking to Bill.

b. Mary Bil yich'i' yalti'igi baa a'konisin

Mary Bill 3.to 3.3.talkCOMP 3.about 1.be aware
I am aware that Mary is talking to Bill.

How accurate are these translations? Is there something missing in

the English translation? Are these /-go/ clauses in fact structurally

identical to the /-I/ clauses? In other words, can the /-go/ complementizer

mark sentential NP's? And, if so, are /-go/ and /- / in those cases seman-

tically identical?
 232

3.2.2. The Structure of the /-go/ Clauses

I am going to argue that /-go/ clauses may in fact be sentential

subjects and objects exactly as the above English translations suggest.

I will use two general types of argumentation. The first is based on the

assumptions that there must always be a subject, and a subject is by

definition an NP. If there is nothing else which can be the subject,

then the /-go/ clause must be. However, since pronouns generally delete

in Navajo, a surface subject is not always present in well-formed sentences.

It is thus necessary to show that a subject could not have been deleted by

pronominalization in order to prove that there is no other possible subject.

The important thing to notice is that, after pronominalization deletion,

a sentence is still complete. While the specific reference may be lost,

the pronoun itself is recoverable through agreement markers in the verb.

In contrast, an elliptical sentence, while not completely ungrammatical,

is an incomplete sentence, more like a phrase. A missing subject or ob-

ject is not recoverable through other information in that sentence. Con-

text alone provides the missing subject or object.

The second type of argument is also based on morphology. If there

is an agreement marker requiring an NP and there is nothing else which

could have triggered the agreement, then the /-go/ clause must have done

so, and the /-go/ clause must therefore be an NP.

Notice that neither type of argument proves that the /-go/ clause

must be an NP in underlying structure, or, more specifically, a subject

or an object. One possibility is that there are subjectless underlying

 233

structures and all /-go/ clauses are dependent adverbial clauses in the

underlying structure. A transformation may then attract the adverbial

clause into subject or object position.

Arguing from agreement markers also does not conclusively prove

the structure of the underlying phrase marker. Agreement could obviously

take place after the transformation which attracted a subordinate clause

into subject or object position.

In the following sections I will be comparing /-go/ and /-I/

clauses and arguing that the differences that are found are not due to

structural differences. While I cannot yet motivate an underlying struc-

ture, I believe it will be sufficient to show that /-go/ and /-f/ clauses

are structurally identical at some point in the derivation.

3.2.3. What Is the Subject?

Consider the sentences in (12) above. I have already shown that

the /-i/ clause in (12b) is a sentential subject (see Chapter 1). If the

/-go/ clause in (12a) is not, analogously, also a subject, then what is

the subject? Recall that /-shil/ is a postposition meaning with me. It

is therefore clear that it is not the surface subject. The only other

word appearing in (12a) is the verb. Thus, if the /-go/ clause is not

the subject, then the subject must have been deleted. Since pronouns

generally delete in Navajo, it is reasonable to hypothesize that the

subject has been deleted. We can test this hypothesis by deleting the

/-go/ clause and seeing if we still have a complete sentence. Recall

 234

that when adverbial /-go/ clauses are left out, the sentence remains well-

formed, i.e., non-elliptical, even when the subject is a pronoun which has

been deleted (cf. examples (1)-(4)). The following example demonstrates

that the hypothesis that a subject has been deleted by pronominalization

is incorrect.

16a. *shil bedhzin

1.with 3.be known

Sentence (16a) can be used as an elliptical phrase. In that sense

the /*/is misleading. For our purposes, however, it is not a complete

sentence. It is interpreted as meaning

16b. e shil b eeho'zin

that 1.with 3.be known
I know about that.

In other words, it is interpreted as a phrase without a subject.

The incompleteness of (16a) argues that (12a) must have a subject.

Since there is nothing else in (12a) which could be the subject, nor is

the subject a deleted pronoun, the /-go/ clause must be the subject. The

structure of (12a) therefore seems to be identical to (12b).

3.2.4. Morphological Evidence For /-go/ Clauses As Subjects

Recall that Navajo verbs may have overt subject markers. One

example is the locative marker /ho-/. Thus we find the following:

 235

17. shil ya' hootte'h

1.with 3.loc.be good
I like it at that place.

18. shil ho'zho

1.with 3.loc.be well
I am happy.

Contrast these with the verbs we are considering which do not have

the /ho-/ subject markers:

19. *shi;L yI'at'eh

1.with 3.be good

20. *shi;L nizhdn

1.with 3.be nice

Once again, the incompleteness of (19) and (20) demonstrates that there

is no deleted pronominal subject. If we add a subject, the postposition-

verb combinations become grammatical.

21. i shil ya't'eh

that 1.with 3.be good
I like that.

It must therefore be the lack of a subject that is causing the ungrammati-

cality or strong feeling of incompleteness.

Now contrast the translations of /-go/ clauses in the following

pairs of sentences which differ only in the presence or absence of the

locative subject marker /ho-/:

 236

22a. Bil kin-lanidi naalnishgo Mary bil ya 't'4'h

Bill Flagstaff.at 3.workCOMP Mary 3.with 3.be good
Mary likes it that Bill works in Flagstaff.

b. Bil kinlanidi naalnishgo Mary bil ya' hoot'te6h

Bill Flagstaff.at 3.workCOMP Mary 3.with 3.loc.be good
Because Bill works in Flagstaff, Mary likes it there.
*Mary likes it that Bill works in Flagstaff.

23a. Ja'an shaarayaago shil nizhoni

John 1.3.P.comeCOMP 1.with 3.be nice
I'm glad that John came to see me.

b. Jaan shaaniyaago shil hdzh6

John 1.3.P.comeCOMP 1.with 3.locibe well

Because John came to see me, I'm happy.
*I'm happy that John came to see me.

With the verbs that have no other subject, the /-go/ clause is translated

as an integral part of the sentence, as a subject. If there is another

subject, the /-go/ gets its normal adverbial translation. Again, it seems

that we must postulate two functions for /-go/, as a sentential NP marker

as well as an adverbial marker.

Recall that /-i/ cannot be an adverbial marker. We can reinforce

the above conclusion by replacing /-go/ with /-I/ in the above examples.

As we expect, the sentences are good only in those cases in which /-go/
clauses have been postulated to be NP's.

24a. Bil kin-1nidi naalnishig I Mary bil ya'at'66eh

Mary is happy that Bill works in Flagstaff.

b. *Bil kinlanidi naalnishigi Mary bil y' Ihoot' h

 237

25a. Ja'an shaaniya'hfgif shi;L nizhoni

I'm glad that John came to see me.

b. *Ja'an shaana{ya-hig i shil hdzh6

3.2.5. /-go/ Clauses As Objects

My arguments that /-go/ clauses can be objects are based on

morphological evidence. I will use the verb and the negative to show that

/-go/ clauses can be an integral part of the sentence as well as separate

adverbial clauses.

The most obvious cases are those with verb agreement. /yi-/ and

its variant /bi-/ appear in the verb only when there is a direct object

with a specific reference. Consider the following:

26. Mary naldzid

Mary 3.afraid
Mary is afraid.

27. Mary Ja'an yinaldzid

Mary John 3.3.afraid
Mary is afraid of John.

28. *Mary Ja'an naldzid

Mary John 3.afraid

Sentences (26)-(28) demonstrate that the /yi-/ on /naldzid/ is in fact the

direct object marker. /yi-/ occurs instead of /bi-/ because the subject

is third person.6 /naldzid/ also marks objects when the subject is non-

third person; in this case, with the prefix /bi-/.

 238

29. nisdzid
I am afraid.

30. Ja'an bina'sdzid

I am afraid of John.

It is thus perfectly clear that if /bi-/-/yi-/ occur, there must

be an object. Now consider the following:

31. Mary na'albpsgo yina/ldzid

Mary 3.driveCOMP 3.3.afraid
Mary is afraid of driving.

32. na'asbgsgo binasdzid

1.driveCOMP 3.1.afraid
I am afraid of driving.

We know from the morphology that there is an object. From the translation,

we see that the /-go/ clause must be that object. Further proof that it

is in fact the /-go/ clause which is the object is provided by the fact

that (32) can be an answer to the question

33. ha' at ' sh binaninldzid

what Q 3.2.afraid
What are you afraid of?

I will conclude my argument that /-go/ clauses are NP's with the

following sentence:

34. doo la' hone'e nijighaago diists'a'da

neg someone inside 4.walk aroundCOMP 1.hear neg
I don't hear anyone walking around inside.
 239

Notice first the position of the negative. The two-part negative surrounds

both the /-go/ clause and the matrix verb. While it is true that /doo/ gets

placed to enclose the entire scope of the negation, Navajo does not permit

the two parts of the negative (/doo/ and /da/) to be placed in different

S's (cf. section 2.4.5). The position of the negative therefore provides

more evidence that the /-go/ clause is not a separate subordinate clause.

Since adverbials are separate clauses, the negative demonstrates that /-go/

clauses are not always adverbials, but may be NP's. The use of the nega-

tive is therefore consistent with my claim that the /-go/ clause is the

object in (34).

Another approach to this sentence would be to claim that /la'/

(someone) and not the whole /-go/ clause is in fact the object of hear.

The sentence would then translate more like "It's not the case that I hear

someone (when he's)walking around inside." Consider the following:

35. doo la' hodiists'a'da

neg someone 4.1.hear neg
I don't hear anyone.

36. *doo la' diists'a'da

neg someone 1.hear neg

7
/,a'/ requires a so-called fourth person marker. Since there is none in

(35), the /-go/ clause and not /la'/ must be the object of /diists'a/.

I have argued so extensively for /-go/ clauses as both NP's and

adverbs for two reasons. First, there seems to be a strong feeling on

the part of Navajo speakers that /-go/ clauses are always adverbial. When
 240

presented with a sentence containing a /-go/ clause, the Navajo speakers

I have consulted first try to translate it as a subordinate adverbial

clause. It is only when such an attempt fails that they will translate

a /-go/ clause as an NP. Second, in section 3.1.2, I discussed the ques-

tion of whether /-go/ has any semantic content, a content somehow related

to its use as an adverbial marker. If, however, /-go/ is also a nominal

marker, then it is necessary to give up the idea that it has any semantic

content. I will return in section 3.3.6 to the semantics of /-go/. I

have so far argued that there is an area in which /-go/ and /- / overlap.

The next step is to differentiate the two within that area. Doing so

will, I believe, clarify the function of /-go/ and therefore the problem

of its semantics.

3.3. Are /-go/ and /- / Clauses Semantically Identical?

3.3.1. Differentiating /-go/ and /-'/

While I have just argued in detail that /-go/ can be an NP comple-

mentizer, I have not yet looked closely at the semantics of /-go/ when it

is an NP complementizer. To demonstrate that /-go/ and /- / can mark

structurally identical clauses, I considered examples in which /-go/ and

-i/ seem to be interchangeable. It is precisely this identity, however,

which makes it difficult to distinguish the two complementizers semantical-

ly. Thus, I will now contrast sentences in which the matrix verbs permit

only /-go/ on sentential NP's with ones which permit only /-1/. The dif-

ferences in the matrix verbs provide clues to the differences in the

 241

complementizers. Consider the following:

37a. Mary na'albgesgo yinaldzid

Mary 3.drive COMP 3.3.afraid
Mary is afraid of driving.

b. *Mary na'albasigif yinaldzid8

Mary 3.drive COMP 3.3.afraid

38a. shil4' altso dindoneelgo dzaa

1.horses all 3.die COMP 1.P.imagine
I imagined that all my livestock would die.

b. *shilf altso dfndon41g a'dzaa

1.horses all 3.die COMP 1.imagine
I imagined the fact that all of my livestock would die.

39a. adad0' naho'4tanigfi chtlini'

yesterday 3.P.rain COMP 1.P.bring out the fact
I brought out the fact that it rained yesterday.

b. *ad C, Ld dL' nah6 it ago chtfin/1(

L.

yesterday 3.P.rain COMP 1.P.bring out the fact

40a. Bil Jaan Mary yiyiiltsa'nig bi'diit'a

Bill John Mary 3.3.P.saw COMP 3.bother
It bothers Bill that John saw Mary.

b. *Bil Jaan Mary yiyiilts/pgo bi'diit 'l

Bill John Mary 3.3.P.saw COMP 1.bother

41a. ada/dd4' Ja'an Mary yich'i' haadzi'igli ta a akdneehee

yesterday John Mary 3.to 3.3.P.talkCOMP just 3.be thus
It's good that John talked to Mary yesterday.
 242

b. *ad 4d-4' Jaan Mary yich'i' haadzii'go t'aIka akneehee

yesterday John Mary 3.to 3.3.P.talkCOMP just 3.be thus

From the above examples, we can hypothesize as a first approximation

that the sentences with /-i/ presuppose the truth of the complement of
9
/-i/. In contrast, the /-go/ complements require no such presupposition.

In (37b), for example, Mary is not afraid of the fact of her driving. One

cannot fear a fact, only the driving. Nor as we see in (38b) can one

imagine facts. On the other hand, one can point out a fact, consider a

fact to be good, or be bothered by a fact, as in (39)-(41). The following

provides further evidence that the complements of /-// but not /-go/ must

be presupposed:

42. shile'echaa'
LLC
nahal'ingo shil t' e 'chidi yilwo- 1
1.dog 3.barkCOMP 1.with 1.imagine pst car 3.go by it happened
I imagined that my dog was barking, but it turned out that a car was
going by.

43. shill' altso dinoonelego baa niseyeel, doo t 'Ia aanif

1.horses all 3.die.COMP 3.about 3.1.P.dream and just true

a kh66t' iid ad jdd'

thus.3.P.happen yesterday
I dreamt that all my horses would die, and yesterday that truly happened.

44.*naho' 1t nig ch' ini'a, ko ndi hane' gif yooch' fd et'e

3.P.rain COMP 1.P.bring out the fact but tell REL lie 3.be
I brought out the fact that it had rained, but what was said is a lie.

45.*naholtanigii ch'ini'a', do hane' gi t'a'a' aan

3.P.rain COMP 1.P.bring out the fact and tell REL just true
I brought out the fact that it had rained, and what was said is true.
 243

It is appropriate to deny what is imagined. It is also fine to assert

that what was dreamt is in fact true. However, it is contradictory to

deny what has already been presupposed to be true. And it is inappro-

priate to assert that what is presupposed is true because it is redundant

to assert what is already assumed. Further explication thus confirms the

hypothesis that sentences with /-I/ involve presuppositions about the

complement.

3.3.2. The Semantics of /- /

It is too strong to claim that sentences with /-i/ always presuppose

the truth of their complement. Consider the following:

46. Jaan din nilinigii yooch'iid &t'

John Navajo 3.be.COMP lie 3.be
That John is a Navajo is a lie.

The translation of sentence (46) suggests that in fact there is no pre-

supposition involved at all. However, it is the English translation that

is at fault. As we know, surface that has numerous meanings. Thus, one

could substitute "the fact that", "the story that", "the idea that", etc.

A more accurate translation of (46) would be, "the idea.that John is a

Navajo is a lie", or, better still, "this business about John's being a

Navajo is a lie". So far as I know, there are no Navajo nouns for "fact",

"idea", "business" to clarify the meaning. Notice, however, that the

semantics of "is a lie" requires referring back to a proposition. Con-

sider the following English sentences:

 244

47. *John's being a Navajo is a lie.

48. The claim that John is a Navajo is a lie.

In other words, the speaker is not assuming the truth of the pro-

position, but is assuming that there is a proposition under discussion.

What is being assumed is a contextual reference. My Navajo consultant

has said, for example, that (46) could be taken as an elided form of (49).

49. Jaan dine nil" ha 'n"n g yooch'iid at'e

John Navajo 3.be story.3.sayCOMP lie 3.be
The story about John's being a Navajo is a lie.

Sentence (49) has no direct English translation. It signifies that the

content of what was said (John is a Navajo) is a lie. What we take as

given is that something was said. /-// in (46) specifies that the pro-

position that John is a Navajo has previously been a part of the discourse.

On the basis of this added data, we must modify our hypothesis

about complements of /-1/ to say that they presuppose the existence of

either a fact or a previously discussed proposition. Notice that this

characterization of the possible presupposition involves two logically

distinct notions, factivity and discourse reference. There is no a priori

reason why a language should combine the two. In fact, the two factors

are normally considered separately. There is thus no generally accepted

term for the union of the two. However, others have also argued that the

concepts need to be combined. Pope (1972), for example, argues that

definite NP's and factive S's (as well as generic NP's) share the property
 245

she designates as definiteness and represents as a TH marker generated in

specifiers, i.e., determiners of NP's and complementizers of S's. The

crucial semantic property shared by TH categories is that of being appro-

priate only when anaphoric (p. 7). The definition of anaphoric used by

Pope is the broader than usual one first introduced by Kuno (1972): any

NP whose entry in the discourse registry is uniquely determined is ana-

phoric. Unlike the more common definition, Kuno's concept of anaphora

includes elements not previously mentioned in the discourse so long as

they are known or familiar to both speaker and hearer. Within Pope's

framework, the crucial factor about elements within the complement of

/-I/ is not that they are presupposed true, but that they are presupposed

identifiable.

Although further refinement of the relationship between presupposi-

tion and /-"/ is necessary, the problems become clearer when the referen-
tial hypothesis is examined through the functional notion of semantic

subordination. In turning to this new perspective, I propose the follow-

ing working version of the Referential Hypothesis:

Complements of /-I/ always presuppose the existence of either

a fact or a previously discussed proposition. In other words,

whatever is in the scope of /-I/ must be anaphoric.

The term scope of L/ is defined as follows:

If /-I'/ commands X asymmetrically, then X is in the scope of /-4/.10

 246

As I will explain in section 3.5.5, the spatial enclitics which

attach to /-1/ (as discussed in Chapter 1) must not be included in the

scope of /-i/. There are two ways to describe this fact. One is to say

that X is in the scope of /-i/ only if /-i/ commands X and X is to the

left of /- /. To avoid using linear order in addition to the concept of

command, I have given the definition in terms of the second alternative.

This alternative relies on the condition that X may not command /-I/.
Notice, however, that under the assumption that the /-i/ and the spatial

enclitic form one constituent, it is possible to use this condition only

if /- / (i.e., the COMP) is Chomsky-adjoined to the S node. That is,

given a tree like (A), everything in S will command /-K/.

(A)
S0

COMP

Yet it is precisely the material in S which is referential and therefore

must be included in the scope of /-i/. The tree in (B) solves this

problem.
 247

(B) S

COMP

Since the COMP node is Chomsky-adjoined to S2, COMP commands S2 , but nothing

in S2 commands the COMP. Thus, as stated, the definition of the scope of

/-i/ implies that the structure in (B) exists for Navajo.

In contrast to the Referential Hypothesis required by /- /, the /-go/

complementizer requires no presuppositions. Notice that a contrast between

-1/ and /-go/ has brought forth requirements for /-i/ clauses. So far,

however, the uses of /-go/ have not been defined any better. Although the

ungrammatical sentences with /-go/ in this section suggest that /-go/ pre-

cludes any presuppositions, we have already seen examples which show that

this statement is too strong. In (12a), for instance, /-go/ occurs with

/-;L belhozin/ (know). I return to this issue in 3.3.5. For the present,

however, notice that /-go/ is still only a semantically empty syntactic

subordinator.
 248

3.3.3. /-1/ and Semantic Subordination

Another way to understand the semantics of /- / is to view this

complementizer as specifying that everything within its scope will be a

semantically subordinate part of an entire sentence. I am using the term

"semantically subordinate" as employed in Erteschik (1972). In that work,

Erteschik contrasts this notion with "semantically dominant". An S is

semantically dominant "if it is not presupposed and does not have contextu-

al reference" (p. 22). An embedded S is semantically subordinate if it is

presupposed or has contextual reference. (By definition, no simple sentence

can be subordinate.) Intuitively, it is the dominant and not the subordi-

nate part of the sentence which it is natural to comment on. An S is

semantically subordinate because its content is already known and is

therefore less important in terms of the message being communicated.

Notice that the two criteria for semantic subordination--contextual

reference and presupposition--are precisely the characteristics required

of everything within the scope of /- /. Erteschik has argued that these

two characteristics should be subsumed under one concept because they

affect the possible syntactic forms for a given underlying structure.

Her dominance condition on extraction states that "extraction can occur

only out of clauses or phrases which can be considered dominant in some

context" (p. 27). I will argue that the dominance/subordination contrast

characterizes the difference in meaning between /-go/ and /-I/ clauses in

Navajo and also affects syntactic processes in that language, among them

question formation.
 249

While, according to Erteschik's theory, syntactically subordinate

clauses need not be semantically subordinate, /-i/ clauses will always be

semantically subordinate because of the semantics of the complementizer.

In contrast, /-go/ clauses may be semantically dominant. This is because

that complementizer is semantically empty and therefore requires no pre-

suppositions or contextual reference. It may therefore be possible to

emphasize either the /-go/ clause or the matrix clause. Erteschik has

devised some tests to determine when an S is semantically dominant. They

are based on "the principle that it is only the dominant part of the

sentence which it is natural to comment on" (p. 16). One test says that

if an S can be denied, that S must be semantically dominant. Erteschik

implies, and I assume, that if a given test is appropriate to a sentence

type and a particular clause fails that test, then that clause cannot be

semantically dominant. The following examples therefore demonstrate that

my claims about /-go/ and /-'/ and dominance are correct:

50. Bil Jsan dine' nilini yinishdl4 n'igo doo akot'e'eda

Bill John Navajo 3.be.COMP 1.believe 3.sayCOMP neg thus.be.neg
It's not right that Bill said he believes that John is a Navajo.

(a) Bil e' doo hwididooniilda

Bill that neg thus.say.neg
Bill wouldn't say that.

(b) #Jan yee' doo dine nilnda

John emph neg Navajo 3.be.neg
John isn't a Navajo.
 250

51. Bil Ja'an dine' nil{'go yinishdla nigo doo kot 'e'eda
Bill John Navajo 3.beCOMP 1.believe 3.say.COMP neg thus.be.neg
It is not right that Bill said he believes that John is a Navajo.

4
(a) Bil e doo hwididooniilda
Bill that neg thus.3.say.neg
Bill wouldn't say that.

(b) Jaan yee' doo dine nilida

John emph neg Navajo 3.be.neg
John isn't a Navajo.

The fact that the /-i/ clause in example (50) cannot be denied is proof

that it cannot be semantically dominant. It cannot be semantically domi-

nant because the /-i/ identifies it as previously-mentioned information.

In (51), however, the idea that John is a Navajo could be new information

(i.e., neither previously discussed in the conversation nor presupposed to

be true), and therefore semantically dominant. As such, it can be negated.

Since (51) has two possible negations, it is said to be ambiguous

with respect to dominance. In other words, /-go/ is neutral with respect

to dominance: it is the context and not the /-go/ which determines the

dominance. This point is even more obvious when a sentence is put into

two different contexts. Consider the following:

52. Bil: Mary ch'e'h hanisht4. h gb6la i ya

Mary in vain 1.seek where.toQ 3.P.go
I can't find Mary. Where did she go?

Jaan: kintahg' i ydago yinii'

town.to 3.P.goCOMP 1.P.hear
I heard she's going to town.
 251

53. Bil: ji }da' Mary kintahgoo iiya", ni-ish blehozin

day.time pst Mary town to 3.P.go 2.with.Q 3.be known
Mary went to town today. Did you know?

Jaan: auu' Mary kintahgoo' iiyaago sid ts '' . ha' at 'I

fila
yes Mary town to 3.P.go.COMP 1.P.hear what Q
Yes, I heard Mary went to town today. Why (do you ask)?

The /-go/ clause in John's response in (52) answers the question

and is therefore the semantically dominant clause. In (53), however, the

/-go/ clause is repeating old information and is therefore relatively un-

important. In fact, a more natural answer would be simply, "Yes, I've

heard".11

In summary, the notion of semantic subordination seems to describe

the /-i/ complementizer: /-i/ marks everything it dominates as semantical-

ly subordinate. /-go/, on the other hand, is semantically empty. It

therefore permits context to determine dominance.

3.3.4. An Explanation of Earlier Translations

In light of the referential and dominance differences between /-go/

and /-/ clauses, let us now reconsider the earlier examples in which the

two types of clauses seemed to be identical.

13a. Mary kinla'nigo'o' lyaago yishniih

Mary Flagstaff.to 3.P.goCOMP 1.hear
I hear that Mary has gone to Flagstaff.
 252

13b. Mary kin;lnigoo' yahigi yishniih

Mary Flagstaff.to 3.P.goCOMP 1.hear
I hear that Mary has gone to Flagstaff.

Since /-go/ clauses may be either semantically subordinate or dominant,

there are two ways in which these two sentences could be considered dif-

ferent. First is the case where the embedded clause must be semantically

dominant. Thus (13a) but not (13b) could answer the following direct

question:

54. ha'ago6l Mary ilyaago niniih 1 2

where.to.Q Mary 3.P.goCOMP 3.hear
Where do you hear that Mary went?

(54) is not asking whether you heard about something. It is asking for

information about where Mary has gone. (54) could not be answered

55. *auu' yinii'

Yes, I've heard.

but could be answered

56. kinlanigo iiya

She went to Flagstaff.

Therefore, as an answer to (54), the embedded and not the matrix clause in

(13) must be semantically dominant. Since such an interpretation contra-

dicts the semantics of /- /, (13b) cannot be an answer to (54). Since

/-go/ is a semantically empty syntactic subordinator, either S may be

 253

semantically dominant in (13a) and therefore (13a) is an appropriate re-

sponse to (54).

The second way of contrasting (13a) and (13b) is to consider a

question they could both answer. (57) is such a question.

57. ha 'at ' illa niniih

what Q 2.hear
What did you hear?

Although both (13a) and (13b) talk about hearing that Mary went to Flag-

staff, the two convey different information. In (13a), the speaker need

not know whether or not Mary has gone to Flagstaff. He is neutrally re-

porting what had been said by others. In (13b), the speaker is reporting

what he considers to be the truth, what he had heard from reliable sources.

We therefore find the following contrast:

58a. Mary kin]a/nigo6' iiyaago yishniih, ako ndi Mary hooghandi sida' nisin
Mary Flagstaff.to 3.P.goCOMP 1.hear but Mary home at 3.sit 1.think
I hear that Mary has gone to Flagstaff, but believe she's at home.

b.*Mary kin]nigod ifiyahgii yishniih, ko ndi Mary hooghandi sida nisin

I heard about the fact that Mary has gone to Flagstaff, but I think
she's at home.

Example (13) is therefore consistent with the semantics required by

/-1/. It is not so clear, however, how (13b) relates to the dominance hy-

pothesis. Since as an answer to (57), the embedded clause in (13b) contains

the information requested, it cannot be considered semantically subordinate.

 254

Yet I have claimed that /-,/ clauses are always semantically subordinate.

There are two aspects to this problem. First, it is often thought

that complements of factive verbs are necessarily assumed to be true by

both the speaker and the hearer. However, Karttunen (1973) has pointed

out that factive verbs can be used to convey information. Consider the

following:

59. MIT regrets that it cannot accommodate children at graduation.

(59) would be found in a letter to graduates in order to inform them not

to bring children to graduation. It does not assume that the graduates

already knew that information. The use of a factive verb indicates that

the speaker, or writer, assumes the truth of the embedded proposition.

(59) is thus an indirect way of conveying a fact. Since the subordinate

S is conveying the new and most important information in the sentence, (59)

is thus a case where the complement of a factive verb is semantically

dominant. We must therefore refine the notion of semantic subordination

to include only those cases where the speaker assumes that the presupposi-

tion is assumed by the hearer as well.

3.3.5. Refining the Referential Hypothesis

Although not all clauses with /-5/ can be subsumed under the defini-

tion of semantic subordination, those that are not semantically subordinate

behave differently from semantically dominant clauses. Notice that it is

the entire subordinate clause of (13a) which is the answer to (57). There
 255

is a difference between concentrating on the whole /-i/ clause and a part

of it. I will return to this point below. For the present, note that when

a constituent within the clause is the answer to a question, as in (58), it

is impossible to use /-i/. These facts can be described mechanically by

recalling that /- / clauses are sentential NP's. As an NP, i.e., as a

constituent of the main clause, an /-/ clause is part of the dominant main

clause. But as an NP, the clause must also be taken as a single element,

an element the speaker indicates as having a unique reference, a fact.

Because English factive complements like that in (59) must be con-

sidered semantically dominant, e.g., because of extraction facts, there is

no way of defining semantic subordination to include all /-i/ clauses.

However, unlike the English equivalents, Navajo /-5/ clauses never act

syntactically like semantically dominant clauses, except in the case where

the clause is considered as a whole. With respect to extraction facts, for

example, /-i/ clauses may not be considered semantically dominant. For

this reason, I will continue to use the semantic dominance/subordination

contrast to help describe the Navajo facts. Also for this reason, the

other concepts developed for English are not exactly appropriate for

Navajo. The concept of anaphoricity developed by Pope coheres with the

notion of semantic subordination since it also assumes that the reference

is known by the hearer as well as the speaker. /- / on the other hand is

more like the complement of factive verbs in that its complement must be

presupposed anaphoric only by the speaker. I must thus modify my defini-

tion of the semantics associated with /-I/.

 256

The Referential Hypothesis

/-I/ denotes reference. The reference is either to a fact or

to a discourse. Therefore, whatever is in the scope of /-I/

is assumed referential for the speaker. The assumption of

reference is generally shared by the hearer, but need not be.

It should now be clear how an /-i/ clause can seem semantically dominant

when taken as a whole, i.e., as an NP. It is only when the information

is known to both the speaker and the hearer, either through presupposition

or previous discourse, that a clause must be semantically subordinate.

However, /-i/ must denote reference only with respect to the speaker.

When the reference is not shared with the hearer, an /- / clause need not
be semantically subordinate according to Erteschik's definition. However,
with respect to syntactic rules, anything that is referential in Navajo

will behave as if semantically subordinate.

Let us now return to the other examples in which /-go/ and /-I/
seemed to be identical.

12a. ch'e'etiingo'' naaneego shi;l b'hozin

entrance to 3.play.COMP 1.with 3.be known
I know that he's playing outside.

b. ch'e'etiingoo/ naanehigii shi-L behozin

entrance to 3.playCOMP 1.with 3.be known
I know that he's playing outside.

So far as I can tell, there is no difference in meaning between these two

when the matrix clause is semantically subordinate. However, in considering

 257

the meaning of the verb know, the reason for the identity in this particular

case should become clear. To know something is by definition to be aware

of a fact. 13
Since in (12) /-1 bee'h'zin/ is unqualified, there is no dif-

ference between knowing and knowing a fact. /- / therefore provides no

new information, so the use of the two complementizers overlaps.

One logical possibility is that the use of /-go/ in (12a) is a

method of qualifying /-- bedh6zin/, so that (12a) would be better trans-

lated as "I think I know...". So far as I am aware, this is not the case.

In all the other cases where /-go/ has been used to cast doubt, the doubt

has not been about the accuracy of the matrix verb, but about commitment

to the embedded proposition. In (58), for example, the speaker is not

saying he's not sure what he heard, but that he's not sure he believes

that what he heard is in fact true. To put doubt on the complement of

/-I b46hdzin/ requires doubting the assertion involved in the matrix verb

as well. Apparently the use of /-go/ is insufficient to indicate such

doubt. However, when /-;L be'ehozin/ is itself qualified, the difference

between /-go/ and /- / embedded under it becomes clear. Consider the

following questions and negations:

60a. ch'e'etiing6dd naane'ego doo shil bee'ho'zinda

entrance to 3.playCOMP neg 1.with 3.be known.neg
I don't know whether he's playing in the yard.

b. *ch'e'etiingo naandhagi doo shil be'h'zinda

entrance to 3.playCOMP neg 1.with 3.be known.neg
I don't know the fact that he is playing in the yard.
 258

61. ch'l'e6tiing66 naane'higi doo Mary bi] b66h/zinda

entrance to 3.playCOMP neg Mary 3.with 3.be known.neg
Mary doesn't know (about the fact) that he is playing in the yard.

62a. ch'e'etiing6 d naanlego/sh nil be4hdzin

entrance to 3.playCOMP.Q 2.with 3.be known
Do you know if he's playing in the yard?

b. ch''e'tiing 6 naanehigiish nil beshozin

entrance to 3.playCOMP.Q 2.with 3.be known
Do you know about (the fact) that he's playing in the yard?

The (b) sentences are perfectly consistent with my claim. (60b) is ungram-

matical because it is a contradiction to simultaneously assert and deny a

proposition. One cannot be in the state of knowing and not knowing simul-

taneously. It is however possible to explain that someone else does not

know the fact. As expected, (61) is grammatical although (60b) is not.

(62b) is questioning whether you are aware of a particular fact.

The (a) sentences are also perfectly consistent. They involve no

presuppositions. They are neutral questions.

3.3.6. /-go/ Is Semantically Empty

Sentences (60a) and (62a) are the first cases given in which

sentential complements have been translated by if instead of that. The

use of if with /-go/ is reminiscent of the adverbial /-go/ clauses dis-

cussed in section 3.1.1. In fact, it is possible to get a temporal reading

from an NP /-go/ complement as well as from the adverbial /-go/.14

 259

63. Ja'an o'ta'd'o' yo'' an lyeedgo shil b//hozin

John school.from away 3.P.runCOMP 1.with 3.be known

I know when John will run away from school.

Sentence (63) may be the most obvious evidence of the identity. From

everything said so far about /-go/ clauses, it should be clear that it

is the same /-go/ which appears in subordinate adverbials and embedded

sentential NP's. We must therefore give up the notion suggested in

section 3.1 that /-go/ contains the notion "adverbial". /-go/, in con-

trast to /-i/, does not have a consistent meaning. I therefore propose

that, unlike /-i/, /-go/ has no semantic content. It is a subordinator

creating a logical connection between a dependent proposition and the rest

of the sentence. To say that two S's are logically connected is to require

some possible-world relationship between them, specifying only that one S

is dependent on the other. In the adverbial cases, there is nothing about

the /-go/ clause which requires it to be an adverb; it is its relationship

to the independent clause which causes that interpretation, i.e., the fact

that there are other subjects and complements. I suggest that the tendency

to expect /-go/ clauses to be adverbs is based on the experience that they

are normally used that way, and not on the semantics or syntax of /-go/.

Since /-go/ is a semantically empty logical connective, it involves

no presupposition of reality. Thus, as in English "if...then" sentences,

a condition may never be fulfilled. The /-go/ by itself provides no clues

about the speaker's belief that the conditions will be fulfilled. In

English one may use when instead of if to imply that one expects the con-

ditions to be met. Contrast the following English sentences:

 260

64. If you rope the horse, my father will buy it.

65. When you rope the horse,my father will buy it.

66. If you rope the horse, and I don't think you will, my
father will buy it.

67. *When you rope the horse, and I don't think you will, my
father will buy it.

Both (64) and (65) require you to rope the horse in order for my father to

be committed to buying it. The major difference is in the speaker's ex-

pectations, as can be seen from (66) and (67). Since /-go/ itself is

neutral on expectations, it can be interpreted either way. The equivalent

Navajo sentences are vague on expectations. Therefore, the Navajo examples

which I previously labelled as ambiguous are in fact simply vague.

One way to specify expectations in Navajo is to add the clitic

/-sh'/ to the /-go/ clause. While the precise uses and meaning of /-shfV

must still be investigated, the following contrast is clear:

68. 94 ' doo yinishggdgK6 shi deeshool1 5

horse neg 3.2.P.tame.COMPneg I 1.F.tame
If you do not break in the horse, I will.

69. 141,e' doo yinishoodgddshfi sh deeshyol

horse neg 3.2.P.tameCOMPneg.possible I 1.F.tame
If you cannot break in the horse, I will.

(68) says nothing about whether the speaker believes that the condition

(your not breaking in the horse) will be met. It is also neutral about
 261

how this condition might not be met. You may not even try to tame the

horse. Sentence (69), on the other hand, would be uttered only when the

speaker expects you to attempt to break in the horse. The speaker is

doubting your ability to do it. He is really saying, "if you should fail,

I'll do it", or "when you fail to break in the horse, I'll take over".

Notice that the translations of (68) and (69) demonstrate that

English if can be used neutrally, much like /-go/. Notice also that it

is the auxiliary verb (do vs. can) which provides the shift in meaning.

As we saw in section 3.1, Navajo also uses its verb to provide information

not carried by the complementizer. The temporal and aspectual relationship

between the two clauses is crucial in distinguishing among because, before,

when, if, etc.

In summary, /-go/ can be viewed as a way of pointing out that one

proposition depends on another without spelling out the dependency. A

sentence may be left vague, or other factors can be used to fill in the

information. These factors include clitics, verb morphology, and the

nature of the independent verb, as we saw with /-;I be'h6zin/, (know).

3.4. Syntactic Evidence For the Referential Hypothesis

Let us now turn to some uses of two Navajo particles, /-ish/, a

question clitic, and /hanii/, a negation particle. It will be seen that

the behavior of these particles is different when used with different

complementizers. The referential hypothesis explains this difference

in behavior.
 262

3.4.1. Clitics Determine Focus and Presupposition

In Chapter 2, I discussed the fact that Navajo places particles in

what has been called clitic or second position. It is not clear to me

exactly how second position should be defined because there are many

places in a complex sentence into which the particles may be placed. In

investigating the problem of defining second position, I discovered that

the yes/no question clitic need not go into second position at all.

Kenneth Hale and Ellavina Perkins, investigating the uses of one negation

particle /hanii/, discovered that Navajo has the option of placing /hanii/

on any constituent in the sentence in order to focus on that constituent.

The result is that the rest of the S becomes presupposed. It seems that

for /-ish/, second position is a neutral position, stylistically natural

and requiring no presuppositions. However, /-ish/ may also behave like

/hanii/. Consider the following:

70. Jaanish 1i' biztal

John.Q horse 3.3.P.kick
(i) Did the horse kick John?
(ii) Is it John that the horse kicked?

71. Jaan 1i 'ishbiztal

John horse.Q 3.3.P.kick
Is it the horse that kicked John?

6
72. Jaan -4 ' biztal shl
John horse 3.3.P.kick.Q
Is it kicking that the horse did to John?
 263

73. ashkii hanii ;1i' nabiilgo'

boy neg horse 3.3.P.throw
It's not the boy the horse threw.

74. ashkii 111' hanii nabi go'

boy horse neg 3.3.P.throw
It's not the horse that threw the boy.

The result of clitic placement in simple sentences in Navajo appears to

be analogous to clefting in English. Except when in neutral second posi-

tion, the focus is on the constituent immediately to the left of the clitic

and the rest of the sentence is presupposed. Thus, translation (i) of

sentence (70) is perfectly neutral. It requires only a "yes" or "no" as

an answer. If, in contrast, (71) were to be answered, "no", it would be

considered an incomplete answer. The other person would respond, "well,

what did then?". Since (71) presupposes that something kicked John, a

denial implies that the speaker knows what did. A simple "no" is thus

withholding information. Instead, one would expect an answer like (75).

75. nda, Ja'an dzaan e'z ga' bizta;L

no John mule emph 3.3.P.kick
No, it's the mule that kicked John.

or simply,

76. nda, dzaanssz ga'

No, it's the mule.

Since John's being kicked is already presupposed, it is unnecessary to

repeat that information.

 264

In a similar fashion, a speaker uses /hanii/ instead of the neutral

negative /doo... da/ when a proposition has already been asserted and the

speaker wants to deny part of the proposition. Sentences (78) or (79)

would be natural responses to (77).

77. ashkii 11' nabiflgo'

boy horse 3.3.P.throw
The horse threw the boy.

78. ashkii ;LI' hanii nabi lgo', dzaan 4e'z ga'

boy horse neg 3.3.P.throw mule emph
It's not the horse that threw the boy, it's the mule.

79. ashkii hanii ; ' nabi go', at'e''d ga'

boy neg horse 3.3.P.throw girl emph
It's not the boy the horse threw, it's the girl.

Notice once again that it is unnecessary to repeat the entire sentence when

the denial is being explained. What is not repeated is precisely what is

still being presupposed. The following non-sentence confirms that what is

left unsaid is precisely what is being presupposed:

80. *ashiike dibe hanii yinoolkaL, at ' e'eke ga'

boys sheep neg 3.3.herd girls emph
It's not the sheep the boys are herding, it's the girls.

81. ashiike dib hanii yinoo-bkal, tl'IzI ga'

boys sheep neg 3.3.herd goats emph
It's not the sheep the boys are herding, it's the goats.

The first part of (80) denies that it is sheep the boys are herding. In

addition, however, it presupposes that the boys are herding something. The
 265

assumption, therefore, is that the second part of the sentence need not

repeat /ashiike/ or /yinoolkal/ because they are presupposed. The verb

-kal/ can only take certain animals as its object. Thus, one herds

sheep and goats but not girls. /at'ele/ could be a subject of /yinoolkal/.

The fact that /at Ie'ke/ is an incorrect continuation while /tl' z/ in (81)

is fine, shows that it is the object and not the subject which must be

substituted for. In other words, it is the denied NP which is being ex-

plained and the presupposed NP which is being deleted.

In summary, we have seen that in simple sentences the /-Ish/ and

/hanii/ can be placed after any constituent in order to focus on that

constituent and presuppose the rest of the sentence. Now let us turn to

complex sentences.

3.4.2. Complex Sentences With /-go!

When the embedded S has a /-go/ complementizer, /-/sh/ and /hanii/

can be placed after any constituent in the embedded clause in order to

focus on that constituent, just as they could in simple sentences. Thus,

we find the following possibilities:

82. Jaanish ;l4' bizta;Lgo yinidld

John.Q horse 3.3.P.kickCOMP 2.believe
(i) Do you believe that the horse kicked John?
(ii) Do you believe that the horse kicked John?

83. Jaan ii'ish biztalgo yinidla

John horse Q 3.3.P.kickCOMP 2.believe
Do you believe that the horse kicked John?
 266

84. Ja'an lI' biztalgolsh yinidl1

John horse 3.3.P.kickCOMP.Q 2.believe
Do you believe that the horse kicked John?
?Do you believe that the horse kicked John?

85. Ja'an hanii 1-' biztalgo

C
yinishdl4
L

John neg horse 3.3.P.kickCOMP 1.believe

I don't believe that the horse kicked John.

86. Jaan Ili' hanii biztalgo yinishdl

John horse neg 3.3.P.kickCOMP 1.believe
I don't believe that the horse that kicked John.

87. Jaan Ii' biztalgo hanii yinishdla"

CL L
John horse 3.3.P.kickCOMP neg 1.believe
I don't believe that the horse kicked John.
?I don't believe that the horse kicked John.

The effect of /hanii/ is more complicated in embedded clauses than

it is in simple sentences. The element to the left of /hanii/ is still the

focus. However, the focus is no longer what is being denied. Since I am

concerned here with focus and presupposition rather than the scope of the

denial, suffice it to say that regardless of where /hanii/ is placed in

the embedded clause, the scope of the denial is the entire embedded clause.

More precisely, the semantic effect of /hanii/ "is clearly to deny the

reason, i.e., the rational connection between the subordination and the

main clause" (p. 19).19 What is germane to the present discussion is the

fact that /hanii/ has the same effect on embedded /-go/ clauses as it does

on simple sentences in that it can focus on any element and in doing so

 267

create a presupposition. The following example from Perkins (1973) demon-

strates the presupposition through an appropriate continuation:

(her (15))
88. Mary hanii nfyaago baa shil hozho, Jaan ga'
Mary neg 3.P.comeCOMP 3.about 1.with 3.happy John emph
It's not that Mary came that I'm happy about, but (that) John (did).

/Jaan/ is being contrasted to the focus /Mary/. Since the rest of the sen-

tence is presupposed, it is not repeated.

In sum, in terms of focus and presupposition, /hanii/ has the same

effect in embedded /-go/ clauses as it does in simple sentences.

3.4.3. Complex Sentences With /- /

It is within an embedded S with an /-1/ complementizer that the

deviance occurs. Consider the following:

89. Ja'ansh -1-1 bizta]=fgif yinidlj

John.Q horse 3.3.P.kickCOMP 2.believe
Do you believe that the horse kicked John?
*Do you believe that the horse kicked John?

90. *Ja'an 1'i'sh biztaligf yinidla

John horse.Q 3.3.P.kickCOMP 2.believe
Do you believe the fact that the horse kicked John?

91. Ja'an 1i'

L t,
bizta lig ish yinidl4
John horse 3.3.P.kickCOMP.Q 2.believe
Do you believe that the horse kicked John?
 268

92. *Ja'an hanii 1" ' biztallgi yinishdla(

John neg horse 3.3.P.kickCOMP 1.believe
I don't believe the fact that the horse kicked John.

93. *Ja'an ;11'I-L hanii bizta;6 gli yinishdlai.

John horse neg 3.3.P.kickCOMP 1.believe
I don't believe the fact that the horse kicked John.

94. Jaan ;Li'

LL
biztaLigi hanii yinishdlai.
John horse 3.3.P.kickCOMP neg 1.believe
I don't believe (this business about) the horse's kicking John.

Notice that the only good question sentences are those in which /- sh/ acts

as a neutral question marker: /-ish/ is not focusing on a particular ele-

ment within the proposition. The only grammatical sentences with /hanii/

are the ones in which there is no focus within the embedded clause. When

/hanii/ is placed after the embedded clause (e.g., (94)), the entire pro-

position is being focused on, as well as denied. In other words, neither

particle may divide up an S within an /- / complementizer into a presupposed

part and a part focused for purposes of questioning or denying. It is

only the entire proposition that can be questioned or denied by use of

-ish/ or /hanii/ if that proposition has an /-f/ complementizer.

These facts contrast with those discussed in sections 3.4.1 and

3.4.2. The ungrammatical sentences here are identical to the good sen-

tences in section 3.4.2 except for the complementizer. We must therefore

assume that it is the choice of the complementizer and not the placement

of /hanii/ or /-Ish/ which is causing the ungrammaticality.

 269

Recall now the referential hypothesis: everything within the scope

of /-i/ must be anaphoric for the speaker, assuming the existence of a

fact or a proposition in discourse. In other words, everything within the

scope of /-i/ must be considered as a whole. Sentence (89), for example,

is a restatement of a claim that has been made (the horse kicked John) and

an inquiry as to whether you believe that claim. It is like saying

95. e-sh yinidl4C

that.Q 2.believe
Do you believe that?

96. hanel'g ish yinidlK

tell.REL.Q 2.believe
Do you believe what is said?

One must either accept or deny the whole. He may then, if he wishes, pick

apart the proposition, explaining which parts he believes and which he does

not.

To put it another way, once an entire proposition is assumed, how is

it possible to then deny or question a part? This restatement is more

easily explicated when what is being referred to is a fact:

97. ashkii 11/' nabiflgo'g Ish nil be'eh ozin

boy horse 3.3.P.throwCOMP.Q 2.you 3.be known
Do you know about the horse throwing the boy?

Sentence (97) states that the horse threw the boy and inquires whether you

are aware of that fact. If something is stated as the truth, how can part

of that truth then be questioned or denied?

 270

Although the above examples suggest that /hanii/ and /-Ish/ can

never occur in certain positions when used with /-i/ because these particles

require a focus as well as a presupposition, this is only partially true.

While not apparent in the above examples, the crucial factor which causes

the ungrammaticality is not that there is a focus within the embedded

clause, but that this use of a focus excludes the focus from the presup-

position. The following example demonstrates that it is possible to have

a focus within an /-i/ clause:

98. Jaan 1pQ ga' biztalg yinishdla

John horse emph 3.3.P.kickCOMP 1.believe
I believe the fact that it is the horse that kicked John.

Example (98) is more transparent because it involves neither a question nor

a negative. Although horse is being emphasized, John's having been kicked

by that horse is not being denied or questioned. It is therefore possible

to include horse in the presupposition. In contrast, (93) requires a pre-

supposition that something kicked John, but denies that the horse did it.

Thus, horse must be excluded from the presupposition.

It is the meaning of the matrix verb in (93) which requires that

horse be excluded from the presupposition. That is, I claim that a focus

is not necessarily eliminated from the presupposition, but is separated

from the presupposition in (93), for example, because it is impossible to

believe that the horse kicked John and simultaneously to deny that it was

the horse who kicked John. In other words, in this case, denying something

about the horse with respect to the matrix verb entails separating horse
 271

from the presupposition that something kicked John. However, when it is

possible to deny the identification of an element with respect to the

main verb but still include that element in the presupposition, the use

of /-i/ forces the interpretation in which the focus is included in the

presupposition. Contrast the following with (93):

99. Jaan 1*' hanii bizta;L g f yinii', dzaaneez ga'

John horse neg 3.3.P.kickCOMP 1.P.hear mule emph
I didn't hear about the horse kicking John, (but about) the mule.

(99) is possible because it does not deny that the horse kicked John. It

denies only that I had heard about it. In fact, my consultant explained

that (99) presupposes that the horse did kick John. She contrasted (99)

with (100), explaining that with /-i/ both the horse and the mule kicked

John, but when /-i/ is replaced by /-go/, only the mule kicked John.

100. Ja'an ;L"' hanii biztalgo yinii', dzaan'ez ga'

John horse neg 3.3.P.kickCOMP 1.P.hear mule emph
I didn't hear about the horse kicking John, (but about) the mule.

Notice that the English translations are the same in both cases.

This is because such a distinction cannot be made in an analogous fashion

in English. However, English relative clauses do exemplify this point.

Consider the following Navajo relative clause and its English translation:

101. ;11'
LI.
hanii nabi ;lgo'e'q ashkii shik'ihod i'4
horse neg 3.3.P.throw.REL boy 1.3.P.blame
It's not the boy whom the horse threw who blamed me.
 272

In (101) the proposition "the boy fell off the horse" is not being negated.

On the contrary, it is being presupposed. It is also being presupposed

that someone blamed me. What is being denied is that the boy, identified

as the one who fell off the horse, did the blaming.

Although there is a focus in the embedded clause, that focus is not

excluded from the referential proposition in the /-I/ clause.20 (Recall

that /-eeJ is the past tense variant of the /- / complementizer.) It is

therefore clear that it is not the presence of a focus within a referential

clause which causes the sentence to be ungrammatical, but the division of

the clause into two parts.

In summary, it is not only consistent with the semantics of /-a/

that there are restrictions on the placement of particles, but it is neces-

sary for them to be so restricted. A particle cannot separate an element

in the embedded clause from the presupposition because the semantics

associated with /-i/ requires that the entire clause have the same re-

ference, whether to a fact or to a discourse.

3.5. Embedded Questions With /-// and /-go/

Having described and contrasted /-go/ and /- /, I will now turn to

Navajo question formation, contrasting the syntactic forms possible with

/-go/ to those discussed in Chapter 1 with /- /, and show how the referen-

tial hypothesis and the concept of semantic dominance/subordination account

for the facts.

 273

3.5.1. WH Questions In /-go/ Adverbials

Since /-go/ clauses are most commonly adverbs, I will begin the

survey of /-go/ questions by examining WH questions from /-go/ adverbials.

Consider the following:

102. nizhe'e' ha'ago6'' iiyaago nicha

2.father where.to.Q 3.P.goCOMP 2.cry
Where did your father go that you are crying?
Because your father went where, you are crying?

103. ha'la h66t 'idgo na ha'chi'

what.Q 3.3.P.happenCOMP 2.for 3.be angry
What happened that you are angry?
Because what happened, you are angry?

104. haagoshg' dinyaago yinicha

where.to.Q 2.P.goCOMP 2.P.cry
Where are you going that you cried?
Because you are going where, you cried?

Since these sentences have no direct English translation, I have given two

approximations. The first one provides the correct meaning in grammatical

English, but misrepresents the syntax of the /-go/ clause. The second one

gives the proper syntax to the /-go/ clause, but no ordinary English ques-

tion can be formed.

We know that the /-go/ clause in (102) is adverbial because (102)

is analogous to (105), exactly like the adverbials we have already dis-

cussed.
 274

105. kinlanigoo iiyaago yishcha

Flagstaff.to 3.P.comeCOMP 1.cry
I'm crying because he went to Flagstaff.

(105) in fact is a perfect answer to (102). In addition, we can see that

the /-go/ clause is not being used as either a subject or object by de-

leting it:

106. yishcha
I am crying.

As usual, the subject pronoun is deleted. However, there is nothing else

incomplete about (106) as there would be if the /-go/ clause were an ob-

ject rather than an adverbial clause.

These examples are significant for several reasons. For one, these

are the first examples of WH words embedded in indirect discourse sub-

ordinate clauses where both the WH word and the interrogative particle

actually appear. For another, these questions are direct questions rather

than indirect questions. And last, it is important to notice that there

is no direct English translation.

The English translations of (102)-(104) are inadequate semantically

because they don't clarify the full extent of the questions. In (103),

for example, the question is not only "what happened?", but what happened,

and how is its happening connected to your being angry? The closest English

in intent is really, "why are you angry?".

The crucial point is that what is being questioned is not only the

WH word, but its connection to the main clause. This phenomenon is the
 275

same one found by Perkins (1973) in her investigation of /hanii/. As she

explains, "in embedded clauses, /hanii/, insofar as its scope is concerned,

behaves as if it were actually attached to the whole clause, rather than

being a part of the clause" (p. 19). Analogously, insofar as its scope is

concerned, the WH word behaves as if it were actually attached to the whole

clause, although syntactically it is a part of it. Recall that I defined

/-go/ as a logical connective, syntactically subordinating one clause to

another and thereby requiring a real world connection between the clauses.

That connection, and therefore the /-go/, plays a crucial role in the mean-

ing of the question. That is, the main clause enters into the question in-

sofar as it is connected to the event in the /-go/ clause.

In sum, the fact that it is possible to get a normal WH question

in Navajo adverbial clauses is a function of the nature of the /-go/ com-

plementizer. It is furthermore, only because /-go/ is a semantically

empty complementizer that the question can have the meaning that it does.

3.5.2. WH Questions With /-go/ As Contrasted To /-I/

Claiming that the WH questions above are possible because of the

nature of the /-go/ complementizer entails that such questions should be

possible in sentential complements with /-go/ as well as in adverbial

clauses. It also implies that such questions should not be possible if

/-I/ is substituted for /-go/ (in those environments where both comple-
mentizers are normally possible). The following examples substantiate

these claims:
 276

107a. Jaan Mary 1 haadi deeshnishgo bil beeh6zin

John Mary Q where.at 3.P.workCOMP 3.with 3.be known
Where does John know Mary to work?

b. *Jaan Mary i1'haadi deeshnishigif bil be h6zin

lO8a. hil h6ne'4 naaghalago nil beehozin

who.Q inside 3.walk aroundCOMP 2.you 3.be known
Who do you know to be walking around inside?

b. *ha'il hone'e naaghahig ni be'h zin

109a. Mary l' h4 4g iiyaago sidin/t s' a'

Mary Q where.to 3.P.goCOMP 2.P.hear
Where did you hear Mary to be going?

b. *Mary li haagoo yahigi sid nits''

lla. ha'at'lila biniinaa Jaan Mary yich'i' haadzii'go nil beehozin

what Q reason John Mary 3.to 3.P.talkCOMP 2.with 3.be known

Why do you know John to have talked to Mary?

b. *halat ' la biniinaa Ja'an Mary yich'i' haadzi' gii nil b6(h6zin

lila. h/dda' ll Jaan Mary yich'i' haadzli'go nil be'6hdzin

when Q John Mary 3.to 3.P.talkCOMP 2.with 3.be known
When do you know John to have talked to Mary?

b. *had4/j'la' Ja/an Mary yich'i' haadzi igii nil bsehozin

112a. haala' yit 'lego Jaan dine' bizaad yihool' a'go nil be'hzin

how Q way John people language 3.3.P.learnCOMP 2.with3.be known

How did John learn Navajo, do you know?

b. *haal yit'eego Jaan din(' bizaad yihoo-la'ig nil b/4hdzin

 277

The above examples have all the crucial properties of the WH questions

from adverbial clauses: they are direct questions; as far as the scope

of the WH word is concerned, the WH word is interpreted as if it were a

part of the matrix; both the WH word and the interrogative particle are

present; and such questions are either impossible or not normal in English.

While not apparent in the previous examples, notice also the crucial pro-

perty that no movement of the WH word or Q particle is required.

Just as /-I/ cannot replace /-go/ in the above WH questions, /-go/

cannot replace /-I/ in the syntactic form required for embedded WH ques-

tions with /-I/. I argued in Chapter 1 that embedded WH questions with

and
/-i/ are indirect questions formed through deletion of the WH word

movement of the spatial enclitic, if there is one, to the /-i/. Thus we

find the following:

113a. Ja'an naanehdiish nil be'hozin

John 3.playCOMPat.Q 2.with 3.be known
Do you know where John is playing?

Although /-go/ is also a possible complementizer for clauses embedded

under /-- beh6zin/, it is impossible to substitute /-go/ for /-// in the

above sentence:

113b. *Jaan naanlegodi'sh nil beehozin

 278

3.5.3. Difference In Meaning

Why should different complementizers require different syntactic

forms for embedded questions? The answer, I believe, is related to a

difference in the two types of questions: embedded questions with /-I/

are indirect questions only; those with /-go/ are direct questions.

Answers to questions with /-i/ but not with /-go/ require an answer about

whether the person knows the information. Contrast the following re-

sponses:

114. Q: Jaan ha'goola' deeya'ago niI beehozin

John where.to.Q 3.P.goCOMP 2.with 3.be known
Where do you know John to be going?

R: doo deeyahigooda
neg 3.P.goCOMPto.neg
He's .not going anywhere.

115. Q: Ja'an deeyahig sh nil beehozin

John 3.P.goCOMPto.Q 2.with 3.be known
Do you know where John is going?

R: auut shil be'h/zin. doo deeya/higoo'da

yes 1.with 3.be known neg 3.P.goCOMPto.neg
Yes, I know. He's not going anywhere.

(115) i s really made up of two questions, the direct yes/no question,

"do you know?" and the indirect form of the question "where is John going?".

The yes/no question is signified by /- sh/. Note that there is no /-Ish/

in (114). (114 ) contains only one question word, /haag6'/, to where.
 279

It is this word which changes form from (114) to (115), from a direct to

an indirect question.

The difference between the two syntactic forms should thus be re-

lated to the difference between direct and indirect questions. Why should

embedded questions be direct with /-go/ but indirect with /-//?

Once again, I believe that the semantics associated with /-i/ but

not /-go/ is responsible for this seemingly strange behavior. Taking

first the problem of why direct questions are impossible with /- /, I

claim that it is the semantics associated with that complementizer which

precludes well-formed direct questions. Recall that /-i/ clauses are

normally semantically subordinate. In a direct question, however, it is

the embedded S which is semantically dominant. Since /-go/ clauses, but

not /-i/ clauses, can be semantically dominant, it follows that the former

but not the latter would contain direct questions.

3.5.4. /-1/ and Deleted WH Words

The dominance relations permitted by the different complementizers

explain why we find both direct and indirect embedded questions in Navajo.

We still need to ask, however, why we find different syntactic forms, and,

more specifically, why the WH word deletes when embedded under /-f/. If

the complementizer together with a yes/no question clitic /-ish/ dis-

tinguishes between direct and indirect questions, why must the WH word

also delete? Deletion of question words is not a particularly common

process. It is a difficult thing to do unless the language has another

 280

way of designating what question is being asked. Thus, if we were to

delete the WH words in the following English sentences,

116a. Do you know where John went?

b. Do you know why John went?

c. Do you know how John went?

d. Do you know when John went?

(117) would be the result:

117. Do you know John went?

However, the hearer of sentence (117) could only interpret it to mean "do

you know that John went?". If (117) were considered to have a deleted WH

word, it would be impossible to distinguish among the alternatives in (116).

Navajo has a partial solution to this problem. If a WH word con-

tains a clitic, the clitic does not delete with it, but moves to comple-

mentizer position. Thus, the /goo/ in (115) tells us that the deleted

question is /haa/ (where). There are, however, many WH words which do

not contain clitics, who, what, how, why. In fact, it is only possible

to form indirect questions with /-i/ with the first two. Navajo minimizes

ambiguity by not forming certain indirect questions.

But why should it be necessary in the first place to delete the

WH word and thereby create problems?

Once again, I believe that the semantics associated with /-I/ is

responsible for this seemingly strange behavior. Recall that the

 281

referential hypothesis states that everything within the scope of //

must be anaphoric, i.e., referring to either a fact or a discourse con-

text. It is not the case, however, that question words are normally

anaphoric from the speaker's point of view. That is, the speaker is not

able to identify the reference of the WH word. Notice that claiming a

speaker is unable to identify a reference makes no claims about whether

the speaker presupposes that there is in fact a reference. The English

question (118) is in fact ambiguous (or perhaps vague) about the speaker's

assumptions.

118. Do you know where John is going?

It may be used with the knowledge that John is going somewhere. Thus, if

Bill and Mary are walking down the street and see John pull out of his

driveway, one might turn to the other and utter (118). In this case, the

speaker assumes that where does have a reference, but may have no assump-

tions about the identity of the reference. On the other hand, if Mary is

trying to find out what John is going to be doing, she might ask Bill

(118) without any assumption that John is in fact going anywhere. In this

case, (118) could be paraphrased, "Do you know if John is going anywhere,

and if so, where?".

I have no reason to assume that (114) is any different from (118)

with respect to the speaker's presupposition about whether there is a

unique reference for the WH word. The use of these questions therefore

requires a refinement of the term referential. To presuppose that there

is a reference is insufficient. It seems that actual knowledge of the

 282

reference is required. It is clear that the speaker does not have know-

ledge of the reference for where in (118). It is also true that the

speaker need not presuppose that his companion has that knowledge in

order to ask the question. As in the first situation described above,

Bill may assume that where has a reference but have no reason to assume

that his companion knows any more than he. It would still be natural to

ask Mary (118) in order to find out whether she has any such knowledge.

In fact, if Bill presupposes that Mary knows the reference, it is more

natural to ask a direct question, either (119) or (120).

119. Where is John going?

120. Where do you know John to be going?

(120) is, of course, more natural in Navajo (cf. (114)).

In sum, I define referential as follows:

An element X is referential if (i) the speaker has actual

knowledge of X through the discourse or through real world

knowledge or (ii) the speaker presupposes that the hearer

has actual knowledge of X.

Now let us return to our consideration of WH words embedded under

/-I/. If the WH word were to remain in the embedded S, the semantic in-

terpretation assigned to it by /- / would say that the question word does

have a reference. Yet we have just said that the question word is to be

interpreted without a reference. The sentence would therefore be

 283

ungrammatical by being semantically anomalous. The sentence would not be

anomalous if there were no WH word. One possible explanation is that

there is a dummy element and not a WH word in the underlying structure.

I have argued against this solution in Chapter 1. Since there is a WH

word in underlying structure, and its presence presents a potential

contradiction, Navajo uses another possibility, deleting the question

word. It is because of the potential semantic contradiction that the

question word deletes. Notice that when enclitic movement takes place

in (115), the spatial enclitic is interpreted as being outside the scope

of /-i/. (See the following section for more evidence.) It was thus

necessary to define the scope of /-]/ in section 3.3.2 to exclude the

spatial enclitic. That is, no part of the question is left in a position

requiring a referential interpretation.21

3.5.5. WH Words With Presuppositions

As we have seen, not all questions are neutral questions. We also

use questions to elicit more information, as in the situation where John

is seen driving out of his driveway. In English it is only the context

which elucidates whether or not the speaker is assuming that there is a

specific reference. In Navajo different clitics are used with the WH

word to convey this information. Thus, in addition to indirect questions

with deleted WH words, we find:

121. Jaan ha' a't 'fsh L L-

yinizinig ish ni- b eh'ozin

John what.I don't know 3.3.wantCOMP.Q 1.with 3.be known

Do you know what John wants?
 284

122. Jaan hia'goo'sh deeyihiigish nisk beehozin

John where.to.I don't know 3.go COMP.Q 2.with 3.be known
Do you know where John is going?

The precise meaning and uses of /-sh-i/ must still be investigated.22In

this particular context, /-shg/ is often translated, "I don't know".

Thus (121) could be paraphrased, "John wants I-don't-know-what. Do you

know?" /ha'at'l'sh/ means something as well as what. Thus, perhaps a

better paraphrase is, "John wants something-I-don't-know-what. Do you

know?".

From what has been said so far, /ha'at' shff/ does not seem to be

referential. How then can it stay within the scope of /-1/? While I do

not have substantial evidence, I claim that /ha'lt'i shil / is in fact

referential. The following provides some evidence that /WHshif/ can be

used referentially:

ha / deeyaigif hi
123. Jan ch'aa deeyada'f, "'shj

John particle 3.P.goCOMP where.to.perhaps 3.P.goCOMP 1.with

beehodooz ;k
3.F.know

If John goes out traveling, I want to know where he is going.

In (123) /ha/a/go'sh4/ is referring to the place John may go to. It is

referential because it has been identified by the previous clause. Notice

however that since it is not certain in (123) that John will be traveling,

/haaigdo'shg/ here does not presuppose that a destination actually exists.

 285

Returning now to (121), I claim that /ha' t'ffshj4/ is referential

in the sense that the speaker is indicating that he presupposes that the

addressee knows or will know the reference. As discussed earlier, /-sh(/

is also used to express a challenge. I suggest that (121) really indicates

that John wants something that I cannot identify but I think you can, so

will you tell me? Notice that, given this interpretation, sentences like

(121) need not be indirect questions at all, but only a yes/no question

containing a referential PRO form. The translations are certainly not

sufficient proof that they are indirect questions. However, whether or

not they are indirect questions is a separate problem from determining

whether they are consistent with the referential hypothesis.

Navajo has still another form for what, /ha'tt' ida/. Although the

facts are not as yet completely clear, it seems that this variant is em-

ployed when the speaker not only assumes there exists a reference, but in

fact knows what that reference is. Thus, when an indirect question is

employed either to discover whether another person shares the speaker's

information or to convey that information is being withheld, /ha'tt' ida/

may be used. For example:

124. Ja'an ha't'ida nayiisnii'g ish nil beehzin? chidi nayiisnii'.

John what -da 3.3.P.buy.COMP.Q 2.with 3.be known car 3.3.P.buy
Do you know what thing John bought? He bought a car.

125. Ja'an ha'a'tI'ida yinizinig 1 Mary bil be4hozin, ndi doo e' niida

John what -da 3.3.wants Mary 3.with 3.be known but neg thab 3.say.neg
Mary knows what John wants, but she isn't saying.
 286

There is another, although marginally grammatical, type of indirect

question involving a specific reference for the WH word. This type in-

cludes cases in which the spatial enclitic movement transformation has

failed to move the enclitic out of its own S.

126a. naagha$ ninedd'gdd' deeshad;L

3.come 3.say.pstCOMPfrom.to 1.F.go

I will go to where he said she is from.

b. ?naaghahidy' ninag/ deesha-

3.comeCOMPfrom 3.say.pstCOMPto 1.F.go

I'm going to the place where he said she is from.

Sentence (126a) is the predicted output. However, there is a reluctance

among some speakers to move a spatial enclitic out of its S when there are

two non-identical third persons. Thus, some speakers will accept (126b).

However, if good, (126b) means that the place (where) has previously been

mentioned. There is, of course, no such constraint on (126a). In other

words, if for some reason a spatial enclitic does remain within the scope

of /-I/, it will be interpreted by the /-I/ to have a reference, just like

anything else within the scope of /-I/.

In summary, there are some types of WH words which appear within

the scope of /-i/. These WH words are semantically consistent with the

referential hypothesis. And leaving these WH words within the embedded

S is also syntactically grammatical. It is the existence of the referen-

tial hypothesis within Navajo which prompted an alternate method of in-

direct question formation.

 287

3.5.6. /-go/ and Indirect Questions

So far I have tried to account for why there are embedded direct

questions with /-go/ but not /-a/ and why the WH word deletes in embedded

indirect questions with /-i/. The final possibility to be considered is

embedded indirect questions with /-go/.

As already demonstrated, /-go/ is vague with respect to dominance,

allowing its clause to be interpreted as either dominant or subordinate.

In terms of possible dominance relationships, one would therefore expect

indirect as well as direct questions with /-go/. I have already shown

that it is impossible to form indirect questions with /-go/ the way they

are formed with /-a/ (cf. (113b)). In fact, it is unclear whether it is

possible to form the equivalent questions with /-go/ at all. If I ask

for a translation from the English, the answer is always a sentence with

I/2 with one exception. The only indirect question type that is

generally accepted is the whether question.

127. ch'e''etiing6 6' naaneegofsh nil beh6zin

entrance to 3.playCOMP.Q 2.with 3.be known
Do you know if he is playing in the yard?

128. Jaanish deeshnishgo nil be'h'zin

John.Q 3.P.workCOMP 2.with 3.be known
Do you know if John has started working?

Notice that the above examples do not actually contain a WH word. We

know, however, that WH words do not delete within /-go/ clauses. However,

there is no reason to postulate an underlying WH word for these cases.

The conditional interpretation is derived from the fact that /-go/ does

not require a presupposition. Since /-go/ is vague with respect to pre-

supposition, these sentences should have two possible interpretations,

the second being a factive interpretation.24 In fact, (127) could also

be translated "do you know that he is playing outside?". The latter

interpretation could also be rendered through the use of /-I/ instead of

/-go/. However, because of the semantics associated with /-I/, whether

questions are impossible with that complementizer.

Given that deletion of the WH word is impossible with the /-go/

complementizer, there are two other possible ways of forming indirect

questions. One is to embed the WH word and Q particle under a yes/no

question, as in (129).

129. ?*Bilish Mary hde yo' eelwodgo yiyiinii'

Bill.Q Mary where.from.Q away 3.P.runCOMP 3.3.P.hear

Although not always rejected, I have found no person who consistently

accepts (129). I therefore consider (129) ungrammatical.

We know from Chapter 2 that it is possible to have both a yes/no

question and a WH question in one sentence if they are in different S's.

I believe the problem with (129) is that the embedded question cannot be

interpreted as an indirect question. Indirect questions are found in

semantically subordinate clauses. However, the WH question particle /la/

singles out the element it is attached to, thereby requiring that clause

to be semantically dominant (recall the discussion in Chapter 2). If the

 289

clause is semantically dominant, the question will be interpreted as a

direct rather than an indirect question. This line of reasoning is sub-

stantiated by the fact that, when accepted, (129) is not answered /auut/

(yes) or /nda/ (no) as an indirect question would be.

Having two direct questions in one sentence was possible in the

direct discourse cases because the questions were from different points

of view, i.e., each question was being asked by a different person.

Since /-go/ clauses are indirect discourse clauses, the point of view

is consistently that of the speaker, just as in English. Thus, (129) is

ungrammatical, just as it would be in English. I have given no transla-

tion of (129) because there is no English form. A close approximation

would be, "*did Bill hear and where that Mary ran away from?".

Recall that the WH question in indirect questions with /-I/ does

not contain a question particle: the only question particle is the yes/no

particle /- sh/. One might therefore expect that an indirect question with

a /-go/ complementizer would be formed with a WH word but no question

particle, as in (130).

130. Bilish Mary haade(g' yd'o' eelwodgo yiyiinii'

Bill.Q Mary where.from away 3.P.runCOMP 3.3.P.hear

Such sentences are occasionally accepted, but the response rarely includes

"yes" or "no". (131) was given as a response to (130).

131. Alta'de' yo'o'' eelwodigii yiyifinii'

school.from away 3.P.runCOMP 3.3.P.hear
I heard that he ran away from school.
 290

It therefore seems that the yes/no particle is being taken as the WH

question particle /-sh/, easy to do since /lsh/ is not a natural cluster,

causing a pronunciation almost like that of /lish/. Since the initial

question particle /dat/ occurs in yes/no questions, but not WH questions,

adding /da'/ to (129)-(130) should clarify whether such double questions

are possible. However, the sentences still create confusion. More in-

vestigation is thus needed before any real claims can be made.

Pending further investigation, I will assume that indirect questions

are impossible with /-go/. It is therefore necessary to explain why such

questions are impossible. I will assume that question particles occur

with direct questions only, and therefore that (129) is ungrammatical.

In other words, I am claiming that if indirect questions were possible

with /-go/, they would be formed either by deletion of the WH word (with
25
enclitic raising) as in indirect questions with /-i/, or by leaving the

WH word in the embedded clause, but without a particle. It is therefore

necessary to account for the fact that these alternatives are impossible.

It is possible to account for the first case either by using some

kind of surface filter or by keeping the WH deletion and/or enclitic

raising from applying to /-go/ clauses. I opt for using /-i/ to trigger

the transformations since I claim that it is the semantics associated

with that complementizer which requires the transformation.

The second case is more difficult to account for since WH words

do occur embedded under /-go/. They occur in direct questions with the

WH question particles. However, so far as I know, WH words do not occur

 291

alone in underlying form. They occur either with an interrogative particle,

6
with /-shJ/, with /-da/, or as a preverbal particle.2 It may be there-

fore that indirect questions with /-go/ are impossible because, while WH

words require a particle, there is no particle for indirect questions.

There is no compelling explanation for why indirect questions with

a WH word embedded under /-go/ are impossible. However, the absence of a

clear explanation correlates with our uncertainty as to whether such

questions are in face impossible. Since /-go/ is a semantically empty

logical connective, there is no semantic reason to rule out these questions.

More investigation is required to see what other factors could be involved.

3.5.7. /-go/ and Unbounded Movement

I suggested that the question particle in (130) might be inter-

preted as the WH question particle. Notice, however, that the particle

in (130) is in the matrix, not the embedded S. For (130) to be taken as

a direct WH question, it is necessary to allow the question particle to

move out of the embedded S, just as it does in the direct discourse cases

discussed in Chapter 2. While not everyone who allows movement out of

direct discourse S's also allows movement out of /-go/ clauses, those

who accepted (130) also accepted the following:

132. Bil la Mary ha(dded' yo' ' eelwodgo yiy nii'

Bill Q Mary where.from away 3.P.runCOMP 3.3.P.hear
27
Where did Bill hear that Mary ran away from?
 292

133. ha'adil& Bil Jaan naaneego bil b hozin

where.at.Q Bill John 3.playCOMP 3.with 3.be known
Where did Bill hear that John is playing?

134. h ag66l' Bil Mary yfiyago yidiizts' '

where.to.Q Bill Mary 3.P.comeCOMP 3.3.P.hear
Where did Bill hear that Mary went?

135. Bil la haad ' Mary yo'o' eelwodgo yiy inii'

Bill Q where.from Mary away 3.P.runCOMP 3.3.P.hear
Where did Bill hear that Mary ran away from?

Since /l// is clearly a WH question particle, (132) is a clear case of

WH particle movement, leaving the WH word within the embedded /-go/ clause.

(133)-(134) are examples of both the WH word and question particle moving

out of the embedded clause. Note that in (135), the WH word appears to

have moved, but it has not ended up in initial position. In other words,

all the possibilities that occur with direct discourse sentences can also

occur for some speakers with /-go/ clauses. The difference is, of course,

that these are the only possibilities for direct questions from embedded

direct discourse clauses, while the normal method for question formation

from /-go/ clauses is to leave the WH word and question particle within

the embedded clause. While I have no proof, it would not be surprising

to discover that the leftward movement out of /-go/ clauses is a generali-

zation from the movement out of direct discourse clauses. That is, I am

suggesting that unbounded leftward movement exists in Navajo for a reason:

because of the use of direct discourse, it would be impossible for strict

constructionists to ask certain questions without some device to exclude

 293

the WH word from the direct discourse point of view interpretation. I

suggest that, whatever the proper formulation of the rule may be, there

is a universal question movement rule which is applied in Navajo for just

these cases. That rule is not normally used in SOV, COMP-final languages

but exists in Navajo, I claim, because of the problem created by the use

of direct discourse. However, once a rule is applied in a particular

language, there is no reason for it not to be generalized to other cases

in that language. What is of interest is that this unbounded movement

rule is at times generalized for /-go/ clauses, but never for /- / clauses,

even though it is the latter which cause problems. This difference should

follow from a universal theory of movement. While I cannot provide a com-

plete theory, I will make some claims about such a theory in the Conclusion.
 294

FOOTNOTES TO CHAPTER 3

1. While deletion is impossible, unbounded leftward movement is possible

for some people. Also, it is still not clear whether indirect ques-

tions exist with /-go/. See section 3.5.6.

2. See Kaufman (forthcoming) for a detailed account of how adverbial

clauses are interpreted.

3. See Chapter 1. There is also a form of this complementizer, /-I/

without the /-g /, which appears in adverbials. Also, /-gli/ does

not occur if /-I/ is followed by a spatial enclitic. I will be

ignoring such clauses in this chapter.

4. This sentence is grammatical when /-i/ is a relative clause marker.

The sentence then means, "the horse you roped my father will buy".

5. Since I argue that not all /-go/ clauses are adverbial, the node ADV

does not in fact help to characterize that complementizer. See

section 3.3.4.

6. /yi-/ and /bi-/ both occur with the third person subject and object

combination. Navajo has a subject object inversion transformation.

When the object precedes the subject, /bi-/ instead of /yi-/ appears

on the verb. For further discussion, see Hale (1973).

7. The use of the fourth person is much more complicated than my ex-

planation would suggest. The /ho-/, for example, is necessary only

 295

when /la'/ refers to a person. Thus (36) is good with the meaning

"I didn't hear one of them".

8. This sentence is good with the relative clause reading, "Mary is

afraid of the one who is driving".

9. It is too strong to claim that /-go/ precludes any presuppositions

since /-go/ occurs with /-1 b e'hozin/. I have no explanation for

why it can occur there, but discuss the problem further in section

3.3.4.

10. "Command" is defined as follows: "Node A 'commands' another node B

if (1) neither A nor B dominates the other; and (2) the S-node that

most immediately dominates A also dominates B" (Langacker (1969a),

167).
Note that as defined, the head of a relative clause is not

within the scope of /-1/. Navajo relative clauses can be formed

either through deletion of the head NP (preferable) or through back-

wards deletion of the antecedent. In the former case only the entire

relative clause is within the scope of /-i/, yet I know of no dif-

ferences in meaning. The reason could be that all definite NP's are

anaphoric and the head of a relative clause is definite.

It is possible to define the scope of /-I/ to include the head

of a relative clause: one could define the scope over an NP instead

of an S. I have rejected this approach for several reasons. First,

command is a more natural concept because it is found throughout

 296

languages. Second, there is an /- / which occurs on some adverbial

clauses, which are S's, but not NP's, and I argue that it is the same

complementizer. Third and most important, the spatial enclitic that

attaches to /-1/ must be considered out of its scope. It would be

impossible to include the head of a relative clause and exclude the

enclitics.

11. It may in fact be more natural to use /- / here because it contains

more information. /-go/, however, is certainly not incorrect. I

discuss this point further below.

12. I discuss this type of question in section 3.5.1. The English

translation is misleading, because the Navajo means, "where did

Mary go that you heard?",and not "where do you hear...?".

13. What must be explained is why /-go/ is permitted at all. In other

words, what makes /-1 be'ehd'zin/ different from the true factives

which do not allow /-go/ complements? This problem, however, is

not peculiar to Navajo. Know in English is identical in behavior.

(Ross' "Conjunctive and Disjunctive Questions," talk first given at

University of California at Berkeley, Fall, 1971.)

14. While I have been told of such readings by more than one person,

there seems to be something strange about them. Many such examples

were presented in Vetter (1968), but in checking his data, I found

most people I asked disagreed with his translations. Much more

work is needed in this particular area.

 297

15. /go'o/ here is an alternative of /dago/ (neg.COMP).

16. It is not clear whether this sentence in addition has a neutral

reading, "did John kick the horse?". Ellavina Perkins is currently

working on this problem. Cf. also Perkins (1973).

17. I have probably left out some possible readings. A ? in this case

does not mean marginally grammatical so much as do not know whether

it is a proper reading. Cf. Perkins (1973) for further information.

18. See Hale (forthcoming A) for a comparison between the scope of /hanii/

and English contradictory intonation. He explains why the scope of

/hanii/ must be the entire embedded S. The use of /hanii/ is very

similar to the English use of contradictory intonation with negation,

for which see Liberman and Sag (1974).

19. It is not clear whether the presence of /hanii/ in the embedded clause

forces a presupposition about the matrix clause as well. See Perkins

(1973).

20. Notice that these sentences require a new definition of focus and/or

presupposition. Chomsky (1969) defines the presupposition so as to

exclude the focus from it, i.e., as "an expression derived by replac-

ing the focus by a variable" (26). Since in these cases the focus is

part of the presupposition, at least one of the definitions must be

changed. (See Chapter 2, footnote 40 for another problem with these

concepts.)
 298

21. The other logical possibility is that the WH word itself moves out-

side the scope of the /-1/. In the Conclusion I discuss some hypo-

theses concerning the claim that WH words will never move rightward,

and try to account for why the WH word does not move leftward out of

the scope of /-I/.

22. See the discussion of example (69).

23. This is true if a sentence with /-i/ exists. If not, I am given a

direct question with /-go/ as the closest Navajo equivalent. The

fact that Navajo has these direct question sentences which cannot be

said in English creates added difficulty in sorting out the exact

meaning of the questions, and makes it particularly difficult to

use translations to determine meaning.

24. It is also not clear whether these sentences can be used temporally

as well. (128) has also been translated "Do you know when John will

start working?" (see footnote 14).

25. This method is definitely ungrammatical, a fact which must be accounted

for even if indirect questions should exist with another formulation.

26. This argument is possible only in a theory in which the question par-

ticles are generated with the WH word rather than in a pre-S node.

27. The English translation is misleading because the Navajo is not asking

where Bill heard a fact, but where Mary ran away from, according to

what Bill heard.

 299

CONCLUSION

I have shown in this dissertation that both unbounded rightward and

unbounded leftward movement exist in Navajo. It is therefore necessary for

theories of movement like Bresnan's and Chomsky's to be modified to account

for their existence. In addition, some new criteria must be found to ex-

plain when and in which direction unbounded movement will occur. I believe

it is premature to propose any final alternative to Bresnan's Complemen-

tizerAttraction Universal which was meant to account for the fact that only

unbounded leftward movement had been found in natural languages. Many more

verb final languages will have to be investigated before some general

principle can be found to describe which languages will evidence unbounded

movement. I have claimed for Navajo that the existence of unbounded move-

ment is related to interpretation problems: movement occurs to remove an

element from the scope of a governing element (verb or complementizer)

which would generate an undesired or semantically anomolous interpretation.

Such a requirement is certainly not universal--it does not hold for English,

for example. However, it may be that some related requirements will be

found for the other SOV, COMP-final languages which have unbounded movement.

On the basis of data now available, it is possible to propose some

tentative hypotheses about the direction in which movement will occur. I

believe that it is significant that unbounded leftward movement in Navajo

is at time acceptable out of /-go/ clauses. Erteschik's condition on ex-

traction states that "extraction can occur only out of clauses or phrases
 300

which can be considered dominant in some context" (27). Since /-go/ but

not /- / clauses can be dominant, it follows that if there is extraction,

it should be from /-go/ but not /- / clauses. However, as stated,

Erteschik's condition does not differentiate leftward movement from

rightward movement. If Erteschik's condition is to be used to account

for the fact that unbounded leftward movement does not occur from /- /
clauses, it must be stated so as to be irrelevant to indirect questions

with /-i/ since in that case spatial enclitics may be extracted, but

rightward, not leftward. Such an exclusion is most easily accomplished

by restricting her principle to leftward extraction. I suggest that such

a restriction is not ad hoc, and that the correlation between the concept

of semantic dominance and cases of leftward movement is significant. I

propose that the following two conditions--one syntactic, the other

semantic--are involved in determining whether movement will be rightward

or leftward.

(i) The most natural direction for unbounded movement is

toward the clause-boundary marker, i.e., COMP position.

(ii) Semantically prominent elements will never move rightward.

Condition (i) is a weaker version of the claims made by Bach, Baker,

Bresnan, Chomsky, et al. It says that when semantic factors do not

interfere (condition (ii)), the direction of the movement will be deter-

mined by word order. Since the COMP in English will always be to the

left of an embedded element, movement will always be leftward. In Navajo,

however, the COMP will always be to the right of an embedded element.

 301

Condition (i) therefore predicts that if movement occurs in Navajo, it will

be rightward. However, as condition (ii) states, there are semantic fac-

tors which will keep elements from moving rightward.

Langacker (1974) uses the term "prominence" to describe this in-

tuitive notion, stating that within a P-marker, node A will be more

prominent than node B if A commands B asymmetrically, if A precedes B, or

if A bears heavier stress than B." Movement rules therefore "have the

effect of altering the relative prominence of constituents" (649). Un-

bounded leftward movement makes node A more prominent because it changes

both the primacy relation of "precede" and "command" in a positive fashion.

Unbounded rightward movement has little or no effect on prominence because

there is a trade-off between "precede" and "command".2 The concept of

prominence is very similar to Erteschik's concept of semantic dominance.

Since semantic dominance refers only to the clause out of which an element

may move, and not to the element itself, I have introduced the term

"prominence" to refer to the element which moves. Combining Erteschik's

concept with Langacker's, clauses which are semantically subordinate can-

not contain elements which are prominent. Since unbounded leftward

movement creates prominence, an element which is not part of a semantically

dominant clause cannot be moved leftward. Since unbounded rightward move-

ment does not create prominence, an element which should be prominent

cannot be moved rightward.

Considering questions, WH words are more prominent in direct than

indirect questions. Or, to put it another way, if a WH word is prominent

 302

within an S (either because of movement or the use of a particle, the

Navajo equivalent to English stress), it will be interpreted as a direct

rather than an indirect question word.

As has often been noted, in indirect questions in English the WH

word moves leftward but not to the matrix S, i.e., not to initial posi-

tion. In Langacker's terms, WH words in indirect questions are not

prominent the way they are in direct questions. As Langacker states,

the correct generalization about a questioned NP is that "it is moved

to the beginning of the interrogative clause with which it is semantically

associated" (Langacker (1969), 81). In more general terms, it marks off

the scope of the question. That is, WH words in English move leftward

less because of condition (ii) above than because of condition (i). I

suggest that what makes COMP position significant is that the COMP is a

clause-boundary marker. Moving a WH word to COMP position sets off the

scope of the question by marking the beginning of the clause containing

the question.

Notice that in English (as in all the classic cases of unbounded

movement) both condition (i) and condition (ii) result in leftward move-

ment. Thus, the two conditions were never specifically differentiated.

I hypothesize for Navajo that the two conditions plus the scope

factors account for the different syntactic forms found in embedded

questions. Thus, in direct discourse sentences, the WH word is always

prominent because the WH question is always a direct question (from some-

body's point of view). In terms of prominence, movement is not necessary

 303

because the interrogative particle singles out the WH word as prominent

(i.e., particles in Navajo are equivalent in this sense to stress in

English). However, the semantic interpretation found in direct discourse

requires movement to produce certain meanings (e.g., direct questions

from the speaker's point of view). The choice of which S a WH question

particle will move to is determined by the scope of the verb, since it

is the verb which determines the point of view: movement is out of the

scope of one verb into that of another. Notice that the element governing

the scope determines the extent but not the direction of the movement.

Thus, in direct discourse, the verb is the governing element. If move-

ment were toward it, we would find rightward movement, not leftward. It

is the prominence factor which causes the leftward movement. Because WH

words in direct questions are prominent they cannot move rightward (con-

dition (ii)). Since the WH word must move because of scope factors, it

moves leftward.

In questions with /-go/, movement is not necessary because question

particles also create prominence. Since /-go/ clauses can be semantically

dominant, it is possible for an element to remain within the /-go/ clause

and still be prominent. However, since in these cases the entire sentence

is within the scope of the question, the WH word may also occur in any S

in the sentence. I am not claiming for Navajo that the scope of a ques-

tion must be delimited by a WH word (as in English), only that a WH word

remains within the interrogative segment of the sentence with which it is

semantically associated.
 304

Given an underlying form with a WH word embedded in an /-i/ clause,

if the non-referential WH word remains, it will be interpreted as referen-

tial, creating a conflict. The WH word must therefore be removed from the

scope of the /-/. If the WH word moves unboundedly leftward to initial

position, it will be considered prominent, also creating conflicts. The

sentence will be interpreted as a direct rather than indirect WH question

and the scope of the WH question will be the entire S. Since /i/ clauses

are not semantically dominant, it is impossible for an element within its

clause to control the scope of the entire sentence. If the WH word moved

leftward to the beginning of its clause, as in English indirect questions,

it would still be interpreted as within the scope of the /-i/. That is,

Navajo marks the end rather than the beginning of a clause with a comple-

mentizer. Since embedded S's are center-embedded, if an element is moved

to the front of the lower clause, there is no way of telling whether it

is in the embedded S or the matrix S. If NP. in (A) is moved to the front

3
of Si, it could be adjoined either to S1 as in (B), or to S2 as in (C)

(A)

(
 305

(B) O

NP V

. NP V C MP

(C)

(NP) . P
M

producing the same linear string. Since Navajo marks the end rather than

the beginning of a clause, when an element is moved rightward it is clear

whether or not the element has moved out of the embedded S. For Navajo,

condition (i) could be stated either in terms of the verb or the COMP since

both are used to mark clauses. (Recall that direct discourse S's have no

overt complementizer.) I have chosen to state condition (i) in terms of

COMP rather than the governing verb since it is in general COMP position

which marks one boundary of an embedded clause. Notice that such a state-

ment does not commit me to postulating underlying COMP's for direct dis-

course verbs since in those cases movement is always leftward anyway

(because of condition (ii)).

 306

Returning to indirect questions, the final factor to consider is,

given the rightward movement, why does the WH word delete leaving only the

spatial enclitic? One approach would be to say that WH words are by

definition prominent and therefore rightward movement is in conflict with

the prominence requirement. Another approach is to say that movement

beyond the scope of the /-i/ is also movement beyond the scope of the

question and a WH word must remain within the interrogative segment of

the sentence with which it is semantically associated. These approaches

are simply ways of stating that, for whatever reason, unbounded rightward

movement of WH words is difficult to process. However discussed, in terms

of constraints on possible transformations, constraints on possible deriva-

tions, functional sentence perspective, behavioral principles, etc., the

fact that certain types of elements tend to occur throughout languages

near the beginning rather than the end of a sentence has long been con-

sidered significant. So far as I know, no truly formal criteria have

been proposed to determine what elements will appear early in a sentence.

Solely syntactic criteria are clearly inadequate. On the other hand,

concepts such as topic-comment and prominence are really intuitive notions

couched in more formal terminology to claim that what comes first sets the

stage. Yet, despite their shortcomings, these notions correctly take

semantic factors into account. I believe that such considerations are

crucial.

In sum, we are as yet unable to predict when a language will use

unbounded movement transformations. However, it seems that when conditions

 307

(i) and (ii) are in conflict, the language is less likely to use unbounded

rules. There are two major factors which lead to movement, prominence and

semantic interpretation due to scope factors. Unbounded rightward movement

is less common than unbounded leftward movement because prominence factors

require the latter. I suggest that since the most natural movement in SOV,

COMP-final languages is rightward (condition (i)), these languages will

avoid leftward movement by marking an element as prominent without movement

if possible, by using particles for example. However, scope factors some-

times demand movement. In such cases, the direction of the movement is

determined by a syntactic criterion, word order. Therefore a universal

theory of movement should take into consideration both syntactic and seman-

tic factors. Further refinement of these factors will depend upon examina-

tion of movement in a wide variety of languages.

 308

FOOTNOTES TO CONCLUSION

1. Recall the one case where /-I/ clauses were considered dominant

(section 3.3.4). I claimed, however, that this case is different

from the English cases. In other words, the rider "in some context"

is not necessary for Navajo.

2. Langacker says "precede" and "command" neutralize prominence. It

seems to me, however, that unbounded rightward movement decreases

prominence.

3. I use the notation (NP) to describe the fact that S may be either

an object or a subject, and when an object, it may or may not be

an NP.

4. Recall that in the cases of WH movement in direct discourse sentences,

many sentences were ambiguous. I claim that if there were a pre-S

COMP, and movement were to COMP position, these sentences would not

be ambiguous. The ambiguity therefore argues against generating WH

question particles separately in a pre-S node.

 309

REFERENCES

Akajian, A. and S. Anderson (1968) "How To Do Things With Words, or

Navajo Made Harder," MS, MIT.

Bach, E. (1971) "Questions," Linguistic Inquiry II, 153-166.

Baker, L. (1970) "Notes on the Description of the English Questions:

The Role of the Abstract Q Morpheme," Foundations of Language 6,

197-219.

Benally, T. (1973) "Saad T'll AlIgag At'eego Bee Na'idikidigif," Papers

On Navajo Linguistics, D.B.A., Inc., Ganado, Arizona.

Brecht, R. D. (1972) "Modal Reference and Presupposition," North-East

Linguistics Society III.

Bresnan, J. (1970) "On Complementizers: Toward a Syntactic Theory of

Complement Type," Foundations Of Language 6, 297-312.

(1972) Theory of Complementation In English Syntax, unpublished

Doctoral dissertation, MIT.

Brogan, P. (n.d.) ''Indirect Questions In English," MS, MIT.

Chomsky, N. (1965) Aspects of the Theory of Syntax, MIT Press, Cambridge,

Mass.

(1968) "Remarks On Nominalization," reproduced by the Indiana

University Linguistics Club.

(1969) "Deep Structure, Surface Structure, and Semantic Inter-

pretation," reproduced by the Indiana University Linguistics Club.

(1971) "Conditions On Transformations," reproduced by the Indiana

University Linguistics Club.

 310

and M. Halle (1968) The Sound Pattern of English, Harper and Row,

New York.

Creamer, M. H. (1974) "Ranking In Navajo Nouns," Din' Bizaad Nanfl'ih/

Navajo Language Review I, 29-38.

Elgin, P. A. S. (1973) Some Topics In Navajo Syntax, unpublished Doctoral

dissertation, University of California, San Diego.

Emonds, J. (1970) Root and Structure Preserving Transformations, unpublished

Doctoral dissertation, MIT.

Erteschik (Shir), N. (1973) On the Nature of Island Constraints, un-

published Doctoral dissertation, MIT.

Firbas, J. (1962) "Notes On the Function of the Sentence In the Act of

Communication," Journal of Studies of the Philosophical Faculty

of the University of Brno, Series A, no. 10.

(1964) "On Defining the Theme In Functional Sentence Analysis,"

(publication reference unavailable).

Greenberg, J. (1962) "Some Universals of Grammar With Particular Reference

to the Ordering of Meaningful Elements," in Universals of Language,

second edition, MIT Press, Cambridge, Mass.

Grosu, A. (1973) "On the Status of the So-Called Right Roof Constraint,"

Language 49, 294-311.

Haile, B. (1942) Learning Navaho, Vol. II, St. Michaels Press, Saint

Michaels, Arizona.

Hale, K. (1971) "Gaps In Grammars and Cultures," MS, MIT.

(1973) "A Note On Subject Object Inversion in Navajo," in B. Kachru

et al., Issues In Linguistics, University of Illinois Press, Urbana,

 311

Illinois.

,(Forthcoming A) "A Remark On English Intonation and Navajo

Particles."

,(Forthcoming B) "The Navajo Particle / /."

and P. Platero (1973) "Aspects of Navajo Anaphora: Relativiza-

tion and Pronominalization," MS, MIT [in Dine Bizaad Ninfl'igh/

Navajo Language Review I, 9-28].

Jackendoff, R. (1972) Semantic Interpretation In Generative Grammar,

MIT Press, Cambridge, Mass.

Jespersen, 0. (1927) A Modern English Grammar On Historical Principles,

vol. IV, George Allen and Union Ltd., London, England.

Kari, J. (1973) Navajo Verb Prefix Phonology, unpublished Doctoral

dissertation, University of New Mexico.

Karttunen, L. (1971) "Some Observations On Factivity," Papers In Linguis-

tics IV:1.

(1973) "Remarks On Presupposition," presented at the Texas Con-

ference On Performances, Conversational Implicature, and Presup-

positions.

Kaufman, E. (Forthcoming) "Navajo Adverbial Clauses and the Semantics of

the Complementizers."

Kiparsky, C. and P. Kiparsky (1970) "Fact," in Bierwisch and Heidolph,

Progress In Linguistics, Mouton and Co., The Hague.

Kuno, S. (1972) "Functional Sentence Perspective," Linguistic Inquiry

III, 269-321.
 312

(1972a) "Pronominalization, Reflexivization, and Direct Discourse,"

Linguistic Inquiry III, 161-196.

(1974) "The Position of Relative Clauses and Conjunctions,"

Linguistic Inquiry V, 117-136.

Landar, H. (1961) "A Note On Preferred Arrangements of Navajo Words,"

IJAL 27, 175-177.

Langacker, R. (1969) "An Analysis of English Questions," MS, University

of California, San Diego.

(1969a) "On Pronominalization and the Chain of Command," in

D. A. Reibel and S. A. Shane, eds., Modern Studies In English,

Prentice-Hall, Englewood Cliffs, N.J.

(1974) "Movement Rules In Functional Perspective," Language 50,

63o-664.

Liberman, M. and I. Sag (1974) "Prosodic Form and Discourse Function,"

Chicago Linguistics Society 10, 416-427.

Perkins, Ellavina (1973) "The Semantics and Syntax of the Navajo Particle

of Constituent Negation," MS, MIT [in Dind'Bizaad Nanil'4ih/Navajo

Language Review I, 53-63.

Platero, P. (1974) "The Navajo Relative Clause," IJAL 40, 202-246.

Pope, E. (1972) Questions and Answers In English, unpublished Doctoral

dissertation, MIT.

Reichard, Gladys (1951) Navajo Grammar, Publication of the American

Ethnological Society, 21.

Reinhart, T. (1973) "Point of View In Sentences Containing Parentheticals,"

MS, MIT.
 313

Ross, J. R. (1966) "A Proposed Rule of Tree-Pruning," in Mathematical

Linguistics and Automatic Translation, Report NSF-17, the Computa-

tion Laboratory of Harvard University.

(1967) Constraints On Variables In Syntax, unpublished Doctoral

dissertation, MIT.

Sapir, E. (1936) "A Type of Athabaskan Relative," IJAL 2, 136-143.

Satyanarayana, P. and K. V. Subbarao (1973) "Are Rightward Movement Rules

Upward Bounded?" in Studies In the Linguistic Sciences 3.1, Uni-

versity of Illinois, Urbana, Illinois.

Schacter, P. (1973) "Focus and Relativization," Language 49, 19-46.

Sgall, P., E. Hajicova, and E. Benesova (1973) Topic, Focus and Generative

Semantics, Scriptor Verlag GmbH, Kronberg Taunus.

Stanley, R. J. (1963) The Phonology of the Navajo Verb, unpublished

Doctoral dissertation, MIT.

Vetter, D. (1968) "A Preliminary Formulation of the Navajo System of

Complementizers," MS, MIT.

Wall, L. and W. Morgan (1958) Navajo-English Dictionary, U.S. Department

of Interior, Branch of Education, Window Rock, Arizona.

Wilson, A. (1969) Breakthrough Navajo, The University of New Mexico,

Gallup Branch, Gallup, N.M.

Young, R. and W. Morgan (n.d.) The Function and Signification of Certain

Navaho Particles, Indian School Printing Department, Phoenix, Ariz.

(1951) A Vocabulary of Colloquial Navajo, United States Indian

Service, Phoenix, Arizona.

(1971) The Navajo Language, Deseret Book Company, Utah.

 314

BIOGRAPHICAL NOTE

The author, born in New York, N.Y. on April 23, 1946, lived in

Roslyn, N.Y. during her youth and graduated from Friends Academy High

School in June, 1964. She attended Brandeis University, graduating

magna cum laude in June, 1968 with a major in philosophy and a chronic

case of linguistophilia contracted from Jay Keyser and Dave Perlmutter.

Under their guidance she began graduate work, taking time off to get

married and accompany her husband to the wilds of Belgium and New York

before matriculating at M.I.T. in September, 1971. She has taught at

the Dine' Bi'lto' Association Linguistics Workshop on the Navajo reserva-

tion, and hopes for a university career. Her article "Navajo Spatial

Enclitics: A Case for Unbounded Rightward Movement," appeared in

Linguistic Inquiry (Volume V, Number 4). She is currently rediscovering

life with her husband Jay, baby son Ian, and proud canine Waffle after

a long hiatus, the product of which you have just read.