The phonological hierarchy of the White Mountain dialect of Western Apache
Item Type text; Dissertation-Reproduction (electronic)
Authors Greenfeld, Philip John, 1943-
Publisher The University of Arizona.
Rights Copyright © is held by the author. Digital access to this material is made possible by the University Libraries, University of Arizona. Further transmission, reproduction or presentation (such as public display or performance) of protected items is prohibited except with permission of the author.
Download date 04/10/2021 08:40:28
Link to Item http://hdl.handle.net/10150/597066 THE PHONOLOGICAL HIERARCHY OF THE
WHITE MOUNTAIN DIALECT OF WESTERN APACHE
by
Philip John Greenfeld
A Dissertation Submitted to the Faculty of the
DEPARTMENT OF ANTHROPOLOGY
In Partial Fulfillment of the Requirements For the Degree of
DOCTOR OF PHILOSOPHY
In the Graduate College
THE UNIVERSITY OF ARIZONA
19 7 2 THE UNIVERSITY OF ARIZONA
GRADUATE COLLEGE
I hereby recommend that this dissertation prepared under my direction by ______Philip John Greenfeld______entitled THE PHONOLOGICAL HIERARCHY OF THE WHITE
MOUNTAIN DIALECT OF WESTERN APACHE______be accepted as fulfilling the dissertation requirement of the degree of Doctor of Philosophy______
Dissertation Director Date
After inspection of the final copy of the dissertation, the following members of the Final Examination Committee concur in its approval and recommend its acceptance:*
7? _____
L ~ > 3 - 7 ^
This approval and acceptance is contingent on the candidate!s adequate performance and defense of this dissertation at the final oral examination. The inclusion of this sheet bound into the library copy of the dissertation is evidence of satisfactory performance at the final examination. STATEMENT BY AUTHOR
This dissertation has been submitted in partial fulfillment of requirements for an advanced degree at The University of Arizona and is deposited in the University Library to be made available to borrowers under rules of the Library.
Brief quotations from this dissertation are allow able without special permission, provided that accurate acknowledgment of source is made. Requests for permission for extended quotation from or reproduction of this manu script in whole or in part may be granted by the head of the major department or the Dean of the Graduate College when in his judgment the proposed use of the material is in the interests of scholarship. In all other instances, however, permission must be obtained from the author.
SIGNED: 7 ^ 7 PREFACE
This study is based upon 13 months of field work on the Fort Apache Indian Reservation, during which time my wife and I lived in housing supplied by the Whiteriver
Public Schools near the community of Seven Mile. Prior to this residence on the reservation, a preliminary study was carried out using one informant in seminar sessions at The University of Arizona.
The field work was financed by various sources which include: The Comins Fund of the Department of
Anthropology, the Doris Duke Foundation American Indian
Oral History Project, and The University of Arizona Pre-
Doc toral Fellowship program. Their support is deeply appreciated.
My interests in linguistic anthropology were first aroused by Dr. Harland Padfield, currently with Oregon
State University; and Dr. Kenneth Hale of M.I.T. My major source of inspiration and training in phonology was
Dr. Joseph Grimes of Cornell University and the Summer
Institute of Linguistics. Many of his ideas reveal them selves here; I hope in a way which does them justice.
This interest and training was further whetted by the teaching and close personal friendship of Dr. Paul R.
iii iv
Turner, and finally channeled to the study of White Moun
tain Apache through the leadership and courses of Dr. Keith
H. Basso. Without his willingness to trust a somewhat
naive graduate student with the people he himself is so
close to, the whole thing would never have been started;
and without the training, inspiration, and confidence of
all these men it would have never been finished.
One other man who deserves special mention is the
late Dr. Edward P. Dozier, who was originally a member of
my doctoral committee. Although he did not live to see
the conclusion of my work, I feel that he had a very real
input into its production. The thoroughness of his own
work and the gentle quality of his life as I observed it
in his teaching had a real effect on me and I felt a deep
personal sorrow with his death.
The actual field work could not have been carried
out without the approval of the White Mountain Apache
Tribal Council, and its chairman at the time of the re
search, Mr. Bonnie Lupe. Even with this approval if it
had not been for the kind of cooperation of Mr. Vincent
Altaha of Whiteriver and my Apache family, the Clarence
Hawkins of East Fork, I could have accomplished nothing.
Their patient efforts to teach me the Apache language can
not be praised highly enough, and any inadequacies in the
final description are due to my inability to learn rather V than their abilities to teach. I would also like to thank Anthony DeClay, Sanni Elgo, and Kathy Begaye for their help as linguistic informants. I especially want to thank the people of East Fork for their kind reception of me and my project, and their tolerance of my presence in their community for the year of field work. One fur ther member of the Apache community deserves special mention, and that is Mrs. Mary V. Riley, Councilwoman from
Seven Mile. Her help and friendship contributed much to my work and our stay on the reservation.
There, of course, were others who contributed to the field work experience. Mr. Dick Cooley continually reminded me of all I did not know, and Mrs. Elizabeth
Cooley's deep personal involvement with the Apache people served as a model and goal for my own relations with them.
The school district with their provision of a job for my wife and a place to live also deserves our thanks, and the intellectual stimulation of fellow anthropologist,
Dr. Michael W. Everett made the field experience especially rewarding.
I also wish to thank Dr. Timothy Smith of the
University of California at San Diego for his generous help in the area of acoustic phonetics. Both his advice and laboratory apparatus are greatly appreciated. Simi larly, Dr. Harry Hoijer of The University of California Vi at Los Angeles generously sent copies of some of his field notes for which I am thankful. And, of course, I want to thank Mrs. Hazel Gillie for the final typing of a most difficult manuscript, and all the problems associated with it.
Finally, I cannot put into words the deep appre ciation and debt which I owe to my wife, Sally. She supported us through the entire graduate school period, suffered through course work in linguistics in order to understand my work better, and was responsible for all but the final typing of this dissertation. It is in many respects hers as much as mine for she has aided at every
step of the way, and I can never repay her for her love and confidence in me. TABLE OF CONTENTS
Page
LIST OF T A B L E S ...... xi
ABSTRACT ...... xii
1. INTRODUCTION 1
The Theoretical B a s e ...... 3 Goals of the S t u d y ...... The Structure of Language ...... The Phonological Hierarchy ...... 10 The Phoneme ...... 10 The Syllable ...... 14 The Foot ...... 16 The Contour ...... 17- Other Higher Level Units ...... 18 Hierarchical Arrangement ...... 18 Syntax and Phonology ...... 19 Phonology and the Ethnography of Speaking . 20
2. THE PEOPLE, THE LANGUAGE, THE LITERATURE, THE INFORMANTS ...... 22
The P e o p l e ...... 23 Number of Speakers...... 24 The Language Situation at Fort Apache . . 26 The Language...... 2? Comparative Studies ...... 28 The Literature...... 29 Studies in Navajo ...... 33 Other Apachean Languages ...... 37 The Informants...... 38 Data C o l l e c t i o n ...... 39
3. UMA PHONETICS ...... 41
C onto id P h o n e s ...... 42 S t o p s ...... 42 Fricatives...... 47 Affricates...... 49 S e m i v o w e l ...... 51 N a s a l s ...... 51
vii viii
TABLE OF CONTENTS— Continued
Page
Vocoid Phones ...... Vocoids ...... Prosodic Features and Boundary Phenomena Nasalization ...... Boundary Markers ...... Stress ...... T o n e ...... Intonation ...... L e n g t h ...... Summary ...... 4. PHONOLOGY AND ITS RELATION TO GRAMMAR AND LEXICON IN W M A ...... 57
Conditioning Environments ...... 57 Phonologically Conditioned Variation and Distribution ...... 58 Grammatically Conditioned Variation and Distribution ...... 61 WMA Morphology ...... 62 Phonological Variants Conditioned by Grammatical Classes ...... 70 Lexically Conditioned Variation and Dis tribution ...... 73 Lexical Environments in WMA Phonology . 77 S u m m a r y ...... 83
5. THE PHONEME LEVEL ...... 84
The Phonemes of White Mountain Apache . . . 85 Phonemic Contrasts ...... 89 C o n s o n a n t s ...... 92 Simple Stops ...... 92 The Stops: Comments and Comparisons 106 Fricatives ...... 122 The Fricatives: Comments and Compari sons ...... 136 A f f r i c a t e s ...... 144 Resonants ...... 149 The Resonants: Comments and Compari-; s o n s ...... 153 Vowels ...... 154 Vowel Phonemes ...... 154 Allophonic Distribution ...... 158 ix ,TABLE OF CONTENTS— Continued
Page
The Vowels: Comments and Comparisons . 168 Summary...... 1?6
6. HIGHER PHONOLOGICAL UNITS: THE SYLLABLE . . . 177
Syllable Structure ...... 178 Boundaries ...... 178 Syllable.Boundary Rules ...... 183 Nuclei...... 184 Syllable Types ...... 191 Prosodic Elements ...... 205 T o n e ...... 205 Stress ...... 212 Length ...... 223 Nasalization ...... 227 Prosodic Features and the Feature Mode 227 Phonotactics ...... 229 Phonotactic Limitations ...... 235 Problem Areas ...... 241 Inadequacies in the Data ...... 248 Summary ...... 249 7. HIGHER PHONOLOGICAL LEVELS: THE FOOT, METER, LINE, AND C O N T O U R ...... 251
The Foot Level ...... 251 Foot S t r u c t u r e ...... 252 Foot Boundaries...... 253 Foot P r o s o d i e s ...... 257 The Foot and the M e t e r ...... 266 A Sample T e x t ...... 267 The Meter L e v e l ...... 273 Meter Structure...... 273 Meter Prosodies...... 274 Meter Types ...... 277 Meters as Contrastive Units ...... 281 The Meter and the Ethnography of Speaking . 281 The Line L e v e l ...... 282 Line Structure ...... 283 Line P r o s o d i e s ...... 284 Line T y p e s ...... 285 The Contour L e v e l ...... 287 Contour Structure ...... 288
I X !• TABLE OP CONTENTS— Continued
Page
Discussion and Summary ...... 289 Higher Level Units in Other Athapaskan L a n g u a g e s ...... 289 Summary...... 291
8. CONCLUSION...... 293
' Summary ...... 293 Significant Aspects of WMA Phonology . . 295 Points of Confusion, Ambiguity, and Inadequate D a t a ...... 296 The Study in the Light of its Goals . . 298 Practical Application ...... 299 A Proposed Orthography ...... 300 Present Orthography ...... 301
LIST OP REFERENCES...... 303 LIST OF TABLES
Table Page
1. WMA phonetic inventory ...... 43
2. The phonemes of the White Mountain dialect of Western A p a c h e ...... 86
3. WMA phonemes with universal distinctive feature specifications ...... 88
4. Possible consonant sequences ...... 231
5. 'D' effect stem initial changes in seven Apachean languages ...... 246
xi ABSTRACT
White Mountain Apache is a dialect of the Apachean branch of the Athapaskan stock of the Na-Dene family of
North American Indian languages. It is spoken by three to four thousand speakers who live on the Fort Apache Indian
Reservation in north-central Arizona.
This dissertation is a description of the phono logical system of this language using Pike's tri-modal model of language structure. In using this model it em phasizes the notion that language is composed of units which relate to each other in hierarchically arranged levels. The units have features, manifestations, and dis tribution.
In the phonological system of White Mountain Apache there are at least six levels of units: phoneme, syllable, foot, meter, line, and contour.
At the phoneme level there are 34 contrastive seg mental phonemes, 28 consonants, two resonants, and four vowels. To adequately describe these phonemes, grammatical and lexical as well as phonological information must be used.
There is one syllable structure with two variants, one of which occurs at the nucleus of a phonological foot.
xii xiii and the other which occurs elsewhere. These two types further correlate with the syntactic classes of stem and non-stem morphemes. Syllables are further characterized by the prosodic features of stress, tone, nasalization, and length. Stress and length are totally predictable with reference to type and phonemic shape, while tone and nasalization are generally distinctive non-predictable features. Tone, although basically a non-predictable feature, has certain phonetic variants which relate to syllable shape.
Syllable shape itself is conditioned or re stricted with relation to adjacent syllables, the gramma tical class of the morphemes represented in the syllables, and the types of syntactic boundaries between such mor phemes. These restrictions upon syllable shape are repre sented in a number of general phonotactic rules.
The foot level is represented by a unit composed of a cohesive series of syllables centered upon a nuclear syllable which carries primary stress. This nuclear syl lable is always the grammatical stem syllable. The foot has prosodic features which include rhythm, tonality, and timing. Its variant forms relate to whether it has onset and coda syllables, and its boundaries are marked by a variety of phonetic phenomena. xiv
The meter level is represented generally by a unit composed Of a series of feet centered around a more heav ily stressed foot, and bounded by audible pause. There are four contrastive meter forms: introductory, normal, parenthetical, and transitional. These are distinguished by differences in their prosodic features of timing and tonality. It is suggested in the dissertation that the locus of differences in discourse form in White Mountain
Apache is to be found in the meter, and there are probably other types of meters in the language which are not iden tified in this study. In relating the phonological hier archy to discourse type in this manner, the dissertation provides the basis of a more complete ethnography of speaking for this society.
The line level is represented by a series of meters which build to a peak, or fall to a conclusion.
These are labeled as non-final and final lines respec tively. These units are bounded by longer pause and some times breath.
Finally, the contour is composed of at least one final line and may consist of one or more non-final lines followed by a final line. It is a unit most generally bounded by long pause and breath.
In the conclusion a practical orthography for
White Mountain Apache is suggested. CHAPTER 1
INTRODUCTION
This is a description of the sound system of the
White Mountain dialect of Western Apache. It is a study in linguistic anthropology. It accepts a definition of culture as being the ideational framework which underlies the phenomenal level of behavior exhibited in the various life ways of the peoples of the world (Goodenough 1964:
11-12). As such, it is a premise of this study that part of this ideational level, perhaps its most basic and un conscious part, is the language spoken by the members of any particular society, and that a description of a lan guage or any part of a language is not only a contribution to linguistics, but a contribution to ethnography.
Given this approach an anthropological study of the language spoken by any society in the world must con form to the major criterion of good ethnography, i.e., it must be a statement of what one has to know in order to act
(in this case speak) appropriately in any role recognized by that society (Goodenough 1957). Such an ethnographic description should come as close as possible to stating what people unconsciously.know about their culture, but it 1 2 need not claim to be a direct equivalent of that competence.
That is, it need not claim cognitive validity, it need only provide for equivalent behavior.
I have phrased this introductory statement in the foregoing manner for a variety of reasons. First, because
I am genuinely skeptical that cognitively valid models are possible. It is a fairly well accepted conclusion that there is no such thing as cognitive sharing among the mem bers of any society (Wallace 1970: 27-32). Rather, each person in a society generates acceptable behavior on the basis of equivalence structures. Even the notion of compe tence as introduced by Chomsky (19651 3-4) is directed at a non-existent "ideal speaker-listener.11 This being the case competence models must by definition be only struc tural models, but it is not the case that structural models can be wholly arbitrary. They must never contradict cog nitive judgments when these native judgments are available, and they must conform to what we know about the nature of human language.
Second, in talking about a competence model of language, we are dealing with that part of culture which is probably the farthest from the conscious level, and therefore the least amenable to testing for cognitive va lidity. Therefore, linguistic descriptions probably cap ture far less in the way of cognition than do ethnographic models dealing with other kinds of cultural knowledge. 3 I present this description, then, as an hypothesis which I have formulated on the basis of an impressionistic auditory observation of the phenomenal level. I believe it
fully and consistently accounts for what an Apache has to know in order to produce utterances in his language, but
it does not necessarily duplicate this process.
If it does this, it may be considered to be de
scriptively adequate. Furthermore, since I make use of
certain general hypotheses about the nature of language
(e.g., there are obligatory contrastive units such as con
sonants and vowels, phonemic units are composed of bundles
of features which do not act independently and change only
in certain ways relative to their environments, natural
classes of phonemes exist and are defined by their commonly
shared features, etc.) it may be said that this description
is in some sense explanatory of White Mountain Apache (WMA)
phonology (Chomsky 196?: 29; Postal 1968: 53ff•)•
The Theoretical Base
The form that this description takes is essentially
that of what has been termed taxonomic linguistics. Rea
sons for this are contained in what I conceive of as the
basic assumptions and goals of this study.
Goals of the Study
1. There is a genuine belief on my part that descrip
tions which are essentially taxonomic (structural or item and arrangement) in form reveal interesting and significant facts about language which generative or process models tend to ignore. Among these is the notion of units, hier archy, and structure. This is not to deny that processes are also very much a part of language, but both approaches
reveal significant findings about languages and language and seem to be complementary rather than mutually exclusive models of description. For example, the validity of the
autonomous phoneme as a unit has recently been reaffirmed
by at least one transformationalist (Schane 1971).
2. The job of taxonomic description is prerequisite
to process description. As Grimes (1969: 18) comments:
As a stage in the task of finding out what is going on in a language, however, a taxonomy of sounds and a taxonomy of output grammatical forms are not only a good idea; any linguist who tries to work without them is in the same position as a thief who tries to rob a bank without first casing the joint.
That process phonologies use taxonomic studies as a base
line from which to work is witnessed by Stanley's (1969a)
use of Hoijer's (Sapir and Hoijer 1967), and Young and
Morgan's (1964) taxonomic descriptions of Navajo. Since
this is the first complete treatment of WKA phonology, I
too hope it can function as the baseline for further lin
guistic studies.
3. One purpose of this study is to provide a descrip
tion for the use of cultural anthropologists in their
studies of the Western Apache. Along with this is the 5 development of a scientific orthography for the transcrip tion of Apache linguistic materials. Cultural anthropolo gists have been working with the Apache for a number of years, but a cursory check of the published material is enough to demonstrate the inconsistency of the transcribed material. A taxonomic description using segmental phonemes provides the necessary orthography, and provides the eth nographer with a clear description which he can make use of.
4. This study is somewhat theoretically innovative in that it attempts to provide as complete a picture as pos
sible not only of WKA segmental phonemes but also complex phonological units which are strings of phonemes. The taxonomic framework seems to be the best one available currently to handle these higher level phonological units.
I do not mean to imply that this is the first study to in clude this kind of description, but a check of the litera ture easily demonstrates the relative lack of attention paid these levels in most phonological statements. Fur thermore, by emphasizing these higher level patterns I am describing essentially what Hymes (1962; 1964) calls
"message form," an important aspect of any speech situa
tion and directly relevant to an ethnography of speaking.
Thus one of the goals of this study is to provide a basis not only for further linguistic studies, but studies which are concerned with this area of Western Apache ethnography. 6
5. My basic goal is to provide a good ethnographic description of the sound system of V/MA. I grant that the process theory reveals many regularities in phonological patterning (the analogous patterning in allophonic mani
festations of phonemes in the same class, for instance),
but this approach to linguistic analysis is in a constant
state of flux. It has not even been able to decide upon what are the universal classificatory distinctive features
(Harms 1968: 22). Since this study is not intended to
develop major theoretical innovations in linguistic theory,
but is aimed toward providing a sound description of WHA
the taxonomic framework seems to provide the best form with which to integrate a variety of notions all of which
can contribute towards an accurate picture of the sound
system of V/MA.
The Structure of Language
This study is based upon a theory of language
structure which posits three interlocking hierarchies. A
phonological hierarchy, which is the manifestation of lan
guage; a grammatical hierarchy which deals with its struc
ture and distributional properties and a lexical or seman
tic hierarchy, which may be considered as providing its
distinctive features (Pike 196?: 14]).
This study is concerned primarily with the form
or manifestation of a particular language. But it must be 7 clearly understood that these hierarchies are not separate components which relate to each other in some kind of flow chart configuration, but are three interlocking and highly interdependent units (Pike 1958), no one of which can be taken as primary. In this assumption lies perhaps the primary difference between this theory of language struc ture and that of the transformational-generative approach, the latter holding that syntax is primary with phonology a function of or a medium for syntax, and semantics the in terpretation of syntax (Chomsky 1965: 16).
Various linguists have argued for some time that language is an interlocking and interdependent structure
(Pike 1958; Longacre 1964), and recently Turner (1970a,
1970b) has presented evidence which indicates that cer tainly phonology is necessary to the interpretation of syntactic structures as much as vice versa. Even within the transformational school some writers are beginning to see the interlocking nature of phonology and syntax as opposed to the totally dependent nature of phonology.
Cook (1971a) has demonstrated for Sarsi (also spelled
Sarcee), a northern Athapaskan language, that some phono logical rules must be ordered prior to certain syntactic rules in order to correctly predict the assignment of tone to surface syntactic structures.
Just as to how complete the interlocking between these two is seems to be relative from language to 8 language. Certainly in many languages phonology is rela tively dependent upon grammar, phonological junctures paralleling precisely grammatical ones, yet in other lan guages phonological cues are the prime markers of syntac tic boundaries, and in others they comprise two separate systems. Pike (196?: 569-73) gives a rather complete re view of the possibilities here.
Language then considered in this way is composed of three hierarchies of levels of units. Each level in a hierarchy is made up of strings of lower level units which function as groups at higher levels. Each of these units like language itself has three modes which serve to characterize it with respect to other units at the same level or other levels in the hierarchy. Again, these three modes are the manifestation mode, the variant forms which the unit can have; the feature mode, the distinctive or contrastive features which distinguish it from other units; and the distribution mode, the locations in which that unit can occur (Pike 1956). Linguistic units are thus defined in terms of what forms they take, what iden
tifies them, and where they occur. This is their value
(D e Sa.ussure 1966), the unit being equal to the sum of its manifestations, features, and distribution, or U = M F D
(Pike 1956).
This view of language presumes that all three hierarchies are formed in an analogous manner, the lowest 9 level units in each hierarchy combining to form higher
level units according to specified sets of rules as de
fined in the distribution mode. Such a view also assumes
that units of language are segmentable, or have boundaries.
Of course this latter is only a theoretical assumption.
Acoustic analysis of speech demonstrates that not all
units (especially phonological units) are clearly defined.
Units often shade into each other with ambiguous boundaries
(Grimes 1969: 70). But there are generally enough units
at any one level which are clearly defined that they can
be used as models to assign the ambiguous unit boundaries
on the basis of analogy (Grimes 1969: 70). The clear cut
cases may be considered to exemplify the unit in question
in its unaffected form, and the ambiguous ones to be the
result of environmental conditioning (Grimes 1969: 70).
These unclear cases are defined by the manifestation mode
in terms of rules which specify the variant forms the unit
can and does take in regard to the distribution mode. So
far as the phonological hierarchy is concerned, these
rules are allophonic rules in the case of the units at the
lowest level of the hierarchy, the phoneme; and phonotac-
tic or morphophonemic rules in the case of syllables and
other higher level phonological units. 10
The Phonological Hierarchy
Boundaries in the phonological hierarchy are marked by a number of devices. Pause, loudness, breath, echo vowels, aspiration, voicelessness, etc., all help to identify phonological units. I would like to briefly dis cuss the kinds of units which might be found in the pho nological hierarchy, and the possible structures they might have.
The Phoneme
The basic unit in the phonological hierarchy is the segmental phoneme. Its structure is that of a simul
taneous bundle of features which are actualized as motor, acoustical, or auditory phenomena (Jakobson and Halle
1956: 8). Some of these features are distinctive and
serve to identify the phoneme as different from other pho nemes. These comprise its feature mode. Others are pre dictable or redundant and vary in relation to features of adjoining phonemes and serve to support distinctive fea
tures in certain environments. These variations in the
structure of the phoneme constitute the manifestation mode
of the phoneme. Finally, the locations in which the pho neme can and does occur constitute its distribution mode.
The Naturalness Condition
It seems unquestionable that the phoneme as a unit
has at least a degree of cognitive validity (Sapir 1949a), 11 and thus must be recognized in some form as part of a
basic theory of language. It also seems that because of
the psychological validity of this unit the structure of
the phoneme is not arbitrary. Phonemes are not the way
they are simply because the linguist prefers to make them
so, nor are phonemes the way they are because it makes a
simpler or more elegant system. Phonemes are structured
within constraints set by the nature of language, a struc
ture which is ultimately rooted in their phonetic manifes
tations.
What I am talking about here is what Postal (1968:
55) terms, "the Naturalness Condition." It is difficult
to state an explicit definition, but to quote Postal,
"The Naturalness Condition is concerned with a fundamental
fact about human language . . . namely, the relation be
tween phonological and phonetic structures is a natural
one"(Postal 1968: 56, emphasis his). That is, the rela
tion between phonological and phonetic structures is
subject to a set of limited alternatives which tend to be
manifested over and over again in the languages of the
world. Phonemitizations which disregard phonetic simi
larity, linear relations among phonemes, phonetic plausa-
bility, and surface structure facts violate this condition
and produce phonemic systems which are neat but wholly
arbitrary. Linguistic description is like a game, it must 12 be played by rules which are contained in the basic struc ture of language; part of the present problem is, however, that we are still not sure of what all these rules or constraints upon play are. But as they become more ap parent, we must conform to them in order to have descrip tions which adequately characterize human languages.
Most modern linguistic theories have recognized this fact. The transformational approach has even given us the term "Naturalness Condition," while Pike (19^3;
194?a) has insisted upon this for some time under the names of phonetic plausability and phonetic similarity. This insistence upon some type of a natural relation between phonetics and phonemics thus eliminates any approach which treats separate features of different phones as manifesta tions of "phonemes," as is suggested by Harris (1944).
Following Harris' suggestion, for example, one linguist
(Durbin 1964) working on San Carlos Apache has suggested a phoneme with glottal stop, voicing, and glottalization as its allophones simply because the glottis is in some way involved in all three. Such an approach may produce a simpler system, but it is wholly arbitrary and does nothing to relate one language to another in some over-all theory. Such a view might justify making a phoneme of alveolar articulation with variant forms in a stop as op posed to a fricative if it would make a simpler system. 13 There seems to be no support for the psychological validity of such units given what we know experimentally about how people react to segmental phones (Jakobson and Halle 1956:
3-6). Such assumptions about the nature of phonemes seem to be wholly unmotivated in observable language data. In fact, further use of such notions leads to the invention of phonetically unrealized segments whose sole function is to mark other neighboring segments for feature changes, such as Lamb apparently does in his stratificational gram mar if we can believe Postal's interpretation (Postal 1968:
38ff). In this respect transformational and Pikian taxo nomic linguistics seems to see eye to eye. In both views phonemes have some reality and although features can change relative to their environments, features remain part of the phoneme and not variants of the phoneme.
Granting then that there is some validity to con sidering phonemes as units composed of simultaneous but unsegmentable features, the question seems to be just how directly phonemic units are related to the phonetic level.
Are they only indirectly related through a long series of processual changes, or are they directly related through a simple statement of their environmental variants. In this study the latter position is adopted, but this does not mean that the other view is wholly incompatable and should be rejected out of hand. Item and process are the 14 complement, not the antithesis, of item and arrangement
(Hockett 1954).
The phoneme, then, is a unit at the lowest level in the phonological hierarchy of a language. It is mani fested by a variety of specific phonetic forms which are conditioned by their environment. This environment may be defined either in terms of the phonological hierarchy, or because the hierarchies interlock in terms of either of the other two linguistic subsystems (Pike 194?a, 194?b,
1958; Postal 1968).
Units higher than the phoneme have boundaries, peaks or nuclei, and sometimes onsets and codas. In fact, above the syllable level they may have a structure of several peaks.
The description of these units involves a state ment concerning the structural arrangement of the lower level units which combine to form the higher ones, prosodic elements which accompany this arrangement, and character istics of their peaks and boundaries as these elements are part of their feature, manifestation, and distribution modes.
The Syllable
Phonemes are arranged into syllables. Syllables contain at least one phoneme with resonant characteristics and are produced with a maximum expulsion of air from the 15 lungs in the form of a chest pulse (Abercrombie 196?: 35-
36). This chest pulse is the peak of the syllable, and marginal areas in between chest pulses are the onsets and codas of syllables (Grimes 1969: 69). Syllables as units have features. These features consist of their structure and any phonological qualities which pertain to the syl lable as a whole, rather than to any one of its parts.
We may call these features prosodies, and note that all higher phonological units have both structural features, and prosodic features. The notion of prosodic features is perhaps most clearly defined by Firth (1957)• The structural features of the syllable are its general form in terms of the arrangement of specific phonemes. The pro sodic features may consist of tone, length, stress, and what are generally called suprasegmental phonemes. These constitute the feature mode. It is important to note that these prosodic elements are attributes of the syllable as a unit, and not of any phoneme in the syllable.
The manifestation mode of the syllable consists of variations in syllable shape relative to neighboring syllables and other factors which are predictable, while the distribution mode consists of a description of the positions particular syllable forms can have in higher phonological units. Simply because certain syllable forms are possible in short unisyllable utterances does not mean that these 16 same syllable shapes are possible in longer multisyllable utterances. Syllable shapes change relative to their en vironments just as phonemes change relative to theirs. A description of the manifestation mode of a syllable con sists of a series of rules which specify possible syllable shape sequences containing no significant higher unit boundaries. These rules are usually called phonotactic rules when they apply to all possible phoneme sequences which are conditioned by purely phonological factors, and morphophonemic rules when they apply to certain syllable shapes defined in terms of grammatical or lexical class membership. Since in this study I argue that even in a taxonomic description of language grammatical and lexical
environments are relevant to phonological variation be cause of the interlocking nature of the three hierarchies,
I can see no need to specify two sets of such rules. So far as I am concerned, phonotactic and morphophonemic rules are essentially the same, the former dealing with syllable shapes relative to phonological environments, and the latter with syllable shapes relative to grammatical environments.
The Foot
Generally syllables are arranged in linear patterns characterized by certain timing features which are a re
sult of the ordering of length and stress prosodies. These 17 patterns are units of rhythm which can be called feet.
The feature mode of the foot includes structural features such as the kind of syllables and the ordering of length or stress syllable prosodies which make them up and other prosodic elements and boundary markers like hesitation, slowing of production, decrescendo, nasalization, vowel harmony, consonant harmony, or perhaps sharpness of ar ticulation (Grimes 1969• 6?)•
The manifestation mode includes actual patterns of these feature properties, and the distribution mode in volves their location with respect to like units in higher phonological levels.
The Contour
Feet in all languages seem to be arranged into patterns characterized by final intonation on the last foot or feet prior to breath or long pause. These breath segments or contours have structural features which include arrangement of feet and stress patterns as well as pro sodic features which may include beside final intonation such things as final aspiration, echo vowels, devoicing of final syllables, and decrescendo.
The manifestation mode of these units may be dif ferent structures relative to different discourse types or cultural function, and the distribution mode may relate to 18 their placement at the beginning or end of some particular discourse type, etc.
Other Higher Level Units
Many languages also seem to have several levels between the foot and the contour. In WMA, for example, I find two which I term the meter and the line. The meter is composed in WMA of from one to three feet bounded by short but observable pause. Whether meter is the best term for this unit is debatable. Grimes (1969: 33) sug gests that the use of the term 'meter* should imply a highly regular system of stresses or rhythmic units such as in poetry, and I am not sure that WMA would meet this standard, but since no other term seems available I will continue to use it.
The line is another higher level unit. It is in termediate between the meter and the contour. The line in
WMA is a series of meters bounded by fairly long pause.
The nucleus of the line is a syllable with extra loud stress, which if high tone is extra high. It has at least two forms, a medial form which occurs in the initial and medial part of contours, and a final form which forms the final part of a contour.
Hierarchical Arrangement Because of the hierarchical arrangement of units in language, boundaries of higher level units coincide with 19 those of lower level units, the higher one being dominant.
For example, the end of a contour is also the end of a line, meter, foot, syllable and phoneme, and since the contour is the highest level represented, it boundary characteristics will be readily apparent while those of lower level units will be obscured. Similarly at the nucleus of any higher level unit will also be the nucleus of all lower level units. Thus, the nuclear syllable of a contour will also be the nuclear syllable of a line, meter, and foot. This also will obscure 'normal* charac
teristics of the nuclei of these lower level units. The
boundary markers and other prosodic elements of higher
level units are likely to obliterate or accentuate the
prosodic elements of lower level units.
Syntax and Phchblogy
We might roughly call these levels of foot, meter,
line, and contour the phonological word, phrase, clause,
and sentence in that usually there is some kind of rela
tionship between these phonological units and the various
levels of the grammatical hierarchy. As noted above, how
ever, this correspondence is not always isomorphic.
It is interesting that so far as taxonomic lin
guistics is concerned, Grimes (1969• 18-19) maintains that
phonological taxonomy in the purest sense is to be kept
independent of grammar, and that there is a second 20 taxonomic stage which relates the two in what he calls a morphophonemic taxonomy. The next stage "beyond this is systematic phonology.
Perhaps then, in Grimes' view, this is not a study in pure taxonomic linguistics, for I follow more closely
Pike's idea that even in the most "basic level of descrip tive phonology the larger phonological and grammatical environment is also relevant (Pike 1947a: 89). In taking this view I vwill rely upon purely phonological information to explain as much as possible, but where grammatical facts or even lexical facts are known to be relevant I will make use of them. This follows from the notion of interlocking hierarchies, it being impossible to totally exclude the other hierarchies from the description of any one hier archy (Pike 1947b). I will continue this discussion of the relation between phonology, syntax, and lexicon in greater detail in Chapter 4, giving explicit reasons why grammatical and lexical environments are necessary in the treatment of WMA phonology.
Phonology and the Ethnography of Speaking
The notion of phonological hierarchy certainly can be related to current concerns in cultural anthropology.
Hymes (1962; 1964) has discussed the possibility of answer ing the question, "What does one have to know in order to 21 use language in a community?11 This implies more than just
the generation of sentences which are enunciated correctly,
but also involves the knowledge of when to speak, what to
say, and how or in what form to say it. The notion of
message form is essentially a product of the phonological
hierarchy. What are the kinds of feet, lines, and contours
that make appropriate speech in any given role on any given
occasion. The phonological hierarchy then extends to dis
course form or type and is a part of the ethnography of
speaking.
Recognizing that discourse form is the highest
level of the phonological hierarchy and connects this
hierarchy directly to other cultural factors, I would say
that no complete statement of phonology then can theoreti
cally exclude this level. For practical reasons, however,
I will not attempt any description beyond that of the con
tour, setting up the entire description in such a way
though that further investigation may later add this
dimension to the study. CHAPTER 2
THE PEOPLE, THE LANGUAGE,
THE LITERATURE, THE INFORMANTS
In the American Southwest there are a number of societies which are linguistically and culturally related to people who live in Canada who speak varieties of a language stock known as Athapaskan (also spelled Athabaskan, Atha bascan, etc.). This stock is part of the Na-Dene family which may also include the northwest coast languages Eyak,
Tlingit, and Haida (Krauss 1964: 128). These southern
Athapaskans speak what is now called Apachean (Hoijer 1943:
38); and consist presently of groups known ethnographically as the Kiowa-Apache (K-A), Jicarilla Apache (Jic), Mesca- lero Apache (Kesc), Chiracahua Apache (Chir), Navajo (Nav, also spelled Navaho), and Western Apache (V/A [Goodwin 1942:
1 ; Hoijer 1943: 38]). It is probable that all these soci eties except the Kiowa-Apache speak languages which are mutually intelligible within a short period of adjustment.
There were once other societies and languages within this family, such as the Lipan Apache, for instance, but they are now extinct or have been completely assimilated by other Apachean groups. This study deals with one dialect of Western Apache.
22 23
The People
The Western Apache consisted of five groups which in turn were divided into 20 geographic divisions that can he called hands (Goodwin 1942: 4-7). The five groups each seem to have had a low level of integration based upon a feeling of linguistic and cultural commonality (Goodwin
1942: 7). Even though each group spoke a different dia lect, it seems probable that these dialects were closer to each other than any one was to any of the neighboring
Apachean dialects. It is on the basis of this linguistic similarity plus two other shared cultural traits that ethnographers have come to call them collectively the
Western Apache. These other traits include the develop ment or incipient development of matrilineal clans, and the practice of agriculture as a significant supplement to hunting and gathering. This development of agriculture has been casually related to the development of the clan system (Kaut 1957? Pox 1967! 91-2). For further socio logical and ethnographic information see Goodwin 0.942;
1971), Basso (1969; 1970) and Everett (1971).
One of the five groups which composed the Western
Apache, the White Mountain group, consisted of two bands; the Eastern White Mountain band and the Western Mountain band, It is the dialect spoken by descendants of these two bands that I will be concerned with. They presently 24 live in the eastern portion of the Fort Apache Indian
Reservation primarily around the town of Whiteriver in north-central Arizona. WMA's closest Apachean relatives are the dialects spoken today in Cibecue, Camp Verde,
Payson, and on the San Carlos Reservation. Just how di vergent these other dialects of Western Apache are as com pared to WMA is not precisely known. Only one study has been done comparing any of these dialects, that of Hill
(1963) for White Mountain and San Carlos. However, her study is relatively ad hoc and limited to minor phonolo gical correspondences, some of which according to my data do not seem valid. Probably the most extensive differences are morphological in nature, such as the use by speakers of WMA of /nohlf/ for the first person plural as opposed to the preference for /nee/ for the same morpheme by speakers of San Carlos; or the use of /log/ for 'fish1 by
speakers of San Carlos, but the preference of /pis/ by speakers of WMA for that morpheme. There are, of course,
some phonological differences as well, but not as many as say, between Western Apache and Navajo. The most striking difference is the correspondence between WMA /d/ and San
Carlos /n/ or /nd/ (Hill 1963).
Number of Speakers
There are no accurate population statistics which make any attempt to classify present day Western Apaches 25 by linguistic group, but Everett (n.d.) lists the follow ing community populations for the Port Apache Reservation.
Whiteriver 1730 Cibecue 902
East Fork 723 Canyon Day 661
Seven Mile 280
North Fork 252
Cedar Creek 230
Carrizo 129 These figures are not particularly useful linguistically however, because population movements, particularly toward
Whiteriver from other areas of the reservation have served to mix dialect and group boundaries. In fact, many des cendants of White Mountain bands are presently living near
Blyas on the San Carlos Reservation (Goodwin 19*4-2: 579).
The most recent population statistics for the en tire Fort Apache Reservation places the population at
5*855 (U.S. Department of Commerce 1971: 20) as of 1969, slightly more than the figure of 5*159 taken from the tribal rolls for 1968 by Everett (1971: 53)• On the basis of all these figures I would guess that there are between three and four thousand speakers of WMA alive today. Un like some American Indian languages, it seems reasonable to suppose that it will not become extinct for some time. 26
The Language Situation at Fort Apache
Most Apache children acquire Apache as a first language, and since the majority attend public school on the reservation and return to their homes each night they have opportunity to maintain this ability. In fact, even with Headstart Programs, most of the first grade is de voted to teaching English as a second language, a job which continues throughout the entire school program.
This has produced a situation in which almost all
younger Apaches are bilingual and even in the older age
groups bilingualism probably approaches 70 percent or bet
ter. It is my impression that the only monolingual adults
are a few very old men who escaped attending boarding
schools in their youth and some women, a few perhaps as young as 40.
Each language has its place, however. English is
the language of the schools, BIA, and other ‘official1
business. Apache is the language of the home, school
ground, and ceremonial. This functional differentiation
with almost complete bilingualism has been termed diglossia
with bilingualism (Ferguson 1959; Fishman 1971: 73ff).
Fishman feels that such a situation may be rather
stable since each language has its own function and is
associated with specific roles and situations (Fishman
1971: 79). Whether this stability will continue for a 27 long period of time probably depends upon whether Apaches try and maintain separate and distinctive Apache roles, or try and emulate Anglo society to such an extent that all roles, including those associated with the home and family become modeled after Anglo patterns.
The Language
As I mentioned above, WHA is related to languages spoken in Canada. Glottochronology dates the separation of the Apachean languages from those of Canada in the neighborhood of 1,000 years ago, with a range of 600 to
1400 years divergence depending upon whether one accepts
Hoijer's original calculation (Hoijer 1956: 321) or Hymes‘ recalculation (Hymes 1957: 293)• It glottochronology has any validity, and this is somewhat dubious (cf. Goss 1965;
1966), then we may suppose that the ancestors of the present day Apacheans began to leave Canada around A.D.
900 or 1,000. Archaeological evidence does not give serious support to their entrada to the Southwest until the 14001s (Hester 1962; Vivian i960) when tree-ring dates on hogans place the Navajos in the northwest corner of New
Mexico. The conclusion of this migration probably did not take place until 1720 when historical records verify that the Jicarilla Apache were finally pushed off the plains into New Mexico by the Commanches (Secoy 1951)• It seems probable then, that the Western Apache were becoming a 28 linguistically and culturally separate people from the
Navajo, who are perhaps their closest cultural relatives
(Kaut 1957), by the 1600's. It is questionable as to when they arrived in north-central Arizona. One historian,
Schroeder (1955-1956; 1963),favors a very late arrival ca. 1800 (1963: 1?); while another, Forbes (1966: 336)» argues for an extremely early one, ca. 1400, that just does not fit the archaeological evidence.
Their late arrival is further supported by the linguistic evidence of the mutual intelligibility of most of the Apachean languages, as well as glottochronological measurements which place no more than 460 years between any of them except Kiowa-Apache (Hymes 1957: 294). This is a distance not even great enough to place them in the separate language category which is usually arbitrarily set at 500 years divergence (Goss 1966).
Comparative Studies
Studies using the comparative method also support the general closeness of relationship between the Apachean languages and thus their probable recent Southwestern en- trada. Hoijer (1938) finds that the Apachean languages may be classed into an eastern and a western group. The western group being composed of Navajo, Western Apache,
Chiricahua, and Mescalero; with Jicarilla, Lipan, and
Kiowa-Apache being assigned to the eastern group. The 29 most important marker in this division is the development of Proto-Athapaskan */t/ into /t/ in the western group,
/k/ in Jicarilla and Lipan, and /k/ alternating with /c/ in Kiowa-Apache (Hoijer 1938: 78). Even more recently
Hoijer (1971) has argued that the eastern and western groups minus Kiowa-Apache form merely a series of dialects.
The whole now being divided up into an eastern group com posed of Jicarilla and Lipan; the western group he origi nally enumerated; and a separate language, Kiowa-Apache, a classification which further stresses the "basic homo geneity of the Southern Athapaskan languages" (Hoijer 1938:
78).
The Literature
There has been no thorough published or unpub lished study of WHA phonology and only one study of its syntax. The syntactic study was done in the tagmemic form by Edgerton (1963), and is a classic example of this item and arrangement model. The transcriptions associated with
Edgerton’s study seem at times to be more phonetic than phonemic in form, however, several phonemes being symbol ized such as /w/, /u/, and /W/ which I find to be non- contrastive. Probably the reason for these extra phonemic symbols is that Edgerton along with Faith Hill are both
SIL workers who moved from Navajo to Apache, a fact which influenced their interpretation of the YJMA phonemic system. 30
The only scientific study of the phonology of one of the Western Apache dialects is a doctoral dissertation by Marshall Durbin on San Carlos (Durbin 1964). I find
Durbin's work unsatisfactory for a variety of reasons.
First, his attention to detail, particularly pho netic detail is scanty. For instance, he ignores pre nasalized stops of any kind when other scholars such as
Hill (1963) and Hoijer (1938) note that /%d/ is one of the distinguishing features of San Carlos. He also ig nores the complementary distribution of voiced and voice less unaspirate stops, and insists upon contrasts between
[h] and [x] and between [w] and [g] neither of which seem valid contrasts in WMA, and thus are probably not valid in San Carlos.
Second, Durbin's approach is rigidly structural
(he was a student of Trager's) and follows Harris' com- ponential features approach which produces a highly arbi trary system that in no way conforms to phonological naturalness. For example, Durbin (1964: 25-26) posits a glottal phoneme which he indicates with the symbol /'/.
It has five variants which using his phonetic symbols are
[?"'], unreleased glottal stop, [?], tense released glottal stop, [3"], lengthened tense glottal stop, and [^], glottal trill (voicing; Durbin 1964: 25-26). Two of these
[%] and [A3 are interesting in that the former is said to 31 occur simultaneously with stops and thus is actualized as the pharyngeal air mechanism and the latter occurs simul taneously with certain fricatives to produce voiced frica tives. Thus the phone [s], a voiced alveolar grooved fricative, is represented phonemically as /s'/ while the phone [t?], a voiceless alveolar stop with pharyngeal air, is represented phonemically as /t'/• This same logic does not carry over into the vowels, however. Vowels are not mentioned as something plus voicing.
Durbin continues along these same lines and sets up a /y/ phoneme with an allophone [-], palatalization.
By such means he eliminates all glottalized stops and alveopalatal fricatives and affricates from the phonemic
inventory and comes up with a neat but arbitrary list of phonemes including 14 of what he calls segmental, semi-
segmental, and vowel phonemes; three phonemic junctures;
and six suprasegmental phonemes.
Further indication of the arbitrary manner in which he approaches phonology is indicated by a footnote
which I quote:
There is an alternate and perhaps better solution for the data given on page 1?. Instead of placing and the three glottal stop variants in the same phoneme, />/, and then stating that /»/ occurs with /s/, /!/, and /x/ to form the voiced contrasts, it would also be possible to consider [z, 1 (vd), y] as the base forms, all voiced, and then apply £], a devoicing phone which would be a variant of /h/ to these voiced 32
forms in order to derive the voiceless contrasts. Thus, /s/ would equal /zh/ and /s»/ would be equivalent to /z/ (Durbin 1964: 68).
■ Along with this arbitrary manner of treating allo- phonic relations Durbin sticks to a rigidly phonological
set of conditioning environments. As I will show, stress
in WMA and San Carlos too for that matter, is predictable from grammatical information. Durbin recognizes this fact, but refuses to make use of it. He states, "again,
it may be pointed out that in morphology, one can say that the heaviest stress falls on what is called the stem mor pheme, but there is no phonological criteria by which to
identify the stem morpheme" (Durbin 1964: 42).
Third, in his attempt to shorten his list of pho nemes, Durbin refuses to accept affricates as complex unit phonemes. Instead he describes the language as having a variety of consonant clusters; and thus a very complex syl
lable pattern; sometimes with as many as three syllable ini
tial consonants (Durbin 1964: 48). Apachean affricates are
anything but consonant clusters, there being no random pat
terns of consonant sequences anywhere in the language to
even suggest such an interpretation. In fact, Apaches have
real troubles with English consonant clusters giving support
to the notion that cognitively Apaches do not view these as
clusters, but as unit phones. For example, one of my best
informants could not produce the final English cluster 33 /st/ and consistently pronounced the word 'past' as /pees/.
Yet English words with final affricates like 'pitch' give
Apaches no trouble even though affricates in WMA do not occur in syllable final position. Evidence like this seems to support the interpretation of Apachean affricates as unit phonemes and tends to invalidate Durbin's approach.
A fourth criticism of Durbin may be focused upon his treatment of higher level phonological units. He is determined to reduce the system to as few phonemes as possible and in so doing he produces a unilevel account which squeezes the prosodies of higher levels down to the segment level in the form of juncture phonemes which mark breaks between phonological units (Durbin 1964: 36). The most serious problem here is that in positing juncture phonemes, he seems to attribute them to the word as if they were a canonical part of that word's phonological shape.
More comment on this problem is relevant under my treatment of WMA vowels, for Durbin makes juncture a cru cial factor in explaining vowel allophones.
Studies in Navajo
Going somewhat further afield I should mention various studies of Navajo, and their relevance for the study of WMA. The most basic work on Navajo has been carried out by Hoijer (Hoijer 1945a; Sapir and Hoijer 34 196?). Hoijer1s work is notable for its pioneering stance, quality, detail, and completeness. He is weak, however, in elegance and organization as has been recently pointed out by Stanley (1969b). Stanley's extremely negative re view can perhaps be marked off to transformational ego tism and we can accept as a more accurate appraisal of
this work the review by Krauss (1970). Certainly, Hoijer deserves credit for not being afraid at an early date to consider grammatical environments as relevant to phono logical conditioning although, as Krauss points out for
Hoijer's most recent work, somewhat inconsistently (Krauss
1970: 221). Hoijer may also be criticized for his incon
sistent treatment of syllabic 'n' and his insistence upon
contrasts between /y/ and /g/ as well as between /x/ and
/h/ in Navajo. Furthermore, his work warrants criticism
for its complete ignoral of stress in any of the Apachean
languages. All of these points are trite, however, when
observed in the light of his genuine contributions to
Apachean studies. His studies must be evaluated in the
light of his theoretical training and the era in which
most of them have been published. They stand as a basis
for further work and have been so used.
A second student of Navajo is Herbert Landar. In
the introduction to his work on Navajo syntax (Landar
1963: 7-8) he outlines a phonological system which treats 35 affricates, aspirated, and glottalized phonemes as con sonant clusters and introduces a unique treatment of the vowel system as having six rather than four vowel phonemes.
He presents his evidence for this six vowel treatment in a footnote to another study, but as I will point out later in the section on the vowels, the evidence is not very convincing (bandar, Ervin, and Horowitz I960: 371:n.8). bandar also has written a paper on stress (bandar 1959) which suggests that there are several degrees of phonemic stress in Navajo, but his evidence does not seem particu larly valid if grammatical factors are considered.
Recently there have been two other studies of
Navajo, both by transformationalists (Seville 1969;
Stanley 1969a). Although these have claimed to be studies of the phonology they hardly treat allophonic variation or the phonetic level, since Hoijer's work in this area is more than adequate. Rather, they are concerned with what is usually called morphophonemics; what I consider phono- tactics with grammatical conditioning.
One of these recent studies (Seville 1969) suffers some of the same arbitrary defects that Durbin's work on
San Carlos does. For example, Saville (1969: 50-51) wishes to eliminate nasalization of vowels as a phonemic property.
In order to do this she capitalized upon the fact that vowels become nasalized before syllable final /n/, a pho netic effect which does not occur in WMA. However, since 36
there are pairs like tin 'ice* and tjf 'a living object lies,'and tl_ 'hurt1 she cannot make nasalization allopho- nic straightaway. To accomplish her goal she follows a
suggestion made by Harris (1945: 241-42) and places a
final /m/ after all vowels which are nasalized but not
followed by /n/. She uses /m/ because it never occurs
'syllable finally. Thus, the morpheme 'a living object
lies' is phonemically tim in her system. This enables her
to write a rule which states that vowels become nasal be
fore nasals, which is followed by a second rule which
states that final /m/ goes to zero in all environments.
Such gymnastics are for the most part unsupported in the
surface structure and seem to violate the naturalness con
dition.
She also follows the consonant cluster interpre
tation of the glottalized stops, aspirated stops, and
affricates, a highly questionable proposition as I have
already pointed out (Saville 1969: 39-40).
A much more complete study of Navajo is that by
Stanley (1969a). It will probably serve as a model for
some time to come of the generative approach to phonology,
and is remarkable for its insights into Navajo phonology.
Stanley's study is essentially concerned with the morpho
phonemics of the verb complex, and he points out as regu
larities many facts which have been considered by Hoijer 37 and others as unpredictable irregularities. His arguments are well motivated in the surface structure of the lan guage and as one might expect from a student of Chomsky's conform well to the naturalness condition.
Other Apachean Languages
I must mention two other works which are of im portance here. First, there is Hoijer's study of Ghirica- hua Apache (Hoijer 1946c). It is typical of Hoijer’s work on the Apachean languages in general and is a good structural description. I will cite it frequently where relevant.
Hoijer has also done a number of studies comparing these languages both by glottochronology and the compara tive method, and he has published a number of things de scribing the over-all structure of Apachean morphology.
All of these I will refer to from time to time.
Finally, Bittle (1963) has provided us with a recent structural description of Kiowa-Apache. It follows
Hoijer’s approach as exemplified in his works on Navajo and Chiricahua, and although it is quite brief, it is use ful for comparative material.
There, of course, have been numerous other studies done on various Athapaskan languages. I will not mention here all of those which I have consulted and feel are rele vant, but will comment upon them as the need arises. 38
The Informants
This study is based primarily upon data elicited from two men both in their late 50's, Vincent Altaha and
Clarence Hawkins. Mr. Altaha lives in Whiteriver and is the son of a rather famous Apache leader of the last cen tury, R 14.
He and his family are originally from the Cedar
Creek area, which was inhabited in pre-reservation days by part of the Western White Mountain band (Goodwin 1942:
578, 651). The present family name, Altaha, is probably an English corruption of R 14‘s Indian name, ^kgolt^gye I * thin one1 (Goodwin's transcription); one form of this word is /’alt’SShe/. Mr. Altaha is an army veteran of t World War II, and a past tribal judge. He is highly accul- turated but totally fluent in WMA. His English is good, but there is considerable interference from Apache in all three hierarchies: phonological, syntactic, and lexi cal (Weinreich 1967• 1). I do not think he ever told me what his clan was, as he tends to emphasize his descent from R 14, both because of his father's fame, and because
Mr. Altaha seems to prefer Anglo patterns of kinship reckoning.
Clarence Hawkins is the matrilineal grandson of a Mexican captive woman who lived in East Fork in the late
1800's. As such, his clan is /nakaiye/ 'Mexican ones (lit. 39 they move around from place to place ones).1 He comes from a family which has lived in East Fork for at least three generations, and he is married to a woman of the majority clan in East Fork, /naadol^oosn/ 'slender peak standing up people.1 His wife's people have farmed in East Fork for over 100 years, and were part of the Eastern White
Mountain band.
Of secondary importance as informants were Mr.
Hawkins' wife, Florence, who is also in her late 50's and is monolingual; Anthony DeClay, a teenager from Whiteriver;
Anderson Hawkins, Mr. Hawkins' second youngest son who is in his early teens; Sanni Elgo, a girl in her early 20's from Blyas who is descended from one of the White Mountain bands, and a Navajo girl in her 20's, Kathy Begaye.
Data Collection
The majority of the data from the segment level of the phonology was collected as words or short utterances in response to questions in Apache such as /ndeebiyaati haat^ii holzee d£l/ 'Apache that language, what is it
(that item) called7 ' Frames of various kinds were also used, especially with verb forms to test for tonal and
stress variation. Especially helpful was the surrounding
frame /do____dah/ 'negative' and the following frame
/_____doleel/ '. . . i t will be.' With nouns the preceding frame /si_____ / 'my . . .' was often used. (
40
Thus, material for the segmental phonology, syl lable, and foot levels was selected primarily from indi vidually elicited words and phrases produced at normal or sometimes lento speed. But for the meter, line, and contour I tape recorded text materials.
These texts were stories told primarily by Mr.
Altaha, most of which have been recorded in sources such as Goodwin's Myths and Tales of the White Mountain Apache
(1939).
Apaches themselves would say then that I have'ana lyzed the sound system of /ndee biyaati/ 'Apache, his language' primarily from what would be called /’aadilgo/
'he is saying it,' but that I conclude my analysis from a way of speaking called /baanaagodi’I/ 'that which he tells to him.' As mentioned above, it is my hope that once the phonology of UMA has been described in terms of discourse types like these, further studies using this as a base form can make clear how the units I identify and describe vary in other message settings. This would make for a complete statement of WHA message forms in an ethno graphy of speaking (Hymes 1962: 25). CHAPTER 3
WMA PHONETICS
This chapter consists in the main of a list of the phones observed in WMA and an explanation of the sym bols used throughout the rest of the thesis.
I will present the material in this order. First, a description of. all contoid phones, these being phones with at least partial obstruction in the oral cavity.
Following this I will present a description of the vocoids, central oral resonants (Pike 19^3). Finally, under a section entitled prosodic features and boundary phenomena
I will describe other phonetic features and their symboli zation.
I will include not only a description of each phone in terms of articulatory phonetics, but also com ments designed to place the sound in some auditory context when it is necessary. For the contoids this is not so important, but it does seem important for the vocoids as articulatory descriptive approaches to vocoids are inade quate to say the least.
41 On the following pages Table 1 gives in outline form the WMA phonetic inventory and a partial description of the phonetic features of each phone. Note that all V/KA sounds are produced with egressive air stream mechanisms.
Contoid Phones
Stops
Erab3 is a voiced bilabial pre-nasalized stop. A com parable phone in Kiowa-Apache has been described as a stop with a light nasal attack (Bittle 1963: 79). There are no sounds in English which compare with it. The nasalization is often so slight that it may sound like a very completely voiced bilabial stop, but careful listening usually indi cates the nasal articulation as well.
[b] is a voiced bilabial stop.
[p] is a voiceless bilabial stop
[p^] is a voiceless bilabial aspirated stop.
[d] is a voiced alveolar stop.
[3] is a voiced alveolar flapped stop. Its articula tion is quite rapid, sometimes approaching Spanish flapped
'r'.
[t] is a voiceless alveolar stop.
[th] is a voiceless alveolar aspirated stop. Its as piration tends to be rather heavy, probably heavier than that of most English stops. Table 1. WMA phonetic inventory.
Bi Alveo- labial Dental Alveolar palatal Velar Pharyngeal Glottal
Stops Voiced' b d g Prenasal mb Flapped a Labalized gw Voiceless 9 Unaspirated P t k Labalized kw Aspirated ph th kh Labalized khw Glottalized t? k? Unreleased t — k= 9 =
Fricatives Voiced te y g Labalized gW Palatalized gy Velarized w Grooved z z Lateral Apical 1 Lamino i Table 1. WMA phonetic inventory- -continued
Bi- Alveo- labial Dental Alveolar palatal Velar Pharyngeal Glottal
Voiceless P Y x h Labalized xw Palatalized xY Backed X Grooved s s Lateral a, i
Affricates Voiced dz dz Semi-voiced tz tz Lateral tl Voiceless Aspirated ts*1 tsh Lateral Aspirated tl Glottalized ts ts’ Lateral Glottalized tl
Semi-Vowel Voiced w
Nasals Voiced m syllabic i 9 ? Table 1. WMA phonetic inventory— continued
Bi- Alveo- labial Dental Alveolar palatal Velar Pharyngeal Glottal Voiced non- syllabic m n
Front Central Unrounded Unrounded Rounded Vocoids High Close i u Open u Mid Close e e o Open & Low Close as a 3 Open a 46
[t91 is a voiceless alveolar glottalized stop. Glot tal and alveolar stoppage being almost simultaneously released. It is not.so heavily aspirated as [th] and sometimes is difficult for the English ear to identify as a glottalized stop.
[t=] is a voiceless alveolar unreleased stop.
[g] is a voiced velar stop.
[gw] is a voiced velar labialized stop. There is no comparable English sound or sequence.
[kw J is a voiceless velar labialized stop. It is simi lar to the initial consonant cluster in English 1quit,1 but the labialization is much shorter in duration. Also it is not aspirated.
[k^] is a voiceless velar aspirated stop. It is con siderably stronger than an English ’k*.
[k^wj ig a voiceless velar aspirated labialized stop.
It is similar to the initial consonant cluster in English
•queen.•
[k9] is a voiceless velar glottalized stop. There is no comparable English phone. The glottal release is al
most simultaneous with the velar release, and quite lenis.
[k=] is a voiceless velar unreleased stop.
[9] is a voiceless glottal stop. It is quite strongly
articulated most of the time and is easily observable both
before and between vocoids. 47
[?=] is a voiceless glottal unreleased stop.
Fricatives
[b] is a voiced bilabial flat fricative. It is com parable to the Spanish bilabial in caballo, 'horse.1
[pi is a voiceless bilabial flat fricative. It is similar to the English 'f1 sound, but is bilabial rather than labiodental.
fwl is a voiced bilabial rounded fricative with second ary velar articulation. It is similar to English 'w* as in 'won't,' but has much heavier labial friction, as well as a velar-friction which is overshadowed by the labial friction. Most English speakers would mistake it for
English 'w‘.
[W] is a voiceless bilabial rounded fricative. It is similar to the way in which some speakers of English pro nounce the initial phoneme in the word 'whet.'
[z] is a voiced alveolar grooved fricative.
[s] is a voiceless alveolar grooved fricative.
[z] is a voiced alveopalatal grooved fricative.
[si is a voiceless alveopalatal grooved fricative.
[g] is a voiced velar fricative. There is no English equivalent.
[g^l is a voiced velar palatalized fricative. One might easily mistake it for an English 'y' the palataliza tion is so strong. • 1 48
[gw] is a voiced velar labialized fricative. The
velar friction is heavier than the labial friction.
[x] is a voiceless velar fricative. The friction is
not particularly harsh, and so it is easily mistaken for
an 'h*.
[xy] is a voiceless palatalized velar fricative. The
velar friction on this phone is considerably further for
ward of that for plain [x], and the friction producing
constriction runs almost the entire area of the top of the
oral cavity from the velum to the alveolar ridge. The
friction is quite pronounced and easily identifiable. The
friction is produced with the tongue mid, the tip and
blade being turned down toward the lower front teeth.
[xw] is a voiceless velar labialized fricative. The
velar friction is quite pronounced and the secondary labial
articulation is also easily audible.
[x] is a voiceless velar backed fricative. The velar
friction is much stronger in this phone than in plain [x].
[h] is a voiceless pharyngeal fricative. It has rela
tively little friction and has secondary articulation in
the oral cavity relative to the shape of the following
vowel if in syllable initial position, or the preceding
vowel if in syllable final position. It occasionally
occurs between two syllabilic nasals, in which case the
secondary oral constriction takes that of a mid central
vocoid. 49
[l] is a voiced 1amino-alveolar lateral fricative. It is similar to the first sound in English 1 little,' but has considerably more friction in its production.
[l] is a voiced apical-alveolar lateral fricative.
It is similar to the final *1 * in English ’little,’ but is not quite so syllabic.
[ll is a voiceless alveolar lateral fricative. It has heavy lateral friction.
m is a voiceless dental lateral fricative. This phone is produced much further front than plain [l], and sometimes even approaches the English interdental frica tive [0].
[Y] is a voiceless alveopalatal flat fricative. It has relatively great obstruction in the palatal area, pro duced with the blade of the tongue.
[y] is a voiced alveopalatal flat fricative. It is similar to English 'y*, but has much greater friction, produced by the tongue blade in proximity to the palatal area.
Affricates [dz] is a voiced alveolar grooved affricate.
[tz] is a semi-voiced alveolar grooved affricate. The voicing of this phone begins part way through the fricative release, producing a mostly voiceless, but nevertheless partially voiced phone. 50
[ts^] is a voiceless alveolar grooved aspirated affri cate. The aspiration is relatively heavy.
[ts?] is a voiceless alveolar grooved glottalized affricate. This phone is relatively lenis, making it easy to miss the fact that it is produced with pharyngeal air.
The glottal release follows immediately the fricative re lease.
[dz] is a voiced alveopalatal grooved affricate.
[tz] is a semi-voiced alveopalatal grooved affricate.
This phone like [tz] is only partially voiced, with voicing beginning part of the way through the articulation
of the fricative.
[ts*1] is a voiceless alveopalatal grooved aspirated
affricate. The aspiration is quite heavy.
[ts7] is a voiceless alveopalatal grooved glottalized
affricate. It is relatively lenis, and the glottal re
lease occurs immediately after the fricative release.
[dl] is a voiced alveolar laterally released affricate,
[tl] is a semi-voiced alveolar laterally released
affricate. Although this phone is essentially voiceless,
voicing does begin part way through the lateral. It has
relatively light friction.
[tl] is a voiceless alveolar laterally released as
pirated affricate. I describe this phone as aspirated
because relative to its counterpart, [tl], it has extreme 51 friction and a very fortis production. The heavy escape of air produces what might be termed aspiration. Cer tainly, it is the equivalent of the aspirated series in the stops and grooved affricates.
[tl93 Is a voiceless alveolar laterally released glottalized affricate. The glottal release takes place immediately after the lateral release.
Semivowel
[w] is a voiced bilabial semivowel. It is the non- syllabic version of the high close back rounded vocoid
[u], and is phonetically quite different from the contoid [w].
Nasals
[m] is a voiced bilabial nasal.
[m] is a voiced bilabial syllabic nasal. It is held considerably longer than plain [m], and is the peak of a chest pulse, not the onset like [m],
[n] is a voiced alveolar nasal.
[a] is a voiced alveolar syllabic nasal. It is held
longer than [n], and is the peak of a chest pulse.
[^] is a voiced velar syllabic nasal. It is like [m] and [n] except that it is produced in the velar area. 52
Vocoid Phones
Vocoids
[i] is a high close front unrounded vocoid. It is similar to the vowel in English 'beat* but Apache [i] is not glided.
[lJ is a high open front unrounded vocoid. It is simi lar to the vowel in English 'bit1 but Apache [u] is more close than English.
[e] is a mid close front unrounded vocoid as in
English 'bait' but not glided. It also is not as close as English [ej.
[&] is a mid open front unrounded vocoid like the vowel in 'bet,' however, this vowel is more close than English
[&]. Apache [e] and [&] are very similar and often vary freely.
[a=Q is a low close front unrounded vocoid like English
'mat.' In Apache this vowel occurs only in the environ ment of nasalization. There is some question in my mind
as to whether this vowel is actually different in tongue height from [&]. I think that it is possible that the
extra resonance added by the nasalization may change the
auditory effect to produce the sound [se).
[a] is a low close central unrounded vocoid as in
'cot.' Apache [a] is probably more central than English
[a]. 53 [a] is a low open central unrounded vocoid. There is no sound in English which compares well with this phone.
[%)] is a low close back rounded vocoid as in * bought.1
The rounding which accompanies this vocoid is very slight.
[ol is a mid open central unrounded vocoid similar to
the vowel in 'but,1 however Apache [9] is slightly more
open than English [9].
[u] is a high close back rounded vocoid as in 'boot1
but is not glided.
[vl is a high open back rounded vocoid. An English
approximation is the vowel in 'book,' however Apache [u] is more close than English [vl. [o] is a mid close back rounded vocoid similar to the
vowel in 'boat.' The rounding of this vocoid is not as
great as that of [u] and [u] above, nor is it as great as
English [o].
Prosodic Features and Boundary Phenomena
Nasalization
Any vocoid or [hi with a single [,] underneath it
is produced with nasal as well as oral articulation. Note
that I use this symbol in two other ways as well, with '11
to indicate the apical as opposed to the laminal, and with
the nasals to indicate syllabicity. 54
Boundary Markers
The following symbols indicate pause and other kinds of boundaries: [+] is a hesitation, or slowing of the speech apparatus without pause. It indicates a foot boundary. [/] indicates a short, but observable pause.
[//] indicates a longer pause, the longer the pause, the more slashes. [#1 indicates a non-final intonation con tour. [##] indicates a final intonation contour. [B] indicates breath. Thus, at the end of a contour we might find [##///B] to indicate the phonetic nature of the boundary. Occasionally unreleased stops will have a final release into aspiration. This I indicated by a raised ‘h* as for example [t=hJ. This does not mean that the stop was aspirated, just that the release was held for a noticable period of time, and then released audibly.
Occasionally [?=] will also be released, not just in as piration but in the form of a vocoid articulation (echo vowel). This I symbolize by raising the vowel symbol.
For example, is an unreleased glottal stop with a short high front echo vowel. See Chapter 7 for further details.
Stress
C 1] is primary stress. Its actual phonetic loudness varies with a variety of factors I will discuss in Chapter 6. 55
["] is secondary stress.
[*J is tertiary stress.
Quatinary stress is unmarked, every VJMA syllable having stress, but unmarked ones being the lightest.
Tone
For the time being I will use ["] to indicate high
tone, and leave low tone unmarked. Phonetically tone seems
to vary relative to syllable shape, etc., but to simplify
phonetic transcriptions I will use what essentially are
the phonemic symbols. See the section on tone in Chapter
6 for specific information.
Intonation
I will not indicate intonation contours except
where they are relevant. To symbolize them I will use
lines drawn over the phonetic data, the relative height
of the line indicating the relative height of the tone and
descrescendo.
Length
There are a variety of phonemena which can be
termed length. Vocoids, for example, not only contrast
in terms of length, but also vary relative to syllable
shape. Contoids also are phonetically lengthened after
syllables which end in a vocoid. I will not indicate
these throughout the phonetic transcription, rather I will 56 discuss length as a prosody of the syllable. Generally
I will use the [:] or ['] to indicate length. Chapters
6 and 7 contain more information on these points.
Summary
In this chapter I have presented the symbols used to transcribe V/HA at the phonetic level, and described them in terms of articulatory phonetics and auditory analogies.
This concludes the discussion of VJMA phonetics.
The next chapter, Chapter 4, will be concerned with how these phones are conditioned by phonological, grammatical, and lexical environments. If
CHAPTER 4
PHONOLOGY AND ITS RELATION
TO GRAMMAR AND LEXICON IN WMA
In Chapter 1 I discussed how the three hierarchies of language are interrelated and interlocked. This chap ter is a further discussion of this interrelatedness as it specifically applies to WMA, and is a justification for the use of certain non-phonological environments in the conditioning of allophonic variation in the segmental pho nemes of WMA.
Conditioning Environments
Environments relevant in WMA to the conditioning of allophones and the description of their distribution are as follows:
1. Immediate phonetic environment: adjacent phones.
2. Location in the syllable: initial, nuclear, or
final.
3. Location in the foot and meter: initial, medial,
or final.
4. Occurrence in a phonologically defined syllable
type: open/closed, heterorganic/homorganic, nasal/non nasal, simple/complex.
57 58
5. Occurrence in a grammatically defined syllable type: stem or non-stem.
6. Occurrence in a lexically defined morpheme class:
English loans, Spanish loans or culturally significant terms.
For conditioning of phonemes the higher phonolo gical unit of maximum significance is the foot, a unit defined by stress and speed of utterance. Phonological conditioning of allophones does not take place over foot boundaries. See Chapter 6 for further information on high er phonological units, and Chapter 3 for information on the symbolism used to indicate their presence.
These five environments may be classed into three sources of conditioning: the first three are phonological, the next is grammatical, and the last is lexical. Along with Postal (1968: 135) I will assert that these environ ments can be rated in an order from least arbitrary (pho nological) through most arbitrary (lexical) in terms of how each may be accounted for in a more general theory of language.
Phonologically Conditioned Variation and Distribution
Phonological conditioning of allophones is the dominant process in any language. It is motivated out of a relatively limited set of possibilities which are 59 subject to generalization at some significant level cross- linguistically (Postal 1968). This set of possibilities in its most general form is recognized in the work of structural linguists by the contention that sounds in linear sequence tend to influence each other, and that some sounds, because they are phonetically similar, are more likely candidates for inclusion in one phoneme (Pike
1947a). Viewed in this manner it is a generalization that an alveolar nasal and a glottalized velar stop will never be assigned to the same phoneme because they are not phonetically similar, but an alveolar nasal and a velar nasal are likely to be.
Within the work of transformational linguistics this set of possibilities has been even more explicitly stated in terms of what they refer to as "general lin guistic theory," from which are derived such specific predictions as: if there are voiceless stops in a lan guage these stops will be actualized as voiced stops in intervocalic position if there is otherwise no voiced- voiceless contrast in the stop system (Postal 1968: 184-
185). Further generalizations of this nature are associ ated with the theory of markedness (Postal 1968).
• Whether stated as a universal derived from some general theory, or simply implied as a result of phonetic similarity, phonetic plausability, linear arrangement, 60 and linguistic intuition, phonological conditioning is not arbitrary, but is due to the general nature of sound sys tems or to the biological limitations of the vocal appa ratus rather than being the specific unique property of a given language. It is, of course, the only kind of conditioning admitted into strictly structural accounts of a language (Bloch and Trager 1942).
However, to accept the proposition that phonolo gical environments are the only relevant linguistic en vironments for the conditioning of phonological regulari ties denies both the interdependence and interpenetration of the sub-systems of language. It also implies that the phonological hierarchy is a closed system. A claim which
is not true of any other part of language or culture.
As Longacre (1964: 8) indicated, "To describe a language exhaustively . . . three volumes are needed: a
phonological statement, a grammatical statement, and a highly sophisticated dictionary." No one is complete with
out the other and all are interdependent. Thus, if infor
mation is stated in the dictionary about the lexical and
grammatical category of a morpheme, this information can
be used to predict certain phonological facts. In answer
to the question then "What does one have to know in order
to speak appropriately?" in some and perhaps all languages one has to know certain kinds of grammatical and lexical 61 information in order to generate the proper phonological material. Since this information is stated in the grammar or the dictionary it may be considered a given, and is relevant throughout the description of the entire inter dependent system. Since one has to start somewhere, how
ever, if the dictionary is not available at the time the phonological statement is put together it would seem plaus-
able to include it on an ad hoc basis pending formal state
ment in the later part of the description.
On the basis of this latter argument then I will
use grammatical environments and some lexical ones to
account for facts in the phonological system; even though
it is heresy, in terms of traditional structural linguistic
dogma.
In view of the model of languages outlined in Chap
ter 1, a model which is essentially derived from Pike; the
use of grammatical environments is entirely consistent
with the interpenetration idea of the model as well as
the general assumptions of the model's author (Pike 194?a;
1947b; 1952; 1958) and those of other tagmemic grammarians
(Longacre 1964: 8).
Grammatically Conditioned Variation and Distribution
Unlike phonologically conditioned variation, gram
matically conditioned variation is less predictable in 62 terms of a general theory of language because it is de pendent upon the unique grammatical features of specific languages. In this sense it is more arbitrary than phono- logically conditioned variation (Postal 1968: 135)» but, since grammatical features in languages are themselves subject to cross-linguistic generalization, one should not consider this kind of conditioning wholly random. To ex plain why grammatical information is relevant to the pho nemic description of WMA I must present a brief grammati cal sketch of the language, particularly the morphology.
WMA Morphology
In the literature on the Apachean languages there is some inconsistency concerning the classification of independent grammatical elements or words. Hoijer (1943c:
193) in an early publication indicates there are only three
"form classes:" nouns, verbs, and particles. 'Supposedly only nouns and verbs can be inflected while "particles are never inflected" (Hoijer 1945c: 194). Durbin, writing on San Carlos Apache (1964: 74) follows Hoijer's lead and similarily posits only three classes; again insisting that particles are not inflected. Hoijer (Sapir and Hoijer
1967) in a later publication on Navajo seems to revise his classification for in that study he lists in chapter titles four classes: nouns, verbs, postpositions, and particles. 63 Hoijer has also noted that Apachean morphemes may be grouped into three classes: prefixes, stems, and en clitics (Sapir and Hoijer 196?: 13)• According to Hoijer morpheme types combine to form the word types mentioned above, each word having at least a stem and polymorphemic words including one or more prefixes and/or one or more
enclitics (Sapir and Hoijer 196?: 135 Hoijer 1945c: 193-
194).
Hoijer has also in at least one publication re
ported that there are at least two prefix types; deriva
tional and paradigmatic. Derivational "are loosely incor
porated into the word" while paradigmatic are "those which
stand closer to the stem and are often fused together"
(Hoijer 1938: 75)• Apparently what Hoijer was referring
to in this publication as derivational prefixes were the
postpositions, which he has since reclassed as a separate
form class or word type (Sapir and Hoijer 196?).
In WMA my own analysis indicates there are four
morpheme classes which I call prefixes, stems, suffixes,
and enclitics. All words in the language are composed of
morphemes which can be designated as one of these four.
The class of suffixes is very small including primarily
the relative morphemes and their allomorphs /-n/ and /-ii/
which make verbs into nouns, and nouns into nominal ex
pressions. They seem to have different properties than 6k the enclitics in that they do not have a CV shape, never stand alone, and always seem to be the last of the morpho logical elements which are attached to the nucleus.
In WMA I find that the classification of grammati cal word types into only three classes is inadequate. In addition to nouns and verbs, not only must postpositions be considered a separate class; but also numbers, connec tives , pronouns, adverbals, and enclitic centered words.
These other word types being what have traditionally been called 'particles1 by Apachean analysts. All of these word types are potentially morphemically complex, although nouns, verbs, and postpositions have a much greater poten tial for inflection than do the so called 'particles.' It should be understood, however, that these 'particles' in cluding the enclitic centered words are in fact grammati cally words and can stand alone.
The WMA grammatical word may be defined with the following ideal tagmemic formula: word - £ inflectional onset:prefix(s) t nucleus:stem £ inflectional coda:en- elitic(s) and/or suffix. This is to be read as a word equals an optional inflectional onset filled by one or more prefixes with an obligatory nucleus filled by a stem and an optional inflectional coda filled by one or more enclitics and/or a suffix. From here on I will often refer to the stem syllable. By this I am referring to the morpheme 65 syllable functioning in the nuclear position of one of the grammatical word types. This means that enclitic mor phemes sometimes are considered stems for phonological purposes and at other times are not, relative to their grammatical function.
The most important grammatical word types in the language are nouns, verbs, and postpositions. These gener ally are composed of monosyllabic stems of tne shape
CV(V)(C) with various prefixes and/or suffixes as our ideal formula indicates. I will briefly outline the spe cific nature of these three types and comment on the other so-called ’particles.'
Verbs
Verbs are the most complex having perhaps as many as 13 prefix positions (Edgerton 1963: 107; Hoijer 1945b;
1945c; 1946a; 1946b; 1948a; 1949) including direct object, subject, mode, tense, and number slots. For example:
[*thaadaadL'kLs] /taadaadigis/ 'they begin to bathe' is composed of /taa-/ 'adverbal prefix, meaning unknown,'
/daa-/ 'pi.,' /di-/ 'inceptive,' /0/ '0 class verb,' /gis/
'bathe' a stem.
Verbs, of course, break down into a number of classes. What is generally called the neuter class con tains verbs which act as adjectives in some cases, but as 66 states of being elsewhere, /ndeez/ 'it is tali' can be inflected /nsdeez/ 'I am tali' for example.
Verbs can also be made into nominal expressions.
Thus /ndeezn/ 'the one who is tall.'
Nouns
Nouns are not nearly so complex, being essentially limited to a possessive prefix and a relative suffix. Noun stems break down into a number of sub-classes such as pos- sessable, obligatorily possessed (body parts and kinship terms), and obligatorily unpossessed (wild animals).
These last can be possessed only by use of a relative pro nominal, as in the case where the animal has been killed by a hunter. Some examples: /ll£’/ 'horse,' it ii [*sL'l££?=] 'my horse;' but only [*s t.'t^aah] /sitaa/ 'my father' not */taa/ 'father;' and C u b a ’s] /^ba?/ 'fox,' or ['mba?= 4 's£*yCCh] /^ba? slyee/ 'the fox which is mine,' but not */si^ba?/ 'my fox.' Nouns can also add a relative suffix, thus using some of the above noun stems we find: /sill£9£/ 1 the one which is my horse,1 /sitaa9£/ 111 'the one who is my father,' etc.
Nouns which refer to places are followed by loca tive enclitics. For example: [*sc'khii"yuh] /sikiiy6/ 11 1 1 'my house that place, ' [_ "ts^t^l'gw vzlly [*yaa^go'kai"zi?=] /yaangokaizi? / 'it lies white to 67 there1 (Canyon Day), [*’1'sii'It^z^ytih] /’Isiileezyfi/ 'salt it lies that place1 (a location on Arivaipa Creek), ['t*1Sks9s"gti:l /teksasgee/ 'Texas at.1 Postpositions Postpositions are structured like nouns. The fact that they are not simply verbal prefixes as Hoijer appar ently first believed can be easily demonstrated. In the first place many of them and perhaps all of them can stand alone. Thus we find /saa/ 1 to me,' /sil/ 'with me,' /biL’aah/ 'underneath him,' /biyaa/ 'below it,' /biyi’/ 'in it,' /bika?/ 'on it.' Secondly, like nouns they take pronominal prefixes in this manner: /sifct&h/ 'under me, ' /biL^Sah/ 'under him,' /nit’ddh/ 'under you,' /saa/ 'for me,' /baa/ 'for him,' /naa/ 'for you,' etc. Thirdly, like nominal place names they may be fol lowed by locative enclitics /y6/ 'there, that place,' /gee/ 'at,' / ziV 'to there;' which are grammatically and phonologically subordinate to them. Compare the following with the place names dis cussed under the section on nouns: [*pL'tl?d&h"&l?=] /bil.^Sdhzi'V 'under him to there, ' [*pi'y9"gef*yuh] /biyagelyd/ 'beneath it at that place.' This is not to imply that postpositions are a homogeneous class. In fact they break down into at least '68 two subclasses. Some postpositional stems begin with vowels and thus when inflected with a pronominal prefix form one syllable. The majority are in the other class and have the CV(V)(C) stem shape of most noun and verb stems. Other Word Types • The other word types mentioned above such as en clitic centered words, numbers, connectives, etc., are what Hoijer terms particles. All of these can be inflected in various ways, but their possibilities for such expan sion are definitely limited. The indefinite pronoun /la7/ 'some' for example can be made into a number, /dala’a/ 'one;' a connective, /la7££/ 'and;' or can stand alone. The enclitic, /y6/ 'there, that place,' can take an inter rogative prefix, /haa-/, to form /haiy6/ 'where,' or it can even stand at the beginning of a string of enclitics each of which is apparently a separate grammatical word, but which together form a lexeme which is phonologically one foot such as /ydgeheeyS/ 'further on.' The relative pronoun /7an/ 'third person' forms /haa’dn/ 'who,' and I suspect that even /9§n/ itself is internally complex in that /-n/ generally is a relative suffix used for animate objects. The negative enclitic /dah/ 'no' can stand alone, or is used with a prefix /do-/ to form /dodah/ 'negative.' 69 Even the pronouns combine with a form of the relative suffix, A l l / to form /slyee/ 'mine,1 /nfyee/ 1 yours,1 etc. Note, however, that enclitics when not functioning as the stem of their own grammatical word type usually attach phonologically to the preceding verb. This means they have an alternating status; sometimes being subor dinate to a verb, and other times being independent. Their phonological status reflects this grammatical alter nation and creates an ambiguity in the phonological de scription. That is, enclitics are sometimes considered part of the non-stem syllables which surround a stem, and other times are considered stem syllables. Because of this ambiguous position different ana lysts have treated them sometimes as separate words and sometimes as suffixes. Bandar (19632 11, 18-19) treats them as separate words; but Hoijer, in most of his publi cations, has treated them as suffixes, and thus writes them without any space between them and the preceding verb. This accurately reflects their phonological attachment but it is comparable to writing all English infinitives as one word. Hoijer (Sapir and Hoijer 196?: 112) acknowledges the possibility of considering enclitics as separate words and has this to say about such a solution. Although we have chosen to regard enclitics as bound morphemes, it is equally possible to class them as words. Either decision is arbitrary 70 (and presents its own problems), but it is per haps significant that enclitics, like many of the function words of English, occur only in combination with words and phrases and function for the most part as syntactic links between words or between phrases. I will treat them as grammatically separate func tion words, but recognize their changing phonological status. This changing status is always predictable if enough syntactic information is available, and I will assume that it is. Phonological Variants Conditioned by Grammatical Classes The division of WMA morphemes into prefixes, stems, suffixes, and enclitics which form the word types outlined above is necessary to account for the following kinds of allophonic and phonemic distribution phenomena. 1. Unaspirate stops and affricates become voiced in voiced environments except in stem initial position. Fur thermore, stem initial consonants are always clearly, pho netically lengthened after CV(V) prefix syllables while non-stem consonants tend to vary in terms of length. This latter phenomena I handle as a prosody of the foot level. This observation has been confirmed by a limited sample of WMA utterances analyzed on a sound spectrograph. For example, the utterance [*niibi'p:a9= + *sl't:aah] *1 sit here waiting for him1 consists of two phonological feet, each of which contains a stem. The spectrograph 71 shows a definite voicing of the bilabial stop in the second syllable, but voicelessness and gemination of the bilabial stop in the third syllable, and the alveolar stop in the last syllable. Morphologically this phrase breaks down into /nii-/ 'definite place,' /bi-/ 'third person,' /ba/ 'wait for' (postpositional stem), /si-/ 'perfective mode,' and /taah/ 'sit' (verb stem). Notice that this voicing in intervocalic position is what we might expect in terms of the phonological universal cited above as stated by Paul Postal (1968: 184-85), but that the gramma tical feature of stem functioning as word nucleus affects the articulation of these stops in stem initial position. Not only are the unaspirate stops voiced in non stem initial voiced environments, but velar stops often lack complete closure and are reduced to [g] intervocali- cally or to f~w~] before /o/. Alveolar stops are sometimes flapped ([3]) as well between back vowels; and occasionally labial stops are fricated intervocalically ([b]). What seems to happen is that for stem syllables, because of their semantic and syntactic load in communi cation, speakers of the language slow their articulation so that voiceless stops retain their true character; but in other locations stops are likely to be slurred into their phonological environments. 72 2. This latter analysis also explains another phono logical observation which is predictable in stem sylla bles. Namely, that syllable nuclei are longer (relative to syllable shape constraints) than in other syllables. Often a long vowel in a non-stem syllable will be as long as a short vowel in a stem syllable, while short vowels in non-stem syllables are about half as long as short vowels in stem syllables. 3. Primary stress falls on stem syllables functioning as foot and word nuclei, while other stresses are predict able in terms of a combination of grammatical and phonolo gical information. 4. Allotones are predictable in terms of the phonemic shape of the stem syllable, with the tone of non-stem syllables being intermediate to that of the stem syllable to which they are grammatically and phonologically at tached. 5. Stem syllables seem to condition certain other kinds of consonant variation also. For example [x] occurs only in stem syllables, while [h] occurs elsewhere. 6. Stem syllables have a different or rather more diverse phonotactic structure than do non-stem syllables. Thus, for example, /I»V: occurs only in stem syllable ini tial position, and /d/ and /g/ when they occur in syllable final position do so only in stem syllables. 73 This differential placement and behavior of con sonants in stem syllables has been recognized for some time in diachronic Apachean studies. Hoijer (1938: 75- 76) notes that proto-phonemes have developed differently in stem syllables as compared to affix syllables. This latter fact means that a difference between stem and non-stem environments must be recognized in dia chronic studies in the Apachean languages. Since this is the case it would seem inconsistent to accept a grammati cal environment for conditioning diachronic change in this language, and then reject the same kind of environment for conditioning synchronic variation. This suggests that in generating utterances Apaches use grammatical cues to determine what to stress, what to devoice, etc. Assuming that adequate grammatical infor mation is available, to insist that stress or other pre dictable phonological phenomena be marked would be to make a claim that Apaches had to memorize certain kinds of in formation twice when it is adequate to posit that they memorize it only once. Lexically Conditioned Variation and Distribution I have argued that the phonological system of a language is subject to conditioning from grammatical sources as well as phonological sources. It stands to 74 reason that if the phonological hierarchy is not in fact a closed system that it may he influenced by still another subsystem of language, the lexical hierarchy. Lexical conditioning is the most arbitrary of all conditioning sources for the phonology of a language for it is a product of the unique history of the. language and the culture of which the language is a part.' Lexically conditioned variation consists essentially of a residue of forms which do not fit the dominant patterns of the language. To acknowledge these exceptions as motivated by lexical facts is to accept the phonological system as a part of the culture and to account for its form in a unified theoretical framework applicable to all of the systems and sub-systems of culture. Exceptions to the phonologically and grammatically conditioned regularities in a language usually occur from two sources: (1) loan words which introduce new units or arrangements of units into the system; or (2) lexical categories with special cultural significance related to social differentiation or perhaps ritual or ceremonial contexts. The UMA data seem to indicate an interesting form of the latter kind of lexical conditioning. Exceptions such as the above have been handled in various ways by structural linguists. The most extreme position is to view things like phonemic contrasts intro duced through loan words and present only in loan words 75 as having basically changed the phonemic system. Thus, even though two phones are in complementary distribution in all native words in the language, because the people begin to use one phone in a place where it contrasts with the other, the linguistic assertion "once a phoneme always a phoneme" is invoked by the analyst and he rewrites all the data in terms of the new contrast. A second alternative is to ignore loan words and treat them as the result of bilingualism and interference of some sort. This eliminates them from consideration and maintains a homogenous wholly "native" phonemic system. Still a third way is to assume that native speakers in the process of adding loan words to their language have in some sense become bilingual, and have learned part of a second phonemic system. Thus, when loan words or other irregular forms are uttered, the second or "co-existent11 phonemic system is called into play, and the original sys tem is not considered altered (Fries and Pike 1949). In accepting this solution, the linguist marks in the phonemic orthography when the second system is in operation. This approach to the problem is more desirable except for the fact that it seems to posit a division in the speakers" minds which may not be justified. A similar but more natural solution to the problem is suggested by Postal (1968: 114-139). Phonological 76 irregularities such as occur in loan words are simply to be marked as a phonological property of those few forms in the lexicon to which they are peculiar. In so doing the linguist is saying that the native speaker has to mem orize certain specific facts which are associated with a small number of cases. Within his transformational framework Postal (1968: 123-24) suggests that one way to do this is to mark certain morphemes in the lexicon with the feature speci fication [+ native] or [- native]. Some phonological rules are then blocked because they do not apply in [-native] environments. There seems to be no reason why the same thing cannot be done within an autonomous phonological framework. Particularly if we consider language to be an integral part of culture and part of the description of a total culture. This does not mean that if we use a label like English loan or Spanish loan as an environment that we are actually implying that native speakers have a cognitive awareness of the source of the forms in question. They may simply be aware that the general rules of the language do not apply to some cases and they must be memorized as exceptions. As linguists, however, we may be able to ex plain the source in a contact language, or some other cul tural fact. In building our model we must mark these 77 exceptions, and if the historical reason is known we may explain it in explicit terms. Lexical Environments in ... WMA Phonology We are forced to use lexical environments in dis cussing WKA phonology because of several phonological facts. Loan Words In Spanish loan words stress is predictable in terms other than the morphological stem explanation pre sented above. Consider the following data: [ 1loos*koh] 11 am funny1 [*loon*kohl 'you are funny1 [*loo*koh] 'he is funny' ['luu-'daa-’kohl 1 they are funny' These forms are derived from the Spanish word loco 'crazy.' Note that the syllable [koh] is considered to be the verb stem, as the prefixes for subject and number are inserted before it, and that the first syllable seems to be con sidered some kind of prefix also, perhaps an adverbal pre fix, as is often found in WMA verbs. But, primary stress occurs on the first syllable, not the stem; yet the sup posed stem initial stop [k] is a voiceless phone as we would expect a stem initial stop to be. 78 Another problematic form is ['p3*k9*sii] 'cow' from Spanish bacas. In this case the second syllable is inflected as a stem, a noun stem, with the relative suffix /-££/ being added. Yet the stress remains on the initial syllable, but the consonant beginning the syllable appar ently identified as the grammatical stem is voiceless as a stem initial unaspirate stop should be. I lack specific information on the inflectional possibilities of other Spanish loans, but generally they seem to have their own set of rules for predicting stress, which apparently are not connected with what Apaches iden tify as their grammatical stems. What seems to have happened is that polysyllabic Spanish words as they en tered the language were broken down into syllables, and each syllable was assigned to a morpheme or morpheme class. Once a stem waa decided on, this stem could be prefixed or suffixed as native words can, but primary stress re mained on the syllable in which it was originally placed in Spanish. This means that stress is predictable in these words, but not in the same way that it is predictable in other Apache words. It also means that the grammatical stem serves as a conditioning environment for certain allophones no matter if it has primary stress or not. I include below a list of all Spanish loans in the language along with the English gloss and the Spanish 79 source. We may consider these forms to be a special case, subject to special rules for stress and perhaps other phonological qualities. (Note that many of these are terms for playing cards which were formerly used for Span ish decks, but today are also applied to American decks. Of. Greenfeld 1971 for further details.) English Spanish Apache Gloss Source [*9i'ydG*sn] /9iy66sn/ 'God' Dios [ 1’SS^ts^th^ /?££ce/ 'axe' hache [*so'xWddA] /sah66£/ 'casket' cajon [*tshol’xWuu-nll] /^alh66n£l/ 'that which is a casket' ca.ion [ 'z£5*liih3 /z£51ii/ 'money* real [ 'Ptt'*soh] /•peeso/ 'money,' 'dollar' peso [ 1 ?oo*3oh] /?oodo/ 'gold,' 'hearts' oro [*mGs1k55'"59hj /masgSSda/ 'silk* mascara ['P5n] /b£A/ 'bread' C'tEh] /deh/ 'tea' M [ 1p£5*pas] /b£5bas/ 'potatoe * papas [*PL'l&d#] /bil66n/ 'cantelope' melSn [*m9's££*n9h] /masSSna/ 'apple' mansana [*SG ’g^Gi^yah] /sagSlya/ 'onion' cebolia [^sa1tii^yah] /sad££ya/ 'watermelon' sand£a [ *tshl£*l6h] /c££le/ 'chile' chile [*sa'Idd^dah] /salSSda/ 'soldier' soldado English Spanish Apache Gloss Source [ 'p9*ks*sii] /bSgasii/ 'cow* bacas [ 'tdu*liih] /d661ii/ 'bull' toro ['kud^ts^ii] /g66cii/ 'Pig' cochi [ 'pSSs^’tos] /bSSsdos/ 'clubs' bastos ['kdu^pas] /g66bas/ 'diamonds' copas [*’es'pSS^Sah] /^espSSda/ 'spades,' 'lance * espadas ['p5£^yo] /b5£yo/ 'scarf* bayo or bayeta? [ ,955s] /’55s/ 'ace* as C'ZMh] /Z1I/ 'king' rez ['sd#*toh] / s66dah/ 'jack* sujeto de valet [1loo*koh] /loogo/ 'he's funny' loco [*mGskG'lth] /masgale/ 'Jicarilla Apaches?' Mescalero I further include all English loan words I recorded, as they are also subject at times to special treatment. (The relative scarcity of English words which have been phono- logically integrated into the language and may be classed as loan words, given the length and intensity of contact, seems to support Sapir's hypothesis that the Athapaskan languages are structurally resistant to borrowing [Sapir 1949b: 1969.]) 81 English Probable Apache Gloss Source [ 'phis] /pis/ 'fish' fish ['phl'-silh] /pisll/ 'that which is fish a fish1 [*naal'piil] /naalbiil/ 'car' automobile [*naa1pii"IfIh] /naabiilil/ 'that which is automobile a car' E*p5d'tsniihl /bSdcii/ 'Apache' Apache [ 'pfoh] /bla/ 'beer' beer ['ktdih] /gfdi/ 'cat' kitty? [ ,su"g£h] /soge/ 'sugar' sugar [*tsh2l9'm9"nff] /celamanff/ 'Chinese' Chinaman Phonemes with Restricted Distributions A second problem which demands a lexical solution relates to distribution of some phonemes. The two pho nemes /p/ and /mb/ are one case in point, /p/ only occurs in a loan word from English [ ’p^is] /pis/ 1 fish;1 and /mb/ also has an extremely limited distribution in that it occurs only in two words /^ba?/ 1 fox,1 and /mb6h/ 'owl' and their derivatives. Each of these will be more fully considered when the individual phonemes are discussed in the next chapter, but it seems highly unlikely that we should consider these as possible syllable structure pat terns which can be used in the creation of other words in the language. The fact that they do not have the same 82 status as other phonemes in the system is because of their lexical associations and should be duly noted in the pho nemic statement. A second case centers around the phonemes /%/and /!/. Hoijer (1948b: 343) claims that in Chiricahua the only words which have these two phonemes in initial posi tion are the loan words /l6ogo/ ‘crazy* and /zligo/ ‘rich' from Spanish loco and rico respectively. He feels that with the borrowing of Spanish words a new syllable possi bility entered the language, restricted solely to these loan words. Whether this is true for WMA is debatable. Certainly these two phonemes only occur rarely in word initial position but I have found them there in a few na tive words as well as loan words. Native words which be gin with them are /leni?/ 'customary past time,' and / z i V 'to there.' Both these words are enclitics, which rarely occur alone. Informants will repeat them alone, however, which indicates that there are native structural analogies for the occurrence of /!/ and /z/ in other classes of mor phemes as well; all this indicating that the borrowing of Spanish loans with these two initial phonemes may not have introduced any new patterns to WMA, and perhaps not to Chiricahua either. The one piece of conflicting evidence here is that some speakers of WMA do not use /!/ in initial position on 83 the loan word /loogo/, but replace it with /!/. Only one monolingual woman and her young son did this, but this is suggestive that initial voiceless laterals are an older pattern. I feel, however, that the willingness of most speakers to place /!/ and /z/ in initial position in both loan words, and some native words indicates that these are and were possible syllable patterns in the language, and that they are not limited to the lexical class of loan words. Summary In this chapter I have tried to justify the use of other than phonological environments for the. description of VJMA phonemes. In so doing I have tried to point out phonological facts of the language which seem to be re lated to grammatical and lexical facts, and therefore re quire statement in terms of their grammatical and lexical associations. To this end I have presented a brief mor phological sketch and some information on certain lexical aspects of the language. I now will proceed with the description of the consonant and vowel phonemes mentioning grammatical and lexical environments as are relevant. CHAPTER 5 THE PHONEME LEVEL In terms of the model of language discussed in ■ Chapter 1, this chapter is devoted to the phoneme level of the WHA phonological hierarchy. Here I will attempt to describe the value of each phonemic segment in the lan guage. As stated above, I consider the value of a phoneme is stated by describing its allophones or positional vari ants (manifestation mode); its distinctive features or values along the parameters with which contrasts in the language are defined (feature mode); and its occurrence or possible locations with respect to other phonemes in the syllable, foot, meter, line, and contour (distribution mode [pike 1956: 6?0; 196?: 85-91]). This latter also includes a statement as to its phonemic class membership which is determined by how it functions in syllables and higher phonological units. Since the distribution mode is what relates the phoneme to higher phonological units, this aspect of pho nological value will be stated only in part in this chap ter; more information being presented in Chapter 6 when we consider the structure of higher phonological units. 84 85 The Phonemes of White Mountain Apache This analysis posits 28 consonants, two resonants, and four vowel phonemes for WKA,, making a total of 3^ pho nemes. By using other theoretical frameworks such as Harris' componential features method, a simpler system could conceivably account for the data (Harris 1944); but such an analysis would lack the psychological or phonolo gical naturalness of the approach taken here. The studies of Durbin (1964), bandar (1963), and Saville (1969) have already been mentioned which more or less follow such a componential features approach, and I have specified my criticism of them. Table 2 lists the phonemes and the articulatory phonetic contrastive parameters of the YJKA phonemic system. Note the symmetry in the three way contrast in the stopped segments between unaspirated, aspirated, and glottalized. Further note the contrastive relation between these stopped segments and the true continuant fricatives which occur at the same points of articulation. It should be pointed out that Harris (1945) was the first to chart an Apachean phonemic system in this fashion to emphasize these regularities. The designation and division of these phonemes in to consonants, resonants, and vowels refers to certain phonetic qualities and their function at syllable level. Table 2. The phonemes of the White Mountain dialect of Western Apache Consonants Resonants Stopped Fricatives Nasals Un Glottal- Pre- Voice aspirated Aspirated ized nasal less Voiced Voiced Labial b P mb m Alveolar Plain d t t’ n Grooved 3 s z Lateral L L L? 1 1 Alveopalatal Plain y Grooved 3 c c? s z Velar g k k? h 8 Glottal *? Vowels Front Back High i o Low e a • n; 00 os 8? Consonants may never function as syllable nuclei, reso nants may or may not, depending upon where they occur, and vowels always function as syllable nuclei. In universal distinctive features I would assign consonants the fea tures [+ cons, - voc], resonants the features [+ cons, 4 voc], and vowels the features [- cons, 4 voc]. Since W M has no phonemic glides or semivowels there are no segments which I would designate as [- cons, - voc] (Halle 1964: 32?)• Justification for not treating /y/ as a semi vowel, but as a fricative will become apparent when I discuss its manifestations and distribution with relation to other fricative phonemes. Although this study is not being presented in the process framework of transformational phonology, Table 3 is included as suggestive of a distinctive feature analy sis in terms of universal distinctive features (Chomsky and Halle 1968). I should mention that although this pre sentation generally follows Chomsky and Halle's sugges tions as to the classification of these phonemes there is not total agreement with their description in The Sound Pattern of English. For example /9/ in WMA patterns like a voiceless unaspirated stop. I have therefore classified it as [t cons, - voc] while Chomsky and Halle (1968: 303) classify it as a glide. I have made other adjustments where I felt them necessary in terms of the systematic facts of WMA. T a b l e 3. WMA phonemes with universal distinctive feature specifications mb b p d t t?j s J c c?z s L L # 1 1 g k k?? g h y m n i e a o -e + 4. ♦ + 1 1 1 1 1 1 i 1 1 1 i 1 1 1 1 1 1 1 1 1 1 1 1 1 V o c a l i c 1 1 1 Consonantal C o n t i n u a n t Delayed Release ------+++--+++--+4+ L a x 1 1 1 1 I 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 I I 1 1 1 1 1 1 * 4 4 L a t e r a l «*• 4 4 4 1 1 1 1 1 1 1 1 1 1 1 1 1 I 1 1 1 1 1 1 1 1 1 1 1 1 1 1 t 1 N a s a l 1 4 4 4 4 4 4 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 G l o t t a l 1 A n t e r io r 4-4-4- + f444444 — — — "• — + 4444 — — — — — — — 44 — — — — C o ro n a l H ig h ^ 4, 4^ ^ 44^ . 444^ 4.^.^444444444 — 444— 44ww4 4 4 4 4 4 4 f 1 1 1 1 1 1 1 1 1 1 1 l 1 1 1 1 1 1 1 1 1 1 1 1 1 B ack 1 1 89 Phonemic Contrasts The following are examples of contrasts between those phonemes which are phonetically the most similar to each other. b/p /bees/ 'knife;1 /pis/ 'fish' b/m /bibe9/ 'her breast;' /bime9/ 'her baby' mb/b /mba9/ 'fox;' /baa/ 'to or for third person' mb/m /9S1 mba9/ 'that fox;' /simaa/ 'my mother' d/n /d£l/ 'this;' /n££/ 1 second person' d/t /d££/ 'this;1 /tii/ 'ice' I I d/t9 /d££/ 'this;' /t9iis/ 'cottonwood' t/t9 /tii/ 'ice;' /t9iis/ 'cottonwood' 11 1/1 /lees/ 'dirt;' /leni9/ 'past time' 1/L /dala9S/ 'one;' /ditiis/ 'he passes out' 1/L9 /loog/ 'muscle;' /L9ool/ 'rope' L/L9 /disLiis/ 'I pass out;' /gosL9is/ 'mud' t9/L9 /t9iis/ 'cottonwood;' /L9iis/ 'snake' L/i /9aadiiLees/ 'he polishes;' /diLiis/ 'he passes out' L/L9 /Lo9/ 'bird;' /L9oh/ 'grass' L/l /hiiLa9/ 'he drinks;' /bilaa9/ 'his penis' I 1/L9 /silaah/ 'my sister;' /siL955h/ 'under me' 1/L /doleel/ 'it will happen;' /diLiis/ 'he nesses out' 1/s /9aal/ 'it is finished;' /9aas/ 'friend' j// / jil/ ‘mountain;' //iih/ 'stick' j/z /jil/ 'mountain;1 /zas/ 'snow' j/s /jaa/ 'here;' /sSan/ 'old woman' j/j /jaage/ 'right here;' /jaage/ 'antelope' /bin§jig/ 'I fear him;' /gos^’ig/ 'seven' j/s /jil/ 'mountain;' /sil/ 'with me' j/z /jil/ 'mountain;' /zl£/ 'king' 3/S’ /jaa/ 'here;' /c’aa/ 'hat' 3/ 5 /diji£/ 'today;' /licii/ 'red' i i 3 / 5 ’ /sijaa’/ 'my ear;' /sic’aa/ 'my hat' 3/ 5 /jaage/ 'antelope;' /zSSli/ ‘money' 3/s /di3i£/ 'today;' /’£sii/ 'salt' it it c/5’ /caa/ 'beaver;' /c’aa/ 'hat' 5// /licii/ 'red;' /bi/i?/ 'his daughter' 5//’ /licii/ 'red;' /bi/’in/ 'his bone' c/s /caa/ 'beaver;' /saa/ 'to or for me' c/z /nco9/ 'it is bad;' /nz6/ 'it is good' c/s /caa/ 'beaver;' /sSan/ 'old woman' c/z /licii/ 'red;' /?izis/ 'bag' ///’ /bi/i?/ 'his daughter;' /bi/’in/ 'his bone' //s //al/ 'needle;' /s5an/ 'old woman' //z //al/ 'needle;' /zas/ 'snow' //s //al/ 'needle;' /sas/ 'bear' //z /li/6g/ 'yellow;' /lizin/ 'black' s/z /le’isis/ 'shirt;' /’izis/ 'sack' s/s /sSan/ 'old woman;' /saa/ 'to or for me' 91 s/z /s5an/ 'old vzoman;1 /z55li/ 'money1 s/s /zas/ 'snow;1 /sas/ 'bear' z/z /bize?/ 'his mouth;' /bizi7/ 'his name' z/y /bize’/ 'his mouth;' /biye’/ 'his son' s/z /hos/ 'castus;' /hoz/ 'wrinkled' z/y /lizin/ 'black;' /diyin/ 'medicine man' g/y /biyin/ 'his song;' /bigil/ 'his saddle' g/g /sigSn/ 'my back;' /sigan/ 'my arm' g/h /gaazi?/ 'December;' /hai/ 'winter' iii g/k /bigeez/ 'its eggs;' /bikeh/ 'his feet' g/k? /?ig&n/ 'back;' /’ik’an/ 'flour' g/k /goo/ 'thus;' /ko?/ 'fire' g/k? /gad/ 'cedar;' /k?ad/ 'soon' k/k? /bikii/ 'his house;' /bik?ii/ 'his relative' it ?/d /ko?/ 'fire;' /sigod/ 'my knee' ?/g /silo?/ 'my reins;' /silog/ 'my muscle' ?/k? /’aal/ 'the end;' /k?ad/ 'soon' ?/t? /naa'?e?i?ees/ 'getting her ready;' /nlt?ee/ 'it is good' m/n /hiisne?/ 'I carry several soft items;' /sime?/ 'my baby' ?/h /silaa?/ 'my penis;' /silaah/ 'my sister' i/e /sini?/ 'my land;' /hiisne?/ 'I carry them' e/a /ndeez/ 'it is tall;' /ndaaz/ 'it is heavy' a/o /gad/ 'cedar;' /sigod/ 'my knee' 92 Consonants Although the previous chart of the phonemes is suggestive of the symmetrical aspects of the segment level it is more convenient to break down the system into the categories simple stops, fricatives, and affricates for description. This being the case, I will proceed in the order listed. Simple Stops There are three series of simple stop phonemes in WMA, and one problematic stop which is not part of a series. Bilabial Alveolar Velar Glottal Pre-nasal mb Unaspirate b d g ’ Aspirate P t k Glottalized t? k’ The three series contrast in terms of force of production; the unaspirated being lenis, the glottalized being intermediate, and the aspirated being fortis. Voic ing is not the distinctive feature in this contrast, al though the unaspirate series has voiced allophones. In fact, historically it appears as if the redundant feature of voice was once purely allophonic even in the fricatives. Depending upon which theoretical school to which one sub scribes, the difference may be considered in terms of 93 actual force, or in the time of onset of voicing (Kim 1965; Lisker and Abramson 1964) as there is apparently some disagreement among acoustic phoneticians about the exact phonetic nature of the difference between unaspirate, aspirate, and glottalized stops. The problematic stop mentioned about is /mb/. It contrasts with the other stops in terms of the fact that it is always voiced and in meter initial position is pre nasalized. The problem it presents for the analysis, and its implications for Apachean historical linguistics are discussed thoroughly below. Unaspirate Stops WMA unaspirated stops /b, d, g, ’/ contrast with each other in terms of point of articulation. They seem to form a natural class of phonemes not only on the basis of their phonetic similarity, but because of several par allels in their manifestation and distribution modes. First, they are the only stops which can occur in syllable final position. Second, they all have unreleased voice less allophones. Third, in stem syllable final position, they are often totally or partially assimilated by an alveolar phoneme which follows in the same foot. For ex ample, the following three verbs end in /d g 9/ : /naasgod/ •I am strong,* /naa?iziig/ 'he works, /hiiLaV 'he drinks,' but in the negatively inflected form these final consonants 94 are deleted, the result being /donaasgodah/ ‘I am strong,1 /donaa ^iziidah/ *he does not work,* /dohiiLadah/ 'he does 1 not drink.' Phonetically before a nasal they also often assimilate to phonetic [dl. See below under [d] for ex amples of this. I will say more about alternations such as these in the next chapter. /b/, /d/, and /g/ are voiceless in most positions, however, to the untrained English ear they often give the auditory impression of voicing in all but syllable final position. This is probably because they are phonetically close to so called voiced stops /b/, /d/, and /g/ in English, phonemes which are partially or completely de- voiced in initial and final positions. WMA /b d g/ have voiced allophones in non-stem voiced environments in normal speech, but in lento (slow, precise) speech they are voice less in all positions. Because of the difficulty in identifying their exact phonetic nature one report on another Apachean lan guage has labeled these three members of the series as intermediate in terms of voicing (Bittle 196): 77). As mentioned earlier tests on a sound spectrograph of a severely limited sample of utterances containing these phonemes indicate that in non-stem voiced environments they are voiced, and confirms their voiceless status else where. The use of the symbols b d g to represent these 95 essentially voiceless phonemes is traditional in Apachean linguistics, going back at least to Sapir's 1936 article, Internal Linguistic Evidence Suggestive of the Northern Origin of the Navaho. It has the following to recommend it: (a) its traditional usage by Athapaskanists. They know the phonetic value of these symbols and are unlikely to assign them an improper interpretation, (b) These pho nemes do have voiced allophones, (c) The use of the sum- bols p _t k would require the aspirate series to be repre sented with diagraphs t^ k^. This being the case, I will continue to use b a g for these unaspirate phonemes. /b/ is an unaspirated bilabial stop. It is atypi cal of the unaspirated series in that its distribution is limited to syllable initial position, occurring there in both stem and non-stem syllables before all vowels. It has allophones of [b] [te] and [p]. The allophone [b] occurs in syllable initial posi tion in non-stem syllables in voiced environments. Examples: [*taabL'gah] 'because of them,' [*niibi'pa?=] 'he waits for him.' The allophone [b] occasionally appears in rapid speech in non-stem syllable initial position in vocoid environments as an alternate to [p]. Examples: [*kWus'tdi"b&ds*tos] or [*k^vs't&f + 'paas*tos] 'nine of clubs,' ["thatil«kai,,y£Sfe(.'gy(.l] or ["thMl'kai"ysi + *Pu,gytl3 'donkey's saddle.' 96 The allophone [pi occurs elsewhere. Specifically: 1. Stem syllable initial position. Examples: [*lL*pah] 'it is gray-brown,1 [*tL'pool] 1 it is round,1 [*m'plll 'he,1 [%hii'ptEs] 'he boils it,' [*8L'pLt=] 'my stomach,' C*9v'pt,?=] 'someone's breast. ' 2. In other environments after voicelessness. Ex amples : [*pJmL?=] 'her baby,' [*p&?il'tshd&*dn] 'made yellow people,' [*pL'tl?aah] 'under him,' [*p&9vl't&"9sh] 'hammer.' /d/ is an unaspirated alveolar stop. It occurs in both syllable initial and syllable final positions in stem syllables, but only in syllable initial position in non stem syllables. It occurs before all vowels.~ In stem final position it varies freely with /g/ in many (all?) stems both from speaker to speaker and within the speech of the same speaker. However, any given speaker seems to prefer one over the other in a majority of utterances. The use of /g/ over /d/ as a stem final phoneme has in the past been cited as one of the primary features distinguishing the San Carlos from the White Mountain dialect (Hill 1963: 150). My data indicate that while there may well be a preference in WMA for the use of the stem final /g/, there is still a considerable amount of free variation. 97 The following are all from the same informant on the same day and were elicited successively: /hiisd’id/ 'I am rubbing* /?aadisc?ig/ 'I rub it1 /9aadaadic?ig/ ‘they rub it1 /’aadaanc’id/ ‘we rub it‘ Of the five WMA informants who collaborated with me, /g/■ was used more often than /d/ by. four of them; but none of them seemed to consider the difference important enough to correct me when I used one they did not customarily use. On the basis of this evidence it seems unlikely that this alternation can be considered a significant marker distinguishing WMA from San Carlos Apache. /d/ has allophones of [t=], [d], [d] and [t]. The allophones of /d/ are produced with slight lip rounding before allophones of /o/. This rounding does not give the auditory impression of labialization. The allophone [t=l occurs in stem syllable foot final position. Examples: [ ‘%Lt=] ‘smoke, 1 [*sUlfitit=l ‘my muscle, 1 [*"nlts9iit=3 'wind, ’ [*9aa‘k?C6t=] ‘when everything gets ripe1 (October), [*naascn‘toot=] ‘you are wet,' ["9Iina‘soot=] 'missionary (lit. shirt drags down),* C^naa’Ls'ziit=l 'I work,1 [*naa9L s ‘kh9t=] 'I sew.1 The allophone [d] occurs in the following environ ments : 98 1. Before a nasal. Examples: [*si'?&&*dg] 'the one who is my wife, 1 ["t?lfs'khaa*cLn] 1 cottonwood standing people,1 [*tGs'tshli*dn] 'horizontally red people, ' [11 t^uu^dn'ts^ii9®] 'whiskey' (lit. water which is red), [*p£’i.l'ts^du^diij 'made yellow people.' Note that [d] occurs in this en vironment when the suffix /-n/ is added to a stem which ends in a stop. For example, the uninflected forms for some of the above are: [*si'?&&t=] 'my wife,' [*lv'tshii?=] 'it is red,' and [*lL'tsh&kz] it is yellow.' 2. Non-stem initial position after a voiced conson ant. Examples: [*ton't&&z*dah] 'it's not heavy,' [*ton'tl9CzMah] 'it's not hard.' In this posi tion [d] varies freely with [t] depending upon how quickly the last syllable follows the voiced con sonant, i.e., whether it is one foot or two. 3* Non-stem syllable initial position intervocalically adjacent to a front vowel. Examples: [^’ido'9aal] 'he dances around,' [*thaadL'kis] 'he begins to bathe,' [ 'k^^dchj 'right here,' ['kL"dih] 'cat, ' [ 'k?9"dih] 'very soon,' [*haadLl'z6&h] 'he spits.' The allophone [3] occurs in non-stem syllable initial position between back vowels. Examples: []'9oo*’abh] 'gold,' On'zo*aah] 'it's no good,' [>so'l55*aah] 'soldier,1 [*ton'tshaa*3ah] 'it's not big.' 99 The allophone [t] occurs elsewhere. Specifically: 1. Stem syllable initial position. Examples: [*n'taah] 'you sit down,1 [*nt&&h] 'man,1 [*s Utaa] 'my eye,' [*n'tiih] 'it hurts,' C*si9*-l't<5‘?=] 'me too,' ["sLl'tff*goh] 'he told me,' [*?Cs't2h] 'I wipe.' 2. In other environments after voicelessness. Examples: [*tL'lLk=] 'it burns, ' [^^Ll'k^ah -v *taar^'6wUll 'soap, ' ["p(_ltaa?Ll'zLs] 'they will dance with them,' [*tL'pool] 'it is round,' [*tu'tziih] 'today,' [*tohii'k?aal*tah] 'he is not walking, ' [*to*k?6 ^ ^naai}.'k ^ £ s^tah 1 'it didn't 1 " fall,' ['kot= +- "pLl't^e'^fh] 'cedar berry.' 3. Stem syllable final before a consonant in the next syllable with no intervening foot boundary. (Very few consonants can occur after /d/ phonotactically, however.) Examples: [*n'zaat"Y&h] 'far away,' ['k?9t"Yu*goh] 'now days.' The phoneme /g/ is an unaspirated velar stop. It occurs in both stem syllable initial and final positions. It also occurs in non-stem syllable initial position. Its alternation with /d/ in stem syllable final position has already been mentioned. It occurs before all vowels. /g/ has allophones of [k=], [gw], [g], [g], [gw ], [w], [x], [kw], and [k]. The allophone [k=] occurs in 100 stem syllable foot final position. Examples: [*lL'tshCk=l 'yellow,1 [*y3l't^aks] 'he counts,1 [*hiin 1 ts^t!k=l 'you hear it,1 [*sL'kok=l 'my knee,1 [*sL'zLk=] 'my liver,1 [*9aadcs'ts?Lk=l 'I scratch.' The allophone [g^l occurs in non-stem syllables between a vowel and the [v] allophone of /o/. ([u] occurs in simple nuclei: (a) before any consonant in the same syllable except /’/ and /h/, or /g/ when the first con sonant in the syllable is also /g/; (b). in open syllables before a consonant in the following syllable in the presence of high tone in free variation with [u]; (c) occasionally in open syllables before a consonant in the next syllable varying freely with [o].) Examples of [gw ] include: [*naagwus1 kv,ot=] 'I hoe,' [^naagWyl't^iihl 'rain,' [*naagWvl'zc6h] 'he plays.' [gw3 sometimes alternates with [w], [gw3> or [g]' The allophone [g] occurs in non-stem syllable initial position in other voiced environments. Examples: [*lL'tshuu*gih] 'that which is yellow,' ["yaarjgo'kai] 'Canyon Day,' ["sLl'tfi*goh] 'he tells me,' ['tl?&*gol 'tonight,‘ [*nl't?&E*goh] 'it is good,' ['ld2*goh] 'many,' [*sL'td*gih] ‘that which is hot,' [*taago't?C&h] 'how goes it?' This allophone also alternates with [g] inter- vocalically, or with [w] before /o/. In fact in rapid speech it sometimes drops out entirely producing phonetic vowel clusters. For example: ["sLl'tfol ‘he tells me,‘ [ ‘lao] 'many,1 [*tao‘t?&Ch] 'how goes it?, [tl?^ol 'to night. ' Note that [g], [g], [gw], and [w] vary freely or drop out in non-stem intervocalic environments producing a situation which cannot be described adequately by an absolute statement of occurrence. The allophone [kw ] occurs in stem syllable initial position or after voicelessness in non-stem syllables be fore the [v] allophone of /o/. Examples: [*naas'kwut=3 'I am strong, ' [**p£Enaago'kv,rUt=] 'hoe,' [*kwus'tvk=3 'it's hot, ‘ [*kwi/s'k?az1 'it's cold. ' In lento speech this labialization before [u] is likely to disappear, or before other allophones of /o/ it may make an occasional appearance in rapid speech; but this is unusual. I have recorded things like [*sL'kWok=] 'my knee,' however. The allophone [x] free varies with [k] in rapid speech in non-stem syllable initial position after a voiceless fricative. Examples: [*pL'k?€h*xo] or [*pi_'k?2h*ko] 'according to,' [*ya'y6&h*ko] or [*ya'y&dh*xo] 'seriously,' [^naa't^S51*xo] or [*naa'thaal*ko] 'meaning unknown.' The allophone [kl occurs elsewhere. Specifically: 1. In stem syllable initial position before vowel phones other than [u]. Examples: [*sL'kenl 102 'my arm, ' [*lL'kai] 'white, 1 [*tL.'kLz] 'it's crooked,' [*?Ll'k&t=J 'I stab it,' ['kot=] 1 cedar,' |_*t^Sanaas'kus3 11 wash it, 1 [^naahas'kahj 'I am dry,' [*haago'k&t=] 'he digs,' ['kah] 'rabbit.' 2. In other environments after voicelessness. Ex amples [*?Llk&r)'k&&.s] 1 to cut in two, ' [^kontCz'nonl 'ten,' ["pLlko'^vl] 1it floats.' To be as accurate as possible, it is likely that /g/ also has fronted and backed allophones relative to the following vowel as has been reported for Navajo (Hoijer 1945a: 9), Chiricahua (Hoijer 1946c: 58), and Kiowa-Apache (Bittle 1963: 78). These differences are not audible, however, and I have recorded them nowhere in my data. This is probably because such variation is so much like English that I did not notice it. /?/ is a voiceless glottal stop. It occurs in initial and final positions in both stem and non-stem syllables, but in syllable final position its occurrence seems rare in other than foot final syllables. This is because syllable final /?/ is often either deleted or replaced by /d/ before a consonant or resonant in the next syllable so long as no foot boundary intervenes. /9/ has allophones of [?=] and [?]. [9=] occurs in final position in foot final syllables. Examples: [*siy&?=] 'my son,' [*sL'z&?=] 'my mouth,' [*koos'liI9=l 103 •I live,1 [*n'tsho?=] 'it's bad, ' ['tl6?= + *pL't’« ’ss] 'bird's wing,' [^st'gwoo?=] 'my teeth,' ["16'ni9=] 'cus tomary past time,' [*ts9L'tii*l&n ’=1 1 they said.' The allophone [?] occurs elsewhere. Specifically: 1. In stem syllable initial position. Examples: ['9&ds] 'ace,' [*?i.s'?fl9=] 'I see it,' [*taas'?aah] 'when,' [*si'9aat=] 'my wife,' [*haadi?aah] 'he sings.' 2. In non-stem syllable initial position. Examples: ['?& & ^?ih] 'prairie dog,' [*naa?&9i'?&s] 1 I 'getting her ready,' [*?9't;;"n3h] 'yesterday,' [9L'ts^unl 'bone,' [*9aa'xy&*y&*96h] 'thank you,' ["si9Ll't&?=] 'me too.' 3. Syllable final position before a consonant in the next syllable with no intervening foot boundary. There are relatively few cases like this, but it is a possibility. Examples: [*tEhS?'naat=] 'right (direction),' [*?L'tla9*koh] 'he is drink ing,' [^naa’dts'k^ooj 'I iron,‘ [si'^ii9'ts^oh] 'my jacket,' []%st,,k^1&x.9 'zoos] 'my shoes. ‘ Aspirate Stops There are three aspirated stops, /p t k/. They contrast at the bilabial, alveolar, and velar points of articulation, /p/ is a recently introduced member of the series about which more will be said later. 104 /p/ is an aspirated bilabial stop. It has one allophone [p^]. It occurs only in two English words in the data: [ 'phis] 'fish,* and ['p^L^silh] 'that fish.' /t/ is a voiceless aspirated alveolar stop. It occurs only in syllable initial position before all vowels. Its occurrence in non-stem syllables is infrequent. It has only one allophone [t^], although the force of the aspiration may vary relative to the following vowel and whether in a stem or non-stem syllable. There also may be slight lip rounding before /of, but it is not audi torily noticable, contrary to the case in Navajo. Examples in non-stem syllables include: [*t^Sadt'kls3 'he bathes,' and [*thgts?aa?=] 'basket.' Examples in stem initial position are: ['t^duh] 'water,' [*si_thaah] 'my father,' [*sb't^9=] 'my forehead,' ['t^ih] 'ice,' [ * n ' 1 it is wide, ' ['t^S^zihJ ‘turkey, ' ['th&a*gih] 'three,' [•>•?(. 't^du'?=] 1 soup.' /k/ is a voiceless aspirated velar stop. It occurs only in both stem and non-stem syllable initial position before all vowels. Its occurrence in non-stem syllables is rare, /k/ has allophones of [k%] and [k^wl . The aspiration may be stronger in stem syllables. The allo phone [k^w] occurs in syllable initial position before [vl. It alternates with [k*1] in slower repetitions when [u] alternates with [o] and [u]. Examples: [^k^d'ts^sks] 105 or [*kho'tsh9k=] ‘his (fourth person) rib, ' [ ,k^lwullzi,?=3 or [ ‘kW " z i 9=] ‘right here, ' [ ‘khw#"&&9=] or [ ‘kh&"d29=]| 'right here.' The allophone [k^] occurs elsewhere. Examples in stem initial position include: ['k^iih] 'house,‘ [ ‘k^ooh] 'thus,' [k^duh] 'here,' ['k^o^s] 'fire,' [^^is'k^iin] 'boy, ' [*si.'k£h] 'my foot,' [*naa?Ll'kbo9=] 'he swims.' In non-stem syllables we find: [^k^o'pLk:] 'his stomach,' [*kbo'gWg^] 'his back.' Glottalized Stops There are only two glottalized stops, /t9 k?/. /t9/ is a voiceless alveolar glottalized stop. It seems to occur only in stem syllable initial position before all vowels. Its only allophone is [t9]. Examples: ['t9£isl 1 tree, cottonwood,' [*nl't9Ch] 'good,' [*pu't9ohl 'it's next,' [*n&s't99n] 'food, fruit,' [^nsnt’ln] 'chief,' [*taa't9oh] 'a shade,' [*naa't9oh] 'tobacco,' [*?L't95.t*dn] 'girl.' /k9/ is a voiceless alveolar glottalized stop. It seems to occur only in stem syllable initial position be fore all vowels. Its only allophone is [k9]. Examples: [*pL'k9Ls] 'his sibling-parallel cousin of the same sex,' [*hiis'k98l] 'I am coming,' [ 'k9at=3 1 soon,' [*sL'k9us3 'my neck, ‘ [*yaa'k?oos3 'cloud,' [*-kwos'k99z3 'it's cold.' 106 Pre-Nasal Stop /mb/ is a voiced bilabial pre-nasalized stop. It occurs in stem initial position in a limited number of words dealing with wild animals. It has two allophones, [b], and [Mb], It sometimes seems to vary freely with allophones of the phoneme /b/. The allophone [bl occurs in stem syllable initial position in meter medial environments. Examples: [ 195£ + 'ba9=l ’that fox,1 [*ho’tif + ’ba?=] 'this fox.1 The allophone [mb3 occurs in meter initial stem syllable initial position. Examples: [ 1mbuh3 ’owl,’ ['rnba‘?»3 'fox,' [ ’Mba?to’tl?L"2:lh3 'blue fox.' In the word for owl [mb3 alternates occasionally with [p3. Thus we hear ['mbtih3 or ['pdh3. Cm^3 also in frequently varies with [b3 in meter initial position as in this example: ['""ba9 ' tshoh3 or ["ba?'tshoh3 'wolf.' This concludes the description of the stop pho nemes of UMA. The S t o p s : Comments and Comparisons The stop system of UMA is asymmetrical and prob lematic in a number of ways. I should like to consider these problems here. Neutralization Since the aspirate and glottalized series have no allophones which occur in syllable final positions, and 10? the unaspirate series does; and since [t=] and [k=] are phonetically similar to all the syllable initial alveolar and velar stops respectively, we might say that they could as easily be assigned to the aspirates or glottals as to the unaspirate phonemes. I have assigned them to the unaspirate phonemes on the basis of the following facts. 1. The unreleased final stop phones are most phoneti cally similar to the unaspirate initial stops. 2. /?/ which parallels this unaspirate series in other ways, has the allophone [?=] in syllable final posi tion, and the assignment of this unreleased phone to / V is a relatively safe assumption. 3. The addition of a suffix to syllables ending in an unreleased stop results in their becoming phonetically [d] or [g]; not [th], [kh ], or [t?], [k?]. For example: ["’“'si9 aat=J ’my wife1 becomes [*si9aa*dn] 'the one who is my wife;’ and [*lL'ts^ dk=] 'yellow' becomes [^lu'ts^uu^gih] 'that which is yellow.' Phonemic Overlap The phonemes /b/ and /mb/ both share the allophone [b]. This raises the question if in fact /mb/ is a sepa rate phoneme. I would first point out that although they share an allophone there is no question about when [b] is an 108 allophone of one as opposed to the other. When [b] is an allophone of /b/ it is in non-stem syllable initial posi tion. When [b] is an allophone of /mb/ it is in stem syllable initial position in the center of a phonological meter. This is overlap, not intersection. See Pike (196? 301) for a discussion of the difference between these phonological relationships. Nevertheless, is there any data which clearly point to a contrast between /mb/ and /b/? The following pairs seem to indicate this contrast. In the first [mb] and [p] are in stem syllable initial position: [•mba’s] 'fox' ['paahl 'for him' A sound spectrograph clearly indicates that /mb/ is pho netically [rab3 in this position. It also indicates that /b/ is clearly [p] in this position, even when preceded by a voiced nasal, as in the second pair: ['^ba?=] 'fox' [*m'pai"yen] 'middle aged man1 and finally they contrast in meter medial stem initial environments as the next pair indicate: + 'ba?=] 'that fox' [*nffbi.'pa?=] 'he waits here for him' 109 Voicing and Voicelessness So far as I am aware this is the first description of an Apachean language which has attempted to account precisely for the variation between voicing and voiceless ness in the unaspirate stops, however, one description of a northern Athapaskan language has done so (Howren 1969: 37). In Dogrib, the unaspirate stop, /d/, is generally lenis and voiceless or only weakly voiced (Howren 1969: 16). In the speech of younger people it is phonetically [tl. There is also a phone [r], which only occurs in prefix syllables. Howren assigns this phone to the pho neme /d/, along with [t], or [dj in stem initial position; [r1 contrasts with [dj in prefix syllables, however, and so Howren assigns prefix [dj to the aspirated stop /t/. In other words, /d/ is unaspirated and has allophones of [dj or [t] in stem initial position and [r] in prefix initial position; while /t/ is aspirate and has allophones of [t^j or [t?j in stem initial position, and [dj in pre fix initial position. The assignment of the voiced con tinuant or flap [r] to /d/ parallels closely to what I have described in WMA for all the unaspirate stops and adds support to the analysis in that it demonstrates that there is a tendency for stops in Athapaskan languages to become more lenis in non-stem syllables. As I will point out later on, this lenition of stops in non-stem syllables n o creates a second case of neutralization or better overlap between /g/, /h/, and /g/ in non-stem syllables. Lexically Restricted Distribution There are two cases in the stops of lexically re stricted phonemes: /p/ and /mb/. The phoneme /p/ occurs only in two English derived loans cited above. Other writers on dialects of Western Apache.have also reported /p/ as a phoneme limited to loan words. In Tonto Apache, a dialect spoken near Prescott, Arizona, Mierau (1963’ 10) reports an aspirated 'p' in the word /?apos/ 'apple,1 again a likely loan word from English (perhaps via Yavapai?). (WMA still used /maasSSna/ a loan from Spanish.) Durbin (1964: 2?0 reports several aspirated bilabial stops in SC, in loan words. Specifi cally: ['p^as "ts^ii] ‘Apache,' and ['k^anp^as] 'potatoe‘ from Spanish camotes 'sweet potatoes.' (In WMA the word 1Apache' is used only when speaking English and when speaking Apache they use the word /ndee/. One monolingual old man used (/“'pSS11s^iihl in a text, but with a very defi nitely unaspirated initial stop. The word for potatoe in WMA is [^nS'm;;nsCh] 'that which is round' from /nSmaas/ 'it is round,' or [ 'p55*pas] from Spanish papas, the latter again with definitely unaspirated stops.) It is interesting that WMA use the loan word /pis/ 'fish' for there is another word, /log/ which is used in Ill San Carlos, and WMA speakers knOw of it. Yet people con stantly say, "/log/ is a San Carlos word; the word in our language is /pis/." It is also interesting to note that monolingual speakers replace the English [f] with [p%], thus adding a new phoneme to the system where there is a hole in the WMA phonemic pattern. The interpretation of [f] as [p*1] is in line with Weinreich1s (196?: 220) hypothesis that borrowing languages will often interpret a sound which is not in their lan guage in terms of a hole in their own system which has some of the feature specifications of the borrowed sound. Thus [f] is voiceless, labial, and continuant; while [p^] is voiceless, labial, and aspirated. The continuency of [f] being reinterpreted as aspiration. The borrowing of the word 'fish' with its accom panying reinterpretation of the English phoneme /f/ as Apache /p/ apparently happened some time ago for it ap pears in a word list collected in September of 1873 from a member of the Arivaipa band (Gatschet 1879: 405, 449; 1970: 108). The phoneme /mb/ presents even more of a problem than /p/. There at least are the /k/ and /t/ phonemes of which /p/ completes the series; but /mb/ stands alone. I would like to present a hypothesis along with supporting evidence which I believe accounts for the presence of this unique phoneme in WMA. 112 An Hypothesis Explaining /#%/ It is my hypothesis that /mb/ is an archaic pho neme which is a reflex of proto-Athapaskan */w/. I be lieve it is a residue from an intermediate stage in the development of the Athapaskan languages to modern Apachean. /mb/ probably at one time had a wider if not universal distribution in stems which today begin with /m/ or /b/. It then split into and was slowly replaced by these latter phonemes. /mb/1s existence today in WMA is evidence of the gradual nature of linguistic change, and the relation between message form and message content in the speech event. This is because its present existence in VJMA can be explained in purely cultural terms. Namely, it occurs in a very limited class of words, all of which occur as the names of the principle characters in sacred myths and tales. It is their significance as names for ritual characters which has preserved an older phonemic shape, and detered them from succumbing to the change that other forms in the language have undergone. This hypothesis is based upon the following evidence: Proto-Athapaskan Labials. There is some ambiguity in the literature on proto-Athapaskan as to what the na ture of the phonemic system was with regard to the bi labials. Li (1930: 7) notes that, "Athabaskan is charac terized by the lack of labials except for one which 113 appears in some dialects . . . as m-, in some others . . . as b- and in still other dialects . . . as a bilabial spirant/#.-." He reconstructs the proto-phoneme as *b- (m-), Hoijer (1938) follows his lead in this and refuses to decide upon the exact character of the proto-Athapaskan labial phoneme. Krauss (1964: 122) reconstructs this phoneme as */w/, taking into account, I suppose, that in some of the Athapaskan languages /w/ is a reflex of this proto-Athapaskan labial. Since I have not made a complete study of all the relevant data I am not prepared to com ment upon the nature of the proto-Athapaskan phonemic system, but will yield to Krauss1 interpretation. The Character in WMA Myths. Many WKA myths and tales concern themselves with the doings of 1 owl1 and 'coyote. 1 (I usually translate [mba‘?=3 as 'fox' since all of my informants gloss it as such, but will refer to it here as 'coyote' in line with Goodwin 1939.) In these tales /Mba?^?oos5/ 'coyote,' /mb6coh/ 'big owl,' /Mba?coh/ 'wolf,' /mba7 bi^eegoo’ii/ 'bob cat,' and /^ba? nteeleh/ 'badger' all have roles; with coyote being the most important and owl less so (Goodwin 1939: vi). Goodwin says, Tales may be divided into two major and two minor cycles. "Creation" and "Coyote" are the major ones. The minor cycles are 'Big Owl" and the "ga'n" (Goodwin 1939: vi). 114 He further states. Two animals have their personalities sharply drawn: Coyote and Big Owl (Goodwin 1939’ viii). The Significance of WHA Myths. The tales fall into two classes: those that explain the origin of cere monies and holy powers; and tales that have to do with the creation of specific customs (Goodwin 1939: vii). The ones that deal with holy things are told only to relatives or those who have the power associated with the tales. The rest are told to entertain (and enculturate) children. All of them are supposed to be told only in the winter be cause the 'powers' these stories are about are not around then and will not hear about themselves and punish the narrator and his family (Goodwin 1939• vi). This means that the tales have ritual significance and probably are somewhat immune to radical change, not only in their content, but in their form. This is implied when Goodwin (1939: x) notes that even European contact has had little impact on the tales. The Comparative Evidence. Comparative evidence for possible former existence of an /mb/ phoneme comes from two sources. First, probable cognate forms of /^ba?/ in the Apachean languages. Navajo (Young and Morgan /ma’ii/ 'coyote' 1964: 2 :21) « White Mountain Apache • /^ba?/ 'fox' 115 San Carlos /mb a V 'coyote1 (This form was supplied by Dr. Hoijer from his field notes. ) Chiricahua (Hoijer 1946c: /^bai?/ [mb] 'coyote1 59) Mescalero (Hoijer and Opler 1938: 176) /mai’coi/ [mb] 'wolf' Jicarilla (Goddard 1911: 117) /baitso/ 'wolf' Kiowa-Apache (Bittle 1963: /ba?/ 'fox' 78) This gives a correspondence of: Nav. WMA SC Chir Mesc Jic K-A m mb mb mb mb b b Note that in Chiricahua and Mescalero, /mb/ does not occur in any words other than 'coyote' and its derivatives (Hoijer 1946c: 59; Hoijer and Opler 1938: 3); but in San Carlos, according to Dr. Hoijer, it also occurs in [Mbu?] 'owl' and £rabe'?3 'baby. ' Other comparative evidence which is relevant is the general sound correspondence of bilabial nasals, bi labial stops, and pre-nasalized stops found throughout the Athapaskan languages. Take for example the cognate forms of the word meaning 'metal, knife, arrowhead.' Navajo (Young and Morgan /bees/ 'knife, metal' 1964: 25) White Mountain Apache /bees/ 'knife, metal' San Carlos (Durbin 1964: /bees/ 'knife, metal* 80) 116 Chiricahua (Hoijer /bees/ 'knife, metal 1946c: 58) Kiowa-Apache (Bittle 1963: 86) /bees/ 'metal' Slave (Howard 1963: 44) /nbeh/ [mb] 'metal1 Sarsi (Hoijer and Joel /mas/ 'knife' 1963: 66) Dogrib (Howren 1969: 46) /me/ Cmb] 'knife' Chipewyan (Haas 1968: 173) /bis/ 'knife' Bear River (Goddard 1929: /besci/ 'arrowhead' 296) This gives a sound correspondence of: Nav WMA SC Chir K-A Slave Sarsi Dogrib Chip b b b b b nb m m [mb] b This set of reflexes demonstrates that for the most part, proto-Athapaskan */w/ has become /m/ or /b/ in all the Athapaskan languages, but that in a few cases it is /mb/. The Evidence from Spanish Loans. Loan words from Spanish were borrowed by WMA, San Carlos, and Chiricahua which began with some type of voiced bilabial. WMA and San Carlos borrowed vacas 'cows.' In WMA the Spanish voiced bilabial was assigned to the /b/ phoneme producing /bSgasii/, but in San Carlos (including the dialect spoken in Cibecue) it was assigned to /m/ and became /mSkasii/ (field notes). How general this is in San Carlos is in some doubt since Durbin reports the word as beginning with a /b/ (Durbin 1964: 81). But a word list collected in 1873 confirms that members of the Aravaipa band were using an /m/ in this word at that time (Gatschet 1879: 445), and another one collected in the same time period cites it as sim=bek6ga (Gatschet 1970: 107) hinting that perhaps it was at first interpreted as /mb/. More will be said about these sources below. Chiricahua borrowed the word /madeya/ 'bottle1 from Spanish botella. They assigned the Spanish phone their phoneme /m/ (Hoijer 1939: 110). Early Evidence of the More General Use of /mb/. I have come across two word lists collected in the 1870's which give the impression that words which today are defi nitely pronounced with an initial /b/ were then pro nounced with an /m/ or an /mb/. A word list collected by G. K. Gilbert in 1871 from an Arivaipa band informant lists the following forms written in this manner (Gatschet 1879: 445): Probable Gilbert1s His Modern Rendition Gloss Cognate m-ba-cho 'wolf /mba’coh/ m-ba-rotl-je 'fox' /mta?doL?izi/ m-pi 'deer' /b£i/1 1 This list seems to indicate that not only were the words for 'wolf' and 'fox' produced with a pre-nasalized labial 118 stop, but one other word 'deer' which today unambiguously begin with /b/ also had this phoneme. Still other evidence comes from a second word list, this one collected by a Dr. Johahan B. White at Camp Apache (Port Apache) in the early 1870's (Gatschet 1970). This list includes the following forms which are of in terest: Page in Probable His Gatschet Modern White's List Gloss 1876 Connate bartschu 'wolf' 107 /Mba?coh/ sim=bekoga 'cow' 107 /sibSgasii/ mega 'teeth' 101 /bigoo’/ me-tsartsche be-tsartsche 'child' 99 /bizaazi/ What is relevant here is the use of m in the third person prefix before 'teeth,' the alternation of m and b in the third person prefix before 'child,1 and the m=b in the word for 'cow.' Between the 1870's and the present period we have records of Western Apache from two sources: Pliny Earle Goddard and Grenville Goodwin. Goddard's word suggests that by 1904, /mb/ had probably become /b/ or /m/ in al most all forms. And in those forms where /mb/ was still in use it varied freely with /m/ and/or /b/. For example, he gives the word for cow in San Carlos as ba ga ci (Goddard 1919: 328), the word for deer 119 as bi? (1919*. 269) and the word for coyote as ma or ba (1919: 352) or m ba (1919: 244). In WMA he gives the word for deer as bi (1920: 402), but apparently does not record the word for cow. He gives the word for coyote as ba?. ma tc?os e . and mba? ts?o se (1920: 410). The word for owl he records as bo? or m bo? (1920: 430). Since he does not record the forms for cow or deer beginning with some type of nasal-stop se quence as was the case in the material cited earlier,;from the 1870's, I think one can assume that these forms were no longer used in the early 1900's. Furthermore, since the words for coyote and owl are given with two or three variants I think it is safe to assume that the process:df change from /mb/ to /m/ or /b/ had begun to work on:these forms too. Especially when one considers that Goddard gives the /mb/ forms far fewer times than the /b/ or /m/ forms. These data would seem to imply that words * other than the current ones were pronounced with an initial /mb/, /m/ alternating with /b/, or /b/ alternating with /mb/ during the 1870's. Goodwin (1942: 88) records one other word with an /mb/ that today begins with /b/. He gives the word for Yavapai as mbatc1 * yu*m&*. which he believes to be an English loan. 120 The above seem to support my hypothesis. -The re constructed phoneme */w/ provides a base phoneme which became a pre-nasalized stop. This split into a .stop, /b/., and a nasal, /m/, the change affecting most of-the forms which originally began with */w/. The change was gradual, however, and some forms changed before others. .Particu larly slow to change were the names of individuals -in the myths and tales because of their ritual significance. The modern comparative evidence supports the .idea :that./mb/ is on its way out as a phoneme in the Apachean "languages.. But its presence in some Northern Athapaskan^languages :in cognate words suggests the plausability:of my assumption that in VJNA it once had a wider distribution. The way the various Apachean languages handled the voiced Spanish bi labial phoneme indicates that the change was still going on during Spanish contact. The old word"lists provide evidence that this sound change was still not complete in the 1800*s as some words which today are definitely begun with initial /b/, began with an /mb/ at"thatrtime;"but by Goddard's time the change was essentially complete. Finally the alternation between /b/ and /mb/ today is evi dence that the change will probably be complete-in another generation or so. A Problem with the Hypothesis The major problem with this hypothesis relates to the source of the words in which /mb/:occurs. Comparative 121 evidence reveals that both /mb6h/ and /mbaVmay hot be extremely old Athapaskan words. /mb6h/ does not seem to be found outside of W M and San Carlos Apache. Although /mba'?/ has cognates throughout the Apachean dialects, possible cognates in the California Athapaskan languages are missing. In the Northern Athapaskan languages words are found which appear to be cognates, but the case is not clear cut. For example, in Slave the word for ‘weasel1 is /nSnb5h/ (Howard 19631 44), in Sarsi the word for ‘wolf* is /magazlna9/ (Hoijer and Joel 1963• 74), and in Chipe- wyan the word for ‘white fox' is /ts’iba/ (Haas 1968: 1?2). One cannot be sure that these are cognates, however, for they can be morphemically analyzed whereas /^ba9/ cannot. Thus, Sarsi 'wolf is a compound consisting of /mi-ga-zin- a9/ 'its-hair~black-that-which-is1 or 'the one whose hair is black.1 The Chipewyan word contains the stem for 'gray,' /-ba/. Only the Slave word has no morphemic analysis provided in its source, but one could assume that it consists of a prefix, /na-/, plus a stem /^bah/. The latter again perhabs being the stem 'gray.' If, as this evidence suggests, WHA 'fox' and 'owl' are relatively recent borrowings my hypothesis is only supported if one assumes that they were borrowed by the Apachean languages before the proto-phoneme */w/ diverged into its different reflexes, in which case these words 122 can be used as evidence for the phonology of the language at the time of their being borrowed. Whatever the value of this hypothesis in explain ing the presence of /mb/ in the WMA stop system, the fact that it occurs in only a small class of wild animal words would seem to place it in position of being a lexically special phoneme, whether its source is archaic residue or language contact. Other Fre-nasalized Stops Another related question which I have not even broached here is the occurrence in some of the Apachean languages of a pre-nasalized alveolar stop. WMA does not have this phoneme, but Chiricahua (Hoijer 1946c: 59) does; as does San Carlos in free variation with /d/ (Hill 1963* 149-150)♦ Just what historical process introduced an /*%/ phoneme to these two languages in the first place, its relation to /mb/, and why /^d/ does not occur in WMA are questions which I cannot answer. Fricatives We may characterize WMA fricatives as alveolar, alveopalatal, and velar, with contrasts between plain, lateral, and grooved. There is also a general contrast between pairs of fricatives at the same point of articu lation in terms of voicing. Again, as in the stops, 123 however, it might be well to think of voicing as only a manifestation of a contrast between fortis and lenis seg ments; for occasionally even the voiced fricatives seem to be only partially voiced, particularly in meter final position. The fricatives are as follows: Alveolar Alveopalatal Velar Plain Voiced y g Voiceless h Lateral Voiced 1 Voiceless 1 Grooved Voiced z z Voiceless s s Special Characteristics of the Fricatives I first wish to consider all but /y/ as a separate natural class with special characteristics. In other words, the following: /s s z z g h 1 1/. It may seem somewhat strange at first glance to consider these phonemes as a natural class with other things in common besides their fricative articulation, but they pattern alike in several ways. This similar patterning is exhibited in: 1. Each has a voiced or voiceless counterpart. 2. The voiced series is severely limited in its dis tribution. Members of this series occur only rarely in voiceless environments including meter initial position and seldom or never in syllable final position. In fact, 124 there are not large numbers of examples with voiced frica tives in the data; yet there are enough to give us good minimal or analogous pairs of evidence of contrast. Thus reviewing some examples of these contrasts we find: s/z /sSan/ 1 old woman‘ /zas/ 1 snow1 s/z /9£lsii/ 1 salt1 /9izii/ 1 bark1 i» h/g /hai/ 'winter' /gaaziV 'December' 1/1 /?aagosle9/ 'I'll make it' /hiisliz/ 'I urinate‘ 3.' These same contrastive pairs partake of a morpho- phonemic alternation in stem syllable initial position. For example: s ~ z /naagolsee/ 'I play' /naagolzee/ 'he plays' s ~ z /haadiseeh/ 'I spit' /haadilzeeh/ 'he spits' h ~ g /’ilhoz/ 'It itches' /hiigoz/ 'it is itching' 1 - 1 /HI9/ 'horse' /sil££9/ 'my horse1 M II This alternation occurs in most of- the stems in the lan guage, but is just irregular enough on the surface to pro duce the kind of contrasts listed above. Given these contrasts there can be little doubt about the separate phonemic status of the two series, but I will discuss later two attempts which transformational phenologists have made to eliminate one of them in the phonologies of Dogrib and Navajo, and the possibilities for this in WMA. It might be mentioned by the way that Krauss (1964: 122) does not reconstruct the voiced series for proto- Athapaskan. /y/ is not part of this sub-class, but has 125 other characteristics that place it in the larger frica tive class in WMA. I will discuss /y/ separately below under comments and comparisons. Phonemic Description The phoneme /s/ is a voiceless alveolar grooved fricative. It has only one allophone, [s] which tends to be lengthened at meter boundaries. VJe shall consider such lengthening to be a secondary phenomena and not relevant to the segment level. The other fricatives are also sub ject to this lengthening and the voiced or lenis ones tend to be devoiced during the extra length. [s] occurs in both stem and non-stem syllables in both initial and final positions. Specifically it occurs . in: 1. Stem syllable initial positions. Examples: ['sdan] 1 old woman,' [*?L'saa?=] ‘can,1 [ ‘so"Wff] •wickiup poles, ‘ [*3=&?L‘sLs] 1 shirt. 1 2. Non-stem syllable initial position. Examples: [*SLzS,paalln£h] ‘my lips,' [*nasLn‘taah] 'you are sitting,' [*SL‘k?3z] 'it is cold,' [^t^suu'k^aij 'you (pi.) are going.' 3. Syllable final position. Examples: [ 'zas] 'snow,' [*naahas'kah] 'I am dry,' [*9t-'zt.sJ 'sack,' [ 't^Os] 'bottle,' [*PLnLs'tzck=] 'I fear 126 him, ‘ [*y6&k?oosJ 'cloud, 1 C^'haas'old man, ' [*-kWvs' td£h] 'nine,' [*nEst?Gn] 'fruit.' [s] occasionally alternates with [z] in non-stem syllable final position immediately preceding a stem ini tial stop. Thus I have recorded: [*kwos'k’az] or [*kv;Uz'k’azl 'it is cold (referring to the weather). ' The phoneme /z/ is a voiced alveolar grooved frica tive. It has only one allophone [z]. It occurs in both stem and non-stem syllables in both initial and final positions, but its occurrence in non-stem syllables or syllable final positions is rare. Also rare is its occur rence in meter initial position. In fact, only one word in my data exhibits this phoneme consistently in syllable initial position when spoken in isolation. This is the word ['z9sJ 'snow.' Specifically we. find it in: 1. Stem syllable initial position. Examples: [ 'zasj 'snow,' [*si'zaak=] 'my tongue,' [*s l 'z Uc=3 'my liver, ' [*'s l s L'zt.l] 'my hair, ' [*su'z&?=] 'my mouth,' [*nul'zool] 'it swells.' 2. Non-stem syllable initial position. Examples: [*sL'k99"zfhJ 'it is cold,' [*s&"z9'paa"nth] 'my lips,' [sG't&O'zf] 'mule.' 3* Syllable final position. Examples: [*kooz'lii’-] 'he is born, ' [*!fl,iz' tl^ih] 'dam, ' [*konez'nan] 1 ten, ' ['*naahaz 1 t^aah] 'they are sitting, ' 12? 1 tall,1 [*n'taaz] 'heavy,1 [^naagodts'k^-z] 'he is crazy, 1 ['khii 4- *kooz'tLl] ' tovm. ' The phoneme /s/ is a voiceless alveopalatal grooved fricative. It has only one allophone [s]. It occurs in stem and non-stem syllables in both initial and final positions. Specifically: 1. Stem syllable initial position. Examples: ['sas] 'bear,' ['sil] ‘first person,' ["9Iin9'soot=J 'missionary,' [?ff'sii] 'salt,' [*"-ko'sas h] 'dog.' 2. Non-stem syllable initial position. Examples: [*t&sdu'?aas] 'we (duo.) are going away,' [*^ts^1 a' a"s£h] 'children,' [*pL'kL"seh] 'her cane.' 3. Syllable final position. Examples: [ 'p££s] 'metal, knife,1 [*hiis'p5&s] 'I boil,' ['x^us] 'cactus,' [*si'?aas] 'my friend/ [%?iis'?H?=j 'I see it, ' [^has'k^h] 'he is mean,1 ['ts^oos] 'worm.' The phoneme /z/ is a voiced alveopalatal grooved fricative. It has only one allophone, [z]. It does not seem to occur in prefix syllables, but only in stem and enclitic or suffix syllables. In stem syllables it occurs in both initial and final position, but in enclitic 128 syllables only in initial position. It occurs in very few words in meter initial position, usually being in foot medial environments. Specifically: 1. Stem initial position. Examples: ['zilh] 'king,1 ['zda*li] 'money,1 [*tLl'z5e " 9& h ] 'San Carlos Apache,' [*li'zLn] 'black,' [*n'zoh] 'good,' [*sc'zfh] 'my name,' [*?L'ziih] 'bark,' [*sL'zaa"zih] 'my little one.' 2. Non-stem syllable initial position. Examples: [*sL'tL"z&h] 'my younger parallel cousin,' [*ko'lL"zih] 'skunk,' ['t^9"zih] 1 turkey,1 [*pL't3=9aah"zi?=] 'under him to there,' [*to'ti?L"zih] 'that which is blue.1 3. Syllable final position. Examples: [^naa’i. 1'tl z] 'he irrigates,' [*pL'z@z] 1 its little one,' [ lxVfUz3 'wrinkled, ' [*?L'gy&€z] 'egg,' [*hiis'lLzl 'I urinate,1 ['ts^uz] 'wood,' [*to'tl?Lz] 'blue.‘ The phoneme /!/ is a voiceless alveolar lateral fricative. It occurs in both stem and non-stem syllables in both initial and final positions. It has two allo- phones, [l] and [jl]. [l] occurs intervocalically when the following vowel is a front vowel, or occasionally in free variation with [l] in meter final position. For example: ['k9t-+"pil'ths"l&h] 'red cedar berry,' [ l,p§m"bll2£-'zo"l£hl 'baking powder, ' [*ko'yaa"_lfhl 'Geronimo, 1 [ , 1. Meter initial, syllable initial position where it sometimes varies with [tl]. Examples ['111?=] or ['tlfl?=l 'horse, 1 [*ll'zt-n] or [^tlL'Etn] 'black.' 2. Stem initial position. Examples: [ 'luk=] 'fish,' [*t9'la"?ah] 'one,' [*?9gwu s '!€?=] 'I'll do it,' ['lE£s] 'dirt,' ['ia"ll] 'and.' 3. Non-stem syllable initial position. Examples: [•"•It-'pah] 'gray, ' [*lt-'kai] 'white, ' [*lL'k?aa?=] 'fat,' [*l&?Ll'ts^oo] 'he lies. ' 4. Syllable final position. Examples: 'k?9sn] 'witch,' [^?onl'tok=] 'school' (from the verb to count), [*naagwu l ‘t^iih] 'rain,' [^tLl'xYLi] I I 1 'black,' ['ptl] 'with him,' [*to'lE&l] 'it will be.1 The phoneme /!/ is a voiced alveolar lateral fricative. It has allophones of [l] and [l], /!/ occurs in both stem and non-stem syllables in both syllable ini tial and final positions, [l] occurs syllable finally, and [l] occurs elsewhere. Examples of [l] in syllable final position include: [■*kwy 1 'z£.£] 'at that place,' [*tL'pool] 'round, ' [ ' ts9ul] ‘flower, ' [*st.'tsh ££*?9l] 'pillow,' [*?Ll'zLs] 'he i i dances,' [*naanLl'z6s] 'he hunts,' [*naagwu1 'zee] 'he plays, ' [*nLl'zoo3=] 'it swells.' 130 The allophone [l] occurs specifically as follows: 1. Stem syllable initial position. Examples: [*sL.'lii?=l 'my horse,' [*9L'loh] 1 string,1 [*n'lLk=] 'burn,1 [^su'laah] 'my sibling of the opposite sex,' [*to'l&&l] 'it will happen,' ["l&'ni?=] 'customary past time.' 2. Non-stem syllable initial position. Examples: ['z&&*lih] 'money,' [^sLla'^aas] 'my friend,' • [*sLl8'k9n] 'my fingernail,' ['too"l&h] 'butter fly, ' [*n.8'pii"lih] 'car. ' The phoneme /h/ is a voiceless velar fricative. It has allophones of [hi, [p], [w], [xY], [xvH, [x], and [xl. The allophones [pi, [w], and [h] vary freely after the vowel /o/ when before /!/ or /e/, usually in non-stem syllable initial position. Examples: [ ' so "VJ££l or [ ‘so"h££l 'wickiup poles,' [*no'Wfi], or [*no'pif] or [*no'hIi] 'we.' This latter morpheme also occurs as [*no'Wi*yee] 'ours,' [*no'WCl] 'with us,' [^taanohCgahl 'because of us,' [*nop(.'1119=1 'our horse.' The articu- 11 lation [p] seems to be preferred by younger speakers, while older ones use [w] or [hi. The allophone [Vi] also occurs in non-stem sylla bles in meter initial position before /o/, again varying freely with [hi. For example [*hvl'zEE] or [^Wul'z62] 'it is called,' [*hyl't£i] or ['"'VJO1't£i] 'you two are jealous.' It also has been observed in stem syllable 131 initial position before [uu] as in [*tsh 'VJilunl or [*ts^9ehuun] 'casket,1 ['h##*ht*go] or [ 'Uuu*h6*go] 1hurry up.1 Note that Edgerton (1963: 146) records the latter word as /Weheego/. Her rendition is probably the result of a contraction produced by rapid speech. There seems to be no reason to posit a separate phoneme /Vi/, since all examples in my data are associated with the vowel /o/. The allophone [xwl occurs in stem syllable initial position after /o/ and before another back vowel where it varies with [x]. [xl, and [hi as in [*to'xWaa*dah] or [*to'xaa*dah] or [*to'haa*dah] 'nothing.1 It also occurs in stem initial position before the [u] allophone of the vowel /o/. Examples of this latter environment include: [ 'xwUs3 'cactus, ' [^'^Ul'xwds] 'he sleeps, ' [*9aal'xWuz] 'he tickles.' The allophone [x^] occurs before a front vowel in stem syllable initial position so long as it is not pre ceded by /o/. Again, it often varies with [xl and [hj. Examples include: [*?9'xy&&*y&*?&hl 'thank you,' [*?LdLl'xy%Zh] 'he kills it,' ['xYE-Si] 'pack,' [*hiis'xyttl] 'I pack it,' [*tLl'xyn] 'black.' The allophone [x] is most likely to occur before /a/ in stem syllable initial position so long as it is not preceded by /o/. It, too, varies with [x] and [h]. 132 Examples include: [ *xaij 'winter,‘ [^^Ll'xes] 'he bites,' [*9aal'xaa"ntt*goh] 'near by.' The allophone [h] occurs elsewhere. Specifically: 1. In free variation with any other allophone of the phoneme. Examples are given above. 2. In non-stem syllable initial position before a vowel so long as it is not adjacent to the vowel /o/. Generally, when it occurs in this position there is friction in the pharyngeal area, but the oral cavity is in the shape of the succeeding vowel. Examples: [*hai'yuh] 'where,' [*haa't?Ii] 'what,' [*hiis'lLz] 'I urinate,' [^t&h6?naat- ] 'right (direction),' [*hii'tloh] 'he laughs,' [*todaahLl'ts?3*dah] 'it's not noisy (nothing is heard),' [*sL'sii*y&*hIl] 'that which is mine,' [•“•to'?5a'nii"n5£'"*hihl 'a long time ago.' 3. In non-stem syllable initial position before a nasal. In this case the syllable prosody of nasalization, [i], also characterizes [h]. Examples: [*?Ls'ts^n"hn] 'woman, ' ,[*haas 'thl5>hn] I It II 'old man.' 4. In syllable final position (as in the case of the syllabilic nasals above, when the syllable nucleus is nasalized the final [h] is also produced with nasal release). Examples: [*p£ih] 'deer,' ['tah] 133 •no,1 ['kah] 'rabbit, 1 [*pi1 tl'?aah"Yuh] 'under him that place,' ['t^ih] 'ice.' Often meter final open syllables will have a final [h] which disappears when the same morpheme occurs in meter medial position followed either by another syllable in the same foot, or by a foot boundary with its lack of total pause. Thus, for example, /y6/ 'that place' when it comes before meter boundary (pause) often is phoneti cally [ytih] as in [/*hai'ytih/l 'where.' On the other hand, when not followed by a meter boundary it looses the [h], as in [/•"'hai'yu + ^dcn'yaah/] 'where are you going?' This has been called an "inorganic 'h'" by Hoijer (19^5a-: 27). Syllable final [h] then functions as a boundary marker, but not in a consistent fashion. Some speakers seem to attach it to meter final open syllables more con sistently than others. Thus, Hoijer's statement for Navajo, that "all open syllables in the final position of the word or utterance are followed by an inorganic h," does not apply to WMA (Hoijer 1945a: 27). The phoneme [g] is a voiced velar fricative. It has allophones of [g], [w], [gw ], and [gY]. It occurs only in stem syllable initial position. The allophone [w] generally occurs adjacent to a back vowel and often free varies with [g] or [g#]. Specifically: 134 1. After /o/. Examples: [*ko'wah] ‘wickiup,1 [*lo‘wLl] * saddle,' [^taano'wahj 'because of us.1 2. Intervocalically or meter initial before a simple nucleus filled by the vowel /a/. Examples: [*W88‘tEls] ‘meadowlark,1 [*sl‘gan] or [^sL'wan] 'my back,' [*taabL'gah] or [^taabL'woh] 'because of them,' [*taa'w9h] 'because.' 3. Before a complex nucleus filled by nasalized /a/. Examples: [ 'waa"zi9=l or ['g^aa"zi?=] 'December,' The allophone [gw] occurs before /o/. Examples: [*sL'gWoo?=] 'my teeth,' [*pL'g^Ul] 'its marrow,' ["ts^Ll gwUz] 'V/hiteriver, ' [4-s^'gwu.iny^h3 'my maternal grandmother,' [*haa'gW&] 'dove.' The allophone [g^] occurs before a front vowel so long as /o/ does not precede it. Examples: [*9L'gy&&z] 'egg,' [%hii'gy££l] 'he packs,' [*pL'gyLl] 'its saddle,' [*t&z'gyiih] 'he rides.' The allophone [g] occurs elsewhere. Specifically: 1. Intervocalically before complex nuclei filled by /a/. Examples: [*naa'g)j] 'he roams,' [*sLdaa'g?39=] 'my beard,' [*tshG'g)3"slh] 'chil dren. ' 2. Before nuclei filled by /a/ when preceded by a consonant. Examples: [*taasLlgos] 'we all bite,' ["Pi.l'ga"s£hJ 'turnips.' 135 The phoneme /y/ is a voiced alveopalatal flat fricative. It has two allophones, [y] and [%]. It occurs in both stem and non-stem syllable initial position. The allophone [Y] occurs after voiceless conson ants except [?] in non-stem syllables. There are few ex amples of this in the data. The most clearly attested cases are [*pL'tl?aah"Yuh] 1under him,1 [*n'zaat"Yuh] 1in the distance.1 The allophone [y] occurs elsewhere. Specifically: 1. Stem syllable initial position. Examples: [*si'yoo?] •my bead,1 [^ti'yin] ‘medicine man,' [*9ii‘yS9=] 'he eats,' [*si‘y&9=] 'my son,' [*ti'yaah] 'he is going,' ['yaa] 'louse,' ['yoo?] 'bead.' 2. Non-stem initial position except after voiceless consonants other than [">]. Examples [^si.'g^^ui*y&h] 'my maternal grandmother, ' [ ' s£'“*yClJ 'mine, ' [*yi)s't&?=] 'come here,' [*yaa't^ii] 'language,' [*niyts'tshL*ni"hll] 'woman who has become a mother,' [*ts^Ll'tI*y€h] 'walnut,' [*siy&'taa?] 'my chin.' 136 The Fricatives: Comments and Comparisons Overlap My first comment concerns the problem of overlap between the realizations of the velar fricatives /h/ and /g/ in stem'initial position, and the realizations of the velar stop /g/ in non-stem syllable initial positon. Note that there can be little doubt that [gw], [w], [x], and [g] in non-stem syllables are allophones of /g/. Words in which these phones appear in non-stem syllables when said very slowly replace them with [g] or [kj. Also, there seem to be no cases where allophones of /g/ or the [x] allophone of /h/ occur in non-stem syllables. Since the phonemic occurrence of /g/ does not intersect with /g/ or /h/ this sharing of allophones may be viewed as overlap, not neutralization or intersection. The Lack of /w/ I should like to point out that there is no /w/ phoneme in WMA. All examples of phonetic fw~| or [w] are allophones of the velar phonemes /g/ and /g/ or the vowel /o/. This is quite different from Navajo which has a definite /w/ - /g/ - /g/ contrast. Historically, however, it seems likely that Navajo /w/ probably came from the proto-Athapaskan labial */w/ for we have the WMA-Navajo cognates /’ime9/ 'someone's baby' and /’awee/ 'someone's baby,' respectively. 137 Fricative Morphophonemics I have presented evidence that there is in fact a contrast between voiced and voiceless fricatives in WMA even though the voiced ones are somewhat limited in their distribution. Historically, it may be that at one time there was only one series of fricatives in Athapaskan. At least Krauss (1964: 122) reconstructs proto-Athapaskan in this manner. Two recent studies shed light on this prob lem. In Dogrib, a Northern Athapaskan language, Howren (1968; 1969) has similar phonological facts, except that the case there is much more clear-cut. There are no voiced fricatives (apparently) in stem final position and, but for one exception there are also no voiced fricatives in non-stem syllable initial position either. The one excep tion is that Dogrib has two phones [w] and [W] which con trast in forms like [nawize] ‘the two of us hunt,* and [naWize] ‘the two of us have hunted.' However, Howren is able to show that the phone [w] in prefix syllable is not an allophone of stem initial [w], but is rather an allo- phone of stem initial [m] (note what was stated above about the ra-w correspondence in WMA and Navajo). He then shows how many verb stems have the voiced/voiceless 138 morphophonemic alternation which I have mentioned in WKA, but like V/MA not all verb stems are part of this pattern. There are still forms like [nahze] where a voiced frica tive occurs in the environment of a voiceless fricative. He further used comparative evidence to show that such cases in verb stems are cognates of forms in other Atha- paskan languages where a verb classifier consisting of a voiced lateral, /!/, is the last prefix before the stem. He posits then that this voiced lateral is part of the underlying representation of those verb stems which are not part of the alternation. (The voiced lateral itself is apparently a product of a voiceless lateral plus a fol lowing vowel which collapses.) He is thus able to take a phonemic system which has obvious contrasts between voiced and voiceless fricatives on the surface of the following kind: 1 s s h M 1 z z g w and reduced it to: 1 s s h W by two rules. The first he called "The Spirant-Voicing Rule (SVR).11 Assuming that all stem-initial fricatives are basically voiceless the rule is stated as follows: [-Voi] --- > [+ Yoi] / [+ Voi] stem ( tCns -Voc 4Cnt 139 which is read as any segment which is minus voicing is to be rewritten as plus voiced in the environment after a voiced segment and stem boundary when the segment in ques tion is also consonantal, non-vocalic, and continuent. That is, voiceless fricatives become voiced in stem ini tial position after voiced segments. A second rule then deletes the classifier /!/ from the surface structure of the language. It seems somewhat presumptuous for Howren to posit such an underlying structure in what supposedly is a model of competence when so far as is reported to us there are no examples on the surface of voiced /!/ classifiers in any of the verb stem paradigms for which this segment must be inserted in the underlying structure. It makes one wonder how the children ever get the idea that it is a voiced lateral which is in the underlying structure. It also seems that since this classifier must be inserted in the lexical entries, it functions merely as a marker for exceptions to the SVR. If this is the only reason for introducing the classifier, why disguise it in what is supposedly a phonologically natural way, when in fact there is nothing in the surface structure to support its presence other than the morphophonemic voiced-voiceless alternation. If evidence for its presence must come from other languages which are obviously not known by speakers 140 of Dogrib it seems an arbitrary device which works, but is not in keeping with the naturalness condition. If the lexical entries were simply marked as being exceptions to the SVR it would seem that both morphopho- nemic generalization which applies to most of the language, and a second generalization about competence, that certain forms have to be learned as exceptions, could also be captured. This is not nearly as elegant, but certainly seems to be more 'correct.1 These same kinds of phonological facts occur in Navajo, and Stanley (1969a) has also offered a solution which in the underlying form eliminates an entire series of fricatives, this time the voiceless ones. He reduces a system of: x 1 s s g 1 z E to simply: g 1 z z (Stanley 1969a: 18, 246). The case in Navajo is more like MHA in that voiced fricatives as well as voiceless ones occur in stem final positions, and there are ample examples of contrasts be tween the two series in a variety of positions. Stanley argues that much of the irregularity in Navajo verb stems is in fact regularity. He notes that there are a series of suffixes which follow verb stems 141 and mark the mode of the verb. These suffixes and other underlying parts of verb stems assimilate or undergo cluster simplification and account for different forms of verb stems in different modes (vowel length is also in cluded here as is the contrast in tone and nasalization) and easily give us a case where only voiceless fricatives are in stem final position. On the other hand, Stanley presents evidence that there are only two classifiers of verb stems, /d/ and /!/, rather than the four usually be lieved present by Athapaskanists, /0 1 1 d/ (Hoijer 1946b); and that the /!/ classifier, which was used by Howren as his triggering device for SVR, is a combination of the /d/ plus the /!/ classifier. This ultimately gives only one series of underlying fricatives, which are voiced in stem initial position, and devoiced in stem final position. His evidence is too complex to summarize here, but his suggestions are intriguing. All of his underlying repre sentations seem to be well motivated in terms of surface structure phenomena. At least there seem to be examples of at least a limited sort in the surface structure for most of his postulations. There is little doubt in my mind that many of his suggestions could be applied to WKA. I still have some doubts even about Stanley’s work in terms of its serving as a competence model, but, it must be admitted that he carefully justifies each postulated rule and underlying representation with good phonetic 142 facts which are apparent elsewhere in the language, and he is readily willing to admit that many forms in the lexicon are exceptions to certain rules, and must be marked as such. Stanley's postulation of a basically voiced series of fricatives which become devoiced in certain locations seems to be more natural than Howren1s because he relies extensively upon patterns in the phonetic data of Navajo. Certainly more valid than one based primarily upon compar ative evidence. Deep structure models must always be con strained by the surface structure facts in order to be plausable statements of competence. My own analysis of VJMA fricatives retains the sur face structure contrast between voiced and voiceless fricatives. I believe that it can be argued that this contrast has a certain amount of cognative validity. Evi dence for this comes from the comment of one informant while I was eliciting some forms which alternated between voiced and voiceless fricatives. Referring to the change in the word 'horse1 from /Iff?/ to /stiff?/ in the pos- i i sessed form he said, "It's not the same there, it changes.11 This overt recognition of the change from voiceless to voiced would seem to indicate that Apaches do not consider voiced and voiceless fricatives to have a merely allophonic relationship, but a phonemic contrastive one. The Position of /y/ as a Fricative Since /y/ does not seem to enter into this general fricative morphophonemic alternation nor contrast with its voiceless counterpart, it may seem strange to consider it as a fricative. I feel, however, if we examine it more closely that there are good reasons for classing it with these phonemes. First, note that /y/ clearly contrasts with /g/ in YJMA, unlike Navajo where this contrast is neutralized be fore front vowels (Sapir and Hoijer 196?: 10). Thus, in WMA we have /diyin/ 'medicine man' and /bigil/ 'its saddle as well as /siyoo9/ 'my bead' and /sigooV 'my teeth.' Second, phonetically it is closer to a fricative than to the English semi-vowel /y/. Third, even though it does not partake of the mor phophonemic alternation characteristic of the rest of the fricatives it does partake of an allophonic one ([Y] in certain voiceless environments, [y] elsewhere) as well as a very similar morphophonemic alternation which I would like to describe now. In the environment after the fricatives /s/ and /s/ where other voiced fricatives are usually voiceless, stem-initial /y/ is reduced to 0. For example: /besll/ i 'I push,' /?iis&?/ 'I eat, ' /nSSdisol/ 'he breathes;' /benyll/ 'you push,' /?iiya?/ 'he eats,' /naSdiyol/ 'he i « 144 "breathes;1 /yedaayfl/ 'they push,' /daa?iiya?/ 'they eat,' 1 I /daanSSdiyol/ 'they breathe.1 In each case it is obvious that the base form of the stem should begin with a /y/, but in the first person it changes to /s/ or /s/. What is happening is that when the sequences /s/ + /y/ and /s/ * /y/ occur because of syntactic arrangement (s-~ s- repre sent first person), the /y/ devoices so completely as to be lost in the friction of the preceding grooved fricative. This is certainly a comparable alternation to that noted for the other fricatives. This type of patterning would seem to support the classification in this language of /y/ as a fricative. Affricates WMA has nine affricates. They are divided into two series which correspond to the grooved alveolar and alveopalatal fricatives, and a third which corresponds to the lateral fricatives. WMA affricates also contrast in terms of their manner of articulation, so that considering them from that point of view they correspond to the unas pirate, aspirate, and glottalized stops. They are as follows: Alveolar Alveopalatal Lateral Unaspirate j J L Aspirate c L Glottalized c? L? 145 The affricates are phonetically combinations of the stops /d t t’/ plus the fricatives /z z s s 1 1/. The unaspirate series consists of the lenis stop /d/ plus the lenis fricatives /z z 1/. Phonetically, these lenis frica tives vary from fully voiced to partially voiced, but they are never fully voiceless. It is for this reason that I represent them phonetically in most positions as [zl, [£!, and [l]. In this point I disagree with Bittle (1963* 82) who reports that Kiowa-Apache unaspirate affricates are combinations of /d/ plus /s s 1/. However, the as pirate series may be viewed as being a combination of /t/ plus /s si/; and the glottalized series as being /t7/ plus these latter fricatives. Affricates never appear in syllable final position, and are rare in any but stem syllables. The unaspirate series seems to be characterized by the same stem/non-stem voiced/voiceless variation that characterizes the unas pirate stops, but the occurrence of these phones in non stem syllables is so rare that I cannot make this claim with any degree of certainty. Phonemic Description The phoneme /j/ is a voiceless unaspirate alveolar affricate. It occurs in both stem and non-stem syllable initial position, but its occurrence in non-stem syllables is extremely rare, there being only two examples in my A 146 data. It has two allophones [tz] and [dz]. Occasionally this phoneme seems to alternate with /z/ in a given loca tion particularly in meter initial position. The allo- phone [tz] occurs in stem syllable initial position. Ex amples : [ 1tztl] 'mountain,1 ['t^)?h] 'here,' [*pLnts't_ztk=] 'I fear him, ' [*haadLs'tziis] 'I pull it,' [*tzLs'xy&h] 'I kill it.' [dz] occurs in inter-vocalic position in non-stem syllables. I find only one example: [*taadzit'xy&hl 'they kill it.' The phoneme /i/ is a voiceless aspirated alveolar affricate. It occurs only in stem syllable initial posi tion. In word initial position it is sometimes varied with /s/. It is generally more common than /j/. /$£/ has only one allophone of [ts*1].. Examples: ['ts^cfch] ' stone,' [*sL'tshi?=] 'my daughter,' [*naal'ts^oos] 'paper,1 ['ts^al] 'needle,' [^pE’c1'ts^oos] 'beans,‘ [*si'ts^8k=] 'my rib,' [*sanl'ts^oos] 'you give me a flat object,' I [ 'ts^ih] 'stick,' ['"'IL'ts^vk=] 'yellow, ‘ [*tf s'ts^aa] 'he dies.' The phoneme /V?/ is a voiceless alveolar global ized affricate. It occurs only in stem syllable initial position. It has only one allophone, [ts9]. It is rela tively rare in the data, and can easily be mistaken for / i f . Examples: [limba9 'ts9oos£h] fox,' [*s L'ts9(.n] 'my bone,' [*9al'ts9C*seh] 1 small,' [^th6'ts9aa9=] 'basket,' 14? [naagodul• ts’oo^sn] 'clan name, 1 [*kWi/s'ts9(*gfh] 'seven, ' [*9its9oos] 'feather,' 'ts,?aa9=3 'plate.' The phoneme /j/ is a voiceless unaspirate alveo- palatal affricate. It occurs both in stem and non-stem syllable initial position. It has allophones of [dz] and [tzl. [t_z] occurs in stem initial position. Examples include: [*tL'tzif] 'today,' [*sL'tzaak=] 'my leg,' ['tSaa/'g&h] 'antelope (legged one),' [*su'tzil*l^h] 'my chest,' [*sL'tzaa9=] 'my ear,' [' t_zdus] 'vagina.' [dz] occurs in non-stem syllables. Examples: [*kodzt*taal] 'dance,' ["haadzcl'kus] 'lightning.' The phoneme /c/ is a voiceless aspirated alveo- palatal affricate. It has only one allophone, [ts^] which occurs in both stem and non-stem syllable initial position. Specifically: 1. In stem syllable initial position. Examples: ['ts^aah] 'beaver,' [*n'ts^aa] 'big,' [*11'ts^oh] 'red, ' [*ts^£.z] 'wood, ' ["ts^Ll'pai" y£h] 'oak, ' [*s L'ts^IIh] 'my nose.' II *” 2. In non-stem syllable initial position. Examples: [*tsha'gJJ"sCh] 'children,' [ 'kuu,"bshih] 'pig.' The phoneme /c9/ is a voiceless alveopalatal glot- talized affricate. It occurs in both stem and non-stem syllable initial position. It has only one allophone [ts9]. Specifically: 148 1. In stem syllable initial position. Examples: ['ts^aah] ‘hat,1 [ ‘ts?Ll] 'flower,' [*nts?iit=] 'wind,1 ['ts^oos] 'worm,' [*tL'ts?Ifz] 'it is rough.' 2. In non-stem syllable initial position. Examples: [*ts?L'tiIh] 'they say,' [*ts?i'y3nl 'food.' The phoneme /L/ is a voiceless alveolar lateral affricate. It has allophones of [tl] and [dl]. It occurs only in initial position in stem syllables and enclitic syllables; but never in prefix syllables, [tl] occurs in stem syllable initial position. Examples: ['tlp?=l 'bird,' [*hiis'tla9=] 'I drink,' ['“^ts'tlai] 'five,1 [*hii'tloh] 'he laughs, ' [‘"’’aadii'tlEes] 'he polishes. ' [dl] occurs only once in the data in intervocalic syllable initial position in an enclitic syllable. There it is the product of the phonotactic combination of a stem final /d/ with an enclitic initial /!/. Example: [*to'kh&E*dl2"ni?=] 'she was begging.' This phoneme is sometimes heard as [gl] by English speakers upon first observation. The phoneme /L/ is a voiceless alveolar aspirated lateral affricate. It has only one allophone, [tl]. It occurs only in stem syllable initial position. Words that begin with /!/ occasionally seem to vary freely with /L/, but the preferred form is definitely /!/. Examples: 149 [*naanl1tli 'he irrigates, ' [*tL'tliis] 'he passes out,' [*kodll'tlaah] 1it turns up,1 [^naa’L 1 go’s f %,?Ls,tlo"hfchj 'frog.' The phoneme /IP/ is a voiceless alveolar global ized lateral affricate. It has only one al 1 optione, [tl?]. It occurs only in stem syllable Initial position. Examples: [^to'tl^lz] 'blue,' ['tl?ool] 'rope,' ['tl?oh] 'grass, 1 [*s(.'tl9aah"Ydh] 'under me, 1 ['tl,£‘?=3 'night, ' [^kWjg'tl^Ls] 'mud,‘ [*naa9Ll'tl?oo"lZh] 'spider (the one that makes rope?),' [*sL'tl?aazf?=] 'my pants (my below ones),' ['tl’iisj 'snake.' This concludes the affricates and the section on the consonants. We will next turn to the second major class of phonemes in UMA, the resonants. R e s o n a n t s The resonants are a small class composed of the two nasal phonemes /m/ and /n/. I describe them as a separate class, because unlike the consonants they can be syllabic, can carry tone, can be stressed, and can be the center of their own chest pulse. Unlike the vowels, however, the resonants do not have these characteristics consistently, but relative to their position in the foot. For this reason I have specified that they are [+ cons, + voc]. It should be mentioned that only /n/, however, can occur in a position where it is syllabic. That /m/ 150 cannot, I must assume, is an accident of phonotactics, a not unexpected one given its apparent historical recency and limited occurrence. Phonemic Description The phoneme /m/ is a voiced bilabial nasal. It occurs only in stem syllable initial position in native words, but in loan words it also occurs in non-stem syl lable initial position. Its historical derivation from proto-Apachean */w/ has already been discussed. It has only one allophone, [m]. It is quite rare, occurring only before /a/ and /e/. Examples: [*s'.'mt,?=] 'my baby, ' [*sL'maa] 'my mother,' [^ma'sddneh] 'apple,' [*m9s'kad*o9h] 'silk,1 [*n3'mwseh] 'potatoe (the round one).' The phoneme /n/ is a voiced alveolar nasal. It has allophones of [n], [n], [y], and [m]. It occurs in both stem and non-stem syllable initial position, and forms the peak of its own syllable before consonants and foot boundaries. The allophone [n] occurs in syllable initial posi tion in both stem and non-stem syllables. Specifically: 1. Stem initial position. Examples: [*sL'nii?=] 'my face,' [*t&'nii9=] 'it is sharp,' [*sL'nfi9=] 'my nostril,' [*ni’=] 'ground,' [*hiis'n&?=] »I carry it,' ['nlfh] 'you,' ['nil] 'with you,' 151 [*naahii'naah] ‘it moves,1 ,nii"geh] ‘new. ‘ 2. Non-stem syllable initial position. Examples: [*nL'nii9=] 'your face,' [*naal‘piil] 'car,' [^no't^oh] 'tobacco,' [*n9't§9=] 'corn,' [*nbn't?3n] 'leader,' [^naanul,zzt'"g£h] 'sahuaro cactus,' [*n&.s't9%l 'fruit.' The allophone [nl occurs following a vowel or con sonant before a consonant except /b/ in the same or next syllable, or in foot final position after either a con sonant or a vowel. It never precedes a vowel. Before a velar stop it alternates with [^]. It is often (always?) derived from the syllable [ni]. Thus many words contain ing syllabic 'n' may be pronounced either with [n] or [ni], the vowel sometimes being reduced to syllabicity in the nasal. Before /b/ this reduction is further followed by assimilation to produce [m]. [n] always has stress and tone and is the center of its own chest pulse. Specifi cally it occurs in the following environments. 1. Before a consonant in the same syllable. Examples: [^nl't’ tth] 'it's good,' [%PLgons‘9aahl 'I am learning,' [*9onl'tak=] 'you count,' [*naan!'ts?oos] 'paper,' ['l&n?=] or ["l&'ni?=] 'past time,' [*haasns'kh&h] 'I fight,' ["k^6'>"paanl•k'i’os] ‘November (fire gives him no heat),1 [^ns1t££z] ‘I am tall.' 152 2. Before a consonant in the next syllable. Examples: [^nSfihn1maas] ‘he is crazy,1 I [*ndaasCl‘x9s] or [*ncdaasLl‘x9s] 'we bite,1 I • • [*n‘taah] * sit down,‘ [*n‘ts^aa"hlh] 'big,1 [*n‘taaz] 1 it is heavy,' [*n‘z6h] 1 it is good,1 I II [*ni‘9Ifh + "la’n'nth] 'you are carrying your shirt,' [*n'ts?iit=] 'wind, ' [*9iin'tsh(.k=] 'you hear it,' [*naasin'toot=] 'you are wet,' [*naanf't!9ook=] or [*naan't!9ook=] 'I am getting tired of it.' 3. Alternating with [r;] before a velar stop. Examples: [*t^aadLn'kis] or [^t^aadi^'kis] 'you bathe, ‘ [*t(.n'kis] or [*tL^'kis] 'you are lazy, ' [*yaango'kai] or [*yaa^go'kai] 'Canyon Day,' [*haan'k^GhJ or [^haarj'k^th] ‘you fight.' 4. After a consonant, not succeeded by a vowel. Examples: [*?i'yuu*sn] 'God,' [*sc'k9C*sn] 'my sibling of the same sex,' [*n‘t^t^zn] 'the one who is tall, ' ' ts*1Sn"hn] 1 the one who is a woman,' [*si?Sa*dn] 'the one who is my wife,1 [*tLl'xy&*ln] 'the one who is black,1 I ~ [^haasns'k^Zh] 'I fight,' [*naahn'maas] 'he is crazy.' 5. After a vowel when not succeeded by a vowel. Examples ["l&'ni9=] or ['l&n?=] 'customary past 153 time,1 [*9Ls'tshgn] or [^’l s 1ts^0*ni] ‘woman,1 [ ‘s&an] or [ ‘saa*ni] ‘old woman,' 'xwuun3 or [*tshal'xW#u*nI] 'casket,1 [^nant^sn] 'leader,' ['pan] 'bread,' [ '^an] 'they,' [*lL'zin] 'black,' [*ts?l'ydn] 1 food,' [^pL^ts^fts’tn] or [*pL"tshs.'ts9<."nlh] 'his head.' The allophone [m] occurs before /b/. It is some- I times derived from the syllable [nij like [n] or other times is derived directly from [n]. Examples: [*m'pil] 'third person' from [*nL'pil], [4:*m'pai"y5n] 'middle aged man, ' from [*ni'pai"ydn], [p5m,lb^-l6 L 1 zoo1112.] 'baking pow der' from [ 'p5n] 'bread' and [ ’pci] 'with it,' [*p£ 'pumpkin' and [*pc'tddh] 'its seed.' The Resonants: Comments and Comparisons There is no contrast between the syllabic and non- syllabic varieties of /n/. When this phoneme occurs before a vowel it is non-syllabic; when it occurs else where it is syllabic. A similar case pertains in Navajo, Hoijer (1943a: 11) noted in his first publication on Navajo that final nasals had a syllabic quality, yet at that time he set up a separate syllabic nasal phoneme. He again noted this in his more recent publication (Sapir and Hoijer 196?: 8), but there he did what I have done and included the syllabic variety as an allophone of /n/. 154 O v e r l a p Note that again we have overlap in that /n/ has a syllabic b i l a b i a l [ml as an all option e. Sin c e [m] o nly occurs in syllable initial position, and [ml occurs only before /b/ the occurrence of either is completely pre dictable, and we have preserved biuniqueness. The logic for assigning the syllabic [m] to /n/ rather than /m/ is I further justified when it is remembered that every example of syllabic [m] found in the data is derived from what in other phonotactic situations is [ni] or [n]. V o w e l s In the following analysis I have tried to present as complete and detailed a description of an Apachean vowel system as any in existence, and in so doing have to some extent sacrificed the elegance which often seems to accompany phonemic statements. I have done this because I am convinced that many of the dramatically simple and aesthetically delightful descriptions which appear in the professional literature have been arrived at by slighting the data to one extent or another. Vowels are extremely variable in all languages and to treat them otherwise may be elegant, but it is also inaccurate. Vowel Phonemes White Mountain Apache has four vowel phonemes which occur in either complex syllable nuclei containing 155 clusters of two vowels, or alone in simple nuclei. Each vowel carries one of the two phonemic tones. All of the four phonemic vowels may also occur with a prosody of nasalization which seems best treated as an environment which in some cases affects the phonetic quality of the vowel. These vowel phonemes may be represented in terms of their contrastive features by the following chart: Front Back High i 0 Low e a Articulatory and Auditory Description Initial auditory impressions of the language pre sent the linguist with a large number of phonetic vowel qualities. Qualities which for the most part have been ignored as "free variation" in the description of other Apachean languages (Bittle 1963: 84; Hoijer 1946c: 63; Sapir and Hoijer 196?: 11-12), or else have been treated in what appears to be overly complex manner (Lander, Erwin, and Horowitz I960: 371:n.8). Durbin's (1964: 36-39) description of the vowels in San Carlos Apache is perhaps the most accurate, but even he seems to ignore the presence of some allophones such as [vl and [a]. Furthermore, his description of vowel 156 phonetics is somewhat unclear and he introduces as vowel allophones phenomena which I feel are not relevant to the segment level, but instead are better treated as boundary . phenomena related to the foot. Using Pike's (1947a: 5) system of vowel descrip- tion, Western Apache vocoids may be charted in this manner: Front Central Back Unrounded Unrounded Bounded Close A High i Open ly Close /c\ Mid / Open Close rV 33 , Low Open The circles indicate what phonetic ranges are in cluded within the four vowel phonemes. These variants are both free and conditioned. Nasalization All Apache vowel phonemes can occur both orally and nasally. Nasalization does not seem to carry a heavy functional load and when it does occur it is relatively easy to overlook, for to the native speaker of English many Apache utterances seem to be produced with a slight nasal quality. But its actual presence or absence is readily detectable by native speakers of Apache as the following instance shows. 157 I was speaking about the wife of my main informant and calling her /simaa/ ‘my mother.1 (We were on a kinterm basis.) The people listening thought me very funny and were laughing at me because I was accidentally allowing the stem initial nasal to affect the following vowels pro ducing /simaa/ instead of /simaa/. The reason they were i i laughing is that /maa/ is 'moo,‘ the noise a cow makes. II Some examples of the contrast between nasal and non-nasal vowels in minimal or analogous environments are as follows: E'sn] '1' VS, [s££] 'summer' [*si'?fl] 'my shirt' vs. [*hii?I19=] 'he sees ['Pii] 'he' vs. ['p£lh] 'deer' 1 1 1 [ * s L ‘tzil] 'my h e a r t ' vs. [*s t.' tshflh] 'my n o s e One reason for the general nasal quality of many Apache words mentioned above is that the nasalization of a final syllable or syllabic nasal often affects the vowel . quality of the entire word or at least of the preceding syllable. I have not noted this as a consistent pattern, however. This phenomenon seems to be more in terms of free variation rather than a hard and fast rule. Durbin (1964: 32) disputes this, however, and claims that in San Carlos at least, all vowels before a nasal are nasalized. Having established the phonemic importance of nasalization on the basis of the phonetic data cited I 158 will proceed to treat it as a prosody of the syllable, symbolized by a [ f] under the co-occurring vowel or vocoid. Allophonic Distribution The phoneme /!/ is a high front vowel. It has allophones of [ ] and [i]. The allophone [t] occurs in the following environments. 1. Simple syllable nuclei before any consonant pho neme in the same syllable except /?/ and /h/. Examples: ['lCk=] 1 smoke,1 [*hiis'ts?Lt=] *1 rub,* ['ts?Ll] 'flower,' ['pill 'with,' ['*tLk'kLs] 'I am lazy,' [*pL'kls] 'her cane,' ['ts^Lz] 'wood,' [*hiis'lLz] 'I urinate.' 2. Simple nuclei before a nasal. Examples: ["tL^'kLs] 'you are lazy,' ["?L'ts?Ln] 'bone,' ["?Ln'ta?=] 'Whiteman,' ["?Ln't&&] 'Indian,' ["^Cn't^Ln] 'road.' 3. Simple nuclei before any consonant or nasal in the next syllable except /?/, /h/, and /y/. (After Z9/ in this environment [i] often alternates with [(_].) Examples: [^IL'pah] 'brown,' [*sL't^aa] 'my father,' [*ha9 t 't^ool] 'it spills out,' [•"'sL'maal 'my mother, ' [*?L:gWoo9] 'someone's teeth,' [*sL'taa] 'he sits,' [*sL'zaak=] 'my tongue, ' [*sL'zLk=] 'my liver, ' [*9 L.'sii] 'salt, ' [•"Pl'nii9=] 'his face.' 159 4. Same as environments 1 through 3 in the presence of nasalization. Examples: [?aas'ts?L*s&h] 'small, 1 [ 11 tobCgwon1 ts^t^dahj 11 don't know. ' it. i The allophone /i/ occurs elsewhere. Specifically, in the following environments. 1. All complex nuclei. Examples: ['t?iis] 'cotton wood, ' [tl^iis] 'snake,' [sL'nii9=] 'my face,' [naal'piil] 'car,' ['^ii] 'shirt,' ['sill 'me,' [ts’ii'yan] 'food,' ['xai] 'winter,' [sL'zail 'sand,' [naas'k^ui] 'I vomit,' [n'tui] 'mountain lion,' ['pi a] 'beer,' [iL'kai] 'white,' ['zll] 'king (a card).1 2. Simple nuclei before /?/ and /h/ in the same syl lable; / V , /h/, and /y/ in the next syllable; or foot boundaries. Examples: ['ni?=] 'earth,' ['nil] 'you,' [*naadi?'sook=] 'he drags it,' ['si + 'z<-k=] 'my liver' (said slowly), [*si'y6?=] 'my son,' [^ti'yunl 'medicine man,' ['tihl 'this,' [*9L'zih] 'cedar bark,' [%1 l 'ts^uu*gihl 1it is yellow,' ['si t 'nii?=] 'my face,' (said slowly), ["si'^fi] 'my shirt,' [*pi'?d&t=] 'his wife,' [*haadi'?aah] 'he sings.' 3. Same as environments 1 and 2 in the presence of nasalization. Examples: ['pfih] 'deer,' ['s££] III II 'summer, ' [ ' k^ih] 'house, 1 [^‘hiis'?££'?=3 1 ^ see it ii 160 it,1 [*nS5s(s1tiit=l 'I am full,1 [*haan'zffsl 'you pull,' ['pf9=l 'bear grass,' [^sa^n't^ih] 'you hand me a long thin object,' ['t^ih] 'ice,' ['ts^ih] 1 stick.' ii The phoneme /e/ is a low front vowel. It has allophones of [e], [se ] and [£]. It is not as commonly found in the data as other vowel phonemes, and in fact .Stanley (1969a: 157) suggests for Navajo that it is derived from another vowel at least in the deep structure and is not pontrastive. He thus proposes a three vowel system. I will comment further on this possibility below. The allophone [e] alternates with [&] especially in complex nuclei not followed by any consonant in the same syllable when high tone is present, [e] is also ob served occasionally in other environments in which [&] occurs. Since [e] has no consistent appearance I will only list a few examples where it has been observed. Keep in mind that [s] has also been noted in these same forms. Examples: [*n'lee 4 *?Ll'thii] 'grave,‘ [*nl't?ee] 'good,1 [*leego'tzaa] 'year,' [*s57n'leel 'you hand me two ii i i objects,1 ["yuW&'hee*ydh] 'over there,' ['ts^ee] 'stone.' The allophone [se] occurs in nuclei not followed by any consonant in the same syllable other than /h/, in the presence of nasalization and high tone. Examples: [*96 & *9ih] 'prairie dog, ' [*til'z&*9&h] 'San Carlos • i iii 161 Apache, ’ £*9aas1 ts7 t*scfe hi 1 small,1 [*'?aagWon'ts^66*s& h] iii i ii i i 'narrow,' [*7ii'y66*h&h] 'few.' ii ii i The allophone [g] occurs elsewhere. Specifically in the following environments. *1. All nuclei followed by any consonant in the same syllable. Examples : [ 'p e e s 3 ' k n i f e , ' [l££sl 'dirt, ' [-"-ko'th £t3=l 'wide, ' [*n't&tzl 'long, ' [*pel'g ^"slhl 'pliers,' [*76s'tla"7ihl 'five,' [^pcda'ya"gDil 'his clothes,' ['tl7&7=] 'night,' [^pu'ts^€.s^=] 'his tail, ' [*si'y&?=] 'my son,1 [*naa7i'76s] 'getting her ready,1 [*n&s't7Snl 'fruit, ' [*7Ll'k&t=] 'I stab,' [^sc'^3'p^^"n^hl i i 'my lips.' 2. Simple nuclei followed by any consonant in the next syllable. Examples: [*s L"s&'ts7Ln] 'my head,' [*t&'nli7=] 'sharp,' [*y£daa'ycll 'they push, ' [*tEht?'n&&t=] 'right hand, 1 [*l&7L].'ts^oo] 'he lies,' [^pAn'sCl] 'I push,' [^pEnaago'kwut=] 'hoe.' 3. Complex nuclei not followed by any consonant in the same syllable in the presence of low or rising tone. Examples: ["n't&&] 'man,' [*pL'p6&"zthl 'his brother.' 4. In the presence of nasalization except for high tone nuclei not followed by any consonant in the 162 same syllable. Examples: [^nSal'z it] 'I hunt,1 [^’ u'gyfctz] 'egg.' The phoneme /a/-is a low back vowel. It has allo- phones of [?], [o], and [a], and [a]. The allophone [?] occurs in the following environments. 1. Complex homorganic nuclei after /g/. Examples: [naa'g?)] 'he roams around,' [*sLdaa'g3J9=] 'my beard,' [^ts^'g;)"s^e] 'children. 2. Complex homorganic nuclei not followed by any con sonant in the same syllable except /?/ or /h/ in the presence of nasalization. In this environment it sometimes alternates with [a]. Examples: 'leaf,' ['t);] 'spring,' [w)2"zi9=] 'December, ' [*9L].'khl>)] 'sugar, ' [^yuut^a'xSjl I I * I ! _ 'Navajo, ' [^s (-"za 'p^j "n£h] 'my lips.1 3. In a few words with complex homorganic nuclei after / V in free variation with [al. Examples: [*si'?oos] or [*si'?aas] 'my friend, cross-cousin,1 [*si'9oot=] or [*si'9&&t=] 'my wife.' The allophone [a] occurs in the following environ ments. 1. Simple syllable nuclei before any consonant pho neme in the same syllable except /?/ and /h/. (Note, however, that it also occasionally free varies even before /?/ and /h/.) Examples: ['zos] 163 1 snow,1 [ 1sas] 'bear,' [*9L'k?9z] 'sugar cane,' [*pL'zsz] 'its offspring,' ['ts^al] 'needle,' ['k?at=] 'soon,' ["’iin'ts^§k=l 'you hear.' 2. Simple nuclei before any nasal not followed by a vowel. Examples: ['pan] 'bread,1 [*nan't^^nl 'leader,' ['’an] 'that person,' [*s L'kan] 'my arm, ' [*""sL'gan] 'my back, ' [*ts’ii'y3n] 'food.1 3. Simple syllable nuclei before any consonant in the next syllable except /’/, so long as no foot boundary intervenes. Here too it occasionally varies with [a], especially when preceded by [’]. Examples: [*’9'%&’=] 'but,' ['k’ a"dfh] 'now,' [*::‘S0'xVIdun] 'casket, ' [^ydut^a'xjj] 'Navajo. ' 1 • I I 4. Simple nuclei before any consonant in the same syllable or any consonant in the next syllable, so long as there is no intervening foot boundary, in the presence of nasalization. Examples: ['sa’=] 'out in the sun,' [*naa'ta’=] 'corn,' [*pc'gah] 'because of him,' [•::"si.'ts^1ak=] 'my rib,' [*99lfdd*hZh] 'thin,' [*naahas'kah] 'I am dry,1 [*SL'kha’=] 'my husband.' 5. In complex heterorganic nuclei after another vowel. Example: [ 'pi 3 ] 'beer.' This occurs in only one word, a loan from English. The allophone [a] occurs elsewhere. Specifically, in the following environments. 164 1. Complex homorganic nuclei except after /g/. Examples: [*sL'maal 'my mother,' [*s't^aa] 'my father,' [*sC'taa] 'my eye,' [%t lope, ' ['sdan] 'old lady,' ['?&&s] 'ace (card term),*[*sL'tzaak=] 'my leg,' [^t^a'ts'?aa •basket.1 2. Initially in complex heterorganic nuclei. Examples: ['?&i] 'that,' ["tshiinaa'gai] 'wagon,' ["tl’ohna'gall 'wheat,' ['?ao] 'yes.' 3. Simple nuclei before /?/ and /h/ in the same or next syllable so long as there is no intervening foot boundary. Examples: [*:-t5la"‘?5h3 'one, ' ['mba9*=] 'fox, ' ['sd*?ah] 'old, ' ['la"?ii] 'and,' [•"-993. he hi 'thin, ' [^sL't^d^s] 'forehead, ' ['tah] 'no,' ['kah] rabbit,' [ 'p5 + 'ka*sii] 'cow' (slow repetition). 4. Complex nuclei followed by any consonant phoneme in the same syllable in the presence of nasaliza tion so long as it is not preceded by /g/. Examples: ['^aall 'that is all.' Pew examples i i occur in the data. The phoneme /o/ is a high back rounded vowel. It has allophones of [v], [u], [o], and [w]. The allophone [u] occurs in simple nuclei before any consonant in the same syllable except /?/ and /h/; or /g/ when preceded by 165 /g/. Examples: [*"'kwvs 1 tdk=] ‘hot,1 £*kwVs'k^az] ‘cold,* [*xw<5s3 ‘cactus,1 [*?Ll'xWus] 'he sleeps,1 [*yaa‘k?us] 'cloud,' C"tpL'sw v3=3 'it's marrow,' t^naas'kw Ut=3 'I am fast,' ['lUk=3 'fish,' [*?Ll'xWuz3 'it tickles,' [*hvl'z&&| 1 it is called,' 'tak=3 'he counts,' ['p&&s 4 *9Cs't?iyk=3 'arrow head, ' Also [*st'kwut=3 'my knee' but note that when this word ends with a /g/ it is phonetically [•"•sL'kok-3. This allophone also occasionally varies freely with [u3 or [o3 in simple nuclei before a consonant or nasal in the next syllable. For example: ['kd"d&9=3 or £ 'kw6 ,,d£9=3 'right here, ' £limbu'ts^oh3 or []l,mbT&'ts^oh3 'Big Owl, ' C^khwo • tshak=3 or ['"k^o'ts^1ak=3 'his (fourth person) rib,' [*9vnl't^ak=3 or [^9onl'tak^j 'school.' I I : The allophone [u3 occurs in the following environ ments. 1. Initially in complex heterorganic nuclei alter nating with [w3. Examplesr [*haas't^d£3 or [*haas't^wi3 'old man,' [*n'td£3 or [*n'tw£3 'mountain,' [*taa?9'khdl3 or [*taa99'khwi3 'how much, ' [[^naasii'k^u£3 or [[^naasii'k^£3 ‘I am vomiting.' Note that the alternation of [w3 and [u3 in the environment k*1 i . creates the phonetic sequence [k^w3 which is very phonetically similar to the labalized 166 allophone of /k/, [k^w]. For example compare the words [*khw O’ts^sk*^ ‘his (fourth person) rib1 and [*taa?9'khwl] 'how much,' where the single phone and the sequence are almost identical. Slow repetition of each reveals, how ever, that the former is actually a labialization of the velar stop, while the latter is actually a vowel which comes between the consonant and a following vowel. As in other such ambiguous cases which seem to violate bi uniqueries s , since each occurs in a specific environment, it is in fact overlap, not intersection. Thus the phone [kriw] occurs before [t>], and the sequence [k^w] occurs before [i]. 2. Complex hcmorganic nuclei in the presence of high tone. Examples: [*luudaa'koh] 'they are crazy,' ['thudl 'water,' ['tuu*lih] 'bull,' [*huu'hE*goh] 'hurry,' ['tzuus] 'vagina,' ['yuu*sn] 'God,' ['kuu*ts^ih] 'pig,' [*sLts^uu] 'my paternal grand mother, ' ['k^du] 'here,' [*pL‘tridu9=] 'his water,' [*tuu't?&*hehl 'around here,' [*yudtha'xS3] • 1 1 'Navajo, ' [*t^1dul'kai"y£h]] 'burrow,' [^ss'x^duri] 'casket.' 3. Simple nuclei before /?/ and /h/ in the same syl lable in the presence of high tone. Examples: [*hai'yuh] 'where,' []'mbdh3 'owl,' [^haa'g^dh] 'dove,' ["sii?Ll'tu?=] 'me too.' In this 16? environment also as noted above [u] may alternate with [o] before any consonant in the next syllable in the presence of high tone. The allophone [o] occurs elsewhere. Specifically in the following environments. 1. Complex homorganic nuclei in the presence of low tone. Examples: ['too"lEh] ‘butterfly,1 [*tL‘pool] 'round,' [*sL'g^oo?=] 'my teeth,' ['yoo?=] 'head,« [*?iis'tl?oo] 'I tie,' [^naal'ts^oos] 'paper,' [*?iin.9'soot=] ‘mission ary, ' [ 'look®3 'sore,' [*too?aa'nii"na?=] 'a long time ago,' ['kool 'thus,' ['tl’ooi] 'rope.' 2. Simple syllables before any consonant in the next syllable, or /’/, or /h/ in the same syllable in the presence of low tone. Examples: [*">o'p&9=] 'pinon nut,' [^konfcz'nan] 'ten,' [*lo'g*Ll] 'saddle,' [*no'Wli] 'we,1 [*xto'xwaa*x'dahl 'nothing,1 [*9ii'loh] 'string,' ['ti’oh] 'grass,' [*sLl'tfI*goh] 'she is telling me,' [*naiyo't^iil] 'bring it,' [*?onl'th@k=] 'school.' 3« All environments in the presence of nasalization. Examples: [ ,kh b‘?s] 'fire,' ['t6?=] 'fly, ' [%hiis'ts^oos] 'I kiss,' [*n'k?66s] 'butter,' [^n'zo"nih] 'it is all right,' E^nSa?tt.s 'k^ool 'I iron it,' [#naa'sot=] 'lizard. i 168 4. After another vowel as the result of contraction. Examples: [ 'tl?€.o] 'night' from ['tl?C.go], [?ao] 'yes1 from [*ha'9S?o], [daot’eC.] 'what's happen ing' from ['daagot^Zfl, [*taos'?aah] 'yesterday' from [*taagWus'?aah], [*?aos'tiih] 'it's all right' from [*9aagWus'tiih]. The Vowels: Comments and Comparisons Upon inspection of the preceding section it will be noticed that the major allophones of each vowel phoneme are for the most part simply close and open or tense and lax varieties of vocoids in four different areas of the standard vowel chart. Generally the tense varieties occur in open syllables, while the lax occur in short closed ones. However, one point that may be confusing is the fact that in some utterances a morpheme may occur with a close allophone, and in other utterances the same morpheme will have an open allophone of the same phoneme. This is caused by the occurrence of a foot boundary between mor phemes. This often takes place in separate repetitions of the same word. Foot boundaries occur between morphemes when the speed of utterance is slowed and equal or nearly equal stress is placed upon both the stem and affixed morphemes. I will deal more fully with this under the 169 chapter on higher phonological units, but for the time being we can compare it with the English phrase, the boy. In English when this occurs as a single foot, as it usually does, it has the phonemic shape /da1boy/. If we say it more slowly with emphasis upon each morpheme, however, it becomes /'di + 'boy/. In Apache 'my father' may be either [^sL't^aa] or ['si r 't^aa], and so on. Another problem area in Apache is the back vowels. Especially the vowel /o/. This phoneme is subject to variation in almost any environment. Some of this varia tion may be accounted for by special rules which apply to \ one ideolect or one highly localized dialect. For in stance, some informants seem to have a rule which substi tutes [o] for [u] in simple closed nuclei in the presence of low tone. Others seem to use at all times either [o] or [i/l before / V , no matter what tone is present. Still others use [o] in the environment [k__k=], but use [uj when they pronounce the word with a final [t=]. Thus [*sL'kok=] or [*sL'kot=] or [^sc'kwut=3 'my knee,' all appear in the data. At times one is tempted to say that there is no consistent pattern of occurrence for allophones of /o/. For example, I have the word for 'bug' recorded as ['ts^oos], ['ts^/s], ['ts?os], and ['ts^uusl; all from the same informant. This kind of variation is again 170 related to the speed of the utterance. I have noticed that [u] is heard as [u], and [(_] as [i] even in simple closed nuclei when spoken very deliberately. This may be because the slower repetition causes the syllable peak to change from a simple nucleus to a complex homorganic nucleus, the type of peak in which we usually find the close allophones [u] and [i]. The Apachean Vowel Problem Western Apache is not alone in this variety. All of the Apachean languages exhibit a complex distribution of vowel allophones along with a large amount of seemingly random variation. It is this complexity and variation which I term the Apachean vowel problem. There have been three solutions offered by other analysts in attempts to account for this problem. One way is to treat the phonetic differences be tween close and open allophones, and the overlap caused by variation and/or extremely restricted environments, as contrast. Landar (1963: 7) describes Navajo as having six phonemic vowels: two front, two central, and two back. He feels that there is contrast in analogous environments between [a] and [o], and [ol and [u] (Landar, Ervin, and Horowitz I960: 371:n.8). He cites the forms /^a'gas/ 'one's aorta' and /sa'gas/ 'my aorta' as evidence for the first contrast, and /'?auu?/ 'yes' and /'yoo?/ 'bead' as 171 evidence of the second. I cannot accept these as analo gous pairs. In the first pair / V is not comparable to /s/ as a preceding environment. In fact, in Apache the close [a] often varies with open [a] after / V as in [*9o'kh##] or [*?a'kh#d] 'over there.' In the second pair I find it difficult to compare /y/ with /?/. The two words occur in Apache as ['?ao] 'yes,' and ['yoo?=] 'bead.' I am confident that the phonemic solution I propose for Apache, could, with minor adjustments, be made acceptable for Navajo also. The second solution has been to treat the variation as if there were no pattern whatsoever in the occurrence of vowel allophones. Hoijer (1946c: 63) calls the vari ation among allophones of the vowel /a/ in Chiricahua "apparently haphazard," and the variation among allophones of theyvowel /o/ "entirely random." For Navajo, Hoijer (1945a: 26; Sapir and Hoijer 196?: 11-12) treats these vowels in a similar fashion. Like treatment is given by Bittle (1963: 840) to the vowels in Kiowa-Apache. He notes, however, that in the case of particular words a "systematic preference" is often shown by a given speaker (Bittle 19632 84:n.3). To treat this variation as random seems to be skirting the issue. As the description above demonstrates the variation in these allophones is definitely patterned, 172 albeit complex. What may have given the impression of free variation in the past is the fact that often a single informant will have a set of rules which applies only to his ideolect. This could produce what Bittle termed "sys tematic preference.11 The set of rules given above are to some extent a synthesis, but after many hours of checking and rechecking with my informants I feel that they account for the data rather well. The third solution to this problem is that offered by Durbin (1964: 36) for San Carlos Apache. The data he is trying to account for is similar to that which Bittle, Hoijer, Bandar, and myself are concerned with in other Apachean languages. In his account, however, he describes the close vowels as being glided and writes them as [e^l, [i^] and [uul. He is particularly interested in those environments I have described as "nuclei not followed by any consonant in the same syllable," (open nuclei) which he notes sometimes have glided vowels and at other times have non-glided vowels. His solution consists in positing a phoneme of internal open transition or juncture which apparently is canonically present in some forms, but absent in others. He offers the following evidence (Durbin 1964: 35-36). [e3-], [i3-] and [uul occur utterance final in all cases if no consonant follows in the same syl lable. A problem is posed, however, in the 173 following data. In ["‘Zi1 "lung", the final [i1] occurs in its expected, environment. In.the compound ["?i"tS-iii'soleei], •heart", [ii] does not occur in its expected environment since before /s/, we would ordinarily expect [i]. This phenomenon has a high frequency of occurrence when morpheme boundaries . . . and syllable boundaries coincide. It also occurs at unexpected places in slow speech as in P'jsi1 1 sit *’], "my liver" as contrasted with ["si'se? i], my mouth". These two are compared with ["si 'pit "O 'my stomach". On the basis of the above, it is either neces sary to establish [ei], [ii] and [uul all as sepa rate phonemes which contrast with [ej, [i], and [u] or to establish [i], [e] and [u] as belonging to three separate phonemes along with a phoneme of internal open transition which would contrast with its own absence. His solution is puzzling in the following respects. First, the allophohes in question often occur in "unex pected places in slow speech." My own interpretation has been that slow speech lengthens vowels and breaks up utter ances into more than one foot. Both of these phenomena are likely to affect vowel quality. Second and more impor tant, is his definition of contrast. Nowhere does he present data which demonstrate a meaning change which hinges on the presence or absence of either the vowel allophones in question or the juncture phoneme. As it is generally defined, contrast occurs when one item is sub stituted for another, and the new item changes the quality of the utterance to produce a sequence with a different meaning. In Apache the number of feet in an utterance does not seem to change the meaning of the utterance other 174 than to place differing amounts of emphasis upon the con stituent morphemes or words. I fail to see how a morpheme which means 'my1 and has a glided vowel in slow speech can contrast with the same morpheme in another word which apparently is not pronounced slowly. In order to further explore Durbin's solution I asked my informants the same sets of words: /sizig/ 'my liver,' /size9/ 'my mouth,' /sibig/ 'my stomach.' In all three cases there was allophonic variation between fast and slow repetitions of the word. This indicates that different allophones do occur in relation to foot bound aries. But it also indicates that foot boundaries are not the canonical property of some words, but not others. This removes these boundaries from the segment level of the phonological hierarchy. The description of Western Apache vowels which I have proposed thus accepts the fact that variation in vowel allophones occurs before phonological boundaries, but it does not accept the phonemic status of these boundaries. Furthermore, it takes note of other kinds of allophons variation, but accounts for this variation in terms of special rules rather than reliance on rather inexplicit statements about random variation. 175 The Status of /e/ Finally, I would like to comment upon the phonemic status of /e/. Stanley (1969a: 157) notes that in Navajo /e/ when it occurs in non-stem syllables can be shown to be derived from /a/ or /o/, and that in stem syllables it often alternates with other vowels. Stanley also points out the extremely interesting fact that the phonetic range of the back vowel /o/ is equal to the phonetic range of the front vowels /e/ and /i/ together, and that what I have called a back vowel,/a/, also has a wide phonetic range. (Along with this it might be noted that /a/ is not rounded like /o/ and in fact is back only in that it is not front.) In terms of phonetic and phonemic symmetry then we would do well if we could eliminate /e/ as a sepa rate phoneme. Since I have not gone extensively into the morphology I do not have the data to do what Stanley has done for Navajo, however, his suggestion is worth keeping in mind, as many of the same facts which he cites for Navajo seem to apply for WKA. For example, in some WMA stems /e/ alternates with /i/, thus I have recorded both /hiisbees/ and /hiisbiis/ for *1 boil;1 or the stem /-geel/ 1 to carry a pack,* occurs as /-gil/ in the word •saddle. 1 176 Summary WMA has three classes of segmental phones: con sonants, resonants, and vowels. There are 28 consonants, two resonants, and four vowels. There is a possibility of reducing this inventory by four consonants and one vowel if one accepts the transformational approach to phonology. I will next turn my attention to the way in which these phonemes are arranged to form the next level of the hierarchy, the syllable. CHAPTER 6 HIGHER PHONOLOGICAL UNITS: THE SYLLABLE In the taxonomic theory which is the basis of this description, phonemes combine into syllables which are considered to be the units of the second level of the pho nological hierarchy. In WMA a syllable may generally be considered to have a consonantal onset which is obligatory unless the peak is a syllabic nasal; a peak or nucleus composed of a short vowel, a long vowel (which I consider to be a double vowel), or a short vowel plus a second vowel, either /i o a/; an optional coda composed of one of the consonantal set /’hdgllsszz/; along with the prosodic elements of length, stress, tone, and sometimes nasalization. In the following discussion I will describe the structure of the various syllable types in WMA through the use of a series of syllable structure rules similar to those which make up the phrase structure of a transforma tional grammar. These rules will be generally ordered and context restricted. This should not be taken as a serious theoretical attempt to introduce formalism (in the sense of symbolic rules) to the description. It is merely a 177 178 more convenient but somewhat ad hoc way of describing pat terns present in the data. The accompanying prosodic ele ments require a more lengthy phonetic treatment which I will present following the discussion of syllable struc ture . Syllable Structure The phonemic structure of a syllable in WMA may be correlated with the phonetic existence of a chest pulse. This chest pulse gives rise to the auditory prominence of the vowel, vowels, or resonant which forms the nucleus of the syllable. This nucleus is then surrounded by areas of less auditory prominence which compose the onset and/or coda of the syllable in which the syllable boundaries lie. Boundaries I have not written syllable boundaries in the data because comparison and analysis of clear cut cases (mono syllabic words for example) reveals that •phonemic syllable boundaries are completely predictable in terms of a few simple rules. In the last chapter these phonemic bound aries were taken as a given, but their basis is explained in this chapter. In arriving at a definition of the pho nemic syllable we must deal with the following phonetic problems. First, in many syllables the boundaries are phonetically ambiguous. This latter is true in the case 179 of what are phonemically open syllables followed by a con sonant in the next syllable. In these syllables the syl lable boundary falls phonetically in the middle of the consonant because consonants in such environments are phonetically lengthened. Second, there are the affricates. They are a problem in that they may be interpreted as either consonant clusters or as single segments. Third, there are the nasal plus alveolar or bilabial stop se quences which also may be considered as being either single segments or sequences of segments. And finally, there are the syllabic nasals. A decision must be made about the boundaries of syllables in which they occur. Long; Consonants One of the characteristics of the Apachean lan guages is that consonants which begin a syllable, when preceded by an open syllable will be phonetically length ened. In such cases it is difficult to interpret whether there is a consonant sequence with a syllable boundary between two identical consonants, or whether it is a single consonant. Clearly unambiguous monosyllabic words provide a clue, however. When prefixed by syllables which end in a vowel they lengthen their initial consonant in the presence of such a vowel. For example: 'horse* becomes [*stl.'lii’=] 'my horse, 1 and [ 'tU] 'blood, ' becomes [*st.t.'tt-l] 'my blood.' On this basis we may 180 further interpret all lengthened consonants as single seg ments, and say that the phonemic point of syllable division is before them even though the phonetic point of syllable division is within them. I will say more about this in the next chapter under the section on the foot level prosody of timing. Affricates So far as the affricates are concerned, if we examine a number of single syllable words which clearly begin and/or end in unambiguous phonemes we find that no where in the data are there random sequences of consonant phonemes. All such words conform clearly to a /CV(V)(b)/ pattern. For example: [•s£l] 'I' H'ko] 'thus' ['k?9t=] 'now' [ 'look=] 'sore' ['sag] 'bear' [ 'xwus] 'cactus' ['!££?=] 'horse' ['thiih] ' ice' 1 1 tit This fact argues for the treatment of the affricates as single segments. To do otherwise would seriously compli cate syllable patterns in the language. Thus words like [ 1 tsh& £"] 1 stone, * [*tl?oh] ’grass,' [ts^aal 'beaver, 1 [ts?aa] 'hat,' ['tlo?=] 'bird,' [tzaa"ge] 'antelope,' [*tL'tliis] 'he passes out,1 [*sL'ts?Ln] 'my bone,' that contain affricates are to be interpreted as having single segments and are written phoneraically as //ee/, A 7oh/, 181 /caa/, /c?aa/, /Lo?/, /jaage/, /diLiis/, /si^’in/ respec tively. Nasal Plus Stop I have recorded words such as [*n'taa] 1 sit down,‘ [*m'pii] 1 third person,1 [*m'pai"y3n] 'middle aged man,1 [*n'tEg] 'man,' [*n'ta?=] 'Whiteman,' and [*n't€8z] 1 I I 'tall.' ■ Again, a decision as to whether they are single segments or sequences of segments must be made. These are fairly easily taken care of, however. For example, reverse sequences of the phonemes /n/ plus /d/ occur as in [*si?dd*dn] 'the one who is my wife,' or ["t^uu^dn'ts^ii] I 1 'whiskey.' I have already demonstrated that the sequence /m/ plus /b/ actually contrasts with the phonetically dif ferent segment /mb/ (cf. the /mb/ phoneme in Chapter 5)• Finally, there are the phonetic facts that the nasal in all these cases is long, receives its own stress, is the nucleus of its own chest pulse and has its own tonal quality. In other words, it is syllabic. These facts demonstrate that nasal plus consonant sequences are in fact sequences, not single segments. Syllabic Nasals The last problem in assigning syllable boundaries concerns the location of the boundary when the nasal fills the nucleus of the syllable. There are in the data 182 various sequences of nasal plus consonant or consonant plus nasal such as [sn] in [*?i'yu&*sn] 'Jesus,1 [dn] as in [*si?6&*dn] 'my wife,' [nil such as in [*nl't^e-t-l 'it is good,' [ns] as in [^pigons'^aa] 'I am learning,' [nz] as in [*n'zo] 'it is fine.' These, too, may be handled on the same basis as other syllable divisions. There are no consonant clusters so in the case of [nCCV] the break comes between the two consonants. Thus in the example 'it is good,' the syllable division would be [*n!.'t?&2]. There are also no syl lables which begin in only a vowel, so in the case of [nCV] the break comes before the consonant. This would divide the word 'it is fine' as [*n.'zo]. In the case of [VCn] sequences there seems to be a parallel to [VCV] se quences, particularly where they arise as the result of a syllabic nasal being suffixed to a stem. For example, in the words above for 'Jesus' and 'the one who is my wife' the syllabic nasal is a relative suffix sometimes used for human nouns. This is parallel to the relativizer /-!£/, and its allomorphs, which also is frequently suffixed to morphemes which end in consonants. Look at the following cases: ['kLsl 'cane,' which becomes [*pL.'kL."s e ] 'that which is her cane,' [*to.'tl?Lzl 'blue' which becomes [*to.'tl9 L."zll] 'that which is blue,' ['k9ot=1 'soon' which becomes [ 'k9(_. "d££] 'right now.' In each case the 183 stem final consonant upon being followed by a vocalic syl lable nucleus ceases to function as the coda to the stem syllable and becomes the onset to the suffix syllable. By analogy then it would seem that a stem syllable, or any syllable for that matter which ends in a consonant and to which is attached a syllabic nasal, may be considered to have lost its final consonant, the point of syllable division coming before the consonant, producing /Cn/ syl lables. Thus the syllable division of 'Jesus' is [*?i.'yud.*sn] and in 'the one who is my wife1 is [*si.'9dd.*dn]. This means that syllables with resonants at their nucleus are generally like other syllables in terms of their form, the exception to this being that such syl lables need not have a consonant onset when they occur in foot initial position, or immediately after another vowel or resonant. This produces /nC/ and /n/ syllable types as in [^ns.'tscz] 'I am tali' or [n.'tctz] 'he is tall.' I I Syllable Boundary Rules With respect to consonants and resonants, phoneme syllable boundaries in WMA can be correctly placed in terms of the following set of rules: 184 2 (ClBftO (• 3. RR ----^ R.R These rules should be read as: (1) a consonant or resonant followed by a vowel has a syllable boundary before it (the period indicates the boundary); (2) a sequence of a resonant optionally preceded by a consonant and followed by a foot boundary, consonant, or resonant, has a syllable boundary before it and the preceding consonant if any; (3) a resonant followed by a resonant has a syllable boundary between them. Nuclei There is the further problem in WMA that the chest pulse between these boundaries is not always channeled through the same vowel quality or is sometimes held longer than others. (Length in this sense should not be confused with the phonetic features of length discussed below which are predictable.) I have to this point described these syllables as being simple and complex, the complex variety being either heterorganic or homorganic. I indicated by the way I wrote them phonetically that I considered all complex nuclei to be composed of two vowels either of iden tical or non-identical qualities. I now wish to justify this assumption. 185 There is ample evidence in the data that what .1 have called homorganic complex nuclei contrast with simple nuclei. There are minimal pairs like: /sinii9/ 'my face1 /siniV 'my land' Or there are numerous analogous pairs like: 1. /sitaa/ 'my father' /sitS?/ 'my forehead* 2. /ndaa/ 'sit down' /nda?/ 'White man* 3. /?it?aa?/ ' its leaf 1 I /nadS?/ 'corn' 1 4. /siloog/ 'my muscle' /log/ 'fish' There are further numerous examples of words with phonetically diverse syllable nuclei including: [*lL'kal] 'white,' ['?af] 'that,' [*hai'yul 'where,' [^pcnail'paa] 'he is driving,' [*nl&i'sLsl 'shirt,' [^nl't^&o] 'it's being good,' [*n'tdf] 'mountain lion,' [*taa'khuf] 'how much, 1 [^naasL'k^uI] 'he vomits, ' [^sL'g%i"ye] 'my mater nal grandmother,' [*tao't?&t] 'how are you,' [*sul't£6] 'he is telling me.' These represent combinations of /a/ plus /i/ and /o/, /e/ plus /i/ and /o/, and /o/ plus /!/. It is obvious that many of these are formed by contraction or deletion of intervening consonants, but the fact 186 remains that they are acceptable normal speech sequences in the language. Given just these pieces of information one could argue that a VV interpretation is incorrect. That the long homorganic nuclei should be written as /i:/, /e :/, /a:/, and /o:/; and that the clusters should be written as either vowels plus glides, i .e., /ay/, /oy/, / ^ /, /aw/, /ew/> iw/, or as vowels plus semi-vowels, i.e., /ay/, /oy/, /ey/, /aw/, /ew/, /iw/; or as six more phonemes, namely /a1/, /e5-/, /o1/, /aV» /eu/, /iu/. There are actually two problems here, then. First, what is the best interpretation of complex heterorganic nuclei, and second, what is the best interpretation of complex homorganic nuclei. Heterorganic Nuclei These are more important for it is from this source that we interpret the others. For these nuclei evidence in favor of the VV interpretation comes from: Symmetry. If we treat the vowels which come first in these nuclei as being followed by either vowel or semi vowel glides, or as being followed by semivowels; in order to be consistent we should also treat long or doubled /i/ and /o/ the same way. For example, if we say that ['^al] 'that one' is phonemically /’ay/ or /’ay/ or /’&V, then we should treat ['sii] 'first person' as phonemically 18? /si#/ or /s£y/ or /s£V. Similarly, if we treat [*nais'paa] 'I am driving1 as phonemically /naysbaa/ or i i ii /naysbaa/, or /na1sb^§,/, then we should treat [*?iis'££] 11 see it1 as /?iys?££/ or /’i^sli/ or /?i^s?££/. To do it ii it so would mean that only /a/ and /e/ could occur in the lengthened or doubled form, all other "complex" nuclei being composed of vowel plus glide or semivowel. The point is homorganic nuclei which end in high back or high front vocoids cannot be treated any differently than are heterorganic nuclei. And if they are treated in this way a system is produced which has long or double low vowels, but no comparable long or doubled high vowels. Of course to consider the third segment in [*?ao] 'yes,' and similar syllables, as a semivowel would further complicate the syllable patterns and phonemic system in that a /w/ phoneme, an entirely new phoneme, would have to be added to the system. Phonetics. I hear nothing in these complex heter organic syllables which does not sound like a vowel. In some cases the second vowel has a different tone from the first producing a tonal glide. (See following section on tone for examples.) Economy. A solution which posits a series of glided phonemes such as /a5-/, /eV, /oV, /au/, /eu/, and /iu/ , adds six new phonemes to the system which does not seem to be necessary. 188 Loan Words. There is only one English loan word which is relevant here and it is /bia/ ‘beer.1 It is definitely a one syllable word. Yet to accept this as a single syllable introduces a new vowel cluster to the lan guage, one which is the reverse of one of the suspect clusters. It would seem that Apaches upon hearing the final English semivowel in ‘beer,‘ fit this into a W syl lable pattern. Its presence in the language thus seems to support the W interpretation. Other Apachean Languages. Various vowel clusters and combinations have been reported and accepted as a pos sibility in Navajo (Hoijer 19^8:28; Sapir and Hoijer 196711) and Chiricahua (Hoijer 1946c: 65). It seems to be currently agreed, judging from the literature, that syllable nuclei in these languages may be composed of more than one vowel, although not all combinations of vowels are possible, /i/ and /o/ are the most common second vowels in other Apachean vowel clusters. Homorreanic Nuclei It is primarily on the basis of heterorganic nuclei patterns that I interpret lengthened vowels to W clusters as well. However, there are other reasons. Phonetics. Again there seems to be no phonetic reason to consider these as different from short vowels in sequence. They have the same vowel quality as short 189 vowels in open syllables, and there are occasionally se quences where the tones on the two vowels are different, demonstrating their doubled vowel status (see following section on tone for examples). Morphoohonemic Evidence. There are a limited number of forms in the language which, uninflected, have a definitely short or simple nuclei. For example: [^pL'zaz] /bizaz/ ‘its offspring' [*sL‘tshak=] /sigfag/ ‘my rib' [*%L‘tshuk=] /lijz(6g/ 'yellow' But when these are inflected with a relativizer they be come: [*pL'zaa"zii] /bizaazii/ 'its little one' C*sL'ts*1aa"g££] /si^aag££/ 'those which are my 11 *1 ribs' [*lL'ts^uugii] /li^oogii/ 'that which is yellow' My interpretation of this fact is that the vowel is doubled. If we consider there to be two series of vowels, /V/ and /V:/ we have to posit a rule which changes one phoneme into another. If, however, we consider long nuclei to be vowel clusters we can write the rule so that the stem vowel is simply copied. It must be stated as: ^1^1^2 ---- > CiViViCg /__££, with the understanding that it applies only to this class of stems. Economy. To posit long as well as short vowels in the system would increase the number of vowels from four 190 to eight, which have, at least in certain positions, the same vowel qualities. For example, short nuclei when open have the same quality as long nuclei. Other Anachean Languages. Generally linguists working in the Apachean languages have accepted that long vowels are actually VV clusters of identical quality. -If they have not gone along with this interpretation in a literal sense they have at least done so for convenience in rule writing. Stanley (1969a: 39), who believes vowel length to be predictable in Navajo within the frameworkcdf a transformational model notes:' We will always represent long vowels by doubling the vowel letter. Still, it should be clearly understood that this is an orthographic device only; long vowels are single segments as far as the rules we will present are concerned. Durbin (1964: 38) uses this interpretation for San Carlos Apache and bases it upon the possibility of vowels in ;the same syllable having different tones. Hoijer again.for Navajo (Sapir and Hoijer 1967i 10) apparently accepts quite literally a double vowel interpretation when he states with regard to long vowels that they are: Clusters which are made up of identical pho nemes: aa, ee, ii, 0 0 , aa, ee, ii, and 0 0 . These are pronounced lik&'loA^ m6AophthoA£s and have a duration approximately twice that, of a single vowel. Similarly Bittle (19631 84) notes for Kiowa-Apache, A syllabic may consist of one or two vowel phonemes . . . . In all two-vowelled syllables, 191 the vowels are identical, and the syllabic is pronounced like a single long vowel. Finally, it might be mentioned that the generally accepted orthography for Navajo, that of Young and Morgan (1964: 1), uses the doubled vowel representation. That this is accepted by people who speak a language very simi lar to WMA would seem to argue that if at least it is not cognitively valid, neither is it particularly objection able. On the basis of these arguments then, I have arrived at the above interpretation of WMA syllable struc ture. Having accepted such an interpretation I will now consider possible manifestations of this structure. Syllable Types Like the allophones of different phonemes, syl lables in WMA seem best described in terms of stem and non-stem types. The essential differences between the two types are found in: (1) Differences in prosodies. In non-stem syllables various prosodic elements are not so prominent. (2) Differences in syllable nuclei. Non-stem syllables may have as their nucleus syllabic nasals while stem syllables always have vowels. (3) Differences in manifestating forms. Stem syllables seem to have a greater number of manifestating forms. Whether this is actually a significant fact of the language, or merely a 192 function of the fact that most non-stem syllables repre sent morphemes which are members of closed classes and thus are limited as to novel combinations is debatable. (4) Differences in phonotactic changes. Stem syllables are characterized by more significant phonotactic and morphophonemic changes than non-stem syllables. We may describe the structure of syllable types in WMA with the following set of rules: 1. Stem S -j> V1 (V2)(C2 ). + Stem Prosodies 2. Non-stem S ■ > :e 2' Vi (V2>(c4> +Non-stem .P/ Prosodies (C5 )B(C6) 3. C2 > /s, s, z, z, h, 1, 1, d, g, V 4. R-j_ > /m, n/ 5* Rg > [/m, n/ /___ loan words J /n/ elsewhere 193 6. //p/ /___the loan word for \ 'fish' \ /mb/ /___the archaic morphemes\ for 'owl' and \ 'coyote' 1 /b, d, g, t, k, t’, k? s, z, j, /___V1 (V2 ) ( h, g, y, 1,1, L, L, L V /b, d, g, ?, t, k, f , k? s, z, j, c, c?, /___V (V ) /s; h, g, y, 1, 1 L, L, L V /b, d, g, ?, elsewhere t, k, t?, k?, ^ f z| j| ^i i ^ n Zi z, J, c, c h, g, y, 1, 1 L, L, L V 7. % /’, h, s, s, z, 1, 1/ 8. Cj / _ V i ( V g ) /---Vi(V2 ) _ elsewhere z, J c9, h, y, 1 1, L/ 9. C5 -> /d, s, s, z, z, 1, 1, 9, h/ 10. c 6 -> /l, s, *?, s/ 11. Vo /a/ /loan word for 'beer'_ /i, o, V^/ elsewhere 12. Vi 71/ /_____ /a/ ^ /i, e, a, o/ elsewhere) The Rules Rule 1. Rule 1 is intended to account for the basic structure of a stem syllable. The onset is composed of a resonant functioning as a consonant, or a true con sonant. This is followed by a nucleus composed of either 195 one or two vowels. Finally, there is an optional coda composed of a consonant. This unit is preceded and fol lowed by syllable boundaries indicated by the periods whose placement conforms to the syllable boundary rules given earlier in the chapter. This basic formula produces the following general syllable types which are exemplified by words listed from the data. (In polysyllabic words I have underlined the relevant syllable; a practice I will follow throughout the rest of the chapter.) Types Example c w c /bees/ 'knife' cvc /k?ad/ 'soon' cv /&£/ 'this' cvv /sl£/ 'I' RVVC /sinii?/ 'my face' RW /nSmaase/ 'potatoe' RVC /nil/ 'with you' RV /la’nne/ 'you are carrying it Rule 2 . Provides the basic forms for the non-stem syllable. Note that one difference between the stem and non-stem syllable is that the non-stem syllable may have an R, CR, or RC shape. This formula generates the follow ing general syllable types which are exemplified by words listed from the data: 196 Type Example CWC /hiisliz/ 'I urinate' CVC /disgis/ 'I am lazy' CV /digis/ 'he is lazy' cvv /taadaadigis/ 1 they bathe' OR /’is^’anhn/ 1 that particular woman' R /nzo/ 'it's good' RC /nlt’ee/ 'it's good' RVVC /naisbaa/ 'I am driving' 1 1 RVC /nazbaa/ 'he raids' RV /nSmaase/ 'potatoe' CRC /hasnskeh/ •I fight' Rule 3. Specifies the phonemic set from which stem syllable final consonants may be drawn. It is of interest that all of the fricatives except /g/ and /y/ may occur here as may all of the unaspirate stops except /b/. Otherwise no other consonant may occur in the surface structure in this position. This rule produces syllables with the following forms: Form Example CVV’ /sinii’/ 'my face' CWh /haadiseeh/ •I spit' C W s /c’oos/ 'bug' CVVs . /yaak’oos/ 'cloud' CVVz /’igeez/ 'egg' 197 Form Example C W z /ndeez/ 'long' CVV1 /nilzool/ 'it swells' CVV1 /dibool/ 'round1 C W d /naasood/ 'it drags' CVVg /haiyinihiySSg;/ 'he is driving them away CV? /simeV 'my baby' CVh /tih/ 'see' CVs /zas/ 'snow' CVz /naagodesgiz/ 'he is crazy' CVs /sas/ 'bear' CVz /ciz/ 'wood' CV1 /dil/ 'blood' CV1 /cm/ 'flower' CVd /sibid/ 'my stomach* CVg /log/ 'fish' It seems likely, however, that with some stems, other consonants may occur in stem final position in the canonical form. Compare the following: 1. /nak££/ 1 two1 2. /^esLal/ 'five' /nadin/ 'twenty' /’esLa’din/ 'fifty' In both sets the cardinal numerals seem to end in the mor pheme /-££/ 'that which is,' but when they become decimals they drop the final /-££/ leaving a stem to which is suf fixed /-din/. In the word for 1 fifty' an underlying /?/, 198 which is part of the stem ‘ five1 (/’e s L a V ) , tut is de leted in /’esLal/ "because of contraction, becomes appar ent. In the word for 1 twenty * however, we see no sign of the /k/ in /naakli/. It is apparently the final phoneme in the stem 1 two1 (/naak/) but because of the surface structure constraints upon syllable final phonemes, it disappears. Cook (1971b) called my attention to this. Rule 4 . Provides for stem syllable initial reso nants. There are no limitations upon these resonants but it should be repeated here that /m/ is relatively rare, and actually does not occur in the data before the vowels /i/ and /o/. I am making the assumption, however, that this is coincidental and that there is no potential re striction upon the distribution of /m/. If there is an actual restriction upon its distribution the rules would need to be reordered, with this rule placed after the vowels have been introduced. Then this rule would be stated so that R^ could be /m/ only before /e/ and /a/, but /n/ before all vowels. Rule 5. Gives the forms which resonants take in non-stem syllables. In non-stem syllables /m/ occurs only in loan words while /n/ may occur in either. Thus the rule is structured to make this lexical restriction upon /m/ clear. As to whether /n/ will then be syllabic or non-syllabic is handled by the allophonic and syllable division rules. 199 Rule 6 , Specifies the limitations for the two ■special" phonemes /p/ and /mb/ in the first part of the rule, and then goes on to place a special restriction upon the co-occurrence of grooved fricatives and affricates in the and C2 positions. This restriction is simply that the grooved obstruents in one syllable must be either all alveolar or all alveopalatal. Thus syllables found in words like: /sas/ •bear1 /ciz/ •wood• /c’oos/ •worm• - /jos/ 1vagina• /’ilzis/ •he dances' are possible, and so are syllables in words like: /zas/ • snow' /?izis/ •bag' /naalgfoos/ •paper' /9 igloos/ •feather' /haadis.jiis/ •I pull it' Of course syllables which end in /s z s z/ can begin in any of the other consonants as well, and sylla bles which begin in any of the grooved obstruents can end in any of the other consonants. But syllables like: 200 are impossible. Finally, the last section of Rule 6 allows all other possible stem syllable sequences providing that none of the preceding conditions exist. Examples of these may be found in Chapter 5* Rule 7. Gives the possible phonemes which may function as the coda of vowel centered non-stem syllables. Note that this is the same set as for stem syllables minus /d g z/. Nowhere in the data do these appear syllable finally in non-stem syllables, at least so far as the sur face representation of such syllables is concerned. The absence of /d/ at this point is not particularly surprising since it apparently occurs in non-stem syllable final posi tion in few of the Athapaskan languages, being deleted historically by a process known as the ‘D* effect (Howren 1971; Hoijer 1945c; 1946b). This ‘D1 effect is currently viewed as a morphophonemic process in Athapaskan which takes prefix final /d/ and either assimilates or deletes it before stem initial consonants (Howren 1971: 99). Since /d/ does not occur on the surface prefix syllable finally in UMA, Rule 7 is correctly stated as is. However, if we take into consideration the fact that prefix morphemes in the lexicon sometimes must be given a canonical form which includes a final /d/ in order to account for 'D' effect alternations in stem initial consonants, the rule should be revised to add /d/ to the set of possible non-stem 201 final consonants. More will be said about the 'D1 effect later in this chapter under the discussion of phonotactic and morphophonemic rules. With this stipulation in mind, I think we can otherwise accept as being accurate the limitation upon the occurrence of J?J and /g/ in this position, and posit only non-stem syllables of the following types in the surface structure of the language. Types Example GW? /naa7diskoo/ 'he irons' \ CWh /daahsistiih/ 'I put a long, slender 1 1 object down‘ CVVs /taanaasgis/ 'I wash it' CVVs /7aasdii/ 'I speak' CVVz /naazbaah/ 'he raids' CVV1 /7aalhoz/ 'it tickles' CW1 /naal^oos/ 'paper' CV7 /la7na7ne7ne/ 'do you have any7 ' CVh /leehohkooh/ 'you come together' CVs /haadasbih/ 'it's full' Rule 8. Relates to the filler of the onset slot in vowel centered non-stem nuclei. This is the same set as the stem syllable minus g L L? t7 k? mb p/. Again, whether their absence is due to pattern or chance is difficult to determine. Certainly /mb/ and /p/ are due to pattern; the others however may be due to chance. It 202 may also be mentioned that there are no examples in the data of /L/ or /z/ occurring in prefix syllables, but only in enclitic or suffix syllables. The generality of such limitations must also remain open to question. Examples can be found under the relevant sections in Chapter 5* The first two parts of Rule 8 describe the same limitations upon grooved fricatives as was previously noted for stem syllables. There are relatively few ex amples in the data of non-stem syllables which contain two grooved fricatives, but the ones that are present seem to confirm this limitation. For example: /?an sishas/ 'he bites him' and /mb£i ndasistaah/ 'they (several) bit you,1 /jisheh/ 'I kill it,' /nSdslstiid/ 'I am full.' Rule 9 . Indicates the limited set of consonantal phonemes which may serve as onsets to a syllable which has a resonant as its nucleus. These are primarily syllables which result from the addition of the suffix /-n/ 'rela- tivizer (human)' to stems which end in consonants, although such syllables occasionally occur before the stem as well. Note that it is the basic stem-final set minus /g/. /g/ apparently does not occur prior to syllabic nasals, but assimilates to /d/ in that environment. /?/ at times also assimilates to /d/ when it occurs in stem final position prior to syllabic nasals, but /?n/ syllables occur in other non-stem positions. For further information see the 203 discussion of the /d/ phoneme in Chapter 5* Examples of syllables of this type from my data include: Forms Examoles / w /nat?oh la?sa9nt?ii/ 'give me one cigarette1 1 1 /dn/ /si’aadn/ 'the one who is my wife' /sn/ /’iytiusn/ 'God' /zn/ /ndeezn/ 'the tall one' /hn/ /?is/9&nhn/ 'that particular woman A n / /dilhiln/ 'the black one' /sn/ /hasnke/ 'you fight' I have no examples of /zn/, or /In/ but I assume they are possible because there are stems which end in these phonemes to which /-n/ could be attached. For ex ample /goliz££/ means 'that which is a skunk,1 but it seems reasonable to suppose that /golizn/ means 'skunk person,' thus producing a /zn/ syllable. Rule 10. Provides for the phonemic filler of the coda of resonant centered syllables. These usually occur only as prefixes and seem severly limited in actual form. Forms and examples are as follows: Forms Examples /ns/ /nsdeez/ 'I am tali' /nl/ /nlt’ee/ 'it's good' /ns/ /bigons’aa/ 'I am learning' / n V /len?/ 'customary past It is doubtful whether other forms can occur. 204 Note that this rule combined with- Rule 9 also pro vides for syllables of the shape: ) Few of these occur in the data. Whether they all are pos sible is open to question. But the existence of /hasnskeh/ 'I fight1 and /n&idnlgfoos/ 'he passes a paper like object1 certainly indicates such a possibility. Rule 11. Specifies a limitation on the vowel which may function in the second vowel position. First it pro vides that the vowel /a/ may appear in the loan word for beer, and second that the vowels /o/ and /i/ may occur; or that the first vowel may be copied no matter what it is. Rule 12. Places the vowel /i/ before /a/, a nucleus which only occurs in the word for 'beer.' It then specifies that the first vowel may be any one of the four vowels in the system in other possible second vowel en vironments. This produces all the double vowels as well as the heterorganic clusters /ei/, /ai/, /oi/, /io/, /eo/, and /ao/. It actually specifies that some homorganic clusters may come from two sources. One from the redupli cation of the first vowel, and the other from the vowel 205 sequence /!/ plus /i/ or /o/ plus /o/. This would, account for the possibility that in fact some long vowels are pro duced by contraption, with the deletion of an intervening consonant, while others are "really" long nuclei. Ex amples of vowel clusters have already been given. This concludes the discussion of syllable types and syllable structure. Since all syllables have prosodic features as well as structural features, I will now turn to a consideration of the prosodies of syllables in WKA. Prosodic Elements All syllables in WMA are characterized by prosodic elements of stress, tone, length, and the presence or absence of nasalisation. Tone There can be no doubt that WMA is a tone language, although compared to Chinese, Yoruba, or Mazateco it has a relatively simple tone system. There are but two phonemic tones in WMA, both of which have predictable allotones. Minimal or analogous pairs are not hard to find; for example: /bidSS/ * its seed' vs. /bidaa/ 'his eye' /yS5/ 'louse' vs. /yaa/ 'sky' /sinii’/ 'my nostril' vs. /sinii9/ 'my face' /diy55/ 'I go' vs. /diyaa/ 'he goes' 206 As can be seen in these examples, WKA exhibits both inherent contrasting tone as is evidenced by the first three examples, and tone which functions to indicate a grammatical or syntactic difference, as is evidenced by the last example. Following Cook (1971a: 168-69) I will term the former lexical tone and the later paradigmatic tone. The two phonemic tones in VJKA are quite variable even within the same phonological foot. I noticed this fact when first working with the language since I was not consistent in writing tone on the same word upon different repetitions. I thought that it was simply because I was not yet accustomed to a tone language. Usually my mis takes consisted of writing high tone on some morphemes when said in isolation, which in the environment of true high tone morphemes were heard to be definitely low tone. Thus the stem in /simaa/ 'my mother' sounded high, but in. /simaa’ii/ 'the one who is my mother' with a high tone suffix the stem sounded quite low. I came to realize that these mistakes were caused by phonetic variations in tones. A clue to just what these variants were was offered by an exercise in phonetic and phonemic tone analysis based upon Western Apache which I discovered in a workbook in phono logical analysis (Robinson 1970: 138-39). The data for this problem was collected by Faith Hill. Her solution 20? called for the prediction of phonetic tone in Western Apache on the basis of the phonemic shape of the stem syl lable. In the problem she recorded six phonetic tones on the stem syllable; three low and three high. High was written as a [l] and low as a [6]. The solution to the data presented was that tones [l] and [4] fell on open syllables, [2] and [5] fell on closed syllables ending in voiced consonants, and [3] and [6], the lowest, fell on closed syllables ending in voiceless consonants. My own description does not agree exactly with Hill's. I feel that the insistence on a consistent six way difference in phonetic tones on stem syllables is too rigid. I used Hill's suggestion and elicited verb forms in the following frame /______doleel/ •. . » it will happen.' In close to one hundred verb forms I did not hear consistently the pattern she reports. Rather, I heard, relative differences between syllables in the same phono logical foot. These relative differences may be summar ized as the following general principles: 1. Prefix syllables tend to be intermediately low or high. 2. Open stem and coda syllables tend to be higher than any other syllables. 3. Low prefix syllables tend to be higher when adja cent to high tone syllables. 208 4. Stems which end in voiceless fricatives tend to be lower than other syllables. It must be noted, however, that these tendencies pertain only within one phonological foot, for feet at different points in a phonological line may be absolutely higher or lower than other feet with essentially the same syllabic makeup. At the risk of being over specific, to illustrate these principles, I, too, will use a number system. I will symbolize high tone, which is marked phonemically, with /"/, with the numbers [^], [^], and [&]. I will symbolize low tone with the numbers [1], [2], and [3]. Different sequences of numbers in successive syllables or in the same syllable represent slight dips or rises in tone. They cannot be taken to represent absolute tone levels, however. Examples of Principle 1 . The following phonologi cal feet illustrate the intermediate character of prefix syllables: [* *s'2*maa33 /simaa/ 'my mother' [f *n^'taaz22 /ndaaz/ 'it's heavy' [+ ^ko^'t^EElll 4] /goteel/ 'it's wide' [+ *'nai^3dnl2«tshoogll 4] /naidnl^oos/ 'he gives a paper-like object' [f *naa^gul^'zuh5 4] /nSSgolzdh/ 'she sweeps' 209 [+ hi5.33• f 1 /hii’ii/ ’he sees ’ ii “ i i [+ *’Ll^kk&^rj3'kh£.£.s5.5 +] /’llkehkees/ ‘he cuts it1 Examples of Principle 2. The following pairs of phonological feet serve to illustrate how open stem syl lables tend to be higher than other syllables. In each case the first member of the pair ends in an open syllable of either high or low tone while the second ends in a closed syllable of either high or low tone while the second ends in a closed syllable of the same phonemic tone, but of course relatively lower phonetic tone. 1. [+ *si.2iraaa33 /simaa/ 'my mother1 [+ *naa22hn4’maas-^ +] /naahnmaas/ 'he spins around' 2. [4- ’t^aa33 +] /sitaa/ 'my father' [f *ko^ ’ t^aal^-1 /gotaal/ 'a dance’ 3- [+ * h a a 22& i 2 1 ,?aa33 +] /haadi’aa/ 'he sings' [t. *’ii^ d o 2 ” aali:L t] /^xldo’aal/ 'he dances' 4. [•v 1 t^uu^^ +] /t65/ 'water' [+ ^ p L 2 1 t^UU^s^ 4] /bit66?/ 'his water1 5. [+ * s L 3 *pii66 /sebll/ 'eight' [4 * h i i s 3 3 »piis4 4 /hiisbiis/ 'I boil it' 6 . [+ *sii33,taa66 4] /siidaa/ 'he sits' [4 hii33'kaal^ 4] /hiigaal/ 'he comes' Examples of Principle 3. The following feet illus trate how low tone syllables are slightly higher in tone when adjacent to high tone syllables. The relevant syl lables are underlined. 210 [+ *?ii22dii22n3'tlaak=^ +] •it shines' T — {*+ *taa22p-o31 k£. &t = ^ +] 'they dig' [4. *taa22cbii3»?aa^^ 4J 'they sent them' [4 *taa.22hui3' + 1 'they are called' [4 "nLl'ko2'yE&66 fl 'he worries' Examples of Principle 4 . The pairs presented under Principle 2 adequately illustrate the lower tone found on syllables which end in voiceless fricatives. Location in the Line. The entirely relative na ture of the phonetic tones described here is more clearly exemplified if we look at some of the verb forms along with superimposed lines indicating high and low tonal re lationships over an entire phonological line. Note that the last foot in the line tends to be absolutely lower than any other foot in the line. For example: [/*993'khuu66 ^naa^go^nl^‘k^aa’*^ f ^to^lZ^^##] 'over there he will make war' [ A n ^ ' t££z3 + *to2^l&2ll##] 'he will be tali' [/*taa2wo3 »k€.£-t=Zj' + "-to2 'lEti1#,7/] 'jbhev will dig it ' [/'P9n36/*?ai-52yiil'ie9=2 + *to^'l&El^##] 'she will make the bread' [/*?Ll2 'xWUsl 4- -"•t02' l£21##] 'he will sleep* 211 The Causes of Tone Variation. The causes which underlie these slight differences in tone are quite ob viously stress and phonetic syllable length. As I will point out, the stem syllable, where most of the tonal variation takes place, receives primary stress and varies in length according to its shape. The longest syllables are open, which corresponds to where the highest tones appear. Syllables which end in voiced consonants are next in length, and syllables which end in voiceless fricatives are shortest of all. As one might expect there is thus a correlation between higher tone and length of voicing be cause both are a function of the vocal cords. Tonal Glides. There are also tonal glides in WMA. ' Upon investigation these usually seem to be the result of the contraction of two syllables (and morphemes) with dif ferent tones into one syllable to produce the glide. Thus all tonally uneven syllables have complex nuclei. For example: [•*iL2«pai36]] which is a shortened form of [*lL2paa3"hi&] 'that which is brown,1 or [^nl^,t,eo^3]] which is a form of [^nl^•t? "go^3 1 it is being good,' or [*9gs2'tlai36] 'five*which seems to be a form of [*96s2'tlaa339ii6G] 'that which is five,1 or [_1saa^3n3] 'old woman' which is probably derived from ['sa^*?a3n3] 'the one (human) who is old.' Tone in Other Apachean Languages. The whole no tion of tone in Apachean languages is another area in 212 which the literature seems conflicting. Recently Saville (1969: 55) and Stanley (1969a: 19, 39) have "both suggested that tonal contrasts are not distinctive in Navajo. Their solution is to posit that all high tone syllables are the result of the absorption of a final consonant (Saville uses an imaginary velar nasal to trigger her rule) and Stanley with what seems to be much sounder motivation uses /n/ or /?/.) Again, as in other transformational accounts of modern languages, these authors have relied upon com parative evidence. Many cognate forms between Mattole, Hupa, and Navajo reveal that in places where Navajo has high tone, these Pacific Coast Athapaskan languages have final nasals (Saville 1969: 5^) • Languages as diverse as Chinese and Mayan suggest the further linguistic possi bility that high tone may be associated with lost voiceless final consonants, and Stanley (1969a) applies these possibilities in Navajo. Whether the same sugges tions will work for VJMA is questionable; but there is some evidence which suggests the possibility. This evidence consists of the general phonetic tendency which I have noted for stem syllables that end in a vowel to be pho netically higher in tone than those that do not. Stress The second prosodity of the syllable in UHA is stress. Every syllable in VJMA is accompanied by one of 213 four degrees of stress. Unlike tone, however, stress is predictable on the basis of grammatical facts: stem or non-stem, and/or phonological facts: syllable shape, tone and location in the phonological foot. But stress like tone is also subject to a considerable degree of variation, particularly in post-stem syllables. The following rules are again best considered as rules of thumb, rather than an exact account of stress possibilities. Things like emphasis, location in the line and contour, and speed of utterance all affect stress to produce this variation. Primary Stress Primary stress ([']) falls on the stem syllable of a grammatical word, which is the peak of a phonological foot. Primary stress also falls on all syllables produced in isolation, and thus any word produced very slowly syl lable by syllable will have as many primary stresses as it has syllables. (The same could probably be said of any language if an informant were asked to produce a word syllable by syllable.) Secondary Stress Secondary stress (["]) falls on noun, verb, and postpositional stems when they function as subordinate parts of compound words. It also falls on post stem syl lables with a CV(V) or CR form when: (1) they end in a 214 glottal stop; (2) they are high tone preceded by low tone; (3) they are preceded by a syllable which ends in a con sonant. Again, secondary stress is placed on stems in terms of their grammatical function, but it is placed on post stem syllables (enclitics) in terms of their environ ment or phonemic shape. Occasionally, some syllables which are stems in one compound appear in other compounds without even secondary stress. This apparently has something to do with their being enclitics. See the section below en titled "Changes in Stress" for further information. Tertiary Stress Tertiary stress ([*]) falls on the first syllable of the phonological foot so long as that syllable is a grammatical prefix. It also falls on post stem syllables (other than syllabic nasals) which do not meet the cri teria stated above for the assignment of secondary stress. Quaternary Stress Quaternary stress, which is unmarked, occurs else where. This is actually a degree of stress, however, the rhythmic pattern of the language being one where every syllable has a ‘beat1 of at least some intensity. This stress pattern produces phonological feet which often have a light but evenly stressed onset, heavily stressed peak, and unevenly stressed coda. 215 Much like the tones, the intensity of these stresses varies relative to the location of the syllable in higher level phonological units. This is related to the idea (discussed in the next chapter in more detail) that the nucleus and boundary markers of higher phonolo gical units override the markers of lower level units. Other factors seem to affect stress loudness too. Stresses seem to be louder or softer relative to the phonemic make up of the syllable; especially with regard to the phoneme which is in syllable initial position. Thus, there seems to be a general progression from loudest to softest for syllables which begin with: a. Glottalized phonemes or glottal stop. b. Aspirated phonemes. c. Resonants. d. Fricatives. e. Unaspirates (other than glottal stop). Other principles relating to syllable loudness include: a. Syllables which begin with affricates are louder than those that begin with stops. b. High tone syllables are louder than low tone syl lables. c. Syllables with final glottal stop are louder than those with other consonants or no consonant. Again, I must note that these are tendencies based upon 216 comparative hearing. Only some kind of acoustic test could demonstrate these impressions conclusively. Listed below are examples especially selected to call attention to the various environments mentioned in these rules of thumb. For primary stress which always falls on the stem syllable the following environments are relevant: 1. Stems which comprise single syllable words. Ex amples: [ ,k*1o‘?=3 'fire,' [ ' ts^t] ' stone, ' ['lif9=l 'horse,1 ['?&!] 'that,' [ 's£l] 'I,' ['plf] 'he,' ['ts^ih] 'stick,' 1 shirt,' ['sLl] 'with me,' ['t^du] 'water,1 ['paa] 'to him.' 2. Stems with prefixes. Examples: [^sl1ts^t'my stone,1 [^sL'ts^ih] 'my stick,' [*sLil££9=] 'my horse,1 [*n'zo] 'it is good,' [*sL'z%Q 'my mouth,1 [^t^aadaadL'kLs] 'they are bathing,1 [naagUs'tii] 'I tell it,' [*to?dd'nii"n&%0 'a long time ago,' [*tdd'tii] 'it's a fact,' [*?LlkS^'k%Es] 'he cuts it in two,' [*si?££] 'my shirt,' [*hai'yuh] 'where.' 3. Stems with suffixes or enclitics. Examples: ['t£a*gi] 'three,1 ['ldd*go] 'many,' ['t^@*zi] 'chicken, ' [ ' t_zaa"gGh] 'antelope,' [ 'k?G"d£h] 'now,' ['kho"?£] 'that which is a fire,' 21? [n1t£t*9inh£f] 'that particular man,1 ['?&£ n"9£f] 1 that particular one,1 [*9ls'tsh@ n"hn] 'that par ticular woman,' [*pi'ya"g2*yu*ld] ‘below it at that place,' [*naa'g55*go] 'he is moving.' For secondary stress examples in these environments are most relevant: 1. Compounds. Examples: [“ts^tt'k^o"9!]] 'that which is a match,' (compare with 'stone' and 'fire' above), [*sL"z9'paa"n&] 'my lips,' (compare with 'my mouth' above), ["t^azi-*k93*ni3 'turkey' (com pare with 'chicken above), ["ts^fne'gjo]'wagon' (compare with 'stick' and 'he is moving' above), ["9iina'soot=] 'missionary' (compare with 'shirt1 and 'it drags' above), ["t^dulL'paa"9!!] 'corn beer' (compare with 'water' and 'brown' above), ["stln'zo] 'she loves me1 (compare 'with me' and 'it is good'), ["paanaagds'tii9=] 'he tells it to him' (compare 'to him' and 'he tells it' above). 2. Post stem syllables which end in a glottal stop. Examples: [*n'lii"ni9=l 'he was,' [*ts9L'tf£"ni9=] ' they said,' [*"*pt' tl9aah"zi9=] 'beneath it at that place.' 3. Post stem syllables which are higher in tone than the preceding syllable. Examples: E*to955'nii"n59=3 'a long time ago,' 218 [*naagonl,k^1aad"nS'?=3 'he was making war,1 [*n'zo"nfl] 'it's all right,' [['k^1o"viJ ‘that which is a fire,' [*9Ls'ts^9 n"hn] 'that particu lar lady,' ['?dln"9fl] 'that particular one.' Post stem syllables which succeed syllables which end in a consonant. Examples: [*ton'tE&z"dah] 'he is not tall,' [*togo't££l"dah] 'it is not wide,' [*ko't%El"koh] 'it is getting wide.' Examples of tertiary and quaternary stress would seem to be self evident from the material already cited in this and the preceding chapter. Loan Words and Stress I mentioned earlier that the grammatical stem was important in assigning stress to words in the language, but that there was one class of words, loan words, where it was not possible to predict stress on this basis. For Spanish loans we must use another set of rules. In these words primary stress falls on the high tone syl lable; or if there is no high tone upon the first syllable in the word. Once we identify this syllable, stress on the remaining syllables seems to follow the statements made for the rest of the language. The English loans at first seem to be much more irregular, but upon inspection it is seen that they usually seem to be composed of a stem plus an optional 219 relative suffix. If it is assumed that the syllable to which the relative can be attached is the stem one has no trouble in assigning stress just as in native words. See the examples in Chapter 3* Changes in Stress The only changes I have observed in stress in any given morpheme seem to center around its syntactic role in compounds. Compare these three: O ’fs'tlai] 'five' [*?6s"tla?'tm] 'fifty' [*9&s"tlaa*dn'la9=] 'fifty-one' In the first word primary stress falls on the stem for 'five,' /’esLa9/. In the second a new compound is formed and the decimal marker /-din/ (probably an enclitic) is stressed. Finally, still another compound is formed, and the decimal marker is not only reduced in stress, but con tracted as well, with primary stress going to the new grammatical nucleus, /la’/ 'one.' A similar change seems to happen in the word ["lf'ni9=] 'customary past time.' It is a compound formed of two enclitics. At times infor mants give it in a contracted form apparently related to the syntactic importance of its constituent morphemes. When in this contracted form the second syllable loses stress and the word becomes ['lfn?=]. The change in stress seems to be syntactically motivated, however, 220 something which could be predicted given the proper gram matical information. Other Apachean Languages and Stress Of all the problems encountered in the Apachean languages, the problem of stress is perhaps the one in which there is the most disagreement in the literature. Hoijer, in several publications explicitly states that Apachean words consist of "evenly stressed syllables.11 For example, 11 The Navaho utterance, as it appears in our material, is made up of one or more evenly stressed syl lables" (Sapir and Hoijer 196?: 3, emphasis added). And again, referring to the entire sub-stock, he says, "An utterance in the Apachean languages, phonetically con sidered, is a succession of evenly stressed syllables each of which possesses a clearly defined pitch accent or tone" (Hoijer 19^3: 38, emphasis added). Speaking about Kiowa- Apache, Bittle (1963: 76) generally supports Hoijer1s view. He says, "Taken as a phonetic entity, the Kiowa-Apache utterance consists of a series of more or less evenly stressed syllables, each one of which bears a characteris tic tone accent." The reason for the denial by Hoijer and Bittle of stress in these Apachean languages possibly related to what I mentioned about every syllable having some stress, 221 and these stresses being subject to a considerable amount of variation. Also, the fact that stress is not phonemic may explain their statements. ; On the other hand, Durbin (1964: 42) for San Carlos Apache, Landar (1959) and Hung (1959) for Navajo, and Hill (Durbin 1964: 42; Robinson 1970: 138-39) for Western Apache generally have all recognized that phonetically at least these languages are composed of anything but "evenly stressed syllables.11 Beyond this point, however, they generally disagree. Hung (1959) finds stress to be non- phonemic and predictable primarily on the basis of syl lable shape and tone. She ignores grammatical facts. Her rules seem extremely complicated and since I have already noted in describing the distribution of the segmental pho nemes that the stem syllable is the most variable in pho nemic shape, it is not surprising that her predictions generally identify the stem syllable as receiving primary stress. Landar (1959) also agrees that Navajo has stress; but he presents evidence which he feels definitely demon strates that it is phonemic. He feels there are three degrees of stress, and his evidence consists of minimal or analogous pairs. Even a visual inspection of most of Landar1s forms by someone familiar with the morphology indicates that in most cases the primary stress falls on the stem syllable. A re-elicitation by myself from a Navajo speaker of some of Landar1s data confirmed that 222 primary stress did in fact fall on stem syllable, and that the loudness of.secondary stress seemed definitely condi tioned by syllable shape or other factors. Furthermore, in approximately 15 hours of Navajo elicitation with around 300 utterances I found no examples of primary stress which were not on the stem syllable. Durbin (196^), I believe, has presented one of the most accurate treatments of stress in an Apachean language insofar as he recognizes several degrees of stress and notes that the general stress pattern of the language is one of secondary-primary-weak (Durbin 1964: 43). This does not agree exactly with my findings, as I note a much greater tendency for syllables after the primary stressed syllable to be more heavily stressed than syllables before it. Durbin, however, considers these stresses to be pho nemic because of the extremely restricted theoretical approach he takes, not allowing any conditioning environ ments other than phonological ones. He says for example: "Again, it must be pointed out that in morphology, one can say that the heaviest stress falls on what is called the stem morpheme, but there are no phonological criteria by which to identify the stem morpheme"(Durbin 1964: 42). I, of course, agree that there are no phonological criteria by which to identify the stem morpheme. But, as I have already argued, since this morpheme must evenutally be 223 identified in the syntax and lexicon it is pointlessly redundant to also mark it for stress. Length Syllable nuclei vary in actual phonetic length relative to two factors. The first is location in the foot, stem syllables generally being longer than non-stem syllables. The second is the general syllable shape. Open syllables and syllables which end in voiced phones tend to be longer than ones closed with voiceless phones. Prefix syllables which are bound on either side by a voiceless consonant are likely to have nuclei so short that they are barely audible and occasionally may be dropped. Thus I have recorded the word for woman as [s*tshan] as well as [*9&s'tsh^nl, the word for five'as [s’tlai] as well as [*?&s't%ai]the word for 'corn beer1 as [ "t^uul 'pal3 or [11 thtfull 'paa"hf] and the word for 'me, too' as [*sLlitd?=] or [*si?Ll'tu9=]. This extremely short syllable nucleus in such environments may explain why Hoijer (1946c: 62) reports that in Chiricahua Apache there are consonant clusters composed of /sd/ and /sd/ which are found only as syllable and word initials. He gives examples of /sde's-te'l/ 'it has begun to stream* and /sdi-?i'/ 'he begins to carry them.' He also reports consonant clusters composed of /?/ plus /s/, /s/ and /!/ which occur word medially in the environment V_____ ,CV. The examples he cites are: /l6«?s-co/ 'wood rat,* /bi-He’s-ke/ 'his children,' /da’l-k’eh/ 'both of them.' In both these cases what I think he has heard, or perhaps not heard, is the extremely short syllable nucleus which occurs between two voiceless consonants in non-stem syl lables. Here are some examples of sets of both simple and complex syllable nuclei in descending order of phonetic length. Note that syllables which end in /’/ tend to be longer than those which end in other voiceless stops. (The relevant nuclei are underlined.) Simple nuclei: 1. /nzd/ 'it is good' /hoz/ 'tickle' /si’ildSV 'me too' /log/ 'fish' /hos/ 'cactus' 2. /tazi/ 'chicken' /hiisliz/ 'I urinate' /siniV 'my land' /sibid/ 'my stomach' /gosL?is/ 'mud' 225 Complex nuclei - 1. /ligai/ 'it is white1 /haadi’aa/ 'he sings' /ndaaz/ •it is heavy' /bit’a a V 'its leaf' /si’aad/ 'my wife' /si’aas/ 'my friend' 2. /holzee/ 1 it is called /ndeez/ 'it is long' /bilgoyeeV 'he worries' /daagogeed/ 'they dig' /hiisbees/ 'I boil it' If I were to symbolize these extremely slight length differences phonetically I would use [:] for the longest nuclei, ['] for the shorter, and unmarked for the shortest. Thus open syllables would be like this with all length factors symbolized: [*n'z:6:] 1 it is good,* closed that ended in a voiced fricative like this Q ,xwV:z3 1 tickle,1 closed that ended in a voiceless top like this: [ 'lv*k»1 'fish,‘ and closed that ended in a voiceless fricative like this: [ 'xwvs] 'cactus.' This ignores slight differences between syllables which both begin and end in a voiceless consonant, or between syllables which end in a glottal stop compared to those which end in an other voiceless stop. 226 Obviously, to have symbolized all these phonetic facts throughout this presentation would be cumbersome; and again it should be remembered that these observations are not accurate scientific measurements based upon elec tronic measuring devices, but impressions which have been summarized into relative differences. So far as vowel or nucleus length goes these ob servations are probably not just pecularities of VJKA, but are phonetic universals. Certainly it should be expected that the voicing of part of a syllable should affect other parts and that voiced final phonemes should contribute to the length of voicing in the nucleus. Vowel Length in Other Acachean Languages I have already speculated on the origin of some consonant clusters in Chiricahua Apache which are sus piciously similar to WKA syllables composed of extremely short nuclei. In Navajo, Hill (Robinson 1970: 168-69) also re ports that the length of vowel nuclei are related to sur rounding consonants. She, however, seems to limit the difference in length to long vowels only; which she feels are either half again as long as short vowels when before voiceless consonants and double the length of short vowels before voiced consonants and stops. The data she presents 22? have probably been somewhat regularized, for she does not represent final stops as being voiceless and unreleased and therefore assigns equal length to syllables which end in such stops and to syllables which end in voiced frica tives. Young and Morgan (1964: 1) also report different phonetic vowel lengths for Navajo, but say the longest occur in syllables before voiceless stops like /d/, or /?/. They make no mention of voiced fricatives or open syllables. Nasalization I have already discussed the importance of nasal ization under the section on vowels in Chapter 5« Like tone, it is a contrastive feature of the syllable which cannot be predicted and therefore is distinctive. Prosodic Features and the Feature Mode Part of the definition of the syllable as a unit involves a description of those phonological features which apply to that unit as a whole rather than to any one of the string of phonemes which compose it. Since all of these features are phonetic, it is sometimes hard to make a decision as to which level in the hierarchy they may be assigned. In WMA I have defined tone, length, stress and nasalization as being attributes of syllables as a whole 228 rather than to any one of their parts. As such they are part of the feature mode of the syllable. Prosodic features of higher level phonological units may be considered to be either distinctive or re dundant just as features of phonemes may be considered to be distinctive or redundant. For example, I mentioned above that the contrast between the stops in WMA was in terms of force of articulation, yet the unaspirate series had the feature of voicing in at least some positions. Voicing in this case is a redundant feature. It serves to reduplicate the fortis/lenis distinction which is con tained in the phonetic difference between aspiration and non-aspiration. At the syllable level we may consider the features of stress and length to be redundant features be cause they are totally, predictable, but we must consider tone and nasalization to be the distinctive prosodic fea tures of syllables in WKA. They then are part of the feature mode, while length and stress are part of the manifestation mode. The balance of the feature mode of syllables in WMA is made up of the basic syllable types while the balance of the manifestation mode is made up of the specific combinations of phonemes which can manifest these types in the language. This concludes my presenta tion of the feature and manifestation mode of the WMA syllable. 229 Next, and concluding my discussion of the syllable, I will turn my attention to the distribution mode: the relations that syllable shapes have to each other in linear succession. This is the area of phonology which I call phonotactics. Fhonotactics To emphasize the syllable as a unit, I would de fine phonotactics as possible syllable shapes in sequence. Thus, its primary concern is with phoneme sequences across syllable boundaries. But syllable shapes like phonemes are modified by several different kinds of environments. There are first of all purely phonological ones. These are described by general phonotactic rules which hold throughout the entire language and specify certain pos sible and impossible phoneme sequences across all syllable boundaries. Second, syllable shapes of certain morpheme classes change relative to their arrangement next to other morpheme classes. These are usually called morphophonemic changes, but essentially they are described by phonotactic rules which are limited to only certain syllable bound aries. In order for these rules to apply, there must not only be phonolotical specifications, but grammatical or perhaps lexical ones as well. Thus the syllable as a unit in terms of its manifestation mode can be described in much the same way as the phoneme. 230 At the present time I am not totally prepared to describe all of WMA’s grammatically restricted phonotactic rules. Instead, I will try to concentrate on the general problem of phoneme sequences across syllable boundaries within the same foot. The method I will use to describe these sequences is a table (Table 4) listing possible syl lable codas in relation to syllable onsets. Syllables not ending in a coda (i.e., ending in a vowel or syllabic resonant) apparently have few restrictions on them which I will handle separately. In Table 4 I have listed across the top all the permissible fillers of onset slots. Down the side are all permissible fillers of coda slots. Sequences which are permissible as attested in the data are marked with an 'x.1 These are illustrated below with an example of each possible sequence. Sequences which are thought to be possible, but not attested in the data are given a star in the table. Such a star indicates a lack of an example. Sequences which are thought to be impossible are entered with a numeral which corresponds to a phonotactic rule which follows under the heading "Phonotactic Limitations." Finally, sequences which present special problems are left blank. These latter are often cases which have some gram matical specification which is not completely clear to me. I will comment on them specifically, however, in the fol lowing discussion under the heading "Problem Areas." Table 4. Possible consonant sequences. Syllable Final mb b E d t A z s A c c? z s L X. 1.9 1 3: K k k? ’ & h z m n z 5 X 5 X X X -X- 3 1 1 1 1 1 * XX * X * •X- X X * * * X 5 5 X 5 X XXX X X X 3 1 1 1 1 1 ** X * X X X ' X ^ X %* Z 5 % 5 X * % 1 1 1 1 1 * K •X- 3 •Jr & % * * X •X* ^ % * X * s 5 X 5 X XX 1 1 1 1 1 XX XX 3 XX X X X X X X X X X X 3= 5 X 5 XXX X X X X X * X X * X XX X 3 X X X X X X X * •*• 1 5 * 5 X * X # X X X * XX * X #* * 1 * XX * * X * X * X d 5 5 3 4 X X g 5 5 2 3 2 h 5 * 5 # to X * * * * X * XX ** * X X X •X- * * 3 ** X 9 5 5 X * * * X * 4^* X * X X * * * * # % 3 X XX * X ro VJ H 232 Examples of /z/ clusters: zb /naazbaah/ 'he raids,' zd /yiidaadezdil/ 'they fired it,' zt /naahaztaah/ 1 they are sitting,' zt? /godezb^iiy6/ 'meaning unknovm,' z.i 'no example,' z^ 'no example,' z^9 'no example, ' zs. 'no example,' zL 'no example,' zL 'no example,' zL%, /9izL9ih/ 'dam,' zl /gozl££9/ 'birthday,' zl 'no example' zg /ndeezgo/ 'it's getting longer,' zk9 'no example,' zk 'no example,' z% /nil^i9bilgoz9da/ 'forest,' zg /naadezgiih/ 'he rides horseback,' zh 'no example,' z^r 'no example, zm 'no example,' zn /goneznan/ 'ten.' Examples of /s/ clusters: sb /haadasbih/ 'it's fall,' sd /dok9enaank9eesdah/ 'he didn't fall down,' /gosdSk/ 'hot,' st /naadestinyS/ 'he crosses the road,' st9 /9ist9od/ 'arrow head, ' /nest9an/ 'fruit, ' s.i /haadis.jiis/ 'I pull it,' /naadesgfaah/ 'he is going back,' s^9 /Lo9is^9oose/ 'mouse,' sz /honszee/ 'I am called,' sL 'no example,' sL 'no example,' sL9 'no example' si 'no example,' si. 'no example,' s£ /taanaasgis/ 'I wash it,' sk /naadeskai/ 'they were going,' sk9 /gosk9az/ 'cold,' s^ /daas9aah/ 'when,' sg 'no example,' sh /.jishedi/ 'I kill it,' sm 'no example,' sn 'no example.' Examples of /z/ clusters: zb 'no example,' zd /donaalLizdah/ 'he is not irrigating,' jzt 'no example,' ■ zt9 'no example,' zj 'no example,' 5c 'no example,' zc9 'no example,' 5s 'no example,' 5L 'no example,' zL 'no 233 example, 1 zL? ‘no example, ' zl. ‘no example, ‘ zl ‘no example,‘ zg /naanlLSgo/ 'he is irrigating,' zk 'no example,' zk? 'no example,' z% ,no example,1 z£ 'no example,1 zh 'no example,' zy /^ilgozyo/ 'Whiteriver,' zm 'no example,' zn 'no example.' Examples of /s/ clusters: sb /hiisbees/ 'I bait it,' jsd /?aasdii/ 'I speak,' st /naastii/ 'I give it to you,1 st? /naast?iizih/ 'Zuni,' sj /gasjis/ 'a kind of bird,' /?aagoscise/ 'near,' sc? /?aasc?ise/ 'short,' sz /bihaszee?/ 'I am fishing,' sL /?esLai/ 'five,' sL /naa?aansLiz/ 'I irrigate,' sL? /gosL?is/ 'mud,' si /?iislah/ 'I cook it,' £1 /?aago§le?/ 'I'll do it,' sg /benaagosgod/ 'hoe,' sk /naa?iskad/ 'I sew,' sk? /hiisk?a3=/ 'I am coming, ' _|2 /haadis?a&/ 'I sing,' /sisgdahi/ 'robin,' sh /hiiSheel/ 'I carry it in a pack,' sm 'no example,1 sn /hiisne?/ 'I carry them.' Examples of /!/ clusters: lb /?ilbees/ 'he boils,1 Id /si?ild6?/ 'me too,' It /yaltii/ 'he talks,' It? /nlt?ee/ 'it's good, ' l.j /nadaahil.iig/ 1 they are afraid of it,' /y6g6nl/ih/ 'he knows it,' 1/? /dodaahil/?adah/ 'it's not noisy,' Iz /nlzig/ 'rotten,' Is /naagolsee/ 'I play,' 'no example,' 1c /nlcoh/ 'it's bad,' 1c? /dogonolc?5ddah/ 'it's not noisy,' 15 'no example.' Is /ganilsood/ 'you split it,' 1L 'no example,' 1L /naanlLiz/ 'he irrigates.' 1£? /naa?ilL?iigl/ 'gopher,' 11 /?illaahS/ 'that which is cooked,' Ig /?ilgenkees/ 'to cut in two,' 234 Ik /naagonikaa9/ 'he is going to war,' Ik? /ko?baanlk?as/ 'November,' 12. /daanl?aa/ 'they sing,' Ig / ^ilga?d/ 'East Fork people,' Ih /?aalhoz/ 'it tickles,' 1^ /lehnaadilyd/ 'cemetary,' 1m 'no example,' In 'no example.' Examples of /!/ clusters: lb 'no example,' Id /’ehooldii/ 'jealous,' _lt 'no example,' It? /bilhiilt?e?/ 'to go by car,' Ij 'no example,' Id /naal^oos/ 'paper,' lg(? /naagodol^?oosn/ 'slender peak standing up people,' Iz /nolzee/ 'he is called,' JLs 'no example, ' 1c /nilcl? bilgoz?aa/ 'forest,' 1c? 'no example,' Iz /naanolzeege/ 'saguaro cactus,' 1£ 'no example,' 1L 'no example,' 1L 'no example,1 1L? /naa?ilL?oh/ 'he packs it,' 11 'no example,‘ lg /haajilgis/ 'lightning,' Ik /9ilkad/ 'sugar1 Ik? 'no example,' 12 'no example,' Ig /ceelgoz/ 'White- river, ' Ih 'no example,' 1% /dozoolyda? ?adgolnidah/ 'it's not fit to eat,' 1m 'no example,' In /benilyilni?/ 'you pushed him.' Examples of /d/ clusters: dy /nzaadyd/ 'here this place,' dn /si?aadn£/ 'the one who is my wife.' Examples of /h/ clusters: hb 'no examples,' hd 'no examples, ht 'no examples,' ht? /daht?Sh/ 'shade,1 hj 'no example, ' h^ 'no example, ' hgf? 'no example, ' hz 'no example,' hs 'no example,' hj 'no example,' he /ndoheoh/ 'mountain lion,' he? 'no example,' hz /biL?5ahzi?/ 'under him to there,' hs /daahsistiin/ 'I put it down,' hL 'no 235 examples,1 hL 'no examples,' hL^ 'no examples,1 hi 'no examples,' hi /t65hlilii/ 'coffee,' hg /yik^ehgo/ 'accord ing to,' hk /leehohkaah/ 'you come together,' hk? 'no example,' h% 'no example,' hg 'no example,.' hy /biL^aahydh/ 'under him that place,' hm 'no example,* hn /lehnaadilyS/ 'cemetary.' Examples of /?/ clusters: 'no example,* 2d /naa2diskoo/ *1 iron it,' 21 'no example, * ?t? 'no example, ' 2J. 'no example,' 2 A lno example,' /mba2^2ooseh/ 'fox,' 2_z 'no examples, * 2§. 'no examples, ' 21 'n0 examples, ' 9c /si?ii?coh/ 'my jacket,' ?c? 'no examples,' 21 /sikee^zoos/ 'my shoes,' 21 /la^s&^nne/ 'give me some,' 2L ,n0 example,1 2%, 'no examples,' 'no examples,' 21. 'no examples,' 73, 'no examples,1 2E 'no examples,' 21 'no examples,' ?k? 'no examples,' 2£ /bike?f=ah/ 'its root,' 21 /ba?heszine/ ' a black bead,' 2% /biyi^yee?/ 'outside it,' 2™ ,n0 examples, 1 2ll /L?e?na?ai/ 'moon. ' Phonotactic Limitations Rule 1, Grooved Fricative Assimilation (GFA) t Where C is any grooved alveopalatal fricative •x or affricate, and v/here \ C is any grooved alveo lar fricative or affri cate. This phonotactic rule specifies that there is assimilation 236 on the part of grooved fricatives in syllable final posi tion to the point of articulation of following grooved fricatives or affricates. Evidence for this rule consists of the following kinds of data. Compare: /Lo’is^ooseh/ 'mouse1 (lit. small mouse) i i i /Louiseoh/ 'pack rat' (lit. big mouse) or /naahiis/e/ 'I smoke it1 /hiisc?id/ 11 rub it1 Actually, the rule is somewhat too strong in the form I have stated it. There are probably some grammatical limi tations involved. The rule may be limited to prefix syl lable final grooved fricatives before other prefix initial grooved consonants for there is nothing in the data to indicate that stem final fricatives in compounds or stem final fricatives prior to enclitics are subject to the rule. Rule 2, /?,/ Replacement (g R). g-- .7 d /--- g ] This rule is not hard to defend. We have already seen that /g/ occurs as a syllable final only in stem syllables. Furthermore, the free variation between /g/ and /d/ in this position has also already been pointed out. What seems to happen is that in stem final position before either a /d/ or /n/ in the next syllable, /g/ systemati cally changes to /d/. It is thus not found as the first 237 consonant in what theoretically would be /g/ plus /d/ or /n/ sequences. Motivation for this rule is found in these forms: 1. /li^6g/ * yellow * /be’ilj^oSdn/ 'made yellow people' 2. /si’aag/ 'my wife' /si9£adn/ 'the one who is my wife 3. /naa’iziig/ 'he is working' /donaa 9iziidah/ 'he is not working' from /do + naa’iziid + dah/ 4. /tSSgii/ 'three' /ta&din/ 'thirty' from /ta&d 4- din/ 5. /gos/9Ig/ 'seven' /gos/9Idin/ 'seventy' from /gos/9id f din/ In 1 and 2 the process is relatively straightforward, /g/ being replaced by /d/ in both cases. In the rest of the cases a second process described in Rule 3 below, known as Identical Segment Deletion, deletes one of the. two /d/ phonemes in sequence, to produce the appropriate surface structures. (A recent article by Cook (1971b) reminded me of these last two examples.) 238 Rule 3, Identical Segment Deletion (ISP) ci ---- > 0 /_____ ,ci This rule specifies that there cannot be consonant clus ters which are composed of identical consonants. Evidence for this rule is partially derived from examples like those cited above under the g R rule. Other evidence is found in the following material. Compare: 1. /naasgod/ ‘I am strong' /donaasgodah/ 1I am not strong' from /do + naasgod + dah/ 2. /haadilzeeh/ 'he spits' /haadiseeh/ 'I spit' from /haadis > zeeh/ 3* /naa’iziig/ 'he works' /naa?iziigo/ 'he is working' from /naa’iziig + go/ 4. /’^naagot’iil/ 'thus he acts' /^anSagot^iileni9/ 'thus he used to act' from /?an&agot?iil + leni9/ In case 1 the /-dah/ is added and one /d/ is deleted. In case 2 the prefix /s-/ 1 first person1 replaces the /!-/ classifier. The stem initial /z/ becomes voiceless (/s/) and ISD goes into effect to produce the final single /s/ in 'I spit.1 In case 3 the enclitic /-go/ ‘on going 239 action1 is added to the stem and the final /g/ of the stem is dropped. In case 4 the enclitics. /leni'V 'customary past time1 are added to a verb ending in /!/. The initial /!/ of the enclitic phrase becomes /!/ and the /!.!/ se quence becomes simply /!/. In slow speech where a foot or meter boundary intervenes this is /’anaagotiil leni?/, [*?anaago'tiil + "le'ni?=]. i i These changes have little phonetic effect. This is because consonants in syllable initial position when preceded by a syllable which ends in a vowel are already phonetically lengthened and could be written [VC^'C^V], when considered phonetically. Thus ISD must be in the description; otherwise I would be implying that there is a phonetic difference between lengthened consonants and double consonants, which is not the case. Rule 4, /!/ Formation (LF). d + 1 ---- > L This rule is motivated by one example in the data. In one of the texts the word /naadSkeed/ 'she begs1 combines with the enclitics /leni?/ 'customary past time' to form /naaddkeeLen9/ 'she was keeping on begging.' This rule seems to be related to a phonotactic rule which occurs at initial stem boundaries, usually called the 'D' effect. More will be said about this later in the chapter. 240 Rule 5t Distribution Restrictions What I have termed Rule 5 on Table 4 is not actu ally a phonotactic rule as Rules 1 through 4 have been, but concerns the fact that some syllable initial consonants can never appear after a syllable final consonant in the same foot because of the limited number of forms they appear in. The consonants /mb/ and /p/ are such cases. Since they only appear in three different noun stems there is no chance for them to be preceded by anything other than a vowel in the same foot. Thus, they are not possible sources of consonant sequences in the language. Vowel-Consonant Sequence Limitations. I stated above that there were no limitations upon the sequence V.C., however, this has one possible exception. I will introduce this as another phonotactic rule and call it Vowel Replacement (VR). VXVX ---- > Vyi/ .y Where VXVX is a homorganic complex nucleus. There are a number of things in the data which indicate that such a limitation exists. All of the long nuclei which occur before /y/ are of the form /Vi/. Thus are found: 1. /haiy6/ ‘where1 2. /sigoiye/ ‘my maternal grandmother' 3. /naakaiye/ 'Mexican' 241 4. /naiyotiil/ 'he carries it' 1 1 5. /haiyolk?aal/ 'he comes here' 6. /biyegeiyo/ 'below there that place' 7. /?ak6iyaa?/ 'below that place over there Added to this , is the fact that some of the morphemes in volved have alternate syllable shapes in other environ ments. Thus, for example, /hai-/ in 1 occurs as /haa-/ in /haa?an/ 1who * and /haat?£l/ 'what;1 or /-gel/ in 6. appears as /-gee/ in /teksasgee/ 'at Texas;' or /?akdl/ in 7 appears as /?ak66/ 'over there,' elsewhere. On the basis of these data the rule seems well motivated. Problem Areas As is evident from Table 4, there are a variety of sequences where there are no data to indicate whether they are possible or not, and there are some sequences which in terms of general patterning seem possible, but which suffer from a conspicuous lack of examples. Both of these I would term problem areas and I will spend the remaining part of the chapter discussing what information is avail able with regard to certain of these. Problems with /h, ?/ Initial Sequences In normal speech syllable final /h/ and /?/ when followed by another consonant in the next syllable tend to be produced so rapidly that they are extremely hard to 242 hear. Often it appears they are deleted and therefore I have few examples in my data. Consider the form /’isLaa’/ i i 'I drink1 with a definite final glottal stop. In the negative form /do?isLaadah/ 'I don't drink,' the stem i i final glottal stop does not seem evident. Those examples I do have seem to have been the re sult of relatively slow repetitions, and they may not be representative, for in a syllable by syllable utterance of a word some informants seem to put in a /?/ or an /h/ after every syllable. I think that the sequences are generally possible, but because of the phonetic nature of /?/ and /h/ are hard to hear or are slurred over. Grooved Fricatives Plus /y/ The problems surrounding the sequences /sy/ and /sy/ are somewhat harder to explain. I have already pre sented evidence that /y/ is deleted or absorbed in some verb stem initial positions^when preceded by /s/ or /s/. Whether this holds for all stem initial /y/ phonemes, and for enclitic initial ones I cannot say. There is one hint elsewhere in the data which seems to argue against such sequences, and it relates to the voicing of the final grooved fricative before a /y/ initial enclitic. The word for dirt is /lees/. This word is apparently derived from a verb stem which means 'it is lying* for we find the compound /?ilsiileezy6/ 'salt, it is lying that place' 24] which is the place name of a spot in Arivaipa canyon. In the compound the final /s/ of /lees/ becomes /z/. Whether this morphophonemic shift is related to the fact that if it had not happened the /y/ would have disappeared, or whether it is related to the fact that /lees/ in its ver bal form requires a /z/ because of some paradigmatic dif ference I cannot say. What I can say is that I have no examples of voiceless grooved fricatives followed by /y/ in the next syllable and in places where one might expect such a sequence the final fricative is voiced. Taking these facts into consideration I would say that the evi dence is against such a sequence, but not so strongly as to justify the postulation of a phonotactic rule at this time. /d, g/ initial clusters and the *0* Effect The most complex of all these problem areas is that of /d/ and /g/ plus other consonants. I have already pointed out that /d/ and /g/ alternate in stem final posi tion in many if not all words in the language. Further more, it is unquestionable that /d/ and /g/ do not occur finally in prefix syllables, at least so far as the sur face structure of the language is concerned. On the other hand, as mentioned earlier, there is reason to postulate a syllable final /d/ in the canonical form of many prefix morphemes on the basis of certain morphophonemic changes 244 which take place in stem initial position. Again, as noted earlier these changes have been termed the 'D1 effect. 'D1 effect changes in stem initial consonants are only relevant here in that they suggest some possibilities with regard to stem final /d/ and what happens when stems which end in /d/ are followed in the same foot by en clitics which begin in a consonant. For this reason, I will briefly summarize the literature which reports on the ‘D1 effect and then give additional data from YIMA which outlines some of its 'D1 effect changes. The ,Dt Effect There are two sources of stem initial consonant changes known as the 'D1 effect. The first of these is when a verb is of the 'd1 class and is thus prefixed with a /d-/ classifier just prior to the stem. This classifier is supposedly one of four which occur in the Athapaskan languages, the other three being /!-/, /!-/, and /0-/. Stanley (1969a:1?) argues that there are only two classi fiers, /!-/ and /d-/ which together form /!-/ or are absent or deleted when /0/ occurs. These classifiers seem to in some way be related to the voice of the verb; serv ing to change active verbs into the passive, causing in transitive verbs to become transitive, or causing transi tive verbs to become intransitive (Young and Morgan 1964: 45-46). 245 The second source of the 'D1 effect is the first person duoplural subject morpheme whose canonical shape for WHA is /iid-/ (Hoijer 1945c; Edgerton 1963: 142). Generally for the Apachean languages when either of these sources of a /d/ phoneme occurs before a stem initial consonant, changes take place depending upon the consonant. These changes are summarized in Table 5 which is adapted from Hoijer (1945c: 200). Unfortunately, I do not have complete data on what this prefix final /d/ does in VJMA, but judging from the examples found in the data the WKA 'D' effect does not vary significantly from its Apachean neighbors. Compare the following examples: Examples 1. /haadaadi?aah/ 'they sing' • - /haadaant9 aah/ 'we two sing' d ^ 9— — 9-19 2. /daa^iiyaV 'they eat' 1 /haa^iida9/ 'we two eat' d 1- y— — -> d /yedaayll/ 'they push' 3. 1 /bedaanjil/ 'we two push' d 4- y— -- > j 4. /daahilziih/ 'they stand up' 1 1 /daa^iijiih/ 'we two stand up' d 4. z— -- > j 5. /haadaadi1z e eh/ 'they spit' ! /haadaanjeeh/ 'we two spit' d + z— 1 6. /daa/aah/ 'they die' /daa?ii/aah/ 'we two die' d 4- Cj---- -> C 246 Table 5* 'D* effect stem initial changes in seven Apachean languages. Chiricahua Stem Kescalero Initial San Jicarilla Kiowa- 'D1 Consonant Navajo Carlos Linan Anache d + ’ ___ ^ t? t? t? t’ 9 d 4 m —— m m hm hm 9 9 & * ri --- x. n n hn hn d + y --- d or j d or j d or j d + g --- g d g g d + z -— ^ j j j j d 4 z -— > 0 J J J or d d 4 1 -— ^ L LL L d 4 Ci--- ^ Ci ci Ci Ci Where C^ is not one of the other consonants listed. X 24? 7. /’aadaadi^’ig/ 'they rub1 /?aadaangf?ig/ 'we two rub' d + Cj-.— Ci 8. /’aadaadiLees/ 'they polish' /’aadanLees/ 'we two polish' d -r Ci- — ->Ci 9. /daadildoh/ 'they shoot' O 1 1 •H /daanldoh/ 'we two shoot' d + Ci- 1 10. /daa’iikal/ 'they hit it' /daa’iikal/ 'we two hit it1 d r- C —*—— Ci Obviously the pattern is similar to the others as cited by Hoijer. Since there is no doubt that WKA's closest rela tive is San Carlos one might assume that WKA probably follows San Carlos generally in terms of the 'D* effect. /d/ and /%/ Initial Sequences: A Proposed Solution It seems likely that except for /dy/ and /dn/, that /g/ and /d/ initial sequences are impossible in this language even at the boundary between stems and enclitics. The evidence in favor of such a solution is as follows: 1. The 'D1 effect suggests that there is a tendency to prohibit such sequences at another point in the phono- tactics. Furthermore, the LF rule demonstrates that one •D' effect type change occurs at stem final boundaries. 2. The absence in the data of any cases of /g/ fol lowed by a consonant in the next syllable plus the g R rule seem to indicate that /g/ cannot occur initially in 248 sequences. If this is true the g R rule should be re stated as: g d /_ Where is not /g/. If this assumption is true, the only thing which remains is to make explicit how such sequences are avoided when stems which end in /g/ and /d/ are suffixed in the same foot by enclitics which begin in a consonant. Since, to my knowledge, enclitics only begin in the consonants /s, z, 1, 1, y, d, g, ?, h/ all that needs to be accounted for are sequences of /d/ or /g/ plus these. The ISO rule takes care of /d.d/ and /g.g/. The restatement of g R takes care of all /g/ initial combinations. This leaves only /d/ initial combinations to be dealt with. The LP rule takes care of /d.l/, and by analogy probably the other 'D1 effect rules can be extended to account for /d.s/, /d.z/ and /d.l/. Since /d.y/ can occur it does not concern us here. This leaves only /d.?/ and /d.h/. It is possible that these two follow the 'D1 effect pat tern. Inadequacies in the Data This last section on phonotactic possibilities has had to remain in the realm of hypothesis because of holes in my data. These holes are there because of two factors: first, I did not become aware of these problems until 249 after I had left the field. I therefore did not systema tically collect data to shed light on them. Second, as far as final /d/ and /g/ are concerned, there seems to be a definite tendency that when stems end in a stop, any enclitics which may follow do not follow in the same foot. Thus, stem final stops are often unreleased and followed by clear pause, the mark of a foot boundary, while follow ing enclitics are given primary stress indicating they are a separate foot. Relevant examples are thus hard to find. Summary In this chapter I have presented data and rules which define the features, manifestations, and distribu tion of the WMA syllable. It was argued that the syllable in WMA is distributed in the phonological foot as either a stem syllable (nucleus), or as a non-stem syllable (onset or coda). That in each of these locations it had certain structural and prosodic features which serve to character ize it. In the latter part of the chapter, data was pre sented concerning the various possible manifestations of the syllable with regard to other syllabic environments within the same phonological foot. A set of phonotactic rules was postulated to account for regularities observed in syllable shapes in sequence, rules which would account for the various manifestations of observed syllables. I 250 tried to emphasize at this point the essential uniformity and relationship between phonotactics and morphophonemics. It was noted, however, that the data were inadequate in places, and that certain syllable shapes could not be totally accounted for at this time. Consequently, cases which were inadequately documented were discussed but only suggestions were made as to their possible solution. \ CHAPTER 7 HIGHER PHONOLOGICAL LEVELS: THE FOOT, METER, LINE, AND CONTOUR In this chapter I will consider the features, mani festations and distributional possibilities of the other higher phonological units of WHA. As mentioned in my introduction, I find it neces sary to posit at least four levels between the syllable and the discourse type in order to account for the data. (1) The foot, a unit of syllables with a primary stressed syllable as its nucleus optionally surrounded by a variety of phonetic markers at its boundaries. (2) The meter, a unit of feet, with an extra loud primary stress on the nucleus of the foot at its peak and pause or hesitation at its boundaries. (3) The line, a unit of meters with a final or non-final intonation on its last meter and longer pause at its boundaries. (4) The contour, two or three lines which have long pause and usually breath at start and finish, and whose last line has final intonation. The Foot Level The phonological foot in most languages is a unit of stress and timing which combines syllables into what 251 252 are sometimes called phonological words. They may or may not correspond closely to grammatical words probably de pending upon the syntactic nature of any given language. In WMA the fit is fairly close between feet and grammatical words. The exception to this one to one cor respondence involves certain kinds of postpositions (those with a - W structure) and enclitics which, although they can stand as separate words, are nevertheless often in corporated into one phonological foot with a noun or verb they syntactically modify. Foot Structure The WMA phonological foot is composed of an op tional onset made up of syllables with secondary, tertiary, or quaternary stress; a nucleus composed of a grammatical stem and its accompanying primary stress; and an optional coda made up of syllables with secondary or tertiary stress. The assignment of stress to the syllables in the foot is dependent for the most part on the grammatical nature of the morpheme or morphemes in the syllable, a fact which was discussed in the last chapter. Since the structure of phonological feet in terms of number of syllables, the specific arrangement of second ary, tertiary, and quaternary stresses in the onset, and the number of coda syllables is dependent upon syntactic considerations; I cannot be as explicit about foot 253 structure as I was about syllable structure. This being the case I will merely outline and give examples of some of the possibilities. Such an outline would include pho nological feet with polysyllabic onsets, polysyllabic • . codas, monosyllabic onsets, monosyllabic codas, com binations of these latter possibilities, or feet with no onsets or codas. These latter would, of course, be simple stems which form individual phonological feet. In my earlier discussion of stress possibilities I presented a variety of foot types which I believe ade quately illustrate this point. Foot Boundaries For the most part foot boundaries are marked by phonetic cues which indicate that the end of one phonolo gical unit has been reached, and that another is beginning. These entail: (1) allophonic variations which do not occur in other phonological environments; (2) the presence of particular allophonic variants in environments which usually condition other allophones; (3) the occurrence of impossible phonotactic sequences. Note that here I am not arguing that these phonetic cues are conditioned by the presence of a boundary unit, rather I am saying that the speaker is aware of the higher units of the language, and when the end of one unit is reached allophonic and phono- tactic rules which ordinarily apply within phonological 254 feet are blocked or better irrelevant. Thus, it is the higher unit which is part of the language, and the bound aries themselves are not units (like juncture phonemes) only spaces between units. Markers Involving Unique Allophones The first kind of phonetic marker concerns itself primarily with the unaspirate stops. In foot final posi tion they are unreleased, a phonetic fact which is simply a brief hesitation at the point of articulation of the foot final stop. Examples of such foot final stops may be found under the sections describing unreleased /d/, /g/, and /?/. Similarly, unaspirate stops in foot initial position, even though preceded by voiced phones in the preceding foot, are voiceless. For example: [/*sL*kW6k= + *n'tii + *to'l&21##] 'my knee with hurt,' [/'p££ + *taadUl'toh#] 'they shoot, ' [/*todi»'ts’llz + 'jtah#] 'it's not rough.' In all three of these cases the final foot begins with a voiceless stop even though preceded by a voiced consonant or vowel in the preceding foot. A second phenomenon which was not treated as allo- phonic but is a peculiarity of some phonemes at foot boundaries is phonetic lengthening. Syllable final frica tives before foot boundaries seem to be slightly longer than usual, thus in 'it's not rough' above the final /£/ is held a little longer, and the final /z/ in [/*n'ttzz + 255 'it will be longer' is also held a little longer. This same phonetic effect is encountered at the pauses which mark meter, line, and contour boundaries, except before such pauses the voiced fricatives tend to drop off into voicelessness before articulation is stopped. Alloohones in Inappropriate Places The second major marker of a foot boundary is the presence of an allophone which if no foot boundary were present would not occur in that environment. For instance, /n/ has the allophone [m] before /b/ in the same phonolo gical foot, but not over foot boundaries. In locations where a foot boundary intervenes between /n/ and /b/ in sequence, /n/ will have its post vocalic form [n], instead of the expected [m] before /b/. Thus, the word 'pumpkin seed' is a compound with only one phonological foot and is phonetically [/*'pS.9l3=11 k*1 smbl '15Sh/] while the phrase 1 old woman's cane' (the cane used in the girl's puberty rite) is two feet in length and is phonetically [/'sdan t *pL'ku"s&h/]. Another example of this kind of boundary phenome non is found in the allophones of vowels in simple nuclei. At foot boundaries vowels in simple nuclei are close even though followed by a consonant in the first syllable of the next foot. This was discussed and exemplified in Chapter 5 under the section on vowel allophones. As I 256 shall point out below, this is because the foot prosody of timing does not apply over foot boundaries; and thus does not unite such syllables in the cohesive relation found within the foot. Impossible Consonant Sequences Finally, all of the impossible consonant sequences indicated in the last chapter occur with intervening foot boundaries, including identical consonants in succession. Ambiguous or Unmarked Boundaries There is one remaining problem about identifying the extent of the phonological foot, and this is the possi bility that on either side of a foot boundary there are no unusual phonological markers to indicate its presence. This is essentially the same kind of problem encountered in deciding where the boundaries of syllables or phonemes are, and the same method is used; the comparison of easily defined cases with ambiguous ones. Since the nucleus of the foot is always a stem morpheme and receives primary stress, ambiguous cases can be resolved by looking for the most logical syntactic break between two foot nuclei. This, plus differences in the foot prosodies of length, timing, and rhythm which occur in the onsets and codas of feet, will easily isolate the boundary even though it is not characterized by any of the phonetic features noted in the preceding section. 257 Foot Prosodies The phonological foot in WKA has four prosodic features: (1) rhythm, which includes the sequence and intensity of stresses; (2) tonality, the relative heights of tones in sequence; (3) timing, the rapidity of stresses in sequence, the amount of time allotted to each syllable in the foot, and the cohesiveness of the syllables; and (4) consonant harmony, the agreement of all grooved ob struents in any one word within a foot in terms of point of articulation. Thus, prosodies at foot level include stress pat terns, not just stresses; tone patterns, not just tone per se; and timing of a series of syllables, not just length of a single syllable. Obviously, however, I have had to deal with the basic source of these prosodic pat terns already in the discussion of the syllable, so the main task here will be to simply sort out what can be assigned to the syllable level, and what is characteristic of the sequence of syllables called the foot. Rhythm In describing the syllable I discussed what stresses would be assigned to what kinds of syllables. A series of syllables then arranged in sequence with their respective stresses provides the foot with its rhythm prosody. Generally, as noted earlier, the phonological 258 foot is characterized by a rhythmic pattern of weak onset stresses, primary nucleus stress, and moderate coda stresses. There are, of course, as many specific variants in this pattern as there are foot structure possibilities compound words having heavier onset stresses, for example. Like the section on foot structure rhythmic patterns have already been exemplified under the section on stress in the preceding chapter. Tonality For the most part this prosody has also been dis cussed under the section in the last chapter on tone. The main feature of the tonal prosody at the foot level is the tendency for feet to have a marked stair step-like tonal sequence caused by the phonetic intermediacy of most onset syllables as compared to stem and coda syllables. Using the number tone notation introduced in the last chapter, here are some examples of this step-like quality: [+ *thaa5dL2'kigl 'he bathes,1 5 5 2 1; or [+*to2?a5a5'nii2"n3?=4 +.] ‘a long time ago,1 2 5 5 2 4. In two-syllable citation forms there often are 2 1 , 2 3, 3 4, and 3 6 sequences since many prefix syllables are low tone. In fact, it is these kinds of two syllable forms that make the intermediate nature of prefix tone most noticable. For example: [+*si2'y&9ml +] 'my son,' (j-*s<-2'maa33+3 'my mother,1 [+*si3'tha9=^4.] 'my forehead, i 259 * eight.* This slight difference then be tween onset and stem/coda tones is the tonality prosody of the WKA foot. Timing The foot is a sequence of cohesive close-knit syl lables, and the primary prosody which produces this co hesiveness is what I call timing. Timing can be observed phonetically in two ways. First, syllable nuclei in the onset part of the foot are somewhat shorter than the one in the peak, and those in the coda while also shorter than the peak syllable are somewhat longer than those in the onset syllables. The peak or nuclear syllable is longest of all. This gives the foot in its maximum form a series of rapidly articulated onset syllables, a slow precisely articulated peak, and a moderately slow coda. It is because of this difference in timing that onset vowels even when in comparable syllable forms are shorter than peak vowels of the same quality. Thus, for example, it is likely that short onset vowels in syllables beginning and ending in voiceless consonants may not always be identifiable as to vowel quality, while a foot peak or nucleus vowel in the identical environment is short, but always identifiable (this was discussed under "Length" in the last chapter). This timing factor does not negate what was stated in the last chapter about vowel 260 length relative to syllable shape, it merely provides a different basis for defining phonetically long and short syllable nuclei in stem and non-stem syllables within the phonological foot. The second phonetic result of the timing prosody has to do with consonants which begin syllables preceded by open syllables. Within the phonological foot, such consonants are phonetically lengthened, so that the point of phonetic syllable division is within them rather than before them. I have already discussed where the phonemic point of syllable division is. Here I simply wish to discuss the phonetic facts involved with this consonant lengthening. This phonetic lengthening is most prominently ob served in stem initials, but depending upon general speed of articulation can also be observed to a greater or lesser extent in non-stem syllable initial consonants as well. The fact that non-stem unaspirate stops and affricates are voiced in non-stem syllables, probably obscures this lengthening and also may be a result of the more rapid articulation of onset and coda syllables. As with stress, the particular initial phoneme involved also plays a role in its prominence; thus syllables which begin in an affri cate are most easily observed as having a lengthened ini tial consonant with stops, resonants, and fricatives following, in approximately.that order. 261 For example in a word like /si^ag/ 'my rib,1 pro duced usually as one foot, the stem initial /$,/ is heard as [t.ts*1] with the syllable break coming in the middle of the affricate. Similarly in /libah/ 'brown1 the stem initial /b/ is held long enough that you hear the begin ning of the stop as clearly distinct from the release of the stop. It could be written as [p.p] and be phonetically correct with the syllable break actually occurring in the middle of the stop. However,.since the phone is really not rearticulated, but merely held for a period of time, perhaps the best way to represent this would be with a length mark [:]. Thus, 1brown' is [*lL'p:ah] and 'rib' is [*sL'tsh:ok=] with other examples being represented I accordingly. I did not present the data in earlier chap ters with this phonetic lengthening symbolized. This was in order to simplify typing and to indicate that such pho netic lengthening is a prosody of the foot level, not an attribute of an individual phoneme. It is caused in my judgment because of the timing prosody, not because of phonemic environment. It is a feature of the foot, not the phoneme. This timing factor does, however, also affect the vowel phoneme in the preceding open syllable nucleus. It is the factor which conditions the presence of [(.], [o], and [v] in simple open syllables followed by a consonant 262 in the following syllable. In other words, it causes these syllables to be phonetically closed even though they are not phonemically closed. The only time this phonetic lengthening of con sonants is not found is when the utterance is produced so slowly so as to make syllables into separate phonological feet. I have already commented on this above in the sec tion on vowels. The Timing Prosody in Other Apachean Languages Phonetic lengthening of syllable initial consonants after open syllables has been noted by Hoijer (1945a: 8 ) for Navajo and Chiricahua (Hoijer 1946c: 57); by Durbin (1964: 44) for San Carlos; and by Bittle (1963: 77) for Kiowa-Apache. These men generally report what I have described above except that Hoijer and Bittle limit this phonetic lengthening to consonants following short voweled (simple nucleus) open syllables while Durbin specifies that it applies to consonants following all open syllables. To repeat my observation, in V/MA consonants are phoneti cally lengthened after all open syllables including syllabic nasals, but stem syllable initial consonants display this prosodic fact much more clearly and consis tently than do non-stem syllable initial consonants, which are articulated much more rapidly, thus obscuring their phonetic lengthening. 263 How accurate these claims are for Chiricahua, San Carlos and Kiowa-Apache I cannot state, but the difference between Navajo and WMA in terms of the occurrence of this prosodic feature seems to be a real difference. I base this last observation upon two facts which have come out of my research. The first is that in eliciting Navajo from a Navajo speaker I observed essentially what Hoijer re ported. Consonants were lengthened after short open syl lables, and not just primarily in stem initial position but throughout the entire foot. Navajo pre- and post-stem syllables are articulated much more precisely than in WMA. Throughout that language phonetic length is much more noticable, as is the general voiceless quality of the stops wherever they are found. Timing in Navajo seems to be slower. The second piece of evidence for this real differ ence is found in informant opinion that WMA is spoken much more quickly than Navajo. Navajo's often make this state ment, and several Navajos living on the Apache reservation voiced this opinion to me during the course of my research. After hearing their language and comparing it to WMA, I would say they are correct in their diagnosis, and would place the locus of the phonetic difference in the two lan guages to differences in the timing prosody of the foot. These observations perhaps underscore the importance of 264 defining higher level units (in this case the foot) in terms not only of their structure, but prosodic features as well. Grooved Obstruent Harmony By grooved obstruent harmony, I am referring to the fact that grooved fricatives and affricates in the on set of the foot assimilate to the point of articulation of any grooved fricatives or affricates in the nucleus of the foot. For illustrative examples of this prosody compare the following sets and note the assimilation of the prefix syllable initial segments: 1. [+Sl'z&?=+] /.size7/ •my mouth‘ [+sl'ps?=+] /sibeV •my breast * 2. [tsL'k^ust] / s i W o s / •my neck' [4sL'kUt»+] /sigod/ •my knee' 3. [4*t£stiti ‘ ^aas-t] /des667aas/ •you two are going* [+^tSsuu'k^ai41 /des66kai/ 'you all are going' 4. [V>ns ,t££z4] /nsdeez/ •I am tali' [4*ns'taa+] /nsdaa/ •I am sitting' Evidence for this prosody being a property of the foot is found in the fact that in words such as these, when said very slowly, the prosody is absent. This is because in lento speech utterances which are ordinarily one foot in length are broken up into several feet. 265 Again, as in the case of the close and open alio- phones of the vowels, the speed of utterance and its de termination of the number of phonological feet is reflected in the prosodic and allophonic facts of the language. A limitation on this prosody is found in stems which form the first part of a compound word and come before stems which contain grooved segments in the same phonological foot. These do not assimilate in this manner. For exam- pie: /sa^nl^pos/ 'give me the paper1 from /sa+ ’nl^oos/ 'to me 4 to handle a flat object' /^eelgoz/ 'Whiteriver' from /^eel 4 goz/ ‘canyons 4 come together' /c266n55s^ooge/ 'yellow jacket' from /c^tiSs 4 nSSs^ooge/ 'bug 4 it is yellow' /cfIdaago?istas/ 'scorpion* from /ell daago’istas/ 'red 4 meaning unknown1 Thus, certain kinds of morpheme boundaries override this prosody as they apparently do in Navajo also (Sapir and Hoijer 196?: 14-16). This prosodic feature of the foot has been en countered in other forms on lower levels. It was pointed out that in terms of syllable structure, that there were no syllables in the language which consisted of both alveolar and alveopalatal grooved obstruents. There are, 266 as well, restrictions upon sequences of these obstruents. In terms of economy, it may seem simpler if one rule was posited which required agreement of all grooved obstruents in the same foot. But I do not think that handling this kind of assimilation at just one level adequately accounts for the processes which are involved. If syllables really are part of one's competence then rules which specify their structure are too. If feet are really a part of one's competence then rules which apply to them are also necessary. Thus, although processes involving the agree ment of obstruents at different levels are obviously re lated, I think that a statement at each level is necessary. The Foot and the Meter As will be noted in the next section, feet have some differences in their prosodic qualities which are conditioned by their location in the meter, and by the type of meter in which they occur. I will not handle these differences in detail here, but merely comment on them generally. Feet which form the nucleus of a meter have over loud stress on their nuclear syllable. Furthermore, both timing and tonality are affected by the type of meter in which a foot is located. The speed of production of the foot is increased in certain types of meter, and the range of contrast between high and low tones is greater in some 26? meters than in others. These may be viewed as differences in the manifestation mode of the foot, and predictable relative to the higher level of unit to which we shall next turn our attention, the meter. Before doing this, however, I wish to present a sample text from which examples of meter, lines, and con tours will be drawn. A Sample Text Because the rest of this chapter is about units found only in continuous speech material, it is essential that some text material be available for illustrative pur poses. For this reason I am introducing the partial transcription of a text. It is representative of the dis course type known as /baanaadodi’I/ 'that which he tell to him' which is a narrative or story telling type of speech. The particular text presented here is the intro duction to a story about /mb6coh haastiihf/ 'old man big owl' entitled /mb6h bilaa7 sidaa/ 'Owl, his penis it sits,' a story recorded by Goodwin (1939: 178) under the title, "Big Owl Chops Off His Penis.11 Essentially, the story ' tells how Big Owl at one time had a penis which was so long that he had to carry it wrapped around himself. This proved so inconvenient that he chopped it off and left it under a pile of rocks which is located near the settlement of Canyon Day, which in the story is called /yaangogSiyS/ 268 'the place of many rabbits.’ The text is organized in this manner. Each contour is numbered to the left, and each line of the contour has a capital letter to the left. Over each meter at its center is a small letter identify ing the meter, and at the beginning of each meter is a letter classifying the meter as to type, a concept which will be explained below. Louder stressed syllables are underlined once, and extra loud syllables twice. Other phonetic symbols are as indicated in Chapter 3* Finally, over each line in the text there is an arrow drawn to represent the intonation contour. Throughout the rest of the chapter, as I have to refer to data in the text, I will cite it by referring to the numbers and letters which identify it. Thus, the first meter in the second line of the fifth contour will be cited as and so on. I. a. 1. A. [B *to"9&&'nii:"n9?"h / ‘ v *yaango1k&ih: / a-long-time-ago at many-rabbits h. c. *kvl lzeillyfih:/# a-place-called B. +■ -pi1 ya llgei*yuh///#?fB at that-place-below-it-over-there 269 I a. , 2. A. "Mb&'tshoh + ^haas1 : n/# owl-big old-man B. *mt'laa?=% ^-ndidul1 tho ^ /*ts?c »tii^n£?"h//##B his-penis he-was-carrying-it they-say N. a. P. b. A. *to"?S&'nii"ne?= / ^ o fl’a,nii"n0’” f === i --- i a-long-time-ago a-long-time-ago * §naagv,o' t?ii"'n&hSi/# 11 it-happened-then N. a. N. b. B. •”-yaaijgwo 1 kaih / *kul'ztt + *pi1 ya11 gel-yuh?//##B many-rabbits a-place-called that-place-below- it-over-there P, a. P . B7 4. A. + *to93 ,nii"n^,?=a / ■$$‘‘?5naagwo ,t2ii"3it*woh:/# II It(was) a-long-time ago it-happened-then T. a. N. b. B. ‘til + ,_s9^=a // ^nlfbL'laa?"^ / this(is) how his-penis N. c. %9ul»z^ + , T. a. N. ^ b. ^ 5. A. 'so’"3 / ^ b L ‘laa’= + »t&Sdaahunl't6tz^l&6bh://##B how his-penis it-was-being-very-long T. a. I. b. T. c. 6. A. •koh:--- / “?££ + *n'?££:h: , / 's3’“— f /# that particular-one (that's) how (it* was) N. ^ a. B. + *n@&'tl?66k= + 'l€h/# he-got-tired (of carrying it around) T. a. P. b. P. c. 7. A. 'koh/ t *n'9£f*y# / 'ya"ge£*y#*ldh//#B that-particular-place it-was-below- there N. a. ' K. b. B. 'ts^ll + *sL'zcn 4 *kvl'z&& / *PL'tl?a&h*Y& + rock it-is-standing a-place-called beneath- ' yaa^V##B there below T. a. 8. A. *99'khqu*lZ n’"h/##B ---- 1 over-there-it-was P. a. 9. A. 'ko + "?a£'sa? +. ^pt'taa + *zL'tZZ/# so that's-how on-this-side he-was-there N . ------a. B. *naaz'teo 4 ^na'p:; 4 'koh//##B ■r he-was-there roamin^-around thusly N. a. ----->> 10. A. *hanaazL'tS 4 ?Ls'ts^1 sn^hiih/# 1 across-from-him this-woman 271 N. a. B. *td&hn'yaa"l& n'?“h///##B i --- i was-coming-along I. a. N. b. 11. A. ‘ko -f °i.£: /^’ls 1 ts*1^ n uhn /# ■ ■ — 1 iii so that particular-woman B. she-was-begiming-to-beg I, a. 12. A. ^haa't’ii: / *?3nl'?ll: what do-you-have there -47 N . a. ^ N . b . 13. A. ■^haa't’ll + *1L / *taats" 1'y^*ni + what good-tasting kind-of-food ^t^aanl?!! + ^t^aano'liih'T////# 1 tii have-you (what) is-it T. A. 14. A. *n'tff*lf*ni9=i///#B she-said I. 15. A. 'an* I I 1 1 that particular-woman 272 N . EL.^ — B n ’- + ‘ZSf/## she-was-saying that 16. A *ko'tfl: / ^tots^i1ya-ni 4 ^^&as'?iil*ta^//##B he-said no-food I-do-not-have-any P. a.— ^ 17. A . ^ n * tii 4 "ts^L'til n/# he-said they-say P. J a. B . '^bd'ts^o 4 *haas'thji n///##B] owl-big old-man 273 The Meter Level The higher phonological unit I call the meter is composed of from one to four phonological feet, separated from like units by audible pause. It corresponds more or less to the grammatical phrase. As I will demonstrate, there seem to be several contrastive types, defined pri marily by their prosodic features. Meter Structure Nucleus At the core of a phonological meter in WMA is a foot which has over loud stress, especially on its nuclear syllable. Examples of these nuclear feet are identified in the preceding text as their nuclear syllables are under lined. Meter nuclei which are also line nuclei have double underlining. Boundaries At its boundaries the meter is marked by audible pause which varies in length from slight hesitation to lengthy silence. This pause is accompanied by a variety of phonetic markers. Thus, in meter final position final unaspirate stops are usually not only unreleased, but are held for a time and then released audibly into either a slight aspiration, or if the stop is a glottal stop, some times into an echo vowel. Examples of this phonetic marker 2?4 may be found in l.A.a., 4.A.a., 4.B.a., and 14.A.a. in the text. A second phonetic marker of meter boundaries is the lengthening of final fricatives, and the devoicing of voiced ones during this added length. For example: in I. A.C., 4.B.C., 6.A.a., and 7.B.b. final /h/ is length ened. This text does not happen to have meters ending in other fricative phonemes, but I have recorded meters end ing in other fricatives which also exhibit this charac teristic. A third phonetic marker of boundaries at this level is the tendency of morphemes which at foot boundaries or foot medially end in a vowel, to add an ninorganic /h/" (cf. page 133 also). Note that most meters in the text which do not end in another consonant do end in /h/. Finally, a fourth phonetic boundary phenomenon occurs with meters which end in a double vowel. There is a marked tendency for these to rise in tone. For examples of this, note the direction of the intonation arrows at meter boundaries in the text, especially in 7»B.a., 9.A.a., II. A.a., 12.A.a., and 16.A.a. Meter Prosodies Timing The primary prosodic feature of the meter is timing. It serves to contrast one meter type with another, and 275 also is somewhat "allophonically" different for any given meter type depending upon where the meter is located in the line and contour. By timing I am referring to speed of utterance; the amount of time given to each syllable; the spacing of feet throughout the meter; and the number of syllables contracted or slurred. Most of these phonetic facts are difficult to represent in a transcription. So, in dis cussing meter types which vary in terms of timing I will X simply describe the differences which I have observed in terms of the musical notion of 1 count1 and point out meters which exemplify each kind of timing. I assume that each syllable in a meter takes one count. Feet in meters which are faster take less than one. Feet in meters which are slower, more. (This does not in any way ignore the length prosody of the syllable mentioned in Chapter 6, or the difference in length between simple and complex nuclei, or even the timing prosody of the foot. This use of ‘count1 is intended to indicate that a complex open nucleus in a meter with one and a half counts to the nucleus is held even longer than the same nucleus given only one count.) On the other hand, the phonetic characteristics of the timing prosody observed as slurring and contraction are easily manifested through comparison of forms. An 276 example of contraction is found by comparing the first foot in 9.B.a. with the first foot in 10.A.a.. In the former foot is found [*naaz'teo] 'he was there,1 while in the latter the same stem and prefix (plus another prefix and minus a contracted last syllable) are present in their full form as f^hanaaz I'tS"! 'across from him.' In this particular case the contraction is caused in 9.B.a. be cause timing at the beginning of a contour final line is slightly quicker than that at the beginning of a contour non-final line. A similar case of contraction occurs in 2.B.a. and 5.A.b. In the former location and sequence [#b] has con tracted to [m], the phonetic forms being [*mC'laa?=] and [*AbL'laa9=] respectively. (Note that the high tone of [A] is transposed to the contracted syllable [ml-].) This- is because the timing in the beginning of a contour final line is faster than it is at other parts of the contour. Slurring is illustrated by comparing the word 'they say' in 2.B.b. with the same word in 17.A.a. In the former case the word clearly begins with a glottalized affricate and is [*ts?L'tff*ni?"h] while in the latter, which occurs in a different meter type, it is heard as [*tsh L'til n], having only an aspirated affricate. Finally, I want to emphasize that differences in timing seem to be due to both significant contrastive 277 meter types, and to variation due to location in line and contour. Pitch Range The other prosodic feature of the meter has to do with the range of pitch between high and low tone. I have already noted the slight differences between pitches on different syllable types. What I am referring to here is that in some meters the distance between a 1 and a 2 tone is greater than in others, and the distance between a high and a low can be either in barely discernible hops, or great sweeps. I have indicated this feature by the length of the rises in the intonation contours drawn over the transcription in the text. For example, in l.A.a. the pitch range is much greater than that for the same mor phemes in 3*A.a. These seem to be the only prosodic features which can be attributed to the meter as a unit. Meter Types It seems that in WMA story material at least, there are four basic kinds of meter which I call normal, parenthetical, introductory, and transition. Since their primary contrastive features are found in the prosodic features mentioned above, and it is difficult to symbolize these features on paper, I will describe the types I have 278 identified and point to certain meters in the text which are of that type. These different meter types are marked in the text N. for normal, P. for parenthetical, I. for introductory, and T. for transition. These abbreviations occur above each meter at its beginning. Normal Meter In normal meters, a full count is given to each syllable. Since the syllables themselves have length dif ferences relative to shape and type, this gives the normal meter a wave-like effect, with each foot composed of short but full onset syllables, a longer peak, and intermediate coda syllables. Tone differences are not extreme, highs being easily differentiated from lows. Normal meter has, however, a certain amount of variation depending upon its location in the line. For example, those at the initial line boundary are produced slightly faster than those line medially. Those near line final boundaries are also affected by the intonational features of the line. I have already called attention to the type of contraction in the initial meter of final lines which is caused by increased speed in that position. Introductory Meter Introductory meters, as the name implies are often the first meter in the phonological contour. But, as I will point out below, this is not always true. If it were. 279 introductory meter could be considered a predictable variant of normal meter, which was conditioned by its con tour environment. This type of meter is characterized by feet whose syllables are more than one count in length. The foot which is actually the nucleus of such meters has syllables which may be described as a count and a half in length. This is particularly noticable if the syllable in question has a complex nucleus to begin with. In these cases I have symbolized this especially long length by the use of a colon. Note particularly l.A.a., 2.A.a., 6.A.b., 11.A.a., and 16.A.a. Introductory meter is also characterized by the tremendous range involved between high and low tone. This is symbolized in the text by long rises or drops in the intonation lines over the meters marked with an I. Parenthetical Meter As the name implies, this type of meter usually occurs when extra phrases of an accessory nature are added to a sentence. It is especially used when 'stock phrases" such as 'they say' are added at the end of a contour, or when some word is repeated more than once. For example, in 3.A.b. the word 'a long time ago' (which is itself a somewhat stock phrase comparable to our 'once upon a time') is repeated with the slight parenthetical addition of the comment 'it happened then.' 280 Parenthetical meter is characterized by feet whose syllables are only given half or three quarters of a count. This rapid speed results in contraction and slurring. For example, in 3.A.b. the first phoneme in the meter comes out as a / V instead of a /d/ and the complex nucleus /55/ in the second syllable is reduced to /£/ (phonetically [$]). This is not to say that such things do not happen in other kinds of meters, it is just that in parenthetical ones they happen much more often. Parenthetical meters are also characterized by relatively little difference between high and low pitches, as I have tried to show with the contour markers over the relevant meters in the text. Transition Meter Transition meter generally is associated with con nectives such as /sa?/ 'what* and /goo/ ‘thus, so.‘ It also occurs with other types of phrases. For example, in 2.B.b. it involves the "stock" phrase 'they say' and in 8.A.a. the phrase 'over there it was.' Transition meter is characterized by a normal one count to the syllable timing, but a narrower pitch range. One might say it has the timing of normal meter, and the pitch range of parenthetical meter. In this way it tends to be exceptionally clear, with little slurring or con traction. 281 Meters as Contrastive Units I have claimed the above four types of meters are contrastive because they are not totally predictable from either the grammatical structure of the utterance, or from the phonological position of the utterance with regard to the line. Evidence that they are not totally predictable from the syntactic function or structure of the morphemes which make them up is the fact that a speaker may take the same or similar morphemes in essentially the same syntac tic position and utter them in different meters. For example, the introductory phrase in ll.A.a., 1 so that,1 is said as an introductory meter, while in f.A.a. the same morphemes are split into two meters, one of which is of the transitional type. Furthermore, the stock phrase * they say1 appears in the text in both parenthetical meter in I?.A.a. and in transitional meter in 2.B.b. Finally, there seems to be no rigid order to the arrangements of types of meter to form the line. Not all lines begin in introductory meters, some end in paren thetical meters, some in normal meters, and some in tran sitional meters. The Meter and the Ethnography of Speaking There is no doubt in my mind that meters are the locus of different discourse styles in UKA. I am sure 282 that one would find other characteristics which would de fine still more meter units in things like prayer, conver sation, and speech making. For instance, in Apache prayers which I have heard, the meters seem to be uniform throughout, rapid but not jumbled, with about three fourths of a count to the foot nucleus and perhaps a whole count for the non-stem syllables. In other words, the actual length of stem and non-stem syllables seems to be more nearly equal. The pitch range also seems equal throughout, with no parenthetical or transitional meters in evidence. Some differences in prayer-type discourse are also probably attributable to the presence of fewer meters to the line, and not so extreme a terminal intona tion at the end of contours. But the primary difference, I am sure, is to be found in the meter. The Line Level The phonological unit called the line is an in- tonational grouping composed of from one to three meters bounded by longer pause and occasionally breath. There are two kinds of line: contour non-final ones and contour final lines. Lines generally correspond to the grammati cal clause. 283 Line Structure Nucleus Like the foot and meter, the line seems to have a nucleus composed of a unit whose stress is relatively louder than any other unit in the string. This nucleus, then, by being the most heavily stressed meter in the line, is also coterminous with the nucleus of the foot and meter in which it occurs. Such line nuclear units are indicated in the text by double underlining on their peak syllable. Boundaries The line is bounded by relatively long pauses, and occasionally even non-final lines have breath taken at their conclusion. But the most common place for breath to occur is after final lines, i.e., at the end of the entire contour. The line, of course, has all of the boundary mark ers found with the meter, and, in fact at line boundaries with their longer pause these features are even more pro minent and likely.to occur. At the boundary of contour final lines, speech becomes much softer as well, and generally trails off into complete silence. 284 Line Prosodies Intonation The primary prosodic feature of the line can be called intonation. Essentially it relates to the height of pitches on different feet in the line. As I noted for the foot, feet in the same line are often absolutely higher or lower than each other. In contour final lines, for example, there is a steady drop in the actual pitch of the highest tones in feet as one gets closer and closer to the end of the line, with a corresponding drop in the lowest tones as well. It could be compared to a musical piece which always kept the same distance between notes, but dropped a tone in each measure the closer that measure was to the end of the song Timing A less important prosodic feature of the line is that of timing. By this I am referring to the amount of pause at meter boundaries, and to some extend the length of a 1 count * on the syllables in the meters which make up the line. The prosody of line timing here somewhat over rides the timing feature of the meter as I noted earlier causing there to be predictable variation in meter types relative to location in the line. 285 Line Types The two line types, non-final and final, may be considered complementary rather than contrastive, in that contours are usually made up of one or more introductory meters and a final meter. If one thinks of the line in this sense, then the differences in the two types are con ditioned by their location in the contour. Final Lines Final lines are characterized by falling intonation so that the last foot in the line is lower in tone than any other foot in the line. Furthermore, each meter and foot in contour final lines is progressively lower than the preceding one, and each is a little softer. They are marked in the data by double contour marks at line ends ([##]), and by falling contour arrows. This combination of intonation along with the tone distinctions which must be maintained in the language some times conspires to form a pattern which I at first thought might be another kind of meter. In lines 7.B., 9.B., and 12.A., most feet are composed of one or two low toned pre fixes, and high tone complex stem syllable nuclei with few or no enclitics in the same foot. This particular se quence of syllable shapes, stresses, tones, and intonation produces an unusual pattern which stands out distinctly from final lines like 6.B., with an assortment of tones 286 on a number of prefixes as well as ending in two enclit ics. The second distinctive feature of the final line is the tendency for the final enclitic on a verb, or some short connective morpheme to be set off in a separate foot, and to drop its pitch very rapidly, even if it is high tone. Take for example the final /?&£/ in 12.A. or the /yaah/ in 7.B. Both drop drastically in tone because of their position, and both are set off as separate feet. Finally, final lines begin with faster timing and then become slower in timing than non-final lines. Non-Final Lines Non-final lines are more difficult to characterize than are contour final ones for they are not so distinc tively marked in terms of intonational properties. They often seem to have a slight rise in pitch, building to a peak at the line boundary; or at least there is no great lowering of pitch as in final lines. Several such non final lines may occur in one contour, the first meter in each successive new line beginning slightly lower than that of the preceding one. At the end of such lines the timing quickens somewhat and the pitch range does not seem to be exceptionally great. High tone final syllables and low tone final syllables are therefore harder to identify. 28? The meter final boundary marker of upward tonal glide on complex syllable nuclei is evident, as in 9 .A., but the intonational decrease of pitch range tends to override other tonal features such as those described with regard to syllable shape. Non-final line boundaries give the impression of incompleteness, and the hearer waits with anticipation for the rest of the lines in the contour. Occasionally a short breath may be taken at non-final line boundaries as in 7.A. The Contour Level The higher phonological unit I call the contour usually consists of at least one non-final line followed by a final line, but the non-final line is not absolutely mandatory. The contour corresponds roughly to the gramma tical sentence, but the match is not always even, especi ally with regards to embedded quotations, and connective words. For example, in the text there are several con tours involved with the various quotations that seem to be syntactically one sentence. Note that contours 13. and 14. seem to be one sentence composed of two contours, as do contours 16. and 17. Both of these involve quotes which are embedded within a sentence. Similarly, connec tive words are sometimes attached to the end of one con tour, and in others seem to be part of the beginning of 288 the next contour. For example, 6.A. begins with a con nective, while 9.A. ends in the same connective; or 5»A. and 6.A. where the same connective is at the beginning of a contour in one case, but medial to a contour in another. These seem to be cases where the syntax is somewhat am biguous, and the phonological unit is the basis of syntac tic interpretation (of. Turner 1970a). Contour Structure Contours may consist of only one line under cer tain conditions, but it is always a final line. This is the case in 5«A., 8 .A., and 12.A. in the text. Apparently non-final lines are not sufficient to manifest the contour level unit. I suspect that if non-final intonation were given to an utterance, listeners would feel that the sen tence were unfinished, whereas if a single final line is presented, a complete contour has been uttered whether or not it is preceded by a non-final line. Nuclei Like other units in the phonological hierarchy, the contour may be considered to have a nucleus. In those contours which consist of two lines this nucleus is the end of the last non-final line before the final line. If, however, the contour is composed of only a final line, the line nucleus may be considered to be the contour nucleus. 289 In complete contours which consist of more than one line, the contour nucleus is the peak from which point final line intonation descends. In contours composed of only one final line, it is the more heavily stressed syllable which makes up the nucleus for the line. Boundaries Contours are usually followed by significant pause and breath. But breath does not always occur. It is tempting to think that in successive contours which in clude several syntacticly related sentences, such as the quotes mentioned above, no breath will usually be taken until the syntactic end of the sentence. This, however, is not completely true, for non-final lines can end in breath, as can contours involving quotes such as 16.A. There does seem to be such a tendency, however. Discussion and Summary Higher Level Units in Other Athapaskan Languages The most detailed description of higher level units in another Athapaskan language is that provided by Howren (1969: 4-8) for Dogrib. He reports that there are two types of what he called intonational phrases (IP's), final and non-final. The final type is characterized by the same kind of pitch lowering that I have described for 290 final lines in WMA. In the Dogrib non-final IP the pitch range tends to narrow, and to rise slightly before the boundary. This latter is similar to what I have noted for syllables with double vowels at meter boundaries and non-final lines. Howren does not present any text mate rial nor a complete discussion of the data, only four short examples. It seems likely that he is describing units which correspond to what I have termed the meter, and the contour final line. He includes no discussion of stress, timing, or other features which serve to define other higher level units. Durbin (1964: 40-41) also includes a short dis cussion of what he calls terminal contours in San Carlos Apache. He notes that falling pitch is the usual intona- tional marker which occurs at the end of an utterance and designates this marker as a phoneme of terminal juncture. This definitely corresponds to what I have called the final line. He also mentions a second type of contour which has rising intonation, however, in his data there apparently are few examples of this kind of phenomenon. Perhaps this is because the only text material he presents is that of a conversation (Durbin 1964: 62) involving relatively simple independent sentences which may not in volve the full contours I have pointed out in the sample text material presented here. 291 Finally, in Navajo Landar (1959) identifies three types of "terminal contours:11 rising, falling, and sus tained. It is my guess from his description that falling corresponds to what I have described as the final line, rising to the introductory meter, and sustained to the non-final line. The general agreement of these studies, that there exists at least a final type of intonation, suggests that linguists armed with comparable assumptions about the structure of the phonological hierarchy might agree as to the existence of the other higher level units such as the foot, meter, and line in the rest of the Athapaskan lan guages. Summary In this chapter I have described the foot, meter, line, and contour as units in the WHA phonological hier archy. It was pointed out that the foot is manifested by a series of syllables, one of which had primary stress. Its features consisted of an optional series of shorter syllables building up to a peak, longer stressed stem syl lables, and optional intermediate length coda syllables. It is distributed either initially, medially, or finally in a unit known as the meter. I further defined a unit called the meter which consists of a series of feet, one of which has above 292 ordinary stress which may have various prosodies of timing and tonality. These prosodies define at least four basic types of meter: introductory, normal, parenthetical, and transitional. These occur distributed in units called phonological lines. The next unit described was the line, a series of meters, which has a nucleus composed of one foot which is extra heavily stressed, and has the prosodic features of either non-final or final intonation. This latter is ex pressed by a falling off of loudness and absolute lowering of tone in successive feet as they approach the line bound ary. The former is expressed by a building of tone and decrease in pitch range. Finally, the contour was defined as being composed of one to several lines, the last always being a final line. The meter has a close relation to grammatical phrases, the line to dependent clauses, and the contour to a series of dependent clauses or an independent clause. CHAPTER 8 CONCLUSION This dissertation has been a description of the phonological hierarchy of the White Mountain dialect of Western Apache using generally Pike's tri-model of lan guage structure. As such it has emphasized the notion that language is composed of units which relate to each other in hierarchially arranged groupings. Summary In White Mountain Apache there are at least six levels: phoneme, syllable, foot, meter, line, and con tour. There are 34 contrastive segmental phonemes: 28 consonants, two resonants, and four vowels. There is one syllable structure with two variants or types based on location in the foot and phonemic struc ture. These types further correlate with the syntactic classes of stem and non-stem morphemes. The syllable is also characterized by prosodic features of stress, tone, nasalization, and length. Stress and length are totally predictable with reference to type and phonemic shape, while tone and nasalization are generally distinctive non- predictable features. Tone, although basically a 293 294 non-predictable feature, has certain phonetic variants which relate to syllable shape. Finally, syllable shape is conditioned or restric ted with relation to adjacent syllables, the grammatical class of the morphemes represented in the syllables, and the types of syntactic boundaries between such morphemes. In connection with such restrictions upon syllable shape a number of general phonotactic rules were suggested and an attempt was made to show the general unity of phonotac tic and morphophonemic rules. The foot level is represented by a unit composed of a cohesive series of syllables centered upon a nuclear or stem syllable which carries primary stress. It has prosodic features which include rhythm, tonality, and timing. Its variant forms relate to whether it has op tional onset and coda syllables, and its boundaries are marked by a variety of phonetic phenomena. The meter level is represented generally by a unit composed of a series of feet centered around a more heavily stressed foot, and bounded by audible pause. There are four contrastive meter forms: introductory, normal, parenthetical, and transitional. These are dis tinguished by differences in their prosodic features of timing and tonality. Using normal meter as a base, it was suggested that these differences could be described 295 by talking about the number of counts to the syllable; normal meter having one count to each syllable, and the others various portions or increases of counts. It was also suggested that the locus of discourse differences was in the meter and probably that other types of meters exist in WMA, which have not been identified. The line level is represented by a series of meters which build to a peak, or fall to a conclusion. These were called the non-final and final line respectively. Lines are bounded by longer pause and sometimes breath. Their distribution in the contour is regular with reference to that higher unit, so in one sense they may be thought of as predictable variants of a single unit. Finally, the contour is composed of at least one final line and may consist of one or more non-final lines followed by a final line. It is a unit most generally bounded by long pause and breath. Significant Aspects of WMA Phonology This study has pointed out the following signifi cant aspects of WMA phonology which should have relevance in classifying it with respect to other Apachean languages, and distinguishing it from other dialects of Apachean and Western Apache in particular. These include; 1. The presence of an archaic phoneme /mb/. 2. The absence of a phoneme /^d/. 296 3. The absence of a phonemic /w/. 4. The complementary relation of voiced and voiceless. unaspirate stops and affricates, and the occasional spirantization of unaspirated stops in certain positions. 5. Final free variation between /d/ and /g/. 6. The foot prosodies of timing and rhythm which ul timately cause the above mentioned voicing of unaspirate obstruents. 7. The correlation of the latter two aspects with the syntactic categories of stem, prefix, and enclitic mor phemes. 8. The predictability of stress, length, and phonetic tonal differences. 9. The existence of four types of meter which served to distinguish discourse types. 10. The existence of two variants of line intonation. Points of Confusion, Ambiguity, and Inadequate Data The most serious problem area seems to be in the section on phonotactic rules, most specifically in the section on /d/ and /g/ initial clusters over syllable boundaries. Since these are stem final consonants, and the only things that can follow them in the same foot are enclitics, the problem could also be stated as: what are the phonotactic restrictions on stem final consonant > 297 enclitic initial consonant sequences? More data needs to be gathered at this point to provide the most general set of phonotactic rules possible for the language. A second problem area relates to the decision to make an enclitic a separate stem, and thus a separate foot, or whether to include it as an adjunct to the preceding verb. Whether this is totally optional or whether phono logical shape of the preceding verb stem, or even location in the line decides this is not apparent in my data. As such, it creates a somewhat ambiguous point as to what is a stem and what is not; and when are certain allophon.es possible, and when not. A third problem relates to the prefix complex it self. I have described generally the surface phonological structure of prefix syllables, but it is known in other Apachean languages that actually prefix syllables are the result of rather complex processes of contraction and de letion (Hoijer 19^5c; 1946a; 1948a; 1948b; 1949). In my description I suggest that prefix syllables may have structural possibilities including both simple and complex nuclei, and opon or closed syllables. Reichard (1951: 9) writing about Navajo, claims that in that language all prefixes have a CV form canonically, and that any surface form which differs from this is the result of contraction. If this is true in Navajo, WMA, with its more rapid onset 298 timing is probably equally guilty, if not more so. To sort out the various processes which produce the surface forms observed calls for an intensive look at YJMA verbs and their prefixes. Finally, the meter types identified here are only suggestive of numerous differences which may occur in dif ferent discourse types in WMA. Further work in this area is necessary. In the definition of all higher level pro sodic patterns, sound spectrographic analysis of the data would be an immense help, and provide a much more accurate basis for classification and description. The Study in the Light of its Goals In Chapter 1, I set forth a series of goals for the study. The study has met those goals in the following manner. It has emphasized the structural and hierarchical nature of the phonemic system of WMA. It has provided a base for further investigation and process type descrip tions. It has provided a description which ethnographers can make use of in both transcribing and writing the lan guage. It has made a significant attempt to be explicit about the nature of higher phonological units as these are the basis of message form. And finally, it has provided % a firm and relatively complete description about a dialect of Apachean about which not much was known by the scien tific community. 299 Practical Application As noted, this study originally had as one of its goals the presentation of a set of phonemic symbols which could be used by ethnographers in the transcription of WMA words in connection with ethnographic research. These "scientific" symbols were presented in Table 3 in the be ginning of Chapter 5* This work has relevance beyond the scientific realm, however. White Mountain Apache is a viable language with a significant body of speakers, and as they continue to strive for separate ethnic identity and the integration achieved through cultural pride and language loyality, written materials should begin to play an important role. For example, it is well known by educational authorities that a child can learn to read more easily in his first language than in his second, and that once he learns in his first, it is easier to teach in his second language. Furthermore, as young Apaches grow up in a totally bilingual environment there will probably be some attempt by the society to stabilize the language, or at least slow the rate of change and pressure from the direction of English. Any such goals demand that WMA have an orthographic system which is both adequate and accurate to reflect only 300 those phonemic differences which are actually a part of the White Mountain dialect, and not include irrelevant San Carlos distinctions, Havajo distinctions, or even English contrasts which do not occur in WHA, A Proposed Orthography Basing most of the symbols and diagraphs upon Young and Morgan's Navajo orthography (1964: iii-vi), I would suggest the following as a possible White Mountain Apache orthographic system: Phoneme Letters Phoneme Lette: /mb/ mb A/ tl AV b A 9 / tl» /P/ P /!/ 1 /d/ d A/ 1 A/ t /g/ g A 9 / t' M k / j / dz A 9 / k ' A/ ts . /v i AV ts1 /g/ gh A/ z A/ h A/ s / y / y /V j /m/ m A/ ch A/ n AV ch' A/ i A/ zh A/ e 301 Phoneme Letters Phoneme Letters /s/ sh /a/ a A/ dl /o/ 0 High tone, /"/, could be marked with an accute accent and low tone left unmarked. Nasalization could be marked as I have been doing it with an apostrophe under the vowel. Thus, nasal 'a* would be ’a 1. Present Orthography There presently is In existence an orthography for WMA, that which is the medium for the Christian New Testa ment as published by the American Bible Society (1969). That orthography is essentially like the one presented here with some.exceptions. These are mainly in the area of under-differentiation and over-differentiation of pho nemes . The problems found in this orthography are as follows: 1. A /w/ phoneme is recognized, something which occurs in Navajo, but is allophonic in WMA. 2. A 'wh' diagraph is used, presumably to represent a A// phoneme. Again, something which is valid for Navajo but not for WMA. 3. A contrast between [u] and [o] is recognized. Something which is valid for English, and may be valid for Navajo, but is not valid for WMA. 302 4. The few words which begin in /mb/ are not recog nized, but are spelled with sometimes an ’m' and sometimes a ’b'. 5. Words that in WKA begin with a /d/ but correspond to San Carlos words which begin with an /n/ or /nd/ are spelled with ‘n1. This tends to place emphasis upon a marker which most White Mountain people feel is definitely a San Carlos linguistic trait. None of these problems are serious, but in terms of a society aiming for language stability and distinc tiveness, they must be considered faults. It would be my suggestion that any people responsible for the production of teaching materials, tribal newspapers, and other WMA written materials consider seriously the orthography I have suggested, and use it in place of other writing sys tems for this language. LIST OF REFERENCES ABERCROMBIE, DAVID 1967 Elements of General Phonetics. Aldine Publish ing Company, Chicago. AMERICAN BIBLE SOCIETY 1 1969 Bik'ehgo*ihi'nan Biyati1 (The New Testament in Western ApacheH The American Bible Society, New York. BASSO, KEITH H. 1969 Western Apache Witchcraft. Anthropological Pacers of The University of Arizona. No. 15. University of Arizona Press, Tucson. 1970 The Cibecue Apache. Holt, Rinehart and Winston, New York. BASSO, KEITH H. and MORRIS E. OPLER (eds.) 1971 Apachean Culture History and Ethnology. Anthro pological Pacers of The University of Arizona. No. 21. University of Arizona Press, Tucson. BITTLE, WILLIAM E. 1963 Kiowa-Apache. In Studies in the Athapaskan Languages. University of California Publications in Linguistics, Vol. 29. University of Cali fornia Press, Berkeley. BLOCH, BERNARD and GEORGE L. TRACER 19^2 Outline of Linguistic Analysis. Linguistic Society of America, The Waverly Press, Inc., Baltimore. CHOMSKY, NOAM 1965 Aspects of the Theory of Syntax. Massachusetts Institute of Technology Press, Cambridge. 303 304 CHOMSKY, NOAM 1967 Current Issues in Linguistic Theory. Mouton & Co., The Hague. CHOMSKY, NOAM and MORRIS HALLE 1968 The Sound Pattern of English. Harper & Row Publishers, New York. COOK, EUNG-DO 1971a Vowels and Tones in Sarcee. Language, Vol. 47, No. 1, pp. 164-179• Baltimore. 1971b Sarcee Numerals. Anthropological Linguistics, Vol. 13, No. 9, pp. 435-441. Bloomington. DeSAUSSURE, FERDINAND 1966 Course in General Linguistics. McGraw-Hill Book Company, New York. DURBIN, MARSHALL ELZA 1964 A Componental Analysis of the San Carlos Dialect of Western Apache: A Study Based on the Analysis of the Phonology. Korohonhonics, and Morphemics. Ph.D. Dissertation, Department of Anthropology, State University of New York at Buffalo, Buffalo. EDGERTON, FAYE E.• 1963 The Tagmemic Analysis of Sentence Structure in Western Apache. In Studies in the Athapaskan Languages. Harry Hoijer and Others. University of California Publications in Linguistics, Vol. 29. University of California Press, Berkeley. EVERETT, MICHAEL W. n.d. Adversaries. Conflict, and Retaliation: Suicide Among the White-Mountain Apache. Ms., Depart ment of Anthropology, University of Arizona, Tucson. 1971 White Mountain Apache Health and Illness: An Ethnographic Study of Medical Decision-Making. Doctoral Dissertation, University of Arizona, Tucson. 305 FERGUSON, CHARLES 1959 Diglossia. Word, Vol. 15, No. 2, pp. 325-340. New York. FIRTH, J. R. 1957 Sounds and Prosodies. Papers in Linguistics, Oxford University Press, London. FISHMAN, JOSHUA A. 1971 Sociolinguistics: A Brief Introduction. Rowley, Newbury House Publishers. FORBES, JACK D. 1966 The Early Western Apache, I3OO-I7OO. Journal of the West, Vol. 5» No. 3» pp. 336-35^* Los Angeles. FOX, ROBIN 1967 Kinship and Marriage. Penguin Books, Baltimore. FRIES, CHARLES and KENNETH PIKE 1949 Coexistent Phonemic Systems. Language. Vol. 25, No. 1, pp. 29-50« Baltimore. GATSCHET, ALBERT S. 1879 Classification into Seven Linguistic Stocks of . Western Indian Dialects Contained in Forty Vocabularies. In Report on United States Geo graphical Surveys West of the One Hundredth Meridian, by First Lieut. George h. Wheeler, Vol. 7, Archaeology, Appendix, Linguistics. Government Printing Office, Washington. 1970 Zwolf Sprachen aus dem Sudwesten Nordamerikas. Anthropological Publications. Oosterhout N.B., Niederlande. GLADWIN, THOMAS and WILLIAM C. STURTEVANT (eds.) 1962 Anthropology and Human Behavior. The Anthro pological Society of Washington, Washington, D.C. 306 GODDARD, PLINY EARLE 1911 Jicarilla Apache Texts. Anthropological Papers of the American Museum of Natural History, Vol. 8. The American fouseum of Natural History, New York. 1919 San Carlos Apache Texts. Anthropological Papers of the American Museum of Natural History, Vol. 24, Part 3« The American Huseum of Natural History, New York. 1920 White Mountain Apache Texts. Anthropological Papers of the American Museum of Natural History, Vol. 24, Part 4. The American Museum of Natural History, New York. 1929 The Bear River Dialect of Athapaskan. University of California Publications in American Archae ology and Ethnology. Vol. 24. No. 5. University of California Press, Berkeley. G00DEN0UGH, WARD H. 1957 Cultural Anthropology and Linguistics. In Report of the Seventh Annual Round Table Meeting on Linguistics and Language Study, Monograph Series on Languages and Linguistics No. 9 . Georgetown University, Washington, D.C. G00DEN0UGH, WARD H. (ed.) 1964 Explorations in Cultural Anthropology: Essays in Honor of George Peter Murdock. McGraw-Hill Book Company, New York. GOODWIN, GRENVILLE 1939 Myths and Tales of the White Mountain Apache. Memoirs of the American Folklore Society, Vol. 33. J. J. Augustin, Publisher, New York. 1942 The Social Organization of the Western Apache. The University of Chicago Press, Chicago. 1971 Western Apache Raiding and Warfare from the Notes of Grenville Goodwin, edited by Keith H. Basso. The University of Arizona Press, Tucson. 307 GOSS, JAKES A. 1965 Ute Linguistics and Anasazi Abandonment of the Four Corners Area. American Antiquity, Vol 31, No. 2, Ft. 2, pp. 73-81. Salt Lake City. 1966 Culture-Historical Inference from Utaztekan Linguistic Evidence. Paper presented to the Plenary Symposium on Utaztekan Prehistory at the Joint Session of the Society for American Archaeology and the Great Basin Anthropological Conference, May 1966. Reno. GREENFELD, PHILIP J. 1971 Playing Card Names in Western Apache. In Notes and Reviews. International Journal of American Linguistics, Vol. 37, No. 3, pp. 195-196. Baltimore. GRIMES, JOSEPH E. 1969 Phonological Analysis: Part One. Summer Insti tute of Linguistics, Inc., Santa Ana. GUMPERZ, JOHN J . and DELL HYHES (eds.) 1964 The Ethnography of Communication. Special Pub lication. American Anthropologist. Vol. 66, No. 6, Pt. 2. Kenasha. HAAS, MARY R. 1968 Notes on a Chipewyan Dialect. International Journal of American Linguistics Vol. 34, No. 3> pp. 163-175' Waverly Press, Baltimore. HALLE, MORRIS 1964 On the Bases of Phonology. In The Structure of Language: Readings in the Philosophy of Language edited by Jerry A. Fordor and Jerrold J. Katz, Prentice-Hall, Inc., Englewood Cliffs. HAMMBL, E. A. (ed.) 1965 Formal Semantic Analysis. Special Publication. American Anthropologist. Vol. 67, No. 5* Pt. 2. Menasha. ' 308 HARMS, ROBERT T. 1968 Introduction to Phonological Theory.' Prentice- Hall, Inc., Englewood Cliffs. HARRIS, ZELLIG S. 194^ Simultaneous Components in Phonology. Language, Vol. 20, No. 3, pp. 181-20$. Waverly Press, Baltimore. 1945 Navaho Phonology and Hoijer‘s Analysis. Inter national Journal of American Linguistics, Vol. 11, No. 4, pp. 239-246. Waveriy Press,Baltimore. HESTER, JAMES J. 1962 Early Navaho Migrations and Acculturation in the Southwest. Museum of New Mexico Papers in Anthropology. No. 6. Museum of New Mexico Press, Santa Fe. HILL, FAITH 1963 Some Comparisons Between the San Carlos and White Mountain Dialects of Western Apache; In Studies in the Athapaskan Languages, by Harry Hoijer and others. University of California Publications in Linguistics, Vol. 29. Univer sity of California Press, Berkeley. HOCKETT, CHARLES 1954 Two Models of Grammatical Description. Word, Vol. 10, No. 2-3, pp. 210-234. New York. HOIJER, HARRY 1938 The Southern Athapaskan Languages. American Anthropologist. Vol. 40, No. 1, pp. 75-8?. Menasha. 1939 Chiricahua Loan-Words from Spanish. Language♦ Vol. 15, No. 2, pp. 110-11$. Waverly Press, Baltimore. 1943 Pitch Accent in the Apachean Languages. Language, Vol. 19, No. 1, pp. 38-41. Waverly Press, Baltimore. 309 HOLIER, HARRY 1945a Mavaho Phonology. University of New Mexico Publications in Anthropology No. 1. The University of New Mexico Press, Albuquerque. 1945b Classificatory Verb Stems in the Apachean Lan guage . International Journal of American Linguistics, Vol. 11, No. 1, pp. 13-23* V/averly Press, Baltimore. 1945c The Apachean Verb, Part I: Verb Structure and Prononinal Prefixes. International Journal of American Linguistics, Vol. 11, No. 4, pp. 193- 203. V/averly Press, Baltimore. 1946a The Apachean Verb, Part II: The Prefixes for Mode and Tense. International Journal of American Linguistics. Vol. 12, No. 1, p. 13* V/averly Press, Baltimore. 1946b The Apachean Verb, Part III: The Classifiers. International Journal of American Linguistics, Vol. 12, No. 2, pp. 51-59* V/averly Press, Baltimore. 1946c Chiricahua Apache. In Linguistic Structures of Native America, edited by Cornelius Osgood. Viking Fund Publications in Anthropology No. 6. The Viking Fund, Inc., New York. 1948a The Apachean Verb, Part IV: Major Form Classes. International Journal of American Linguistics, Vol. 14, No. 4, pp. 247-259. V/averly Press, Baltimore. 1948b Linguistic and Cultural Change. Language, Vol. 24, No. 4, pp. 335-345. Baltimore. 1949 The Apachean Verb, Part V: The Theme and Pre fix Complex. International Journal of American Linguistics. Vol. 15, No. 1, pp. 12-22. V/averly Press, Baltimore. 1956 The Chronology of the Athapaskan Languages. International Journal of American Linguistics. Vol. 22, No. 4, pp. 219-232. V/averly Press, Baltimore. 310 HOIJER, HARRY 1971 The Position of the Apachean Languages in the Athapaskan Stock. In Apachean Culture History and Ethnology, edited by Keith H. Basso and Morris E. Opler. Anthropological Papers of the University of Arizona. No. 21. University of Arizona Press, Tucson. HOIJER, HARRY and JANET JOEL 1963 Sarsi Nouns. In Studies in the Athapaskan Lan guages . by Harry Hoijer and others. University of California Publications in Linguistics, Vol. 29. University of California Press, Berkeley. HOIJER, HARRY and MORRIS E. OPLER 1938 Chiricahua and Mexcalero Apache Texts. Univer sity of Chicago Press, Chicago. HOIJER, HARRY AND OTHERS 1963 Studies in the Athapaskan Languages. University of California Publications in Linguistics. Vol. 29. University of California Press, Berkeley. HOWARD, PHILIP G. 1963 A Preliminary Presentation of Slave Phonemes. In Studies in the Athapaskan Languages, by Harry Hoijer and others. University of California Publications in Linguistics, Vol. 29. Univer sity of California Press, Berkeley. HOWREN, ROBERT 1968 Stem Phonology and Affix Phonology in Dogrib (Northern Athapaskan). In Papers from the 4th Regional Meeting of the Chicago Linguistic Society, edited by B. J. Darden and others. Chicago. 1969 The Phonology of Rae Dogrib. Ms., The Bulletin of the National Museum of Canada. 1971 A Formalization of the Athabaskan ‘D-Effect.1 International Journal of American Linguistics. Vol. 37, No. 2, pp. 96-II3. Waverly Press, Baltimore. 311 HUNG, BEVERLY Y. P. 1959 On the Phonemic Status of Navaho Stress. Anthropological Linguistics, Vol. 1, No. 1, pp. 20-23. Indiana University, Bloomington. HYMES, DELL H. 1957 A Note on Athapaskan Glottochronology. Inter national Journal of American Linguistics, Vol. 23, No. 4, pp. 29I-297. Waverly Press, Baltimore. 1962 The Ethnography of Speaking. In Anthropology and Human Behavior, edited by Thomas Gladwin and William C. Sturtevant. The Anthropological Society of Washington, Washington, D.C. 1964 Introduction: Toward Ethnographies of Communi cation. In The Ethnography of Communication, edited by John J. Gumperz and Dell Hymes. American Anthropologist, Vol. 66, No. 6, Pt. 2, pp. 1-34. Menasha. JAK0BS0N, ROMAN and MORRIS HALLE 1956 Fundamentals of Language. Mouton and Company, The Hague. KAUT, CHARLES R. 1957 The Western Apache Clan System: Its Origins and Development. University of New Mexico Publi cations in Anthropology No. 9. University of New Mexico Press, Albuquerque. KIM, CHIN-W. 1965 On the Autonomy of the Tensity Feature in Stop Classification (With Special Reference to Korean Stops). Word. Vol. 21, No. 3» PP* 339- 359* New York. KRAUSS, MICHAEL E. 1964 Proto-Athapaskan-Eyak and the Problem of Na- Dene: The Phonology. International.Journal of American Linguistics. V0I..3O, No. 2, pp. 118- 131. Waverly Press, Baltimore. 312 KRAUSS, MICHAEL E. 1970 Review of "The Phonology and Morphology of the Navaho Language" by Sapir and Hoijer. Inter national Journal of American Linguistics,Vol. 36, No. 3, pp. 220-228. Uaverly Press, Balti more. LANDAR, HERBERT J. 1959 The Navaho Int©national System. Anthropological Linguistics. Vol, 1, No. 9, pp. 11-19. Bloomington. 1963 Navaho Syntax. Language. Vol. 39, No. 3, Part 2. V/averly Press, Baltimore. LANDAR, HERBERT J., SUSAN M. ERVIN and ARNOLD E. HOROWITZ 1960 Navaho Color Categories. Language, Vol. 36, No. 3, pp. 368-382. V/averly Press, Baltimore. LI, FANG-KUEI 1930 Mattole An Athabaskan Language. The University of Chicago Press, Chicago. LISKER, LEIGH and ARTHUR S. ABRAMSON 1964 A Cross-Language Study of Voicing in Initial Stops: Acoustical Measurements. Word, Vol. 20, No. 3, pp. 384-422. New York. LONGACRE, ROBERT E. 1964 Grammar Discovery Procedures. Mouton & Co., The Hague. MIERAU, ERIC 1963 Concerning Yavapai-Apache Bilingualism. Inter national Journal of American Linguistics, Vol. 29, No. 1, pp. 1-3. Waverly Press, Baltimore. OSGOOD, CORNELIUS (ed.) 1946 Linguistic Structures of Native America. Viking Fund Publications in Anthropology No. 6. The Viking Fund, Inc., New York. 313 PIKE, KENNETH L. 1943 Phonetics: A Critical Analysis of Phonetic Theory and a Technic for Practical Description of Sounds. The University of Michigan Press, Ann Arbor. 1947a Phonemics: A Technique for Reducing Languages to Writing. The University of Michigan Press, Ann Arbor. 1947b Grammatical Prerequisites to Phonemic Analysis. Word. Vol. 3, No. 3> pp. 155-172. New York. 1952 More on Grammatical Prerequisites. Word. Vol. 8, No. 2, pp. 106-121. New York. 1956 Towards a Theory of the Structure of Human Be havior. In Estudios antropologicos publicados en homenaje al doctor Manuel Camio. Sociedad Mexicana de Antropologia, Mexico, D.F. 1958 Interpenetration of Phonology, Morphology and Syntax. In Proceedings of the 8th International Congress of Linguists. Edition De L 1Academic De La Bepublique Fopulaire Roumaine, Bucarest. 1967 Language in Relation to a Unified Theory of The Structure of Human Behavior. Mouton 6 Co., The Hague. POSTAL,, PAUL M. 1968 Aspects of Phonological Theory. Harper & Row, New York. REICHARD, GLADYS A. 1951 Navaho Grammar. Publications of the American Ethnological Society. No. 21. J. J. Augustin Publisher, New York. ROBINSON, DOW F. 1970 Workbook for Phonological Analysis. Summer Institute of Linguistics, Santa Ana. 314 SAPIR, EDWARD 1936 Internal Linguistic Evidence Suggestive of the Northern Origin of the Navaho. American Anthro pologist , Vol. 38, No. 3> PP« 224-235• Menasha. 1949a The Psychological Reality of Phonemes. In Selected Writings of Eduard Sapir in Language. Culture and Personality, edited by David G. I-iandelbaum. University of California Press, Berkeley. 1949b Language: An Introduction to the Study of Speech. Harcourt, Brace & World, Inc., New York. SAPIR, EDWARD and HARRY HOIJER 1967 The Phonology and Morphology of the Navaho Lan guage . University of California Publications in Linguistics No. 5°« University of California Press, Berkeley. SAVILLE, MURIEL R. 1969 Navajo Morphophonemics. University Microfilms Limited, Ann Arbor. SCHANE, SANFORD A. 1971 The Phoneme Revisited. Language, Vol. 47, No. 3, pp. 503-521. Waverly Press, Baltimore. SCHROEDER, ALBERT H. 1955- Fray Marcos de Nizay Coronado and the Yavapai. 1956 New Mexico Historical Review. Vol. 30, No. 4, pp. 265-296; Vol. 31, No. 1, pp. 24-37. Albuquerque. 1963 Navajo and Apache Relationships West of the Rio Grande. El Palacio. Vol. 70, No. 3, PP* 5-23• Santa Fe. SECOY, FRANK R. 1951 The Identity of the "Padouca": An Ethnohistori- cal Analysis. American Anthropologist. Vol. 53, No. 4, pp. 525-542. Menasha. 315 STANLEY, RICHARD 1969a Navaho Phonology. Unpublished paper. 1969b Review of The Phonology and Morphology of the Navaho Language, by Edward Sapir and Harry Hoijer. Language. Vol. 45, No. 4, pp. 927-939. Waverly Press, Baltimore. TURNER, PAUL R . 1970a Grammar for the Speaker and Hearer. Linguistics: An International Review. Vol. 57, pp. 93-99. Kouton & Co., The Hague. 1970b Phonological Corequisites to Grammatical Analy sis. Linguistics: An International Review, Vol. 60, pp. 66-73. Mouton & Co., The Hague. U.S. DEPARTMENT OF COMMERCE 1971 Federal and State Indian Reservations: An Economic Development Administration Handbook. U.S. Government Printing Office, Washington. VIVIAN, GWINN R. I960 The Navajo Archaeology of the Chacra Mesa. Master's Thesis, University of New Mexico, Albuquerque. WALLACE, ANTHONY F. C. 1970 Culture and Personality. Random House, New York. WEINREICH, URIEL 1967 Languages in Contact. Mouton & Co., The Hague. YOUNG, ROBERT W. and WILLIAM MORGAN 1964 The Navaho Language. A Publication of the Edu cation Division, United States Indian Service. Desert Book Company, Salt Lake City. 6 9 7* 7