A segmental phonology of black English 9783111356129, 9783110999983

Citation preview

JANUA LINGUARUM STUDIA MEMORIAE NICOLAI VAN WIJK DEDICATA edenda curat C. H. VAN S C H O O N E V E L D Indiana University

Series Practica, 191

A SEGMENTAL PHONOLOGY OF BLACK ENGLISH

by

P H I L I P A. L U E L S D O R F F

Universität Regensburg

1975 MOUTON THE HAGUE • PARIS

© Copyright 1975 in The Netherlands Mouton & Co. N.V., Publishers, The Hague No part of this book may be translated or reproduced in any form, by print, photoprint, microfilm or any other means, without written permission from the publishers.

L I B R A R Y O F C O N G R E S S CATALOG C A R D N U M B E R 72-94483

Printed in Hungary

ACKNOWLEDGMENTS

The present study was conceived and begun in 1966-67 while I was Project Linguist with the Urban Language Study of the Center for Applied Linguistics in Washington, D.C. To the Director of the Center, Charles A. Ferguson, the Director of the Urban Language Study, Joey L. Dillard, and the Coordinator of Linguists of the Urban Language Study, Marvin D. Loflin, I owe a profound debt of gratitude for providing a free and provocative environment within which to work. The first draft of this work was completed during the summer of 1969 and the second draft during the winter of 1970.1 wish to express my heartfelt appreciation to Charles W. Kreidler for his comments on the first draft which resulted in numerous improvements. To Frances Lamberts and Wendy Uncles I am grateful for graciously volunteering to assist in the preparation of the typescripts. No study exists in vacuo, least of all the following. My debt to Noam Chomsky and Morris Halle and the handful of researchers on Black English, especially William Labov and William A. Stewart, is pervasive. To Marvin D. Loflin, however, I owe much of my intellectual independence and inspiration. Without his initiative, encouragement, and sincere devotion to the pursuit of truth, this study would never have seen the light of day. Last but not least, I wish to thank my principal informant for his unbounded patience and tolerance during the 500 hours of elicitation which provided the empirical foundations upon which this study rests. It is to him that this study is gratefully and respectfully dedicated.

TABLE O F C O N T E N T S

Acknowledgments

5

I Introduction

9

II Dialectology in Generative Grammar

21

III Segment Structure Conditions

32

IV Sequence Structure Conditions

44

V Phonology of the Word

59

VI Phonology of Inflection

73

VII Phonology of Derivation

80

VIII Conclusion

87

Bibliography

90

Index of Names

92

Index of Citation Forms

93

Index of Subjects

99

I INTRODUCTION

0. SUMMARY

In Section 1 I state the purpose of the present study. Section 2 supplies background information relevant to the present study and details several shortcomings inherent in the approach adopted. Section 3 provides a summary statement of the theoretical framework within which the study is written. Section 4 contains a brief discussion of field procedures, stressing the importance of systematic elicitation to the construction of a generative grammar. Section 5 is a review of the literature related to the phonology of Black English and a statement of the major differences between previous conclusions and those presented in this study. The final section is a statement of the way in which the present study is organized.

\. PURPOSE

The purpose of the present study is both empirical and theoretical. On the one hand, I present a description of the segmental phonology of a dialect of English spoken by Black adolescents in the District of Columbia. On the other, I attempt to contribute to the development of phonological and dialectological theory by providing some basis for specifying the ways in which the. dialect under investigation differs phonologically from Standard English. In addition, I hope that the results of this study will prove useful to those interested in preparing pedagogical materials for the teaching of Standard English as a second dialect1.

2. BACKGROUND

The data for this study were collected over a ten month period in 1966-67 while I was Project Linguist with the Urban Language Study of the Center for Applied Linguistics in Washington, D.C. 2 . The principal informant was a fourteen-year-old 1 2

Cf. Luelsdorff (1970). The data elicited support a much broader study than the present work.

10

INTRODUCTION

male native resident of Washington who turned fifteen during the course of the elicitation. Some of the results of the study are generalizable to the speech of other Blacks in the Washington area and in Delaware and New York. The scope of the study and the theoretical framework in which it is written (transformational grammar) necessitated working in depth with one informant over a protracted period of time. Several shortcomings are inherent in this kind of approach. First of all, working with only one principal informant leaves the problem of the extent to which the analysis is generalizable to other portions of the Black community an open question. In this regard, I can only hope that future research will contribute to our understanding of the nature and extent of interpersonal variation in Black speech. Second, it became apparent during the course of this investigation that variant pronunciations existed in the speech of the principal informant for what were intuitively felt by him to be one and the same words. This phenomenon of intrapersonal variation emerged only after an attempt lasting several months to encourage the informant to speak with me as naturally and uninhibitedly as he would with his family and friends. Although it is difficult to assess the extent to which this attempt was successful, the large number of doublets elicited testifies to some measure of success. Typically, the informant described one of the pronunciations as the way he would speak when talking with family and friends and the other as the way he would speak when talking with teachers and strangers. I labeled (unoriginally) the former pronunciation 'casual' and the latter 'careful'. In the vast majority of cases, careful pronunciation corresponds to Standard English. Third, in view of the length of the period of elicitation coupled with the informant's daily exposure to Standard English, it was expected that the informant's speech patterns would change in the direction of more formal Standard English. I attempted to cope with this dynamic situation by the process of socialization mentioned above and by a preliminary survey of the segmental phonology in which some of the major dialect variants were noted. One of the most obvious changes was the articulation of preconsonantal and word-final r's in a dialect which, at the outset of the study, was r-less. 3. THEORETICAL FRAMEWORK

This study is written within the theoretical framework of generative phonology. Of the descriptions of the form and substance of generative phonology, those of Chomsky and Halle (1968), McCawley (1968), Stanley (1967, 1968), and Zwicky (1965) are among the best. A generative phonology is a set of rules which are divided into three subgroups according to their function 3 . The first of these subgroups contains redundancy condi3 Chomsky & Halle (1968) discuss the necessity of a fourth group of rules, called "readjustment rules".

INTRODUCTION

11

tions4 which function to specify redundant feature specifications in phonological segments on the basis of feature information contained within those segments and feature information contained within surrounding segments. The first type of redundancy condition is called a 'segment structure condition' and the second a 'sequence structure condition'. On the basis of the information that a vowel is Back and High, for example, it is possible to predict that it will also be Rounded. Since this prediction depends exclusively upon information contained within the same phonological segment, it is an example of a segment structure condition. Alternatively, if a morpheme begins with two consonants followed by a liquid, it is possible to specify that the first consonant will be s. Since this rule requires feature information in surrounding segments for its proper operation, it is an example of a sequence structure condition. The second subgroup of rules consists of phonological rules which function to assign stress and to add, delete, and rearrange phonological segments. Stress rules depend upon syntactic information for their proper operation and apply cyclically, first to the innermost constituents of the utterance. The labeled brackets surrounding the innermost constituent are then erased and the rules reapply to the next innermost constituent, etc. In some variety of Standard English, for example, there is a Compound Stress Rule which assigns primary stress in nouns to a primary-stressed vowel preceding a vowel with primary stress. All monosyllables are assigned primary stress and, by convention, when a vowel in a constituent is assigned primary stress, the 1 i stresses on all other vowels are reduced by one. Thus, (^(¿black^^board)^)^ remains unaltered by the rules during the first pass through the cycle, and the inneri

I

most brackets are erased yielding (Nblack board)N. During the second cycle, primary stress is assigned to the first primary-stressed vowel by the Compound Stress Rule, 1 2 1 1 yielding (Nblack board) N . Similarly, the noun phrase (JV/>(/4black)^(iVboard)Ar)JVi, remains unaltered during the first pass through the cycle, and the innermost brackets i

I

are erased yielding (^pblack board) NP . During the second pass through the cycle, primary stress is assigned by the Nuclear Stress Rule to the second primary-stressed 2 1 vowel yielding (^pblack board) NP . The second type of phonological rule is illustrated by the rule which deletes the vowel of the plural ending -ez in the environment of a preceding segment which is not both Coronal and Strident5. All the features (excluding Stress) in the redundancy conditions and phonological rules are binary in that they are assigned either the coefficient ' + ' or the coefficient ' — I t is the function of the third subgroup of rules to convert these binary features into integers representing different degrees of the presence of the feature in question. 4 6

Stanley (1967, 1968) places the redundancy rules in the lexicon. Cf. Luelsdorff (1969) for a discussion and justification of this rule.

12

INTRODUCTION

Since high back vowels are more rounded than mid back vowels, for example, the feature Rounded is assigned a higher integer in the case of back vowels which are High than in the case of back vowels which are non High.

4. FIELD PROCEDURES

Research in dialectology has traditionally involved the use of questionnaires as the means of eliciting primary linguistic data in support of an analysis. While this method is of demonstrated value in the area of sociolinguistic research (Labov, 1966; Wolfram, 1969), questionnaires are in principle incapable of yielding data in support of a theory of linguistic competence6. Consider the case of the Detroit Dialect Study (Shuy, Wolfram & Riley, 1968). One of the stated objectives of this study was to describe the specialized linguistic features of the various English-speaking sub-cultures of Detroit, that is, to construct a theory of the various English dialects spoken in Detroit. Following the typical procedure of dialect geographers, a questionnaire was used in order to elicit data in support of this theory. The questionnaire consists of six parts. The first four parts contain various questions designed to elicit protracted discourse. Such questions include: What are your favorite TV programs? Did you ever have a teacher who hollered a lot? What do you do in a day's work? How does your family celebrate the holidays ? The sixth part consists of words, sentences, and a text for oral reading. At best, these questions elicit informant responses containing SOME dialect features, but they fall far short of eliciting the paradigms of grammatical and ungrammatical sentences necessary for constructing a grammar. That this is the case is testified to by the failure of a recent attempt (Fasold, 1969) to construct a grammar fragment based exclusively on the recorded data of the Detroit Dialect Study. In this study, the author was forced to resort to procedures of SYSTEMATIC ELICITATION in order to obtain the paradigms he knew to be relevant to his grammatical analysis. One fundamental difficulty with the questionnaire approach, then, is that it is incapable of providing data in support of a complete and correct linguistic theory. The fifth part of the Detroit questionnaire is designed to elicit lexical differences across dialects in the areas of household terms and food and cooking terms, among others. But here the questions which are asked are asked precisely because dialect differences are known IN ADVANCE to exist with respect to the items on the question6 Unless one wishes to semantically interpret the expression 'competence' in an hitherto unknown way.

INTRODUCTION

13

naire. Thus, a second fundamental difficulty with the questionnaire method is that it is restricted to eliciting known dialect differences, thereby arbitrarily circumscribing the range of dialect differences that might otherwise be uncovered. Because the questionnaire method does not yield data in support of a complete and correct grammar and assumes apriori knowledge of the range of dialect variation, it was not used as a means of collecting data for the present study. Rather, phonetic paradigms were systematically elicited and the principal informant was encouraged to express his linguistic intuitions. Systematic elicitation involves using whatever techniques are feasible for obtaining completely representative data in support of a linguistic theory, including the speaker's judgements concerning the grammaticality of sentences. If the informant is initially unable to distinguish sentences that are ungrammatical and sentences that are false (cf. Fasold, 1969), then he should be tutored so as to be able to make the distinction. Ideally, the informant should be just as linguistically sophisticated as the linguist7.

5. REVIEW OF RELATED LITERATURE

In recent years several publications have appeared containing information on the pronunciation of English by lower-class urban Blacks. The purpose of this section is to summarize some of the contributions of these linguists to our understanding of the phonology of Black English and to state the major differences between the conclusions which they reach and those presented in the present study. Labov (1966) observes that Black English is r-less, not only in the sense that r is not pronounced before other consonants and at the ends of words, but also in the sense that intervocalic r is absent across word boundaries and sometimes absent word internally in words containing intervocalic r in Standard English. Thus, intervocalic r is absent in expressions like four o'clock and in words like Carol, Paris, and terrace. In the present study, no examples of the absence of word-internal intervocalic r were found, but r-lessness was found to extend to the absence of r in a great many words which in Standard English contain r's in clusters before and after weakly stressed vowels. Thus, while r is present in Carol, Paris, and terrace, it is absent from words like professor, protect, wizard, and lizard. Labov further observes that Black English is /-less in the sense that I is often replaced by a back unrounded glide and disappears entirely, especially after back rounded vowels. Labov finds tendencies toward homonyms in pairs such as toll:toe, tool: too, help:hep, all:awe, etc. In the present study, on the other hand, tense and lax vowels were found to neutralize in the environment of a following preconsonantal or word-final I, and underlying preconsonantal and word-final I were found to be phonetically realized as a back off-glide agreeing in tenseness and height with the 7

For an informative report on a linguist's attempt to sophisticate his informant, cf. Hale (1965);

14

INTRODUCTION

preceding (not necessarily immediately preceding) vowel. Thus, the dialect under investigation exhibits a kind of neutralization and vowel harmony not observed by Labov8. Further, the above mentioned pairs reported to be homonyms by Labov were not found to be homonyms in the D.C. data, the vowel of the first word in each pair being longer than that of the second. Labov further notes the general Black English tendency toward simplification of consonant clusters at the ends of words. The list of clusters which simplify is not complete and no rules are given to express this process. In the present study, a complete list of simplifying clusters is given, in addition to a set of cluster simplification rules. Further, although Labov includes -nt among the clusters which simplify, -nt was not found to simplify in the speech of my principal informant. The same is true of underlying -ks. Among the examples of final consonant weakening, Labov includes the frequent deletion of underlying t, d, k, and g, the frequent devoicing of d and g, and the frequent realization of underlying t and k as glottal stop. All of these processes were observed to be operative in the Washington data, with the exception of the deletion of underlying word-final t and k. Moreover, all voiced obstruents and obstruent clusters were found to devoice in utterance-final position. Labov further notes that final m and n are usually phonetically manifested as some degree of nasalization in the preceding vowel. The same conclusion was found to be true of the Washington data. Other phonological features of Black English noted by Labov include the absence of a distinction between i and e before nasals, between ih and eh before r and sometimes /, between uh and oh before r, the monophthongization of ay and aw, the monophthongization of oy before /, the merger of 0 and/and 6 and v word-finally and intervocalically, and the presence of skr, jw and sr in several words which in Standard English contain str and Sr. In the Washington data, there are no examples of the neutralization of lax / and lax e before r or /, but I have observed this phenomenon in St. Louis. I also find no examples of the monophthongization of oy before / or of Standard English morpheme initial str heard as skr. In the speech of the principal informant, words which in Standard English contain initial Sr were pronounced in the Standard way, but I have heard Sr pronounced [sr] in Madison and Chicago. Finally, there are many examples of the neutralization of vowels other than lax i and lax e before r resulting, for example, in fairy being pronounced the same as furry and carriage the same as courage. In the paper under review, Labov was the first to point out that there may be differences in the underlying representation of Standard and Black English noun stems. I quote the relevant passage in extenso: In the most casual and spontaneous speech of the young Negro people whose language we have been examining, the plural //-s// inflection is seldom deleted. It follows the same phonetic rules as in standard English: 1. after sibilants /s, z, §, i / , the regular plural is [iz]; 8

Personal communication, December 1966.

INTRODUCTION

15

2. after other voiceless consonants, [s]; 3. elsewhere, [z]. The regular form of the plural after a word like test, desk, is [s], as in [tests]. If the rules were so ordered that we began with the abstract form //test//, added the //-siI, and then deleted the /t/ in the final phonetic form [tesiz], [gosiz], [tosiz], as the plurals of test, ghost, and toast. A form such as [tesiz] implies an order of the rules which begins with //tes//, or reduces //test// immediately to /tes/. Then the plural //-s// is added, and the phonetic rules give us [tesiz]. It should be emphasized that those speakers who use this form do so consistently, frequently, and in the most careful speech: it is not a mere slip of the tongue. In a section dealing with grammatical correlates of phonological variables, Labov makes it abundantly clear that many of the phonological processes in the dialect cannot be stated without reference to grammatical information. Although this is a commonplace in the literature of generative phonology, I recapitulate several examples of this phenomenon here, since I have not done field work in all the areas in which Black English phonology and grammar interrelate. The future Labov notes that there is no question about the existence of the grammatical category of future in Black English, since will is present in its emphatic or full form. The loss of final /, however, has a serious effect on the realization of future forms: you'll = you they'll = they

he'll - he she'll = she

It is obvious that the rule which vocalizes / (cf. above) in word-final and preconsonantal position does not apply in these cases since the colloquial future is identical with the colloquial present. What is required in these cases is an obligatory / deletion rule which deletes / when it is immediately followed and preceded by a word boundary. This rule applies first in the ordered sequence of rules which operate on underlying I. The copula Labov observes that the copula paradigm is seriously affected by phonological processes : I'm ^ I you're = you he's ? he

we're = we you're = you they're = they

Among the rules which are needed to account for copula deletion is one which deletes the z in he's. In this case, it is sufficient to state that z is deleted in the environment of a following and preceding word-boundary. Labov (1968) reports that wherever Standard English can contract, Black English can delete is and are, and vice-versa, and that wherever Standard English cannot

16

INTRODUCTION

contract, Black English cannot delete is and are, and vice-versa, and that there are no Black English speakers who always or never delete is in those environments in which it can be contracted. Examples: SE Tom's wild *Tom wild.

BE Tom's wild. Tom wild.

You're sick. *You sick.

You're sick. You sick.

Labov proposes rules for contraction and deletion and claims that the problem of ordering these rules cannot be solved without reference to quantitative data. Such a claim, if correct, would result in an enlarged conception of the notions 'rule of grammar' and 'linguistic competence'. It is therefore important that the problem of ordering the deletion and contraction rules be carefully examined in order to determine if Labov's claim is correct. Labov presents four possible relations of order between contraction and deletion: Case 1 l.C 2. D

Case 2 l.D 2. C

Case 3 1. iCl

Case 4 l.C(D)

az — z j . . .

sz-+0f...

az->-fzl/...

3Z

Z

0f. . .

9Z

—

zf...

To)

[0J

In Case 1, contraction occurs first, deletion second. In Case 2 deletion occurs first, contraction second. Labov interprets Case 3 as showing deletion and contraction as simultaneous alternates of the same rule, with only one set of environmental constraints. Labov interprets Case 4 as showing deletion as an extension of contraction, again with only one set of environmental constraints. Consider Case 2, where deletion precedes contraction. It is easy to see that if deletion precedes contraction, the contraction rule will have nothing to contract unless the deletion rule fails to apply. In other words, the operation of the obligatory contraction rule is incorrectly made contingent upon the prior non-application of the optional deletion rule. Consider Case 3, where deletion and contraction are given as simultaneous alternates of the same rule, with only one set of environmental constraints. Here, there are two things to be said: first, the simultaneous rule hypothesis has not been incorporated into the theory of generative phonology9 and, second, the environmental constraints on contraction are different from the environmental constraints on deletion. For example, is does not contract in the environment of a 9

Cf. Chomsky & Halle (1968), 19.

INTRODUCTION

17

preceding word ending in a [+Coronal, -f Strident] segment, but it does delete in this environment. Examples: The fish is good. *The fish's good. The fish good. Thus, Case 3 is both ruled out by the theory of generative phonology and observationally inadequate. Case 4, with deletion as an extension of contraction, is the same as Case 1, except that the environmental constraints in Case 4 are the same, whereas the environmental constraints in Case 1 are different. Case 4 is thus ruled out on the same grounds as Case 3: the environmental constraints on contraction and deletion are different. Having ruled out Cases 2-4, we are left with Case 1, where contraction precedes deletion under different environmental constraints. For the time being, this appears to be the correct claim, and this is the conclusion which Labov arrives at based on a consideration of quantitative data. The foregoing remarks have been directed toward demonstrating the incorrectness of Labov's claim that the problem of ordering the contraction and deletion rules cannot be solved without reference to quantitative data. Since this claim is incorrect, it does not follow that the notions 'rule of grammar' and 'linguistic competence' have to be enlarged so as to include certain quantitative notions. Other aspects of Labov's paper will be considered in Chapter V. Wolfram (1969) discusses several phonological and grammatical features of Black English. Like all previous treatments of the phonology of Black English, Wolfram's discussion is informal and incomplete, but it also embodies several claims about the grammar of Black English which are incorrect. Wolfram (1969, 51, 3.4.1) gives a list of word-final consonant clusters in which the final member of the cluster may be absent. Wolfram's list includes [st], [sp], [sk], [st], [zd], [2d], [ft], [vd], [nd], [md], [Id], [pt], [kt]. Clusters in which voicing or voicelessness is not a defining characteristic of the entire cluster such as [mp] (e.g .jump), [nt] (e.g. count), [It] (e.g. colt), [nk] (e.g. crank), and [lp] (e.g. gulp) are not included in his analysis on the grounds that they do not function in the same way as the above list. Wolfram's discussion of the phonology of word-final clusters embodies two claims, both of which are incorrect: (1) All word-final consonant clusters in which voicing or voicelessness is a defining characteristic of the entire cluster may simplify by losing their final member in certain environments, and (2) All word-final consonant clusters in which voicing or voicelessness is not a defining characteristic of the entire cluster do not simplify by losing their final member in certain environments. As counterexamples to the first claim, I cite [ks] and [ps] (e.g. wax, lapse) and [lb], which never simplify, and as counterexamples to the second claim I cite [rst] and [mpt] (e.g. burst, attempt), which do simplify by losing their final members in certain environments. 2

18

INTRODUCTION

Aside from the observational inadequacy of Wolfram's claims, it makes absolutely no sense to talk about the final member of a PHONETIC cluster being absent. Phonetically, you either have a cluster or you do not have a cluster, there can be no question of the absence of the final member. Fasold (1969) discusses a number of grammatical and phonological features of Black English which differentiate the dialect from Standard English. I will consider only some of those aspects of Fasold's discussion which relate to the pronunciation of Black English. The third person singular present tense marker Fasold and others have noted that the dialect has no third person singular present tense marker. The relevant paradigm is: He walk. The man walk. He don't walk. The man don't walk. He have a bike. The man have a bike. There is thus no phonological evidence of a third person singular present tense suffix in Black English. The simplest solution to this problem entails postulating a rule which rewrites the present tense grammatical category as zero. The -ED suffix The -ED suffix is phonetically realized in certain kinds of grammatical and phonological environments and not realized in others. Some of the relevant examples are: Yesterday I wash my face. Yesterday I washed it. Yesterday I wash it. Yesterday I burn my hand. Yesterday I burned it. Yesterday I burn it. If the -ED suffix is phonetically realized as either t or d (cf. Luelsdorff, 1969, for a discussion of rules for the phonology of English inflection), and preceded by a consonant and followed by a consonant, the t and d are obligatorily deleted. If, on the other hand, the t and d are preceded by a consonant and followed by a VOWEL, they are optionally deleted. In the case of verb stems ending in a vowel, the d realization of the -ED suffix may be optionally deleted by the word-final consonant weakening rule mentioned above.

INTRODUCTION

19

Examples: Yesterday he played it. Yesterday he play it. Apparently, this rule is indifferent to the first sound of the following word. Thus, Yesterday he played the game. Yesterday he play the game. are apparently equally acceptable. In the paper under consideration, Fasold mentions a very interesting example of grammatical constraints on the application of phonological rules: In Negro dialect, the -ed suffix is absent from verbs ending in a vowel less often than for verbs ending in a consonant besides t or d. When a verb ends in one of these two consonants, the -ed suffix (pronounced id) is absent even less often and is absent only under certain restricted circumstances. A few verbs can be used with infinitive phrases or with nominalized verbs (He started crying, and He wanted to go). When such verbs end in / or d (start, want, need, end up), and are used in one of these two constructions, the i of id may be deleted. The verb then ends in dd or td which is then simplified to d. These two operations are common to both standard English and Negro dialect and result in sentences like He stard crying and He wanda go. At this point, in Negro dialect either the consonant cluster simplification rule or the word-final d elimination rule can apply, giving sentences like He sta crying and He en up coming}0 In general, this is the only situation in which the -ed suffix can be absent from a verb ending in t or d.

Footnote 10 states: "These same verbs never lose the -ed suffix when they appear at the end of a clause. Sentences like He wanted to drive, but he could not get the car sta or That was what he need do not occur, r is absent in sta because of pronunciation rules in Negro dialect." Omission offorms of auxiliary have In Black English the contracted form of the auxiliary have may be optionally deleted, yielding I been herefor hours and He gone home already. We thus require a rule which deletes v in the environment of a preceding and following word-boundary. Will and would The contracted forms of will and would may be optionally deleted, yielding sentences like He miss you tomorrow from underlying He'll miss you tomorrow or He'd miss you tomorrow, «/-deletion is effected by the optional word-final d-deletion rule and /-deletion may be accounted for by postulating a rule which deletes / in the environment of a preceding and following word-boundary (cf. above). 2*

20

INTRODUCTION

Auxiliary be As in the case of copula be, the contracted forms of the auxiliary be may be optionally deleted unless the subject is /, in which case the contracted form'm or its full form am are always present. We therefore require a rule which optionally deletes r and z in the environment of a preceding and following word-boundary ( c f a b o v e for a discussion of copula deletion). 6. ORGANIZATION

This study is divided into eight chapters, roughly following the organization of the phonological component of a generative grammar sketched above. Chapter II, "Dialectology in Generative Grammar", surveys several past treatments of the syntax and phonology of dialects within the framework of generative grammar and states the assumptions on which the present study is based. In Chapter III, "Segment Structure Conditions", the phonemes of the dialect are given, defined in terms of their distinctive feature composition, and a set of segment structure conditions are formulated functioning to predict redundant feature specifications. Chapter IV, "Sequence Structure Conditions", gives the distribution of consonants in morpheme initial and morpheme final position and a set of sequence structure conditions predicting values for redundant feature specifications in consonants in clusters. Chapter V, "Phonology of the Word", deals with phonological changes within the word not associated with either inflection or derivation. Chapter VI, "Phonology of Inflection", is restricted to a discussion of aspects of the phonology of inflection, while Chapter VII concerns aspects of the phonology of derivation. The concluding chapter deals with some of the major theoretical issues touched on in the preceding chapters.

II DIALECTOLOGY

IN

GENERATIVE

GRAMMAR

0. SUMMARY In the present chapter I examine several past approaches to the theory of syntactic and phonological variation within generative grammar. The general conclusion of this recent work is that dialects of the same language may differ in their deep structures. This unanticipated conclusion, if correct, entails an approach to the study of intrapersonal and interpersonal variation considerably more individually, i.e. clinically, oriented than past approaches have in fact been. In Section 1 I examine several contributions to the theory of syntactic dialectology. In Section 2 generative contributions to phonological dialectology are reviewed and a distinction is drawn between grammars underlying the speech behavior of individual speakers and grammars which relate the speech behavior of individual speakers. In Section 3, I formulate the principle on which the present study is based and, in Section 4, present seven examples from the phonology of Black English which, when analyzed on the basis of the principle stated in Section 3, entail underlying phonological representations different from those of Standard English.

1. SYNTACTIC DIALECTOLOGY Contributions to the theory of syntactic dialectology are, unfortunately, extremely few in number and, typically, very poor in quality. Exceptions to this rule are t h e recent contributions of E. S. Klima, P. S. Rosenbaum, and M . D . Loflin. Klima (1964) approaches the relationship between grammatical systems and their differences in the following way. The syntactic structure of each system is considered revealed by the set of rules which most economically generates the sentences of t h e system : That set of rules will be designated as its grammar (G). The relationship between one style (Li) and another (L2) will be thought of in terms of the rules (Ei_2) that it is necessary to add as an extension to the grammar (Gi) of Li in order to account for the sentences of L2. A convention will be adopted regarding the place where extension rules may be added to the grammar. They may not be added just anywhere, but must come at the end of certain

22

DIALECTOLOGY IN GENERATIVE GRAMMAR

sets of rules; e.g. extension rules dealing with the case forms of pronouns must come after the set of grammar rules for case in the previous system. By this convention, extension rules are prevented from superseding previous rules. Fundamental structural difference, varying in structure and degree, will be considered to exist between systems Li and L2 when the set of rules G 2 for most economically generating the sentences of L2 is not equivalent to Gi plus its extension EI_ 2 . From the point of view simply of relating coexisting systems, the particular pairing and the direction chosen in extending the grammar of one system to account for the sentences of another are those representing the shortest extension rules, that is, Li is compared with L2, and L2 with L3, rather than Li with L3, because EI_ 2 and E 2 _ 3 are each shorter than a hypothetical EI_ 3 . Similarly, Li rather than L2 is taken as primary in comparing Li and L2 because E]_ 2 is shorter than a hypothetical E 2 _!.

Although motivated by a purely synchronic principle of simplicity (shortness of rules), Klima points out that the order in which the styles are considered does, in fact, recapitulate comparable aspects in the historical development of the English pronouns. Rosenbaum (1964) calls for a characterization of the notion DIALECT in terms of transformational theory and the determination of the level on which dialects differ from one another. Due to the then lack of transformational research in the area of syntactic dialectology, Rosenbaum leaves the problem of deep structural differences between dialects of the same language an open question. He concludes by stressing the need to develop a motivated methodology for describing and relating different linguistic systems and investigating the constraints on dialect divergence. In 1965 Beryl Bailey advanced the bold and exciting hypothesis that the speech of the Southern Negro differs in deep structure from the speech of the Southern White (Bailey, 1965). In support of this hypothesis, Loflin (1967b) presents evidence that certain copulative sentences with be are ambiguous in Nonstandard Negro English (NNE) and that in order to disambiguate such sentences an HABITUATIVE category must be postulated in the grammar of NNE not found in the grammar of Standard English. The HABITUATIVE category appears to have the function of representing a recurring activity engaged in at specific times. Loflin further observes that there is no perfective form in NNE comparable to the one posited for Standard English (have + en): surface realizations of have + en in simple sentences, yes/no questions, tag questions, and nominalizations are totally lacking.

2. PHONOLOGICAL DIALECTOLOGY

Applegate (1961) presents a description of the speech of two children whose "subdialect" deviates from Standard English. It was the only language of the children, and they used it to communicate with English-speaking adults, with each other, and with their playamtes. The children's brother served as a translator in situations where the children could not communicate effectively with adults.

DIALECTOLOGY IN GENERATIVE GRAMMAR

23

Although Applegate refers to the children's dialect as an "autonomous system", he describes it in terms of its deviant relationship to Standard English. One relevant passage is worthy of quotation in extenso: At this point reference to the language of the community is helpful. In adult speech, the form of the plural morpheme, the possessive, and the third-person singular verb morpheme is [s] alternating with [z] in those places where the children have [t] and [d]. Further examination of the children's speech shows that all tricatives are missing. /{/ is articulated as /p/, /v/ is /b/, /s/ is /t/ and /z/ is /d/. The interdental fricatives are also articulated as stops, and /c/ and /j / are /t/ and /d/ respectively. The two series of stops in the children's language are then related in the same way that stops and fricatives are related in adult speech. A rule of the following form, again in terms of distinctive features, can be used to describe this phenomenon. + consonant — vocalic + continuant X

+ consonant — vocalic — continuant X

The application of this rule results in the change of all fricatives, i.e., continuants, to noncontinuants, or stops.

This rule, rather than describing the dialect as an autonomous system, describes it in terms if its deviation from Standard English. The children have no fricatives in their speech, yet underlying fricatives are postulated for their dialect in the interests of relating it to Standard English. While rules of this type may in fact reflect the way in which the dialect was acquired and the knowledge a speaker of the Standard language would have to internalize in order to speak the dialect, they are unjustified in the description of the dialect because there are no forms in the dialect which would support their postulation. A sharp distinction must be made between grammars underlying the speech behavior of individual speakers and grammars relating the speech behavior of individual speakers. The goal of the former is the accurate and complete description of the linguistic competence of selected members of the speech community. The goal of the latter is to relate these descriptions to one another in a maximally simple and illuminating way. The former are statements of the linguistic competence of individual language users; the latter are statements of the similarities and differences in the linguistic knowledge of individuals in the community. Exemplary of the generative approach to dialectology is a study by Halle (1961) in which he considers a hypothetical dialect of English - a dialect with almost exactly those features as described by Applegate in the above-mentioned publication - differing from the standard language in the following two respects: When the standard language has a continuant consonant in noninitial position, the dialect has the cognate noncontinuant (stop) consonant.

24

DIALECTOLOGY IN GENERATIVE GRAMMAR

When the standard language has several identical noncontinuant consonants in a word, the dialect replaces all but the first of these by a glottal stop. Examples : I cuff (cup) gave (Gabe) sauce (sought) lies (lied)

II [kÁp] [gép] [SDt]

[lájd]

puff brave toss dies

III [pÁp] [bréb] [tôt] [dájd]

pup babe taught died

[pÁ?] [bé?] [tó?] [dáj?]

Note that the dialect admits words with several identical noncontinuant consonants (cf. Column II), but that in every one of these examples the second noncontinuant corresponds to a fricative in the Standard language. Halle handles the phonetic peculiarities of this dialect by the following two ordered rules, which do not function in the standard language: (1) If in a word there are several identical nonvocalic, consonantal noncontinuants, all but the first become nonvocalic, nonconsonantal noncontinuants (i.e., glottal stops in distinctive feature terminology). Examples in Column III. (2) In noninitial position, nonvocalic, consonantal continuants become noncontinuant. Halle comments on this solution as follows: I believe that this solution, proposed by Applegate, is preferable to the alternative of postulating a different phonological system for the dialect than for the standard language. It seems to me intuitively more satisfactory to say, as we have done here, that the dialect differs from the standard language only in the relatively minor fact of having two additional low-level rules, rather than to assert - as we should have to do if we rejected the proposed solution that the dialect deviates from the standard language in the much more crucial sense of having either a different phonemic repertoire than the standard language, or of having a strikingly different distribution of phonemes. It must be stressed that in the proposed solution the ordering of the rules is absolutely crucial, for if Rule (1) is allowed to operate after Rule (2) the noncontinuants produced by Rule (2) would be turned into glottal stops by Rule (1); i.e., the examples in Column II could not be accounted for. Without ordering of the rules we are forced to accept the unintuitive alternatives mentioned above.

Halle (1962) characterizes recent work in dialectology as focusing attention on the facts of the utterance and concerning itself primarily with questions of mutual intelligibility of two dialects, the similarities and differences of cognate utterances, of their phoneme repertoires, distributional constraint, etc. Instead of following this procedure, Halle proposes to focus on the grammars of the dialect, that is, on the ordered set of statements that describe the data, rather than on the data directly. Halle considers the case of Pig Latin in order to show that these two approaches are different in fundamental ways. The former approach entails noting the distributional differences of the phonemes of Pig Latin and General American, that infixation is the major morphological device in Pig Latin, and that Pig Latin is incomprehensible to the uninitiated speaker of General American. On the basis of such observation we

DIALECTOLOGY IN GENERATIVE GRAMMAR

25

would be led to falsely conclude that Pig Latin and General American were at best remotely related tongues. The latter approach involving a comparison of the grammars of General American and Pig Latin, results in the observation that Pig Latin contains a morphonemic rule that is absent in the more standard dialects, namely, the shift of initial consonant clusters to the ends of words and the addition of /e/. Since this rule is the only difference between the grammars of Pig Latin and General American, we conclude that Pig Latin is a special dialect of General American, a conclusion which is obviously right. Halle states that this result follows only if instead of concentrating on utterances, we shift primary attention to the grammars that underlie the utterances. Saporta (1965) offers evidence from Spanish dialects to support the view of Halle that the grammatical description of a given dialect may be converted into an adequate description of a related dialect by the addition, deletion, and reordering of a relatively small number of rules. Saporta stresses that the choice of underlying forms and rules is motivated by the desire to account for the greatest number of facts in a manner as straightforward as possible. This desire, however, leads Saporta to postulate underlying representations for a given dialect which are unsupportable on the basis of the primary data of that dialect. As an example, consider Saporta's treatment of the morphonemics of pluralization in Latin American Spanish: Castilian (C)

Latin American (LA)

lunes lunes íápie lápiOes

lúnes lúnes làpis lápises

'Monday' 'Mondays' 'pencil' 'pencils'

For C there is a general rule for the plural expressed in (1). (1)

pi

•

[sIV. e 0/Vs es

This rules states that the plural is represented by /s/ in the environment after all unstressed vowels and after stressed /e/, by 0 after unstressed vowels followed by /s/, and by /es/ elsewhere. Thus the plural of lunes /lúnes/ with final /s/ is also /lúnes/, but the plural of lápiz /lápiG/ with final /0/ is lápices /lápiOes/. Saporta notes that /0/ and /s/ have fallen together in LA and concludes that some modification must be made in the grammar. One alternative is to list for LA all words like lápiz /làpis/ as exceptions to the rule - that is, to say that the plural is represented by zero after unstressed vowel plus /s/ except for a list of words, when this list corresponds exactly

26

DIALECTOLOGY IN GENERATIVE GRAMMAR

to words where C has /0/. The other alternative is to keep (1) unaltered, but add a rule like (2) for LA: (2) [s] Saporta adopts the second alternative, arguing that linguists presumably agree that rules are preferable to lists, even short lists. This presumption, however, even if correct, is irrelevant to the justification of underlying representations. Assuming Rule (1) to be correct, it is an adequate expression of the morphophonemics of pluralization in Castilian Spanish. A rule like (2), in addition to (1), might be found, perhaps in an optional form, in the grammar of a bidialectal speaker of Castilian and Latin American Spanish. But Rule (2) is totally unmotivated on the basis of the data of Latin American Spanish alone. Given these data alone, there is no alternative between a list and a rule: a list must be given. A speaker of C learning the morphophonemics of pluralization in LA must internalize a rule of the sort given in (2). A speaker of LA learning the morphophonemics of pluralization in C, however, is confronted with a different task. First, he will have to learn that the list of exceptions including /lapis/ in his grammar consists of forms ending in /0/ in Castilian Spanish. The underlying representation of the final consonants of these forms will change from /s/ to /0/ and he will internalize (2) as an optional rule. The end products of these two different processes of second dialect acquisition will be the same: identical underlying representations with identical optional rules. The native speaker of Castilian has added an optional rule to his grammar with the acquisition of Latin American Spanish as a second dialect. Having formed the plural of /lapiO/, he applies an optional rule, converting /lapi0es/ to [lapises]. The native speaker of LA has changed his underlying representation and learned an optional rule. Interestingly enough, he must apply more rules in speaking his native dialect after he has learned Castilian than when he was a monodialectal speaker of Latin American Spanish.

3. PRINCIPLES A N D P R O C E D U R E S

The essential feature of the above approaches to phonological dialectology in generative grammar is the principle that the description of a given dialect may be converted into an adequate description of a related dialect by the addition, deletion, and reordering of a relatively small number of rules. Call this the Dependence Principle. In each case examined, however, this principle has led to the postulation of underlying representations for a dialect identical to those of the system from which it is derived, irrespective of whether or not those representations are justifiable on the basis of the dialect data alone. THIS PRINCIPLE HAS THE UNDESIRABLE RESULT OF MAKING THE ADE-

DIALECTOLOGY IN GENERATIVE GRAMMAR

27

QUATE DESCRIPTION OF A DIALECT IMPOSSIBLE WITHOUT RECOURSE TO THE DATA OF

Since generative grammars are representations of the linguistic competence of individual speaker-hearers, this principle must be rejected. In its stead, I offer the principle that each dialect may be adequately described in its own terms, i.e., without reference to the data of related dialects. Call this the Independence Principle. The associated procedure is to produce such descriptions for individual speaker-hearers. These independently motivated individual descriptions may then be compared with an eye to pointing out their similarities and differences, leaving the extent and nature of dialect differences an essentially empirical question.

SOME OTHER VARIETY OF THE LANGUAGE.

4. UNDERLYING PHONOLOGICAL DIFFERENCES

In this Section I present several examples of Black English phonetic data which, when analyzed on the principle that each dialect may be adequately described on its own terms, result in systematic phonemic representations for Black English which are different from those of Standard English. Example A In Black English there are no examples of lax [e] occurring in the environment of an immediately following nasal consonant. Corresponding to Standard English words in which lax [E] occurs in the environment of a following nasal are Black English words in which lax [i] occurs in the environment of a following nasal. Examples:

pen hem

SE

BE

[pen] [hsm]

[pin] [him]

According to the Dependence Principle, the phonology of Black English would differ from the phonology of Standard English in that it would contain a rule which rewrites e as [i] in the environment of a following nasal consonant. According to the Independence Principle, Black English would have an underlying i before nasals in just those words in which Standard English has an underlying e before nasals and the phonology of Black English would contain a Sequence Structure Condition stating that if a nonTense, nonBack, nonLow true vowel occurs in the environment of an immediately following nasal consonant, then the vowel will also be specified High. Thus, according to the Dependence Principle, the underlying representations of Black English and Standard English are the same, and the phonology of Black English differs from that of Standard English in that it contains an additional phonological rule. According to the Independence Principle, the underlying representations of

28

DIALECTOLOGY IN GENERATIVE GRAMMAR

Black English and Standard English are different and the phonology of Black English contains a sequence structure condition which the phonology of Standard English does not. Since I accept the Independence Principle and reject the Dependence Principle, I conclude that there are underlying differences in the phonologies of Standard and Black English. It might be argued that the analysis that I am proposing fails to express the relationship between Standard and Black English. But this relationship is expressed by the rule which rewrites e as i in the environment of a following nasal. This rule is of a formally different character from phonological rules, it does not occur in the phonology of either dialect, and is postulated in order to account for the differences between the two dialects. Example B In Black English there are several words which contain [a] before [r] which have standard English counterparts with either [e] or [ae] before [r]. Examples: SE very fairy carriage

[veri] [fed] [kaerij]

BE [vari] [fori] [karij]

This correspondence between Standard English and Black English is not completely regular, however, for there are some words which contain [e] and [ae] before [r] in both language varieties. Examples:

hairy Larry

SE

BE

[heri] [laeri]

[heri] [laeri]

According to the Dependence Principle, the same underlying representations would be postulated for Black English and Standard English, and a rule would be added to the phonology of English which would have the effect of rewritting e and ae before r as phonetic [a] in the case of words like very, fairy, carriage, Mary, merry, etc. and words like hairy and Larry would be marked as exceptions to this rule. According to the Independence Principle, different underlying representations would be postulated for Standard English and Black English in the case of very, fairy, carriage, etc., but not in the case of hairy and Larry, other things being equal. The two systems would be related to one another by a rule which maps the underlying representations with ae and e before r in Standard English onto Black English underlying representations with a before r, with certain words being marked as exceptions. This rule would not be a phonological rule in either Standard English or Black English, but would merely serve to relate the two systems in an economical and revealing way.

DIALECTOLOGY IN GENERATIVE GRAMMAR

29

Example C In Black English there is a large number of words which do not contain an [r] where an [r] is present in Standard English prevocalically and postvocalically in consonant clusters. Examples: SE wizard custard protect professor

[wizard] [kastard] [pratskt] [prafesar]

BE [wizit] [kastit] [patek] [pafesa]

Following the Dependence Principle, an underlying r would be postulated in those environments in which it is phonetically absent in Black English but phonetically present in Standard English, and a rule would be added to the phonology of Standard English which would delete the r's in Black English in the appropriate environments. According to the Independence Principle, on the other hand, no underlying r's would be postulated for Black English in these environments and the relationship between the two systems would be expressed by a rule which deletes underlying Standard English r in the appropriate environments. Again, this rule would not be a phonological rule in either Standard English or Black English, but would merely serve to relate the two systems in an economical and revealing way. Example D In Black English the world bulb is pronounced [by :p], and, in some varieties of Black English, the word help is pronounced [hep]. According to the Dependence Principle, underlying lb and Ip clusters would be postulated for these words and the Black English pronunciation would be derived by introducing a rule which would delete underlying I in the environment of a following bilabial stop. According to the Independence Principle no underlying clusters would be postulated for these two words, even though there is an /-deletion rule operative in Black English which is independently motivated, and especially frequently operative in the environment of a following bilabial stop. For if we were to posit underlying clusters for bulb and help on the basis of the independently motivated /-deletion rule, there would be nothing preventing us from positing underlying clusters for cub, sub, pep, etc., except the correct but irrelevant observation that bulb and help end in clusters in Standard English, whereas cub, sub, pep, etc., do not. Note that the irrelevance of this observation is a direct consequence of my acceptance of the Independence Principle. Thus, even in the case of underlying representation by means of the application of independently motivated phonological rules, these representations are rejected if they necessarily entail considering Standard English data.

30

DIALECTOLOGY IN GENERATIVE GRAMMAR

Example E It is well-known fact that Black English is r-less. Among other things, this entails the absence of phonetic [r] in word-final position. Examples:

sister father bother smother

SE

BE

[sistar] [faòar] [baòar] [smadar]

[sista] [fava] [bava] [smava]

According to the Dependence Principle, the Black English phonetic representations would be derived from underlying representations containing final r. According to the Independence Principle, however, the Black English pronunciations of bother and smother are derived by a rule which deletes word-final r, where the positing of an underlying r for these words is justified on the basis of the forms bothering and smothering in which [r] is present. Thus, with respect to the feature under discussion, Standard English and Black English would have the same underlying representations. Since father and sister do not have related forms which contain an [r], the only justification for positing an underlying r for these words is the correct but irrelevant observation that they end in r in r-full varieties of Standard English. Thus, the systematic phonemic representations for sister and father differ from those of the /--full varieties of Standard English in that they do not contain an underlying final r, even though if there were an underlying final r it could be deleted at no extra cost by the /--deletion rule which is in the grammar anyway. Example F In Black English there are several dozen monosyllabic and a few polysyllabic words which phonetically contain the low front lax unrounded vowel [a] in the environment of a single voiceless stop. Examples. [nat] [pat] [hat]

not pot hot

[lat] [tap] [fagat]

lot top forgot

In addition, the word knot is pronounced [nat]; and the words heart and part are pronounced [hat] and [pat], respectively. According to the Dependence Principle, an underlying r would be postulated for the systematic phonemic representations of heart and part; heart and part would be marked as exceptions to a general rule for r-less dialects which lengthens the vowel preceding underlying r, and the underlying r would be deleted by a general rule for r-less dialects deleting r preconsonantally after lengthening. According to the Independence Principle, on the other hand, no under-

DIALECTOLOGY IN GENERATIVE GRAMMAR

31

lying r would be postulated for the Black English pronunciations of heart and part, and the underlying representations of heart and part differ from the underlying representations of hot and pot solely in their vowels, i.e. /hat/ heart, /pat/ part vs. Ihaet/ not, /paet/ pot. The postulation of an underlying r in these words can only be justified by recourse to Standard English data, but the consideration of Standard English data is ruled out by the Independence Principle. Example

G

A variety of considerations lead to the conclusion that the underlying representations for the Standard English plural and preterit must be /Vz/ and /Vd/, i.e., that the phonology of these aspects of Standard English inflection must be treated by a deletion rule rather than an epenthesis rule and that these inflectional suffixes end in voiced rather than voiceless consonants in their underlying representations (for justification and discussion cf. Luelsdorff, 1969). The same arguments in support of handling the phonology of the plural and preterit by means of a deletion rule in Standard English may be adduced to support handling the phonology of the plural and preterite by means of a deletion rule in Black English, but the underlying representations of the consonants of these suffixes are different in Black English. The Black English phonetic manifestation of the plural after noun stems ending in [+Coronal, + Strident] consonants is [is] and the phonetic manifestation of the preterit after verb stems ending in [+Coronal, —Continuant] consonants is [it]. According to the Depencence Principle, underlying voiced final consonants would be postulated for Black English and, by appropriately extending the utterance-final devoicing rule discussed in Chapter III, these voiced consonants would devoice. According to the Independence Principle, underlying voiceless consonants would be postulated for the dialect, and Standard English would be related to the dialect by a rule which maps certain underlying voiced consonants onto certain underlying voiceless consonants.

Ill SEGMENT

STRUCTURE

CONDITIONS

0. SUMMARY

In Section 1 I present the articulatory correlates of the distinctive features used in this study. These features are drawn from the universal inventory of phonetic features presented in Chomsky and Halle (1968). In Section 2 I present the distinctive feature matrix for the systematic phonemes of Black English. Here, 12 features are presented as distinctive and 35 segments as systematic phonemic. In Section 3 I present and justify the complete set of segment structure conditions for Black English. The purpose of these conditions is two-fold: they constitute a definition of the set of systematic phonemes of the language and, together with the sequence structure conditions presented in Chapter IV, enable incompletely specified dictionary matrices to select their corresponding completely specified systematic phonemic matrices before the phonological rules begin to apply. In Section 4 I propose a means for determining which subset of the universal set of features is distinctive, apply this proposal in order to justify the postulation of certain features as distinctive for Black English, and pose a problem requiring resolution by subsequent research in phonological theory.

1. THE FEATURES

In this section I present the articulatory correlates of the distinctive features used in this study. In the concluding section of this chapter I attempt to justify the selection of these features by showing that they lead to the maximally economical formulation of certain phonological rules.

Vocalic Vocalic sounds are produced with an oral cavity in which the narrowest constriction does not exceed that of the high vowels [i] and [u] and with the vocal cords positioned so as to allow spontaneous voicing. In producing nonvocalic sounds one or both of these conditions are not satisfied.

SEGMENT STRUCTURE CONDITIONS

33

Consonantal Consonantal sounds are produced with a radical obstruction in the midsagittal section of the vocal tract. Nonconsonantal sounds are produced without such an obstruction. High High sounds are produced by raising the body of the tongue above the level that it occupies in the articulation of the vowel [e] in bed. Nonhigh sounds are produced without such a raising of the body of the tongue. Back Back sounds are produced by retracting the body of the tongue from the position that it occupies in the articulation of [e]. Nonback sounds are produced without such a retraction from this position. Low Low sounds are produced by lowering the body of the tongue below the level that it occupies in the production of [e]. Nonlow sounds are produced without such a lowering of the tongue body. Anterior Anterior sounds are produced with an obstruction located in front of the palatoalveolar region of the mouth. Nonanterior sounds are produced without such an obstruction. Coronal Coronal sounds are produced with the blade of the tongue raised from its neutral position. Noncoronal sounds are produced with the blade of the tongue in its neutral position. Tense Tense sounds are produced with considerable effort by the supraglottal musculature. Nontense sounds are produced with less effort, more rapidly, and less distinctly.

34

SEGMENT STRUCTURE CONDITIONS

Voiced

Voiced sounds are produced with vibrating vocal cords. Nonvoiced sounds are produced with a glottal opening so wide as to prevent vocal cord vibration. Continuant

Continuants are produced with a constriction in the vocal tract which is not so narrow as to prevent the flow of air past the constriction. In the production of noncontinuants the air flow through the mouth is blocked. Nasal

Nasal sounds are produced with a lowered velum which allows air to escape through the nose. In the production of nonnasal sounds the velum is raised so that the air from the lungs escapes only through the mouth. Strident

Strident sounds are acoustically marked by greater noisiness than nonstrident sounds. Obstruent continuants (excepting [0] and [Ò]) are strident, while stops and sonorants are nonstrident. 2. THE SYSTEMATIC PHONEMES In this section I present the distinctive feature matrix for the systematic phonemes of Black English. This matrix consists of 12 rows, each row containing a distinctive feature of the language, and 35 columns, each column corresponding to a systematic TABLE 1 Systematic Phonemes of Black English

ï i ëejëaeûuôoâaywr lpb fvmtds zOôncj â zkgh b+ ++++++++++++++++++++ ++ ++ ++++++ ++++++- + ++ -

Vocalic Consonantal High Back

++++++++++++ ++

Anterior Coronal Tense Voiced Continuant Nasal Strident

+++++++++++++ ++ +++++++++++ +- +- +- +- +- +- +++++- +- - +- +- +- - +- +- +- + ++++++++++++++++- +- ++- +- +- ++- +- +- + +++++++++++++++H ++ b+++ 1- + + + + ++ ++ ++++

SEGMENT STRUCTURE CONDITIONS

35

phoneme. In section 3 the set of systematic phonemes is defined by the complete set of segment structure conditions expressing the constraints on the feature composition of the systematic phonemes. No attempt will be made to justify the postulation of this set of systematic phonemes for Black English except to say that they have been found to be both necessary and sufficient to express the phonological processes presented in this study. 3. SEGMENT STRUCTURE CONDITIONS

In this section I present and justify the complete set of segment structure conditions for Black English. This set of conditions provides a definition of the systematic phonemes of Black English and, together with the set of sequence structure conditions (cf. Chapter IV), fills in all blanks in dictionary matrices before the phonological rules begin to apply. I begin with the observation that all the systematic phonemes in Table 1 that are specified [+Vocalic, —Consonantal] are also specified [—Anterior, —Coronal, -f Voiced, -f Continuant, -Nasal, -Strident]. /(Ci)

+ Vocalic —Consonantal

T(Cx)

—Anterior —Coronal + Voiced + Continuant —Nasal -Strident

No systematic significance is attached to the numerical subscripts accompanying the morpheme structure conditions, since all morpheme structure conditions are unordered. A conceivable alternative to C\ captures the generalization that all segments specified [+Vocalic] are also specified [+Voiced, -f Continuant, —Nasal, —Strident], The current evaluation procedure is based on the number of feature specifications that each morpheme structure condition saves in the dictionary. Since Ci saves 6 feature specifications for each true vowel in the dictionary, whereas the alternative proposal saves 4 feature specifications for each [+Vocalic] segment in the dictionary, it would seem that Ci is to be preferred to the proposed alternative. Next observe that all segments in Table 1 that are specified [+High] are also specified [—Low] and that all segments that are specified [+Low] are also specified [—High], It is clearly impossible to incorporate both these generalizations in the form of segment structure conditions in the grammar of BE without the feature theory becoming ternary. We must therefore choose one or the other segment structure condition, but not both. Since the former condition enables us to save feature specifications 3*

36

SEGMENT STRUCTURE CONDITIONS

for 12 segments, wnereas the latter condition enables us to save feature specifications for only 5 segments, we tentatively opt in favor of the former condition, given here as C 2 . /(C2) [+High] \ T(C 2) [-Low] Since the determination of the true generalizations for the language depends upon the evaluation procedure, and the evaluation procedure depends upon the availability of the dictionary, the determination of the true generalizations depends upon the availability of the dictionary. Since no dictionary is currently available, however, the proposed morpheme structure conditons are best regarded as highly tentative. In regard to Ci, notice that since there are no languages which contain true vowels that are either [—Continuant] or [ +Strident], and apparently no languages with systematic phonemic nasal vowels, the relevant sub-part of Ci should be elevated to the status of a linguistic universal. C 2 also has the status of a linguistic unversal, only for a different reason. While it is theoretically possible for some natural language to have systematic phonemic [ + Vocalic, + Nasal] segments, although no such languages have been found, it is theoretically impossible for segments to be specified [+High, •f Low] because it is physiologically impossible for a segment to be [+High] and I + Low] at the same time. This and similar redundancies point to a serious defect in the present theory, namely, that some of the features are not independent of one another. Notice next that all segments specified [ — Vocalic, — Consonantal] are also specified '[—Anterior, —Coronal, + Tense, +Continuant, —Nasal, — Strident] and that High glides are Voiced and that h is nonBack, Low, and nonVoiced. These three generalizations are given in segment structure conditions C 3 , C4, and C5 respectively. (C 3 )

T(C 3 )

—Vocalic ] —Consonantal] I —Anterior —Coronal +Tense -I- Continuant —Nasal -Strident

—Vocalic —Consonantal + High I r ( C 4 ) [ +Voiced ]

I(C*)

—Vocalic —Consonantal -High \ T(C 5 ) - B a c k + Low —Voiced

/(C.)

The only feature distinguishing the liquids r and / is Anterior for which r is specified '—' and I is specified ' + ' . The specifications for all the remaining features are predictable. This generalization for liquids is given as segment structure condition 6.

SEGMENT STRUCTURE CONDITIONS

/(Ce)

T(Ce)

37

+Vocalic "1 +Consonantal J I -High -Back —Low +Coronal + Tense +Voiced + Continuant —Nasal -Strident

At this point all the possible true generalizations have been given in the form of segment structure conditions for the true vowels, glides, and liquids in Table 1. The remaining segment structure conditions apply exclusively to the true consonants. Note first that all true consonants that are specified [+Voiced] and [—Voiced] are also specified [—Tense] and [-f Tense], respectively, and that all true consonants that are specified [+Tense] and [—Tense] are also specified [—Voiced] and [+Voiced], respectively. There is thus a choice between a segment structure condition which has Voiced as its antecedent term and Tense as its consequent term and a segment structure condition which has Tense as its antecedent term and Voiced as its consequent term. For the time being I have no principled basis for selecting one of these conditions and rejecting the other, so I will arbitrarily choose the former. /(C 7 )

T(C7)

["-Vocalic + Consonantal aVoiced I [-aTense ]

An analogous, but decidable, situation obtains in the case of the features High and Anterior in the true consonants. Here, all true consonants that are specified [—High] and [+High] are also specified [+Anterior] and [—Anterior], respectively, and the true consonants that are specified [+Anterior] and [—Anterior] are also specified [—High] and [+High], respectively. There is thus a choice between a segment structure condition with the feature High in its antecedent and the feature Anterior in its consequence and a segment structure condition with the feature Anterior in its antecedent and the feature High in its consequence. I decide on the former segment structure condition, since full specifications for the feature High are required for the

38

SEGMENT STRUCTURE CONDITIONS

proper operation of C 2 . I(Ca)

f-Vocalic +Consonantal aHigh \ T (C8) [ —aAnterior ] Notice next that all true consonants that are specified [+Back] are also specified [—Coronal, —Continuant, —Nasal, —Strident] and that all High nonBack consonants are specified [+Coronal, —Nasal, + Strident], These two generalizations are given as segment structure conditions 9 and 10. /(C.)

T(C9)

—Vocalic + Consonantal + Back

/(CM)

\ —Coronal —Continuant —Nasal -Strident

—Vocalic + Consonantal + High .-Back i

T (C10)

+ Coronal —Nasal + Strident

Note next that all nonHigh true consonants are also nonBack and nonLow. This constraint is expressed in segment structure condition 11. I (Cil) r -Vocalic +Consonantal .-High I 1 r ( C n ) ["-Back [-Back [ --Low Low

J

Observe further that all true consonants that are specified [—High, —Coronal, —Nasal] agree in stridency with their specifications for continuance. I give this constraint as segment structure condition 12. I(Ci 2 ) [-Vocalic +Consonantal -High —Coronal —Nasal aContinuant i

T (C12) [aStrident

]

SEGMENT STRUCTURE CONDITIONS

39

The final segment structure condition expresses a constraint on the feature composition of nasals. If a true consonant is specified [+Nasal], then it will also be specified [+• Voiced, —Continuant, —Strident], This constraint is given as segment structure condition 13. I(Ciz)

T(C

13)

—Vocalic +Consonantal Nasal I +Voiced —Continuant -Strident

4. JUSTIFICATION OF THE FEATURES

In this section I attempt to justify the selection of the distinctive features postulated for Black English by showing that these features lead to the maximally economical formulation of certain Black English phonological rules. Vocalic and Consonantal The Black English indefinite article a(n) has three different phonetic manifestations: [an], [a?], and [a], [an] occurs in the environment of a following morpheme boundary followed by a vowel: [an3va] another, [a?] occurs in the envronment of a following word boundary followed by a true vowel: [a?aepa] an apple, [a] occurs in the environment of a following word boundary followed by either a glide, or a liquid, or a true consonant: [awos] a wasp, [alaa:] a line, [atuf] a tooth. I account for these facts by postulating the underlying representation an for the indefinite article and by rewriting underlying n as [ ?] in the environment of a following word boundary followed by a true vowel and as 0 in the environment of a following word boundary not followed by a true vowel. Thus: ? / #a_ 0/ #a

#

[+Vocalic, —Consonantal] [ +Vocalic, —Consonantal]

Although there are aspects of this rule which require justification, such arguments are irrelevant to the present discussion. What is important to notice is that the features Vocalic and Consonantal are required by the environmental statement in the rule. Since no simpler environmental statement can be written, I offer this rule as evidence in support of the claim that the features Vocalic and Consonantal are necessary at the systematic phonemic level in the phonology of Black English.

40

SEGMENT STRUCTURE CONDITIONS

Coronal and Strident Nouns in Black English which end in [+Coronal, + Strident] segments phonetically realize the plural grammatical category as [is]. Nouns which do not end in segments which are specified both [+Coronal] and [+ Strident] phonetically realize the plural grammatical category as [s] if the preceding segment is specified [ — Voiced] and as [z] if the preceding segment is specified [+Voiced], irrelevant details omitted. I account for the phonology of pluralization in Black English by postulating a rule which rewrites the grammatical formative PLURAL as the underlying systematic phonemic representation is and a rule which deletes the i in is in the environment of a preceding morpheme boundary which is itself preceded by a segment which is not specified both [+Coronal] and [-Strident]. Thus, Plural — is i -*• 0 I -[+Coronal, + Strident] + This process is much more general than it has been presented here and, again, irrelevant details have been omitted. What is important to notice, however, is that the features Coronal and Strident are required by the environmental statement in the rule. Since there is no single feature nor any other combination of two features which are capable of performing the same task in the rules of the phonology of pluralization, I conclude that the features Coronal and Strident, like the features Vocalic and Consonantal, are necessary at the systematic phonemic level in the phonology of Black English. High, Back, and Tense In the phonology of Black English there is a rule which, irrelevant details omitted, phonetically realizes word-final underlying / as a Back vowel which agrees in Highness and Tenseness with the preceding vowel. Examples: [nidu] needle, [pfku] pickle. [tdby] table, [ksdy] kettle. I account for these facts by a rule which rewrites underlying / in polysyllabic words, as a Back vowel which agrees in Highness and Tenseness with the preceding vowel. Thus, I - [ + Back, aHigh, pTense] / / [aHigh, [iTense] C 0

#

Again, irrelevant details have been omitted from the statement of this rule. Aside from the fact that this vowel harmony justifies the postulation of an underlying /, it is essential to notice that the necessity of this rule also justifies the postulation of the features High, Back, and Tense as distinctive in the phonology of Black English.

41

SEGMENT STRUCTURE CONDITIONS

Voiced and Continuant In Black English the stressed nontense vowels i and e and the stressed tense vowel a break (i.e. become centering diphthongs) in the environment of word-final voiced consonants. In addition, stressed tense a breaks in the environment of a word-final voiceless continuant. Thus, i,e,ce

[—Vocalic, + Consonantal, + Voiced]*

0 - 3 œ

[—Vocalic, + Consonantal, + Continuant] #

In addition to providing additional support for the distinctiveness of features Vocalic and Consonantal, the breaking rule supports the claim that the features Voiced and Continuant are required at the systematic phonemic level of representation. Low In Black English there are several dozen monosyllabic words containing the systematic phoneme a followed by a voiced true consonant. Such words include cub, sub, bud, bug, buzz, bun, etc. If the voiced true consonant following a is specified [ — Nasal], then a is phonetically realized as the unrounded counterpart of [o], i.e. as a mean-mid back unrounded tense vowel. Thus, -f Back + Low —Tense

—Low + Tense

/ #Co —

—Vocalic + Consonantal —Nasal

#

In addition to providing additional support for the claim that Backness and Tenseness are distinctive features, the necessity of this rule supports the claim that Lowness is a distinctive feature as well, for there is no more economical way of isolating the segment a than by means of the features Back, Low, and Tense. Nasal In Black English there is a phonological rule which nasalizes vowels in the environment of a following nasal consonant. In addition, there is a subsequent phonological rule which deletes nasals in word-final position preceding a word beginning with a nonvowel 1 and in utterance-final position. Thus,

{

ml « r

^

| ' 1 — # i s

[ f Vocalic, — Consonantal] 1 i

The negation sign - is used here and elsewhere to capture the generalization that the converse of the conjunction of the features may be expanded in an arbitrary order. 1

42

SEGMENT STRUCTURE CONDITIONS

where S is not dominated by another S. Neither of these rules can be written more economically than by mentioning the feature Nasal, and I offer this as evidence in support of the claim that the feature Nasal is needed at the systematic phonemic level of representation. Anterior In order that the above vowel-harmony rule apply only to systematic phonemic I and not to systematic phonemic r, a set of features is needed which distinguishes / from r. Such a set of features is available, consisting of the single member Anterior, and I offer this as evidence that the feature Anterior is required at the systematic phonemic level of representation. The above discussion is an attempt to justify the selection of a subset of universal features to be used in the systematic phonemic matrices representing the morphemes of Black English. I have argued that if a set of features can be shown to be needed in the statement of a Black English phonological rule by showing that the selection of this set results in the maximally simple statement of the rule, then the features in the set are required at the systematic phonemic level of representation. Implicit in this discussion is an evaluation metric which, other things being equal, values a rule formulation with the fewest features. The question now arises if there are any features not given in Table 1 which would result in the more economical statement of some phonological rule in Black English than the statement possible using only the features in Table 1. In Black English there is a phonological rule which devoices utterance final voiced true consonants which are nonnasal. This rule also applies to nonnasal consonants in clusters. Examples [biak] [biat] [bwp] [liaft] [jv:2t]

big bid bib lived judged

By using only the features presented in Table 1,1 arrive at the following formulation of this phonological rule: [+Voiced] - [-Voiced] / —Vocalic + Consonantal # ]s —Nasal i where S is not dominated by S. This rule requires mentioning three features in its environmental statement: Vocalic, Consonantal, and Nasal.

SEGMENT STRUCTURE CONDITIONS

43

Now there is a more economical way of stating this rule by mentioning the feature Obstruent, a feature not included in Table 1. Thus: [+Voiced] - [-Voiced]/

1 + Obstruent h 11

*h

By using the feature Obstruent, this rule requires mentioning only one feature in its environmental statement. The evaluation metric thus leads us to prefering the second formulation to the first, and the feature Obstruent should therefore be included among the inventory of features at the systematic phonemic level. This entails making the feature Nasal completely redundant, since Nasals are the only segments specified [ -Obstruent, —Continuant]. This would be an acceptable consequence if it could be shown that the feature Nasal were not needed at the systematic phonemic level of representation. But this has in fact been shown to be not true. We are thus faced with the following dilemma: both the feature Nasal and the feature Obstruent are required at the systematic phonemic level of representation. But by including both features at the systematic phonemic level of representation, either all the specifications for the feature Nasal are predictable or all the feature specifications for the feature Obstruent are predictable, but not both. If all the specifications for a feature are predictable, then there is no justification for claiming that the feature is distinctive. Given the present theory of generative phonology, I see no way in which this dilemma can be resolved.

IV SEQUENCE STRUCTURE

CONDITIONS

0. SUMMARY

In Section 1 I present and justify the complete inventory of Black English systematic phonemic consonant clusters in morpheme initial position. In Section 2 I give the complete set of sequence structure conditions for the morpheme-initial clusters. These conditions constitute a definition of the set of possible Black English morphemeinitial consonant clusters. In Section 3 I present and justify the complete inventory of Black English systematic phonemic consonant clusters in morpheme-final position. Section 4 contains the complete set of sequence structure conditons for the morphemefinal clusters. These conditions define the set of possible Black English morphemefinal consonant clusters.

1. INITIAL CONSONANT CLUSTERS: INVENTORY

Table 2 properly contains a complete list of Black English morpheme-initial phonetic consonant clusters, where 'consonant cluster' is defined as a sequence of two or more nonvowels. Note that the Black English columns labeled 'Casual' and 'Careful' contain identical phonetic representations. This means that there is no intrapersonal variation in the speech of the principal informant in the domain of morpheme-initial consonant clusters. Contrast this with the data presented in Table 4 where there is a great deal of intrapersonal variation in the speech of the principal informant in the domain of morpheme-final consonant clusters. The same is true of word-initial and word-final single consonants. The general observation is that there are significant quantitative and qualitative differences between pretonic and posttonic consonantism. Although this is a non-trivial observation, for the time being I can offer no hypothesis which would explain this state of affairs. Consider Table 2, numbers 3, 25, and 32. Here we find that a proper subset of the words which are pronounced with morpheme-initial consonant clusters in Standard English are pronounced with morpheme-initial consonant clusters in Black English.

SEQUENCE STRUCTURE CONDITIONS

45

TABLE 2 Morpheme-Initial Phonetic Consonant Standard English

1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21.

br pr tr fr gr dr kr Or sr St sp sm sk sn sf si Pi kl bl fl gl

Black English Casual Careful br pr tr, tV fr gr dr kr Bar, V sr st sp sm sk sn (-sf-) si Pi kl bl fl gl

br pr tr, tV fr gr dr kr Ôsr, 0V sr st sp sm sk sn (-sf-) si Pi kl bl fl gl

Clusters

Standard English

22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33. 34. 35. 36. 37. 38. 39. 40. 41.

Black English Casual Careful

dw kw gw tw sw hw ~ w 8w fy ky my by py vy

dV kw gw tw, tV sw w

dV kw gw tw, tV sw w

—

—

fy ky my by, bV py vy

Ç str skr spr spi skw sky

Ç str skr spr spi skw

fy ky my by, bV py vy Ç str skr spr spi skw

—

—

Examples : Standard English trunk (suitcase) trunk (elephant's) twelve toward bugle bureau

[trarjk] [traqk] [twelv] [tword] [byugl] [byaro]

Black English [trerjk] [tegk] [tweyv] [tost] [byugu] [blre]

What this means, unfortunately, is that every English word must be elicited from Black English speakers if it is to be correctly entered in the lexicon. In this particular case, those words containing phonetic clusters are entered in the Black English lexicon with underlying systematic phonemic clusters, and those words containing single consonants are entered in the lexicon with underlying single consonants. Next, consider Table 2, numbers 8 and 22. Here we find that corresponding to Standard English [6r] and [dw] are Black English [Oar, 6V] and [dV], respectively.

46

SEQUENCE STRUCTURE CONDITIONS

Examples: Standard English three through dwarf

[Ori] [Öru] [dworf]

Black English [ôari]

m

[doaf]

In general, there are no morpheme-initial clusters in [0r] or [dw] in the speech of the principal informant. Words like Dwight and dwindle were not in his vocabulary. Consider Table 2, numbers 28 and 41. [0w] and [sky] are not given as initial clusters because thwart, thwack, skew, skewer were not in the vocabulary of the principal informant. Even if they had been, there is the possibility that they would not have contained the clusters [0w] and [sky]. As linguists are repeatedly finding out, Black English differs from Standard English in many unexpected and sometimes quite subtle ways. Similarly, [kn] and [gn] are not included in the inventory of morpheme-initial clusters because of the four words know, acknowledge, gnostic and agnostic, only know was in the principal informant's vocabulary. All of the Black English clusters given in Table 2 are postulated as systematic phonemic clusters with no other justification than that they are phonetically present. Phonetic [9] is treated as actualizing underlying hy. I posit systematic phonemic clusters containing y on the grounds that clusters containing [y] are not invariably followed by [u] ([pysr] pure, [kyar] cure) and there is insufficient data for postulating the Vowel Shift Rule for Black English (cf. Chapter VII). The inventory of morpheme-initial systematic phonemic clusters is given in Table 3.

2. INITIAL CONSONANT CLUSTERS: SEQUENCE STRUCTURE CONDITIONS

Consider first the class of trisegmental clusters consisting of spr, spl, str, skr, and skw. The generalization that the longest morpheme initial consonant clusters consist of three segments may be expressed by a sequence structure condition stating that if a morpheme begins with two segments specified [+Consonantal] and is followed by a segment which is specified either [+Consonantal] or [—Vocalic], then the fourth segment is specified [+Vocalic, —Consonantal], i.e. is a true vowel. Next observe that morpheme-initial clusters consisting of three segments contain as their third segment either a liquid or a glide. Although there are no trisegmental initial clusters containing the glide y as their third member, such clusters must not be excluded from the domain of POSSIBLE morpheme-initial clusters on the grounds that there is a trisegmental cluster containing w as its third member, and w and y belong to the same natural class. This generalization may be captured by a morpheme structure condition stating that if a morpheme begins with two segments specified [+Consonantal] and the third segment is specified either [+Consonantal] or [—Vocalic], then the third segment is also specified either [+Vocalic] or [—Consonantal], respectively. In order

47

SEQUENCE STRUCTURE CONDITIONS

that this condition admit clusters containing w and y and reject custers containing h, it must further stipulate that the glide be specified [+High]. Notice next that the first two segments of trisegmental clusters are s and a voiceless stop, respectively. This generalization may be expressed by a sequence structure condition stating that if a morpheme begins with two segments specified [+Consonantal] which are followed by a segment specified either [+Consonantal] or [—Vocalic], then the first two segments are specified [—Sonorant]. It would be a mistake, I think, to allow the consequent of this condition to contain the full complement of features for s and the voiceless stops, since these features may be contained in the consequent of a more general condition to the effect that if a morpheme begins with two segments specified [ — Sonorant], then the first segment is s and the second segment a voiceless stop, sf, of course, excepted. We may collapse all of the above sequence structure conditions by allowing the feature [—Sonorant] to appear in the first two setments of the antecedent and by enclosing the third segment in the antecedent in parentheses. I give this as morpheme structure condition 14: ([+Consonantal]) / (Cu) + [-Sonorant

] [-Sonorant

] ({ ([-Vocalic] J

T (Cu)

})[

>

]

-Vocalic If-Vocalic + Consonantal + Consonantal -High -Low ([+Vocalic] > —Back + Tense ({ }) [+Vocalic —Low —Voiced (["—Consonantal ) —Consonantal +Anterior —Continuant [ + High +Coronal —Nasal + Tense [—Strident —Voiced + Continuant —Nasal + Strident

Notice, by the way, that it would be a mistake to include the features [—High, —Back, —Low, + Coronal, + Tense, + Voiced, -(-Continuant, —Nasal, —Strident] in the consequent for the liquids in column 3 of Cu, the features [—Anterior, —Coronal, + Tense, + Continuant, - L o w , —Nasal, + Voiced, —Strident] in the consequent for the glides in column 3 of Clt, and the features [—Anterior, —Coronal, + Voiced, + Continuant, —Nasal, —Strident] in the consequent for the vowels in column 4 of C14, since these matrices are already accepted by the grammar through Ci, C 2 , C 3 , C 4 , and C 6 (cf. Chapter III, Section 3). Such a procedure is rejected by an evaluation metric which maps generality inversely into simplicity in terms of numbers of features.

48

SEQUENCE STRUCTURE CONDITIONS

C14 captures all the true generalizations about clusters 1-5 and 18-21 in Table 3, with the sole proviso that the morpheme sphere be marked [— Cu ] in the lexicon. TABLE 3 Morpheme-Initial Phonemic Consonant Clusters Cluster 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33. 34. 35. 36.

Examples spr spl str skr skw pr br tr dr kr gr fr Pi bl kl gl

fl

sp st sk sf sm sn Sr si kw gw tw sw fy ky my by py vy hy

spray, spread split, splash straw, string screw, scratch square, squeak pray, proud brown, bread tray, tree drip, dry crowd, crawl gray, grass free, fruit play, please blue, black clue, claw glue, glass flower, fly spend, spoil stay, stand scare, score hemisphere small, smart sneeze, snow shrimp, shredded slow, sleep quick, quiet Gwen twine, twist swim, swell few, fuse cue, cure mule, music bugle, beauty pupil, puny view human, huge

The problem of the predictability of voice in obstruents is discussed later in this chapter. Consider the remaining clusters in Table 3, i.e., all bisegmental clusters other than sp, st, sk, and sf. In each case, excepting my and hy, the first segment is specified [—Sonorant], and the second is specified [+Sonorant]. In order to distinguish such

49

SEQUENCE STRUCTURE CONDITIONS

clusters from morpheme-initial obstruent + vowel sequences, the antecedent in the sequence-structure condition relating to the second segment must contain, in addition to the feature [-f-Sonorant], either the feature specification [+Consonantal] (for liquids), or the feature [—Vocalic] (for glides), or both the features [+Consonantal, —Vocalic] (for nasals). The first generalization, then, is that if a morpheme begins with a segment specified [+Consonantal] which is followed by a segment specified [+Sonorant] and is a nonvowel, i.e. is also specified, adopting Zwicky's notation for expressing the complement of sets, {[+Vocalic, —Consonantal]} ( = Z), then the third segment is specified [+Vocalic, —Consonantal]. A second generalization is that if a morpheme begins with a segment specified [+Consonantal] which is followed by a segment which is specified both [ + Sonorant] and [—Vocalic], then the second segment is specified both [—Consonantal] and [+High]. Next, if a morpheme begins with a segment which is specified [+Consonantal] and is followed by a segment which is specified [+Sonorant] and both [+Consonantal] and [—Vocalic], then the second segment is also specified [+Nasal]. Finally, if the first segment is specified [+Strident], then it is also specified [+Continuant]. All the above generalizations are captured by sequence structure condition 15. I (Cn)

+ Consonantal] (+Strident) J

< (—Vocalic

T (Cu)

[

+ Sonorant ) -(-Vocalic "1 —Consonantal J i

[+Long] / # C0

+Consonantal +Voiced

bid: bid

big: big

#

Some sample derivations: bib: bib 1. biab 2. bw:b 3. bw:p 5*

(Pt) (PIS) (Pu)

1. biad 2. bw:d

3. bw:t

(Pt) (Pia) (P12)

1. bwg 2. bia:g 3. bw.k

(PT) (Pu) (Pu)

68

PHONOLOGY OF THE WORD

7. CONSONANT WEAKENING

As mentioned in Chapter I, underlying d and g in monosyllables in utterance final position in casual style either devoice (cf. Pi 2 ) or are deleted. Moreover, underlying t and k in monosyllables in utterance final position in casual style are actualized as glottal stop. Examples: bid big

[bra :t] [bra :k]

[bra :] [bra:]

[pi?] [pi?]

pit

pick

These generalizations are expressed by P14 and P15, respectively.