Creole Phonology

382 158 14MB

English Pages 0 [197] Year 1981

Report DMCA / Copyright

DOWNLOAD FILE

Polecaj historie

Creole Phonology

Citation preview

Creole Phonology

JANUA LINGUARUM Studia Memoriae Nicolai van Wijk Dedicata edenda curat

C. H. van Schooneveld Indiana University

Series Practica 117

Creole Phonology by

Henri Tinelli

Mouton Publishers The Hague · Paris · New York

Library of Congress Cataloging in Publication Data Tinelli, Henry, 1939Creole phonology. (Janua linguarum: Series practica; 117) Bibliography: p. 1. Creole dialects-Phonology. 2. Creole dialects, French-Haiti-Phonology. I. Title. II. Series. 80-27310 PM7831.T5 417'.2 ISBN 90-279-3048-1

ISBN: 9 0 - 2 7 9 - 3 0 4 8 - 1 © 1981, Mouton Publishers, The Hague, The Netherlands Printed in The Netherlands

Contents Introduction

1

Abbreviations and Symbols

3

Part I

Phonetics

Chapter One

Nasality

Chapter Chapter Chapter Chapter

Mid Vowels r Palatals, Implosives, h Distinctive Features

Part II Chapter Chapter Chapter Chapter Chapter Chapter Chapter Chapter Chapter Chapter Chapter Chapter Chapter Chapter Chapter Chapter Chapter Part III Chapter Chapter Chapter Chapter

Two Three Four Five Underlying Six Seven Eight Nine Ten Eleven Twelve Thirteen Fourteen Fifteen Sixteen Seventeen Eighteen Nineteen Twenty Twenty-One Twenty-Two

Phonology Introduction Systematic Phonemes Natural Sets Segment Redundancy Rules Redundant Segment Definitions Sequence Redundancy Rules The Archiphoneme in Generative Phonology . . . . Lexical Nasality Vowel Sequences Glides Liquids Gemination Initial Clusters Final Clusters Medial Clusters Three-consonant Clusters /h/

Transformational Twenty-Three Twenty-Four Twenty-Five Twenty-Six

7 11 13 14 15

21 23 24 25 31 33 34 40 44 45 47 49 50 52 54 56 57

Phonology

Introduction Vowel Nasalization li-la Nasalization Agglutination

61 63 66 69

Chapter Chapter Chapter Chapter Chapter Chapter Chapter Chapter Part IV Chapter Chapter Chapter Chapter Chapter Chapter Chapter Chapter Chapter Chapter Chapter Chapter Chapter Chapter

Twenty-Seven Twenty-Eight Twenty-Nine Thirty Thirty-One Thirty-Two Thirty-Three Thirty-Four

Splicing Elision and Clustering Vowel Harmony Syllabic Consonants Particle Elision Syllable Definition Other Transformations Rule Ordering

76 78 84 88 92 97 103 108

Introduction Lenition Degauling Glides Syllabic Restructuration Nasalization Deneutralization Palatalization Prenasalization Rounding Cluster Reduction r Laxing and Deletion Minor Creolization Processes Conclusion

119 120 122 126 130 137 148 150 159 165 167 172 174 182

Creolization Thirty-Five Thirty-Six Thirty-Seven Thirty-Ei^it Thirty-Nine Forty Forty-One Forty-Two Forty-Three Forty-Four Forty-Five Forty-Six Forty-Seven Forty-Eight

Notes Bibliography

184 186

Introduction

Creolization is still a relatively undefined term for an historical process which creolists consider of significant interest for the general theory of language. Whereas the events of colonization and slavery in the Caribbean and North America are well documented, the various hypotheses which have been offered to explain the genesis and evolution of creole languages are still in active competition, mostly because they lack the support of substantial evidence. This volume is the result of an attempt to unearth such documentation through a systematic study of creole phonology and morphophonemics. It presents synchronic and diachronic analyses founded on raw data from Haitian Creole and on the limited amount of published materials currently available for the non-French creóles of the Atlantic and the Caribbean. Parts One, Two, and Three examine the phonetic characteristics and present a comprehensive and unified treatment of the underlying and transformational phonology of Haitian Creole. Part Three extends the analysis to English, Spanish, and Portuguese based creóles, with particular attention to a number of sub-systems which are of general interest for creole linguistics. Detailed discussion and progressively more inclusive generalizations about the Atlantic creóles will lead to an analysis of phonological creolization, which is treated in Part Four. Creole studies can be related to three prevalent conceptions of Indo-European based creóles. The "Europeanists" emphasize the role of the base language and find their illustrations mostly in the lexicon, the etymology of grammatical markers or vocabular items, and to a lesser degree in phonology. French creóles are thus treated as Romance languages, as dialects of French or, in a naive approach, as degraded French. The "Africanists" minimize the importance of the lexicon and emphasize syntax and morphology, particularly the aspect, tense, and determiner systems. They tend to make references to the striking structural similarities which are found in creóles with different bases, although the European standpoint can be upheld even in this domain if a monogenetic theory of creolization is adopted. The third conception, perhaps the most widely held in modern creole studies, is that of creóles as Janus-like or hybrid languages with a European lexicon and phonology engrafted upon an African syntax.

2

CREOLE PHONOLOGY

A close analysis of Haitian phonology shows that, if the first approach is adopted, Haitian and the other French creóles could indeed be called Romance languages, just as a systematic analysis of the phonologies of English creóles might show them to be Germanic languages. However, I will also try to show that even in their phonologies Atlantic creóles have never drifted away from West African patterns. As for the third conception, a detailed phonetic and morphophonemic investigation reveals that it is basically correct but simplistic. Both European and African features and systemic shifts have played their roles in the elaboration of creole phonologies. One of the most fundamental aspects of the African imput was a process of amplification which I will call "structural resonance." It affected, in the European language, the sub-systems which were unstable and shifting at the time of the slave trade. That movement was amplified and accelerated if some similar structure or tendency was present in some of the languages spoken by the Africans. Otherwise, it was stopped or dampened. On the same grounds, structures which presented similar features in the base language and the African "substratum" were not merely carried over into the new pidgin but were reinforced and extended to other sub-systems. Thus can be explained the fact that a creole always seems to outdistance its European base in those "resonating" sub-systems and structural shifts. This will be illustrated with data from English, French, and Portuguese creóles. It also provides reasons other than a poorly documented relexification process for ValkhofPs remark that "even when we compare French Creole with Portuguese Creole, two Romance languages, we discover that their creole features are much more outstanding than their common Romance ones" (1966:3). The author gratefully acknowledges a grant-in-aid from the University of Colorado Council on Research and Creative Work.

Abbreviations and Symbols

PHONETIC SYMBOLS Phonetic transcriptions follow the symbolization o f the International Phonetic Association, with a few modifications: s replaces I.P.A. / ζ is the voiced homorganic o f s. ts and dz are the corresponding affricates.

Underlying forms are between slanted lines: / / Surface (phonetic) and intermediate forms are italicized. When the distinction is not clear from the context, square brackets (f ] ) are used. RULES RR

Redundancy rules (Part II).

TR

(Phonological) transformations (Part III).

DR +

Morpheme boundary.

Diaphonie rules (Part IV).

%

Syllable boundary.

# ##

Word Boundary.

α

+ or -

ß

- or +

Higher unit boundary.

LANGUAGES C.

:

Cameroons Pidgin.

Dom. :

Dominica Creole.

E

:

English

Fr.

:

G.

French. Guiana English Creole

4

CREOLE PHONOLOGY

Gu.

Gullah.

HC

Haitian

JC

Jamaican Creole

Κ.

Krio.

Kr.

Kriôl.

Mart. Pap.

Martiniquais. Papiamentu.

Port.

Portuguese.

S.

Saramaccan.

Sp. Sr.

Spanish.

Tr.

Trinidadian.

Sranan.

PARTI

Phonetics

C H A P T E R ONE

Nasality

Haitian Creole has five nasal vowels: Γ (e.g., sigwe 'doze'), ε (me (nil'

in'), 3 (f5n

'here is'), H"

'melt'), u ( ü f j 'oungan's temple'), ε and 5 are slightly higher

than the corresponding French segments, especially after dental consonants e.g., nym 'man', tjmat

'tomato'. Yet they are always distinctly lower than the

oral closed e and o. When a closed e or o is nasalized by a contiguous segment, it is automatically lowered. This rule modifies new loanwords when they are nativized, but it does not apply synchronically, i.e., it is not included in the transformational component

o f the phonology. A segment redundancy rule

( R R 14.7, Part II, chapter 9 ) specifies that nasal vowels which are [ + mid ] in the lexicon are redundantly [ — high ] . Nasalized a, e. and o occur only as superficial segments. They always result from the operation of a phonological transformation across morpheme boundaries—see T R 1 (Part III, chapter 24). Nasal a from underlying /Λ/ (which I will usually transcribe /a/) is substantially higher than its oral counterpart. Hall (1953:17) notes it a, but in the dialect o f Port-au-Prince nasal a is different from a nasalized French a. Both are [ — acute ] , i.e., non front, but French a is central whereas Hn. /a/ is back and low mid. Hn. /a/ is also different from Hn. j in that it is not rounded. Its timbre is thus closer to that of Λ.1 Γ and ü are absent from the phonology o f contemporary French, but they are underlying segments in Haitian Creole, although their lexical visibility has been steadily decreasing through the loss of numerous African morphemes. Goodman (1964:50) calls them "marginal phonemes," and words such as pTga 'be sure not to', sigwe 'doze', witga (or uga) 'voodoo priest', vudü 'voodoo', firn

'movie

have

been

represented

by

several

analysts

with

an

under-

lying /i/ or /u/ followed by a nasal consonant (e.g., /piNga/, /wuNgaN/, /vuNduN/, /fim/). The nasal transformation ( T R

1) could easily be modified to

process isolated morphemes, and the diacritic nasal consonant Ν could then be deleted in most cases by an independently required transformation ( T R 19, Part III, chapter 33). But most o f these morphemes do not display a conditioned alternation with oral i and u, and there is no compelling reason to consider Γ and u foreign to the deep structure o f the language, even though these t w o segments are no longer found in French. Indeed, were it not for the small number of African terms left in modern French creóles, we might say that these

8

PHONETICS

dialects have a strong tendency to prenasalize their consonants after nasal vowels (see Part IV, chapter 43). Whatever may be decided about the surface phonetic shape of the morphemes mentioned above, their real systematic structure is obviously a residual pattern which has been preserved in the few African lexemes left in today's Creole. Papiamentu provides a clear illustration of this creole process, as in the following forms (from Lenz 1928:82,203,253). Pap. kuminda, kumida kaminda nertgí fmrruTìga

P., Sp. Sp. P., Sp. P., (O. Sp.)

comida camino negar formiga

The progressive nasalization of the vowel (¡comida -*• komTda) was followed by the prenasalization of the following consonant ( k o m p d a ) , and the creation of a nasal consonant (komtnda). In numerous other cases, it looks as if the nasalization of a consonant took place over a vowel without affecting it. Pap. nanifi mansa priminti

Sp. nariz P. massa, Sp. masa Sp. prometer

We will see that this does not happen in French creóles (Part II, chapter 13), but that a consonant can be nasalized, or a nasal consonant created, after a vowel which has itself been nasalized. I suspect that the same process occurred in Papiamentu, although it is concealed in Lenz's data by the fact that he chose to note vowel nasality in only a very few forms. Unfortunately, he usually did not mark non-phonemic features, except when he wanted to discuss a phonetic point: Se puede decir que en papiamento toda vocal seguida de nasal se pronuncia más o menos nasal ella misma, de modo que en la escritura corriente no hay necesidad de indicar esta nasalización. En los cuentos de Sillie [his informant] lo hago sólo cuando el hecho me llamó la atención, me cayó al oído. A menudo consonantes nasales se cambian en nasales, cuando hay otra nasal cercana: nanishi = c. [Castillian Spanish] nariz, lamantá = c. levantar (1928: 82). I propose for these forms an evolution in two stages, with the following nasalization of the vowel followed by that of the following consonant', as in komida > Pap. kumida ~ kuminda.

NASALITY

9

Lenz leaves unexplained a large number of other examples of consonant nasalization or creation of new NC's: Pap .trinsjona brongosá ñama prentende etc.

Sp.

traicionar avergonzar llamar Sp. pretender

Here again, a two-stage nasalization has taken place. This time vowel nasalization was not progressive but regressive, and it occurred across a consonant; i.e., it is an example of nasal harmonization. As in French mut3, kalsj^· Hn. mutj, kasJ 'sheep, shorts', I would propose a derivation of the type jamar -*• *jama -*• ñama 'to call'. Similarly, we can posit intermediate forms with a nasal vowel for trihsjona, brdt}gosa, pretende, mendra (medrar < Sp. mejorar), primîhti, mansa, nanifî, etc.. A few of these vowels were noted as nasal by Lenz. The others are probably still nasal in Papiamentu, unless its decreolization has hispanized them back into oral vowels. In other cases, the prenasalization of a consonant cannot be explained by an assimilation, and may be due to the African "substratum": Pap. angúa, hunga, pampel [> ar¡gúa, hîir\ga, pàmpel], Sp. aguja, jugar, papel 'needle, play, paper'. Progressive and regressive nasalization of consonants is common in the languages spoken in the slave trade areas of West Africa. In Fanti, /ρ, t, k, kw/ have variants with nasalized aspiration before underlying nasal vowels and before vowels which have been nasalized by rule (cf. Haitianbuk~anε,from an etymon bukane or bukane), and / h , hw, w/ are nasalized throughout in the same contexts. In addition, / w / has a very short / n / onset (Weimers 1946:12-13). Wolof and Peul have prenasalized consonants which are simple phonological units (even in absolute initial position), and so does Portuguese-based Kriôl (Senegal), in such words as ntede 'understand' n dietatogether', nda 'first'. This is probably nothing more than a structural residue in the New World creóles, which for several centuries have not been like Kriôl in permanent contact with African languages. They should be treated as formes figées, and their structure will be analyzed in the set of morpheme redundancy rules rather than in the transformational component. The latter will of course include a study of nasalization by rule in Haitian (Part III, chapter 24) since, unlike the limited occurrence of underlying /Γ/ and \ u / , nasalization by rule is a highly productive process in French creóles. As for nasal vowels, including high vowels, we have already seen that they are not limited to French-based creóles. In Papiamentu, any vowel is nasalized to some extent—"más o menos nasalisada" (Lenz 1928:87-88)-before a nasal

10

PHONETICS

consonant, and even when no nasal consonant follows, as in kumida. High vowels are no exception to the rule: ür¡, punta, mundi, nür¡ka, ïimpi, etc., Sp. uno, un, punta, monte ( = bosque), nunca. Γ and u can even be found in creóles derived from a language where vowel nasality is non-existent or at least non-phonemic, as for instance English-based Krio (Sierra Leone) or the "folk-speech" of Jamaican Creole. Bailey lists Jamaican data where, in the same idiolect, u, Γ, a, and e are free variants of u, i, a, e: kum ~ ku 'come', im ~ Γ 'he, she', wan ~ wa 'want', pemkin ~ pekin 'pumpkin' (1966:15). Since she does not include them in her list of phonemes, they could be considered classical allophones of the oral phonemes. Yet Le Page (1957:383) mentions the regular occurrence of fully nasal Γ and u in JC phonology, which obviously requires a thorough study based on field work. At least u has been noted in Sranan utterances, and I would not be surprised to read in a future publication that Γ is also a regular underlying segment in that creole, as well as in Gullah. Berry gives a clearer description for Krio with the citation ¡sens ~ sesl 'sense', which illustrates his statement that "post-vocalic / m , n, 17/ have as free variants nasalization of the preceding vowel" (1961:11). He adds in a footnote that "Krio has five nasal vowel phonemes," and lists Γ, u, ε, a, j among the classical phonemes of the language (p. 4). However, these nasal phonemes would not necessarily be "systematic" phonemes in a generative treatment of Krio, and I have not been able to find the published or oral data which would be necessary to decide on the generative systemic status of these five classical phonemes. The fact remains that even English creóles have nasal vowels, including ΐ and u, at least at the phonetic and classical phonemic levels. Valdman (1971:206) and Dejean (1971:587) correctly deny the existence in Haitian of two oral [a] 's (front and back, as in some varieties of French), and of two classical phonemes /a ~ α/, as posited by Ans (1968). Yet, although Ans' oral phonemes a vs. α are an analytic artefact, there are indeed two superficial nasal a's: [ à ] , which is the nasalized form o f / a / , and [Λ] from /à/.

CHAPTER TWO

Mid Vowels

Oral mid vowels in Haitian Creole are normally open in closed syllables, closed in open syllables. Their distribution is thus similar to that of Standard French. But the exceptions to this rule are quite different in the two languages. French has an opposition e vs. ε in accented open syllables (e.g,,/e 'fairy',/ε 'done') and an opposition o vs. ) in accented closed syllables (e.g., sol 'willow', SJI 'ground'). The extent of these oppositions seems to be receding in modern French,2 but French creóles have entirely neutralized them:

Fr.

epol

Hn. zepjl

'shoulder'

otr roz

ht r~>z

'other' 'rose'

agi ε

agle

'English'

Yet both the opposition and the conditioned distribution still exist in Haitian, where they may be even more widespread than in modern French. Some medial consonant clusters leave the preceding syllable open (e.g., pt, vl, sp, st, gz).3 Consequently, the closed varieties of the vowels occur before these clusters:

Fr.

aksete destine eksperjas kcstjJ

Hn. asepte destine esperjas kestjj

'to accept' 'fate' 'experience' 'question'

In other cases, an open J appears in an open syllable, as in GRISE 'size', vjle 'thief, z?rej 'ear'; but it is the result of vowel harmony.4 Stil other formswith open and closed mid vowels in contrast are those where a syllable-final r has been deleted (synchronically, as in fle 'flower', vs. fieri 'decorated with flowers', or diachronically, as in kj, Fr. kir ' b o d y ' , ^ , FT.fleer'flower'). This has created a large class of words where the mid vowels are in contrast in Creole, but not in the French cognates. It has also created a systemic situation which does not obtain in French: a contrast D VS. O in open syllables. On the other hand, the French distinction j vs. o in closed syllables is never found in Haitian:

PHONETICS àkjr pjr po bjr

H n . akj ρj po bj

'again' 'harbor' 'skin' 'side, place'

CHAPTER THREE

r is a velar or post-palatal continuant produced with almost no air turbulence, and always voiced, as opposed to the French velar fricative, which is largely voiceless after a voiceless consonant. I have use r instead of the IPA symbol γ for typographical reasons, r appears in any position, except as the final segment of the syllable or before a consonant. In thousands of words borrowed from French it has disappeared entirely from both surface and underlying representations (e.g., tJti, Fr. forty 'tortoise', kadav, kadavr 'corpse'). In others, it has been retained in underlying final position, as evidenced by its occurrence in m3tr-e 'show' (mjt 'watch'), milatr-es 'mulatto woman' (milat 'mulatto'), mezir-e 'to measure' (mezi 'measure'). r is labialized before labial vowels and w. The resulting segment is noted in the examples as r or rw, although the labiality is normally simultaneous with the lingual articulation of r. Its consonantism may disappear completely, expecially between a stop and w or. u. The resulting segment is then a labial glide which is homorganic with r, i.e., w. A sequence of two w's is always simplified. In the following examples, tro and krwe are underlying representations which occur as surface forms only in the speech of bilinguals, i.e., as gallicisms: Hn. / t r i / -ν trwj ~ twj /krwe/ krwe ~ kwe

'too much' 'believe'

Another index of the systemic and articulatory closeness of r and w in French Creole is the fact that w is consonantalized and delabialized for emphasis, especially after a labial consonant: bwa 'wood' may then become phonetically [b7a] , 5 Words such as wuga 'oungan' or woze 'Roger' normally have an initial w. Some speakers do not hear this w, although it is present even in their own pronunciation of these words. Hall (1953) has both üga and ruga ("oûgâ, roûgâ") in his lexicon. Yet these same speakers hear an r in aro 'over, up' or rote 'height' (sometimes pronounced arwo, rwote, or awo, wo te). Since other speakers hear r in all these forms, I would posit the underlying representations /ruga, roze, àro, rot ε / . However, these facts cast some doubt about the future status of the systemic phonemic distinction r vs. w before labials. It is worth noting that w, like r, never appears in final position in an underlying morpheme.

CHAPTER FOUR

Palatals, Implosives, h

In their tables listing the classical phonemes of French Creole, Hall (1953, dialects of Port-au-Prince and the Gosseline Valley) has no affricates, while Goodman (1958, Trinidad) and Taylor (1947, Dominica) have / δ / and / d z / . Evans (1938, Haitian) notes that "ts and dz are used by peasant speakers where town people use tj and dj" (p. 203). This observation is in part corroborated by my Port-au-Prince data, where ts and dz occur in free alternation with tj and dj only in very few words and idioms such as dzjl 'muzzle' or tsu (in tjbe tsubum 'fall with a crash'). In all the other forms where Goodman, Taylor, and Evans mention ts and dz, I have found tj and dj (e.g., djab 'devil', djjb 'job', dj5n 'John'). The origin of these sequences will be studied in Part IV, chapter 42, as part of a palatalization process which is widespread in all creóles. It is also recurrent in Northern Gallo-Romance. In contemporary Haitian, this palatalization process operates as a phonological transformation on dental stops before i and /, giving ts and dz. Since this effect is produced by a low-level rule, it will normally not be transcribed in the examples. I have found only one implosive stop in my Haitian data, in the word b j k j , where it is in free alternation with the explosive homorganic consonant. Evans, the only creolist to mention their occurrence, cites an implosive b in kjba (name of a plant) and bjkj, and an implosive g in gaga 'witch' and huga 'Voodoo priest'. h occurs very rarely in Haitian, and can be found only in initial position. The "h aspiré" which is actually pronounced in the French dialect spoken by bilinguals is always reflected by r in the creole cognates. Yet, all my informants consistently used h in two or three words (e.g., has 'ax', hen 'hatred', haga 'shed'), and rejected initial r in these words as ungrammatical. I will therefore include h in the set of systematic phonemes of Haitian.

CHAPTER FIVE

Distinctive Features

The set of distinctive features which I have used in this study follows the binary format proposed by Jakobson, Fant, and Halle (1956, 1963) and Halle (1957, 1959), with a few modifications. For instance, the feature "diffuse" has not been used, and the simple opposition [± vocalic], [± consonantal] has been replaced by a system which includes the feature "syllabic." syllabic (syl): Separates the vowels, [+ syl], from all other segments. Liquids and nasal consonants are thus [— syl]. This grouping is "natural," since some segments in these two subclasses can become [+ syl] in strictly specified environments (see Part III, chapter 30). consonantal (ens): Divides the set of systematic phonemes into vowels and glides ([— ens] ) and "true consonants" (stops, fricatives, nasals). This feature corresponds to SPE (The Sound Pattern of English, p. 299) "consonantal." obstruent (obs): [+ obs] indicates either a complete, or a more radical constriction of the vocal tract, and applies to oral stops and fricatives. Nasals, liquids, glides, and vowels are [— obs]. This feature corresponds to SPE "nonsonorant." closure (clos): In [+ clos] segments, the air-flow through the oral cavity is totally or locally (e.g., for I) blocked. They include oral and nasal occlusives, » and /. I am following here SPE (p. 318), where [+ clos] is labeled [— continuant] and [— clos] [+ continuant]. The glides, r, and the vowels are [— clos]. fricative (fri): [+ fri] segments are redundantly [ - syl]. They include the traditional "fricatives" or "constrictives," plus h, without which this feature would be redundant throughout the system since true consonants which are [+ clos] are automatically [— fri] and vice-versa. voiced (ves); [+ ved] has here the traditional meaning of "vocal cord vibrations." [— ved] indicates voiceless segments. Nasal sounds are redundantly [+ ved] in Creole. nasal (nas): [+ nas] corresponds to a low velum position which allows a nasal release of the air-flow. labial (lab): [+ lab] corresponds to SPE "rounded" and roughly to Jakobson's "flat." It is characterized by a narrowing of the labial orifice and by labial projection.

16

PHONETICS

mid, high, and acute (act) are related to the position of the body of the tongue according to two perpendicular axes. The Jakobsonian featural opposition [± diffuse] (or [+ compact]) has been abandoned, mainly because it had different significations for vowels and for consonants. On the back-front scale only two articulatory areas are necessary in Creole, hence the opposition [+ act] versus [— act]. Four articulatory areas are needed for the vowels on the vertical axis. This can be achieved with two features in a binary model. The usual high and low could have been used by abandoning the traditional practice of rejecting the co-occurrence [+ high, + low] while accepting non-high, non-low segments. The featural scheme adopted here includes the following quadruple distinction: [+high, — mid] : Hn./, i,u, u (Fr.i, y, u) [+high, + mid ] : Hn.e, o (Fr.e, φ, o) [— high, + mid ] : Hn. ε, ο, ε , α, 5 (Fr. ε, ce, ο, ε, o, a) [— high, — mid ] : Hn.a (Fr. α, α") [ + mid ] delimits a subset of vowels which includes Hn. /a/ but excludes Fr. /a/. Thus the phonetic difference between these two segments is accounted for in the deep structure. Since there is no opposition in Haitian between mid /à/ (Λ) and a non-existent low ( - h i g h ) /à/, there would be no justification for positing a low α in the deep structure and then "raise" it transformationally. On the other hand, French (low) α must be raised to Creole (mid) α in the diaphonie rules (Part IV). [+ act] sounds are produced with a fronting of the tongue from a central position. Harms (1966, 1968) argues for the concomitant use of "grave," i.e., for a three-point distinction on the front-back dimension. This is not necessary in French Creole. The following charts illustrate the resulting underlying vocalic structures of Haitian and French.

DISTINCTIVE FEATURES

+

+

ac Lite

i / Γ

u ' li

e

0

•a

3I 3

ε I ε

-

a a

+

-

•a Λ

-

acute

i

y

u

e

Φ

0

ε I ε

Ό ε

+



labial

+ +

17

ce / œ

3/ ï

-

,

9 a / à

+

labial



+

-

Ό ε

Haitian

PART II

Underlying Phonology

CHAPTER SIX

Introduction

For the purposes of presentation, the structural constraints which apply to the underlying morphemes of Haitian Creole will be divided into those that deal with the internal structure of individual phonemes (chapters 7, 8, 9, 10) and those which bear upon sequences of phonemes—sequence structure rules (chapters 11 to 22). The distinction here is between context-free (segmental) and context-sensitive (sequential) restrictions, and it does not coincide with the classical paradigmatic vs. syntagmatic distinction. The latter is based on the Saussurian opposition between the lists or paradigms in the linguistic system or langue (e.g., the dictionary of morphemes, the list of phonemes) and the sequence of these elements in actual utterances. The former corresponds to a division, in the lexicon, between the structural limitations which apply to phonemes and those which apply to morphemes. The combination of redundancy-free definitions (chapters 7 and 8) and segment redundancy rules (chapter 9) provides the complete set of phonological segments and natural classes of segments (chapter 10), while sequence redundancy rules (chapters 11 to 22) specify whether a string of phonemes is (or could be) a native (or nativized) underlying morpheme of the language. Taken together, redundancy-free definitions and redundancy rules for segments and sequences form the complete description of the underlying phonological structure of Haitian. Since they also provide the complete definition of the morpheme of this language, they have sometimes been called "morpheme structure conditions." Part II is a detailed treatment of the underlying system (morpheme structure component) which forms a part of the lexicon, and will be followed by the phonological transformations (Part III) which derive the phonetic string from the concatenation of underlying lexical entries in an actual utterance. Saussure's langue-langage opposition has been improved upon by generative linguistics, but not abandoned after all. It has been refined by the addition to the linguistic system (langue) of redundancy rules and of derivational rules or transformations, which describe combinatory processes. In the following chapters, morpheme boundaries are represented by +. For discussions of theoretical topics such as the "true generalization condition," types of rules, and ordering, I refer the reader to Stanley (1968), some of whose proposals have been followed in this study, and to McCawley's treatment of

22

UNDERLYING PHONOLOGY

"static" vs. "dynamic" rules (1970:7, 8). For instance, the matrix represented by Tables 1 and 2 cannot be mapped as a branching diagram, since it includes no feature for which there would be no zero. In this I am following Stanley's proposal rather than Halle's (1959). Had it been set up so as to permit treediagramming, the redundancy-free segment matrix would have lost a great deal of generality. The main motivation given by Halle for the possibility of treediagramming is that it provides a quick way to test the binarity of the theoretical model by revealing any accidental use of blanks as a third feature specification. Greater generalization has been chosen here, rather than facility of verification.

CHAPTER SEVEN

Systematic Phonemes

Tables 1 and 2 provide generative definitions for isolated underlying segments. They list for each systematic phoneme those of its feature specifications which are distinctive throughout the system. The division of the "systematic," or underlying, phonological structure of Haitian Creole into two submatrices (Tables 1 and 2) is due solely to space limitations, and is in itself of no theoretical significance. u

ι

syllabic consonantal obstruent fricative closure nasal high mid acute labial voiced Table 1. Redundancy-free

b t

systematic phonemic matrix

d k

g f v s

syllabic consonantal obstruent fricative closure nasal high mid acute labial voiced Table 2. Redundancy-free

systematic phonemic matrix

(vowels)

u.

m

η ñ

(consonants)

1 r

j

w h

CHAPTER EIGHT

Natural Sets

Each column in Table 3 provides a generative definition, within the complete underlying structure of Haitian phonology, of most of the articulatory subsystems which were recognized in structural linguistics under such labels as fricatives, liquids, glides, etc. As in Tables 1 and 2 (segment definitions), this matrix separates one class from all other classes of the system, and is free of redundancies; it lists only the set of feature specifications which is distinctive. Just as in Tables 1 and 2, each definition (each column) is language-specific: the column for "labio-dentals," for instance, provides a definition of labiodental units in relation to the whole underlying system of Haitian.

Vowel Consonant Stop (oral) 1 Fricative 2 Fricative 3 Nasal C Bilabial C Labio-dental 4 Dental Velar Labio-velar Glide 5 Glide 6 Low Vowel 7 Mid V(owel) Low-mid V High-mid V Front V Back V High V Nasal V

syl +

ens

obs

fri clos ñas 1 high

mid

act

lab

vcd

+

+ +

+ + +

+

+ +

+ +

+ +

+







+



+

_

+

+

+ + +

Table 3. Redundancy-free





+

+ + + + +

+ —

definitions of natural

+ sub-systems



CHAPTER NINE

Segment Redundancy Rules

The complete definitions of segments and natural classes of Haitian Creole will be grouped in Tables 5 and 6 (chapter 10), where each definition is represented by a column of distinctive features. These columns are completed with the redundant features specified by the following set o f redundancy rules: the segment rules of this chapter and the sequence rules of chapter 10. Redundancy rules are distinct from phonological transformations (TR's) in both their function and their location in the grammar. TR's change feature specifications, may add or delete segments, and are included in the transformational component of the phonology. Redundancy rules never change pluses and minuses, but they describe all the redundancies of the underlying system by adding feature specifications. Their function is therefore to complete the lexical entries, and they are part of the lexicon. Just as a complete Chomskyan "grammar" can be viewed as the definition of the phrase "sentence of language x , " the list of systematic phonemes, coupled with the set of segment structure rules, is in effect a series of compact and precise language-specific phonological definitions. They define for a given language the term "systematic phoneme" and all the possible subsystems of the underlying phonological structure (such as velars, fricatives, semi-vowels, etc.). REDUNDANCY RULES (RR's)

R R 1.

1. 2.

+ mid — clos + ñas

3.

+ ñas + lab -

R R 2.

1. 2. 3.

[ + syl

]

[ -

]

high

I + cns [ + fric [ — ved

-

syl

26

UNDERLYING PHONOLOGY

RR 3.

1.

2. 3.

+ + -

+ RR 4.

1. 2.

3.

f +

cns

[

-

cns

]

[

+ obs

]

[

-

]

[ +

syl obs I L _ vcd r + lab L + high + lab • — fric - clos

• + cns

RR 5.

2.

3.

4.

RR 6.

obs clos syl high vcd

>

fric) vcd) fric act j j + lab vcd) ™ vcd + lab + high + act _ fric clos η as act

_ cns + η as — — —

+ +

fric clos fric high act

obs

SEGMENT REDUNDANCY RULES — (+ (+ + + 1

— — — : — + ' + +

1

clos obs) vcd| obs high act obs syl ved syl ens ved ens lab clos

+ clos + syl + nas

2 3 4'

— (+ _(+ +

[

+

fric

]

[

-

fric

]

[

+ clos

]

[

-

]

[

+ nas

] ] ]

obs ved) actJ _ ens

— lab — act 1

— + + — +

21

3

syl nas obs fric ens

— fric (+ lab ) > high)

RR 10.

RR 11.

[

— ens

[

+ fric + high + act + clos

.

.

clos

]

27

28

UNDERLYING PHONOLOGY

RR 12.

1 2 3 4 I 1 5 i i ι 6

obs vcd syl — clos " + high + mid " - high mid ens » L clos» . syl fric — act lab —

[ - ñas

]

RR 13.

[ + high ]

RR 14.

[ - high ]

SEGMENT REDUNDANCY RULES

RR 15.

[

RR 16.

c:

-

syl

Γ +

obs

1

+

fric

.

lab

-I

r — ens

η

L +

lab

-I

r -

obs

"I

+

I

[

-

mid

]

[

+

act

]

[

-

act

]

[

+ lab

]

clos

L— nas [



high

L +

RR 17.

]

J

lab ens ved

Γ + obs -

fric

L+

high hie, ens

[ =

obs

•— clos

Γ: ] syl

high mid

RR 18.

r + L -

obs act

r — clos + L-

η

high J η

high act

J

29

UNDERLYING PHONOLOGY

RR 19.

+

1



2

+ + + +

3 4

+

5

— -

RR 20.

1

2 3

RR 21.

-

act obs fric ens high syl high mid ens obs clos

obs lab j+ high j - fric ens + act obs syl + act

|+

Γ— obs [+ fric

CHAPTER TEN

Redundant Segment Definitions

The definitions given by Tables 4, 5, and 6 result from the operation of the segment redundancy rules upon the redundancy-free matrices of Tables 1 and 2. The pluses and minsus which make up these redundancy-free definitions (Tables 1 and 2) are circled in the following tables.

Vowel Consonant Stop (oral) Fricative Fricative8 Nasal C Bilabial C Labio-dental Dental Velar Labio-velar Glide Glide8 Low V Mid V Low-mid V High-mid V Front V Back V High V Nasal V

syl Θ

ens obs —

-

tri clos

nas

high

mid act

lab

ved +







Θ







-







+ θ

Φ

θ —

+



+

θ θ

— —

ω



+



+



Φ



Θ + + +

Θ θ Θ +

Table 4. Fully-specified

— —

+

θ

θ θ θ

θ θ

-



θ













Φ



_

_

-

_

-

+









-









-































-

θ of natural

Φ θ

Φ Φ —

θ + Φ



Φ

-



+

Φ



+ + +



Φ

+



θ

θ

θ

Φ Φ

θ

Φ





definitions

+

— —

Θ θ

ο θ



θ Φ sub-systems

θ

+

+ +

32

UNDERLYING PHONOLOGY

syllabic consonantal obstruent fricative closure nasal high mid acute labial voiced

1 + θ

e +

ε + θ



θ Φ θ

θ θ +

-

θ

ω ω

®

ω

a Φ

θ θ Φ θ + +



θ θ —









+

+

+

+

0 +

+

Ρ b t syllabic consonantal obstruent fricative closure nasal high mid acute labial voiced

d k g f



Φ Φ θ θ







+ +

φ +

-

θ φ φ —

φ

0 +

U +





-

+

+







φ φ

Φ Φ

Φ θ

Φ

-





Φ +

φ +

ε +

a +



Φ

-

-

θ

+

+

+

(vowels)

ν s ζ

+ + + + + + + + + + + Θ + + + + θ Φ Φ Φ Φ Φ + Φ — θ — — Φ Φ — —

Φ Φ

θ + θ



Table 5. Fully specified systematic phonemes

1 +

U φ

+ + + + φ φ

θ θ

m η ñ 1 r j W h θ θ — + + + + + + Θ θ — θ + + - - - θ θ — — — — — + φ φ — — — Φ Φ Φ Φ θ + — — — — Φ Φ θ φ φ — θ Φ — — + Φ —

I

ζ

φ φ + +



Φ Φ +



Φ





Φ Θ Φ Φ Φ — Φ + + + θ + + + + θ Φ θ Φ θ Φ θ Φ θ φ θ φ —









Table 6. Fully specified systematic phonemes







(consonants)











CHAPTER ELEVEN

Sequence Redundancy Rules

The segment redundancy rules are supplemented by sequence rules, which specify what combinations of phonemes are acceptable in lexical entries. The complete set of redundancy rules realizes the structural constraints which apply to underlying lexical units, and represents the compact, precise, and complete definition of the underlying morpheme in the language. It provides all the combinations of distinctive features which are "grammatical" between two morpheme boundaries in terms of segments and sequences of segments, and produces no ungrammatical combination. The rules are unordered. Each rule operates without reference to any other rule, and a morpheme is processed cyclically until it goes through the whole set with no chnage being made in its phonological structure. It is clear that the Praguian concepts of "archiphonemes" and of "redundancy" have been preserved in this generative treatment of phonology (see chapter 12). Redundancy rules fulfill another function, which is even more crucial in creolizing and decreolizing situations than in ordinary cases of borrowing. They act as a filtering mechanism which nativizes newly borrowed lexical terms. Thus a loanword will be processed by these rules, which act then as transformational rules, changing feature specifications, deleting and adding segments. These creolizing processes are described in Part IV of this volume, with special attention to the English, French, Portuguese, and Spanish creóles of the Atlantic and the Caribbean.

CHAPTER TWELVE

The Archiphoneme in Generative Phonology

+ clos + high act + ved

RR 22. ens Ί nas J i

[ + syl]

high] act

-

Y

-

+

In a former study of Haitian phonology (Tinelli 1970), I included η in the set of systematic phonemes, and described one of its environments as the result of a deep-structural sequential constraint: a following segment could only be g. That treatment introduced a dissymmetry in the sequence redundancy rules; g was singled out as an irregular unit in the set of voiced stops. Another dissymmetry was created in the underlying consonantal structure, which had four nasal segments (m, n, «, η), as opposed to the three-unit subsets b-d-g, p-t-k,f-s-J, v-z-z. Actually, 77, not g, is the exceptional segment, η must still be kept in the underlying system rather than created by the transformational component. Yet it does not belong in the set of systematic phonemes. Phonetic [77] and [g] never occur together at the end of a morpheme. With very few exceptions, phonetic [η] is always in word-final position. In some cases it alternates with [g] when the morpheme is followed by a derivational suffix: Ιο η, log+e 'long, length,' Ιαη, lag+az 'tongue, speech'. When the derivation is not by affixation, g does not surface, as in the compound Ιαη+sat 'cat's tongue' (a flower). In other morphemes, [17] does not alternate: ζοη 'nail', zepcv 'pin', se+domev 'Santo Domingo'. In the first case, both / η / and Igl will be posited in the underlying lexical entry: /1077g/, /lar?g/. If we follow a strict alternation principle, the nonalternating forms must be entered in the lexicon without / g / : / ζ ο η / , / z e p e r j / , /domer?/. Yet, throughout the phonological system of French creóles, 17 is the only nasal consonant which appears solely in final position or before underlying g at a morpheme boundary. In that respect, 17 is exceptional in the set of nasal consonants, and exceptions are normally described through transformations in Chomsky an linguistics.

GENERATIVE PHONOLOGY

35

In any event, this archiphonemic structure, which in similar environments also covers m and n, must be accounted for somewhere in the grammar. A classical description of these structures would include an underlying/n/ or /N/ (/zaNb/, /vaNd/, /laNg/) and derive the superficial forms by transformation: /zaNb/ /vaNdI /laNg/

-> ->

iamb vand lam

-»· ¿am ~ iab + -> van ~ vad + Ιαη ~ l a g +

SUFFIX SUFFIX SUFFIX

η would never occur in the underlying structure of the phonology, /m/ and /n/would be present only in non-alternating lexical entries (e.g., mene, worn, etc.), and they would always be derived by transformation in all other cases. Two rules are necessary for the above derivations: a "placing rule," which will define the place of articulation of /N/ or of the NC archiphoneme, and a deletion rule which will drop m, n, or η when b, d, or g is retained, i.e., before a suffix. The question is whether both rules function in the transformational component. They are quite different as to the type of environments which trigger them. The deletion rule cannot operate unless the morpheme has already been inserted into a string, i.e., before the "surface structure" level: the change depends on the presence or absence of a word boundary or suffix. Since this rule describes a phonotactic or combinatory change, it must apply after lexical insertion, and can only be a transformation. The rule which "places" the articulation of NC is totally independent of higher structures. It defines a phonological combinatory process, as does the deletion rule, and a process which at first sight appears to be a purely phonetic, low-level change. Yet it can be predicted at a much deeper level of the grammar. Before any transformation has applied—i.e., before the surface structure enters the phonological component—and even before lexical insertion. Since it describes a phonological co-occurrence restriction which is predictable in the lexicon within morpheme boundaries, it must be a deep-structural process which occurs in the redundancy component of the phonology, i.e., a sequence redundancy rule. It states that the phonology of French creole includes an NC archiphoneme which is always homorganic with a following voiced stop in every underlying morpheme where it occurs, and that in any new loanword, a nasal consonant will be assimilated to this archiphoneme if it is followed by an underlying voiced stop. Therefore η is present in the underlying phonological structure. It is not one of the redundancy-free systematic phonemes; rather, it is represented in lexical matrices by the NC archiphoneme, and its place of articulation is specified by a deep-structural rule. In that respect, the situation in French Creole is similar to that of Papiamentu, of Spanish (except in Chile and Argentina, as noted by Lenz 1928:77-78),

36

UNDERLYING PHONOLOGY

or of Krio, where Berry notes that "the opposition m # n #17 is neutralized in preconsonantal position: the variant of all three consonants is then a nasal which is homorganic with the following consonant: cf. /abirjkoni/ : / a b i n d u / " (1961: 11). Berry adds that m, n, and η are in free variation in word-final position, and that "confusion of end nasals has a long history: cf. 'town, down, ground' > / t 3 T ¡ , d DTj, gran/, 'lie down, sit down' Κ /lid a m , sidjm/" (ibid.). The fact that this has not occurred in French creóles could be taken as a clue to a deepstructural difference between the creóles based on French and those based on English. Yet I think that the "confusion" in Krio was featural and phonetic rather than rule-governed or deep-structural: the sequence wn became η in some morphemes or allomorphs and m in others, by a fusion of the consonantal nasality of n with the velarity of w ( wn > 17), or with the labiality of that same w (wn > m). In a form like E. 'ground' > Krio /gran/, w has disappeared without influencing n because it preceded a homorganic stop, which maintained its point of articulation. Since French formatives did not have w in that environment, nasal consonants in word-final position were unchanged from French to Creole, except when they were homorganic, in sixteenth or seventeenth century French, with a final stop (e.g., French bombe, langue, bande > Creole b5m, 15"% ban). The underlying organization of consonantal nasality is different in modern French, where [ η ] , when it occurs at all, is not a segment but a short articulatory transition (from NV to a velar stop) resulting from a late transformation. This η is too short to be audible for a native speaker, who in any case might not hear it even if it were longer, since it has no functional reality. In Creole, on the contrary, [77] is a full-sized unit, and native speakers will correct the learner who places or replaces it in an ungrammatical string. On the other hand, the systematic nasal consonants of French and Haitian belong to identical sets: /m/, /n/, and / ñ / . In both languages, /ñ/ and the fricatives If I and /z/ are [+ acute], whereas the corresponding stops / k / and /g / are [— acute]. This front articulation (+ acute) of the "velar" fricatives and nasal consonant is due to a palatalization which took place in Gallo-Romance, and the systemic dissimmetry has been maintained in Creole as it has in French. The fact that Haitian has four nasal consonants at the phonetic level does not entail that all four have the same deep-structural status. We have seen that some morphemes with final [r?] never alternate. This would seemingly force us to posit η as a systematic phoneme, since a child learning his language would never hear a [g] in any occurrence of these formatives. However, a possible conception of the morpheme structure constraints, or redundancy rules, is that they define any unit which is an actual or a POSSIBLE underlying morpheme of a particular language. Thus the set of systematic phonemes and redundancy rules does not at any given time describe the actual lexicon of the language, but rather its lexical STRUCTURE. The absence of suffixal derivation for

GENERATIVE PHONOLOGY

37

non-alternating morphemes with [τ?] is of course due to an accidental gap in the lexicon, or at least to non-phonological causes. The question is therefore whether a newly coined or borrowed cognate would add a suffix to [ζδη], [zepgrj], [se+domo?], etc., or to *[zîg—], * [ z e p s g - ] , * [se+domeg-] · I will assume here that the starred (not evidenced) series would be chosen by native speakers if derivational affixes were added to these morphemes, i.e., that /g/ is present in their knowledge of these underlying forms. A I gl which never surfaces for an accidental, non-phonological reason: the necessary environment is not realized in the current vocabulary of the language. This analysis offers two advantages: it can be disproved with new data, and it leaves open the possibility of integrating newly coined or borrowed forms such as, for instance, *zep§g+e '(?)to pin' or *z?g+e '(?)'. If, as was first claimed by Navarro-Tomás (1953:189), some creóles are now in a process of decreolization, the possibility of inserting such new words must be built into their CURRENT grammar: surely the phonological structure of Haitian will not have been modified if a vocabulary item such as *zepege is borrowed from French "épingler". Rather, it is the current structure of its grammar which will have made that borrowing possible. The model proposed here retains m, n, and i? in the underlying morphemes before they are inserted into the output of the syntactic component, but not in the redundancy-free lexical entry. The mechanism of the NC archiphoneme and the distribution of functions among the various components of lexical phonology are illustrated with the following derivations. A. Lexical

syl ens obs fri clos ñas high mid act lab ved

entries

0 0' 0 + 0 0 + 0 0 0 +

+ 0 0 0 0 0 0 +

0 + 0 0 0 + 0 0 — 0 — 0 0 0

0 0 0 0 + —

0 0 + 0 +

NC b I ¿dm 'leg' ¿Abe 'to cross'

/ Ζ Λ

0 0 0 + 0 0 0 0 0 + +

+ 0 0 0 0 0 0 + — —

0

0 + 0 0 0 + 0 0 0 0 0

0 0 + 0 +

0 0 -

0 +





0 0 + 0 +

0 0 0 0 0

/ν Λ NC à/ vm 'sell' Vàide 'salesman'

0 0 0 0 0 0 0 + —

0

0 + 0 0 0 + 0 0 0 0 0

0 0 0 0 + 0 + 0 —

0 +

a NC gl 'tongue' liïgaj 'speech'

/I

IATI

38

UNDERLYING PHONOLOGY

Syntactic and semantic information has been omitted in the matrices. "NC" stands for the archiphoneme defined by the feature specifications of its column. B, Morpheme structure constraints: output of the segment redundancy

syl ens ob s fri clos ñas high mid act lab ved



+ + +

+







+ + — +









+ + — + + — + — 0 — — + — — + — 0 — — — 0 + + + + + i





λ NC

b



+ + +

+





+

— —

_



+





+



+



+ +







+ +













+

+



+





+





-

+ + 0 —



+





+ + 0



+







+





+

+



+ +





-

+

0 0 +

+

+

ν

a

NC

d

ι







rules

— — —

+

0 0 +

+

λ

NC

g

— —

In terms of segment structure, NC has a redundancy-free definition. But its articulation (height, acuteness, and labiality) is still predictable from the sequence structure constraints. The other segments in the three matrices are completely specified at this point, since no other morpheme structure constraint will apply to them. C. Morpheme structure constraints: output of the sequence redundancy

syl cons ob s fri clos ñas high mid act lab ved



+ + +

+







+





+

+



+



+ +







+ +













+

+







+











+ + +

+

+

+ +



+

+

+

-

-







-





-



+ /

+

+

+

A

V

g















— —

-





+







-



-

+







+ + +

+ + + + —



+



+ + — +





-



+ + + + + +

+ +

+

+

+

ζ

Λ m b

ν

Λ

η

d



+

rules

GENERATIVE PHONOLOGY D. Output of phonological

39

transformations

Once completed, the underlying matrices are placed in the context of the "surface structure," and subjected to the operations of the transformational component described in Part Three of this volume: ZAmb

#

#

VAtld

#

#

lìng

#

#

-*·

->

zKmb + e vAnd+

ε

V(ng+

az

->

TR's TR's TR's TR'S TR's TR's

ûm VAITI

I At) - >

zîtbe

->·

vAde lAgaj

->

'leg' 'sell' 'tongue' 'to cross' 'salesman' 'speech'

CHAPTER THIRTEEN

Lexical Nasality

RR 23.

[ [ +

I

"+

ens "

L+

nas -

+

syl]

A nasal consonant in final position must be preceded by a vowel. This constraint does not apply in non-final position: kadna, promnad. Nasal structures are complex and pervasive in Haitian, as in all the other French creóles and in some of the English-based creóles of the Atlantic and Caribbean (e.g., Gullah). Some of the nasality which is found at the phonetic level is lexical, while in other parts of the same utterance it has a transformational origin. Nasality which is produced through synchronic combinatory processes ("rule-derived" nasality) will be studied in detail in the transformational component of this volume (Part III, chapters 24 and 25). Lexical nasality is present when a vowel is nasal in the Haitian morpheme and its French etymon, or when nasality is the result of a historical change which has been "lexicalized," whereby a French vowel has become nasalized in all occurrences of a given Haitian morpheme. Lexical nasality is automatic however when the vowel is followed in the underlying representation of a morpheme by a nasal consonant. This fact can be described with the following sequence redundancy rule: RR 24. syl " [+

syl] +

nas -

I [+

nas ]

RR 24 has very few exceptions (e.g., km 'horn', kone 'twist of paper', leogan 'Leogane', kamjj 'bus, truck', denje 'last'. In words such as Tmid 'damp',

LEXICAL NASALITY

41

aprëmidi 'afternoon', legón 'vegetable', the vowel which precedes the nasal consonant is nasal, even though it is a high vowel. In p/Y/mr'millet' or bananmï 'ripe banana' (a bird), for instance, the nasality of ?" can easily be missed on the first hearing. It is clearly audible, however, when it in turn nasalizes the following morpheme, as in piffmija'the millet' or bananmlja liwoz 'The ripe-banana bird, it's pink'. Since nasalization across morpheme boundaries does not hop over voiced segments to assimilate following units in Creole, the definite article in these examples is obviously nasalized by the vowel of nii t ripe, sorghum'. The same test confirms auditory impressions with lame 'sea, lake' and ane 'year,' which do not nasalize the singular definite article: lamea 'the sea', tutanea 'throughout the year' (cf. tutlaíunea 'throughout the day'). The insertion of rule 24 into the phonology of Haitian reflects my assumption that new morphemes, borrowed from another language or coined, would be modified to fit this constraint by a native monolingual speaker. Should a regular pattern of oral vowel ~ nasal consonant (within morpheme boundaries and free of language-contact interferences) later emerge to prove this generalization too powerful, RR 24 would have to be modified according to the phonological or syntactic environment which had conditioned such a sequence. According to Jourdain, "in all American creóles, it is always the case that m, n, and gn nasalize the preceding vowel, or the vowel that follows them, and often both" (1956:294). In the terminology of generative grammar, this statement claims two OBLIGATORY rules which respectively nasalize the vowel which precedes and the one which follows m, n, or ñ. These rules are in a disjunctive ordering, i.e., m, n, and ñ nasalize EITHER the preceding vowel OR the vowel which follows them. However, in terms of performance, the disjunctive ordering would "often" become conjunctive,with both rules applying to the same phonological string. I have adopted the regressive nasalization rule as RR 24, a sequence restriction on underlying morphemes. I find the second rule (progressive nasalization) too powerful in view of the large number of morphemes which do not respect such a condition: e.g., lame 'sea, lake' (vs. me 'but, hand'), neg 'man', bone 'early', ne 'be born', me'nun', etc. and despite the fact that in many others a nasal consonant is indeed followed by a nasal vowel: me 'hand', mene 'take, bring', mama 'mother', retane 'go back', mes 'thin' (vs. lames 'mass'), mi"'ripe', sorghum, wall' and so on. The nasality of the consonants in words such as finì 'finish', tune' go back', djnu 'sleep', etc., can extend not only to the following vowel but, even across a morpheme, word, or phrase boundary, to still another segment: li ap νΐηΐ / li fate -*• l αρ νΐηΐ/ Ifate lap νΐηΐ/η fate 'He was singing when he came', li refemëli ako li refemel ako -*• li refemen ako 'He closed it again'. The confusion which often accompanied treatments of nasality in French Creole seems to have had two main causes. One is the lack of a clear distinction between the lexical and syntactic domains. Many creolists have in fact entirely

42

UNDERLYING PHONOLOGY

overlooked processes which take place across syntactic boundaries and have restricted their descriptions to lexical items; Yet, when a formative reaches the superficial level, it may have been modified by a nasalization rule. (See Chapter 25, Part III.) Moreover, it is not always clear whether the standpoint is synchronic or diachronic. For instance, when Jourdain states that a vowel is nasalized in contact with a nasal consonant, she is concerned solely with words at the phonetic level. But is she describing a historical change or a synchronic Haitian process? It would of course be futile, especially in the case of creóles, to draw an impassable line between historical changes and synchronic processes, but transformational theory allows us to conduct two separate analyses as a first stage of the investigation: one for the structure of the lexical entries or underlying morphemes of the language, the other for patterns which are generated by the transformational component—patterns which are the result of combinatory processes between morphemes or between higher grammatical units. These combinatory changes are triggered by specific phonological concatenations of segments and boundaries (and also, probably, by intonation patterns which cannot be adequately described with the current transformational models), by syntactic labels and lexical markings on some classes of morphemes, or by combinations of all of these elements. RR 12 in the segment structure component makes all [+high, + mid] vowels non-nasal in underlying forms, so that e and o are always oral in the lexicon, even when followed by a nasal: RR 24 cannot apply to them. (See Part III, chapter 24 for superficial e and o.) The main effect of RR 24 is a crucial morphophonemic simplification from French to Creole. Virtually all alternations between nasal and non-nasal vowels which were part of the French declensional system (e.g., as gender markers) have been lost in Creole. Derivational alternations have not all disappeared, but this particular alternation pattern, which was twofold in French (oral vowel ~ nasal vowel, followed by nasal consonant ~ zero), has been reduced to nasal consonant ~ zero: Fr. prizj, prizon+je /prizme, Hn. prizj, prizJn+je /prizme 'prison, prisoner'. In a very few cases, Haitian has kept an alternation in the height of the vowel (Fr. vwaze, vwazin+ai, Hn. vwazë, vwazTn+aj 'neighbor, neighborhood'), but the opposition nasal vowel ~ oral vowel has been lost in all cases. Contemporary Standard French has no condition equivalent to Haitian RR 24, as the following examples show: Fr. kart ami plat zam bene kjn&r,

Hn. kan zarra plen (pa)iam bene kins kjne

'cane' 'friend' 'pregnant' (animals) 'never' 'bathe' 'know'

LEXICAL NASALITY

43

However, French phonology contains a vowel-nasalization rule which must operate before the deletion of the final consonant in order to provide superficial alternations of the type: Fr .bo 'good' (m) ~ bon' good' (f), bjn(-cer) 'happiness' pie 'full' (m) ~ plen 'full' ( f ) vwaze 'neighbor' (m) ~ vwazin 'neighbor' ( f ) vwazin(-az) 'neighborhood' The overall result of these diachronic changes is that the superficial nasal vowel can still have two different synchronic sources in twentiethcentury French: it may derive by transformation from an underlying oral vowel (e.g., /ρΐεη/ -> [ρΐε] ~ [ρΐεη]), or it may be inherently nasal, as in / t a p / -»• [ta] 'time'. In Haitian it is always inherently nasal (with the exception of the new nasalization rules analyzed in Part III, chapters 24 and 2 5 ) : / ρ ΐ ε η / -»· [pie ] ~ [ ρ ΐ ε η ] , / ta/ ->· [ t a ] . The end result of the creolizing change is thus a morphophonemic differentiation between French and its creóles. The creolization process has cancelled the rule for vowel nasalization, and "lexicalized" vowel nasality within morpheme boundaries by generalizing the deep-structural origin of these nasal vowels.

CHAPTER FOURTEEN

Vowel Sequences

Sequences of vowels are extremely rare in the lexicon (e.g., pei 'country', mai 'corn', and their free variants peji, maji). Since no sequence of three vowels appears in my data, I have added the following rule: RR 25.

[ + syl ]

[ + syl ]

[

] I

[-

syl ]

CHAPTER FIFTEEN

Glides

RR 26.

— syl

X

— ens [ + syl] An initial vowel must be followed by a vowel. Rule 26 prevents sequences like +jt, +wj, +jr, etc. RR 27.

r-

, π ensj

0

i act ]

The only glide that can occur in final position is / . Therefore, j can be totally and minimally defined, in this position, as [ — ens, — syl], and the remaining features of / are redundantly specified as + or — by the set of segment and sequence redundancy rules. For instance, [+ act] is redundant in the underlying forms of morphemes like kaj, fej, ruj. kaj 'house' is entered in the lexicon in the form of the following matrix: syl ens ob s fri clos ñas high mid act lab ved + k a J

46

UNDERLYING PHONOLOGY

a is specified as being [ + syl ] in this environment by sequence rule 34. This shows that the complete set of segment and sequence redundancy rules is unordered: rules 34 and 27 must apply before other segment rules have completed the definitions of /a/ and /j/. Placing segment and sequence structure rules in two different components would result in widespread systemic redundancy. A glide cannot be preceded by another glide. This rule is in fact a sub-matrix of the gemination rule (RR 29). It applies vacuously to h, since h can be preceded only by a morpheme boundary and followed only by a vowel. Therefore, RR 29 states in effect that the sequences wj and jw are not permitted in lexical entries. In only one alternant of one morpheme can we find jw: jwit ~ —wit 'eight'; —wit appears in diz-wit 'eighteen', or in isolation ('eight'). I have treated here the free alternant jwit as an exception to the gemination rule.

CHAPTER SIXTEEN

Liquids

RR 28.

[

+ ens — obs — nas

] I

-

I + ens — obs — nas

Rule 28 rejects sequences of two liquids in the underlying representations of Haitian morphemes. The symbol ~ indicates here a negative condition. It is of course possible to avoid this negative statement. The class of liquids is defined in Creole by the following specifications: + consonantal — obstruent — nasal Therefore the rule is that a systematic phoneme having these three features must be preceded by a morpheme boundary or by a segment in which at least one of these features has a different specification. The following chart indicates the three classes (x, y, z) of segments which are acceptable in lexical entries before a liquid. (Zero may be either + or —.)

consonantal obstruent nasal

X 0 + 0

y

0 0 +

ζ -

0 0

48

UNDERLYING PHONOLOGY

These facts can be stated in a positive rule: + ens — obs — nas

RR 28.

[ — ens [ + obs [ + nas +

] ] ]

CHAPTER SEVENTEEN

Gemination

RR 29.

— syl α clos β fri 7 act

[ - syl

—α -β —y

]

clos fri act

RR 29 rejects consonant gemination (including liquids and glides) as foreign to the lexical structure of Haitian. Gemination does occur in utterances and in intermediate representations, but always across morpheme or higher boundaries: e.g., I ap pote maze (p + ρ -> [ρ: ] ) 'He will bring some food', pje m fem mal (m + m -*• [m:]). In that respect, French Creole displays a deep-structural organization which is characteristic of Gallo-Romance (and more generally Western Romance). Modern French has reduced lexical gemination, except in highly formal, consciously archaic, or semi-educated styles, and geminates are always derived by rule.

CHAPTER EIGHTEEN

Initial Clusters

RR 30.

+ [ -

syl

]

[ -

syl

]

I Γ - obs L — nas

1 J

If a morpheme begins with a consonant, the second segment is a vowel, a liquid or a glide (except h). This condition eliminates initial clusters such as ps (cf. Fr. "psychologie"), or ft. Whenever words including such clusters within morpheme-boundaries appear in a corpus of Haitian, there are other clues in their phonological, morphological, and syntactic environment which point to a direct borrowing from French. [ - syl

]

[ -

syl

]

; + [ + ens

]

The initial consonant of a morpheme cannot be a glide if it is not followed by a vowel.

ΓΊ

+ r L -

syl

t -

syl

]

obs J 4· [ -

ens

]

An initial liquid or glide must be followed by a vowel, a liquid, or a glide. In a traditional generative model—i.e., if redundancy rules were mixed with phonological transformations—RR 32 and RR29 would stand in a "bleeding relationship", the possible input to 32 being reduced by the application of 29. In the

INITIAL CLUSTERS

51

framework choosen for this study, the ultimate result is identical: sequences like w j are rejected in all environments (within underlying morphemes) by rule 29, and sequences like ml are rejected in initial position by 32 (but not by 29 : such sequences may appear in other positions: e.g., Umlzt, palmis, etc.).

CHAPTER NINETEEN

Final Clusters

RR 33.

[ + fri

X

]

[ + fri]

Y



[

]

[

]

No cluster of two fricatives is found at a boundary. (They are extremely rare even between segments.) RR 34. X L + ensJ I [ + syl

]

A glide in final position is always preceded by a vowel. R R 35.

[ -

syl

]

[ -

syl]

+

I [ + ens

]

In final position, the first member of a consonant cluster cannot be a glide. This rule rejects *jt+, *wm+, etc. RR 36. [ -

syl

]

4[ + obs

]

- obs + ens — nas

FINAL CLUSTERS

53

In a final cluster, / or r has to be preceded by a vowel or by an oral fricative or stop (i.e., by an obstruent, since the fricative h is excluded from final position): e.g., bl, fl, tr, kr, br, etc., but not mr, Ir, jr. In final position, clusters such as zr, ti, il, dl, si, / / do not seem to occur. These clusters are also very rare in French, where they occur mostly in learned words. However, I have not excluded them from the possible structure of underlying Creole morphemes.

CHAPTER TWENTY

Medial Clusters

In clusters of fricatives and oral stops, the two consonants have the same specifications for the feature voiced. Since this is always the case in surface representations, I have stated the co-occurrence constraint as a redundancy rule rather than in a phonological transformation. RR 37 is in fact a conflation of two rules: an obstruent is voiced before a voiced obstruent, and unvoiced before an unvoiced obstruent. These conditions will be numbered 37a and 37b. RR 37 (a/b).

[ + obs

]

[ α ved

]

Γ+ obs"| |_α ved J

RR 37 is true of a sequence of two obstruents (e.g., sf, sk, tj, ks, kt, gz), but it does not apply to all sequences of two [+ ens] segments (e.g., mp, ms, or consonant + liquid). RR 38.

— —

syl obs clos act [ + ens

]

r, w, or h cannot be followed by a stop, fricative, nasal consonant, or liquid. (/ and I can be followed by a [ + ens ] segment: e.g., kajma ~ kajima 'cayman', elmi 'enemy'.) RR 39.

Γ+ obs L+ ved

I J

[ + ens

]

i [ + ved

]

MEDIAL CLUSTERS

55

Whenever a voiced obstruent is followed by a [ + ens], but [ + or — obs ] segment, both are voiced. RR 39 delimits a sub-matrix of 37a. Yet 39 and 37 as a whole have different functions: each carries some information which is not provided by the other. However, 37a does not state anything which is not provided by 39, and can be deleted from the grammar. In this instance, as in most other cases, the use of alphas and betas is not essential for the phonological description. The alphas of RR 37 can therefore be replaced by the specification —, with no loss of accuracy or generality, and 37 ( a / b ) must be rewritten as: RR 37.

[ + obs

]

4[ - ved

Γ+ obs — ved

]

CHAPTER TWENTY-ONE

Three-consonant Clusters

RR 40.

[ - syl

]

[ — syl

]

Γ - syl + ens 4· [ - nas

]

In a sequence of three consonants, if the third is not a glide, it cannot be a nasal either, i.e., it can only be a liquid or an obstruent. RR 41.

[ + obs

]

[ -

syl

]

[ -

syl

]

I [ + fri

]

In a three-consonant cluster, if the first segment is an obstruent, it cannot be an occlusive. The only obstruent which occurs in this environment, in my data and in previous studies, is s. As it is stated, rule 41 reflects my assumption that / , s, ν, z, and ζ are also possible in this position. In final position, the only instance of a three-consonant cluster that has been recorded here and in previous works is str (in the underlying representations). I have assumed that any threemember cluster which appears intervocalically is also possible in final position (still, as everywhere else in Part II, at the underlying level). Should additional data from the same dialect infirm this, some redundancy rules should be modified (by adding a boundary, for instance).

CHAPTER TWENTY-TWO

/h/

— ens

RR 42. Χ

ved ; +

h can be preceded only by a morpheme boundary in a lexical entry. It is always followed by a vowel, but this restriction also applies to w and /, and is already described elsewhere in the grammar (RR 26).

PART III

Transformational Phonology

CHAPTER TWENTY-THREE

Introduction

The following chapters present some of the more interesting phonological processes of Haitian Creole. It will be obvious to students of other French creóles that a large part of this transformational component can be used to describe dialects such as Martinique, Guadeloupe, or Dominica. I have given special attention in the discussion of phonological transformations, as I did in that of the morpheme structure (Part II) and in the diaphonie analysis (Part IV), to certain domains which had been under-studied or overlooked in previous research. The analyses proposed for these areas are based on the scrupulous processing of first-hand and of published data with the modern tools made available by current linguistic theory. They represent an effort to provide a better understanding of the deep-seated regularities of creole phonology, as well as a basis for new questions and further research. Other processes have been given less attention, for instance the well-known lenition of high vowels, or the palatalization of voiceless stops before front vowels. The structural description of every phonological transformation (TR) includes a morpheme boundary, since one of the functions of a TR is to describe a combinatory process between morphemes or higher units. A rule whose domain could not include a morpheme boundary would be a lexical redundancy rule, not a transformation. The TR which nasalizes vowels, for instance, cannot operate unless the vowel is separated from a following nasal segment by a morpheme boundary. This absolute distinction between redundancy rules and transformations ensures that exceptional formatives such as /kan/ 'horn' or /leogan/ 'Léogane' are not transformed into *kon and *leogan. A TR may include additional morpheme boundaries at any point in the string of segments. Most transformations apply to strings which may also include a syllable boundary (%) anywhere. They "disregard" syllable boundaries. In other TR's (e.g., lowering and /--deletion) the presence of % is crucial and obligatory. For presentation purposes, boundaries which are not obligatory present in the string will be omitted in the formulation of the rules. When a combinatory process occurs between two words, but not between two morphemes in the same word, the word boundary is obligatory, and will be included in the structural description of the rule. More than half the transforamtions studied in the following chapters are of this type, i.e., are "word-level"

62

TRANSFORMATIONAL PHONOLOGY

transformations (e.g., elision, li-la nasalization, clitic agglutination). Some of these TR's may also function across higher syntactic boundaries, as in vowel nasalization; such facts will be mentioned and illustrated in the text. The final chapter of Part III deals with the ordering of phonological transformations, and presents a few characteristic derivations. The ordering which I propose was deducted solely from rule inputs and outputs, without any reference to their structures. Since rule outputs are almost always based on actual phonetic strings, this ordering is empirical and to a large extent free from the possible arbitrariness of a particular rule notation. Yet it shows a clearcut functional division between a "block" of rules which do not operate across a word boundary and another block, ordered after the former, in which all the rules may involve different words. This cannot be coincidental. Although more rules should be considered, this partial model seems to indicate that the concept of "cycle" is not necessary to describe the phonology of French creóles.

CHAPTER TWENTY-FOUR

V o w e l Nasalization

Robert A. Hall, Jr. observes that "in Haitian Creole, as in other languages with an African substratum (e.g., Taki-Taki), nasalization is notoriously unstable, and in many instances extends partially or wholly over one or more neighboring sounds" (1950a:476). The last part of this observation is interesting because it suggests that these processes are worth studying beyond the morpheme or the word level. I will show that these extensions are in many cases regular, and that the residual non-patterning nasalization merely reflects our limited knowledge of these structures. A native nasalization process which did not exist in French has been created in Haitian across syntactic boundaries (the spaces in the following transcriptions, when pertinent to the discussion): /di/ ->

[dì] ; /sa/ ->

m ap dijo sa w ap / ε kute sa map dì nu /liI

-

-

'He has gone out' 'himself [ba]

ba li kat kob ba mwen /pa/

'Give him four pennies' 'Give it to me' [pa]

pa li pa m I rele I ^

'I'll tell them' 'What are you doing?' 'Listen to what I'm telling you'

[ΐη

li s J ti ίι mem M

[sa]

'his' (Pn) 'mine' [relè]

li rele jo li rele mama jo

'He called them' 'He called their mother'

64

TRANSFORMATIONAL PHONOLOGY

/pu/

->·

[pu]

pu li piïmwë /f3/

'for him' 'for me -

[fD] 'You have to tell them' Ί have to' tell you'

f j wu dijo f j m di wu

Any underlying oral vowel can thus be nasalized across a morpheme boundary by a following nasal segment. As a first approximation, the following TR can account for this fact. TR 1.

Vowel nasalization

[ + syllabic]

[ + nasal] / —- + ( # )

[ + nasal]

This rule is optional, but frequently applied, especially in connected discourse. It gives Creole sentences their characteristic nasal quality. Whether this transformational nasality is due to a Northern Romance evolutive latency or to a contribution of African structures is not clear. It is probable that, as in other areas of creole phonology, a characteristic feature of Northern Romance was reinforced by the presence of a similar feature in West African languages. This nasalization by rule does not apply to consonants and glides: /li muri/ -»· [1 muri] (but *[n muri] ) 'He died', 'He is dead' /li ap maze/ -*• 'He is eating'

[1 ap maze] (but *[1 am maze])

/kaj mwc / 'my h o u s e d

[kaj mwc ] (but * [kaj" mwe ] )

Other utterances reveal more complex patterns of transformationally generated nasality than the rule above would display. What makes its application appear erratic is not only the nasalization of high and high-mid vowels, but also the optional character of the rule. However, a definite pattern emerges from the data. A vowel can be nasalized by a following segment across a simple morpheme boundary, as in the examples above, but also, still optionally, across higher syntactic boundaries:

VOWEL NASALIZATION

I ba klarmct maie / nepJt salapba

65

li / li refemë η ako /1 ale

'She gave Clarinette some food, and whatever she gave her, she would close it [the door] again and leave' neg la di ¡ m pa ¿am soie o 'The man said: I never thought about you' I have found no instance of a vowel which would be nasalized transformationally by a preceding consonant (e.g., m ap νΐηϊ Ί am coming', but *m ap νΐηΐ, η ale 'We are leaving', but *n ale). It should be noted that, because nasalization by rule makes no substantial change in the height of a vowel, the phonetic level displays three nasal vowels which do not exist in the set of underlying phonemes: [e], [ a ] , and [o]. Apart from the systematic nasal vowels /Γ/, /u/, /ε/, / 3'/, and /λ/, any vowel can be nasalized across a syntactic boundary. The result is that all systematic vowels, nasal and oral, can be found in a nasal form at the phonetic level. Superficial [® ] . [ o ] , and [a] are always derived by rule from underlying /e/, /o/, and /a/, [Λ] is always the phonetic manifestation of the systematic phoneme /A/ (normally noted / a/ in this study), whereas superficial [Γ], [ ü ] , [ε ] , and [d] can represent either a nasal or an oral underlying vowel. These facts account for the apparent dissymmetry between the two varieties of a, one oral (deepstructural and superficial) and two nasal, only one of which, [Λ] , is a nasal phoneme. The descriptive and explanatory power of a generative model, even one with strict surface constraints, is manifested in the fact that the same explanation is valid for superficial [Γ] and [ u ] , which derive from underlying oral or nasal segments.

CHAPTER TWENTY-FIVE

li-la Nasalization

Whereas vowel nasalization by rule is regressive in Haitian, the nasalization of consonants is progressive. It is also syntactically labeled: all the examples that I have found through informants and in previous studies concern the nasalization of the determiners li 3sg and la 'the' (sg) and of the personal pronoun li 3sg. The consonant I does not seem to undergo nasalization, progressive or regressive, in any other morpheme: e.g., m se mun leste Ί was born in Lestere' but *m se mîin neste, na lari ja 'in the street' but *nanarija, ti mun jo sal apil 'The children are very dirty' but *ti mun jo san apil. mwe and nu are always nasal, and the rule applies to them vacuously, but -wu 2sg POSS and -jo 3pl POSS must be rejected by the structural description of the rule, since they are not nasalized. TR2.

li-la nasalization + clos [ + nas

]

# I Det PPn

S.D.: 1 2 3 4 S.C. 1 2 3 + nas

— obs

4 +

nas

Item 4 in this rule need not be specified [ + syllabic ] : li and la are the only determiners and personal pronoun (with mwe and nü, already nasal) that begin with a continuant, and syntactic labeling provides the correct sequence. Note also that the only difference between underlying I and n, in terms of non-redundant features, is the nasality specification. Laterality is a redundant feature throughout the phonological component of this language (as it is in French), and it is inserted by a segment rule whenever a [ + closure, — obstruent ] segment is also [ — nasal ] . Moreover, one formative at least shows superficial free variation between η and nasal I (or η with lateral escape): la ~ na 'in'; this is true of the speech of many monlinguals and most bilinguals.

li-la NASALIZATION fami

li

->- famì

m

'her family' N.B.: «/"confirms the nasality of Tin le m t al lada la

67

-*• lem

t al laclaría

farm.

(TR truncation

-> le m t al lada

ri)

'when I went in there' (lit. into it) I have heard lada n, never lada I for 'in there'. This seems to indicate that has been lexicalized as a unit, and is inserted as such into the lexicon by the child who learns his language, since the rule is no longer optional for this particular sequence. 'In it/him/her' and 'into it/him/her' are often distinguished from 'in there' by the use of the free variants ladani, where the rule has been applied, and lada li, where it has not. Like vowel nasalization, this rule functions across any syntactic boundary, even that of an embedded sentence:

ladan

lipa

djab / se jugaso

lije

lipa

djab / s j

gasjriïje

'He's not a devil. A boy, that's what he is' fak fwa l νΐηΐ /1 ¿wen

neg la

fak fwa

I vim

/ η zw en neg

la

'Every time he came, he met the man' si la ge

ta pu wu ti barian / li pra I -*• sil gë

ta pu wu ti banan

/

riïpra

η

'If he has the time for a banana or two, he takes them' me

fak fwa

I ap νΐηΐ / / fate

me fak fwa

I ap νΐηΐ / η

fate

'But every time, on her way here, she would sing' N.B.: The nasalization of / confirms the nasalTin νΐηΐ, and in revÎm in the next example. I revÎnÎ / / ¿wen

neg la ak f i j a

l τενΐηΐ

/ η ¿wen

neglaak

fija

'She came back and she caught the man with the girl' li re feme

I ako

li re feme

η

ako

'She would shut it again' N.B.: The change from / to η is not due to a regressive action of the a in ako; cf. I ako isit, but *n ako isit 'He is still here'. The reduction of li to I is producted by the elision transformations (see chapter 28). Note that the article la in the next citation is affixed to a complex NP whose head is le 'time/when/while/during', which is therefore a noun. lelapdesanla

lelapdesanna

'when he comes down' (time he PROG-FUT come-down the)

68

TRANSFORMATION PHONOLOGY

mun la -*• mun na 'the person' me pa mwe la -+ me pa m na 'Here is my fiancé' (Deictic POSS-lsg the) There are thus two nasalization transformations in Haitian: one for the regressive nasalization of vowels, the other for the progressive assimilation of the consonant in the determiners li and la, the personal pronoun (subject or object) li, and their reduced variant I. As evidenced in all the examples given above, I is never nasalized alone in li and la: the nasalization of the consonant and that of the vowel are always concomitant in the speech of monolinguals and bilinguals.

CHAPTER TWENTY-SIX

Agglutination

The first segment is optionally deleted in two of the Haitian Creole determiners which begin with a glide or a consonant: the singular definite artilce la with its variants na and la, and the possessive adjective wu ' y ° u sg'. The form jo, which manifests both the plural definite article and the possessive 'their', is left untouched by this change. Perhaps because of the confusion which would result with wu, whose truncated form is frequently lowered to o by a harmonizing mechanism (see chapters 28 and 29), as in me o 'your hand', papa o 'your father', and with an *o from jo. li 'his, her' is the only other determiner whose first segment is a liquid or a glide. It does not agglutinate in the dialect of Portau-Prince, but it does undergo the rule in the Cap Haitian area (Northern Haiti), and also in the Antillean creóles (Goodman's classification, 1964: 39). This truncation of li in Northern Haitian occurs in the same environment (post-vocalically) as that of la in all Haitian dialects, where agglutination is context-free when the determiner/personal pronoun begins with a glide. These facts can be captured by the following rule: TR 3.

Agglutination -

syl

"I

J

+ syl null

/

< [ + syl ] >

# ([Det ppn

-

act

- mid < - lab>

A non-syllabic segment which is the first element of a determiner or an object personal pronoun is deleted if it is followed by a non-front, non-mid vowel, i.e., by a, u, or their nasalized variants, li does not seem to be truncated in Southern Haitian, and we must exclude it from the structural description of the rule with the specification that the vowel must be non-acute, since no other reason has yet been found for this exception. Although truncation in the Northern dialect ismore inclusive than in Port-au-Prince, it is still relatively complex, and the use of angled brakets for this rule is necessary in all the French creóles of the Caribbean, since nu 'us, our' is never truncated. The exceptional character of nu may have the same cause as that of jo, since the truncated forms of

70

TRANSFORMATION PHONOLOGY

wu and nu would be very similar in all cases, and identical when u from wu is nasalized by a preceding segment. The use of angled parentheses follows Harms' convention (1968: 66). When the segment is consonantal (/ in this case), it can be deleted only if it follows a vowel and precedes a non-labial vowel (i.e., a). This convention allows us to formalize the claim that the truncation of determiners and that of object personal pronouns are one and the same process, and that the only restriction to this morphological agglutination is phonetic: glide deletion in these suffixes is context-free, whereas the deletion of a consonant can only take place between two vowels. The only other constraints on this rule are the failure of nu and jo to agglutinate for the reasons stated above and, in Southern Haitian, the exceptional case of li: pje + bwa + la ->- pje + bwa + a (-+pjebwa, see chapter 27) 'the tree' fapo

+ la -* jap o + a (

through a splicing transformation;

fapo wa)

'the hat' mai + la

mai + a

{mata)

'the corn' bato + wu

bato + u

'your boat' farct

+ wu+la

-> faret

+ u+a

(Jaretu wa)

'your cart' (lit., cart-2sg POSS-the sg) al fe

rut + wu

-> al fe rut u

'Clear out!' (lit., go do road-your) kaj + la -*• kaj la ( c f .

*ka;a)

'the house' (no agglutination) pul + la -*• pul la ( c f .

*pula)

'the hen' (no agglutination) In French Guyana, agglutination seems to be possible in any environment for the article, but never for the object pronoun li: legliz-a 'the church' (Horth 1949:17), wom-â 'the man' (Goodman 1964:47). The possessive is not l i , but the preposed so, as in all non-Caribbean French creóles.

AGGLUTINATION

71

Nothing in the previous discussion would apply to the creóles of Mauritius and Louisiana. The definite article and the object pronoun li do not agglutinate in Mauritian, judging by the data provided by Richardson (1963), whose list shows one morph only for each personal pronoun: e.g., lagz.rav li 'Fight him', m fin fu li εη bez Ί gave him a good beating'. In Louisiana, the definite article is suffixed, but Morgan's data (1960, 1964, 1970) do not exhibit agglutination, and he does not mention it in his paragraphs on sandhi and personal pronouns (1964:163, 164): diri la 'the rice field', liv sa la 'this book', tzri li 'bury it', vlope li 'wrap it'. Clitic agglutination seems to be a characteristic of Caribbean creóles exclusively, and its mechanism is most interesting in those varieties which extend it to the object pronoun li, since the operation of the rule when two object pronouns follow the verb is relatively complex and will allow me to propose a much more inclusive generalization for this process in all these dialects. In Dominica, for instance, the language has the same rule as Northern Haitian (data from Taylor 1951). It too agglutinates both li and the definite article la, and it does so according to identical criteria, except when li is suffixed to the 2sg pronoun u or to the 3sg i: jo ba i li 'They gave it to him' (lit., they give-him-it), sa ki ba u li 'Who gave you it?'. In these two examples, li does not agglutinate. But it does when it follows another pronoun, and li and la agglutinate to any morpheme ending with a vowel or with r \ ba mê i vit mwe ba jo i mi i mwe per i liv sa a pei a

'Give-me-it-quickly' Ί gave it to them' (them-it) 'There it (he, she) is' Ί am afraid of him' 'this book' 'the country'

The same structures could be found in nineteenth-century Trinidadian (Thomas 1969:90), and in Martinique (Funck 1950:51) with the possessive li followed by the definite article: mwe pakor di i li (Trinidad) Ί haven't told him yet' (I not-yet tell-him-it) ba i li (Trinidad) 'Give it to her' (give-her-it) legliz la u la we lakrwa i la (Martinique) 'the church whose cross you see' (cross-its-the)

72

TRANSFORMATION PHONOLOGY

The rule seems to make morphologically conditioned exceptions based on the person and number of the first pronoun in the pair, and it is so described by Taylor. Actually, it is a morphophonemic rule only insofar as it applies selectively to the article, the possessive adjective, and the personal pronoun. The underlying regularity is phonetic and syntactic, not morphological. I would even claim that it is still less selective than this. Agglutination actually affects all determiners and all personal pronouns, as rule 3 implies. Its restrictions are always the result of a universal, or of an historical, accident. A universal accident such as the fact that the acoustic spectrum of the liquid I is remarkably similar to that of a vowel, especially the vowel a. I believe that it is this vocalism of I which allows its fusion with a preceding vowel (and, not unexpectedly, with the liquid r in the creóles which have retained it in final position). Thus the formant structure oil has led to the agglutination of those suffixes which, by an historical accident, happen to begin with that consonant. It is significant that the only other clitics which are thus agglutinated to the main unit of the phrase are those which begin with glides, i.e., with segments which are characterized by a normally shorter duration and a vocalic spectrum. This of course eliminates from the agglutination process all the proclitics. Not because they are preposed, but because none of them ends in / or w. Among the suffixed "tool words," those which do not agglutinate are the demonstrative adjective (sa) and all the personal pronouns and possessive adjectives which do not begin with I or with a glide, i.e., la and jo (articles), li, wu, and jo (pronouns and possessive adjectives). We have seen above that jo has not been affected by agglutination, in any creole, in any of its functions. Since the cause of this exception is neither syntactic nor morphological, I assume that it must have been phonetic or semantic, and compounded with the historical accident which derived the 2sg-pl and 3pl pronouns and possessives from French vous and Spanish ellos (or Fr. eux). The non-agglutination of jo is therefore a true exception to the very inclusive agglutinating tendency of clitics in French creóles, and its cause may have been the confusion which would have resulted in the discourse in such an essential distinction as second vs. third persons. The possible confusion in Haitian between nu 2pl and nü lpl does not seem to be such a crucial hindrance to communication, and can perhaps be compared to the lack of distinction between we-inclusive and we-exclusive in the Indo-European languages. The non-agglutination of li in all French creóles except those of northern Haiti and the Windward Islands plus Guadeloupe is thus the only fact that remains unexplained at this point. In the data from Dominica, Trinidad, and Martinique cited earlier, agglutination occurs with only one of the pronouns or determiners in each pair. Here again, despite appearances, the rule is not morphologically conditioned: listing which persons are truncated before or after which other sublist of pronouns

AGGLUTINATION

73

will provide the desired output, but it would not do justice to the deep-seated regularities and evolutive tendencies of creole grammar. A closer look at the data shows that the pair of clitics and its structural environment is scanned from left to right by the rule, which operates on the first clitic that displays the required phonological shape. The rule applies only once in a clitic cluster. For instance, in ba li li 'give-him-it', the first li agglutinates to the verb because the environment is adequate. The second li is then left untouched. In bà mwe li vit 'give-me-it-fast' and mwe ba jo li Ί gave-them-it', mwe does not have the required phonetic structure, and jo is the exception to the rule: the following li can be truncated. In ba mwe jo 'give-me-them', neither clitic meets the structural conditions of the rule, and the only possible outputs are the unchanged bamwz jo, or ba m jo (by an elision transformation, for which see chapter 28). All of this, of course, applies to determiners as well as pronouns: e.g., liv sa la liv sa a 'this book' (-> liv sa: -*• liv sa, by a splicing process, see chapter 27). The emphatic la-a (e.g., liv-la-a 'this particular book') is an interesting counterexample to TR 3. Valdman (ms.) proposes a syntactic distinction between lai (the definite article) and la2, a "déterminatif dèi etique." If the analysis is correct, I assume that the structure of these emphatic noun-phrases is Noun+ la2+lal, and that the "deictic determiner" is not subject to agglutination. However, I cannot think of any syntactic or other evidence for the proposed la2, and would rather see in the leftmost la a reduplication of the unmarked definite article. This reduplication is the morphophonemic manifestation of the abstract unit EMPH(asis), and has to be ordered as a transformation after Clitic agglutination, as in the following derivation: bjkj + la + EMPH Agglutination : b o k j + a + EMPH EMPH reduplication : b jkj + la + a The reverse ordering is possible, but it would imply that la, the phonological shape of EMPH, is marked negatively in the lexicon for Clitic agglutination, perhaps because it is neither a determiner nor a pronoun (and although it is a clitic). We would then be back to an analysis which posits two separate formatives with identical matrices in the lexicon, i.e., to lai and la2. Moreover, la2 would be an unexplained exception to the rule which, as we have seen, applies to all clitics with the required phonological shape, and is not influenced by syntactic function. Other facts make this structure even less clearly understood. For instance, Sylvain claims that la+a occurs in fact only after a consonant. If this is correct, the whole problem is eliminated, since there would no longer by any counterexample to the agglutination rule as I have posited it. She also cites a (seemingly)

74

TRANSFORMATION PHONOLOGY

context-free alternation for the plural emphatic structure: pe+la+jo 'the loaves of bread in question', and ρε+α+jo, which has been processed by Clitic agglutination in a regular fashion. To summarize, I suggest that in the case of clitic clustering the regularity manifested by the agglutination transformation is syntactic only inasmuch as it covers the broad category of phrasal substitutes ("articles" and "pronouns"). The actual underlying regularity is phonological, and entirely independent of syntactic functions. It would therefore by erroneous to say that the rule applies "to the indirect object if it is 2sg-pl or 3sg and otherwise to the direct object if it is 2sg-pl or 3sg." In French creóles, agglutination is a phonotactic transformation which applies optionally and, in a group of (one or) two clitics, affects the leftmost unit which begins with a liquid or a glide (except jo). Since the rule is so inclusive, the term clitic in this description need not be more than a broad cover label for all determiners 1 and all personal pronouns, regardless of their position relative to the head of the syntactic unit. Of all the French creóles, only those spoken in southern Haiti, Guadeloupe, and the Windward Islands, for a reason which is still unknown, have made an exception for the form li, and can be considered more conservative in this respect. It is worth noting that the parallelism between ba u li 'give-you-it' and ba i li 'give-him-it' provides additional evidence for positing an underlying wu, although the truncated form u is statistically more common in connected discourse, and is probably in the process of being lexicalized in Haitian Creole by progressive elimination of superficial wu, just as i from li must have been finally lexicalized as the sole subject form in Dominican Creole. The tendency for clitics to agglutinate to their focal unit is universal in French creóles. The process has clearly been expanding since its inception, and it is playing a vital part in the development of a more complex inflectional morphology. The agglutination of pronouns and determiners is therefore as widespread and deep-seated a characteristic of French creóles as nasalization, and the residual exceptions which I have noted are probably less exceptional than symptomatic of the need for further research based on direct field-work. The agglutination of post-posed clitics through a suffixing stage may be a universal tendency (see Tinelli ms6), but an interesting aspect of this evolution in French creóles is that it is an original development in the contemporary Romance family (if one believes that creóles belong to the base-language family), just as Rumanian innovated by agglutinating the post-posed Latin demonstrative, as shown by the following derivations proposed by Bourciez (1967:580): LATIN

OLD RUMANIAN

MODERN RUMANIAN

capra-( il)la caprae-(il)laei vulpi-(il)laei

*capra-lla *capre-ei *vulpi-ei

capra-a capre-i vulpi-i

'the goat' 'the goat' (Dat.) 'the fox' (Dat.)

AGGLUTINATION

75

On the other hand, modern French now seems to be creating pronominal prefixes, although suffixation is also at work (e.g., Std. Fr. "dis-le lui," nonstd. diz-i 'Tell him'). This tendency is particularly strong in social dialects and a colloquial style which are not yet considered standard: STANDARD FRENCH

NON-STANDARD

D veut sortir. Elle veut sortir. Il ne veut pas. Elle ne veut pas. Il n'y a qu'à le lui dire.

i-νφ sortir a (e/)-νφ sortir i-νφ pa α(ε/)-νφρα

'He wants to go out' 'She wants to go out' 'He doesn't want to' 'She doesn't want to'

jakai-dir

'Tell him'

CHAPTER TWENTY-SEVEN

Splicing

When the singular definite article is truncated after a vowel, its remaining segment (a or a) can be linked phonetically to that vowel in such a way that they fuse into a non-syllabic transition which is a composite of the feature specifications of the two vowels. Since the resulting linkage has all the pertinent features of / and H> when the vowels preceding the article are respectively front and labial, I will note it as if it were a segment. Such is the usual way of presenting "epenthetic glides." In this case, however, as in most instances of so-called glide insertion, the resulting portion of the phonetic chain is normally much shorter than the surface form of an underlying glide, or of a deep-structural vowel (usually i or u) which has lost its syllabicity by transformation. This transition may develop, after a few generations of speakers, into a full-sized segment, and be included as such in the lexical representation of the morpheme (e.g., kanowa, kanoßa from kanoa 'canoe' in some dialects of Costa Rican Spanish; krejjl, tejat, raji, lewJ in Haitian Creole, from French krejl, teatr, air, leJ). This is unlikely to happen in the form under study since the process is due to a synchronic transformation, at a syntactic boundary. Although this transition can be viewed as a phonetic segment (which of course no instrumental analysis could corroborate), such a view in fact represents the opposite of what actually happens. The effect of the splicing rule is not to separate the two vowels by a third segment, but on the contrary to cancel the solution of continuity between them by transforming part of each vowel into, quite literally, a glide from one to the other. TR4.

Splicing rule

+ syl + high a act lab . .-a 1 S.D.: 1 2 3

# I Art

2

[ + syl]

3

SPLICING

S.C.:

r

77

3n

l — syl a act •a lab

iodi a ¿odila fia fila madam ni a -*• madSmnìla maie a -*• maiela tu a tuwa prezida nîi a - prezidanuwa wu retire powa wu retire po a w Sapo a fapo a

'today' 'the girl' 'his wife' (wife-his-the) 'the food' 'the hole' 'our president' 'you skin it' 'the hat'

This transition is not obligatory, and it is most often heard when the vowel preceding the article is high (/, u, and their nasal homorganics). It occurs very seldom when this vowel is low (e.g., rivje a 'the river', kaj me a 'at the nun's home'), and this possibility has been left out of the rule. Splicing does not have to occur when the two vowels in contact are identical (aa or aa), which is of course additional evidence that the process is not epenthesis. Splicing between a or a and the truncated article is sometimes complete and, depending mainly upon the tempo, results in crasis. The following are examples of partial braiding and crasis: ragbwa:, ragbwa tut bagaj sa:, tut ba: j sa prezidaa, prezida:, prezida

'the forest' 'all those things' 'the president'

Reduction of two consecutive a's or i s to one syllable is common in connected discourse, regardless of syntactic labeling: le ofìsje a ap pale avek li -*/ε ofìsje a pale avek li 'when the officer is speaking to him'. Note also that the agglutination of the article (TR 3) does not occur after glides, which are thus treated as consonants, kaj la 'the house' cannot be reduced to *kaj a.

CHAPTER TWENTY-EIGHT

Elision and Clustering

Elision, like nasalization, is pervasive in Haitian Creole. It is optional, but in terms of performance, the rules are used in ordinary connected discourse almost every time the structural descriptions are met. Pronoun reductions have been mentioned by all students of Haitian, but Goodman is the only one who has suggested that the elision processes which apply to personal pronouns, some aspect markers, tense markers, and the negative particle pa, may be separate manifestations of a more general and unified pattern (1964:40-41). His aim however was to find generalizations across language boundaries rather than within each linguistic system, and he did not attempt to formulate rules of any kind for elision as one complex process. I will consider here the various elision rules of Haitian Creole, and what possibilities exist of conflating them. I will also analyze other, little documented areas of creole phonology, and provide evidence for the existence in Haitian of a battery of functionally related transformations which includes elision, syllabification (see chapter 30), and vowel harmony (chapter 29). Subject personal pronouns followed by a vowel are reduced to their first segment: a consonant in the case of / m w e / lsg,/li/ (or [nî]) 3sg, and / n u / l-2pl, a glide with / w u / 2sg and / j o / 3pl. As a first approximation, the following transformation accounts for these changes: TR 5. [ -

ens

]j

2

-

null /

[ -

syl]

-

I mwe av al gade j o / ->· [m av al gade jo] Ί will go and watch them' Ili ap vTn pi pwes/ -»· [1 ap vih pi pwes] 'It is becoming thicker' /nuapralwe/^ 'We shall see'

[napralwe]

J# PPn

[ + syl]

ELISION AND CLUSTERINGS

79

/jo ap kwit patat jo/ -*• [j ap kwit patat jo] 'They are cooking the potatoes' /we ede li/ -»· [w ede li] 'You help her' / wu ase ζ 3ηε li/ ^ [w asez ϊηε 1 ] 'You season it' Valdman would rather make the subject personal pronoun elision rule obligatory before a vowel-initial verb particle: m ap vìrii, but *mwe ap νΐιίΐ Ί am coming' (personal communication). However, the full forms do occur in that environment occasionally: mwz ap vini Ί am coming', li ape soti 'He is going out'. This happens rarely in fast discourse, but with greater frequency as the rate slows down. There is an apparent correlation, which has not yet been tested, between the various applications (in terms of performance) of elision rules to several elements within the same syntactic unit. The possibilities would clearly cover a continuum, from an optional blanket application of the elision rules to all structures between given syntactic boundaries to a total performative independence: pe ki te ape tan nu a (no elision, syllabicated structure) ρε ki te ap tan nua ρε ki t ap tan nu a (*?) ρε ki te ap tan nu a (*?) pe k tap tan nùa (blanket application) 'the priest who had been waiting for us' mwz te ale \νε m te ale we mwè t ale ννε mwe tal we mwe te al we m t al we 'I went to see'

(no elision) (*?) (*?) (*?)

(*?)

(blanket application)

A study of this type might yield grammatical utterances at both ends of the spectrum and progressively less grammatical structures in between, and thus provide a phonological, performative illustration of Ross's concept of grammatical "squishiness." On the other hand, there may well be some conditioning elements for the functioning of the three elision rules, perhaps even for a larger

80

TRANSFORMATION PHONOLOGY

block of transformations which would include Clitic agglutination and ¡i-la nasalization as well. If so, the descriptive device which Labov calls "variable rules" might be the most appropriate mechanism available today to deal with the above data. Whereas / r a w e l , /li/, and / n u / are also reduced before a consonant or a glide, the subject pronouns / w u / and / j o / can undergo elision only in a prevocalic position. We thus have to change rule 5 to the following transformation, which describes the various environments in which personal pronouns suffer elision: TR 6. Elision 1 [ - ens ] j 2

[ + ens ] null / < [ — ens ] > - 1 # [ / - < [ + s y l ] > PPn VP

Subject personal pronouns are optionally reduced to their first segments. The transformation can apply only before a vowel if the first segment is nonconsonantal, but before any segment if the pronoun begins with a consonant:

lem te rive legliz / m te preske areta

(/mwè / ->· [m] )

'When I arrived at the church, I was almost late'

mswivli (J mwE /

[m])

'I followed him'

m νΐη reme I 'I ended up liking him'

m we 5" bagaj Ί saw something' (lit., I saw a thing)

m made o padJ 'I beg your pardon'

epi I feu sistem / u gwmaz (/li / 'And he built a system, a gear'

le I mete kan na 'When he placed the cane . . . '

epi li par et deva nu 'And he appeared before us'

[1] )

ELISION AND CLUSTERING

81

I km / ε sa: 'He can do that' nrivehëf (/nu/ [n]) 'We arrived at Hinche' η sjti 'We went out' /wuweki/oj/ *[w we ki/jj] 'You saw something' I jo pr ale zeremi / ->• *[j pr ale zeremi] 'They are going to Jérémie' This process creates consonant clusters which, in the same environment, would seem foreign to the phonology of contemporary French, as in the following Haitian examples. mwe pr ale na ma/e/ -*• [m pr ale na ma/e] Ί am going to the market' Imwe maze apil zodi ja/ ->- [m maze apil zodija] Ί have eaten a lot today' /mwe te ale we/ Ί had gone to see'

[m t al we]

/mwe fatige ak bagaj saa/ Ί am tired of this'

[m fatige a baaj saa]

/lisjti/ [1 SDti] 'He has gone out' / nuz it nu te fe klere / ^ [nuz3t η te fe klere ] 'We had made some tafia' Clusters involving a sonorant followed by an obstruent do occur in some regional varieties of French (for instance in Southern France), but always after a vowel and never, as they do in Haitian Creole, in the initial position of a phrase or sentence. Moreover, in Southern French, 2 clusters of this type always

82

TRANSFORMATION PHONOLOGY

follow a semi-nasalized vowel, i.e., a vowel with a non-nasal first half or two thirds. The nasal consonant is also homorganic with the oral stop or fricative which immediately follows it. These restrictions clearly show that the string semi-nasal vowel + nasal consonant + homorganic obstruent is the result of a low-level combinatory process in Southern French. Historically, the nasality of the consonant shifted onto the preceding vowel, which became semi-nasal in the southern and northern varieties of French, and later fully nasalized in Standard French. The result was twofold in the latter: the consonant cluster was reduced to its oral stop when the nasal disappeared (e.g., Lat. cantare, Std. Fr. [/ate] ), and the nasal vowels have since been included in the lexicon as systematic phonemes of the language, with nasality a distinctive feature which is manifested over the entire length of the phonetic segment.3 In a synchronic description of these clusters in Southern French, the nasal consonant can also be treated as a prenasalization of the oral consonant, i.e., as a partial shift of nasality from the vowel onto the obstruent. In the following examples, the nasal vowels of Southern French are in fact phonetically nasal only in the last half or third of their duration. Sth. Fr. Std. Fr.

[iletjmbe] [il e tabe] 'He fell'

Sth. Fr. Sth. Fr.

[ze SrjkDr Shtady son 017kl amba] [ze akor atady son 5kl aba] Ί heard his uncle again down there'

On the other hand, these clusters are common in some African languages, even in phrase—or sentence—initial position, just as they occur in Haitian Creole when mwe, li, or nu is elided: e.g., Fanti mpá 'mat, bed', ntd 'twins', nkda 'debt', nsd 'hand' (Weimers 1946:13), Tshiluba (Congo) nzambi 'God', Kongo (Angola) mbwanga 'a powder used as a cure for headache' (Turner 1949:201). Yet it would be hazardous to cite the African "substratum" as the sole influence on the phonology of modern Haitian Creole. Consonant clusters, even in the initial position of a phrase or sentence, are not at all foreign to the phonology of contemporary Standard French, where consonant clustering has been a continuous diachronic phenomenon. A case could in fact be made for the existence of a North Gallo-Romance latency for consonant clustering. This would then be a deep-seated shift which has occurred in the two dialects: Standard French and Haitian Creole. The French examples will be limited to consonant clusters which display the same structure and environment as the Haitian clusters under study: a sonorant followed by an obstruent as the initial segments of a word, phrase, or sentence. Note that in French, as in Haitian, these

ELISON A N D CLUSTERING

83

clusters are the result of an optional elision transformation: ms/0 'Mr., Sir', IbalJ 'the ball', nkurpa 'Don't run', Itapi 'the carpet'. Whether or not the first consonants in these strings are syllabic does not concern us yet (see chapter 30). The interesting fact is that the French phonological structures which we have here are identical to those which occur superficially for Haitian Creole in the same environments. These clusters are the result of a low-level elision process in both modern French and modern Creole, and their existence in the latter cannot be attributed exclusively to a West African substratum. /mwe/, /nu/, and /li/ can also be elided after a vowel when they are possessive adjectives or object pronouns, /jo/ and /wu/ are reduced, as we have just seen, only when they are subjects. TR 7.

Elision 2 (OPT)

[-ens] j 2

^ null /

[+ syl]

] # [ [+ ens] V PPn Ν PosAdj

[prezidaa voje m] (/mwe/ — - > [m] ) The president sent me' /li vin reme li/ [1 vin reme ηϊ] -»• [1 vin re me η] 'He fell in love with her' [m ralo te zamTm na] Ί told my friend' The reduction of the two formatives mwe in the last example is subject to different conditions: the first mwe suffers elision simply because it is in the subject slot, whereas in the second case elision is phonologically conditioned: mwc follows a vowel.

CHAPTER TWENTY-NINE

Vowel Harmony

Ans (1968:48, 90) notes that wu (2sg) is reduced to its first segment after a vowel when it represents an object personal pronoun or a possessive adjective. This would introduce a noticeable lack of symmetry in a process which is otherwise regular and systematically dependent upon the phonological structure of the language: jo, the only other form in this series with the structure glide ( [ - s y l , - ens]) + vowel, suffers no elision in similar environments. Ans goes even further in positing a syllabic w in the following strings, but he does not explain how the syllabic w could be different from the vowel u: / l i t e b D w u / -> [ l i t e b s w ] 'He had kissed you' /ramase tut zafe wu/ -*• [ramase tut zafe w] 'Gather all your things' In the model used for this study, syllabicity (or vocalism) is in fact the sole feature that distinguishes w from u, so that a "syllabic w" could be nothing but u, and w could circularly be called a non-syllabic u. The final segment in each of the examples above is actually the output of the truncation of wu, and not that of an elision transformation. In connected discourse, the u resulting from this rule will normally undergo a low-level height-harmony rule under the influence of the preceding vowel. Vowel harmony applies to u, optionally, but with great frequency in natural speech, whether it is the possessive adjective or the personal pronoun (subject or object): /màzewu/ -»· [maze o] •your food' Ipapawu/ -*• [papa o ~ papa a] 'your father' [m ap kute wu] [m ap kute o] Ί am listening to you'

VOWEL HARMONY

85

[sa pu m f e ave w u ] -> [sa pu m f e ave o ] -*• [sa pu m f e ave 3] 'What am I going to do with you!' [m ap ba d mun avek u] Ί am giving you people (to go) with you' /eske wu maze li/

[eske o maze 1] -*• [ e s k o m a z e l ]

'Do you eat it' /sawufe/ -*• [ s a o f e ] -*• [ s a o f e ] 'What are you doing?' A complex pattern emerges from the data, whereby u, the reduced form of wu, changes optionally to o after any vowel but i, u, ando. When the preceding vowel is [—high] (i.e., e, J, or a), u may be, and often is, lowered to J. These processes can be captured through two separate but ordered rules, both optional: T R 8.

Vowel harmony 1 + — — .+

syl nas mid lab

- high + mid - lab

[ + mid ] /

# [

-

PPn Det

The pronoun or determiner + u + is changed to + o + following a non-high vowel ( e , a, y ) or the unrounded mid vowel e. T R 9.

Vowel harmony 2 — nas + mid + lab

[-high]

/

[-high]

# [

-

Pn Det

The pronoun or determiner + o + is optionally lowered to + d + after a non-high vowel. Another vowel harmony transformation may affect, in turn, the final vowel in certain monosyllabic formatives before ( w ) u , after it has been changed to j . I have not given the underlying strings in the following examples, but only surface forms and intermediate strings that can occur at the superficial level if no other transformation operates.

86

TRANSFORMATION PHONOLOGY

m ap ba omun avek u -*• map bj jmun avek u sa Y fe s? J fe s Χ fe leu rive -> le J rive h J rive h . rive Vhen you get there' le u fin epise li wü mete I buji -*• le 3 h: fin epise li wu mete l buji Once you have spiced it, you boil it' tut sa u vie tut sa o vie -» tut sa ) vie tut SJ Y vie -»· tut SJ: vie 'everything you want' The same type of combinatory process occurs in morphemes like si 'if', di 'say', w e 'see', before wu and ju (indefinite article) or their truncated variant u. A labial assimilation changes si, di, w to su, du, and w, as in: [ s i w u z w c n n î ] -*• [ s u w u z w e n n ï ] -*• [ s u : z w e n n T ] If you meet him' /mwe w e j ù b a g a j / [mwe iïbagaj] [ m w o a b a a j ] -*• [mw3T:ba:j] Ί saw a thing'

[m we j b a g a j ]

[m di (w)u se ju neg ki k ï t pwogre] - * • . . . -*• [ m d u s 5"negkikDtpwogre] 'Let met tell you, he's a man who's against progress' In the last example, ju undergoes the same transformation as wu. It is first reduced to its vowel (although for ju truncation is independent of the phonetic environment), then lowered after a non-high vowel. This change also occurs with great frequency in connected discourse: /sejugraneg/ -»·...-»· 'He is a great man'

[sïgraneg]

/ g e j c j u g e p ki m nde mwe / -*• . . . Ά wasp stung me'

[g ï g e p ki mode m ]

The shortened morpheme u, from wu, can also be lowered to o, even in the absence of any conditioning environment, and a preceding high vowel may in turn be lowered when it is in contact with the allomorph o, as in:

VOWEL HARMONY

87

/siuretiïn ε / - » · . . . - • [se u retune] (or, [ s u w u ] -»• [su:] -»· [ s u ] ) •if you come back' The reduced forms of wu are thus derived by a truncating rule followed by the optional application of one or two vowel-height harmony transformations, rather than by elision. The resulting vowel still carries the stress if it is in phrasefinal position. Papiamentu offers numerous examples of vowel harmony, not only for the nasality feature (see above, Part I), but also for those of height, frontness, and labiality, as in: Pap.

rospondé so sodé, so sedé nogosi midi sigi kustumbra

Sp.

responder suceder negocio medir seguir costumbra

'reply' 'happen' 'business' 'measure' 'follow' 'custom'

To my knowledge, Chataigner is the only creolist who mentions vowel harmony. He has discovered alternations in Portuguese-based Kriôl such as i ~ e 'to be' or ku ~ ko "which', based on the features of the preceding or following vowel across a syntactic boundary: i ditu 'it is a saying', kel e ditu 'that one is a saying', ku bu udja 'which you looked-at' ko no udja Vhich we looked-at' (1963:50). Chataigner adds that, in his opinion, "similar phenomena can be found in wolof: bdgu ma Ί don't want to', and bagu mo ko Ί don't want it' ". The Kriôl harmonization is progressive in the first pair of examples, and regressive in the second pair. More data would of course be needed to uncover even part of the complete rule.

CHAPTER THIRTY

Syllabic Consonants

Syllabicity has usually been of more interest to the africanist than to the student of Indo-European languages. Syllabic consonants are common in many West African languages, where they cannot be overlooked, as they often are in European languages, when they carry a tone. Weimers states that in Fanti consonants are without tone, while vowels are syllabic and carry tones. But r, m, and η can be used as vowels, and in general consonants "are vowels when they have tone" (1946:14-15). Syllabic m and η are found in such forms as mpá 'mat, bed' or ntá 'twins', while they are non-syllabic in mi-mpd 'my bed' and minti Ί don't hear'. These forms and Weimers' different treatments of m in mpá and (mi-) mpá show that he gave the concept of syllabicity a phonotactic definition. Syllabic consonants are also present in Indo-European languages, and some scholars have studied them at length. In his English Pronunciation: 1500-1700, Dobson devotes nearly thirty pages to the evolution of m, n, I, and r . After a brief review of the phonetic circumstances in which a consonant may become syllabic, he writes that "the evidence of the orthoepists makes it obvious that syllabic consonants were a commonly recognized feature of sixteenth- and seventeenth-century pronunciation (1957:889). Yet syllabic consonants have been discovered in French and Portuguese creóles but not, so far, in English Creoles, with the exception of Krio (see infra). For Haitian Creole, Ans (1968) proposes syllabic m, n, and/, which is plausible in view of utterances such as m s)ti Ί went out' or η te /ate asam 'We had sung together'. However, his treatment of m, n, and I (and w, for that matter) as syllables was prompted solely by the fact that each of these phonological segments can also be a complete superficial morpheme in the language, and Ans treats them as a cas d'espèce, or exceptions to the "regular" syllabic patterns of French creóles. Such a description is an ad hoc device which discards the problem but does not do justice to the real structure of the language. On the one hand, k is also the reduced form of a morpheme (ki, subject relative pronoun), but it is never treated as syllabic. On the other hand, Ans grants syllabicity to m, n, and I only when they stand for higher grammatical units, and claims that they are always syllabic in that case, whatever their phonological environment. The syllable is thus given a syntactic definition in French Creole, and the only criterion for determining the syllabicity of those sonorants is syntactic labeling.

SYLLABIC CONSONANTS

89

Only sonorants (i.e., [ — obstruent] segments) can become syllabic in Haitian. For instance, k can never be syllabic, even between consonants (e.g., [g J gep ki m Dde m w e ] , but *[g 3 gep k mode m] ). It is also non-syllabic when in contact with a vowel: [sa k pase], [fi k sdì ale a] 'the girl who just left', m, η, and I become syllabic when they are followed by a consonant and preceded by a sentence boundary or another consonant: /mwe ape zwen tu pu mwe ratre/ Til find a hole to get into'

[m ap zwêh tu ρ m ratre]

[m remake m te sepatik msje apil] Ί noticed that the gentleman liked me very much' /sa pu mwe f e avek wu/ [sa ρ m f e ave d] 'What am I going to do with you?' /li pa ale zwe paske li travaj / -»· [1 pa ale zwe pask 1 travaj ] *He doesn't go and play because he is working' /nuz 3t nu te f e klerh / -*• [nuz Dt η te fe klere ] 'We used to make tafia' /natirelma/ -»· [ntirelma] 'Of course!' /apil mD/e/ [ampil m a / e ] 'Very much so, my good fellow!'

[mpil m a / e ]

On the other hand, sonorants remain non-syllabic when they are preceded or followed by a syllabic segment, even when the sonorant represents in itself a morpheme or a higher syntactic unit, as in pu I sjti 'so that he might go out' (vs. for instance ju pul SDti 'a hen ran out'). TR 10. Syllabification

#

ob s fri J

[ + syl]

— [ + ens]

/ [ + ens ] ( # )

90

TRANSFORMATION PHONOLOGY

Rule 10 could theoretically apply to ñ and r, which are also sonorants, and to w and /, but these segments are never found in these particular environments at any level of the grammer (lexical component, intermediate or surface structures). Chataigner has uncovered an interesting type of "vowel" harmony in KriÔl, by which a syllabic r anticipates the timbre of the vowel in the following syllable. "Ainsi en brdadi, r revêt comme une teinte a, en prsisa une teinte i, en prmete-m une teinte e, en brguna une teinte M " (1963:50). He explains this harmonization by the absence of a schwa in Kriôl, and compares this variable syllabic r to those of non-standard Portuguese (e.g., prmcta, for either permete or promete in Standard Portuguese) and of the Poitevin patois in France, both of which have an r with a schwa-like formant structure. In English-based Krio, the same results have been obtained, although by different means: an English syllabic consonant loses its syllabicity in Krio by the realization of the preceding etymological vowel, modified according to the rule of creolization. The following Krio examples are from Berry (1961:10, 11). /I ~ 1/ 'apple' 'devil' 'angel'

KRIo

/VI / /apul/ /debul/ /endzel/

/n~n/ 'garden' 'addition'

/Vn / /gadin/ /adi/Dn/

/ m ~ 1/ 'rheumatism'

/Vm/ /romatizim/

Yet Krio has tow syllabic consonants, m and rt, whose occurrence is limited to preconsonantal position after pause, i.e., to a subset of the environments where Haitian sonorants are syllabified. Within lexical item boundaries, the redundancy component of Haitian (RR's 30, 31, 32) restricts the composition of consonantal pairs in initial position, and that of triple clusters. For instance, no sequence of two fricatives can begin or end an underlying formative. If a morpheme begins with a consonant, the second segment must be a vowel, a liquid, or a glide. The initial unit of a morpheme cannot be a glide unless it is followed by a vowel. If the segment in the morpheme's initial position is a liquid, it must be followed by a vowel or a glide. The third member of a triple cluster cannot be nasal, and the first must be + fricative if it is an obstruent. Thus the environments which trigger the syllabification of a consonant can be created only by transformations (such as elision).

SYLLABIC CONSONANTS

91

Yet, even in the examples given above, the syllabic status of these sonorants is still open to controversy, precisely because the only segments which can acquire it are sonorants. Should we note for instance a syllabic I in French [lvevu] from /bvevu/ 'Get up' (see Pulgram 1970:100), while s is non-syllabic in [skatyv0] "what you want?' One reason for doing so could be that the group Iv, in initial position, is only created by rule (elision), and does not appear in the underlying forms of the lexicon, whereas sk, like st or sp, is also a deepstructural sequence (as in skalzt 'skeleton', ski 'ski', stabl 'stable', spiral 'spiral'). Thus the liquid I could become syllabic, but the obstruent s remains non-syllabic in all cases. On the other hand, another obstruent, / , is very rarely followed by another obstruent at the beginning of an underlying French morpheme (as in ftimi 'Northerner'). Yet / is extremely frequent in this superficial environment (e.g., fkrwabje Ί think so'). In Haitian, the syllabification rule is in any case isolated from other transformations, except Elision, which feeds it. It has no influence on the lowering of mid-vowels, since its structural description involves only consonants and boundaries. The following derivation illustrates the operation of this block of transformations. "Surface structure" Syllable definition Boundary shift Vowel adjustment Elision Syllabification Erase all % and # boundaries

/po + li + di/ po% + li% + di N.A. N.A. po % + / % + di N.A. poldi 'Its skin is hard'

It is evident that the syllabification rule does not in this case apply to I, which is thus represented in the phonetic string as a consonant, even though it is a whole morpheme in itself. The string reaches the phonetic level as a continuum, where the two syllable "peaks" are marked (by the feature [+ syllabic] ), but not their boundaries.

CHAPTER THIRTY-ONE

Particle Elision

Several morphemes other than personal pronouns and possessive adjectives are subject to the optional deletion of their final vowel: the subject relative pronoun ki, the preverbal particles ape PROG, te ANTERIOR TNS, the form pu 'for or INTENTION (which in my data shows elision after mwe only), eske 'yes-no question', and paske 'because'. /sa ki pase/ -*• [sa k pase] "What's happening?' /kiles ki ape vTnT/ 'Who is coming?'

[kiles k ap vint]

¡ge apil mun ki ape gade/ -*• [ge apil miïn k ap gade ] There were many people who were watching' /fi ki sat ale la/ [ f i k sot ale (j)a] t h e girl who just left' /sa ki vie kola/ [sa k vie kola] 'those who wanted cola' I /ate pu mwe I 'Sing for me'

[/ate ρ mwe ]

When preceded or followed by a vowel, ki can have its final vowel elided. It cannot be reduced betweeen two consonants, nor in the initial slot of a sentence, i.e., as part of an interrogative substitute: [ki sa wu fe] "What are you doing?', but not *[k sa wu f e ] . (ki may be deleted entirely in that position: [sa wu f e ] . ) The future particle va is truncated to a (unless it is the emphatic ava) after a non-syllabic segment. This segment can be an elided personal pronoun. va is thus subject to rules which are different from elision, although they reflect a similar tendency to open syllabication, te and ape can be reduced when they are in contact with a vowel.

PARTICLE ELISION

93

/ j o pa te evite ipil m un/ -»· [jo pa t Sate apil miïn] 'They had not invited many people' /mwe ape maze Ii kufie ja/ -* [m ap maze 1 kufie ja] Ί am eating it right now' / paske wu ape ede m we I [paske w ap ede m ] 'because you are helping me' /mwe te ape baj li ju lesa/ -»· [m t ap ba 1 j ù l e s j ] Ί have been giving him a lesson' /mwe te gade li/ ->• [m te gade 1] (but *[m t gade 1] ) Ί have watched her' /mwe va a/te ju burik/ -»· [m a a/te ju burik] Ί will buy a donkey' ape takes the lexically conditioned form apr before the verb ale. apr and its variant pr (after a syllabic) are thus syntactically independent of ape, and do not undergo the same rules: li pr ale ~ I apr ale ~ I pr ale H e is leaving', eske and paske lose their final vowel only before a syllabic or before a sonorant followed by an obstruent. This elision, like all the others, is optional, but usually applied when its S.D. is met. /li ape /ate paske asefi kota/ [1 ap /ate pask asefi k î t a ] 'She is singing because Asefi is pleased' /li ape tan jo paske jo bezwe laza la/ -»· [1 ap tain jo paske jo bezwè lazà la] He is waiting because they need the money' /eske wu ape S3ti/ -» [eske w ap S3ti] 'Are you going out?' /eske li k5ta/ 'Is she pleased?'

[ eske 1 kata] ->• [esk 1 k Jta]

The verbs kJnz Tcnow', ale 'go', mete 'put', and a few others also participate in the elision process. They have been labeled here VE (eliding verbs). Hall (1953) adds the verbs s iti 'go out', vim 'come', and firn 'finish' to this category. The three forms sdì, νΐη, and fin are also used as aspect prefixes, with the approximate meanings of perfective for fin and recent past for the other two.

94

TRANSFORMATION PHONOLOGY

Another morpheme, met 'may', has an open ε whereas the verb mete ~ met has the closed variety. This morphological specialization parallels that of bj and kote 'side' (nouns) as locative prepositions, and kj 'body', kadav 'corpse', and tet 'head' as reflexive prefixes. In this description, the class VE includes the true verbs and excludes the aspect prefixes sot, vm,fin, and the modal met. To summarize the various elision phenomena which occur with formatives other than personal pronouns and possessive adjectives, ape and VE's can lose their final vowels before any segment (vowel or sonorant). eske and paske only if they precede a syllabic. A close look at the data shows that the conditioning element of these transformations is not merely lexical, as the above description might suggest, but phonotactic. These elisions are structured in such a way that no cluster of three non-syllabic segments can result from their operation. They form with the syllabification rule a functional constellation of transformations. Thus ape and VE, in which the last consonant is always followed by a vowel, can be reduced before any phonological unit, eske and paske, which already include two consecutive consonants, cannot undergo elision unless they are followed by a syllabic, ki and te are reduced to k and t if they are preceded or followed by a syllabic segment.4 Morphemic labeling is still necessary to make the rule operate, since not all the formatives which meet the required S.D. lose their final vowels, but these labels are merely additional restrictions and no lexical subclassification is needed. Rule 11 deletes the final vowels of ape and VE before any segment within sentence boundaries, and those of ki and te after a syllabic. TR 11. [ + syl ]

null / [ + syl ] ( + ) [ - syl ]

+ E

The subscript E labels all the morphemes which can undergo rule 11, and rule 12, which applies to te, eske, paske, and ki before a syllabic: TR 12. [ + syl ] ^ null / [ - syl ] ( + ) [ - syl ]

+ [ + syl ] E

The two rules can be conflated into one phonological transformation, Particle elision.

PARTICLE ELISION

95

TR 13. Particle elision

[ + syl ] -> null /

( +)

[-syl]

+ E

[ + syl ] [ + syl J I




The angled brackets state that if the final vowel of an eliding formative is preceded by two consonants, then a vowel, or a sonorant preceding an obstruent across a morpheme or higher boundary, must follow the particle for the rule to operate. Thus, if the last consonant of the eliding morpheme is preceded by a syllabic segment, the final vowel can be elided regardless of what follows it. I have adopted here the convention that longer expansions precede shorter ones in a synchronic derivation. TR 13 is therefore the conflation of five rules whose environments are: 1)

[-syl] + [-syl]

+ [ + syl] E (ki, pu, and te, after a non-syllabic)

2)

[-syl]

[-syl]

+ [ + syl ] E (eske and paske before vowels)

3) (eske and paske before m, n, or I, followed by obstruent) 4)

[ + syl ] + [ - syl ]

+ E (ki, pu, and te, after a syllabic)

5)

[ + syl ] [ - s y l ] (ape and VE)

+ E

96

TRANSFORMATION PHONOLOGY

The reduction of ke (complementizer or object relative pronoun) has not been treated here. Its use is a structural interference from French, which I have not found in the speech of monolinguals. An example of such gallicism is /mwe te va swete ke wu /wazi ju b t mun/ •+ [ m t a swete k o /wazi jiï b t mùn] Ί wish you had chosen someone else', paske and eske can be reduced before a consonant if it is preceded by a sonorant, which then undergoes the syllabification transformation. Thus in the derivations of [pask 1 te dì* m] 'because he told me' and of [esk m met k3mase] 'May I start?'. I paske li te d ì mwe / TR 6.

Elisioni:

paske l te di mwe TR 7.

/ eske m we met k 3m ase /

eske m met kornase

Elision 2:

paske l te di m

Not applicable

TR 13. Particle elision: pask l te di m

esk m met

kJmase

TR 10. Syllabification: [pask 1 te d ï m]

[esk m met kornase]

CHAPTER THIRTY-TWO

Syllable Definition

The structure of the syllable intervenes in at least two phonological transformations in Haitian, r is deleted when it occurs at the end of the syllable, and oral mid vowels are lowered to low-mid in closed syllables. The syllable would still not have to be defined if these constraints were deep-structural redundancies or characteristics of only a few lexical entries. The following alternations show however that Haitian defines its syllables transformationally and in terms of boundaries, since an underlying r is deleted, and a vowel lowered, in certain phonotactic environments, i.e., in phonological strings produced by the syntactic component in correlation with the underlying shapes of lexical entries. SUPERFICIAL

UNDERLYING

[male ~ maler+e] [ρε ~ per+e] [2we+t ~ zwe] [paz ~ poz+e] [ / ο ~ /od-tje]

/maler/ /per/ /zwe/ /ροζI //od/

'misfortune, poor' fear, cowardly' t o y , to play' 'a break, rested' Ίιοί, saucepan'

A syllable boundary is inserted before a word boundary ( # ) , before the second consonant in a cluster or before an intervocalic consonant or glide, if this cluster, consonant or glide is followed by a vowel. The phonetic alternations above show that the syllabic unit transcends the structure of a morpheme (but not that of a word): as in French, a morpheme may be divided by a syllable boundary, so that the syllable is defined functionally as a transformational, not a deep-systemic, unit. The following examples show the output of the syllable definition transformation before the application of other obligatory rules. # + ma % 1er + % # # + zwe + t+% # # + ροζ + % #

# + ma % le % r + e + % # + zwe + % # # + po % ζ + e + % #

98

TRANSFORMATION PHONOLOGY

The structural strings a, b, and c, below, provide a more precise description of the output of the rule in terms of natural phonological classes:

#

(a) stop fricative nasal C liquid glide

(b)

stop fricative nasal C liquid

%

stop fricative nasal C liquid glide

vowel

(c)

vowel

vowel

Insert a syllable boundary % between a consonant or glide and one or more consonants followed by a vowel, or after a vowel separated from a following vowel by one or no consonant. TR 14. Syllable definition (OBL) + —

null

#

[ + ens ] . [ - syl] + [ + syl] - - ([-syl] ) +

[ + syl ]

(a) (b) (c)

Other morpheme boundaries + (but no other word boundary) may be present anywhere in the phonological sequence. Not all consonant clusters close the preceding syllable, however, as is shown when a high-mid vowel precedes a cluster, in a morpheme (e.g., eskiz 'excuse', sektam 'September', oktjb October', problem 'problem') or in a word (e.g., de+vlope 'develop', de+klase t o degrade'). The following list describes the environments in which a boundary shift will occur. It includes some clusters for which I have not found data, but which I assume would leave the preceding vowel mid-high. The square brackets indicate a corresponding vertical ordering on both sides of %. stop

fricative stop nasal C liquid

fricative vowel

%

(a) (b) , ι consonant » ( J glide J ) vowel

SYLLABLE DEFINITION

99

stop fricative labial C velar C

liquid

(c)

dental C palatal C

(d)

The syllable boundary % will be moved to the left of the cluster. TR 15. Boundary shift (OBL)

+ ens

+ ens

α fri

[ + syl]

_+ obs + obs

+ %

Γ+ clos"] L- act J 1

— — — + +

afri obs nas clos act

(a,b) C [ - syl ] J ) [ + syl ] (c)

(d)

3 4

S.D.: 1 2 3 4 5 6 7 S.C.: 1 4 2 3 5 6 7 The elements between accolades are disjunctively ordered, and [ + ens ] is present in all cases. Rule 14 constitutes the definition of the concepts "closed syllable" and "open syllable" in Haitian. TR 15 describes the exceptions to that definition. This treatment allows a significant simplification in the environment of the rules for vowel lowering (TR 16) and for r-deletion (TR 17). It also provides a natural representation of the linguistic facts: mid vowels are low in closed syllables, and it is at the end of a syllable that r disappears. French has evolved a similar open-syllabic structure from Latin. Although closed syllables are much more frequent in modern French than in any other Romance language or dialect, most of them are due to recent borrowings from Greek, Latin, and English, while others are native (e.g., port 'door') and almost always due to a schwa in Gallo-Romance and the deletion, which is still spreading, of this schwa or "e caduc." There are thus, and always have been, two conflicting tendencies in the syllabic patterns of French: a strongly entrenched open syllabication inherited from Proto-Romance, and a powerful movement towards a closed syllable structure. This movement has all the appearances of a linguistic conspiracy, and is manifested not only in the weakening and gradual disappearance of post-tonic vowels, but also in less obvious historical facts, such as the lack of prosthesis in some recent borrowings (cf. Spanish estadio, esplín,

100

TRANSFORMATION PHONOLOGY

estilo vs. French stade, spleen, style). E. and J. Bourciez claim that 'French syllabication has developed the tendencies of Latin. By syllabicating pa-trem a word which derives from *pat-rem, Latin had already innovated. But it still accepted scrip-si, rec-tum, i.e., a syllable boundary between two consonants which, unlike -tr, could not go together. A new tendency was slowly developed whereby the last consonant of a closed syllable was preferably more 'open' than the initial consonant of the following syllable. Whereas *partire was maintained, factu was changed drastically, and its c became a yod. Later, when ct or pt reappeared in loanwords such as facteur, aptitude, the pronunciation became closer to fa-cteur, a-ptitude. . . " ( 1 9 6 7 : 4 8 ) . Spanish seems to have followed the same direction with its posthetic e, as did French creóles, which developed further that strong Romance tendency toward open syllabication and a basic CV structure. It is not surprising therefore to find the same historical change in other Romance-based creóles: e.g., galufu f o r k ' , pifikelo fisherman' in the Portuguese creole of St. Thomas (Gulf of Guinea), lávulu 'book', pázil(i) 'priest', útulu 'other' in that of Annobón 5 (cf. Port, garfo, pesqueiro, libro, padre, outro), gouroupier 'servant', galeféter t o caulk' in the French creole of Trinidad 6 (cf. French croupier, calfater). Another process which has contributed to open syllables is the addition of a vocalic ending, usually i. It has been at work in the creolization of both Portuguese and English, so that this evidence could probably be used to sustain a monogenetic theory of creolization. I believe, however, that it is part of a general tendency of creolization, due to the contact with a West African syllable typology. In the case of Romance-based creóles, that movement was of course further enhanced by a similar Romance trend. The Portuguese creóles of Annobón and St. Thomas offer examples such as tejji 'three', mazi 'but' (Annobón), St. Th.floli 'flower', ornali 'sea' (Port, tres, mas, flor, o mar), perhaps by analogy with the numerous words with final i from Portuguese e (e.g., dosi 'twelve') which are also "typical of Brazilian Portuguese" (Valkhoff 1966:63). This feature is also common in French creóles. Open syllabication is much less characteristic of modern French than of the other Romance languages, and could not be amplified to the same extent in its creolization. The addition of a final i after a closed syllable can also be found in the English creóles, especially Sranan and Saramaccan where it is a highly regular phenomenon (e.g., Sr. dribi 'drive', wipi, Svhip', dti 'hurt'). Jamaican Creole knows it too, but the current situation in JC tends to show that this process is a characteristic of creolization in general, and that it is much more stable in a "resonating" context (as in the Portuguese creóles) than when the African typology does not encounter a similar structure in the base language, which is the case with English. Cassidy has shown that decreolization in JC involves a return to the English syllabic structure CVC, through the loss of a final vowel, e.g., JC glaas 'glass', tob 'tub', whereas in Sranan they are grási, tobo, and in Twi, which has borrowed

SYLLABLE DEFINITION

101

these two words, the nativized forms are giraasè and topóo. Why JC (or any creole) decreolizes is mostly a research topic for sociologists, political scientists, and economists. How it decreolizes is of direct concern to the various branches of linguistics. It seems that the least stable characteristics are the first to go in a decreolizing situation, and that the two major factors of structural stability in creóles are a high degree of Afro-European pattern resonance, which creates this stability, and any political or socio-linguistic situation (such as the isolation of the linguistic community) which preserves it from possible decreolizing influences. Once English has become the universal trade language, and because of the powerful position of American English in the Americas and the Caribbean, English creóles were naturally the most susceptible to a return, through decreolization, to the common closed syllabication of the Germanic family, although Sranan and Saramaccan, which are more isolated and conservative than Jamaican Creole, still retain the strong CV pattern which they received from the creolizing African languages. Thus English-based creóles exhibit, sometimes within the same phonological structure (e.g., in Jamaican), both creolization (English CVC > Sr., JC CVCV) and decreolization features (JC t CVCV > JC 2 CVC). In the Portuguese creóles, open syllabication is a deepseated feature, and therefore a stable one. As an example of a conservative socio-linguistic situation, we could mention Papiamentu, where, according to Wood, decreolization is now past its peak, because "the Creole language is in a comparatively strong position, and the metropolitan languages [Dutch and Spanish] in a correspondingly weak one. Unless these facts change, it is likely that the degree of Hispanization of Papiamentu will be fixed at approximately its present level" (1972:863). Chataigner (1963) offers a well-conducted analysis of syllabic structure in Kriôl, the Portuguese creole spoken in Senegal, which displays a definite evolution toward open syllabication. He gives a convincing demonstration that Mandingo, among other African languages, has had a decisive influence in that movement. In both Kriôl and Mandingo the basic morpheme structure is CV (CV(CV) ) with, in addition, grammatical tools in V, and CVC with / as the second consonant. Kriôl has other CVC formatives, but the second C is restricted to w, I, r, m, and s, and Chataigner proposes that "if one adopts an African conception of C," the groups V+Glide will be seen as diphthongs, or complex vowels, like nasalized vowels, V+l/r/s as "quasi-diphthongs," and V+m as a nasal vowel. In any event, these CVC formatives are exceptional, and in most cases the Portuguese vocabulary has been processed through the Mandingo syllabic mold. Even the name of the language and the people, Kriôl, was written and correctly syllabicated Carayol by the British in Gambia. Other apparent exceptions, such as -nda, nterga, nganal, mbarka, can be discarded by using the proposed "African conception" of consonants, and Chataigner notes that these structures contain nasalized consonants, i.e., simple phonemes, in Mandingo, in Wolof, in Peul, which are spoken in the same area.

102

TRANSFORMATION PHONOLOGY

We have already seen that the purely African character of these clusters is much less obvious in the French creóles, since modern French has created them in Europe by itself, without any African interference. A crucial difference between Kriôl and the French creóles lies in their base languages. Through the conspiracy of several historical changes noted earlier, French (but not Portuguese) soon lost some of the characteristic open syllabicity of Romance, despite the occurrence of compensatory changes such as syllabic segmentation. To this basic difference must be added a continuous African contact for Kriôl, and the corresponding isolation of Caribbean French creóles, although the African input both in Portuguese creóles and in French creóles is of course undeniable.

CHAPTER THIRTY-THREE

Other Transformations

TR 16. Lowering (OBL) Γ+ m i d l - [-high] I— nas J

I — - [-syljj %

Oral high-mid vowels are lowered in closed syllables. Examples of derivations will be given after the rule for /"-deletion. Rule 16 will not operate if the vowel is followed by a consonant cluster such as those defined by TR 15 (Boundary shift), or to morphemes which display an underlying low-mid vowel in an open syllable (e.g., sosje 'sorcerer', scvjet towel'). In the phonetic output of the grammar, high-mid vowels may occur in what appears to be closed syllables ( e s g p o l di 'its skin is hard'). Chapter 30 explains why those utterances do not constitute counterexamples or exceptions to TR 16. The lowering rule accounts for alternations of the type: INTERMEDIATE OR PHONETIC FORMS

UNDERLYING FORMS

/kler/

kler-, kle kler(-e) poz-,pjz poz-(e) fez-,fe fez(-z)

/ροζ/ /fez/

TR 17. /--deletion (OBL) + ens — clos — obs /ter/ //ar/

null I

(-> ter) -»· te,

in a-te -ma 'burial' (cf. α-ter-e 'bury') -»· Ja, in Ja-z-ma 'load' (cf. far-ct 'cart')

'clear' 'light, polish' 'a break' t o rest' 'do, make' 'maker'

104

TRANSFORMATION PHONOLOGY

/zur/

zu, in ìu-ma

/mirI

mi Vail'

'swear-word' (cf. íur-e 'swear') (cf. mir-aj 'thick wall')

This rule applies to r only. Since it is obligatory and precedes the reduction of final clusters (TR 18), it has already applied to forms such as the following, which undergo TR 18 vacuously: /kofr/ /sikr/

-*•...-*• k)f sik

'chest' 'sugar'

An alternative solution would be to add to TR 18 the environment of TR 17 (/C #), and restrict the latter so that it apply to all consonants except r. TR 18 would then set r apart as completely different from all other consonants, in ALL environments. The analysis chosen here implies that r in contemporary creole still belongs to the class of consonants (and as such is deleted in the environment / C #), and that it differs from other [ + ens ] segments ONLY by the fact that it is also deleted whenever it does not precede a vowel or a glide. TR 18. Final cluster reduction (OBL) [ + ens ] -*• null / [ + ens ]

#

A cluster loses its final consonant in word-final position. Rule 18 includes the regular deletion of I. When a newly-borrowed form does not display the opposition Cl — C2 vs. CI — zero, a rule similar to TR 18 operates on the foreign word before it is lexicalized. This fact is captured in the grammer of the language by a subset of the sequence redundancy rules. Hence TR 18, or some equivalent device, must operate both in the lexicon and in the transformational component. In a synchronic description of Haitian, this constraint bears in effect on clusters whose first element is nasal, and on those whose second element is liquid: /zamb/ /band/ /sandr/ /mebl/ /febl/ /katr/

-> -» -*• -»· ->

¿am ban sand mebl febl kat

-*• -*•

san meb feb

'leg' 'gang, group' 'ash' furniture' Sveak' four'

OTHER TRANSFORMATIONS

105

I have extended the area of operation of the rule to other clusters, in order to reflect the borrowing and nativizing process. TR 19. Unreleased NC deletion (OBL) Γ + ens "I null I

[ - syl ]

+

L + n a s -I

/pänd/

->•

pad(-je) (cf. sis-pan

/ma/and/

->•

maßd(—iz)

/I3r?g/



iDg(-e)

(cf. maSan (cf. /3η

'to hang' 'to hang') 'goods' 'saleswoman') length' 'long')

TR 20. Final consonant deletion (OBL, lexically restricted) [-syl]

null /

&

#

This rule deletes the final consonants of certain morphemes when they are used without a suffix. It is lexically restricted (&), since all morphemes do not undergo this change, even if they meet the environmental conditions of the rule. Yet the triggering factor appears to be in part phonological. The deletion seems to be restricted to consonants with a front and lingual articulation (dentals, alveolars, and palatals), i.e., to Chomsky and Halle's "coronals" (1968: 307). These consonants are all [ + acute ] , and all but two (the palatals / and ff) are [ — high]. The only unit in this set which escapes the rule is /. la gra tafìa Sapo da bra kwa repo / ε

„ prizj fi (r)wa

~ ~ ~ ~ ~ ~ ~ ~ ~ ~

lad-e grad-e tafjat-e Sapot-e dat-is bras-e kwaz-e repoz-wa fez-ε prizm-je (prizm-e) fii-el (r)waj-Jm

'bacon, to lard' 'large, size' tafia, drunkard' liât, covered' tooth, dentist' 'arm, to mix' 'cross, to cross' test, repository' tnake, maker' "prison, prisoner' 'girl, god-child' "king, kingdom'

106

TRANSFORMATION PHONOLOGY

We have seen that a final group composed of a nasal consonant followed by an homorganic stop always loses the stop when the morpheme occurs without a suffix (e.g., bom, ban, /5η). The Structural Description of TR 18 (Final cluster reduction) could thus be treated as a subset of the S.D. of TR 20 (Final consonant deletion), and the two rules conflated. However, derivations such as that of san from /sândr/ show that the deletion of d is due solely to the preceding nasal, and not to the following word boundary which is the distinctive mark of rule 20. Moreover, the ordering of TR 20 before r-deletion, which is independently motivated (see chapter 12), makes such a conflation inadequate. The morpheme /fez/ 'do, make' displays a strange behavior in the transformational component of contemporary Haitian phonology. Its exceptional treatment is manifested at the phonetic level by its open mid vowel in / ε 'do, make' as opposed to the high-mid e in fez-e 'maker'. Words such as fapo or repo do not undergo the lowering transformation because Lowering is ordered after TR 20, which "bleeds" it (see chapter 34). Yet /fez/ loses its final consonant AFTER being lowered to fez, which seems to indicate that lowering has to operate both before and after the deletion of final consonants. A troublesome problem for the current formats of generative grammar: despite the concepts of embedding and cyclically and some attempts at iterative processes, they are not apt to give it an elegant and natural solution. I believe, however, that these facts are not related to rule ordering, but to the underlying ζ of /fez/, which actually behaves as if it were an r. Since r-deletion applies after lowering, e is changed to ε before this "z-sounding" r is deleted. The historical evidence for French supports this solution. Fouché (1952) mentions an assimilation of medial r and z, which started in the thirteenth century in some areas of the oil domain. That change was in effect a weakening of r, an apico-alveolar segment, in intervocalic position. It spread during the fifteenth century to the lower class dialects of Paris, and in 1489, the poet Villon rhymed "chaise" (from chaire < Lat. cathëdra) with "aise." By the middle of the sixteenth century, even the royal court had adopted the new articulation, which so infuriated the most vocal grammarians that a reaction took place. Obviously, by then a few forms with ζ from r had already been lexicalized, since they can still be found in the modern French "chaise," "nasiller" (cf. Fr. narine 'nostril', Spanish nariz 'nose'), and in the feminine suffix -φζ (masculine neje), fret (OF *frwet fret). According to the classification proposed by Gobi, these forms are due to one of two types of conservatism: archaism, or the "conservazione de uno stadio fonetico del francese di un' epoca determinata' '(1933:338). 1 These archaisms are of course not restricted to Haitian: MARTINIQUE

TRINIDAD MASCAREIGNES

fet

vet devwe mirwe sweñe etwet bwet latwel

Fr.

le ver ddvwar mirwar swañe etrwa bwat la twal

•( e/Hn., Fr.

agi z pc fjre

WHEN THE SYLLABLE IS OPEN IN THE FRENCH MORPHEME

Hn.

agle pe fore-(bwa)

'English' 'be quiet' 'forest'

Fr. ¡ e > e / H n . , when followed in French by non-final st, sp, sk( 1 ) (except in esk(a) Yes-no Q-word'), pt, Iv, ks, gz,. Modern Southern French exhibits the same list of "exceptions" (e.g., So. Fr. kestjo, aksepte, espwar, etc.). kcstj? aksepte espwar egzersis septabr elve esklav

Hn.

kestß ~ kesjJ aksepte espwa egzesis septam ' ~ sektam elve esklav

'question' 'accept' 'hope' 'exercise' 'September' 'raise' 'slave'

I have also treated these facts as exceptions to the general rule in the phonological component of modern Haitian (see Part III, chapter 32), and described them by using the concept of closed vs. open syllable as defined by TR 14 and TR 15.

SYLLABIC RESTRUCTURATION Fr./ε > ε / Η η · , Ι Ν

131

SYLLABLES WHICH ARE CLOSED IN HAITIAN OR IN SYLL-

ABLES WHICH WERE OPENED BY THE LOSS O F r

Fr.

fet megr fer le

Hn. fet meg fe let

'celebration' 'sii™' 'expensive' 'milk'

One possible exception to this change is Fr. ¡persm> Hall (1953) has Hn. peson.

pesln/

Hn., although

F r . / o > ; > / H n . , IN CLOSED SYLLABLES

Fr.

fot hot rol po

Hn. fit kit rwjl ρ }t

'mistake' 'coast' 'role' 'pot'

F r . / o > o / H n . , ELSEWHERE

Fr.

fo Jo rido otoer

Hn. fo Jo rido rote

'false' 'hot' 'curtain' 'height'

Fr./ j > o / H n . , when the syllable is open in French. This correspondence is further restricted by the fact that J never occurs at the end of a word in Standard French. Fr.

kirekt fjre

Hn.

korek fore-bwa

'correct' 'forest'

F r . / j > j / H n . , elsewhere, i.e., in syllables which are closed in both French and Haitian, and in Haitian syllables which have become open through the loss of r. Fr.

pjrt k)l bjr

Hn.

pjt lak?l bj

'door' 'glue' 'side'

132

CREOLIZATION

kir mor (Exception:

k? tarn) Fr .¡sortir > soti/Un.,

'body' 'death' 'go out'.)

Two or three words in the language display w for Fr. j , between a labial consonant and a vowel: Fr.

pjct bja

Hn.

'poet' 'boa'

pwet bwa

Fr. φ and œ have been studied in chapter 37. The rule for the unrounding of front labial vowels and glide (DR 3) must be ordered before those which lower (DR 5) or raise (DR 6) mid vowels. DR 4.

Syllabic restructuration + ens

+ ens ob s (a)

[ + obs

] (a)

[ -

]

— nas

%

[ + syl] [ + fri]

(b)

+ clos

fri

[ + syl ]

+ clos (c)

— act

(c) + act

2 Fr.: 1 2 3 4 5 Hn. ("Structural change"):

(b)

4

1 3 2 4 5

A syllable boundary shifts to the front of a medial cluster with the structure liquid + stop or fricative, fricative + other consonant, or labial C + nonlabial C (except r). This restructuration caused the change from French open ε and y to their closed counterparts in Haitian. These clusters are exceptional in the phonology of contemporary Haitian, since a vowel followed by any other consonant cluster is always open. Hence the concept of a phonological syllable was also used in the synchronic description of the language. The modification of syllabic structure represents a diachronic change between sixteenth- or seventeenth-century French and French creóles. Therefore it

SYLLABIC RESTRUCTURATION

133

does not shift the syllable boundary to the front of clusters such as stop or fricative followed by a liquid, or stop followed by a fricative, since that was already a fact in French at that time, as it is to this day. The boundary shift results in the more frequent occurrence of open syllables in Haitian, and is only one illustration of the strong tendency of creóles towards open syllabication. Is this tendency a mere transfer from the African "substratum"? Is it due to a characteristic creolizing process in which the catalytic interference of an African structure has acted as a resonator upon an already existing shift in the system of the base language? Or is it simply a natural evolution of French which continued to take place freely in the New World, protected from the more conservative influences of the educated French community by its social and geographical isolation? It is true that open syllabicity is common in West African languages. Cassidy mentions the fact that "words can end in vowels or n a s a l s . . . in Twi and some other African languages" (1962:271). On the other hand, some linguists feel that there may well be a universal tendency towards open syllabicity. These changes would then not be due directly to either African or French structures, but to the (accidential) manifestation of this universal linguistic tendency. Given the still rudimentary state of the study of linguistic universals, this suggestion offers a weak explanation for the diachronic evolution of Haitian. Cassidy (1962:271-72) notes the same strong tendency in early Jamaican Creole where, however, the ongoing decreolization has virtually reversed the direction of the change, and in contemporary Surinam Creole, where it is still very stable. Phonological creolization in Sranan consisted in subjecting European words to an "African sound-law" in order to make them fit, among other things, the open-syllabic pattern which is so frequent in West Africa. We might add that this relatively simple sound law also applies regularly to Saramaccan. If the word ends in an unstressed vowel, no change occurs in the syllabic structure: E. heavy sister

Sr(anan) hébi sisa

If this unstressed vowel is followed by a consonant, metathesis turns the sequence into an open syllable (in Sranan, not in Saramaccan, where the consonant is simply deleted): E. kettle copper thunder bastard master

Sr.

kétre kjprj dindru básra másra

Sar.

kttz kipu (giita) bása misa

134

CREOLIZATION

people candle

pipai kándra

(símbej kandéja

When the stress is on the last vowel of a word ending in a non nasal consonant, an epithetic vowel harmonizes with the final vowel, either in all its features (51% of instances) or in backness. The rule applies in 71% of instances, according to Cassidy. E. jump cork drive fill (full) shake

Sr.djómpo kórku dríbi fúru séci

S. diipi3 fúu séki

Sometimes both metathesis and epithesis occur, to eliminate a final consonant cluster: E. court work

Sr.krútu wróko

oroko)

These changes do not occur when the word ends in a nasal consonant. Cassidy's explanation is that these phonological changes are due to the African element in the creolizing process, and it would be difficult not to agress with him: English has not displayed such a tendency towards open syllabicity. It seems that the matter is more complex in creóles which are based on French, since the same tendency can be found both in West African languages and in French and the other Romance languages. This tendency is manifested in the history of French by the deletion of word-final voiceless consonants (e.g., des [ d e ] , port [ρ Dr], pot [po]), as it is in Caribbean Spanish by the deletion of word-final s (e.g., Puerto-Rican [lah p w e l t a h ] , Dominican [ l a p w e r t a ] , las puertas 'the doors'). Yet I will not hesitate to extend Cassidy's "African" explanation to French creóles, and not simply because the slaves of English- and French-governed islands came from the same linguistic areas of Africa. On the basis of internal synchronic evidence 4 and diachronic documents which are readily available, the widespread opinion that French is an open-syllabic language is a fallacy. It is true that the deletion of final voiceless obstruents has created a vast number of new open syllables in the lexicon of French. So has the nasalization process, which was followed by the loss of the nasalizing consonant when it was not followed by a vowel. But other major phonological changes have had the opposite result of turning open syllables into closed ones. Some of these changes have been at work for centuries and show no sign of being reversed in contemporary French. The most visible is of course the increasing loss of "e caduc"

SYLLABIC RESTRUCTURATION

135

(schwa), which is the final stage in the weakening of unstressed vowels; a phenomenon very similar in nature to the English neutralization (to schwa) of all vowels, starting with α as in French. French elision has essentially the same function as the deletion of "e caduc," although it applies to a partially different environment. The results of these various changes, some of which are still synchronic rules, are identical in the two languages: English has been a syllable-closing language in modern times, and so has French, perhaps because of earlier influences from the Germanic structures introduced by invaders from Northern Europe. The fact is less obvious in French than in English because countering Romance tendencies or shifts have been at work. In earlier times (e.g., final consonant deletion, starting with s) and, on an irregular basis, today: e.g., non-standard aggregations, as in avfyvwar 'She wants to see', ¡epa Ί don't know', ftadipa Ί am not going to tell you', ivjen 'They are coming', from Standard French elvvwar, zanfajsepa, zdn(a)tal(3jdirepa, ilvjen. Thus French has been oscillating between two opposite directions for several centuries, but the closing tendency is clearly stronger today. 5 The only difference in this domain between English and French creóles can also be found in their base languages, and the African input was undoubtedly the main factor in syllable patterning in all creóles. As for the harmonizing process described by Cassidy in Sranan, the vowel harmony transformations of contemporary French creóles (see Part III, chapter 29) appear to be part of a very different mechanism. Vowel harmony is essentially morphophonemic in Haitian, while it is phonological in Sranan, where it seems to function within morpheme boundaries exclusively, at least if we judge by Cassidy's proposal. No phonological vowel harmony has been found in Haitian, but I would not be surprised if a deep-structural investigation of English creóles revealed harmonizing transformations similar in nature to those of French creóles. DR 5.

Open syllable adjustment nas >

[ + high ]

/

%

+ mid French ε and J are raised to Hn. e and o in open syllables. The context portion of the rule refers to Haitian only: thus clusters of the type st, sk, etc. do not check the preceding syllable, and DR 5 has no exceptions. Fr.

DR 6.

agle eksplike

Hn.

agle esplike

Closed syllable adjustment

'English' 'explain'

136

CREOLIZATION

[ + mid ] Fr.

fot rol kot

>

[-high] Hn.

/ fit ywjl kot

[ — syl ] ! % 'mistake' 'role' 'coast'

DR 6 applies vacuously to low-mid vowels. It has an exact correspondent in TR 18 (Lowering). Both DR 5 and DR 6 are necessary in this diaphonie description. DR 6 is needed to treat the above forms, which in French are exceptions to the rule of distribution of mid vowels (open in closed syllables, closed in open syllables). No open-syllable adjustment exists in modern French creóles as a phonological transformation. These structures are productedby a sequence redundancy rule

CHAPTER FORTY

Nasalization

Haitian Creole has two high nasal vowels, 7 and u, which do not exist in the phonology of modern French. They did occur however in Old French, and have been preserved to this day in some regional dialects (see Fouché 1952:363). Hn.

Rh ptïm pïga nu vudu zurumu

'moon' 'feather' 'be sure not to' 'we/you (pi.)' 'voodoo' 'pumpkin'

Fr./fl > α ( [ Λ ] ) / Η η . We have already seen that Hn. it is a back vowel, and higher than Fr. a. Fr,

Fr./œ >

a-j (f-)t-a-pri

Hn.

aro ( [ Λ γ ο ] ) t'apri ( [tApyi] )

'on top of, over' 'please'

ε/Hn.

Fr. œ has behaved like oral œ and other labial front vowels: it has been unrounded ( t o ε ) in all environments.

F r . œ OCCURS IN ALL POSITIONS; SO DOES H n . ε

Fr.

lœdi bree (bryn)

Hn.

ledi bre (bñn)

'Monday' 'dark'

Hn. ju 'a' (art.) may be derived from Fr. yn, with a dissociation of the features front and labial. It may also be derived from Spanish un. It is interesting to note that the change from œ to ε is currently taking place in modern French, where a large number of speakers have lost the opposition

138

CREO LIZ ΑΤΙΟΝ

œ ~ ε through the unrounding of ce. Similarly i sometimes replaces^ in some varieties of non-standard French. Thus the dissociation of vowel frontness and labiality in French creóles could be considered, from a different perspective, as a return to the original Romance opposition front-unrounded vs. back-rounded. This return to "normal" (or to the less marked units, in the terminology of generative phonology) was precipitated by the creolizing contact. It may be in progress in French, but at a much slower rate, and it has affected so far only the nasal vowel.

F r . / ε , 5 > ε , 3 7 H n . , RESPECTIVELY, IN ALL ENVIRONMENTS

Fr.

pie fjdr kapJ

Hn.

pie fin kapJ

'full, fill' 'to melt' 'coward'

As in French, only the low varieties of mid vowels are nasal in Haitian. Although their articulation is sometimes slightly higher than in French, they remain low vowels. Contact with m seems to cause a slight raising of the vowel; e.g., in nJm, rim. But the modification is not constant. In mun 'person', Fr. o has been raised to u, but the Haitian form may represent a merging of Fr. mJd 'people, world' and words with the same meaning in various African languages (see Sylvain 1936:53, Taylor 1956:404, Goodman 1964:30). The systemic distribution of nasality in creóles is commonly described as anarchic and difficult to account for, synchronically and historically. The implicit conclusion is that creolization is a special type of evolution, one of whose main characteristics is extreme irregularity. Few linguists have tackled the problem, and those who have, from a historical or a synchronic viewpoint, have mentioned the erratic aspects of this phonological domain. A comparative study of nasality in French lexical entries or words and the resulting distribution of nasality in Haitian lexical entries or words shows that the existence of sound correspondences between modern Haitian and modern French is quite clear, as are these correspondences themselves. A generative treatment relating the phonologies of the two languages brings to light certain discrepancies and exceptions which have so far remained concealed in the apparent confusion of the data. Some of those discrepancies may be contemporary reflexes of the past evolutive stages of Haitian. In the complete absence of written records, they could lead to diachronic research, and may help fill the historical vacuum between modern Haitian and the mixture of French and African dialects from which it originated. A diachronic analysis of nasalization has to make a distinction which does not exist in the Morpheme Structure Component of Haitian. Namely, the

NASALIZATION

139

separation, in Haitian formatives, of those nasal vowels which come from French oral vowels, from those which are already nasal in the French etymon. The following examples show that both regressive and progressive assimilations have been at work. These items were not in environments which might have induced vowel nasality by (synchronic) transformation. (For instance, lamo would become tarn j in m ap voje o / al f e f e kJpe lami mene ba mwë Ί am sending you to fetch Father Death for me'.) Regressive nasalization: man ok

Fr.

fíme Isι ßm lestjmak pïïmc lame fwlmí zurumu parie bukane banan buie νΐήΐ kiïma (maji)grene peñe mene / ν «ν fini bwm DR 7.

mañik fidr alyme len jabr l estoma plyme la mer furmi ziro m J pane, panje bukane banan bene, benje v(d)nir kjrria egrene peñe m(a)ne finir b )m

'manioc' 'to melt' 'to switch on ' *wooP 'bedroom' 'stomach 'to pluck' 'sea, lake' 'ant' *pumpkin' 'basket' 'to smoke, broil' 'banana' 'swim, bathe' 'come' 'how' 'husked (corn)' t o comb' 'bring, take' 'to end' 'one-eyed'

Regressive nasalization + ens

[ + syl ]

>

[ + nas ]

/ — + nas

A French vowel was nasalized by a following nasal consonant. Fr.

bukane plyme pane

Hn.

*bukane *pïïme *pañe

Ç>bukXne) Ç>pfïme) Ç>pme)

140

CREOLIZATION

Very few forms have escaped regressive nasalization: e.g., leogan 'Leogane', kamp 'bus, truck', km "horn', kone 'horn-shaped paper bag', dene 'last', pan 'breakdown' (see Valdman 1971:206). Progressive nasalization, on the contrary, has more exceptions than examples: Hn.

meb milet nef viheg male mare mo neges mjde nD meg matrite lam j kadna kana wJma

Fr.

mcebl myle noef vinegr malcer amare mo negres mirdr ner megr matyrite la m)r kadna kanar jmar

'piece of furniture' 'mule' 'new' Vinegar' 'tragedy' 'to Vord' 'woman' 'to bite' 'North' 'sea' 'ripeness' 'death' 'padlock' 'duck' 'lobster'

It would seem that in most cases diachronic nasalization has occurred in a totally haphazard way. We have, for instance, progressive nasalization in reme, mene, me, but not in Gm), vlneg or again lamo. Regressive nasalization took place across an oral segment in kasJ 'pants', ava 'before', mut J 'sheep', but not in the following words: Hn.

madam mate matât bJsmasD kapj bulcanè novam

Fr.

madam mate ma tat (E. boss) mas J kapo bukane njvabr

Vife, Mrs.' 'morning' 'aunt' 'bricklayer' 'coward' 'to broil' 'November'

The forms *madam, *matat, *bjsmasD, *kap3, *bukane, *n jvam are ungrammatical. Progressive nasalization: Hn.

mut J me

Fr.

mutj me

'sheep' 'but'

NASALIZATION

plyme alyme myr peñe kme eme pane z(a)nu areñe, arenje

pilme time mi peñe kme reme pane zenu zareñe •ν

141

'to pluck' 'light (V)' 'wall, ripe' 'to comb' 'know' 'to like, love' 'basket' 'knee' 'spider'

With extremely few exceptions (e.g., rríízik 'music', muri 'die, dead'), progressive nasalization has affected vowels in absolute non-final position only. Thus the diachronic rule would be: [ + syllabic ]

>

[ + nasal ]

between NC and #

This rule would reject froms such as lame, ne, 'born', Gmj, wJma, into a minor class which, for some or no reason, had escaped progressive nasalization. For the linguist who is familiar with French creóles, however, some of those forms, those which have lost a syllable-final r, immediately attract the attention. Taylor (1957:360) noted that in the creole of Martinique, the vowels, ε,-> and a deriving from French er, or, and ar are never replaced by nasal vowels. This is not true of Haitian, which presents counter-evidence such as dirriï 'to sleep' and feme 'to close' (French dormir, ferme). In fact, any vowel, high, low, or mid, may have become nasal in Haitian, even when it was separated by r from the nasal consonant in the French etymon: mi 'wall', vim 'to come', tune 'to go back', kJfimasjj 'religious confirmation' (Fr. myr, v(3)nir, turne, kifirmasjJ). Ans followed up Taylor's treatment (restricting it to final syllables), and tried to use the correlation loss-of-r, non-nasalization of the preceding vowel, in order to show that the lost r has in fact retained as a classical phoneme in the lexical representations of those morphemes. It is this r which, in his "purely synchronic analysis" (1968:57), prevents the nasalization of the vowel. Ans' model, which has been criticized by Valdman (1971:204), is thus based on an underlying r which always disappears in syllable-final position. A non-negligible difference should be noted at this point between an abstract unit which is subject to absolute deletion (syllable-final r never surfaces as phonetic matter), and hypothetical segments which neutralize to superficial [a] in Nupe. In a sense, Nupe [a] could be viewed as the mirror image of an archiphoneme: an "archiphone," which conditions its environments in various ways, and realizes three phonemes. (The Praguian archiphoneme is conditioned by various environments, and realized by two or more phonemes.) A Haitian / r / in syllable-final position would be a non-phone, corresponding to an independently motivated phoneme which would neither assimilate, nor be assimilated by

142

CREOLIZATION

its environment. This hardly constitutes a clear and "natural" synchronic description (even if it could be shown that those f s are in any sense "psychologically real" for monolingual speakers-which I seriously doubt). Why a post-posed segment would prevent the assimilation of a second segment, by a third one twice removed from it, is not obvious. Nor is the relationship between [r] and nasality any more direct. Except for the fact that [r] is non-nasal; but then, why / r / ? Hyman's abstract units absolutely neutralize, although they have been shown to participate, in a regular fashion, to other phonological processes (tone pattern, syllabication), and the features attributed to them are in a direct and natural relationship with the assimilations which are posited. Schane's abstract nasal consonants absolutely delete in many cases (e.g., [va] ~ [vad] ~ [vat] 'sell, sale', [ / a ] ~ [ / a t ] 'song, sing', [da] ~ [dat] 'tooth, dent-', from /vANd/, / / A N t / , /dANt/, 1968:48), but there is a natural relationship between a nasal consonant and the nasalization of a vowel. The explanation for the lack of nasalization in Haitian is purely historical, and it much simpler and more natural than Ans' solution. It does involve r, inasmuch as it is a consonant: as such, it closed the syllable, and thus prevented the nasalization of the vowel (e.g., lame, lamj). In the forms [kàna] 'duck', [n:>] 'North', and [ a m e ] 'bitter', which Ans used as examples, the final r is at least an etymological reality. (nidik 'North' (adj.) shows that no is not / n i r / but / n o d / in the lexicon of contemporary Haitian, at least for bilinguals). Yet this treatment still leaves unexplained the nasal ΐ of mT, v'irvi, firn, etc. (from Fr. myr, vanir, finir), unless one takes their origin to be in the past participles vany and fini (or in the r-less infinitives of the 16th and 17th centuries), and in erratic, inexplicable forms like my. It also forces the analyst to posit in some forms a ghost r which is twice immaterial, since it never existed in the etymon to begin with: e.g., mo kadna, ane, etc. (Fr. mo(t), kadna(s), ane, OF ane). In other words, Ans' synchronic hypothesis presents the double weakness of not being synchronic at all and of not being functional, historically, in many cases. The data show that word-final i and u have been nasalized by a preceding nasal, whether or not they are followed by an r in the French cognate. It is also clear that a and o have not been so nasalized. A can occur in the last syllable in modern French, whether that syllable is open or closed, o is not often found in closed syllables, and there is no form where o is preceded by a nasal and followed by a word-final consonant. French J never occurs in stressed open syllables. Its presence in word-final position in Haitian is due to the deletion of r, and it is never nasalized in that environment, e never appears in a stressed closed syllable in French. (The analysis is here limited to lexical items, morphemes or words, and does not account for combinatory phenomena beyond the word level, as in Fr. ilaapjrtetkwalir 'He brought something to read.') e has generally been nasalized in final position, with a few exceptions (see below),

NASALIZATION

143

and lowered to ε . French ε has also undergone progressive nasalization in absolute final position, i.e., when it was not followed by a consonant, including r, in the French etymon: e.g., me 'but', vs. lame 'sea' (Fr. me, la mer). The following table summarizes the patterns of progressive nasalization of French vowels before a word-final consonant (including r), and in absolute final position. I have left aside the French labial front vowels, assuming that they have already been backed or unrounded at this stage of the derivation.

PROGRESSIVE NASALIZATION IN THE ENVIRONMENT F r . / N C

#

FRENCH

EXAMPLES (Hn.)

O o, a

fwOTlT 'ant' none malfihT 'hawk' zamT 'friend' nu 'we, you' (pi) none zuriïmiï 'pumpkin' ζεηίϊ 'knee' 'smoke' ane *year' lafime 'spider' d ne 'last' zareñe 'pluck' pïïmè 'but' none me 'know' Ιοηε Not applicable, (J never occurs in Fr./NC Not nasalized.

EXCEPTIONS (Hn.)

PROGRESSIVE NASALIZATION IN THE ENVIRONMENT F r . / N C

i

u e o, 3, e, a

C,

fïhï 'end' none vTnT 'come' ranï 'bray' lamiï 'love' (N) none Not applicable, (e never occurs in Fr./NC Not nasalized.

#.)

#

C! #.)

The only vowels which have undergone progressive nasalization in that position are thus i and u, before r exclusively. All other final consonants have blocked progressive nasalization (e.g., lanwit 'night', mil 'thousand', muí 'fly'). It seems that the rule for progressive nasalization operated in two steps. The nasalization of ε (and vacuously, that of J) must be ordered before /--deletion takes place:

144

CREOLIZATION

+ mid >

[ + nas ]

between NC and #

high . That of i and u may be ordered either before r-deletion, in which case it reads: + high — mid

>

[ + nas ]

between NC and (r) #

or after it, and it then operates in the same environment as the nasalization of ε and J : + high >

[ + nas ]

between NC and #

— mid As for the nasalization of e, it could take place with that of i-u or with the assimilation of e-D, since r would occur only in the environment Fr. / #. That fact, correlated with the large number of exceptions in the progressive nasalization of e, and with the different behaviors of the pairs e-J, i-u, suggests a possible reflex of the actual order in which the changes occurred. For all known creóles, diachronic linguistics has virtually no records upon which to rely, except in the present stage of the language, usually Indo-European, which provided most of the lexical items. Research into the actual chronology of change is thus, of necessity, based on formal comparisons of structures in the creole and in the corresponding Indo-European language. It is quite possible that only one progressive nasalization took place in the creolization of French, with environmental restrictions for open vowels. Another assumption is that progressive nasalization does operate in two stages, which may reflect the diachronic: first on low-mid vowels and on some forms with final e (e.g., paite, la fyme), and again, in the same environment but after the deletion of r, to high vowels. We would then have three chronologically ordered rules:

(a) First Progressive Nasalization (ε, and e in some forms) V + mid >

[ + nas ]

between NC and #

— acute

(b) r Deletion (for which see chapter 46)

NASALIZATION

r

>

145

null before syllable boundary

(c) Second Progressive Nasalization (/' and u) V - mid >

[ + nas ]

between NC and #

_+ high Nasal vowel adjustment is ordered after the nasalization rules, lowering e and o to ε and J. (O has suffered regressive, but not progressive, nasalization from French to Haitian.) It has probably been active at all times and still functions as a redundancy rule (but not as a transformation) of contemporary Haitian. Regressive nasalization took place after metathesis (e.g., forms like bwm, fwjnu, Fr. birri, furmí). ddrru, f5m 'shape', njmal 'normal', show that regressive nasalization is to be ordered after r-deletion (although the nasal J of dJrrii could be the result of nasal harmony, as in ava 'before'): Fr. fjrm > *fim > Hn. fjm, as in: Hn. bet la te afJm ju mun 'The animal had the shape of a human being' m afîm / mJfe Ί feel great, my good fellow' Fr. dormir > *d)mi > Hn. dJmT (The deletion of word-final r may have occurred even before French was creolized, in the French '-er' and '-ir' infinitives, and in some nouns and adjectives such as pane, dcrrte. See Fouché 1966:663-80.) However, other forms used by the same monolingual speakers indicate the reverse ordering: Hn. km, kme, dene, whose vowels, like those of elen 'Helen' or leogan, have not suffered regressive nasalization. Are those items exceptions to the regular ordering? Do they reflect two waves of borrowing at different times, separated by either a reversal of the regular order of, or a restriction on, the changes? The fact that nasalization in the creolization process is not a chaotic and unaccountable series of events, but a highly regular albeit complex set of changes, still leaves us with a few forms which can be treated as inexplicable "exceptions," or as clues to the need for further research in this area. Nasalization phenomena have always been regarded as one of the main differences between French and its creóles, and indeed as a characterisitic of creolization. I would rather consider them (at least regressive nasalization) an evolutive feature which is common to French and French-based creóles.

146

CREOLIZATION

Regressive nasalization took place in Haitian whenever the vowel was in contact with a following nasal consonant, whether that consonant was itself followed by a vowel, a boundary, or another consonant. Contemporary French exhibits nasal vowels when in the etymon the vowels are followed by a consonant cluster whose first segment is nasal: LATIN

campu lana camera

(> *tfap) Ç>*tfambra)

FRENCH

HAITIAN

Ja len fabr

Ja lën ßm

'field' 'wool' 'bedroom'

During the 11th c., however, a, o, and e were nasalized in French before any nasal consonant, whatever followed it. Regressive nasalization in French creóles is thus quite close to what took place in Old French. It is also a fact, mentioned by Jourdain, that similar nasalization processes (progressive nasalization especially) can be found in West African languages. Some of those languages, or perhaps an unstable mixture of African structures, had a powerful influence in the evolution of French creóles. Yet there is considerable evidence, in the comparative study of Haitian and French, that this influence was felt most in those areas which already possessed some type of evolutive latency similar to that of French. It seems reasonable, at least in this perspective, to treat Haitian and French as two languages which belong to the Northern West Romance group. The former underwent a crucial contact situation with other, unrelated languages. It can be assumed that the contribution of West African languages to French creóles has been twofold: a direct infusion of patterns (manifested in syntax especially) and of a few lexical items, and a catalytic effect on the evolutive potentials which already existed in the type of French with which the slaves came into contact. DR 8. + + I1+

Nasal vowel adjustment syl nas high! mid)

>

[

+ mid"] - high J

All nasal vowels, except those which are [ + high, — mid ] (i.e., 7 and m), belong to one subset of the vocalic system: [ + mid, — high ] . Thus, words like firn 'movie', ju 'a', mun 'person', do not undergo this rule; e and o (produced by DR 7) are lowered to [ — high ] , while a, borrowed from French or produced by DR 7, is raised to λ. Fr. œ has been bled out of the system by DR 3 (unrounding of front non-consonantals), and Fr. e and o undergo DR 8 vacuously. The result is a nasal vowel structure with two degrees of height: [ + high,

NASALIZATION

147

- mid ] and [ — high, + mid ] . It should be noted that these diphonic rules apply to isolated words or morphemes; we have seen in Part III, chapter 24 that a native nasalization is at work in Haitian, and may produce high-mid or low non-mid nasal vowels (ε, o, a ) in the surface representations.

CHAPTER FORTY-ONE

Deneutralization

In word-final position, French schwa is pronounced only within certain types of consonant clusters (e.g., tabla rJd 'round table', vinegra bfà 'white vinegar', but el kri 'she is shouting'). This schwa never appears in Haitian cognates. Its loss has allowed the deletion of r and I in word-final position (after an obstruent for I).

F r . / a > n u l l / H n . , IN WORD-FINAL POSITION

Fr.

ruz(a) tab! (a ) frer( 3 ) vinegrfd ) d(3)

Hn.

ruz 'red' tab 'table' fre 'brother' vin eg 'vinegar' d- (in dio 'water')

We may reasonably assume that the French etyma would also have lost r and I if final schwa had not been retained in consonant clusters. This is in fact what happens in popular French; e.g., yn tab 'a table', η )t 'our', lot zur 'the other day', pov tip 'poor fellow', fjrmidab V o n d e r f u l ' (Std. Fr. yn tabi, notr, lotre zur, povre tip, fjrmidabl). In the seventheenth century, forms such as kat 'four', njt 'our', vjt 'your' were accepted in Standard French, although syk 'sugar' (cf. Hn. sik) or k jf 'chest' (Hn. k ) f ) were considered substandard. Latin r and I were dropped in French when they become word-final: L. pänärium *prîmâriu cantare *usitilium

OF

*panjer *premjer *tfan ter *util

Fr.

pane pramje fate uti

'basket' 'first' 'sing' 'tool'

r also disappeared in -ir infinitives (e.g., dirmi(r), cf. Hn. domi) and in some substantives ending in oir (e.g., tiroi(r), cf. Hn. tirwa), but was restored by grammarians in the eighteenth century. Creole languages have never been subjected to the normative pressures which exist for French, and it is likely that Haitian has simply undergone freely some of the evolutive processes which have

DENEUTRALIZATION

149

been at work in Northern West Romance ever since Vulgar Latin. Thus the language-contact situation referred to earlier would not be the only factor in the different rates of change in French and in Haitian. In non-final position, Fr. schwa has become e or i: Fr.

b(d)zwe dajr s(a)men v(d)nir ekravis

Hn. bezwc derw ) semen νΐηϊ kribif

'need' 'outside' 'week' 'come' 'crayfish'

More rarely, Creole has u: Fr.

sfdjkwe Hn. suke 'shake' s(d)kus sukus 'shaking (N)' (Cf., in Trinidad, "soucousse, soucrer," and "soucrade" 'a shaking', Thomas 1969:73.) d(e)va duva 'before' (In laruze duva zu 'the morning dew'.)

i is normally the reflex of Fr. schwa when the following syllabic segment is acute, but there are exceptions to this rule: Fr. r( 3jlizjj, Hn. relizj 'religion'. In at least one instance, Fr. schwa has changed to o: Fr. agranaζ, Hn. gwJnaz 'gear', with the subsequent change of r to rw and w, and the nasalization of o to 3.

CHAPTER FORTY-TWO

Palatalization

If ρ and b remain unchanged, except for their deletion in certain clusters, all non-labial stops have been affected to some degree by weakening and palatalizing changes.

F r . / p > p / H n . , IN ALL POSITIONS

Fr.

kape plyzjcer zyp

Hn.

kape pliz/ε zip

'to stand' 'several' 'skirt'

A ρ is still pronounced in Hn. trop 'too much' (after a verb), whereas the ρ in Fr. tro 'to much' (in any context) now surfaces only in liaisons.

Fi./b > b/Hn. Fr.

sabr tobe

Hn.

sab tobe

'sabre' fall'

Fr./t, d > t, d / H n . Hn. t is phonetically similar to Fr. t (dental or dento-alveolar stop), except before a high front vowel or glide, where it is affricated to f 5 . So is d, to dz: Fr.

D R 9.

aiti sorti p(d J ti dy ri

Hn.

aitsi sotsi tsi-(mamaj) dziri

DENTAL TOP LAXING

'Haiti' 'go out' 'children' 'rice'

PALATALIZATION

+ clos + act — ñas

>

[ + strident ]

/

— — + +

151

ens mid high act

t and d are laxed to t8 and dz before high front vowels and glides. Since these affricated segments are predictable in contemporary French creóles, a constraint similar to DR 9 must be included as a low-level rule in the transformational component of Haitian phonology.

Fr./Ä: > k / ì ì n . As in popular Parisian French, Hn. k is normally palatalized before front vowels and glides: Fr.

kite koer pikyr kiiizin la pliii

Hn.

kite,k'ite ke, k'e piki, pik 'i kizin,k'izin lapli, lapl'i

'leave, let' 'heart' 'shot (N)' 'kitchen' 'rain'

Hn.

gwo hg agle

'big' 'ogre' 'English'

Fr./*>*/Hn. Fr.

gro )gr agi E

g is palatalized in Hn. gjil ~ dzDI ~ djjl, Fr. gœl 'mouth'.

Fr./d>d/Hn. d can be palatalized (to d\ dz, or even gj) in one or two words before i or j (djab ~ dzab ~ gjab 'devil'). This phenomenon is more widespread in Louisiana Creole (data from Morgan 1960,1964, 1970): Fr. d]'4> dy ri tjebJ

La.

Je d'iri c'ibo

'God' 'rice' 'hold tight'

152

CREOLIZATION

La. goel P(a)ti ki di -

èε lei (Piti k'i d'i g'im

'tail' 'throat' 'little' 'who' 'say' 'gamecock'

Thomas (1969:3) noted the same process in Trinidad French Creole, as does Funck for Martinique. In the examples for Trinidad, Thomas's spelling 'ch' stands for [ δ ] . cuite culotte (re)culer quinze marquer (em)barquer tranquille pourquoi cricket

Tr.

chuite chilotte chouler chinze mâcher bâcher Mart, twaéil puéi kwiéet

'cooked' 'trousers' 'to recede' 'fifteen' 'to mark' 'to embark' 'quiet' 'why' 'criquet'

The palatalization of velars must have taken place before front rounded vowels had been backed, as shown by "chouler" (and perhaps by "chuite," which has a back glide in twentieth-century French creóles: Hn. kwit). The form "chouler" in nineteenth-century Trinidadian can be related to 6u in the modern Haitian expression t3be éubiïm 'to fall down', and both are derived from Fr. ky 'buttocks', although the word in modern Haitian is the unrelated buda, probably of African origin (cf. Krio an 'arm', onda-án 'armpit').

Fr./x > s / H n . Fr.

syrtu ρ ist

Hn.

situ pis

'especially' 'track'

Fr. s corresponds to Hn. / when / occurs after it or in the following syllable (e.g., Fr. scf, Hn. f e f , 'dry'), and in the word mufe (or raye), Fr. ma s; φ (msj), 'man, Mr.'. An occasional / for s can also be found in the English creóles (e.g., Gullah fifa 'sister').

Fr./z > z / H n .

PALATALIZATION

Fr.

roz poze, rapoze

Hn.

rw3Z, poze

Hn.

kaj kijyj zje, ze

wjz

153

rose 'put, rest'

Among the exceptions are: Fr.

kaz (kelka)foz (Ιεζ)ίφ

'house' 'something' 'eye'

Fr./z > z / H n . (EXCEPTION: FR. bagai, Hn. bagaj, 'thing') Fr.

zirom '5 kaz

Hn.

zurumu kaz

'pumpkin' 'cage'

ζ is frequently weakened to a glide at the end of a Haitian word:. Fr.

Jarz vwajaz

Hn.

/aj vojaj

'load' 'trip'

Thomas mentions this change in Trinidad (1969:6): Fr.

lagaz kaz foz

Tr.

lagaj kaj fjf

These forms are identical in various French creóles of the Caribbean. Since they are due to a lexically-restricted change which was not rule-governed, they must have been transmitted by slavers or by slaves who were transferred from one island to another. Palatalization also seems to be widespread in Spanish and Portuguese creóles, although published materials for Papiamentu are not very informative in this domain and the data which they offer are sometimes contradictory. Wood (1972:861) mentions the changes Sp. / d j , sj > /, / / Pap., as in: Sp.

diente dia viernes siete demasiado

Pap. Jente Jabjerne fete mafa

'tooth' 'Friday' 'seven' Very'

Navarro-Tomás (1953:187) does not have mafa but "masiar" for Sp. demasiado, and judging by his examples, it seems that Papiamentu / derives from Castillian Θ, whereas s would remain unchanged even before i:

154

CREOLIZATION

Sp.

maíz dulce así sopa

Pap. maífi du fi asi sopì

'corn' 'sweet' 'thus' 'soup'

But the data provided by Lenz (1928:81) do not confirm such a distinction: ciento cocina palacio visita siete prisiów diez

Pap.

ferito kufina palafo bifita fete prifon fes

'100' 'kitchen' 'palace' Visit' 'seven' 'prison' 'ten'

ValkhofPs analysis of the Portuguese creóles of the Gulf of Guinea (Annobón, St. Thomas, and Principe) is more detailed. In Annobón and St. Thomas, the phonemes " / and ζ are recent developments of s, and ζ before palatal vowels and semi-vowels" (1966:88). Although a few counter-examples can be fornd even in his descriptions (e.g., ST dozi 'twelve', si 'this', inasi 'you', p. 87), other data illustrate this statement, which is perhaps simply too inclusive: Port, grande tres igreja vestir presente

An.

gazi tejfi gesa biji plezeci

'great' 'three' 'church' 'to dress'(p. 91) 'present' (p. 90)

These few examples are of course insufficient for an in-depth analysis of creolization in Papiamentu and the Portuguese creóles. No systematic study of the former has been conducted since Lenz (1928), and Valkhoff (1966) was intended as a broad presentation of the three Portuguese dialects, which now needs to be supplemented with intensive research in more limited grammatical and phonological areas. k has been preserved in almost all positions in the English creóles. It has been palatalized before front vowels in Sranan, Gullah, and Jamaican, and before a in Jamaican and Krio. 6 E.

keep kiss kill shake

Gul. ci:ρ cis cil

Sr.

ciri séci

PALATALIZATION

E.

take talk cart can carpenter

Sr. JC

155

téci tdci

kyaat kyan kyaapinta

In non-initial position, Gullah has kept k even before front vowels: E.

dark crooked cunning

Gul.

ddki kriikiti skimi

Palatalization has expanded in Jamaican to consonants other than k, and Berry thinks that "there may still be [in Krio] a tendency to palatalize /k, g/ before / a / in English loanwords on the analogy of an early series: 'cat, cask, captain, garlic, girl. . . ' > /kiat, kias, kiapin, gialik, gial. . . / " (1961:11). E.

guard gardent fern baby girl therefore

JC

gyaad gyaadn fyaan biebi gyal Kiro diafo

According to Cassidy and Le Page, "from this very clear evidence of palatalization preserved here we may argue that, about 1700 or 1725 when the folkspeech [of Jamaica] was taking form, these pronunciations must have been if not the dominant ones, at least as common on Englishmen's lips as the unpalatalized pronunciations. . . " (1961:31). The authors then give as the reason for the preservation to this day of this English palatalization an attempt to preserve contrastive pairs such as gaadn ~ gyaadn ( < E. Gordon, garden) or kaad ~ kyaad ( < E. cord, card), where English J and a have regularly merged to / a / in JC. This explanation is not likely to be correct, since there are scores of English words where no such confusion could arise from the vowel merger (e.g., gyal or kyaapinta). One might also wonder why palatalization was not used to distinguish baan ( < E. born) from baan ( < E. barn); after all, consonants other than velar stops are palatalized in JC, but not in English. I would assume that the widespread palatalization of consonants found in JC and other English creóles is a rule-governed phonological change which started in English, survived the creolization period, and has been preserved to this day, as these authors rightfully claim, rather than an archaic structure preserved solely for semantic reasons in a few cases of minimal pairs and for no known reason everywhere else.

156

CREOLIZATION

Moreover, from the strikingly similar evidence found in French creóles and in the history of French, I would think that if English and French palatalized consonants have been preserved with such regularity in their creóles, it is because they were quite familiar to speakers of Ewe, Fanti or Nupe. In Fanti for instance, stops, nasals, and glides are strongly palatalized before /i/, and slightly so before /e/. As in the French creóles, t and d are affricated with sibilant release (f 5 , d z) before i, and /k, g, h, j / display the allophones ó, J, f , ζ before l i , e/ (Weimers 1945). In other words, palatalization as a factor of historical change was already present in the evolution, actual or potential, of English and French. Their creolization gave more momentum t o . that factor because the process was amplified by similar African structures and tendencies, whereas Standard English later went through a depalatalization process, and French palatalization has been steadily expanding to new forms, although much more slowly in Standard French (even in its more colloquial styles) than in its creóles. Dobson thinks that in English "perhaps the palatalized pronunciations were originally vulgar or dialectical and had developed before the seventeenth century, but then forced their way for a time into StE, from which in the nineteenth century they were again expelled" (1957:952). This prevocalic palatalization also spread from Gallo-Romance dialects to Standard French. M.K. Pope (1952:123) notes that the "palatalising movements extended over several centuries," from Late Latin to Late Gallo-Roman. I would add that palatalization has never ceased to be at least potential in Francien, where forms such as k ï 'who' are becoming more frequent today. These processes illustrate an interesting parallelism between antillean and continental (e.g., Louisiana) French creóles, and from a more inclusive viewpoint, between French and English creóles. A comparison with English creóles reveals that palatalization of velar stops is most frequent in Gullah, where it has occurred before i and a:

/ E.

/

begin forgive give again kettle

α

Gul.

bifin f'efiß fi: fin, afin citi

PALATALIZATION

157

cilabaf cdfu casd.ße cándl jâiik jal

calabash cashew cassava candle garlic girl

Although the palatalization of g in 'girl' might be attributed to a former [i] ( < OE girle) later centralized before r, the Gullah formative for 'garlic' ( < OE garleac) cannot be so explained. Velars have not been palatalized in Gullah by any vowel other than i and a; not even by non-high front vowels (cf. ke, kez 'because')· It is also difficult to link palatalization to stress patterns with the limited data that we have, bijin and ejiß would seem to prove that they are not related, but historically the stress has shifted from the second to the first syllable in bijin. Sranan is the only other English creole where the palatalization of velar stops was regular. It occurred to a lesser extent that in Gullah however, and only before i: beg begin forget give calabash cassava candle

Sr.

béji bijin frijiti η krabdsi kasába kdndra

A parallel development is taking place in Papiamentu with the continuing weakening of ö, which is now losing its occlusion: "en la africada eh se observó mayor desarollo del elemento fricativo que en el español ordinario, con la correspondiente reducción de la parte occlusiva: lechi, cuchara, chincha 'chinche'" (Navarro-Tomás 1953:185). This shift is also, and perhaps primarily, a posteriorization of the whole unit tf from dento-palatal to palatal, and is it paralleled by the regular opening of dj into /, as in jenti 'tooth' (Sp. diente). See also J from sj in fete 'seven' (Sp. siete). These facts point to an areal typology similar to those found among unrelated Amerindian languages and in the Balkans, for English-based Gullah and Sranan, and the French and Ibero-Romance creóles of the Americas. Although such a theory would have to explain the non-participation of Jamaican and Guyana Creoles to thse widespread palatalizing processes, it is not unreasonable to think that, among the English creóles, they are most evident in Gullah because of the influence of Louisiana and Caribbean French creóles.

158

CREOLIZATION

Why French creóles almost universally display a much higher amount of palatalization than English-based creóles is best explained by a theory of resonating structures. Proto-Romance, and the modern Romance languages, have always shown strong palatalizing tendencies. Even French, one of the Romance languages most influenced by non-Romance languages. Since palatalization processes are well represented in West African languages, French creóles actualized the latent tendencies of their base language in this domain. No such movement could be amplified by the contact of English with African phonologies, since palatalization is in English an historical event and the stop system of this language, particularly velar stops, has been stable for several centuries. If this hypothesis is correct, the regular palatalization of velars in Gullah can perhaps be explained by an areal interference from Louisiana French Creole, itself influenced by Haitian Creole when thousands of Haitians fled to the continent during the Haitian war of independence. Goodman (1964:131) notes that they introduced into Louisiana Creole Haitian grammatical constructions such as the preposed future tense particle ava, shared exclusively with Haitian, and the possessive structure. Similarly, palatalization may have been introduced in Sranan from the creole of French Guyana, a Bay creole like Haitian (Goodman's classification, 1964: 16), although it is less extensive than in Gullah. The fact that Saramaccan, spoken in the interior of Surinam, is in this instance very different from Sranan is not surprising when one notes that the former has like Krio a lexical, phonemic tone, whereas Sranan is toneless. Yet the presence or absence of tone is a deeper structural difference than that of palatalization. Since tone is distinctive it is more consciously "real" for the speaker. Moreover, "nearly all Saramaccan speakers are familiar with Sranan, and in some areas of close contact between the two groups (such as in Berg en Dal) the two languages have become interintelligible" (Hancock 1969:9). Thus Sranan, located on the coast, may have adopted the French creole palatalization of velars (restricted for the time being before i in Sranan), while Saramaccan, being geographically more isolated, has not yet developed such a tendency. Other English creóles, such as Krio (Sierra Leone) and "Cameroons Pidgin," were protected from that influence by an ocean. As for Guiana Creole (English Guyana), Hancock notes that "because of the heavily forested terrain, settlements within the country are separated from each other and from those in neighbouring countries" (1969:9). Since the forms given here were collected in rural areas of the interior, it is natural that they should not display any palatalization, even though Sranan does in neighboring Surinam. The only creole where the lack of a regular palatalization is unexplained is Jamaican. But the Jamaican forms cited here are from Russell (1868), and the date may be of help in finding an explanation. Indeed, Le Page states in his brief description of JC that "palatalization of any consonant may take place in the vicinity of a front vowel" (1957:383).

CHAPTER FORTY-THREE

Prenasalization

Fr .¡b >mj

Hn., IN WORD-FINAL POSITION AFTER A NASAL VOWEL

zab Hn. tjb kokobr ( >*kokib>) bib asabl ( > *asab >) desabr(> *desab >)

zam torn kjkym bom asam desam

'leg' 'tumb' 'cucumber' 'bomb, saucepan' 'together' 'December'

Fr J d > n / H n . , IN WORD-FINAL POSITION AFTER A NASAL VOWEL

sadr (> *sad > ) desadr (> *desad bad ded vjad amad

Hn. >)

san desan ban kodèn vjan zaman

'ash' 'go down' 'group' 'turkey' 'meat' 'almond'

Fl./g > 17/Hn., IN WORD-FINAL POSITION AFTER A NASAL VOWEL

Fr.

tag K se-djmèg epëgl (> *zepëg>) trißgl ( > *trißg> ) ß g ( {

Hn.

ίατ} Ιοη se -dome ν zepëv trí/αη

'tongue' 'long' 'Santo-Domingo' L • » pin 'triangle'

Since similar processes apply to all three voiced stops, they could be conflated as follows:

160

CREOLIZATION

+ obs -— obs η + clos

>

/

[ + nas ]

#

. + nas J + ved _ A French voiced stop in word-final position after a nasal vowel appears to be changed to the homorganic nasal sound in Haitian Creole. In French, nasal consonants do not occur in this environment ( / C #). The rule is a correct description of the contemporary phonetic data. However, a transformational analysis of Haitian (see TR's 18 and 19 in Part III, chapter 11), and evidence from French dialects (e.g., Southern French bimba, derida, Iir]gd, where the accented vowels are semi-nasalized), point to a different evolution. It is much more likely that both contemporary Standard French and Haitian derive the above morphemes from a proto-language in which these forms were very close to those of modern Southern French. If this assumption is correct, it explains the contemporary alternations X-NV-# ~ X-OV-NC-# in French 7 and X-NV-NC-# ~ X-NV-OC+V-Y in Creole; e.g., Fr. b~), bm 'good' (m, f)', Hn. zam 'leg', zab+e 'to step over'. What actually occurs is that a word-final cluster made of a nasal consonant followed by the homorganic voiced (nonnasal) stop was reduced in both languages, French retaining the oral stop, Haitian the nasal segment. In non-final position, the evolution has been identical in the two languages. Thus the change was not the complete nasalization, but the prenasalization of a final voiced stop, as reflected in DR 10. DR 10. Prenasalization

[ + nas ] 1 Fr.: 1 2 Hn : 1Γ L+

+ obs + clos .+ ved . 2 3 > 2 Π obs nas -I

# 3

2 3

The situation in the English creóles is in part different, and more difficult to evaluate since creolization has had the same effect as the contemporary evolution of these structures in the base language. (C. = Cameroons, G. = Guyana, Gu. = Gullah, JC = Jamaican, S. = Saramaccan, Sr. = Sranan, K. = Krio.)

PRENASALIZATION

OE E.

mb# m# climb dumb

OE/ME τiL E. 7?# strong

E. creóles Krio Others Gu.

m# klem klajm dUm

E. creóles

r?# trar\

c.

long

G„ JC Gu. C., Gu., JC Krio

hang

lom Ιαη Iati hen εν

or Krio S. K„ Sr.

Vr,gV trár¡ga taártga lár¡ga

Sr.

h erigi

161

The addition of an epithetic vowel in K., S., and Sr. has protected the cluster, as it was maintained without change in C., Gu., S. már¡go 'mango' or JC, K. memba 'remember'. In Sranan the change has spread further, since r¡g has been reduced to rj even in non-final position when it was protected by an etymological or an epithetic vowel: E.

strong long sing mango thank (*thang) hang

Sr.

trár\a Ιάηα Sir?/ maria ίάηϊ άηα

'song'

The truncation of voiceless stops, on the other hand, is much less regular. They have been preserved in many words not only by an epithetic vowel, but even in absolute final position: E.

mp# jump

E. er. m# (regular) C. Jom S. Sr.

+ V fómbo fómpo

Irregular G. Jámp JC J)mp Gu. iAmp

162

CREOLIZATION

Ε. τik# think drink

E. nt# hunt

E. cr. C.

u# (regular) tir¡

C.

drin

E. cr. none

n#

+ V S. tinga Sr. dénki S. diingi (Sr. drin i)

S. Sr. C.

+ V hindi mti hanta

Irregular K„ G„ JC, Gu. tink K., G., JC, Gu. drink Irregular K. 3nt G. h mt JC hont Gu. hmt

Although it is difficult to determine whether the changes mb#, ng# > m#, η # were due to creolization or to the momentum of an identical change in English, the case of nd# is unambiguous. Final n d # and even ndV were reduced in the creóles even though the d has been maintained by English: E. nd# sand

E. er. n# (regular) G., JC, Gu. san

stand S. (stand up)

blind

tan

C, G, Gu., JC blajn, bien

S. Sr.

ndV sándu sánti

nV C. Gu. G. K. Sr.

tanip tannp stanip tináp bréni

Irregular none

none

none

The evolution of the sequence V-NC-Vcd stop in final position in English and its creóles is thus remarkably similar to the history of identical clusters in French creóles. English creóles seem to have generalized an evolution started in England several centuries before slavery and creolization took place. The loss of b in mb# was regular in English by the end of the thirteenth century, and was noted by Hart (1569, 1570), Gil (1619), and many other scholars of the sixteenth and seventeenth centuries, in words such as lamb, dumb, thumb, womb, comb, etc. "ng" in Old English was pronounced [r?g] in all positions, "and this is also to be regarded as the normal Middle English pronunciation even in final position" (Dobson 1957:963). [17] seems to have been in ME an allophone of / n / which occurred exclusively before velar consonants, so that some phoneticians did not recognize its separate existence. [17] developed into a phoneme when the conditioning velar was lost. Dobson writes that in the Eastern English

PRENALIZATION

163

dialects "(including no doubt vulgar London English), [g] is lost from final [rçg] from as early as the fourteenth century, whereas StE maintained [i?g] until the late sixteenth" (1957:963). In any case, both mb# and r¡g# were reduced to their nasal elements in English, and these changes are not historically a creolization process. nd#, on the other hand, resisted change in English, and its truncation to d became regular only under the pressure of the creolizing movement. Dobson also notes that "though [n] < final [nd] is exactly parallel [to [rj] < final [r?g] ] , there is a significant difference in the treatment which StE accorded the two pronounciations; in the seventeenth century [τ?] < [17g] is regular and is shown by the most careful sources, whereas [n] < [nd] is recorded chiefly in sources which reflect vulgar pronunciations. . ., and in PresE [r?] < [r?g] alone survives, whereas [nd] has been fully maintained" (1957:964). The reasons which he proposes for this difference of treatment are not backed with convincing evidence. Nevertheless, the interesting fact is that creolization completed the change started in England, and made it regular by generalizing to nd# the truncation of mb# and r¡g#. If it is true that the non-deletion of final d after η was due to the conscious linguistic conservatism of educated English speakers, from the fourteenth century to this day, one could claim that the loss of d parallels that of g and b in the English creóles because of the "vulgar" variety of English spoken by the early slavers, whose lingusitic behavior was much less conservative than that of educated Londoners. In other words, the extention to final nd of the truncation of η g and mb can reasonably be considered an English dialectalism, rather than a creolism. Yet this dialectical hypothesis, although probably correct, is not necessary to explain creole words such as san vs. modern StE saend. The evolution of these clusters in French was thoroughly different from the change that took place in English. What disappeared was not the final stop but the "implosive" nasal consonant, after it had nasalized the preceding vowel: OF

Irngfa) mmd(d) bomb(a)

Mod. Fr.

lJg(d j mJd(a) bob(3)

'long (f)' 'world' 'bomb'

Yet French creóles exhibit a cluster modification which is identical to the evolution of these structures in English, although the vowel was also nasalized (as in French): Hn. Πη, mun, bJm. Thus both groups of creóles, French and English, have lost the word-final voiced stop after a nasal consonant. Such similar treatments in the creóles, as opposed to the different evolutions of their respective base languages, reflect creolization processes caused by similar West African inputs.

164

CREOLIZATION

This truncation occurred in French creóles although it conflicted with the structures of French and with their evolution at the time (towards the nasalization of the vowel and the loss of the nasal consonant). In English, truncation of the creole type had already been taking place for at least two centuries in ηg and mb. Creolization resonated and added momentum to that movement, which extended to the remaining cluster, nd, helped in this by its frequent reduction in some socio-stylistic varieties of English. This area of creole phonology is thus even more interesting than the palatalization of velar stops, since it displays a standardized creole evolution in the face of opposite changes in the base languages.

CHAPTER FORTY-FOUR

Rounding

F r Jr > rw,

w / H n . , BEFORE A LABIAL VOWEL

Thomas noted the same structural change in nineteenth-century Trinidadian; e.g., "foufer mofen tnî fouisson" 'My brother is ill with ague', where "fouèr" and "fouisson" are derived from French frzr and friiS (1969:4). The combinatory factor seems to be here the labial aspect of / , which rounded a weakly articulated velar r or γ. Thomas offers about a dozen forms where the same process has occurred after labial consonants, and adds that when r is followed by o, it is either rounded or suppressed altogether after any consonant: Fr.

frjmaz

Tr.

frote kr?Jy

FT.Ir > rw,

Fr.

fouömaie fomaie foter fouöter cochi couöchi

'cheese' 'to rub' 'crooked'

n u l l / H n . , BEFORE THE LABIAL GLIDE w

roz

Hn.

krjfc efrote krwar rwa

rwoz, WJZ r w ) f , w )f krwjfct, kwjfet frwDte, fwjte krw£, kwe rwa, wa

'rose' 'stone' 'hook' 'saucy' 'believe' 'king'

DR 11. R-rounding + ens + lab -

obs

>

[ + labof ]

/ ens

— clos _

166

CREOLIZATION

r is labialized before a labial vowel or glide (labof = labial offset). Sometimes the labialization is complete. Fr.

Hn.

rjf otœr

rwDf rwote

'stone' 'height'

DR 12. Rounded R deletion (OPT) + — + Fr.

ens obs clos labof krwar rwa

>

null

Hn.

I

krwe, kwe rwa, wa

[ — syl ]

'believe' 'king'

Since DR 12 is ordered after DR 11, its effect is reduced to rounded r before labial glides. This rule is "optional." Fr. r changed to a, and was then nasalized, in one word: Fr. rje, Hn. afe. 'nothing'. This change can be explained by the position of the segment (before a non-syllabic), and by the spectrum of French velar r, whose formant structure is usually very similar to that of a.

CHAPTER FORTY-FIVE

Cluster Reduction

Cluster reduction is widespread in both English and French creóles, where it has affected mostly groups of stops and fricatives or /. The changes appear to have been somewhat irregular, and are still unstable, in some English creóles. I will show, however, that an underlying regularity can be detected.

FRICATIVE + STOP IN FINAL POSITION ENGLISH

KRIO

wasp against first burst must last rest coast breast

waswás ßns ges gens fis φα s bis mDs

GULLAH

GUYANA

WJS gens f^s b)s m?s

SRANAN

JC

ges

gens agéns

bjs músu

CAMEROONS

bds mas laas

res

rzs kos

bres

Similarly, French t has disappeared in word-final position after s, as well as after k: FRENCH

HAITIAN

datisi zyst vest kDmynist palmist pjastr direkt

datis zis ves korrnnis (pje-)palmis Pias direk

'dentist' 'just' 'coat' 'communist' 'palm-tree' 'gourde' 'direct'

After I, English creóles dropped the stop, while French creóles retained it, and lost the liquid r (but not ΐ)\

168

CREOLIZATION

ENGLISH

KRIO

GULLAH

salt old world

SJI

Sil ol ßdl

ol WJ!

FRENCH

HAITIAN

kylt rekDlt kart vert ( f ) port

kilt larekolt kat vet Ρ*

JC

GUYANA

sal wol

Ol wjrl

CAMEROONS SJL

ol w)l

'cult' 'crop' 'card' 'unripe' 'door'

The last three examples are not real clusters, since r is deleted after the vowel in the same syllable. Only Guyana Creole retains t in this position: e.g., E. salt > G. silt. Sranan and Saramaccan also keep it, but only because it has been protected by an epithetic vowel: E. salt > Sr. sówtu, S. sátu. In initial and medial positions, French creóles do not reduce these clusters, whereas a regular change has occurred in all the English creóles when s preceded a voiceless stop: s is deleted before p, t, or k, except when a vowel has been inserted between the fricative and the stop, as in Cameroons Pidgins. E. #sp spoon spoil split

Cr. $p K. pun; C. sipún K. pwel;C. sip31 (G. spajt) K., Gu., JC plit ( + JC plik )

E. #st story street

Cr. #t K., Sr. fori; C., Gu., G. tiri K. trit\Gu., JC, G. tri.t

E. #sk squeeze scratch

Cr. #fc G., JC kwi.z; Gu. kßi:f Κ., Gu., JC kratJ

E. st IV—V yesterday master sister

Cr. t JC jeside K., S., C., JC mdsa\Gu., Sr., S. sisa

mi:ss

There are exceptions to these changes: e.g., 'mosquito', in which all Englishbased Atlantic creóles have preserved the cluster sk, 'sister', with st in Cameroons Pidgin, JC, and Guyana Creole, and the occasional fusion of initial st

CLUSTER REDUCTION

169

into an affricate, especially in Krio, Sranan, Cameroons, and JC (E. stab > K., JC, C. cuk, Sr. )Áku, G. dzuk, G u . - f u k , JC Juk). But the rule applies in most instances. In medial position, the situation seems to be unstable in Krio, Cameroons, and Jamaican, while t is still maintained in Guyana in all occurrences of intervocalic si, judging by Hancock's data (1969:58-59): ENGLISH

KRIO

CAMEROONS

JC

GUYANA

master sister

mdsa, mdsta sisí, sista

mdsa sista

mdsa sista

mdsta sista

These apparent irregularities may also be due to stylistic levels or to characteristics of a creole-to-English continuum. Fr. k is deleted before a fricative between two vowels (Fr. saksjfinist > Hn. sasofonis 'saxophonist'), or followed by a stop (the only examples available are of k before str and sp). It is kept in other clusters. Fr.

ekspjzisjJ cksper ekstrem-Jksjy aksjj esklav biskiii

Hn.

espozisj 3" espe lestrcmosjj aksfi esklav biskwit

'exhibition' 'expert' 'extreme-onction' 'action' 'slave' 'cookie'

Hn. k for Fr. t is rare; e.g., Fr. sim(a)tjzr, Hn. sTmikjc 'churchyard', Fr. (archaic) pit, Hn. pjk 'dumb' (hand). Cf. Jamaican Creole dzerjklman 'gentleman', likl 'little', bikl 'victuals' (Le Page 1957:384), plik 'split' (Hancock 1969:41).

Fr J g > n u l l / H n . , BEFORE ζ

Fr.

egzapl Zgzekyte

Hn.

ezap ezekite

'example' 'execute'

Although the evolution of sp, st, and sk seems to have been slightly different in final position from initial and medial positions, the overall process has been a generalized simplification of these clusters at a boundary by truncation of their more exposed component: the fricative in initial position, the stop at the end of the word. Between vowels, the weaker element, the fricative, was deleted. The change is thus identical in English and in French creóles.

170

CREOLIZATION

F r . / / > n u l l / H n . , IN WORD-FINAL POSITION AFTER A CONSONANT

Fr.

sepl 5kl serkl tabi mcebl trijagl

Fr.// > //Hn., Fr.

Hn.

s e p n J k sek tab m e b trißri

'simple' 'uncle' 'circle' 'table' 'furniture' 'triangle'

ELSEWHERE

plyme seel

Hn. p T i m e sel

Exceptions: Fr. j b l i z e , Hn. o b i z e 'force'; Fr. kalsD,

'pluck' 'alone' Hn. kasJ

'pants'.

This deletion did not occur in the English creóles, perhaps because of the syllabicity of liquids in this type of phonotactic environments, and of the frequent epenthesis which protects these liquids. pineapple tremble middle candle kettle devil people

Gu. p à j n a p l tri:mbl midi candi citi deßl pi.pl G., JC, Gu.

Κ. p a j n ä p u l trimbul midul cändul tikitul début pipul

We have seen that the clusters It and Id are reduced to I in final position. This simplification is reversed when / is followed by a labial consonant at the end of a word. Then the labial is preserved and / deleted: E. help, self, JC h e p , s e f . These data from Le Page (1957:384) are confirmed in Hancock (1969: 39): G., JC, and Gu. h e p , k e p , and C. h é l e p , where / has been protected by an epenthetic vowel. Thus a single rule seems to be at work here. It deletes / before a labial or a dental stop. The history of modern English shows that / has been dropped before velars and labials k (walk, talk, folk, yolk), m (psalm, qualm, holm), ν (halves, calves), and / in the group - a l f (half, calf, halfpenny), while its deletion before t , d , p , and b is exceptional. The creóles have generalized this deletion process to all "implosive" Γ s, whatever consonant followed. Yet it would probably be incorrect to see here primarily an African amplification of the English

CLUSTER REDUCTION

171

change. The latter is well attested in the seventeenth century before m, k, f , v, but orthoepists show that I had also been lost before ρ and b as early as the second half of the century. Since at that time the change had evidently been spreading from the "vulgar" styles for at least several decades, i.e. about the time when creolization was taking place, it is not surprising to find it in Creole, where the lack of a written tradition prevented the latinistic retrieval of the I. I would therefore consider this generalized process in English creóles a dialectal, rather than a creole characteristic. DR 13. Unreleased velar deletion + clos + high < - vcd >

> null I

(%)

+ fri - high < + act >

[

+ obs] clos

DR 13 reduces clusters such as gz, ksp(l), to z, sp(l). Unreleased g is normally deleted before a ( — high ) fricative. (ζ is actually the only fricative which has been found after g so far.) k is deleted before s if s is itself followed by a stop: Fr. ckspozisjj, eksplwa, Hn. espozisß 'exhibition', esplwa 'feat'. Otherwise k is retained (e.g., Fr. aksjj, Hn. aksj5" 'action'). Hn. lestremjsjj (Fr. dcstremJksjj) is an exception, since k is deleted before sf. DR 14. Word-final cluster reduction [ + ens ] > null / [ + obs ] Fr.

katefism korekt

Hn.

katefis korek

# 'catechism' 'correct'

The use of [ + obs ] in the environment portion of the rule excludes final clusters whose first element is nasal (these clusters are prevented by the sequence structure conditions of French), or a liquid: r disappears in this position, and I does not trigger the deletion of a following consonant in French Creoles (e.g., Fr. kylt > Hn. kilt 'cult').

CHAPTER FORTY-SIX

r Laxing and Deletion

F r . / r > null/Hn., EVERYWHERE

EXCEPT

IN SYLLABLE-INITIAL

POSITION,

WHETHER /"WAS ALONE OR IN A CONSONANT CLUSTER IN THE FRENCH ETYMON

In the phonology of modern Haitian, the deletion of r in a cluster leaves the syllable closed for all rule purposes, except when other final consonant deletions have occurred. Fr.

sykr ρ3'dr Jodjer nujjrk pzrdy merkredi

Hn.

sik ρm fodje nujjk pedi mekredi

'sugar' 'lay eggs' 'saucepan' 'USA, New York' 'lose' 'Wednesday'

This loss of r in French creóles has already been discussed at length (Part I; Part III, chapters 30, 33). It can be viewed either as the continuation of a French process which was checked in France by a learnéd reaction but developed freely in Creole, or as the creole amplification of a change which was at work naturally in the base language. The evidence is less revealing in the English creóles, since r had been vocalized and lost before a consonant in English since as early as the end of the thirteenth century (Dobson 1957:967), i.e., long before creolization. The deletion of r has proceeded in the base language to this day, with mainstream American English representing a conservative dialect while R.P. and New England are more progressive in this domain. The result is that the English creóles, Standard English, and the French creóles all have maintained r within word boundaries, but only before vowels. French creóles and some of the English creóles (Krio, Sranan, Saramaccan, Gullah) have lost the r entirely in all other environments, while others (Guyana and JC) have retained a residual vowel length, as has R.P. English. E.

hurt dark

K. at\ JC, G. K., C. JC, Gu.

Sr. dti\ ha:t dak da:k

Gu.

hut

r LAXING AND DELETION

17 3

DR 15. r deletion + ens — obs — clos

>

null /

[ - syl ] o

r is deleted in syllable-final position or before a consonant. DR 16. r laxing + ens — obs — clos

>

[-fri]

r loses most of its turbulent noise component, without however losing its consonantalism. In effect, the French velar fricative r becomes a post-palatal y, whether it has been labialized by DR 11 or not. DR 15 corresponds to a Haitian segment structure condition, since 1) r never alternates with a different segment, and 2) DR 16 is context-free.

CHAPTER FORTY-SEVEN

Minor Creolization Processes

h aspiré

IN HAITIAN

O F \h > r, γ / Η η . ( > NULL, BUT PREVENTS L I A I S O N / F R . )

afe (hacher) ele (héler) o (haut) air (haïr) ape (happer) enir (hennir)

Hn.

rafe rele ro raji rape rani

'chop' 'call' 'tall' 'hate' 'snatch' 'bray'

This change did not occur in Hn. haf (Fr. a f , "hache") 'ax', and Hn. hen (Fr. ε η, "haine") 'hatred'. On the other hand, a verb such as eme ("aimer"), which never had an initial h in French, is reme 'to love' in Creole (while the related Fr. amur displays the agglutinated article I in Hn. lamu 'love'). DR 17. h aspiré "— ens _— ved

>

"+ ens " _+ ved

/ #

h changed to r (or to γ ) in initial position. Another description of this evolution can be given by omitting DR 17 entirely, if the change took place before creolization in some French dialect, and was simply carried over into Creole. Hn. r corresponds in the words listed above to French h aspiré, which is no longer pronounced in Standard French, but still prevents liaison with the preceding morpheme. Léon's study of h and r in Norman French and in Canadian French underlines the systemic and phonetic closeness of h and r (a velar sound) in contemporary Norman French: "A l'audition, le h normand donne l'impression d'une consonne émise non pas au niveau du larynx, mais plutôt dans la région du pharynx. Il semble, le plus souvent, qu'une forte constriction se produise grâce à une élévation du dos de la l a n g u e . . . On peut penser qu'il s'agit là d'un h de type 'consonantique', par opposition au h canadien qui

MINOR CREOLIZATION PROCESS

175

semble bien produit au niveau du larynx, grâce à un élargissement du canal, ce qui classerait le phonème dans une catégorie 'vocalique'" (1967:127). Léon's hypothesis may provide an explanation for Hn. r derived from an Old French h or from a dialectal French h. We can infer from his study that OF went through two different changes. It became less and less consonantal 8 in Canadian French and in Standard French, whereas its articulation moved into the oral cavity in Norman French and in Haitian. In some Canadian French dialects, h is still pronounced (Léon, p. 140), but its function is that of ajuncture. In Standard French, it is no longer pronounced, but has left a residual juncture phenomenon (the absence of liaison, and the non-elision of schwa in an immediately preceding determiner). In Haitian, as in Norman French, it has come to be articulated with the back of the tongue; we thus find a pharyngeal h (genetically very close to r) in Norman French, and a velar "h", i.e., in Portau-Prince Haitian. In both dialects " h " has become a true consonant, [ — syllabic, + consonantal ] in my featural system. The opposition h vs. r is limited in Norman French (Léon, p. 128). It is almost non-existent in Haitian, where h is residual in the morphemes where it appears. Thus the relationship between h and r, both systematically and phonetically, is strikingly similar in Norman French and Haitian on the one hand, and on the other hand in Standard French and Canadian French.

BILABIAL FRICATIVES IN GULLAH

Except in isolated formatives such as for instance Hn. fwal < Fr. ¡(d)val 'horse', fricative consonants have been kept unchanged in all creóles. The only rule-governed exception is Gullah, where labio-dentals have become bilabial in all environments, following an Ewe rather than a Spanish model (data from Hancock 1969): flog fowl softly elephant tough knife angry never move

Gu.

φlag φαηΐ ϊζφΐΐ οΐΐφόπι ί&φ ηαίφ ßeks neßa mu:ß

This change has very few exceptions, and they are themselves irregular: e.g., ji:, φεβίβ 'give, forgive', φαί ~ φα]β 'five'. This bilabiality is certainly of African

176

CREOLIZATION

origin. It would be interesting, however, to find out why it is represented in Gullah alone, judging by published sources, and to some extent in Papiamentu, where it seems to be irregularly distributed (see Navarro-Tomás 1953:184), and is perhaps disappearing. Gullah is also exceptional in its treatment of the English bilabials w and b. w has changed to the fricative β in all positions except final. Whether the sound has retained its velarity is not noted by students of Gullah. E.

twelve always one what everywhere

Gu.

tßel 5:ßez 0¡i η ßdt eßsße

The evolution of b also seems to be highly regular, despite a form such as bdhdjm 'behind', found in Hancock, b has been weakened to a fricative only before e, and sometimes i (with numerous exceptions, perhaps due to a broad transscription—see Hancock 1969:36-67). It was protected by a following liquid. bad better big black blind brush

Gu. ßad ßzra ßig blak blajn bref

From a purely systemic analysis, it would seem that this change should have been paralleled by d>b and g > y in the same environments. Yet, not only did this not happen, but in Gullah, as in all the English creóles and in some varieties of "Black American English,"