Phonology and Morphology of Creole Languages [Reprint 2014 ed.] 9783110929560, 9783484304789

Citation preview

Linguistische Arbeiten

478

Herausgegeben von Hans Altmann, Peter Blumenthal, Hans Jürgen Heringer, Ingo Plag, Beatrice Primus und Richard Wiese

Phonology and Morphology of Creole Languages Edited by Ingo Plag

Max Niemeyer Verlag Tübingen 2003

Bibliografische Information der Deutschen Bibliothek Die Deutsche Bibliothek verzeichnet diese Publikation in der Deutschen Nationalbibliografie; detaillierte bibliografische Daten sind im Internet über http://dnb.ddb.de abrufbar. ISBN 3-484-30478-2

ISSN 0344-6727

© Max Niemeyer Verlag GmbH, Tübingen 2003 http://www. niemeyer. de Das Werk einschließlich aller seiner Teile ist urheberrechtlich geschützt. Jede Verwertung außerhalb der engen Grenzen des Urheberrechtsgesetzes ist ohne Zustimmung des Verlages unzulässig und strafbar. Das gilt insbesondere für Vervielfältigungen, Übersetzungen, Mikroverfilmungen und die Einspeicherung und Verarbeitung in elektronischen Systemen. Printed in Germany. Gedruckt auf alterungsbeständigem Papier. Druck: Hanf Buch- und Mediendruck GmbH, Pfungstadt Einband: Industriebuchbinderei Nädele, Nehren

Table of Contents

Preface Introduction

VII IX

Section 1: Segments and syllables Christian Uffmann Markedness, faithfulness and creolization: The retention of the unmarked

3

Albert Valdman and Iskra Iskrova A new look at nasalization in Haitian Creole

25

Emmanuel Nikiema and Parth Bhatt Two types of R deletion in Haitian Creole

43

Sabine Lappe and Ingo Plag Rules vs. analogy: Modeling variation in word-final epenthesis in Sranan

71

Norval Smith New evidence from the past: To epenthesize or not to epenthesize? That is the question...91 Emmanuel Schang Syllabic structure and creolization in Saotomense

109

Section 2: Stress, tone and intonation Anne-Marie Brousseau The accentual system of Haitian Creole: The role of transfer and markedness values

123

David Sutcliffe African American English suprasegmentale: A study of pitch patterns in the Black English of the United States

147

Section 3: Morphophonology Winford James The role of tone and rhyme structure in the organisation of grammatical morphemes in Tobagonian

165

VI Shelome Gooden Prosodic contrast in Jamaican Creole reduplication

193

Thomas Β. Klein Syllable structure and lexical markedness in creole morphophonology: Determiner allomorphy in Haitian and elsewhere

209

Section 4: Derivational morphology Margot van den Berg Early 18th century Sranan -man

231

Patrick Steinkrüger Morphological processes of word formation in Chabacano (Philippine Spanish Creole)..253 Nicholas Faraclas The -pela suffix in Tok Pisin and the notion of 'simplicity' in pidgin and Creole languages: What happens to morphology under contact?

269

Section 5: Inflectional morphology Tonjes Veenstra What verbal morphology can tell us about creole genesis: the case of French-related Creoles

293

Marlyse Baptista Inflectional plural marking in pidgins and Creoles: A comparative study

315

Alain Kihm Inflectional categories in creole languages

333

Subject Index

365

Language Index

369

Author Index

371

Preface

This volume contains a collection of 17 papers originally presented at the 'International Workshop on the Phonology and Morphology of Creole Languages' held at the University of Siegen, August 22-24, 2001. Four other papers from the workshop are published in Yearbook of Morphology 2002, in a special section on the morphology of Creole languages, edited by the same editor. All papers in this volume have been anonymously reviewed by co-contributors or by outside reviewers. The editor is very grateful to all contributors and the outside reviewers for offering generously their time and expertise. The quality of the papers has benefited from the constructive peer criticism and the cross-fertilization of ideas. I would also like to extend my deepest gratitude to the people who have helped in various ways in the preparation of this volume: Maria Braun and Sabine Lappe for their advice on editorial matters, and Mareile Schramm, Gisela Schwung and Guido Bongard for producing the camera-ready copy. Those readers who have ever copy-edited papers on phonology or converted electronic files containing such papers know what these people must have gone through. Finally, I thank the Deutsche Forschungsgemeinschaft, the Universität Siegen and the Fachbereich Sprach-, Literatur- und Medienwissenschaften for their financial support. Siegen, January 2003

Introduction

Until very recently, phonology and morphology have been neglected areas in the study of Creole languages. Available phonological studies were largely confined to segmental aspects, and questions of syllable structure, stress, tone and intonation seemed to be beyond the interests of most creolists and were, apart from the textbook claim that Creoles have a preference for simple CV syllable structures, largely ignored. Similarly, inflectional morphology has generally been held to be marginal in these languages, if not totally absent, and word-formation to be restricted to a few exclusively regular processes. Together with its sister volume, Plag (2003), this collection of articles aims at redressing this imbalance in the popularity of topics in creolistics. The contributions present intriguing data and new analyses that call into question the above-mentioned claims. Furthermore, this volume demonstrates that the study of the phonological and morphological systems can give us crucial insights into one of the major questions of creole studies, i.e. the question of how these languages and their grammars come about. The volume is organized into 5 sections, each focusing on particular aspects of the respective subsystems. In section 1 ('Segments and syllables'), Christian Uffmann proposes a general theory of the influence of substrates and universale in creolization, arguing that the effects of these two forces can be adequately modeled and teased apart in an optimalitytheoretical approach. Uffmann illustrates this with data from vowel inventories and syllable structures. The two papers by Emmanuel Nikiema and Parth Bhatt, and Albert Valdman and Iskra Iskrova, respectively, deal with segmental phonological alternations in Haitian, the realization of Irl and nasalization, providing detailed empirical and theoretical analyses of the pertinent phenomena. They both show that assumptions about syllable structure, underlying representations and lexical exceptions need to be integrated in order to account for the intriguing facts of Haitian Creole. Sabine Lappe and Ingo Plag's as well as Norval Smith's contribution investigate the shape and emergence of syllable structure in early varieties of Surinamese Creoles. Both papers have implications for phonological theory. With their analogical model of wordfinal epenthetic vowels in Early Sranan, Lappe and Plag problematize the notion of rule in phonology, and Smith introduces the notion of 'dependent syllable' to explain the variability of vowel epenthesis in complex onsets in the different early Surinamese Creoles. Emmanuel Schang presents little known data from Portuguese-based Sao Tomense Creole with interesting cases of metathesis and other syllable-related processes, arguing that the distinction between changes occurring in the creolization process and those occurring later is often impossible to make. The two papers in section 2 deal with 'Stress, tone and intonation'. Starting out with clear assumptions from a parametric theory of stress assignment and from the full transfer/full access mode hypothesis taken from second language acquisition theory, AnneMarie Brousseau develops a systematic comparison of Haitian stress with the stress and tone system of its major substrate language. It turns out that the hybrid properties of the Haitian prosodic system can be explained as the result of markedness resettings as they

χ

Ingo Plag

occur in the creation of interlanguages in second language acquisition (see also Veenstra's paper on inflection for a similar point). David Sutcliffe's paper deals with tonal and intonational patterns as found in the wellknown ex-slave recordings from the Southern United States. He shows that there are striking similarities between the pitch-accent patterns found in the anglophone Caribbean Creoles and African American Vernacular English, supporting the Creole influence on the development of this variety. The first paper of the following section ('Morphophonology') is concerned with the role of suprasegmental structure (i.e. tone, stress and syllable structure) in the determination of grammatical morphemes in Tobagonian Creole English. Winford James describes intricate patterns of auxiliaries, negators, determiners and suffixes, and demonstrates how tone, stress, rhyme structure, syntactic category, semantics, and syntax combine to organize these morphemes into various subsystems. Pitch patterns are in the focus of Shelome Gooden's paper on the prosodic differences between two segmentally identical types of reduplication in Jamaican Creole, distributive {swelswel 'swollen in different areas') and intensive (swelswel 'very swollen'). Based on the acoustic analysis of experimental data, a phonological analysis is proposed according to which monosyllabic and disyllabic unreduplicated words have a HL pattern. Under reduplication, this HL pattern is realized differently, depending on the semantic property of the output form, and depending on the number of syllables of the base word. The third paper in this section, written by Thomas Klein, investigates the notoriously intricate allomorphy of the definite determiner in Haitian creole. Postposed /laJ shows what could be called anti-markedness effects, in that the CV allomorph [la] is chosen with stems ending in consonants or glides (pitit la 'the child'), whereas the V allomorph [a] appears with stems ending in vowels (ru a 'the wheel'). He offers a solution to this problem in an optimality-theoretical model in which morphemes may carry constraint violations as part of the input representation to the effect that they are distinguished in part by the presence versus absence of such violations. Section 4 is devoted to 'Derivational morphology'. Margot van den Berg's article is a detailed study of the nature of the word-forming element -man ('-man', '-er'), as it occurs in the earliest sources of Sranan. According to van den Berg, the available evidence shows that this element has functioned as a derivational suffix already in the earliest stages of the language. Patrick Steinkrüger's paper gives the first systematic overview of word-formation processes in the Spanish-based Philippine Creole Chabacano, showing an intriguing mixture of morphemes and processes as contributed by the different input languages. The nature of this mixture is a serious challenge for theories of creolization. The Tok Pisin multi-faceted suffix -pela is the topic of Nicholas Faraclas' contribution. He argues that the behavior of -pela in associative noun phrases and other constructions is in most cases much more complex than the behavior of any markers used with a similar function in the superstate or substrate languages, thus showing that Creole grammars cannot be called 'simple' in any meaningful sense of the word. In the first paper of the final section ('Inflection'), Tonjes Veenstra investigates the form and function of long vs. short verb forms in French Creoles, comparing it to Basic Varieties as they occur in natural second language acquisition. He concludes that the formal properties of verbal morphology in these Creoles can be attributed to processes of second language

Introduction

XI

acquisition with the first generation of adult speakers, but that the formal opposition between long and short forms acquires its creole-specific function only after the target shift from superstrate to Creole by the second generation of speakers. The marking of plurality with nouns is dealt with in the article by Marlyse Baptista. Her cross-linguistic, corpus-based study of six Creole languages shows that, first, affixational plural marking is not at all uncommon, and that, second, such plural marking is subject to certain licensing conditions, involving animacy, definiteness and tense as determining factors. The final contribution of this volume is Alain Kihm's formal account of inflectional properties in a number of Creole languages, arguing that the lack of inflectional morphology should not be confused with the lack of inflectional categories. Based on these considerations Kihm comes to the conclusion that the controversy about the putative lack of inflectional morphology may be pointless below the observational level, where it is inconclusive. To summarize, the articles in this collection open up new perspectives for the study of creole languages, both empirically and theoretically. The contributions draw our attention to a host of hitherto unknown phenomena, propose new analyses of well-known data, and indicate promising directions for our theoretical thinking, hopefully stimulating further research on the phonology and morphology of these languages.

Reference

Plag, Ingo (ed.) (2003): The morphology of Creole languages. Special section of Yearbook Morphology 2002. - Dordrecht: Kluwer.

of

Section 1: Segments and syllables

Christian

Ujfmann

Markedness, faithfulness and creolization: The retention of the unmarked

1. Introduction1 One of the perennial questions and debates in Creole studies concerns the problem of the origin of the linguistic features found in Creoles (see e.g. Muysken and Smith 1986). Linked to this is another question as to how or whether pidgins and Creoles favor unmarked options in their grammar. More recently, Parkvall (2001) took up the debate again, claiming that these questions have somewhat moved out of the focus of research. In this paper, I will therefore try to shed new light on these questions from a theoretical viewpoint. In particular, I will focus on the question whether Optimality Theory (McCarthy and Prince 1993, Prince and Smolensky 1993; henceforth OT), the probably most influential theory of grammar today, allows us to gain new insights into problems of creolization. I will propose that OT is in fact uniquely suited to understand issues of Creole formation. Within this framework, substratist and universalist theories can be unified in a novel fashion. It can also be explained why certain features appear in a Creole; the central claim of this paper will be that those features that are relatively unmarked will show up in the Creole. In extending existing theories of acquisition in OT (Tesar and Smolensky 1996, 2000, Hayes 1999), I will propose a formal account of substrate leveling, where those substrate subgrammars are systematically retained that are relatively unmarked compared to competing subgrammars. Creoles thus display what in OT literature is referred to as the 'emergence of the unmarked' (McCarthy and Prince 1994), rephrased here as the retention of the unmarked. Seen from this perspective, Creole grammars are substrate-induced. The choice of substrate features, however, is governed by markedness considerations, for which a straightforward explanation will be offered. The remainder of this paper is organized as follows: In §2 a brief introduction to OT will be provided, which may be skipped by readers familiar with this framework. How OT can unify substratist and universalist theories will be explained in §3. After introducing theories of acquisition in §4, the main theory, how substrate leveling forces retention of the unmarked, will be outlined in §5. An example of how this principle interacts with demographics will be presented in §6, and §7 will address some residual issues.

1

Many thanks for comments, suggestions and discussion to Birgit Alber, Parth Bhatt, Anne-Marie Brousseau, Wolfram Keller, Claire Lefebvre, Bettina Migge, Ingo Plag and audiences at the Siegen conference and the universities of Marburg, Toronto and UQAM. All remaining errors are my own.

Christian Uffmann

4 2. I n t r o d u c t i o n t o O p t i m a l i t y T h e o r y

The development of OT (McCarthy and Prince 1993, Prince and Smolensky 1993) meant a major paradigm shift in phonological theory. OT no longer assumes a set of ordered rules which transforms an underlying representation into an output form (or surface representation). Instead of serial derivations, there is only one level of derivation, on which we find a set of parallel constraints on the well-formedness of output forms. Hence, the viewpoint changes from the input, which has to undergo some changes, to the output, the well-formedness of which is determined by a set of ranked and violable constraints. In short, the mechanism works as follows: From an input (the underlying form) a set of potential output candidates is generated by the GENerator function. This set of candidates is then evaluated by the EVALuator function, which contains the constraint ranking. The candidate with the fewest violations of high-ranked constraints wins and is selected as the output form. This function is commonly represented in tableau form, as in the model tableau in (1). (1)

A schematic evaluation tableau Input (UR) Candidate a Candidate b

CONSTRAINT X

CONSTRAINT υ

*i *

In the left-hand column, we find (a) the input (the underlying representation) and (b) (a selection of) the candidates provided by GEN. The remaining columns contain the constraints, ranked from left to right. An asterisk indicates a constraint violation, i.e. the nonsatisfaction of the constraint in question. An exclamation mark after an asterisk means that the violation is lethal, i.e. the violation of this constraint throws the respective candidate out of the race. Shaded cells indicate that the constraint is irrelevant for the outcome (because selection has already taken place). The winning candidate is represented by the pointed finger symbol. In this example, candidate (a) loses to candidate (b) because it violates constraint X, which is ranked above constraint Y. The violation of constraint Y is irrelevant because candidate (b) remains the only candidate in the race. Reversing this ranking would mean that a different candidate, candidate (a), would be selected, because now satisfaction of constraint Y would be more important than satisfaction of constraint X. In OT, it is assumed that constraints are universal, so that differences between languages can be accounted for solely by different rankings of the same set of constraints. Reranking establishes a different grammar, and a different candidate will emerge as optimal. Another important feature of OT relevant to the argument developed here is that a distinction drawn in previous theories is collapsed, the distinction between on the one hand rules that account for active alternations and on the other hand lexical constraints that define, for example, phoneme inventories and phonotactics in the lexicon. In OT, phoneme inventories and restrictions o n t h e combinability of s e g m e n t s are also e x c l u s i v e l y ac-

counted for by a constraint ranking.

Markedness, faithfulness

and

creolization

5

2.1. Markedness and faithfulness Regarding the nature of constraints, two major classes of constraints can be distinguished, markedness constraints and faithfulness constraints. Faithfulness constraints (F) demand that the input be preserved faithfully, that no changes take place between input and output in other words, the output should be just like the input. Markedness constraints (M), on the other hand, demand that the output be unmarked in the sense of traditional theories of (universal) markedness (e.g. Greenberg 1966, Jakobson 1969). A marked structure in the input should thus not appear in the output. Below, lists of typical faithfulness and markedness constraints that are well-established in OT literature are given: (2)

Typical faithfulness constraints (in Correspondence Theory, McCarthy and Prince 1995) MAX DEP IDENT(F)

(3)

NO deletion No epenthesis No change in feature specification

Typical markedness constraints (cf. Kager 1999) * DORSAL *[+rd, -bk] NOCODA ONSET *CLASH

NO dorsal segments No front round vowels Syllables do not have codas Syllables have onsets NO adjacent stressed syllables

Obviously, constraints of the two classes may be in conflict, for they often make opposing demands. Faithfulness constraints demand that nothing be changed while markedness constraints demand maximal unmarkedness of an output, even at the expense of applying changes to the input. This is why constraints are ranked: Ranking faithfulness constraints and markedness constraints with respect to each other resolves this conflict, telling the speaker what is to be seen as more important in her language, a faithful input-to-output mapping, or unmarked structure in the output. If a markedness constraint outranks the corresponding faithfulness constraint, the result will be that a candidate which is unmarked with respect to this feature will be selected, for candidates must not violate this markedness constraint. On the other hand, if the faithfulness constraint outranks the markedness constraint, markedness considerations do not play a role in establishing the output form. The grammar then simply demands that the input must not be altered with respect to some structure.

6

Christian Uffmann

3. The creole analogy

This distinction between markedness and faithfulness constraints is reminiscent of a perennial discussion in creole studies, the discussion between so-called substratists and universalists. According to the universalists, Creoles are instantiations of the default (unmarked) options of Universal Grammar. The most influential theory here is Bickerton's (1981) Language Bioprogram Hypothesis, according to which Creoles are formed by children. Due to insufficient input in their multilingual environment, they resort to their innate bioprogram, the standard settings of Universal Grammar. This view is challenged by substratists who believe that creole structures come predominantly from the substrate languages, for example through relexification of superstate phonetic strings, as one of the more radical substratist theories asserts (Lefebvre 1986, 1998). In other words, universalists claim that Creoles represent the default settings of Universal Grammar, while substratists argue that Creole grammars are faithful preservations of their substrate grammars. In the following, I will propose that these two views are not mutually incompatible, but that O T offers an elegant solution to reconcile the opposing sides. In OT, universals are integrated into the grammar in the form of markedness constraints, while the notion of substrate retention (or transfer) can be modeled as the retention of a substrate ranking. Thus, substratist and universalist theories can converge in an OT account.

3.1. Towards a unification of theories In their analysis of Sranan syllable structure, Alber and Plag (2001) show how this convergence can come about. They argue that Sranan syllable structure is both, universally unmarked (owing to its tendency towards CV syllables) and also found in the substrate languages. In their O T analysis, this is explained by high-ranked constraints on syllable structure (in particular, the CODACOND constraint, ONSET and constraints on sonority) which outrank the respective faithfulness constraints. This high ranking of markedness constraints at the same time formalizes the universalist hypothesis (calling for the high ranking of markedness constraints) but can also be straightforwardly explained as the retention of a substrate ranking. While Alber and Plag notice this ambiguity and point out that creole structures can be both substrate-induced and unmarked at the same time, they do not exploit this observation to its full potential. In the remainder of this article, I will therefore show that this observation can be generalized into a novel view of creolization. The generalization is that if some creole subgrammar has a markedness constraint outranking the respective faithfulness constraint and the substrate grammar has the same ranking, then the two schools, universalism and substratism, converge. The universalist view is supported by the unmarkedness of the output structure, while the substratist view is supported by the identity of creole and substrate rankings. So, if and then

the creole grammar has Μ » F the substrate grammar has Μ » F substratism and universalism converge.

Markedness, faithfulness

and

creolization

7

Within OT, the seeming incompatibility of the two views can thus be overcome in a dialectical fashion because universals, formalized as markedness constraints, are a central part of grammar itself and not a principle operating outside the grammar proper. It follows then that the emergence of unmarked structures in Creoles can also be conceptualized as the retention of unmarked substrate structures. A small case study will make this point clearer.

3.2. The emergence of the Haitian Creole vowel inventory As mentioned above, phoneme systems, too, are defined by output constraints in OT. If markedness constraints on certain classes of sounds outrank a faithfulness constraint, these sounds will not appear in the output - a less marked (unfaithful) sound will be selected instead. Therefore, I will discuss Creole vowel systems as fairly simple instances of constraint interaction to illustrate my point, beginning with a discussion of the vowel inventory of Haitian Creole (HC), comparing it to both, superstate and substrate vowel inventories. Starting with a comparison of the vowel inventories of French and HC, 2 the most conspicuous difference lies in the absence of front rounded vowels from the HC system, as can be seen in (4): The vowels /y, 0, ce/ are missing from the HC inventory. Therefore, Haitian is relatively unmarked vis-ä-vis French, as front rounded vowels are commonly seen as relatively marked segments (e.g. Chomsky and Halle 1968). (4)

French vowel system 3 1 e ε a

y 0 os/a

u 0 0 a

HC vowel system (Tinelli 1981) I e ε

u ο 0 a

A look at the major substrate language of HC, Fongbe (Lefebvre 1986, 1998), however, reveals that in Fongbe, too, front rounded vowels are missing. Its vowel system is essentially the same as that of HC and also represents the unmarked option with respect to the presence or absence of front rounded vowels.

2

Only oral vowels are compared here, for reasons of simplicity as well as the uncertain status of underlying nasal vowels in both languages. Phonetically, HC has five nasal vowels, the same as Fongbe, although [ü] and [i] are marginal in HC. With respect to surface nasal vowels, the HC system is thus like Fongbe as well but unlike French. It might represent a marked option vis-ä-vis the superstrate, but nevertheless the complexity of the nasal vowel system is licensed by the substrate.

3

The vowel inventory given here is that of 17th century popular French (Lefebvre 1998). Note a second difference where HC is unlike French but like Fongbe and at the same time unmarked, viz. the number of low vowels. French (in some variants until today) has two low vowels, whereas HC and Fongbe have only one.

8

(5)

Christian Uffmann

Fongbe vowel system (Capo 1991) i e ε

u ο ο a

This ambiguous state of affairs - the vowel inventory is at the same time relatively unmarked and identical to the substrate - can be captured precisely in an O T analysis, for markedness is encoded directly into the grammar. Within this framework, we can assume one markedness constraint and one faithfulness constraint, the relative ranking of which will decide whether front rounded vowels are allowed in a grammar or not. The relevant markedness constraint is given in (6): (6)

*[+rd,-bk]

front rounded vowels are disallowed

This constraint militates against the occurrence of front rounded vowels and is assumed to be universal. Whether a language allows front rounded vowels or not will depend on the ranking of this constraint with respect to its faithfulness counterpart, defined in (7): (7)

iDENT(rd)

no change in the specification of the feature [round]

If the faithfulness constraint iDENT(rd) outranks the markedness constraint against front rounded vowels, front rounded vowels in the input will also appear in the output because the demand not to change the specification of [round] is stronger than the demand not to have front rounded vowels in the output. If the markedness constraint, however, outranks iDENT(rd), the occurrence of front rounded vowels in the output will be blocked by this constraint. French is an example of the first option: The occurrence of front rounded vowels in French means that the iDENT(rd) constraint ranks above the *[+rd,-bk] constraint, while the observation that front rounded vowels do not occur in HC and Fongbe (and that front rounded vowels in French etyma are generally unrounded in HC 4 ) can be explained as the reversal of this ranking. The rankings are given in (8-9). (8) (9)

French: HC/Fongbe:

iDENT(rd) *[-bk,+rd]

» *[-bk,+rd] » iDENT(rd)

= =

F » Μ Μ » F

The tableaux in (10) demonstrate the effect of the two rankings. A hypothetical vowel /y/ in the input is treated differently in French and HC/Fongbe. Violation of the markedness constraint *[+rd,-bk] is non-lethal in French because it means satisfaction of the higher-ranked constraint iDENT(rd). The less marked output candidate [i] thus loses in the competition, for violation of the I d e n t constraint is lethal.

4

In a few idiosyncratic cases (e.g. jous from juste) we find backing instead of unrounding (Tinelli 1981 and one anonymous reviewer).

Markedness, faithfulness

and

(10) French ranking /y/

iDENT(rd)

HC/Fon ranking n + r d , -bk] *

[i]

9

creolization

*!

¥

*[+rd, -bk]

iDENT(rd)

[y] *

The situation is different if the ranking is reversed, as in HC and Fongbe: Here, violation of the markedness constraint *[+rd,-bk] is lethal, as the first candidate [y] demonstrates. Instead, the less marked candidate [i] will be selected, at the expense of an IDENT violation. Essentially, HC and Fongbe have a ranking where a markedness constraint outranks a faithfulness constraint, while in French the faithfulness constraint outranks the markedness constraint. The ranking of substrate and Creole is thus identical, which supports the substratist view. On the other hand, the ranking enforces an unmarked output, which supports the universalist view. Both views are therefore accounted for and rolled into one constraint ranking. The account of the HC vowel inventory is thus very similar to that of Sranan syllable structure, as proposed by Alber and Plag (2001), discussed above. The decisive claim which I am going to make in the remainder of this paper, however, is that unmarked substrate structure is systematically retained in creolization. In addition, this retention of the unmarked does not arise from some independent principle such as a bioprogram. Instead, it follows directly from the substrate and general principles of constraint interaction and learning within OT. How this can be conceptualized, will be discussed in the following.

4. T h e o r i e s of acquisition in O T

Before turning to the issue of creolization in OT, we should first turn to a discussion of acquiring grammars in OT from a more general point of view. In OT, the acquisition of a grammar is equivalent to establishing a constraint ranking, the process of which has been formalized as an algorithm by Tesar and Smolensky (1996, 2000). Tesar and Smolensky propose a "Constraint Demotion Algorithm" as a model of first language acquisition. In the beginning all constraints are contained in one stratum {Ci ... C n }, devoid of any initial ranking. The algorithm then establishes a ranking by a process of demoting individual constraints, i.e. by placing them on a lower stratum. Constraints that are violated in a winning candidate are demoted because violation of such a constraint is obviously not important: It can be violated without any consequences for the selection of the winning candidate. This procedure is repeated until all constraints are ranked and the grammar thus generated matches the observations. Several modifications have since been proposed for this model. Gnanadesikan (1995) argues against one single initial stratum. Taking into account actual child data, she finds that the earliest utterances of a child are maximally unmarked. Hence, she proposes two initial strata, one containing the set of markedness constraints and one containing the set of faithfulness constraints, where the set of markedness constraints outranks the set of faithfulness

10

Christian Uffmann

constraints. Because all markedness constraints outrank all faithfulness constraints in the beginning, children start off with unmarked utterances and not with faithful utterances devoid of alternations (but cf. Hale and Reiss 1998 for an alternative view). Hayes (1999) proposes an additional amendment to the original model developed by Tesar and Smolensky. He finds that their algorithm sometimes fails to capture regular alternations, instead producing a grammar which overgenerates faithful input-to-output mappings, due to lack of negative evidence. Given allophonic variation between two sounds, the algorithm will often treat the difference as phonemic, not as a regular alternation. Thus, Hayes adds an important condition to the demotion algorithm: Faithfulness constraints should always be ranked as low as possible if there are several ranking alternatives. This way, variation will be detected and accounted for by an appropriate markedness constraint. Hayes' condition amounts to introducing Occam's Razor into O T acquisition theory, because it enforces a maximally economical inventory of underlying segments, letting variation account for as many output phenomena as possible. In sum, learning can be described as constraint demotion in OT, as demonstrated by Tesar and Smolensky (1996, 2000). Low ranking of faithfulness (or high ranking of markedness) is of universal importance in this process, however, first in distinguishing two initial strata (Gnanadesikan 1995) and second in placing conditions on the demotion algorithm (Hayes 1999).

4.1. Language shift and SLA In second language acquisition (SLA) and other processes of acquisition in language contact, for example language shift, learning obviously proceeds differently from first language acquisition. Most importantly, there is no initial state, an unordered stratum containing all constraints. Instead, there already exists one specific ranking of the universal set of constraints, viz. the speaker's native language. In language shift and SLA, the starting point for establishing a new ranking is thus different, it is an already existing ranking, in other words, an established grammar (Broselow et al. 1998, Grijzenhout and van Rooy 2001). The difference is illustrated in (11). (11)

LI acquisition target grammar

SLA/Language shift

G, /N

G2 A

initial state (one stratum)

G]

(targeting) starting point

G, is shorthand for the first grammar one acquires, one's mother tongue. While G! is the endpoint of LI acquisition, established from a previously unordered set of constraints, it is the starting point for SLA; there a G 2 , the second language grammar, is the target grammar, which has to be established by reranking one's Gi, without recourse to the tabula rasa of the initial state. Learning is thus reranking of G b a set of already ranked constraints.

Markedness, faithfulness and

creolization

11

Let me exemplify this somewhat abstractly. Suppose there is a target grammar (the supers t a t e or just any target of SLA) which has a specific ranking different from the ranking of the same constraints {Χ,Υ,Ζ} in the LI (or the substrate) such that the target grammar has a ranking X » Υ » Ζ and the LI has a ranking Υ » X » Z, then demotion of Y below X is equivalent to the acquisition of a part of the target (or superstate) grammar. So, in SLA and language shift, too, acquisition can be described as the demotion of constraints. This simple mechanism is able to model different outcomes of language shift. Perfect shifting of a language group without retention of substrate features is equivalent to the perfect reranking of one's native grammar, i.e. performing all necessary constraint demotions to match the target grammar. The retention of substrate features, however, means that the reranking was imperfect, that partial rankings of the substrate grammar were retained, although the target grammar was largely acquired. The concept of relexification (Lefebvre 1986, 1998) finally means that no reranking takes place: The lexical forms of a target language are acquired but none of its grammar, that is, none of its specific rankings. The native grammar is retained as a whole. Relexification can thus be seen as the starting point and the basis of the acquisition of another language, for it means the transfer of a ranking.

4.2. The problem of creolization The situation in creolization on the other hand is significantly more complex than that of language shift and cannot be explained alone by this model. First, a one-to-one relation between substrate and superstate is missing. Instead, scenarios of creolization are typically described as multilingual: Speakers of several languages form the incipient Creole community. Second, learning is only partially targeted, if at all. The primary driving force behind creolization is the need of substrate speakers to communicate among themselves (e.g. Thomason and Kaufman 1988). The emergent Creole thus has to serve two communicative functions, intra-group and inter-ethnic. Targeting of the superstate is of subordinate importance in this situation and varies with accessibility of the superstate, which might be fairly accessible in a situation where many superstate speakers are present, living in close proximity with substrate speakers, or it might be essentially inaccessible, for example in Sranan, a case of early superstate withdrawal (again, see e.g. Thomason and Kaufman 1988). One important process in the emergent creole grammar, at least according to the substratists, therefore is leveling of the different substrate grammars (see e.g. Lefebvre 1998). The crucial question then is what happens in leveling. In the following section, I will therefore pursue two questions. First, how do grammars converge in leveling when a target is absent, and what is the OT position on this? Second, how does acquisition - acquisition as constraint demotion - work in a situation of substrate leveling?

12

Christian Uffmann

5. T h e retention o f the u n m a r k e d

The problem with substrate leveling is to determine which language will win with respect to some property of the emerging grammar when substrates diverge. If there is one clearly dominant group in the substrate, either because speakers of that language outnumber speakers of other languages, or because the group is socially prestigious, the solution might be easy; speakers of that language can impose (elements of) their grammar on the rest of the community. Often, however, the substrate is heterogeneous, composed of several languages, of which none is in a dominant position vis-ä-vis other substrate languages. In such cases, I suggest, conflict will be resolved by choosing the least marked grammar of the competing substrate grammars, i.e. the grammar in which faithfulness is ranked lowest. As demonstrated above, low-ranked faithfulness is of utmost importance to successfully acquire a grammar, either by stipulating two initial strata, as proposed by Gnanadesikan (1995), or by placing conditions on the demotion algorithm (Hayes 1999). This principle is also found in scenarios of leveling: The substrate with lowest-ranked faithfulness (as regards some property of language) will be selected, because, universally, low faithfulness will always be preferred over low markedness. Formally, this principle can be paraphrased as follows: (12) Let Si and S 2 be substrate grammars (where Si Φ S 2 ) and S , 2 the unification of these grammars (leveling), and let there be a different ranking for some subgrammar for S t and S 2 . Then, if and then

S, = S2 = S|, 2 =

Μ»F F »Μ Μ» F

This principle means that if there are two substrate grammars, and the grammars differ with respect to some property, such that one grammar has a markedness constraint outranking a faithfulness constraint, and the other grammar has the reverse ranking, then the grammar with the high-ranked markedness constraint, viz. the less marked grammar of the two, will be selected in leveling, which in this example is substrate language 1 (S|).

5.1. The emergence of the Ndyuka vowel inventory An effect of this principle can be found in the development of creole vowel inventories. As a case in point, the Ndyuka vowel inventory may serve as an exemplification of this principle. Above, we discussed the vowel inventory of HC, comparing it to both, substrate and superstrate, and finding that it is relatively unmarked compared to the superstrate. Now let us consider Ndyuka, comparing it to its two major substrate languages, this time disregarding the superstrate. I want to show in this section that again the Creole is unmarked compared to the more complex substrate system and at the same time resembles the vowel inventory of the less marked of the substrate languages.

Markedness, faithfulness and

creolization

13

The Ndyuka vowel system is given in (13). It is a fairly simple five-vowel system of the type commonly found across the world's languages (Lass 1984). (13) Ndyuka vowel system i e

u ο a

The two major substrate languages of Ndyuka, as of all Surinamese Creoles, are Gbe and Kikongo (Arends 1995, among others). At the time of the inception of a Surinamese Plantation Creole, both groups together composed the vast majority of the slave population, and moreover, they were roughly equally numerous, according to the demographic data discussed in Arends (1995). The vowel inventories of both languages are provided in (14). Gbe has a 7- or 8-vowel system, differing between lects and leaving aside nasal vowels (see above in the discussion of HC), while Kikongo has a 5-vowel system identical to that of Ndyuka. The Gbe system is more complex than the Ndyuka system, with an [ATR] distinction among mid vowels and an additional central vowel (which is, however, absent in a number of lects). (14) Gbe vowel system i e ε

9 a

u ο ο

Kikongo vowel system i e

u ο a

The intriguing point is that in other parts of Ndyuka grammar, the importance of the Gbe substrate has been demonstrated repeatedly (e.g. Migge 1998), while the virtual absence of Kikongo features has puzzled researchers. So, why does Ndyuka not have a Gbe-style vowel system if Gbe speakers had such a profound influence on the formation of the Surinamese Creoles? The answer lies in the relative markedness of the vowel systems. The OT analysis proposed now is in many respects very similar to the analysis of HC presented above, and the reader may want to compare the two analyses for a better understanding of the processes operative in Creoles. Again, the difference between the two vowel inventories can be captured by means of the relative ranking of two constraints, one of which is a markedness constraint, the other one a faithfulness constraint. The markedness constraint *[-ATR] militates against contrastively specified [-ATR] (or 'lax') vowels, here /ε, ο/. The corresponding faithfulness constraint, on the other hand, demands that the input value of [ATR] must not change in the output. (15) *[-ATR] (16) IDENT(ATR)

no lax vowels no change in [ATR] values

(markedness constraint) (faithfulness constraint)

Ranking these two constraints with respect to each other generates two possible grammars, one with high markedness, the other one with high faithfulness. High markedness will

Christian Uffmann

14

block the occurrence of [-ATR] vowels in the output, while high faithfulness will ensure that vowels underlyingly specified as [-ATR] will also appear in the output as [-ATR], Both rankings can be found in the Creole and its substrates, as shown in (17)-(19). (17) Gbe:

IDENT(ATR) »

(18) Kikongo:

*[-ATR]

» IDENT(ATR)

*[-ATR]

= Μ» F

(19) Ndyuka:

*[-ATR]

»

= Μ»

IDENT(ATR)

= F »

Μ F

The effect of this ranking can be seen in the following tableaux: In Gbe, high-ranked IDENT(ATR) guarantees that an underlying /ε/ will also appear as [ε] in the output; satisfaction of the IDENT constraint is more important than satisfaction of the lower-ranked markedness constraint. Reversing the ranking, as in Kikongo and Ndyuka, means that a different candidate will be selected, viz. the [+ATR] vowel. Input /ε/ cannot appear in the output, because violation of the markedness constraint *[-ATR] proves lethal for the first, faithful, candidate. (20) Gbe ranking:

Id [ε] [e]

Kikongo/Ndyuka ranking:

IDENT(ATR)

* r - A T R ] *

*!

er-

/ε/

*[-ATRI

[ε]

*!

[e]

IDENT(ATR) *

In sum, the predicted emergence of unmarked structure is indeed found in Ndyuka. Both relevant substrate languages, Gbe and Kikongo, are in conflict with respect to some ranking, and the less marked one succeeds: The vowel inventory of Ndyuka resembles that of Kikongo, not that of Gbe, despite profound Gbe influence in other areas of Ndyuka grammar. 5

5.2. Explaining the selection of unmarked grammars The question, however, is why the relatively unmarked ranking succeeds in leveling. In this section, I will provide four main arguments that show why choosing an unmarked subgrammar should be preferred over choosing the marked one. First, the principle of low faithfulness as a universal principle for successful acquisition has already been mentioned. Hayes (1999) shows its applicability to first language acquisition, and by extension learners might as well use it in other contexts of language learning. In LI acquisition, the learner, if she has the option to either demote a markedness constraint or a faithfulness constraint, will opt for demotion of faithfulness. The situation is essentially the same in leveling. If substrate languages differ, leveling can occur either by demoting a faithfulness constraint or a markedness constraint. Then, the universal principle - if in 5

Bettina Migge (pc.) points out to me that Ndyuka mid vowels may be phonetically realized as [ε, ο]; the rankings in (20) thus refer only to the presence or absence of contrastively specified [ATR] values. N d y u k a speakers will always collapse a distinction between [e] and [e]; the phonetic realization of the mid vowel is hence of minor phonological concern here.

Markedness, faithfulness and

creolization

15

doubt, demote the faithfulness constraint - is transferred from what is a general principle of acquisition. Second, unmarked structure is easier to learn. This is part of the very definition of markedness, which is heavily grounded in observations on how children acquire language, and which structures are acquired first (Jakobson 1969). In SLA, too, markedness effects are frequently observed by researchers (e.g. Eckman 1977, Yavas 1994). Thus, the principle of economy comes into play: Why should learners design a relatively marked grammar, which involves greater cost in terms of learnability, if other options are available? Third, there is another aspect of the principle of economy: That opting for the less marked grammar does very often not involve learning of new structure for anyone. Demotion of a faithfulness constraint below a markedness constraint leads to a grammar whose set of structures is a proper subset of the more marked grammar. In Ndyuka, the five-vowel system is a proper subset of the Gbe system. The Ndyuka/Kikongo system contains exactly the same five vowels as the Gbe system minus two additional mid vowels. Thus, in leveling a Gbe speaker does not have to learn new contrasts or new vowels if she settles on the Kikongo-type system. She can dispense with two vowels but does not need to learn anything new, because the new system is fully contained within the old one. For the individual speaker of the more marked grammar, nothing needs to be done to acquire the new grammar; acquisition simply means not to use certain structures any longer. Over time, the grammar can then be optimized to a grammar producing only a five-vowel system, without the superfluous seven-vowel option, by demotion of faithfulness. For a Kikongo speaker, however, acquisition of the Gbe vowel system means learning new contrasts. The retention of unmarked structure thus minimizes the effort for all speakers involved creating the Creole. Fourth, children acquiring the emerging creole as a first language are exposed to contradicting data in trying to establish the vowel inventory of their native tongue: They will hear five vowels from Kikongo speakers and seven from Gbe speakers. The same lexical item will be pronounced differently by Gbe and Kikongo speakers. The child reasonably then assumes to deal with free variation with respect to the realization of mid vowels and can use the principle of low faithfulness without having to fear that a generalization is lost, as she frequently encounters negative evidence for the hypothesis that [e] and [ε] are different phonemes. The same applies to newly arriving slaves who acquire the creole as a second language. The selection of unmarked grammars is thus a continuous process, repeated as long as substrate languages are present and reinforced by new arrivals. From all the above a prediction follows. The probability of a substrate feature to appear in the creole grammar depends on two factors: First, the relative markedness of that feature. The more marked it is, the less likely will it be retained in the creole grammar. Second, the number of speakers whose grammar contains this feature is relevant: The more speakers have this feature, the more likely it is to be retained. Both factors work in tandem, such that more speakers are needed for a relatively marked feature to survive leveling. The retention of a seven-vowel system in HC, compared to Ndyuka, can thus be explained by the demographic and cultural dominance of Gbe speaking slaves in Haiti (Lefebvre 1998), compared

16

Christian

Uffmann

to Suriname where a sizeable part of the slave population was of Kikongo origin (Arends 1995). 6 This general principle, as outlined above, is what I term the "retention of the unmarked", in analogy to McCarthy and Prince's (1994) principle of the "emergence of the unmarked". McCarthy and Prince show how synchronically the unmarked can emerge in certain grammatical domains, in particular reduplication. I extend their proposal to diachrony and suggest that here, too, there are contexts in which the unmarked can emerge, especially in the context of substrate leveling, where those parts of grammar are retained that are unmarked vis-ä-vis other grammars that participate in leveling.

5.3. Vowel inventories across Creoles But do the predictions made here hold in the light of additional data? I limited myself to vowel inventories in this paper and looked at other Creole languages to see whether they support my claim. Overall, the sample of Creoles I analyzed seems to support the theory. Five- and seven-vowel systems (not taking into account quantity, which, according to Alleyne (1980) seems to be a later development in Creoles anyway) appear to be standard among English, Dutch and French based Creoles.7 Examples of five-vowel systems are found in Ndyuka (Huttar and Huttar 1994), Sranan (Smith 1977), basilectal Belizean Creole (Greene 1999) and Negerhollands (Stolz 1986), while seven vowels are found (among other languages) in HC, Saramaccan (Bakker, Smith and Veenstra 1995), Nigerian Pidgin (Faraclas 1996) and Krio (Fyle and Jones 1980). 8 Holm (1988) comes to similar conclusions, also finding exclusively 5- and 7-vowel systems in the (basilectal) Caribbean Creoles, while Alleyne (1980) sees a five-vowel system as basic for most Caribbean Creoles. The point is that no basilectal Creole seems to reach the complexity of either the supers t a t e or the more complex substrate inventories. English, French and Dutch all have more than seven vowels, and there are a number of substrate languages which also have a more complex system. Many West African languages have nine or ten vowels (Clements 2000), for example Akan or Ijo, which have nine vowels (Ladefoged and Maddieson 1996); yet, no creole shows up with an Akan-type vowel inventory, although the impact of Akan (and its dialects, such as Twi or Fante) on the formation of Creoles and the relatively high number of Akan speakers in slave populations are uncontroversial.

6

7

8

Note that this prediction only holds for cases of substrate leveling. Marked superstrate structure can be acquired as a result of superstrate targeting and is therefore outside the realm of substrate leveling as described in this section (see also section 5.4.). Spanish Creoles are not included because Spanish itself has only five vowels. Hence, a five-vowel system in a Spanish lexifier Creole can as well be superstrate-induced and thus does not lend any evidence to the claim made here. Some Creoles are harder to classify synchronically. Jamaican Creole, for example, could be described as a five-vowel system with quantity (length), where diphthongs and the laxing of short vowels are phonetically derived (cp. Cassidy and LePage 1980 for a discussion). Alternatively, Jamaican could be seen as a more complex system, which, however, is not based on any superstrate or substrate model and thus constitutes an independent development, not a counterexample to my claim.

Markedness, faithfulness and creolization

17

5 . 4 . A sociolinguistic caveat One caveat is in order now, however. The model proposed in this section is a theoretical model o f leveling as a key process in creolization. It does not account for sociolinguistic factors in creolization, and only imperfectly can the complexity o f the sociolinguistic setting be modeled. Hence, it is impossible to determine the outcome o f an ongoing process o f creolization with this model. Accessibility o f the superstate, the homogeneity or heterogeneity o f the substrate, the ratio o f substrate to s u p e r s t a t e speakers and the prestige o f ethnic groups participating in the formation o f the Creole cannot be integrated into this model. This model o f leveling is thus predicted to be seen most clearly only in radical creolization with minimal s u p e r s t a t e influence in all stages o f development, that is minimal influence during the formative period and minimal influence in the subsequent development, for example as a result o f superstate withdrawal, as found in the Surinamese Creoles. T o account for s u p e r s t a t e influence on the Creole (as, for example, in the occurrence o f closed syllables in many Creoles), the model o f leveling has to be combined with the model o f shift and S L A discussed in 4 . 1 . Then, the relative importance o f leveling (as demotion o f faithfulness among substrates) and s u p e r s t a t e learning (as regular constraint demotion) have to be taken into account and weighted individually in each scenario, but the relative weight o f each can often only be established in a post-hoc analysis, looking at the outcome. What has been perceived as a ' c l i n e ' o f creolization (Schneider 1990), can then be conceptualized as the relative impact o f faithfulness demotion in leveling and constraint demotion in supers t a t e acquisition, the two models described in this paper, in Creole formation. Similarly, decreolization can be described straightforwardly using the shift model introduced in 4 . 1 . which now operates on a grammar previously acquired through leveling. This array o f possible outcomes o f creolization demonstrates why the model o f creolization in O T proposed here is to be used with caution. T w o different processes, which cannot simply be unified within one model, mix in an intricate fashion depending on external factors, and while it should be possible to identify these processes in an analysis against the background o f demographic and other sociolinguistic information available, it is very hard to predict when which process will apply. T h e central model o f leveling in particular is very likely to be obscured by superstate influence or independent developments and is thus to be studied best in radical Creoles.

6 . T h e e m e r g e n c e o f t h e u n m a r k e d in s h i f t i n g d e m o g r a p h i c s

A final example will show how the proposed model works as leveling to the unmarked modulo the strength o f individual substrate languages, this time in a diachronic setting after creolization proper, in the further development o f a Creole, due to continued substrate influence. A demographic shift with respect to speakers o f different substrate languages, too, can cause a change in the grammar, because an unmarked option b e c o m e s available. I will show now how this can c o m e about with the example o f the development o f the default epenthetic vowel in Sranan (Plag and Uffmann 2 0 0 0 , Lappe and Plag this volume).

Christian Uffmann

18

Diachronically, the default epenthetic vowel of Sranan changed from /e/ to Iii, as recorded in the earliest sources. In Herlein (1718), /e/ occurs almost exclusively; in van Dyk (1765), the system is mixed, both /i/ and Id occur, while Id is still the predominantly chosen epenthetic vowel. By the end of the 18th century, however, /i/ had largely replaced Id as the epenthetic vowel, as attested in Nepveu (1770) and Schumann (1783). In Modern Sranan, III still is the default epenthetic vowel. The occurrence of Id as a default vowel is somewhat unusual cross-linguistically. In general, high vowels are considered less marked than mid vowels. This universal preference is expressed as a universal ranking of two markedness constraints in Beckman (1998), repeated here in (21): (21) *Mid » *High

mid vowels are more marked than high vowels

This ranking states that universally, the constraint against mid vowels, here called *Mid, outranks the constraint against high vowels, *High. Default epenthetic vowels are generally assumed to be the least marked vowels of the vowel inventory of a language, and in many languages, we find /i/ as the default epenthetic vowel, as in Modern Sranan. So, how could Id ever emerge as a default vowel if it is so marked? The explanation probably rests in differences that can be observed between substrate languages. In two substrate languages, i.e. in Twi and in certain Gbe lects, we find Id as the default vowel in epenthesis (Plag and Uffmann 2000, and references therein). I assume here that there is a language-specific ranking in these two cases which contradicts the more general ranking in (20). This Twi- and Gbe-specific ranking is tentatively formalized in (22). 9 (22) *i » *e

Twi- and Gbe-specific ranking

This specific ranking states that Iii is more marked than Id. Retention of this substrate ranking in the emerging Creole can thus account for the fact that Early Sranan, too, has Id as a default vowel. But why did it change in the course of the 18th century? Interestingly, the demographic factor gives us an explanation of this. Between 1700 and 1750, the vast majority of slave imports came from the Slave and Gold Coasts (Arends 1995:243), meaning that most new arrivals probably spoke Gbe or Twi; languages with default Id constituted a majority in the slave community. Therefore, the marked option could be retained. In the second half of the 18th century, the situation changed dramatically when the Slave Coast ceased to supply slaves to Suriname, and the Gold Coast became a minor supplier. Instead, the Loango region and the Windward coast took over as the major slaving areas. This shift in demographics can explain why the default epenthetic vowel in Sranan changed from Id to HI. Speakers transmitting the marked option became fewer in number; consequently, there was opportunity for the unmarked to emerge, supported by substrate languages that also favored the unmarked option. Kikongo, for example, the chief language spoken by slaves from Loango, has I'll as the default epenthetic vowel (Plag and Uffmann 2000). The marked ranking in (22) could be replaced by 9

This ranking should be seen as a simplified ranking to capture the Twi facts. It is not meant to be of any theoretical import. The true underlying constraint interaction that leads to the preference of /e/ over Iii is presumably more complex.

Markedness, faithfulness and

creolization

19

the unmarked ranking in (21) in Sranan because (a) there was a number of substrate speakers having the unmarked ranking in their native grammars and (b) the marked grammar, once spoken by a majority of slaves, was no longer reinforced, due to the diminished influence of Gbe and Twi.

7. Residual issues

A few final thoughts should be devoted to the question of how the theory presented here fits in with a broader context of Creole phonology. The analysis has been limited (mostly) to vowel inventories, so I would like to open up the perspective now and mention a few residual issues, which, for reasons of space, cannot be discussed in detail here but which still should be addressed. A possible objection against the theory presented here comes from the often made observation that Creoles have a fairly unremarkable phonology, meaning that there do not seem to be many phonological alternations or processes. A lack of alternations, however, is often explained as high-ranked faithfulness constraints, which militate against phonological changes from input to output, viz. alternations. So, does this superficial input-to-output faithfulness not contradict the claim that Creoles have high-ranked markedness constraints? There are three arguments against this objection. First, the view that Creoles do not have much of a phonology worth mentioning is a relic of classical derivational theory, where the active phonology accounts for processes only (in addition to a more passive component which defines the segment inventory and basic phonotactics). However, in OT, segment inventories, too, are defined by constraints. Lexical material often has to undergo major restructuring processes in order to be well-formed in a creole with a more restricted inventory. By accounting for both, alternations and inventories, with the same mechanism (a set of ranked constraints), there are many more unfaithful mappings (as in the nativization of lexical material) than if we just took processes into account, as a traditional, rule-based theory would. Second, phonological processes are frequently morphologically triggered. Alternations often appear at morpheme boundaries. The largely analytic morphology of Creoles, however, removed possible triggers for phonological alternations. Due to the limitations made by the morphology (or lack thereof) in Creoles, possible alternations are thus restricted mostly to phonologically triggered alternations. Third, the alleged lack of alternations becomes a disputable claim against the abovementioned morphological caveat. Alternations of the unmarked kind are not too infrequent among Creoles. Assimilatory palatalization, for example, i.e. palatalization in the context of high front vowels, is frequently found in Creoles (e.g. in HC, Gullah and the Surinamese Creoles, Tinelli 1981) as well as in the substrate languages (Holm 1988). Regressive nasalization, too, is a common feature of Creole phonologies. Alternations that are crosslinguistically frequent (and hence probably unmarked) are thus also common among Creoles. Marked processes, on the other hand, are much less often found in Creoles, as predicted by my proposal. Processes of vowel harmony, for example, which are found in many West

20

Christian

Uffmann

African languages, seem to have been lost almost entirely. 10 Other marked processes, such as dissimilatory palatalization, i.e. palatalization in the context of a low vowel, found in Jamaican Creole, Belizean Creole or Krio, seem to be superstrate-induced and thus do not constitute a counterexample to my theory (Cassidy and LePage 1980). With respect to other processes, it is not quite clear as how marked they should be considered, with research into issues of markedness still going on. The processes underlying determiner allomorphy in HC (cf. Nikiema 1999), for example, could be seen as generating unmarked structure, despite the resulting surface complexity." More research on this topic seems necessary, however, to conclusively answer this question.

8. C o n c l u s i o n

This paper has demonstrated that it is possible to have the cake and eat it, too. The framework of OT can open a new perspective on Creole studies in showing that universalist and substratist theories are not mutually incompatible. I have shown here how the unmarked can emerge in an essentially substratist model of creolization. OT allows us to design a formal model of substrate leveling where unmarked substrate structures are selectively retained. The lack of marked structure in Creoles thus follows directly from the substrate. For any scientific theory it is important to be falsifiable, and one advantage of the theory here is that it is falsifiable. Substratists could falsify this theory by showing that it is not unmarked structure which is retained. A clear counterexample would be one where substrate structures are retained which are (a) universally marked and (b) not shared by a significant proportion (say, a relative majority) of all substrate speakers. Such a case would indeed argue against the view advocated here that unmarked options are preferred in substrate leveling. For universalists, on the other hand, it would be important to show that unmarked structures (outside those simplifications that are a possible result of pidginization and untutored SLA) are found in Creoles which are not found in the substrate, or that unmarkedness in the creole goes beyond the least marked substrate grammar. In sum, the integration of OT into theories of creolization can open a new line of research that goes beyond the established dichotomy of substrates and universals. Further research into different areas of grammar is necessary, however, to demonstrate the validity of this approach. The investigation into Creole phonology provided here suggests that the approach may turn out fruitful for our understanding of the processes behind creolization, though.

10

1

There seems to be a small residue of ATR-harmony in HC (Tinelli 1981) and basilectal Nigerian Pidgin (Faraclas 1996), but generally, vowel harmony is absent from Creoles. The type of harmony found in epenthesis, e.g. in Sranan, is excluded, first, because it does not account for alternations, but also because it is fundamentally different from the known substrate patterns: Substrate languages have height or ATR-harmonies, while the quality of the epenthetic vowel in Sranan is determined by back harmony, which is unattested in the substrate (Uffmann 2001).

' Note that this type of allomorphy does not seem to be restricted to HC but appears elsewhere as well. Gerlach (2001) provides an analysis of the allomorphy of personal pronoun clitics in Portuguese, which looks conspicuously like the HC determiner allomorphy.

Markedness,

faithfulness

and

creolization

21

The questions raised again by Parkvall (2001) can be addressed, and a few answers to his questions have been proposed in this paper already.

References

Alber, Birgit and Ingo Plag (2001): Epenthesis, deletion and the emergence of the optimal syllable in Creole. - In: Lingua 111, 811-840. Alleyne, Mervin (1980): Comparative Afro-American. Ann Arbor: Karoma. Arends, Jacques (1995): Demographic factors in the formation of Sranan. - In: Jacques Arends (ed.): The Early Stages of Creolization, 233-285. Amsterdam: Benjamins. Bakker, Peter, Norval Smith and Tonjes Veenstra (1995): Saramaccan. - In: Jacques Arends, Peter Muysken and N. Smith (eds.): Pidgins and Creoles: An Introduction, 165-178. Amsterdam: Benjamins. Beckman, Jill (1998): Positional Faithfulness. PhD diss. University of Massachusetts, Amherst. Bickerton, Derek (1981): Roots of Language. Ann Arbor: Karoma. Broselow, Ellen, Su-I Chen and Chilin Wang (1998): The emergence of the unmarked in second language phonology. - In: Studies in Second Language Acquisition 20, 261-280. Capo, Hounkpati (1991): A Comparative Phonology of Gbe. Dordrecht: Foris and Gwome: Lado Gbe. Cassidy, Frederick G. and Robert LePage (1980): Dictionary of Jamaican English. 2nd ed. Cambridge University Press. Chomsky, N o a m and Morris Halle (1968): The Sound Pattern of English. N e w York: Harper and Row. Clements, George N. (2000): Phonology. - In: Bernd Heine and Derek Nurse (eds.): African Languages: An Introduction, 123-160. Cambridge University Press. Eckman, Fred (1977): Markedness and the contrastive analysis hypothesis. - In: Language Learning 31, 195-216. Faraclas, Nicholas (1996): Nigerian Pidgin. London: Routledge. Fyle, Clifford and Eldred Jones (1980): A Krio-English Dictionary. Oxford University Press. Gerlach, Birgit (2001): Neutralisierung im Portugiesischen. Talk presented at the DüKöMarSie workshop, University of Cologne, 30/11/2001. Gnanadesikan, Amalia (1995): Markedness and faithfulness constraints in child phonology. Ms. Rutgers University. Greenberg, Joseph (1966): Language Universals. The Hague: Mouton. Greene, Laurie A. (1999): A Grammar ofBelizean Creole. New York: Lang. Grijzenhout, Janet and Bertus van Rooy (2001): Second language phonology: Acquisition through gradual constraint demotion. SFB 282 Working Papers 119, University of Düsseldorf. Hale, Mark and Charles Reiss (1988): Formal and empirical arguments concerning phonological acquisition. - In: Linguistic Inquiry 29, 656-683. Hayes, Bruce (1999): Phonological acquisition in Optimality Theory: The early stages. Ms. UCLA. Herlein, J.D. (1718): Beschrijvinge van de volksplantige Zuriname. Leeuwarden: Injema. Holm, John (1988): Pidgins and Creoles. 2 volumes. Cambridge University Press. Huttar, George and Mary Huttar (1994): Ndyuka. London: Routledge. Jakobson, Roman (1969/1941): Kindersprache, Aphasie und allgemeine Lautgesetze. Frankfurt/M: Suhrkamp. Kager, Rene (1999): Optimality Theory. Cambridge University Press.

22

Christian

Uffmann

Ladefoged, Peter and Ian Maddieson (1996): The Sounds of the World's Languages. Oxford: Blackwell. Lappe, Sabine and Ingo Plag (this volume): Rules vs. analogy: Modeling variation in word-final epenthesis in Sranan, 71-90. Lass, Roger (1984): Phonology. Cambridge University Press. Lefebvre, Claire (1986): Relexification in Creole genesis revisited. - In: Peter Muysken and Norval Smith (eds.): Substrata versus Universals in Creole Genesis, 279-301. Amsterdam: Benjamins. - (1998): Creole Genesis and the Acquisition of Grammar. Cambridge University Press. McCarthy, John and Alan Prince (1993): Prosodic Morphology I: Constraint interaction and satisfaction. Ms. University of Massachusetts and Rutgers University. - (1994): The emergence of the unmarked: Optimality in prosodic morphology. - In: ΝELS 24, 333379. - (1995): Faithfulness and reduplicative identity. - In: Jill Beckman, Laura Walsh Dickey and Suzanne Urbanczyk (eds.): Papers in Optimality Theory (University of Massachusetts Occasional Papers 18), 249-384. Amherst: GLSA. Migge, Bettina (1998): Substrate influence in Creole formation: The origin of give-type serial verb constructions in the Surinamese Plantation Creole. - In: JPCL 13, 215-265. Muysken, Peter and Norval Smith (eds.) (1986): Substrata versus Universals in Creole Genesis. Amsterdam: Benjamins. Nepveu, Jean (1770): Annotatien op de Surinaamsche Beschrijvinge van A° 1718, 274-288. Ms. Municipal Archives, Amsterdam. Nikiema, Emmanuel (1999): De la variation du determinant /la/ dans les crioles hai'tien et st-lucien. In: Lingua 102, 1-25. Parkvall, Mikael (2001): Creolistics and the quest for creoleness: a reply to Claire Lefebvre. - In: JPCL 16, 147-151. Plag, Ingo and Christian Uffmann (2000): Phonological restructuring in Creole: The development of paragoge in Sranan. - In: Ingrid Neumann-Holzschuh and Edgar Schneider (eds.): Degrees of Restructuring in Creole Languages, 310-336. Amsterdam: Benjamins. Prince, Alan and Paul Smolensky (1993): Optimality Theory: constraint interaction in generative grammar. Ms. Rutgers University and University of Colorado (to appear at MIT Press). Schneider, Edgar W. (1990): The cline of creoleness in English-oriented Creoles and semi-creoles of the Caribbean. - In: English World-Wide 11, 79-113. Schumann, Christian Ludwig (1783): Neger-Englisches Wörterbuch. Editio tertia. Ms. Paramaribo / Moravian Archives Utrecht and Paramaribo. Smith, Norval (1977): Vowel epithesis in the Surinam Creoles. - In: Amsterdam Creole Studies 1, 154. Stolz, Thomas (1986): Gibt es das kreolische Sprachwandelmodell? Frankfurt/M: Lang. Tesar, Bruce, and Paul Smolensky (1996): Learnability in Optimality Theory. Technical Report JHUCogSci-96-2, Johns Hopkins University. - (2000): Learnability in Optimality Theory. Cambridge, MA: MIT Press. Thomason, Sarah G. and Terrence Kaufman (1988): Language Contact, Creolization, and Genetic Linguistics. Berkeley, CA: University of California Press. Tinelli, Henri (1981): Creole Phonology. The Hague: Mouton. Uffmann, Christian (2001): Patterns of vowel epenthesis (not only) in Shona loanwords. - In: C. F6ry, A. Dubach Green and R. van de Vijver (eds.): Proceedings of HILP 5, 193-211. University of Potsdam.

Markedness, faithfulness and creolization

23

van Dyk, Pieter (1765?): Nieuwe en nooit bevoorens geziene onderwyzinge in het Basiert Engels, of Neeger Engels, zoo als het in de Hollandsze colonien gebruikt word. Amsterdam: van Egmont. Yavas, Mehmet (ed.) (1994): First and Second Language Phonology. San Diego: Singular.

Christian Uffmann, Institut für Anglistik und Amerikanistik, Uni Marburg, Wilhelm-Röpke-Str. 6D, 35032 Marburg, Email: [email protected]

Albert Valdman and Iskra Iskrova A new look at nasalization in Haitian Creole

1. Introduction 1

The status of nasal vowels in Haitian Creole (HC) and in French lexifier Creoles in general has been the object of a considerable literature. However, its analysis has remained impervious to various types of phonological approaches. The nasal vowel system of HC has posed two major problems: (1) the determination of the total inventory, in particular the status of the high nasal vowels [~i] and [ü]; (2) the analysis of nasal vowels occurring in the context of adjacent or nearby other nasal segments. There is a further issue that has proven even more untractable: the analysis of nasalization phenomena occurring in the post-posed clitics, the third person singular pronoun li and the definite determiner la. Even more complex but relatively unexplored by phonologists are nasalization phenomena associated with the possessive pronoun in northern varieties of the language, for example [figir a mwe] 'alternating with [figiYäm] 'my face', [do a mwe] alternating with [dwäm] 'my back' (Valdman 1978). This article will focus on two issues of the complex situation of nasal vowels in HC: (1) whether nasal vowels need to be posited at the underlying level or whether they are to be derived by phonological rules; (2) various phenomena of nasal diffusion within lexemes and across morphological boundaries. In Section 1, we present the inventory of nasal vowels in HC and describe their distribution. In Section 2, we review previous analyses, all of which either fail to account for empirical data or violate various theoretical principles of current autosegmental phonological theory. In Section 3, we present our own autosegmental analysis, which involves both positing underlying nasal vowels and intramorphemic nasal assimilation rules, some constrained by lexical marking. In Section 4, we deal specifically with nasal assimilation across morpheme boundaries. For the sake of economy, we limit that latter section to phenomena occurring word internally, leaving for subsequent research the more thorny issue of progressive assimilation in postposed clitics, notably the third person li and the definite determiner la, as well as various sandhi phenomena characteristic of northern dialects of HC (Etienne 1974, Valdman 1978). In Section 5, we offer a summarizing conclusion.

We owe a debt of gratitude to our colleague Stuart Davis who generously provided advice and counsel on theoretical issues; of course, only we as authors are responsible for the application of his suggestions to this specific analysis. We would like to thank our reviewers, Parth Bhatt and Emmanuel Nikiema, as well as the editor, for useful comments and suggestions. We also would like to thank Nicolas Andre, Lyonel Jean Desmarattes, and Jacques Pierre who, as native speakers of Haitian Creole, provided some critical examples.

26

Albert Valdman andIskra Iskrova

2. Inventory of nasal vowels in HC

The maximal inventory of nasal vowels of HC comprises five units; its major differences with that of Referential French (RF) is the presence of a pair of high vowels and the absence of the front rounded vowel, see Table (1). There are also significant differences in the phonetic characteristics of the various phonemes that will not be dealt with here. Linguists dealing with the varieties of French currently prefer the term Referential French (franqais de reference) instead of the more widely used label Standard French to underscore the fact that it does not wholly correspond to the speech of educated and cultivated Parisians but represents a more idealized norm (Poirier 2000). In any case, the differences in inventory are presented here merely to underscore differences between the two systems. In no way, do these imply any genetic relationship between them. In other words, we do not claim that HC is derived directly from RF. (1)

Nasal vowel inventories of HC and RF Haitian Creole

Referential French

"ι

ü

e

5

έ

ä

ce

δ ä

The absence of contrastive pairs have led some linguists to consider the high nasal vowels [1] and [0] as allophonic variants (Valdman 1978, Dejean 1980) whereas others arguing from parallel distribution between these vowels and the other three have accorded them phonemic status (Tinelli 1974, Cadely 1994). At this stage of our study we maintain that high vowels may undergo contextual nasalization but we do not have convincing evidence to claim that they are underlyingly nasal. But the major problem in the treatment of nasal vowels remains their presence in the environment of nasal consonants. Unlike in RF where these vowels are absent before wordfinal nasal consonants (thus [bö] 'good (masc.)' contrasts with [bon] 'good (fem.)' and sequences like *[bön] are not permitted), in HC they contrast with oral vowels in that context, see (2). Note that there are no four-way contrastive pairs involving the front unrounded vowel since [Ιεη] and [len] are free variants. Minimal pairs that we have been able to find, such as [νεη] 'our glass' versus [ven] 'vein' are indeed minimal but note that the former member is bimorphemic: [νε] 'glass' + [n] 'our'. Henceforth, examples that appear without slant lines (//) or square brackets ([ ]) should be assumed to be in surface phonological representation, unless otherwise indicated. (2)

Final Position a. mo 'word' pa 'step' 1ε 'hour'

b. nö pä le

'no' 'peacock' 'linseed'

Preceding a nasal consonant c. mon 'hill' d. mön 'world' pan 'breakdown' pan 'to hang' Ιεη 'wool' len 'wool'

Nasalization in Haitian Creole

27

In addition, nasal vowels occur in the immediate environment of nasal consonants in wordinternal syllables, see (3), and across morpheme boundaries in derivations, as in (4). The insertion of /n/ in / kapöne/ and / desine/ will be discussed in Section 4; the vowel change in the latter form falls outside of the scope of this article. (3)

lame 'sea'

(4)

bobin gagän kapö dese

remed 'medicine'

'spool' 'throat' 'coward' 'drawing'

bobin gagän kapö dese

+ + + +

e e e e

>

> > >

käkänay 'charring' bobine gagäne kapöne desine

'to 'to 'to 'to

roll up' hold by the throat' intimidate' draw' 2

The pervasive presence of nasal vowels in the environment of nasal consonants has led phonologists to posit general rules of regressive and progressive nasal assimilation (Hall, 1953, d'Ans 1968, Tinelli 1974). However, not all oral vowels are affected by these putative rules, as shown in the contrastive pairs in (5) and cases of optional nasalization (6). (5)

ne mäj

'nose' 'handle'

(6)

kotönad/kotonad

ne 'knot' m a j 'step in stairs' 'cotton fabric'

fänal/fanal

'lantern'

Besides the cases of optional nasalization, there are minimal or near minimal pairs that show a contrast in polysyllabic words, see (7). (7)

mäti chämot känot mönite

'lie' 'two-story house' 'small boat' 'computer screen'

mati chamo kano monite

'martyr' 'camel' 'shank (of animal)' 'teaching assistant'

3. Existing analyses for regressive assimilation

3.1. The traditional structural analysis Traditional structural descriptions of the nasal vowel system of HC (Tinelli 1974, Valdman 1978) recognize underlying nasal vowels and posit as well a regressive assimilation rule (8) operating word-internally that would account for the data in (9): (8)

2

[-cons]

[+nas] / _ nCVX

Note that in this form in addition to the nasalization in process there is a vowel quality change that will not be discussed in this paper.

28

(9)

Albert Valdman and Iskra Iskrova

[läme] 'sea' [länwit] 'night'

[mena3] 'boy/girl friend' [känänät] 'listless'

[mäniget] 'maneuvering'

A more powerful version of this rule extending the environment to word final syllables has been proposed for Dominican Creole by Taylor (1947) and for HC by d'Ans (1968). In order to prevent the application of the rule to lexemes such as [Jam] 'voodoo charm', resorting to etymological information, these authors invoke a latent post-vocalic /r/. The presence of that consonant makes Rule (10) no longer applicable, and unlike the contrasting form [Jam], the vowel is not nasalized. (10) [-cons] ->• [+nas] /

n#

The underlying r solution has been proposed again by Nikiema and Bhatt (this volume). Valdman (1978) provides several empirical and theoretical arguments against this proposal, the most important of which is that the associated rule would be required to delete the postvocalic r which, furthermore, never surfaces. Another serious problem with the underlying r hypothesis is that not all etymological r's block nasalization. For example, corresponding to RF germe are the free variants [jem] and bem]. According to our colleagues the latter variants could not surface since regressive nasal assimilation is blocked by the underlying segment. Another weakness of the proposal is that it implies that HC lexemes are direct reflexes of RF ones. As Chaudenson (1992) and Valdman (1992) point out the target accessible to the creators of HC was more likely a variety of non-standard dialects and sociolects of French (regional and vernacular varieties). In some of these, as reflected in present-day overseas varieties of French, post-vocalic r was considerably weakened, if not absent. Thus, the creators of HC had no surface indication of the presence of [r]. Furthermore, there exist forms, some cited in Cadely (to appear a), for which one cannot posit an underlying Irl and that do not undergo nasalization: e.g., [lapeson] 'person', [ban] 'servant'. According to Nikiema and Bhatt these should surface as [lapeson] and [bön], respectively. Finally, the underlying r hypothesis might be applicable to Northern HC in which that segment does surface following nouns ending in [ε] and [o] whose RF cognates contain a post-vocalic [R], for example: [se] 'sister' (soeur) alternating with [ser] but [seyäm] 'my sister' and [ko] {corps) alternating with [kor] but [koyäm] 'my body'. In Northern HC when r appears it is a weak glide in final position (Π) and fronted intervocallicaly ([γ]). Unlike the corresponding paratactic construction in what we may refer to as Standard HC (SHC), the Northern HC (NHC) possessive determiner construction contains the linking preposition a, compare: SHC [sem] to NHC [seyäm]. In view of the significant differences between the two varieties, we do not take into account data from NHC in our analysis.

3.2. The problem of variation The regressive assimilation rule (10) is also inadequate because it fails to take into account two empirical facts. As pointed out by Valdman (1978:65) and more recently by Cadely (to appear a), there is extensive variation in the application of the rule. Three types of cases must be recognized:

Nasalization

in Haitian

Creole

29

(11) a. Obligatory application of the rule: [ven] *[νεη] 'vein', [lame] 'wave' (versus [lame] 'old woman') b. Non application: [pan] 'car breakdown', [kanif] 'penknife' c. Free variation: [len] ~ [len] 'wool', [fänal] ~ [fanal] 'lantern' As Cadely (to appear a) underscores, for most lexemes containing the structure XCVNVX the basic lexicographic resources for HC (Valdman et al. 1981, Valdman et al. 1996, Freeman and Laguerre 1998) list both nasalized and non nasalized variants. A frequency list (Vernet and Freeman 1988) shows that absence of nasalization dominates: 876 instances versus 498 showing optional nasalization.

3.3. Previous autosegmental analyses Several efforts have been made within an autosegmental framework to account for nasality in HC (Cadely 1994, to appear-a, to appear-b) or Saint Lucian Creole (SLC) (Bhatt and Nikiema 2000). Cadely (to appear-b) has proposed that nasality is derived in HC from a non-associated nasal consonant following the vowel that, by the nasality parameter, is incorporated into the nucleus. Bhatt and Nikiema come up with a similar proposal featuring an adjacent [n]. According to them, an adjacent consonant can only be linked to the rime, and in this particular case to the nucleus. Cadely's analysis might account for vowel nasalization in French since that language lacks the sequences nasal vowel (V~) plus morpheme final nasal consonants (N) displayed in (2d). However, it encounters several problems in accounting for the more complex HC data. In his proposal the underlying representation of [Jam] and [kanif] are, respectively, /Ja(n)m/ 3 and /kanif/. The former (12a) does not contain any skeletal slot for the /n/, whereas the latter (12b) does. When there is no slot (12a), the nasal attaches to the preceding nucleus by the 'nasality parameter' (Prunet, 1992). 0 I 1 χ

Ν I 1 X

C 1 1 X

b. 0 I 1 X

1 J

l\ 1 a η m

k

|

[Jam]

Ν I 1 X

0 1 1 X

Ν I 1 X

1 a

1 η

1 i

C 1I X I f

[kanif]

In the case of the doublets ([fänal] ~ [fanal]) Cadely proposes two possible configurations: with or without linking (13):

3

The parentheses indicate the presence of an /n/ in the underlying representation that is not anchored in the skeletal structure of the word.

Albert Valdman and Iskra Iskrova

30 (13) a. Ο

Ν

Ο

Ν

C

b. Ο

Ν

Ο

Ν

C

χ

χ

χ

χ

χ

χ

χ

χ

χ

χ

f

a η η

a

f

ι\ I a η η

a

[fänal]

[fanal]

But (13) does not account completely for the two variants because Cadely does not specify under what conditions the floating /(n)/ is or is not to be associated with the nucleus. What his analysis lacks is some rule or principle governing the modalities of the association and a motivation for the variation. From a theoretical perspective the representation of [fänal] as /fa(n)nal/ incurs the violation of two major principles: the Obligatory Contour Principle and the syllabic structure requirements. According to the former (Goldsmith 1976, McCarthy 1988) adjacent identical elements are prohibited. Representations containing two nasals segments in adjacent positions, as is the case for (/fa(n)nal/ and /va(n)n/), are ruled out by the theory. Besides that, a representation in which the nucleus is biconstituent (14a) is better suited for languages like RF that display alternations between a final nasal vowel in non-derived environment and an oral vowel followed by Ν in a derived environment. Such "unpacking" of the feature nasal, a term borrowed from Paradis and Prunet (2000), rarely occurs in HC, and if so it occurs in free variation with a form that preserves the feature nasal on the stem final vowel, as illustrated by derivatives like /padö/ 'pardon' > /padone/padöne/ 'to pardon' and /butö/ 'button' > /butöne/ 'to button'. For this reason we think that HC does not show evidence for biconstituent nuclei. In our view monoconstituent nasal vowels (14b) better reflect the structure of SHC that has not developed unpacking of the nasal feature in derived environments. Our proposal has the added advantage of not imposing on HC the structural features of the lexifier and accounts for the derivational processes within a language specific frame. (14) a. Ο Ν C X X X ν

I\ a n

b. Ο

Ν X

ν

X

C X

a

In a more recent analysis Cadely (to appear-a) assigns an underlying status to HC nasal vowels, but this treatment differs from the previous one only by a terminological variant: naming the floating element a feature rather than a segment. It is not clear what this author means by "nasality is associated with the vowel at the lexical level"; 4 in autosegmental phonology a floating feature is by definition mobile and can associate to various positions within the word, respecting linking and delinking rules. This is not the case in HC, where nasality is associated to one position in particular. The data from the language does not show evidence for the presence of a floating feature.

4

Our translation o f the author's statement: "la nasalite est associee ä la voyelle au niveau lexical"

Nasalization in Haitian Creole

31

Concerning the Bhatt and Nikiema proposal, it does not hold for HC, because unlike in SLC where such pairs as [läm] and [lam] 'blade' are in free variation, there are in HC contrasts in CVN surface syllables: [Jam] 'room' versus [Jam] 'charm'. There are also theoretical problems on which we will comment briefly. These authors postulate two root nodes for the adjacent consonant whereas, as will be shown below, we hypothesize that the two nodes are within the vowel. They derive the nasality feature by default from the VS feature in the nucleus. It is indeed the case that the status of nasality is the subject of controversy among autosegmental phonologists but most consider nasality nodes to differ from other vocalic features. Since we associate nasality with vowels, and not with adjacent, floating, or unassociated consonants, our analysis doesn't run into this sort of problem. Another theoretical problem stems from these authors invoking Government Theory wherein syllables have no branching rimes and codas are replaced by onsets preceding empty nuclei. In their derivation of [3am] and [vjän] (2000:39), they seem to suggest, erroneously in our opinion, that the syllable structure is part of the underlying segmental representation: the latter is composed exclusively of phonic material. Finally, the model Bhatt and Nikiema operate with is extremely complex: onsets that are difficult to motivate are associated with coda positions while at the same time diffusing nasality to the vowel. Then, when the vowel has become nasalized, it spreads its nasality onto the coda!

3.4. Toward an alternative analysis: biconstituency of underlying nasal vowels We propose that vocalic nasality is both underlying and derived in HC. Although autosegmental analyses have shown a marked preference for derived nasality, we claim that the complex HC data we present here provide sufficient evidence for the need to posit underlying V~'s. We follow Paradis and Prunet (2000) who argue that V~'s are underlying in French, as well as in Hindi and Portuguese. They claim that, although V~'s do not have longer durations than oral vowels, they are biconstituent, but at the root level rather than the segmental level. As shown in (15) these vowels consist of one segment with two root nodes: one of them carries the features of the vowel, the second one bears the nasality feature. (15) Segmental tier Root level Nodes 5

(set of features)

V I χ l \ R11I Rn2 • I [-cons]

· I [+nasal]

To support their claim Paradis and Prunet adduce empirical evidence from borrowing. They examined the incorporation of lexical items from donor languages exhibiting different types of nasal vowels into the phonological system of a single recipient language. They show that 5

Rn stands for root node.

Albert Valdman and Iskra Iskrova

32

different patteras appear in the recipient language depending upon the nature of the nasal vowels in the donor language. If the nasal vowels are underlying in the donor language, and if the borrower language does not possess underlying nasal vowels, then they are unpacked in the borrower language, as is the case of the reproduction of French lexemes containing nasal vowels by speakers of Canadian English. If, on the other hand, they are not underlying, i.e., they are the product of nasal assimilation or spreading, then they appear in the recipient language as oral vowels, i.e., the set of features specifying the vowel quality without the nasality feature, as is the case in the reproduction of Malay lexemes by these same Anglophone speakers. These data showing different outcomes according to the underlying value of the input suggest that there is a fundamental difference between underlying and assimilatory vocalic nasality (16).

(16)

English borrowing Malay (progressive nasalization) [rami] 'type of fiber' [näpu] 'musk deer'

[iami] / *[janfin], *[ramin] [nü:pu] / *[nö:pu], *[nD:npu]

French (underlying nasality) [amädin] 'amandine' [ämas] 'in a group'

[aemandin] [anmass]

4. P r o p o s e d a n a l y s i s

There are three types of data involving V~'s in HC that any phonological analysis needs to account for. Here, we present the data that will be discussed in the following four subsections, one for each of the three cases the first is divided into two subparts. (17) a. Obligatory nasalization: i. [Jam] 6 [*Jam] 'room' ii.

fcäm]

[*3am]

'leg'

iii. [lame] 'sea' b. Obligatory non nasalization i. Final position: [pan] [*pän] 'breakdown' ii. Word-internal position: [kanif] [*känif] 'penknife' c. Free variation: [Ιεη] ~ [len] 'wool', [fanal] ~ [fänal] 'lantern' Accordingly, we divide this section in three parts corresponding to each of these three types of data but, as will become clear below, we need to subdivide the first type into two subcategories.

6

Note that, unlike the corresponding lexeme in French, in HC this form never appears with the voiced stop /b/.

Nasalization in Haitian Creole

33

4.1. Underlying nasal vowels It appears superficially that the examples in (17a.i) and (17a.ii) should be derived from the same underlying structures and the operation of the same principles, but in fact they differ fundamentally. We deal first with cases of obligatory surface nasalization like [Jam] [* Jam] 'room'. These are handled by positing underlying V~'s with root nodes characterized in (15) above; these V~'s occur in all environments, including before N, and contrast with corresponding oral vowels (18). (18) a. b. c. d.

[pä] [pät] [mäti] [Jam]

'peacock' 'slope' 'lie' 'room'

versus versus versus versus

[pa] [pat] [mati] [Jam]

'step' 'paste' 'martyr' 'charm'

4.2. Derived nasal consonants Bhatt and Nikiema (2000) describe in SLC a nasalization phenomenon involving the spread of nasality from left to right and from a V to a following oral consonant. This phenomenon, which affects underlying voiced stops forming the coda in the final syllable of root morphemes, also occurs in HC (19). (19) /3äb/ /väd/ /lag/

(a)

3äb + e väd + ε ζ lag + a j

'to step over' 'saleswoman' 'parlance, expression'

(b)

3am 'leg' vän 'to sell' lag 'tongue'

In the form containing the derivational suffix /-aj/, (20a), the voiced stop occurs in the onset of the syllable and it retains the features of the underlying form. However, as shown in (20b), in the root form the consonant is associated with the coda and attracts the nasality of the vowel. The nasal vowel spreads its nasality over the following segment that shares with it a position in the rime. (20) a. Ο

/läg+aj/ > [lägaj] R

Ο

I Ν

ä

R Ν

C

χ

χ

χ

g

a

J

l \ R n l Rn2 [+nas] [•ägaj]

34

Albert Valdman andIskra Iskrova b.

/lag/ > [läi]] Ο

R Ν

χ

χ

I

I

1

ä

C

71

Rnl Rh2 [+nas] im Thus, although the V~'s of [Jäm] *[fam] 'room' and [3am] *[3am] 'leg' appear identical at the surface level and both are underlying, the underlying structures from which they are derived differ: CVN for the former versus CVC, where the coda has the feature specification [+stop, +voice] for the latter.7 This type of alternation offers strong evidence for the existence of underlying nasal vowels, i.e., [+nas] can only be part of the feature specification of the vowel. In the root morpheme there isn't any floating [+nas] feature because in polysyllabic words the nasal feature does not spread leftward beyond the vowel, that is, /desäd/ yields only [desän] and *[desän] or *[nesä] are excluded. Clearly, in the environment of underlying nasal vowels a progressive nasalization rule operates that is restricted within the rime and applies only to voiced obstruent segments. The coda consonant keeps its specification for place of articulation. In the examples provided in (19) RF cognates show the voiced obstruent: /3am/ jambe, /vän/ vendre, /läq/ langue. This fact might suggest that all cases of V~N should be analyzed as sequences of V ~ plus voiced obstruent, and one might question our analysis of /Jam/ as V~N. However, we deem it legitimate to posit an underlying voiced obstruent only when we have evidence in the form of derived forms containing that type of segment. There are in HC numerous cases of forms where such evidence is absent, for example: /desäm/ 'December' dicembre, /fann/ 'to split' fendre, /vyän/ 'meat' viande, /pön/ 'to lay (eggs)' pondre. Cases of words ending in surface final V~N whose French cognates do not show the voiced obstruent that we analyze as having undergone obligatory nasalization (17a.i.) also abound: /madäm/ 'wife, woman' madame, /nam/ 'soul' äme, /nöm/ 'man' homme, /kolön/ 'column' colonne.

7

T h i s phenomenon occurs widely in non-standard varieties o f French: regional varieties, including those in the Americas (Louisiana, Saint-Bart, etc.) and so-called Popular French. In the latter variety, w h i c h in fact characterizes everyday speech, a pronunciation such as [3äm] for jambe is nearly categorical.

Nasalization in Haitian Creole

35

4.3. Obligatory non-nasalization One major difference between HC and SLC is that in the latter Creole surface sequences are absent in which an oral vowel occurs obligatorily before a N. All such cases are analyzed as containing an underlying r that blocks the leftward diffusion of nasality: [lam] 'tear' (RF larme) is inderlyingly /larm/. Thus, matching HC [pan] 'breakdown' SLC shows free variation: [pan] ~ [pan]. The occurrence of lexemes containing obligatory V[-nasal] Ν sequences, in both morpheme-final and non-final syllables, as in (17b), provides additional support to our hypothesis of underlying V~'s; it also makes the application of the Bhatt and Nikiema (2000) solution unworkable for HC.

4.4. Optional nasalisation We account for the different surface outputs of (17a-i) and (17c), namely obligatory versus optional nasalization, by deriving the former from underlying nasality and the latter from an optional regressive assimilation rule. Variable forms, such as those of (17c) are derived as follows (21): (21) a. Ο

Ν

Ο

Ν

I

I

C

b. Ο

f

[fanal]

Ν

a

Ο

Ν

C

n Μ [+nas] [fänal]

In the context of nasal consonants forms containing the diacritic feature may undergo optional nasalization spreading leftward from an adjacent N. Another alternative, that we reject, is optional denasalization. First, the regressive assimilation hypothesis complies with the markedness requirements of Universal Grammar: nasalization of vowels morpheme internally in the context of Ν is unmarked. Second, arguing from a language internal perspective, Cadely (to appear-a) presents evidence that suggests that nasalization in lexical items showing the structure X C V N V X constitutes a marked feature. As mentioned in 3.2 above, such forms where the V is not nasalized before Ν are twice as numerous as those showing nasalization. In addition, by comparing current pronunciations with those that can be deduced from the spelling conventions of texts of the colonial period, a stage of the language that Valdman (1992) labels Saint-Domingue Creole, he shows that the frequency of nasalized forms increased since the formative period of HC. He also refers to an apparent linguistic change in progress noted first by Valdman (1991) involving the extension in nonnasal environments of nasal allomorphs of the definite determiner among bilingual speakers. To conclude, lexemes like [Jam] (17a) and [kanif] (17b) have underlying nasal and oral vowels, respectively, whereas those exhibiting free variation like [fanal] ~ [fänal] (17c) are marked in the lexicon to undergo optional regressive assimilation.

Albert Valdman and Iskra Iskrova

36

To be sure, there are considerable differences among speakers about which lexical items undergo optional nasalization and intra-speaker variation as well. It is no doubt the case that bilingual speakers, more subject to the influence of RF, are likely to show a greater incidence of unmarked variants. Nonetheless, the situation in HC differs markedly from that reported by Bhatt and Nikiema (2000) for SLC. Those native speakers we have consulted and the available lexicographic resources clearly identify lexemes where nasalization is not permitted.

5. N a s a l i z a t i o n in d e r i v a t i o n a l m o r p h o l o g y

Another phenomenon involving the assimilation of nasalization in HC occurs in the derivation of denominal verbs by the adjunction of the verb forming suffix /-e/, the most productive affix in the language, see (4). Nasal assimilation spreading rightward from vowel to vowel across a morpheme boundary also occurs with the definite determiner (22): 8 (22) a. bäk lä

'the bank'

kabän nä 'the bed' pä ä 'the peacock'

b. bak la

'the tray'

chat la pa a

'the cat' 'the step'

The outward similarity between the spreading of nasalization in (4) and (22) would lead one to postulate nasal harmony in which nasality spreads rightward from the nucleus of the root form to the suffix. However, this hypothesis runs afoul of the data in (23) where there is no spreading. These militate against vowel harmony: (23) köt + e aväs + e fläm 9 + e 3äm + e

'account' 'advance' 'flame' 'leg'

> > > >

köte aväse fläbe 3äbe

*köte *aväse *fläbe *3äbe

'to 'to 'to 'to

count' advance' singe' cross over'

The absence of nasalization in the forms in (23) but its presence in those in (4) indicates that it is crucially dependent on the presence of a root-final N. However, as shown in (24) progressive assimilation of nasalization also occurs in lexemes ending with a V~; the data are based on Freeman and Laguerre (1998). 10 (24) a. Ν coda i. Oral variant only dwan 'customs' ii. Variation bläm 'blame'

dedwane 'to clear through customs' blame/bläme 'to blame'

Among younger bilingual speakers nasalization extends to all environments, see Valdman (1991). Recall (4.2) that the root form [fläm] is derived from /fläb/ and [3am] from /3äb/. As would be expected, there is considerable variation among native speakers.

Nasalization

in Haitian Creole

37

iii.Nasalized variant only plim 'feather' b. V~ final i. Variation padö 'pardon' ii. Nasalized variant only butö 'button'

deplime 'to pluck (feathers)'

padone/ padöne 'to pardon' butone 'to button'

c. V~ final with vowel change dese 'drawing' desine/e 'to draw' We provide a unified treatment for these derivatives based on positing a root morphemefinal underlying nasal vowel for the forms in (24b) and (24c). Because the affixation results in an empty onset slot, the nasality feature of the vowel is associated with the following syllable and surfaces as its onset. For the forms in (24a) the existing coda fills the empty onset slot and is therefore resyllabified. This analysis is motivated by the existence of forms like those in (25) not involving nasalization in which a glide is created to repair the formation of a hiatus caused by the affixation of the /-e/ suffix. (25) fizi änwi pläje

'gun' 'trouble' 'floor'

fizije 'to shoot' änwije 'to bother' pläjeje 'to lay a floor'

HC has available several hiatus filling strategies, one of which found in Northern HC involves the insertion of the sonant /r/, as was shown in Section 3.2.; we repeat one of the examples here: [se] 'sister' (soeur) alternating with [ser] but [seyäm] 'my sister'. But the most favored repair strategy takes the form of glide creation (26)." (26) French ocean theater poete aorte

HC osejä tejat powet awot

'ocean' 'theatre' 'poet' 'aorta'

The [+nas] feature of both the coda Ν and that resulting from hiatus repair strategy spreads to the adjacent vocalic suffix. There remains the occurrence of free variation in the suffix and the puzzling case of the denasalization of the root morpheme final vowel in [padone] (24b-i) and [desine] (24c). The progressive assimilation is a feature whose optional or obligatory nature must be marked on particular root morphemes. It does not apply to forms like /dwan/ (24a-i); it is obligatory in forms like /butö/ (24b-ii); and optional in all other cases, see the derivations of [bläme] and [butöne] in (27). As is the case for the optional assimilation rule (4.4.), empirical studies need to be undertaken for a description of the dialectal and/or sociolinguistic determinants of these lexical markings. " As a general rule, the inserted segments remain high on the sonority scale.

38

Albert Valdman and lskra Iskrova

(27) a. bläme O

N

C

+

O

N

A l l / b l ä m

O I e

A b

N

O

N

I I I l ä m e I PI [+nas]

b. butöne O

N

I b

u

O I t

N I ö

+

O

I

*

N

O I e

/

I

N I

b

I u

Γ' [+nas]

O t

N I ö

O

N

I

I PI

[+nas]

We explain the puzzling case of /padö/ by hypothesizing that there are in HC two ways to fill the empty onset. In most instances, the nasal node delinks from the vowel position in order to fill the onset. Alternatively, it only spreads its [+nas] feature onto the empty onset, but still remains associated with the root node of the vowel. The two alternative derivations are indicated in (28). The application of one or the other of these two treatments appears be optional. 12 (28) O I p

N

O I

a

N I

d

+ I

ö

O

N

I

I e

η Rnl [+nas]

R

n

2

R

n

Rnl [+nas]

Rn2 [+nas]

[padöne]

12

The vowel change in [desin-e] (25c) involves a vowel adjustment rule that we leave aside in this paper.

Nasalization

in Haitian

Creole

Ο

Ν

Ο

Ν

ρ

a

d

δ Rnl

39 +

Rn2 [+nas]

Ο

Ν

η

e

Rn Rn [+nas]

[padone] association line spreading line

As is the case for the regressive consonant to vowel assimilation treated in 4.4., the nasal assimilation rule applying to the suffix /-e/ needs to be constrained by lexical marking because it occurs only with that suffix. The nasality feature is not diffused across morpheme boundaries in the case of other vowel initial suffixes, for example: /tete/ 'foolishness' > [tetenad] 'nonsense'; /maköne/ 'to join together' > [makönaj] 'intertwining'. For Cadely (to appear-b) this process is totally optional, for he lists all verb forms with both the oral and nasal variant of the suffix, e.g. [padöne] ~ [padone]. However, Freeman and Laguerre (1998) make a four-way distinction in their listing of -e verbs, see (29). (29) a. Oral variant only: dedwane 'to clear through customs', emosjone 'to be moved' b. Variation with the oral variant listed first: bobine/bobine,

padöne/padöne,

bläme/bläme

'to b l a m e '

mänaj

c. Variation with the nasal variant listed first: vaksinef vaksine 'to vaccinate' d. Nasal variant only: boutöne'io button', chiföne'to wrinkle' Targete and Urciolo uniformly list these verbs as invariably ending with the nasal vowel. However, we consider Freeman and Laguerre as better documented, and we base our analysis on data (presented in 25 above) that underlie this classification.

6. C o n c l u s i o n

In this paper we have started from the traditional view that vocalic nasality is underlying in HC, that is, V~'s must be indicated in the lexicon. A marked feature of HC is that these vowels occur before Ν yielding surface contrasts like [Jam] 'charm' versus [Jam] 'room'. This proposal, which is consonant with theoretical assumptions of autosegmental phonology, also entails positing three different types of nasal assimilation processes: (1) vowel-toconsonant progressive assimilation operating rime internally that produced additional V~N sequences: [fläm] < /fläb/; (2) optional consonant-to-vowel regressive assimilation also morpheme internal operating within root morphemes: [len]~ [Ιεη] 'wool', [fänal] ~ [fanal] 'lantern'; (3) consonant-to-vowel progressive assimilation operating across morpheme

40

Albert

Valdman and Iskra

Iskrova

boundaries associated with onset creation to repair hiatus limited to the verb forming suffix /-e:/: [deplime] < /plim/ 'to pluck (feathers)', [blame / e] < /bläm/ 'to blame', [padone/padöne] < /padö/ 'to pardon'; this complex process is subject to extensive variation, most probably free (i.e. intra-speaker) variation. A superficial review of our approach might suggest that it violates the Obligatory Contour Principle because in a surface form like [fänal] it appears that the [+nasal] feature of the vowel is juxtaposed to a following Ν in a sequence V~N. However in our analysis, following Paradis and Prunet (2000), the nasal feature of the vowel and the contiguous Ν are situated on separate tiers: the two segments are adjacent in the linear sequencing of segments but not at the level of root nodes. Another distinguishing feature of our analysis is that the first and second processes are optional and autonomous and that they affect only previously marked lexical items. It appears that non-inherent nasality is an innovative feature of HC. But the empirical basis of currently available descriptive studies of the language is not robust enough for us to hazard such a generalization. It is no doubt the case that some sort of nasal assimilation, probably vowel harmony, and hiatus repair strategies also contribute to the puzzling allomorphy of the definite determiner, but the treatment of this complex problem is beyond the scope of this study.

References

Parth Bhatt and Nikiema, Emmanuel (2000): Le Statut de la nasalite en Creole de Sainte-Lucie. - In: Revue quebecoise

de linguistique

28, 23-45.

Cadely, Jean-Robert (to appear a): Le Statut de la nasalite en Creole hai'tien. - In: Lingua -

(to appear b): Nasality in Haitian Creole. (1994): Aspects

de la phonologie

du Creole ha'itien. Unpublished doctoral dissertation, U Q A M ,

Montreal. Chaudenson, Robert (1992): Des lies,

des hommes,

des langues:

langues

Creoles,

cultures

Creoles.

Paris: L'Harmattan. d'Ans, Andre-Marcel (1968): Le Creole frangais Dejean, Y v e s (1980): Comment

d'Haiti. La Haye: Mouton.

ecrire le Creole d'Haiti.

Etienne, Gerard (1974): Le Creole du Nord

d'Haiti:

Outremont (Quebec): Collectif Paroles. Etudes

des niveaux

de structure.

Unpublished

dissertation: University o f Strasbourg. Freeman, Bryant and Jowel Laguerre (1998): Haitian-English

dictionary.

Lawrence: University of

Kansas Institute o f Haitian studies. Goldsmith, John (1976): A n overview o f autosegmental phonology. - In: Linguistic

Analysis

2, 23-

68. Hall, Robert Α., Jr. (1953): Haitian

Creole:

grammar-texts-vocabulary.

Memoirs o f the American

Folklore Society 43. McCarthy, John (1988): Feature geometry and dependency: a review. - In: Phonetica

43, 8 4 - 1 0 8 .

Nikiema, Emmanuel and Parth Bhatt (this volume): T w o types o f R deletion in Haitian, 4 3 - 6 9 . Paradis, Carole and Jean-Francois Prunet (2000): Nasal v o w e l s as two segments: evidence from borrowings. — In: Language 76:2, 3 2 4 - 3 5 6 .

Nasalization in Haitian Creole

41

Poirier, Claude (2000): Le frangais de reference et la lexicologie/lexicographie au Quebec. - In: Francard, Michel (ed.): Le frangais de reference, 139-155. Leuven (Belgium): Peeters. Prunet, Jean-Frangois (1992): Spreading and locality domains in phonology. New York: Targete, Jean and Raphael G. Urciolo (1993): Haitian Creole-English Dictionary. Kensington, Md: Dunwoody Press. Taylor, Douglas R. (1947): Phonemes of Caribbean Creole. - In: Word 3, 173-170. Tinelli, Henri (1974): Generative and creolization processes: Nasality in Haitian Creole. - In: Lingua 33, 343-366. Valdman, Albert (1992): On the socio-historical context in the development of Louisiana and SaintDomingue Creoles. - In: Journal of French Language Studies 2, 99-119. - (1991): Decreolization or dialect contact in Haiti? - In: Francis Byrne and Thorn Huebner (eds.): Development and structure of Creole languages, 75-88. Amsterdam/Philadelphia: John Benjamins. - (1978): Le criole: structure, Statut et origine. Paris: Klincksieck. - (1970): Nasalization in Creole French. - In: Proceedings of the sixth international congress of phonetic sciences (Prague 1967). Prague: Academia-Publishing House of the Czechoslovak Academy of Sciences, 967-71. Valdman, Albert, Charles Pooser and Rozevel Jean-Baptiste (1996): A learner's dictionary of Haitian Creole. Bloomington, Indiana: Indiana University Creole Institute. Valdman, Albert, Sarah Yoder, Craig Roberts and Yves Joseph (1981): Haitian Creole-EnglishFrench dictionary. Bloomington, Indiana: Indiana University Creole Institute. Vernet, Pierre and Bryant C. Freeman (1988): Dictionnaire preliminaire des frequences de la langue criole haitienne. Port-au-Prince: Centre de Linguistique Appliquee de l'Universite d'Etat d'Haiti.

Albert Valdman /Iskra Iskrova, Creole Institute, Ballantine Hall 604, Indiana University, Bloomington, IN 47405 USA, Email: [email protected], [email protected]

Emmanuel Nikiema and Parth Bhatt

Two types of R deletion in Haitian Creole

1. Introduction 1

In most French-lexifier Creoles, the distributional patterns of R in segmental strings are fairly similar: all prevocalic Rs are systematically preserved, but post-vocalic and postconsonantal Rs may delete. For example, the word for land is pronounced [te], with a short and lax vowel in Central Haitian Creole (the standardized variety), Martinique and Guadeloupe Creole (Valdman 1978a, Colat-Joliviere 1978, among others), but [teR] in Northern Haitian (Fattier 1998, Valdman 1978a) and [te:R] in Reunionnais with a rather long and tense vowel, followed by a reduced R consonant (Chaudenson 1974, Staudacher-Valliamee 1992). The aim of this paper is to examine the syllabic status of R in Central Haitian Creole (hereafter HC). We will argue that R is in fact present in HC syllabic representation and furthermore that there are two distinct processes of R deletion, namely post-consonantal R deletion as in liv 'book' and post-vocalic R deletion as in te 'land', from the corresponding French etymon 7ivre' and 'terre' respectively. In particular, we show that post-consonantal R deletion is part of a more general process known in many languages as Cluster Simplification as exemplified in alternations such as meg 'thin' and megri 'lose weight'. On the other hand, post-vocalic R deletion is analyzed as resulting from the syllabification of R within the nucleus as is proposed for languages such as English (some varieties, cf. Harris 1994), Bulgarian (Zee 1988) and Quebec French (Nikiema 1999b). The article is organized as follows: Section 2 distinguishes between post-vocalic and post-consonantal R deletion; section 3 shows that even deleted on surface, R remains adjoined to the nucleus in underlying representation; section 4 suggests that post-consonantal R deletion is part of a wider phenomenon called cluster simplification; section 5 explores two possible analyses for the presence or absence of R in surface forms; section 6 presents a government-based analysis of post-consonantal R deletion, whereas section 7 explores post-vocalic R deletion. Concluding remarks are presented in section 8.

2. The two types of R deletion

The (standard) HC data below taken mainly from Cadely (1988, 1994) and Valdman (1978a,b, 1996) show that prevocalic Rs are systematically maintained both word-initially 1

We thank our audience at the International Workshop on the Phonology and Morphology of Creoles for their questions and comments; we are particularly indebted to Albert Valdman and Christian Uffmann for insightful discussions. We would also like to thank Ingo Plag and an anonymous reviewer for their useful comments on our paper.

44

Emmanuel Nikiema and Parth Bhatt

and word-internally (la), whereas the data in (lb) reveal that post-vocalic Rs are deleted from surface representation in both word-internal and word-final position. Deletion of postvocalic R is postulated, based on the comparison with the French etymon from which Creole forms are derived. We discuss this aspect in further detail in section 3. (1)

a. Pre-vocalic Rs are maintained Gloss Fr. Etymon HC [REV] 'dream' reve [Reta] 'late' retard [woj] 'stone' roche [Rivje] 'river' riviere [maRi] 'husband' mari [tRavaj] 'work' travail [pRes] 'prince' prince

b. Post-vocalic Rs are systematically deleted HC Gloss Fr. etymon [ma] 'pond' mare 'land' terre [te] [fwe] 'brother' frere [tu3u] 'always' toujours [kod] 'rope' corde [guvenmä] 'administration' gouvernement parier [pale] 'speech'

Deletion also applies to R in post-consonantal position as can be seen in the data exemplified in (2); the corresponding French etymon is given for comparison purposes. (2)

Post-consonantal R deletion in HC HC Gloss Fr. Etymon [sik] 'sugar' sucre [vät] 'belly' ventre livre [liv] 'book' [let] 'letter' lettre [meg] 'thin' maigre [pwop] 'clean' propre

HC [kat] [lit] [läk] [kof] [lib] [lot]

Gloss 'four' 'liter' 'ink' 'bin' 'free' 'other'

Fr. etymon quatre litre I 'encre coffre libre I 'autre

Post-vocalic R deletion exemplified in (lb) and (3) below is to be distinguished from postconsonantal R deletion (2) for the following reasons: First, post-vocalic R deletion is not only subject to language variability (applies in some French-lexifier Creoles such as those spoken in Central Haiti, St. Lucia, Guadeloupe and Martinique, but not in some others such as those spoken in Northern Haiti, Guyana and La Reunion), but it may also apply in both word-internal (toti 'turtle') and word-final position (five 'brother'). Second, post-consonantal R deletion is limited to word-final position and is obligatory in all varieties described in the literature. Third, cross-comparison of data from other French-lexifier Creoles (see the examples in (4) and (5)) reveals that deletion or assimilation of post-vocalic R may affect the phonetic or phonological properties of the preceding vowel, whereas post-consonantal R deletion does not have such an effect. Finally, post-vocalic and post-consonantal R deletion have to be distinct processes since they may both apply within the same phonological word as in mab 'marble', ab it 'referee' and löd 'order' from the French etymon 'marbre', 'arbitre' and 'ordre' respectively.

Two types of R deletion in Haitian Creole (3)

Post-vocalic R deletion in HC a. word-final position Creole Gloss Fr. Etymon [be] 'butter' beurre [fie] 'flower' fleur [dule] 'pain' douleur [fo] 'strong' fort [Jale] 'heat' chaleur [ri/a] 'wealthy' richard

45

b. word-internal position Creole Gloss Fr. Etymon [toti] 'turtle' tortue [3unal] 'journal' journal [zotej] 'toe' orteil [domi] 'to sleep' dormir [kod] 'rope' corde [bab] · 'beard' barbe

While post-consonantal R deletion is obligatory in all French-lexifier Creoles and limited to word-final position (2), post-vocalic R deletion may not apply in word-final or word-internal positions in some Creoles such as Northern Haitian (Fattier 1998, Valdman 1978a), Guyanais (Barthelemi, 1995) and Reunionnais (Staudacher-Valliamee, 1992). For example, the data in (4) and (5) show that deletion applies in (Central) HC, St. Lucia Creole (Carrington 1984) and Guadeloupe Creole (Hazael-Massieux, 1972), but not in Reunionnais and in Guyanais, translations are ours. Note that post-vocalic Rs are maintained in Guyanais, and slightly vocalized in Reunionnais with a surface compensatory lengthening. (4)

Gua/St-Luc./ Haitian [ma] 'pond' [te] 'land' [fwe] 'brother' [tu3u] 'always' [guvenmä] 'administration' [zeb] 'herb' [pale] 'speech'

Reunionnais 'pond' [ma: R ] R 'land' [te: ] R 'brother' [foe: ] 'always'' [tuzu: R ] [guve: R nmä] 'administ.' 'herb' [ze: R b] 'speech' [pa: R le]

Guyanais [gitaR] [kwijeR] [barb] [boRdiR] [kaRt] [mDRn] [maRb]

'guitar' 'spoon' 'bear' 'border' 'cart' 'mountain' 'marble'

Variability in the surface realization of R is also reported in Hazael-Massieux (1972:183) for some lexical items observed in Guadeloupe Creole: post-vocalic Rs are realized as [w] or replaced by a slightly long vowel. (5)

French arche corne arme morne

Guadeloupe Creole (Hazael-Massieux, 1972:183) [?awf] ~ [?a:J] [kown] ~ [ko:n]/*[kö:n] (no regressive assimilation) [?awm] ~ [?a:m] /*[?a:m] [mown] ~ [mo:n] /*[mö:n]

Stressing that the vowel preceding the nasal consonant is never nasalized in the examples above, Hazael-Massieux (1972: 141) suggested that, in order to account for the absence of automatic nasalization, [w] is to be considered a latent consonant (i.e. the consonant is still present somehow). Since [w] appears to be an allophone of /R/, we extend this hypothesis in suggesting that post-vocalic Rs may be underlying consonants in some Creoles, or more

46

Emmanuel Nikiema and Parth Bhatt

precisely are syllabified in the nucleus. In other words, R is part of the phonological representations of lexical forms. 2 In summary, post-consonantal R deletion applies in all French-lexifier Creoles but only in word-final position: compare liv in (2) where deletion applies to the word-initial clusters travaj and pR§s in ( l a ) where it does not. Post-vocalic R deletion is subject to variability since it may delete in some French-lexifier Creoles and not in others. Moreover, post-vocalic R deletion may apply in both word-internal and word-final positions, contrary to postconsonantal Rs. Finally, post-consonantal R deletion is limited to word-final position, regardless of the preceding vowel or consonant (2). One potential problem with the view presented so far is that R deletion is posited in French-lexifier Creoles based on a comparison with the corresponding French etyma. Although most creolists agree that the lexicon of French Creoles is mainly derived from varieties of French spoken during the 16th and 17th centuries, it does not follow that R is necessarily part of the underlying representation of Creole forms. The question therefore is as follows: what motivates the R deletion hypothesis in (standard) HC? In the next section, we show that in order to account for the HC data, one must postulate an underlying R.

3. R is part o f H C u n d e r l y i n g r e p r e s e n t a t i o n s

One major reason for positing R in the underlying representation of Haitian forms lies in the existence of an opposition between oral and nasal/nasalized vowels in the same phonetic environment as illustrated in (6). Regressive nasalization can apply optionally in (6a), but is systematically blocked in (6b). Comparison of the two sets of data with their corresponding French etyma shows that the only difference is that regressive nasalization is blocked in all cases where there is an R in the corresponding French etymon. (6)

a. Regressive lam [lam] chen [Jen] zami [zami] fami [fami] pom [pom] -

nasalization (optional) [läm] 'wave' {lame) [Jen] 'chain' (chaine) - [zämi] 'friend' (ami) [fami] 'family' (famille) [pöm] 'apple' (pomme)

b. No regressive nasalization [lam] /*[läm] 'tears' [Jam] /*[Jam] 'charm' [amwa] /* [ämwa] 'cabinet' [fom]/*[fom] 'form' [kon] /*[kon] 'horn'

(larme) (charme) (armoire) (forme) (come)

In all cases where nasalization optionally applies (6a), there is no R in the French etymon. However, in all cases where nasalization is blocked, the corresponding French etymon contains R. We therefore posit that there is an underlying (post-vocalic) R in the Creole forms that blocks nasalization of the preceding vowel. Although Valdman (1978a) argues against this view, other authors including Hazael-Massieux (1972) and Cadely (1994) have

2

In this paper, we will only deal with the syllabic representation of R in lexical forms and not with what w e believe to be an epenthetic R which occurs in a complex sandhi context (papaRam 'my father'). These are two separate phenomena.

Two types of R deletion in Haitian

Creole

47

made similar proposals in favor of underlying Rs. One would be hard pressed to explain why nasalization is blocked in (6b) if R is not part of the underlying representation of the lexical items. Another argument for positing R in Haitian underlying representations, is the absence of nasal(ized) vowels that are followed by post-vocalic Rs in both French and French Creoles. 3 As will be shown in section 7.5, we account for this cooccurrence restriction by proposing that post-vocalic Rs are syllabified in the nucleus (with the preceding V), thus occupying the target for nasality and blocking it. We are not claiming that all Creole forms in which blocking of nasalization takes place contain an underlying R. Our position concerning the relation between the presence of an underlying R and the presence or absence of a nasal vowel will be presented in more detail in section 7. For the moment we present a brief outline of our proposal as follows. There are to our knowledge six contexts in which nasal consonants occur in the underlying representations of HC: a) b) c) d) e)

the nasal consonant is in absolute word final position, i.e. VN as in plan. the nasal consonant is followed by an underlying consonant, i.e. VNC (as in dan-dantis) the nasal consonant is followed by a pronounced consonant i.e. V N C 0 as in plant. the nasal consonant occurs between two vowels, i.e. VNV as in emab. the nasal consonant is preceded by a vowel and followed by an empty nucleus, i.e. V N 0 as in pom. f) the nasal is preceded by a vowel and an underlying consonant, i.e.VCN as in mon In cases a), b) and c), the vowel preceding the nasal consonant is always nasal. In cases d) and e) the vowel preceding the nasal consonant is oral but may optionally be nasal (regressive nasalization). In case f) the vowel preceding the consonants is oral and never nasal. With respect to syllabification, we have proposed in Bhatt and Nikiema (2000) that for forms which have patterns a) VN, b) VNC and c) V N C 0 , the nasal consonant in fact syllabifies within the nucleus and does not have its own temporal position. Vowel nasality is thus derived from the presence of the nasal consonant (as we shall show in more detail below). In forms which follow patterns d) VNV and e) V N 0 , however, the nasal consonant is associated to its own temporal position. In general these forms are pronounced with an oral vowel. Nasality may propagate leftwards due to regressive assimilation, but this is not obligatory and is subject to variation between speakers. We propose that in these cases the nasal consonant in fact syllabifies in onset position. It is important to note that the vast majority of cases where optional regressive nasalization occurs are those which end in an empty nucleus. We propose that it is in fact the presence of a final empty nucleus which triggers the possibility of regressive nasalization. In other words, since there is no final full vowel, the nasal consonant is treated as if it were in word-final position. The syllabification of the nasal consonant is thus subject to parametric variation in these cases with some speakers associating the consonant to the onset and not spreading nasality to the vowel, with others allowing the consonant to appear as a rhymal consonant and spreading nasality to the vowel. There are only two exceptions in French known to us, namely the words genre and denrie.

48

Emmanuel Nikiema and Parth

Bhatt

In forms which follow the pattern in f) VCN, however, the vowel is never nasal underlyingly. Furthermore, this vowel is never subject to regressive nasalization. In these forms, the nasal consonant is once again associated to its own temporal position and syllabifies in onset position. Bhatt and Nikiema (2000) analyze nasal vowels in St. Lucia Creole as consisting of an underlying nasal consonant that is adjoined within the nucleus to a vowel. More precisely, we proposed that nasal vowels are the result of spreading of the Spontaneous Voicing node (Avery 1996) from the nasal consonant to the immediately preceding vowel. This is illustrated by forms such as bon [bö] 'good' in which the vowel is always nasal and cannot be oral *[mon]. In this case the nasal consonant syllabifies within the same timing position as the vowel. Similarly we will propose that post-vocalic Rs are associated to the temporal position of the preceding vowel to form a diphthong. It is this R diphthong that blocks nasalization of the vowel in forms such as lame 'armed forces' or chanm 'charm'. Our analysis would indicate that the vowel of the first syllable la should be pronounced as an oral vowel. In other words, in a lexical item such as mon 'hill' or 'mountain' [man] (from French morne), the vowel is always oral and cannot be nasalized by the following nasal consonant. Our claim here is that when an underlying R intervenes between the vowel and the nasal consonant, nasalization of the vowel is blocked. The feature that gives the nasal quality to the vowel cannot 'pass through' the intervening consonant. This does not mean, however, that whenever an oral vowel precedes a nasal consonant there is an underlying R. In some cases, for example, an oral vowel may appear next to a nasal consonant because of syllabification. In forms such as lam [lam]/[läm] 'blade', for example, the vowel is usually oral [lam], but can optionally be pronounced as a nasal [läm] by some speakers. In the form lam [lam], the vowel remains oral even though it is followed by a nasal consonant because the consonant is anchored in onset position. In other words, even though the vowel is followed by a nasal consonant, the two segments syllabify in different syllables. In some cases, the presence of an empty nucleus following the nasal consonant may lead some speakers of St. Lucia Creole to treat the nasal consonant as if it were also within the same syllable as the vowel (i.e. in a coda position), therefore triggering regressive nasalization of the vowel [läm]. In HC we find similar forms where a nasal consonant follows a vowel such as bön [bon] 'female servant' and pan [pan] 'breakdown (of a car)'. In HC, optional nasalization is not possible for these forms since they are always pronounced with an oral vowel. We interpret this observation as indicating that, in HC, the fact that the vowel and the consonant belong to different syllables (a vowel followed by an onset consonant) excludes the spreading of the Spontaneous Voicing node to the vowel. This in turn indicates that the optional regressive nasalization rule depends on the individual speaker. It is essential to notice, however, that in this case the nasal or oral quality of the vowel is not in any way related to the presence or absence of an underlying R, but rather to the syllabification of the nasal consonant. This approach correctly accounts for the possible variations in surface form: a) vowel nasality is obligatory when the nasal consonant is syllabified within the nucleus. b) vowel nasality is blocked when a consonant appears between the vowel and the nasal consonant, c) vowel nasality is optional when the nasal consonant appears in onset position between two vowels or between a vowel and an empty nucleus.

Two types of R deletion in Haitian

Creole

49

Let us now turn briefly to forms such as lame [lame] 'the old woman' (Fr. la mere) vs. lame [lame]/[lame] 'the sea' (Fr. la mer). According to the analysis presented above, the vowel of the initial syllable la should be oral since the nasal consonant is in onset position for both forms. In the form lame [lame]/[läme] 'sea', however, nasality on the first syllable is optional (Valdman and Iskrova, this volume). We believe that in these cases, some speakers have, by analogy, generalized the distinction between oral and nasal vowels in order to produce a lexical distinction. In other words, since words such as mon [mon] 'hill' and monn [mön] 'people' or 'world' are differentiated by the quality of the vowel, speakers may attempt to distinguish between lame 'old women' and lame 'the sea' in the same way (Burstynsky 1997). This difference, however, has more to do with strategies used to distinguish homophonous forms in the lexicon, than with segmental phonology or syllable structure. The underlying syllabic representation of the forms is identical, as is the case for many other homophonous lexical items. The presence of ambiguity is another reason that has lead us to not only propose underlying Rs in HC, but also to distinguish between post-vocalic and post-consonantal R deletion. The data illustrated below show that a surface form such as [kat] (7a) may be derived from either an underlying /kaRt/ 'card' through post-vocalic R deletion, or from an underlying /katR/ 'four' via post-consonantal R deletion. In other words, R deletion may create ambiguity when not identified as post-consonantal or post-vocalic. We therefore distinguish two types of R deletion processes: post-consonantal R deletion which is automatic (applies in all French Creoles to comply with syllabic structure licensing requirements), and postvocalic R deletion which is parametrically determined and which may differ between French Creoles. (7)

a. Post-vocalic or post-consonantal R

b. Deletion of R or other consonants

[kat] < carte / quatre [zeb] < herbe / zebre [sik] < cirque / sucre

[sab] < sable / sabre [lam] < lame /1arme [vit] < vite / vitre

4. Post-consonantal R deletion is part of a wider p h e n o m e n o n

We have shown that post-vocalic and post-consonantal R deletion are two different processes: post-vocalic R deletion does not occur in all French-lexifier Creoles, it may apply in either word-internal or word-final position; it may affect the quality and length of the preceding vowel, and it blocks regressive nasalization. On the other hand, post-consonantal R deletion is obligatory, limited to word-final position and applies in all French-lexifier Creoles. We will now show that post-consonantal R deletion is part of a wider phenomenon known as cluster simplification. Cluster simplification is a deletion process that affects the rightmost member of a word-final sequence of consonants. The phenomenon is observed in all French Creoles, in Colloquial French, in Quebec French (Pupier and Drapeau 1973; Nikiema 1999b), and in other languages such as Catalan (Mascaro 1989) and many varieties of English (Goldsmith 1990). For a detailed and more recent analysis of cluster simpli-

50

Emmanuel Nikiema and Parth Bhatt

fication in HC we refer the reader to Nikiema and Bhatt (1998), and to Cöt

Χ

X

Ν Ν Ν" In conformity with the general spirit of the times, I will assume that both of these aspects of syllabification are universal. The variation in syllable types found in individual languages is then a consequence of the different orderings of various syllabic constraints in an optimality theory model of phonology. I will make no mention of these constraints in this article in order to save space. Interested readers should refer to Smith (2003) for an OT-analysis of the facts discussed in this article, but without the extra evidence now provided by newly discovered early Sranan material (Van den Berg 2000). This new material derives from a recently discovered Sranan text of the 1762 peace treaty between the Dutch colonial authorities in Surinam and the Saramaccan maroon tribe. In order to account for onset and coda clusters I revise Levin's model to admit recursion, replacing at the same time the symbol X - standing for X-slot or skeletal slot - with the symbol R (' Root). Roots can be regarded as equivalent to phonological segments. I recast the syllabification rules in the form of a formal X' schema in Speas's terms (Speas 1990).

New evidence from the past: To epenthesize or not to epenthesize?

(8)

a. Ν b. Ν' c. N"

= = =

95

(R") Ν (R/N") (R/N") N'

Part a. of the schema defines the nucleus as consisting of zero or more roots (X in Levin's terms). I have decided not to limit this to a maximum of two units, as is often done. I consider this to be a task for the constraints in an OT model. Part b. of the schema defines the rhyme (N') in terms of a nucleus (N) - the head - and an optional coda. The coda is conceived of by Levin as the complement of the syllable. Here I introduce the option of recursion - the coda (complement) consisting either of a root (a segment) or a syllable (an N"). In itself the addition of recursion brings this approach closer to the general spirit of the X' model, which of course allows for recursion in syntax. Part c. of the schema defines the syllable (N") as consisting of an optional onset and a rhyme - the head. The onset is conceived of by Levin as the specifier of the syllable. Once again recursion is possible - the onset (specifier) consisting, like the coda, either of a root (a segment) or a syllable (an N"). As mentioned above, the notion of recursion in phonotactic structures was first utilized by Garcia-Bellido (1997). As far as I am aware this was the first use made of recursion in the phonotactic hierarchy. This in itself is not surprising since this hierarchy has not really been developed much as such, or even identified as a separate hierarchy. In the prosodic hierarchy, which I regard as a separate hierarchy from the phonotactic one (see Smith (2000)), we may compare the recursive nature of accent. An accented mora may be the strongest mora in a foot, an accented foot may be the strongest foot in a word, an accented word may be the strongest word in a phonological phrase, and so on. I have adopted Garcia-Bellido's analysis of consonant clusters as dependent syllables. In Smith (2003), I attempted to show that this assumption provides an analysis of the Creole languages of Surinam whereby the differences amongst the various lects become much smaller in theoretical terms, and thereby easier to understand. An important restriction I base on Fikkert (1994) is that non-dependent syllable nodes require to have scope over nonhead elements in dependent syllables obeying the sonority hierarchy (leaving any (epenthetic) nuclei in such syllables out of consideration). As we will see this restricts severely the type of cluster that may be analysed in terms of dependent syllables. In short we are restricted to onset clusters where the first element (the dependent onset) is less sonorous than the second (the dependent coda) and coda clusters where the first element (the dependent onset) is more sonorous than the second (the dependent coda). Typically the clusters so analysed will be onset clusters whose second element is a nasal, liquid or glide, or coda clusters whose first element is a nasal, liquid or glide. The question might well be raised what the limits are on clusters. This question obviously also has a bearing on the formulation of the sonority hierarchy itself - the fewer the number of categories recognized, the smaller the number of consonants permitted in a sequence. Clusters like /gnj-/ in Buchan Scots /gnja:v/ 'gnaw' would seem inherently to be possible intrasyllabic clusters.

96

Norval

Smith

3. T h e Surinam creole epenthesis problem - s o m e initial facts

In order to provide the reader with an easy introduction to the problem at issue I will give a couple of examples of the treatment in some of the creole dialects/languages of Surinam of words of English origin which have an initial liquid cluster in English. In the following table - amplified from (1) - the reflexes in Sranan, Aluku, Kwinti, Ndyuka and Saramaccan are given of the two English words "dry" and "blow". Lect Sranan Aluku Kwinti Ndyuka Saramaccan

CrCrCV(DCV(D-

cvcv-

Example drey dee ~ dele dee ~ del6 dee dee

'dry' 'dry' 'dry' 'dry' 'dry'

ClCrCV(1)CV(1)-

cvcv-

Example bro boo ~ bolo bolo 3 boo boo

'blow' 'blow' 'blow' 'blow' 'blow'

Note that in Sranan we find a neutralization of post-consonantal /r, 1/ of English to Irl in nearly all cases. All the other lects display what are apparently much more deviant reflexes. Aluku (Bilby 1993), in particular, has a very interesting reflex of these liquid clusters. There are always two options. One has an extra low-toned vocalic mora in the former position of the liquid, although as we will see later this correspondence is illusory in the sense that this mora does not correspond to the liquid either segmentally or structurally. The other option involves a lateral liquid preceded by an epenthetic vowel - once again identical in quality to the following vowel. This second option is immediately recognizable as an intermediate stage in the historical derivation of the liquidless variant, not only in Aluku itself, but also in Saramaccan and Ndyuka (cf. Bilby 1993; Smith 1987). In Kwinti (cf. Huttar and Smith 1984; Smith 1993) - we may suspect, following Bilby (1993) that a similar situation pertains or pertained until recently, although the published data are not entirely unambiguous here. The situation in Ndyuka may also have been similar a hundred years ago. Bilby (1993) quotes a written source dating from approximately 1920, utilizing a syllabic orthographic system, the Afaka script (cf. Dubelaar and Pakosie 1988; see also Bilby 1993). In this spelling Ndyuka forms of English words with /l/ or Id generally have a liquid corresponding to the English one. The reflexes of English words with initial clusters frequently (though not always) show a liquid (either / or r) following an echo-vowel. So we find golo-n 'ground' and bo-lo 'blow'. In Saramaccan (cf. Voorhoeve 1961) all non-initial liquids in words of English origin have been lost, and nearly all in words of Portuguese origin. This has happened since the 18th century, when these liquids were still present, as is clear from Schumann's 1778 word-list (Schuchardt 1914).

T h e data available to m e from Kwinti is rather limited. It is mostly drawn from I luttar and Smith (1984). The lack of a form /boo/ here should therefore not be taken as implying the non-existence of such a form.

New evidence from the past: To epenthesize or not to epenthesize?

97

In the modern language however the only trace of the liquid is normally once again an extra vocalic mora (identical in quality to the following vowel) after the initial consonant. This extra mora bears a default low tone (except when influenced by tone-sandhi rules that are outside the scope of this paper). The only language which allows complex onset clusters systematically is Sranan, where in the modern language we find initial liquid clusters corresponding to their English counterparts.

4. Sranan initial liquid clusters - more than English

Now it is a fact of Sranan that many more initial clusters occur in words of English origin than in English itself due to syncope of unstressed syllables. Initial liquid clusters in Sranan go back to three kinds of English phonological structure. (10) a. Words EME 4 EME b. Words EME EME c. Words EME EME

which have an initial liquid cluster in English, e.g. /br5k/ /broko/ 'broke' /kl5s bai/ /krosbey/ 'close by' which have a final or intersyllabic liquid cluster in English, e.g. /kürt/ /krütu/ 'court' /self/ /srefi/ 'self with an initial open syllable followed by a single liquid in English, e.g. /sürinam/ /sranän/ 'Surinam' /Jiliq/ /sren/ 'shilling'

Up till now the evidence from earlier historical stages of Sranan has been that while type (10b) and (10c) items sometimes occur without initial clusters - with an epenthetic vowel intervening between the initial consonant and the liquid, type (10a) items do not. Two schools of thought have existed on this. The first (cf. Sebba 1982) claims that the type a. clusters have always been pronounced as clusters. The second, to which I belong (cf. Smith 1977), claims that in fact these clusters did contain a (brief) epenthetic vowel. However, because the earlier recorders were able to identify these lexical items as English, Dutch, or even Portuguese, they knew that these items started in these lexifier languages with clusters, and therefore retained the clusters in the spellings they employed. Alternately, it has been claimed that the recorders did not hear these vowels because they were so brief (Voorhoeve 1961). The cases of apparent orthographic metathesis in words of the type of (10b) were interpreted by Sebba (1982) as cases of true phonological metathesis. A phonetic basis for VC-metathesis of this type has been proposed in Blevins and Garrett (1998). However, the existence of CVLVC-structures in such words, richly illustrated in Van den Berg (2000), rules this out as a possible option.

4

EME refers to the directly ancestral form in the Early Modern English of London (as argued for in Smith 1987).

Norval Smith

98

Now, some words of the second and third types occur in earlier recordings of Sranan in forms with epenthesized initial clusters. Words of the second type, in particular tend to exhibit three types: a. with CVLC.. structure (as in English); b. with CV(L)VC.. structure; and c. with CLVC.. structure. English

Sranan 1783

court

1798

1855

Modern

kroetoe

kroetoe

krutu

wroko

woroko wroko

wroko

Aluku 1-variant

Saramaccan

a. kuiitu

b.

work

c. kruttu a. worko b.

self

c. wrokko a. b.

help

c. srefi a. helpi

srefie

b. • 5

c.

reppie

wooko

serefi srefi helpi herepi/yerepi yrepi (1856)

selefi

seepi/seei

yelepi

heepi

srefi

yepi/lepi

In the earliest period we tend to find some forms resembling the English model, although the appearance of these is inconsistent. Focke's careful renditions in 1855 include a number of cases where the liquid is intervocalic. Historically, but not chronologically (in terms of the time of recording) these would appear to be intermediate cases. So we can apparently set up the sequence: (12) work(o)

>

woroko

>

wroko

The second stage seems to lie behind the various Maroon lect renditions lacking a liquid, which I give in the table above. Certain complications arise due to difficult initial clusters in Sranan - compare the forms for "help". Once again this aspect is tangential to the present discussion. In the next table I illustrate some words of the third type: English shilling calabash Surinam

5

Sranan 1783 b. shelling c. b. kallabassi c. b. c.

Saramaccan 1798

1855

Modern

sreen calabassie

sren, sring sren

seen kaabäsi krabässi

krabäsi

s ran am

sranan

saanä

reppie and modern Ιέρί result from the simplification of the difficult cluster /yr/.

New evidence from the past: To epenthesize or not to epenthesize?

99

Once again we see that at least some early forms resemble their English models in that they do not have the clusters that appear in modern Sranan. For the a. type of case - with an initial cluster in English - 1 have been compelled up till now to rely on indirect evidence from Schumann's word-list of Saramaccan, as discussed in the next section, which is largely taken from Smith (2003).

5. Evidence from Schumann's recording of Saramaccan (1778)

Smith (2003) quotes two kinds of evidence in support of the "epenthesis-hypothesis". Firstly, there are 18th century recordings of Saramaccan where on the basis of the sourceword in the lexifier language and the modern form of the word it seems to be impossible to accept the 18th century cluster recording as correct. Secondly, there are a few cases where for some reason or other these epenthetic vowels were actually recorded. The examples of the first kind all come from Schumann's 1778 word-list of Saramaccan. However, as Schumann was also responsible for the best 18th century dictionary of Sranan (Schumann 1783), it can be assumed that the same reasoning can be applied to the orthography of both works. Frequently, nearly all corresponding words are spelt the same in the two sources. One very clear example is: Portuguese form poraque

Portuguese meaning 'electric eel'

Saramaccan (1778)

Saramaccan (20th c.)

plakkeh (cf. mod. Sranan prake)

polake pulake6

The first vowel of this (Brazilian) Portuguese word would be pronounced [o] or [u]. Both pronunciations seem also to occur in modern Saramaccan sources. However Schumann gives a spelling with an initial orthographic cluster in his 18th century word-list. If this had been a correct rendition phonologically speaking the modern reflex would have had to be */paake/. In other words, the occurrence of the same vowel in the first syllable, qualitatively different from that in the second syllable, at stage 1 and stage 3 in the history of the word guarantees the existence of this same vowel at stage 2, whatever the orthography might be. One might argue that the "restored" quality at stage 3 is due to phonological universals after all we have here a less prominent syllable, whose labial vowel might be due to the initial labial consonant of the word. This kind of argument, which I would reject for the Surinam Creole initial clusters, is not even possible for the following (non-initial) case where Schumann also recorded a cluster.

6

From the joint Leiden University-Summer Institute of Linguistics-Surinam Ministry of Education Surinam Creole Etymological Project.

100 (15)

Norval Smith Portuguese form escuro

Portuguese meaning 'dark'

Saramaccan (1778) sukru

Reconstructed 1778 form *sukiiru

Saramaccan (20th c.) zugüu, suguu

The reason why I quote this case - despite it not being initial - is that it involves the hightoned equivalent of a Portuguese stressed vowel being omitted by Schumann, resulting in an orthographic cluster which cannot correspond to any kind of phonological reality. One reason for Schumann's mis-recording here may concern precisely the difference between tone and stress. To anyone who might say that this is not relevant for Sranan because there is no evidence that Sranan was ever a real tone language (although as Smith and Adamson (1996) have shown it does make limited use of a high tone under specific circumstances, and also exhibits downstep phenomena) I would point to the existence of Ndyuka, Aluku, Paramaccan and Kwinti, four maroon lects, all of which derive historically from Plantation Sranan, and all of which are tone languages, indicating that Plantation Sranan itself was a tone language throughout the 18th century. The second type of evidence where epenthetic vowels do split up European language clusters is also quite rare. Source-language

Source-word

English

bully-tree

Portuguese

estrela

Portuguese

mostrar

Portuguese

criar

Portuguese

embrulhar

Sranan (1783) Saramaccan (1778) boltri bullitiri

Sranan (20th c.) Saramaccan (20th c.) bortri botii

-

-

terija

teea, te£ja

-

-

mussula (krijä) kiljä bruja7 bulja

musulä kriä (1855) kiiä bruja buujä

Here we once again find our direct evidence only in Schumann's Saramaccan word-list (1778), compiled 5 years before his Sranan dictionary (1783). The point I want to make here is that Schumann writes an epenthetic vowel in a number of cases where the European etymon - which in all these cases contains a cluster - is less obvious. This applies at least to some of these cases. The first example is of direct relevance, seeming to contain as it does a reflex of the English word tree. Note also that Schumann possibly missed the relationship of this word to the English word tree as he glosses it as 'bullentri'. The eighteenth century Saramaccan form terija 'star' has fooled more than one commentator into considering this to be a Spanish-derived form (< estrellä). In fact it can be 7

This form and bulja are given in Schumann (1783) as Djutongo (an 18th century form of speech more or less equivalent to Saramaccan) forms used in Sranan. Only bulja is given for Saramaccan itself, however.

New evidence from the past: To epenthesize or not to epenthesize?

101

easily explained as the result of a liquid dissimilation process, fairly persistent in the history of the Surinam Creoles, applying to the Portuguese word estrela (cf. Smith 1987: 385). What interests us here of course is the fact that it appears as tereja and not treja. Due precisely to Schumann's unfamiliarity with the liquid-dissimilation process whereby a liquid directly following an accented syllable is dropped if a liquid is present earlier in the word paralleled in examples such as Saramaccan (1778) trobi < *t(o)röbili < trouble, bali < *balili < barril (Portuguese 'barrel'), probossi < *p(o)roboseri < provoceren (Dutch 'provoke'), proba < *p(o)robora < polvora (Portuguese 'powder'), he possibly failed to see the connection with the Portuguese word for 'star'. The next word is sufficiently unlike Portuguese mostrar in form and meaning to have resisted identification (in modern Saramaccan the meaning is given as 'stand with crossed legs, run zigzag to make a fool of someone' and in the 18th century 'a bow made while complimenting someone', which corresponds closest in Portuguese to the meaning of the reflexive form mostrar se 'show off, make a public appearance'). Note that this word has preserved a form in modern Saramaccan which is virtually identical with the 18th century rendition. Here the normal liquid-cluster strategy of copying the following vowel for the epenthetic vowel has not been followed - here instead the sibilant-cluster strategy of inserting an epenthetic Iii or /u/ (this last in labial environments) has been followed, the choice being influenced by the preceding /u/. This has aided in the retention of the liquid, which in general in the Surinam maroon Creoles has been retained longer between nonidentical vowels. The two following examples are parallel to each other. The Saramaccan versions both involve a liquid cluster preceding an unstressed vowel syllable which in turn precedes a stressed vowel syllable. In each case Schumann (1778) has noted the first (epenthetic) vowel and ignored the organic unstressed vowel following. I assume that these two words have developed as follows: (17) criar embrulhar

kriär äbruAär

> >

*kiri(j)ä *burujä

> >

*kili(j)ä {kiljä} * bulujä {bulja}

> >

kii(j)ä buujä

To sum up, this evidence is only of direct application to Schumann's 1778 recording of Saramaccan. The relevance of this for his Sranan dictionary of 1783 can never be more than suggestive. It is for this reason that the evidence revealed by the 1762 peace treaty is so interesting.

6. The evidence of the 1762 peace treaty with the Saramaccan Maroons

The Sranan version of this treaty was published in Hoogbergen and Ροΐϊηιέ (2000). This edition is unfortunately rather inaccurate in some points, and I employ the more careful rendition provided by Van den Berg (2000). This text provides us with evidence of variation between forms with initial liquid clusters and forms with this cluster broken up in English lexical items of all three types.

102

Norval

Smith

(18) English

Aluku S'maccan

Sranan 1762

c.1765

1783

1798

1855

Modern

broke creek

brokke

brokko

bröko

bröko

kriekie

kriki

kriki

klosieby

krösibei

krosbiy

kiiki

klossi by

klossibai

c. granman

granman

granman

a. courtoe

koerte

grand-

b. garanman

man

[galän]

kuütu

c.

kroete

kruttu

a.

worke

worko

kroetoe

wrokko

c.

help

kröetoe

krütu

woroko

b. self

gaamä

granman

b. couroutoe work

boöko

b. korosie bay c.

court

broko

b. c. criqiii

close-by

ant bolöko

b. boroko c. broko

1-vari-

wroko

wröko

woöko wröko

a. selfi serdfi

b. serefie c. sreffie

sleffi/zlifi srefi

a. helpie

helpi

srefie

helpi

srifi

selöfi

se6pi/se6i

yelöpi

heipi

srifi

helpi herepi/

b.

yerepi c. shilling

b. serem

schelling

b.

kalebassi

yrepi (1856) yipi/ldpi

sreen

sren

shelling

c. calabash

reppie

kallabassi

se6n sren

cala-

kaabäsi

bassie c. Surinam

krabässi

krabäsi

sranäm

sranän

b. Saranam c.

saanä

Note that particular passages are typified by a greater frequency of orthographic epenthesis than others. This I presume does not so much reflect a variation in the pronunciation as an uncertainty about how to record words of the types studied in this paper.

New evidence from the past: To epenthesize or not to epenthesize?

103

7. Theoretical structural analysis

Intrasyllabic consonant clusters will be defined as involving recursive syllables appearing at the onset and coda positions as defined by Levin's X' approach.8 (19)

N" Ν"

N' Ν

Ν"

The dependent N"-nodes represent consonant clusters. As noted by Garcia-Bellido (1997) there is a contrast in Spanish between the following forms: [parado] [parado]

(20) a. prado b. parado

'prairie' 'stopped'

There is a parasitic vowel appearing between the members of the initial cluster in (20b). In terms of my approach these two words would appear as follows: (21) a.

Ν"

N"

X I Ν" N'

/I R N'

/ l \ I R Ν R Ν

Ν

R

R

!

I d ο

p a r a

b.

Ν" / I R N'

N" / I R N'

Ν

Ν

Ν

I R

I R

I R

a

d o

p a

8

Ν" / I R N'

[p"rado]

r

[parado]

This analysis can in fact easily be extended to larger phonotactic units involving inter-syllabic clusters as well. Compare my analysis of Leurbost Gaelic syllable structure (Smith 1999).

104

Nerval Smith

In other words what is being stated here is that there is always a potential for such secondary epenthetic elements in clusters. It will depend on the facts of the language concerned how these clusters are realized - with or without an epenthetic vowel. In itself the occurrence of such an epenthetic vowel will not alter the status of the syllable concerned as a recursive syllable. Note however that epenthetic vowels in dependent syllables do not count for the Sonority Sequencing Principle while the consonants do. When such epenthetic vowels appear or not will depend on the constraint-ranking in the particular language or dialect. They may be compulsory in certain clusters, and disallowed in others. They may be restricted to certain speech-styles. In English there is even an informal spelling convention representing the appearance of certain epenthetic vowels in case of extreme emphasis. This is seen in the "uh" in (22). (22) Normal /pli:z/

Orthography please

Extreme Emphasis /psli:z/

Orthography 9

For the Surinam Creoles I claim in Smith (2003) that this reduces the basic distinction between the representations of blow in Aluku /bolo/, and in Sranan / b r o / (cf.l, ignoring the difference in the liquids) to a fairly trivial one - whether the nucleus position in the dependent onset-syllable is realized or not. N"

(23) Ν"

N'

/l\ I

N" Ν"

N'

/l\ I

R Ν R Ν

R Ν R Ν

R

R

b ο Aluku

1

r

ο

'blow'

Sranan

What am I actually claiming here? After all Aluku is supposedly a language of which it could be said, pretheoretically-speaking, that there are no codas. Now it appears that the claim is being made that in a particularly marked type of syllable there are coda-elements. To this the riposte could be made that if it is necessary to have codas then it is, firstly, preferable to have vowelly elements in them (e.g. sonorants), and, secondly, if we have to have a coda what better place for them than in an onset! Also it remains the case of course that no superordinate syllable has a coda. The next logical development is to a stage where there are no codas at all. We find this stage in the non-1 variant of Aluku, and in Saramaccan.

9

Taken from a notice on the library door at the Department of Linguistics of the University of Massachusetts at Amherst. This spelling with "uh" represents a fairly general convention in American English, appearing commonly in comics, for example.

New evidence from the past: To epenthesize or not to epenthesize?

(24)

N" N"

N"

N'

/ l \

105

Ν" / I R Ν

I

Ν' I Ν

R Ν R Ν R R I 1) ο 1 ο Aluku 1

b >

ο

'blow'

Aluku 2

In other words the marked element in the dependent syllable - the /l/ - has now been lost at the expense of deleting the coda. Note, however, that we still have a dependent syllable the epenthetic /o/ remains subsidiary to the following vowel in the superordinate nucleus. Its quality is derivative, and in Saramaccan at least the corresponding vowel is subject to tone-sandhi. Here I have adduced early evidence from Sranan showing that precisely the same syllabic patterns as are found in Aluku (the 1-variant) are possible in Sranan. I suggest in Smith (2003) that such patterns represent a direct reflection of patterns found in the Gbe languages, where various "clusters" are broken up by brief epenthetic vowels identical to the following organic vowel, as is now evidenced from all the Surinam Creoles.

8. G b e clusters

In Fon liquid-clusters involve /, except those with an initial coronal, which take either / or r. There is in fact no contrast between / and r in Fon (or any other Gbe language), just one phoneme which we might call l\l as this is its usual value, as in fact is also the case in most Surinam Creoles. Höftmann (1993: 48) says that preceding / a weak (non-nasalized) copy of the following vowel is heard. This is not the case before the /--variant: (25) bio fie flin dlo

= = = =

[b°lo] [fie] [Hi] [d°lo] ~ [dro]

The status of these as underlying phonological clusters is guaranteed by the option of /r/ with coronals. If there was an input rather than an output vowel intervening, this phonotactic relationship would be most unusual. In Gen there is also a single liquid phoneme l\l with phonetic reflexes / and r. As far as clusters are concerned the situation is similar to that in Fon, except that only r appears fol-

106

Nerval

Smith

lowing coronals. In Gen we also find a brief epenthetic vowel, but it appears in phonetic O - c lusters, and not in C/-clusters - the opposite situation from in Fon (Bole-Richard 1983). We can draw no overall conclusions from the distribution of epenthesis in clusters in these two lects, i.e. one dialect of Fon, and Gen, especially as the facts are so different. Various Gbe lects were spoken by slaves brought to Surinam, but we do not yet have access to descriptions of the phonetics of all of these. What is clear is that epenthesis occurs both in Surinam and on the (historical) Slave Coast, and this may well be connected.

9. C o n c l u s i o n

There is new evidence from an early Sranan source, datable to 1762, which makes clear that all three types of initial liquid cluster in modern Sranan can probably be traced back to a stage where epenthetic vowels divided up the consonants in these "clusters". At the same time the theory of syllable structure adduced here reduces the difference between epenthesized and non-epenthesized clusters, making the diachronic and synchronic relationship between the two types more comprehensible. We owe a great debt, both to the early recorders of Sranan for recording so much data, and the modern (re)searchers who keep turning up earlier and earlier evidence pertinent to the history of Sranan and its relatives.

References

Bilby, Kenneth M. (1993): Latent intervocalic liquids in Aluku: Links to the phonological past of a Maroon Creole. - In: Francis Byrne and John Holm (eds.): Atlantic meets Pacific: A global view of pidginization and creolization, 25-35. Amsterdam: John Benjamins Publishing Company. Blevins, Juliette and Andrew Garett (1998): The origins of consonant-vowel metathesis. - In: Language 74 (3), 508-555. Βοΐέ-Richard, Remy (1983): Systematique phonologique et grammaticale d'un parier Ewe: le Gen-Mina du Sud-Togo et Sud-Benin. Paris: Editions L'Harmattan. De Saussure, Ferdinand (1915 [1959]): Course in general linguistics. New York: Philosophical Library. Dubelaar, Cornells Ν. and Andre Pakosie (1988): Seven notes in Afaka script. - In: Nieuwe WestIndische Gids/New West Indian Guide 62, 146-164. Fikkert, Paula (1994): On the acquisition of prosodic structure. The Hague: Holland Academic Graphics. Garcia-Bellido, P. (1997): Functional/lexical categories: The Spanish syllable structure. [Unpublished paper presented at the HILP3 Phonology Conference, Free University, Amsterdam, January 1997.] Höftmann, Hildegard (1993): Grammatik des Fon. Leipzig: Langenscheidt. Hoogbergen, Wim and Thomas Polime (2000): De Saramakaanse vrede in het Sranantongo. - In: OSO 19.

New evidence from the past: To epenthesize

or not to

epenthesize?

107

Huttar, George L. and Norval S.H. Smith (1984): The development of the liquids in Kwinti. - In: Peter C. Muysken and Norval S.H. Smith (eds.): Amsterdam Creole Studies VII, 21-30. Levin, Juliette (1987): Constraints on syllabification in French: Eliminating truncation rules. - In: David Birdsong and Jean-Pierre Montreuil (eds.): Advances in Romance linguistics. Dordrecht: Foris Publications. Lu Fayan (601): Qie Yun (rhyming dictionary). Schuchardt, Hugo (1914): Die Sprache der Saramakkaneger in Surinam (= Verhandelingen der Koninklijke Akademie van Wetenschappen te Amsterdam, afdeling letterkunde, nieuwe reeks XIV, 6). Amsterdam: Johannes Müller. Schumann, Christian J. (1778): Saramaccan Deutsches Wörter-Buch. [MS, Moravian Brethren, Bambey, Surinam. Reproduced in Schuchardt 1914.] - (1783): Neger-Englisches Wörter-Buch. [MS., Moravian Brethren, Paramaribo, Surinam. Reproduced in: Andre A. Kramp (1983): Early Creole lexicography: A study of C. L. Schumann's manuscript dictionary of Sranan. Unpublished D.Litt. dissertation, State University of Leiden.] Sebba, Mark (1982): Metathesis, epenthesis and liquids in Surinam Creole. - In: Amsterdam Creole Studies 4: 23-37. [Norval Smith (ed.).] Selkirk, Elisabeth O. (1984): On the major class features and syllable theory. - In Mark Aronoff and Richard T. Oehrle (eds.), Language sound structure, 107-136. Cambridge, Mass.: MIT Press. Smith, Norval S.H. (1977): The development of the liquids in the Surinam Creoles. - In: Amsterdam Creole Studies 1: 31-54. - (1987): The genesis of the Creole languages of Surinam. [Unpublished D.Litt. dissertation, University of Amsterdam.] - (1993): Optional liquids in Kwinti: A link with East Surinam Maroon Creoles? [Lecture, SPCL conference, Amsterdam, June 1993.] - (1999): A preliminary account of some aspects of Leurbost Gaelic syllable structure. - In: Harry van der Hulst and Nancy Ritter (eds.): The syllable: Views and facts, 577-630. Berlin: Mouton de Gruyter. -

(2000): Lenition: the search for a relevant hierarchical framework. [LOT special session on Segmental Markedness and Syllabic Position, at the 8th Manchester Phonology Meeting, University of Manchester, 19/5/00.] - (2003): Evidence for recursive syllable structures in Aluku and Sranan. - In: Dany Adone (ed.): Recent developments in Creole studies. Tübingen, Niemeyer Verlag. Smith, Norval S.H. and Lilian Adamson (1996): Tonal phenomena in Sranan (with L. Adamson). [Lecture, Westminster Creolistics Workshop, University of Westminster. 29-31/3/96.] Speas, Margaret J. (1990): Phrase structure in natural language. Dordrecht: Kluwer Academic Publishers. Unknown (n.d.): Joint Leiden University-Summer Institute of Linguistics-Surinam Ministry of Education Surinam Creole Etymological Dictionary Project. Van den Berg, Margot (2000): Revised text of the 1762 treaty between the Dutch and the Saramaccans. [MS, University of Amsterdam.] Voorhoeve, Jan (1961): Le ton et la grammaire dans le Saramaccan. - In: Word 15: 436-445. Wang, Jialing and Norval S.H. Smith (1997): Introduction. - In: Wang Jialing and Norval S.H. Smith (eds.): Studies in Chinese phonology, 1-11. Berlin: Mouton de Gruyter.

Norval Smith, Dept. of Theoretical Linguistics/Amsterdam Centre for Language and Communication, University of Amsterdam, Spuistraat 210, 1012 VT Amsterdam, The Netherlands, Email: [email protected]

Emmanuel

Schang

Syllabic structure and creolization in Saotomense

1. Introduction

This paper presents data from two related Portuguese-based Creoles from Säo Τοηιέ Island, called Saotomense and Angolar. Its goal is to discuss the origin of two phenomena of Saotomense regarding the syllable shape: complex onsets (section 4) and the incorporation of liquid consonants (section 5). Before investigating these points, I present briefly the history of Säo Tome Island (section 2) and discuss the methodological problems (section 3) involved in the investigation of the origins of the syllabic structure of Saotomense. I conclude that the examination of the facts of Saotomense suggests that innovations in creolization are often hard to separate from subsequent diachronic developments.

2. A brief historical survey

Saotomense is the main Creole language spoken on Säo Tome Island (Republica Democrätica de Säo Τοηιέ e Principe) in the Gulf of Guinea.1 In addition to Saotomense, also called Forro, another Creole, Angolar, is spoken on the island by the fishermen of several 'praias' (beaches). The Angolars are former maroons gathered in the mountain, the Pico. In a nutshell, the Portuguese found the uninhabited islands of the Gulf of Guinea (Säo Τοηιέ, Principe, Annobön) around 1470. They settled them with prisoners, farmers, soldiers from Portugal and slaves from the Benin and Manicongo kingdoms on the mainland. The settlement began in 1485 and we have clues of existing slave rebellions from 1510. It is now widely acknowledged that the first settlement of Säo Τοηιέ was constituted by incomers from the Kwa/Edo-speaking area and in a second stage, Bantu speakers (mainly from Angola) formed the majority. It is important to notice that until recently the growth of the population of the island has not been due to birth but due to the influx of new people from outside (Caldeira 1999). The four Creole languages of the gulf of Guinea (GG), Fa d'Ambo (Annobön), Lung'ie (Principe), Saotomense and Angolar are closely related even if they are not mutually intelligible. The close relationship between these creole languages has been questioned and several hypothesis have been proposed (Lorenzino 1998, Ferraz 1976, Maurer 1992, Schang 2000, Roug6 and Schang 2001, Hagemeijer and Parkwall 2001). All converge on

1

I would like to thank the editor, an anonymous reviewer and the team 'contact de langues' of LLACAN for their comments on a previous version of this paper and especially Jean-Louis Roug6 for his comments and examples.

110

Emmanuel Schang

attributing to the Kwa-Edo languages a great importance in the birth of the Gulf of Guinea Creoles. However, the question of the nature of the link between Saotomense and Angolar remains uncertain. Are the common points between Saotomense and Angolar the result of the recent influence of Saotomense on Angolar or the result of the divergence from a common ancestral language? We know from historical works (see Caldeira 1999 for a synthesis) that the plantations of the island were quite autonomous and constituted small baronies during the first centuries and that the differences between the life in the city (Povoa^äo, i.e. Säo Tome City) and in the plantations were immense, to the effect that the contact between slaves, freedmen and Portuguese was quite limited for large parts of the population. Unfortunately, there is no source of Early Saotomense we can rely on. Unlike for Sranan (see, for example, Alber and Plag 2001 on syllable structure) it is impossible to study early texts. Schuchardt (1882) are the oldest reliable sources and the language described therein looks basically the same as Modern Saotomense. So, we can only try to reconstruct what happened from the beginning to the 19th century (and present) situation. In the 19th century, the Bantu workers developed several languages in the plantations 2 (for instance Lingwa Monte Cafe or Fa Uba Budo, two different languages) known under the generic name 'Tonga', but they did not develop a creole language. This situation (long after the genesis of Saotomense) provides us a good example of the diversity of languages that can develop in a plantation context. In the first decades of the settlement of the island, various varieties seem to have been created on the different plantations. These varieties progressively disappeared as the Creole (Early Saotomense), being the language of the freedmen, became the dominant language (besides Portuguese). This Early Saotomense and the varieties that maintained in the various plantations is what we called in Rouge and Schang (2001) the pre-creole state (PCS from now on). PCS is highly parallel to what is called Tonga nowadays, not in its structure, but in the fact that it incorporates various linguistic entities: the ancestor of Modern Saotomense and a range of local varieties that eventually disappeared. The PCS is a theoretical construct allowing us to distinguish Modern Saotomense from the linguistic varieties of the initial periods of language contact. It is a useful tool because it embraces the different varieties of the island at a crucial moment: the beginning of the settlement.

3. M e t h o d o l o g i c a l p r o b l e m s

All the clues we can gather on the history of Säo Tome Island converge on the claim that a creole language was quickly established on the island (in the thirty first years). This quick process of formation is different in its very nature from the normal evolution of languages. Usually, it takes many more decades (see Ruhlen 1994) for a language to diverge from its source. Manessy (1995) insists on the difference between creole language genesis and the 'normal' evolution of language, focusing on the concept of autonomy of the Creole towards the lexifier language. Once established, Saotomense had its own way of evolution. How2

Mainly bantu-based languages. See Rouge (1992) for a description of these languages.

Syllabic structure and creolization in Saotomense

111

ever, the process of constitution itself (i.e. genesis) of the new language must be distinguished from its subsequent independent evolution. If the reader allows me a comparison with syntax, I think that we cannot envisage the genesis of Saotomense neither via radical relexification (i.e. relexification as the unique process of creation of the syntactic features of Saotomense) nor via diachronic/dialectal evolution (as in Chaudenson 1992) as the only phenomenon giving birth to Saotomense. These considerations about syntax (Schang 2000) parallel those about phonology: I think that the explanation of the present form of Saotomense words mixes necessarily historical evolution and constraints on the incorporation of Portuguese words. But then, it is sometimes difficult to separate 'normal' evolution from incorporation during creolization. If we take history seriously, we must admit that the great number of adult Bantu slaves arriving after the creation of the Saotomense, and then after the PCS, might have change the language. There is no reason to think that once the Saotomense was created, it was impervious to further influences of L2 learning. For the new incomers, Saotomense was the target-language, the language of the local dominant class (except for the minority of Portuguese speakers, see Caldeira 1999 for an explanation of the social situation). Logically, we might have different processes for the Portuguese words incorporation in Saotomense: - words incorporated by the creators of Saotomense since the beginning (early 1500). - words incorporated by Saotomense speakers (Saotomense as L I ) . - words incorporated by Bantu or Kwa speaking slaves (Saotomense as L2). - other possibilities (one can hope it represents only marginal occurrences). In this regard, complex onsets in Saotomense and liquid incorporation presented in the following sections illustrate the complexity of the incorporation of Portuguese words in Saotomense.

4. W h e r e d o c o m p l e x o n s e t s c o m e f r o m ?

We know that Saotomense is five centuries old and that the socio-historical situation is very complex. But despite all this complexity, we can find a great homogeneity in the syllabic structure of Saotomense words. Like many other Creole languages, Saotomense and Angolar are languages with open syllables only (see Lorenzino 1998 or Schang 2000 for an argumentation). The main difference between Saotomense and Angolar in the syllable shape is that Saotomense allows complex onsets while this is quite rare in Angolar. This phenomenon is highlighted in Lorenzino (1998:85), as illustrated in the following examples:

Emmanuel Schang

112

(1) Portuguese trabalhar desgraga vermelho pretu

Angolar

Saotomense

gloss

taba disiga0a veme peetu

tlaba di3glasa vleme pletu

'to work' 'misfortune' 'red' 'black'

We can find complex onsets in Saotomense in monosyllabic words, and in word initial and other positions in polysyllabic words, as illustrated in (2):3 fka 'mojka di'Jtinu li'3boa '/tlada 'Jkola Jtlo'visu

T M A marker (tense+aspect) 'fly' 'fate, destiny' 'Lisbon' 'road, street'. 'school' 'work'

A first explanation of this fact is that Saotomense preserves the Portuguese etymon's consonant cluster. Nevertheless, if we look in details at the Saotomense words, we can find another explanation that shows that some complex onsets might have another origin. If w e compare Saotomense words with their correspondent in Angolar, we can see that epenthesis is used in Angolar to break up the consonant cluster:

(3) Portuguese Lisboa escola

Angolar

gloss

liziboa sikola

'Lisbon' 'school'

In Angolar, epenthesis is preferred. One can say that this is a major difference between Saotomense and Angolar. But there are clues that this has not always been the case. There is a regular correspondence between Angolar and Saotomense. Saotomense [J] often corresponds to Angolar [si]:

(4) Portuguese

3

Saotomense

Angolar

gloss

pescador

pijkado

pisikado

'fisherman'

casca mosquito escrever estaca

kajka mijkitu Jkleve Jtaka

kasika nsikitu / misikitu sikeve sitaka

'peel' 'mosquito' 'to write' 'stake'

Words are presented with an accent mark, when necessary. I assume that Saotomense (and Angolar) are accent languages. This assumption is contested in several works (Maurer 1992 and Lorenzino 1998 for instance). But in the absence of minimal pairs and of any kind of exploration of the tone system, I assume with Ferraz (1979) that Saotomense and Angolar are free-pitch accent languages (as proposed in Ferraz and Trail 1981 for Principe Creole).

Syllabic structure and creolization in Saotomense

113

Epenthesis is a regular process of preservation of the Portuguese consonants in Angolar, except for liquids, where deletion is preferred (a problem that will be addressed in section 5). We can reasonably think that Angolar and Saotomense had two different strategies to incorporate the Portuguese words. But if we look carefully, we can find examples in Saotomense and Angolar that do not conform to this prediction: (5) Portuguese espelho esperto vespera

Saotomense supe supetu besupla

Angolar 9upe Gupetu be0upa

gloss 'mirror' 'clever' 'previous day'

To conform with the predictions I have made just before, esperto 'clever, smart' ; espelho 'mirror' and vispera 'the previous day' should have been incorporated as [Jpetu]; [Jpeju] and [bejpela] in Saotomense. However, it is not the case and one can wonder why. There is no reason to think that there is 'something' in the Portuguese words in (5) that prohibits the preservation of [f] in Saotomense. Recall that in Modern Portuguese espelho is pronounced [appB/Cu] but this has not always been the case (Teyssier 1980), and is still not the case in Brazilian Portuguese (where the graph before consonant is pronounced M). Hence, as claimed in Roug6 and Schang (2001) with other arguments, the Portuguese words in (4) and (5) may have been incorporated in Saotomense in a previous stage (the PCS) with an epenthesis, and Angolar conserved this form. Now the explanation is rather clear, the string [si] of the PCS words (created via epenthesis as in the present Angolar form) was reinterpreted in [J] in Modern Saotomense, probably under Portuguese influence. The examples in (5) do not conform to this pattern since the epenthetic vowel is not [i] but [u] (probably a case of labial attraction due to the final consonant). At least one example illustrates the regular change [si] > [J] even if the Portuguese etymon contained [si] and not [J]:Port.: cirurgiäo 'surgeon' > [Jtli'30] 'surgeon/doctor' Another argument that completes this point is the regular tendency to palatalization of the consonant before a front vowel in Saotomense: Port: passear > paxa or [pajja]; Port: cinza > [Jinza] or [Jinja], Of course, recent Portuguese words can now be incorporated directly with [J/3] in Ci and [-continuant] in C 2 in a complex onset as it is a authorized sequence. It can be claimed, then, that this evolution is not directly the result of the conformity to Portuguese, a symbol of decreolization, but rather as a sign of autonomous evolution of Saotomense from the PCS. We have seen that once Saotomense was created, it follows its own diachronic development.

114

Emmanuel

Schang

5. The liquids in Saotomense

The incorporation of Portuguese liquids into Saotomense is a complex phenomenon. Saotomense retained only the liquid /l/ as did Fa d'Ambu and Angolar. In contrast, Lung'ie has both Irl and III. As Ferraz (1987:288) explained: "In the major area of liquid clusters, it is Saotomense (Saotomense) that is more marked than the other three [Gulf of Guinea Portuguese Creoles], as it shows by far the largest number of such clusters. With these exceptions, the four Creoles all have open syllables; since /1C/ would close the syllable, Saotomense has the sequence/Cl/ only, keeping the syllable open." The three Portuguese liquids [r, 1, κ] are retained (where retained) as /l/ in the Saotomense words. Let us look in detail at what happens to Portuguese liquids. The first option is that they deleted, as illustrated with the examples in (6): Portuguese

Pedro padre mostrar Alfredo quintal barriga claridade marca barba orvalho

Saotomense 'ρε du 'pade mu'sa (also [mo'Jtla]) 'fedu kin'te 'bega ke'dad3i 'meka 'beba love

gloss 'Peter' 'priest' 'to show' 'Alfred' 'garden' 'stomach' 'clarity' 'mark' 'beard' 'dew'

We find here various cases: deletion occurs when the liquid is in intervocalic position {barriga), when the liquid is in coda position (quintal) or in a consonant cluster {padre). The second possibility is that the Portuguese liquid is preserved, in which case it can be preserved: - in the same position as that of the Portuguese word - in a different position, via: a. epenthesis of a vowel b. metathesis The following examples illustrate the three environments:

Syllabic structure and creolization in Saotomense

115

Liquid preserved in the same: position as in the Portuguese etymon Portuguese brasa crescente febre outro encontrar estrada igreja flecha gritar atrds escola arroz bule

Saotomense 'blasa kle'set/i feble (also [flebe]) 'otlo 'kön'tla 'Jtlada 'gleza 'fle/a gli'ta •tlaji 'Jkola 'loso buli

gloss 'ember' 'crescent' 'fever' 'another' 'to find' 'road' 'church' 'arrow' 'to shout' 'behind' 'school' 'rice' 'teapot'

The generalization that emerges from (7) is that if liquids are preserved they occur in complex or simplex onsets. As can be seen by comparison with the deletion cases in (6), being in onset position is a necessary, but not a sufficient condition for the survival of the liquid in situ. Liquids in word-final position trigger epenthesis, to the effect that the preserved liquid ends up in onset position: (8)

Liquid preserved in word final position Portuguese doutor quintal sinal flor

Saotomense dotolo kintali (also [kinte]) Ji'nali floli

gloss 'doctor' 'garden' 'sign' 'flower'

In consonant clusters, we often find epenthesis (within an across syllables): (9)

Liquid preserved in epenthesized clusters Portuguese crer cru alma

Saotomense ke'le ku'lu 'zalima

gloss 'to believe' 'raw' 'soul'

Metathesis is avoided in these cases. For epenthesis, recall that it was a regular phenomenon (cf. § 4) in Early Saotomense and it is still widely attested in Angolar to break consonant clusters. Intervocalic / is allowed in the Gulf of Guinea Creoles so that epenthesis makes an excellent repair for many more words.

116

Emmanuel Schang

Finally, we turn to cases of metathesis. In the examples of metathesis discussed below, I will assume that for verbs the final r of the infinitive form is the source of the metathesized liquid. Metathesis involving the final r of the infinitive form is not completely regular (for instance, we have kume < comer "to eat" and not klume). Hence, in such cases the third person singular form of the present tense or preterit (preterito perfeito) are the most likely source forms. Three different cases of metathesis can be distinguished. First, there is metathesis with shift of the liquid from coda to onset within a syllable: (10) Tautosyllabic coda-onset metathesis Portuguese carväo falcäo calcanhar cercar pulga vergar vergonha vermelho torcer magro servigo erder orvalho culpa carpinteiro lagartixa bordar jardim cirurgiäo (surgeon) pulmäo aborrecer barbeiro barquinho calgao calmaria carregar carregador corpo forquinha/forquilha folga firmar

Saotomense kla'vö fla'kö klakajiö Jtle'ka 'pluga vle'ga vle'gojia vle'me tlo'se mla'gu Jli'visu (also [ftlo'visu]) le'da Ιο'νε 'klupa klapTtelu (also [kapin'telu]) laga'tlisa bo'dla 3a'dn /tli'30 plu'mö blo'se bla'belu bla'kl kla'sö klama'ja kla'ga klaga'do 'klopo flo'kjä 'floga 'flima

gloss 'coal' 'falcon' 'heel' 'to surround' 'flea' 'to bend' 'shame' 'red' 'to twist' 'thin' 'work' 'to burn' 'dew' 'fault' 'carpenter' 'lizard' 'to embroider' 'garden' 'sorcerer' 'lung' 'to annoy' 'barber' 'boat' 'trousers' 'calm' 'to load' 'porter' 'body' 'little pitchfork' 'game, joke' 'to stop, to sign'

Syllabic structure and creolization in Saotomense

Portuguese formiga subir salvar ferver

Saotomense flo'miga su'bli Jtla'va fle'be

117 gloss 'ant' 'to go up' 'to save' 'to boil'

Second, there is metathesis with shift of the liquid from the coda of one syllable to the onset of the following syllable, illustrated in (11): (11) Cross-syllabic coda-onset metathesis Portuguese lagartixa bor dar jardim artemisia cirurgiäo

Saotomense laga'tlisa bo'dla 3a'dfi atli'mi3a Jtli'30

gloss 'lizard' 'to embroider' 'garden' 'artemisia' (plant species) 'surgeon'

Third, there are miscellaneous other cases, including intriguing instances of long-distance metathesis: (12) Other cases of metathesis Portuguese conversar degredo magro vidro segredo navegar febre dentro

Saotomense klöve'sa dle'gedu 'mlagu 'vlidu Jtle'gedu navle'ga 'flebe (also [Teble]) 'dlentu (also [ 'nglentu])

gloss 'to converse 'exile' 'thin' 'glass' 'secret' 'to sail' 'fever' 'into'

Another example of this type is the strange form flomenta4 'to massage' coming from a old Portuguese word fomentu 'cataplasm', via the infinitive morpheme of the verb fomentar (which has a different sense nowadays). We can find at least a Bantu word which presents metathesis: klisenge 'axe' that comes from the Kikongo word ki-sengele. In terms of number of types affected by deletion, epenthesis and metathesis, we can observe that words with metathesis are not more numerous than words from other categories, but they are present in significant number. They are presented here in greater numbers because they seem to represent a more marked way of syllable structure adaptation than the more wide-spread processes of epenthesis and deletion.

4

Many thanks to J-L Rougd, who sent me these (and other) examples, taken from his forthcoming etymological dictionary of the Portuguese Atlantic Creoles.

118

Emmanuel Schang

The examples given before illustrate the fact that the incorporation of the Saotomense words is a rather complex phenomenon, as four strategies apply: metathesis (with three distinct cases), epenthesis, deletion and preservation. But it is probably even more complex if we consider the possible evolution of words like [bega] and [meka]. As pointed out in Rouge and Schang (2001), some examples might presuppose a former step *baliga > bega , *malika > meka. This is confirmed by the pair kinte / kintali, where the second was found in a typical njambi (voodoo) song. Kintali is considered by my informant as the more ancient form. Kinte is then probably a more recent form, not directly derived from the Portuguese quintal but from the Early Saotomense form kintali (the one still maintained in voodoo songs). The segments 1/r were not (not always) retained in an intervocalic context (mainly between /a/ and /if), a phenomenon already noticed in Lorenzino (1998:86). We can here distinguish two steps in the history of these words: incorporation with epenthesis and evolution with deletion of the intervocalic liquid. Blevins and Garrett (1998) analysed long-distance metathesis in the Le Havre dialect of French and claimed that: "[...] long-distance movement processes evidently move segments or features into syllables which are more prominent: stress syllables, as in Colville; or initial syllables as with Bagneres-de-Luchon French r and Ancient Greek h. [...] This perceptual reinterpretation involves segments and features which are drawn to positions of prosodic prominence."

What is surprising with Saotomense however is that we do not find an attraction of the liquid towards k as in Le Havre French metathesis (Latin: kämera > *kambra > crambo 'room'). As we can see from the examples in (12), even nasals precede the liquid. And even more surprising is the fact that the liquid is moved to positions that are not prominent and not word-initial (as in [navle'ga] 'to sail'). Putting aside [navle'ga], the Saotomense data are generally in accordance with Blevins and Garrett's theory because it seems clear that mainly two places are offered for long-distance metathesis: the word-initial syllable or a stressed syllable.

6. C o n c l u s i o n

The nature of the processes involved in creolization is an important issue in Creole studies. Two major kinds of explanation are usually presented to account for creolization: substratists claim that Creoles preserve the substrate language structure while universalists claim that the Creole languages display an unmarked (universal) state of affairs. Creole languages are rarely analysed per se, as languages able to develop without reference to the base-language. Of course, the reason is obvious: the base-language (here the Portuguese) represents a model for a part of the population. However, for older Creole languages like Saotomense, autonomous evolution should be taken in account. In this paper I provided examples that do not fit perfectly with the view that the Creole forms can be analysed as a direct mapping of input (Portuguese words) and output (Saotomense words). I argued for an explanation that takes into account the diachrony.

Syllabic structure and creolization in Saotomense

119

Thus, the behavior of onset clusters cannot be explained without reference to processes that must have happened after creolization. Furthermore I have shown that a variety of processes (preservation, deletion, vowel epenthesis and three different metatheses) apply to the Portuguese liquid in its incorporation into Saotomense. These various processes are consistent with the hypothesis of the PCS, i.e. the diversity in the incorporation processes matches the diversity of influences in the genesis of the creole language . Of course, it is difficult to draw the line between creolization (as an abrupt process of incorporation of words) and subsequent language change. But there is no reason to believe that creole studies should be simpler than studies of non-creole languages.

References

Alber, Birgit and Ingo Plag (2001): Epenthesis, deletion and the emergence of the optimal syllable in creole. - In: Lingua 111,811-840. Blevins, Juliette and Andrew Garrett (1998): The origins of consonant-vowel metathesis. - In: Language 74, 3, 508-556. Caldeira, Arlindo Manuel (1999): Mulheres, sexualidade e casamento em Säo Tome e Principe. Lisboa: Cosmos. Chaudenson, Robert (1992): Des lies, des langues, des hommes. Paris: 'Harmattan Ferraz, Luis Ivens (1976): The origin and development of four Creoles in the Gulf of Guinea. - In: African Studies 35.1.76. - (1979): The Creole of Säo Tome. Johannesburg: Witwatersrand University Press. - (1987): The liquid in the Gulf of Guinea Creoles. - In: African Studies 46.2.87, 287-295. Ferraz, Luiz Ivens and Anthony Trail (1981): The Interpretation of one in Principense Creole. - In: Studies in African Linguistics 12 (2). Hagemeijer, Tjerk and Mikael Parkvall (2001): The origin of the Gulf of Guinea Creoles. Paper presented at the ACBLPE Congress, Coimbra, June 2001. Lorenzino, Gerardo (1998): The Angolar Creole Portuguese of Säo Tome. Lincom Europa. Manessy, Gabriel (1995): Creoles, pidgins, Varietes vehiculaires. Proces et genese. CNRS editions. Maurer, Philippe (1992): L'apport lexical bantou en angolar. - In: Afrikanistische Arbeitspapiere 29, 163-174. Rouge, Jean-Louis (forthcoming): Dictionnaire etymologique des Creolesportugais d'Afrique. - (1992): Les langues des Tongas. Adas do Coloquio sobre crioulos de base lexical portuguesa. Lisbon: Colibri. 171-176 Rouge, Jean-Louis and Emmanuel Schang (2001): Elements de comparaison de la lunga ngola et de la lungwa santome: sur la piste d'un pre-criole. Paper presented at the ACBLPE Congress, Coimbra, June 2001. Ruhlen, Merrit (1994): The origins of language. Tracing the Evolution of the Mother Tongue. (1997) for the translation in French. Schang, Emmanuel (2000): L 'emergence des Creoles portugais du golfe de Guinee. Lille: Les Presses du Septentrion.

120

Emmanuel

Schang

Schuchardt, Hugo (1882): Kreolische Studien I. Über das Negerportugiesische von S. Thome (Westafrika), Sitzungsberichte der kaiserlichen Akademie der Wissenschaften zu Wien 101 (2), 889917. Teyssier, Paul (1980) Histoire de la langueportugaise. P.U.F.: collection 'Que-sais-je ?'

Emmanuel Schang, Universite d'Orleans, Faculte des Lettres, Langues et Sciences Humaines 10, rue de Tours, 45072 Orleans Cedex 2, France, Email: Emmanuel. [email protected]

Section 2: Stress, tone and intonation

Anne-Marie Brousseau The accentual system of Haitian Creole: The role of transfer and markedness values

1. Introduction 1

Most of the papers in this volume offer completed analyses of phonological or morphological phenomena in Creole languages. My contribution is somewhat different: it proposes some key elements of a research program rather than an actual analysis. The paper provides a comparison of the stress system of Haitian with those of its source languages: the tone system of Fongbe and the stress system of French. It is a first step toward the formulation of a scenario that would explain how the stress system of Haitian has evolved from the phrasal stress system of French on the one hand, and from the tonal system of Fongbe (and other Gbe lects) on the other. This scenario would identify the respective role of the main substrate language(s) and of Universal Grammar in the genesis of a radical Creole such as Haitian. The basic assumption behind such a scenario is that the emergence of Haitian is a case of second language acquisition (cf. Anderson 1983). This is consistent with Singler's convincing demonstration that only adults could have been the architects of the Creole in Haiti (Singler 1993). This is also in line with Valdman (1983) who defines creolization "as the crystallization and subsequent elaboration of an autonomous interlanguage system" (p. 214). The phonological systems of Haitian appear as a compromise between those of its contributing languages. While Haitian has the same vowel inventory as Fongbe, it has a consonantal inventory that is very similar to French. The syllable structures of Haitian are less complex than those of French (e.g. no consonant clusters in Haitian), but not to the point that they equate with the simplicity of those of Fongbe (e.g. closed syllables in Haitian). The accentual system is also a kind of hybrid: it is a stress system (like French), realized at the level of the phonological word (like Fongbe) and quantity sensitive (unlike French or Fongbe). This hybrid character of Haitian is similar to what is typically found in interlanguages, in the more straightforward cases of second language acquisition, especially for phonology and even more especially for prosody (cf. among others, Broselow 1983 for syllabic structures, Archibald 1998a,b and Devonish 1989 for stress). The similarities are too striking to constitute a mere coincidence. Wode (1986) proposes that transfer occurs in all contact situations, since it is a major component of language processing abilities. This entails that for any given pair of L1-L2, the same transfer types will be

1

I thank the participants of the workshop for stimulating discussion over the first draft of this paper. I thank also the anonymous reviewers who provided me with very keen and cogent advice on some theoretical aspects of the paper. Of course, any shortcomings of interpretation are my sole responsibility.

124

Anne-Marie

Brousseau

found across age groups as across contact situations. Examining the way English consonants and vowels are realized in various contact situations (such as naturalistic L2 acquisition, tutored L2 acquisition, relearning of a language, and pidgins), Wode shows that the substitution patterns are the same in all situations. Therefore, the transfer regularities may be explained by the only set of factors invariant enough: the language background of the speaker. Different speakers who share a common language background use the same substitution patterns in different situations. If a Creole equates to an interlanguage, the next step is to determine what are the mechanisms that set the grammar of the Creole/interlanguage apart from both the substrate/first language and the superstrate/second language. From a creolistic standpoint, and from a second language acquisition standpoint, these mechanisms may involve transfer from the substrate/Ll and general properties determined by Universal Grammar. Of course, there is an ongoing debate in both fields (there again, mere coincidence?) as to whether a sensible scenario of Creole/interlanguage development can be explained mostly, even solely, by transfer or universals. I will adopt an intermediate view, which combines both (cf. Alber and Plag 2001, Uffmann, this volume for a similar view on the phonology of Creoles). From the standpoint of second language acquisition, I will adopt the "Full Transfer/Full Access Mode Hypothesis" (see section 2.1). From a creolistic point of view, I will adopt the "Transfer to Somewhere Principle" of Andersen (1983) and I will follow Mufwene (1991), extending to phonological strategies what he proposes for morphosyntactic strategies: A grammatical form will occur consistently and to a significant extent in the interlanguage as a result of transfer if and only if (1) natural acquisition principles are consistent with the LI structure or (2) there already exists within the L2 input the potential for (mis-)generalization from the input to produce the same form or structure. (Andersen 1983: 182). [Pidgins and Creoles] seem to inherit their morphosyntactic strategies from some of the languages of the contact situation, a formative strategy predicted by USLAH [Universals of second language acquisition hypothesis]. In this, case 'universals' may be interpreted as the regulatory principles that determine which grammatical features of the lexifier or of some other languages of the contact situation will be selected by a [Pidgin or Creole], (Mufwene 1991:137)

The paper is organized as follows. Section 2 provides the theoretical background on second language acquisition (henceforth SLA) and on markedness conditions. Section 3 presents the set of parameters relevant to accentual systems. Section 4 presents the accentual systems of Haitian, French and Fongbe, and characterizes these systems in terms of parameter sets. These parameter sets are compared in section 5 and their respective markedness values are evaluated, in order to account for the hybrid system of Haitian. Section 6 provides a brief conclusion.

The accentual system of Haitian Creole

125

2. Theoretical a s s u m p t i o n s

2.1. Second language acquisition Theories that aim at explaining the development of an interlanguage in second language acquisition may be subsumed under two well-known labels: the contrastive analysis (or Full Transfer Hypothesis) and the L2=L1 hypothesis (or Full Access Mode Hypothesis). The first attributes to transfer from LI the bulk of the differences between the interlanguage and the target language, while the second ascribes these differences to universals involved in both LI and L2 acquisition. The contrastive analysis has been criticized and plainly rejected by some, refined and defended by others. In any case, the hypothesis accounts for a large part of the phonological evidence (cf. for instance Altenberg and Vago 1987, Broselow 1983, Ioup 1984). The L2=L1 hypothesis states that "errors" in L2 production parallel the developmental errors of children in LI acquisition contexts. In a markedness perspective, this amounts to saying that marked properties will be more difficult to acquire. Many studies on L2 acquisition (cf. Flynn 1987, Flynn et al. 1993, White 1989, Youssef and Mazurkewich 1998 among others) have shown the validity of this hypothesis, particularly for syntax, while others (cf. for instance Clahsen 1988, Clahsen and Muysken 1989) have argued that L2 acquisition by adults is governed by principles independent of Universal Grammar (henceforth UG). Overall, it seems that while LI properties hinder the acquisition of features of the target language in phonology, the LI has a more moderate influence in the domain of syntax. Recent developments in SLA theory point to a reconciliation of these two opposing views. Most scholars (cf. Hyltenstam 1987 and the references therein, Wode 1986) would now agree on the following characterization of interlanguages. First, interlanguages are natural languages: they follow principles and constraints found in natural languages, and they reflect language universals. Second, the typological nature of both LI and L2 determines the structural character of the interlanguage. Third, transfer is an integral part of second language learning and it is constrained by formal properties of LI and L2. In other words, universals act as guiding principles in the process of developing an interlanguage, which also has recourse to the grammar of LI. This reconciled view is best expressed by the Full Transfer/Full Access Mode Hypothesis (Schwartz and Spouse 1996, White 2000). According to this hypothesis, second language learners assume the grammar of LI as a starting point (Full Transfer) and progressively restructure this initial grammar using principles and parameters of UG (Full Access). Learners maintain the parameter settings of LI in the interlanguage unless they are exposed to evidence to the contrary: positive evidence in the target language triggers the resetting of the parameters in the interlanguage. In this view, an interlanguage is the "combination of (incorrect) L1 parameter settings, and the effect of Universal Grammar manifested in parameters that have been reset to the (correct) L2 value" (Archibald 1998a: 283). Since learners have full access to UG, the parallel developments between LI and L2 are accounted for: UG guides both types of acquisition in a similar fashion. But at the same time, LI and L2 acquisition show many differences in their respective developments: the principles of UG do not allow L2 acquisition to develop as efficiently for adults as for children.

126

Anne-Marie

Brousseau

This can be attributed to the fact that L2 acquisition involves parameter resetting while LI involves parameter setting (cf. Flynn 1987, Flynn and al. 1993, Youssef and Mazurkewich 1998). Since the Full Transfer/Full Access Mode Hypothesis has proven very successful in accounting for phonological aspects of interlanguages, and especially for prosody (cf. Archibald 1998a, 1998b, Broselow and Finer 1991, Youssef and Mazurkewich 1998), I will assume it without further discussion.

2.2. Universals and Markedness In the generative tradition, the notion of markedness is dependent upon the notion of Universal Grammar. The grammar of a particular language is a manifestation of UG: it is the result of fixing the values - marked or unmarked - of parameters left open in UG. The generative view encompasses a strong formulation of the notion of markedness, based on implicational relationships (cf Greenberg 1966). When the existence of a category A implies the existence of a category B, A is more marked than Β (e.g. if a language has voiced stops, it also has voiceless stops). This implicational relationship is the basic notion behind the Subset principle of Wexler and Manzini (1987): when the two values of a parameter generate structures in a subset/superset relation, the default value corresponds to the subset setting. In other words, general options are less marked than restrictive ones. Parameter values that conform to the Subset principle provide the most robust cases for markedness values. To determine the values of parameters that do not involve subset relations, a weaker formulation of markedness is required; one based on the frequency criterion (cf. again Greenberg 1966). The default value of a parameter is the one that corresponds to the most frequent option (within and/or across languages). This weaker notion is necessary for all parameters referring to directionality, many of which define the grammar of accentual phonology. Markedness conditions act as guiding principles in SLA: they determine the order in which structures are acquired and they constrain the transfer from LI. Hyltenstam (1987) shows that in the early stages of acquisition, the interlanguage is immune to LI and L2 influences: unmarked categories are preferred. This explains the similarities found in all L1-L2 situations. In later stages, differences arise, which reflect the various interactions between the markedness values of the two languages. Thus, Hyltenstam (1987) makes the following predictions for a given parameter or set of parameters: 1) when both LI and L2 are unmarked, the acquisition is straightforward; 2) when LI is unmarked and L2 is marked, the unmarked category is transferred and, often, will remain for a while as a feature of the interlanguage; 3) when LI is marked and L2 is unmarked, transfer is not likely to occur and, if it does, the marked category will not remain; 4) when both LI and L2 are marked, the unmarked value can occur, but will be abandoned rapidly in favor of the marked value. These predictions reflect a widespread point of view in SLA and markedness research. For instance, according to Eckman (1977) and Kellerman (1987), learners transfer in L2 only those structures of LI that are unmarked or relatively less marked. Markedness conditions also guide the process of parameter resetting, given the kind of evidence needed to reset the value of a parameter. It is easier for the learner to reset a value from the subset (unmarked value) to the superset (marked value) than vice-versa, since only

The accentual system of Haitian Creole

127

positive evidence is needed in this case (Dresher and Kaye 1990, White 1989). Thus, when LI is unmarked and L2 is marked for a given parameter, the interlanguage can show one value or the other: either the unmarked value transferred from LI remains, or the marked value of L2 is adopted, following parameter resetting.

3. Metrical parameters and markedness values

Dresher and Kaye (1990) have proposed a precise formulation of the metrical parameters needed to account for stress systems across languages, including their markedness values. From the set they propose, seven parameters (PI to P8A) are needed to characterize the stress systems of Haitian and French. The last parameter (P8) is taken from Hayes (1995). (Default values appear in boldface throughout the paper.) (1)

Metrical parameters and default values PI: The word tree is strong on the [Left/Right] P2: Feet are [Binary/Unbounded] P3: Feet are built from the [Left/Right] P4: Feet are strong on the [Left/Right] P5: Feet are quantity-sensitive [Yes/No] P6: Feet are quantity-sensitive to the [Nucleus/Rhyme] P8A: There is an extrametrical syllable [Yes/No] P8: The extrametrical syllable is on the [Left/Right]

The default value is straightforward in three cases, according to the Subset principle. By default, feet are unbounded (P2), quantity-insensitive (P5), and they do not dominate any extrametrical syllable (P8A). For directional parameters, the default is determined based on frequency. By default, the word tree is strong on the left (PI), as are the feet (P4), and foot construction starts on the left (P3). In systems showing extrametricality, the extrametrical syllable is on the right (P8). For systems where feet are quantity-sensitive, the default value of P6 is more controversial; Dresher and Kaye merely suggest that sensitivity to the nucleus is the default value. Note that the Subset principle does not provide a clear answer in this case. One well-known peculiarity of the French stress system is that all but the last words lose their stress when combined within phrases. To characterize such a system, two additional parameters are needed. I propose the following, based on discussions in Hayes (1995) and Halle and Vergnaud (1987). (2)

Other relevant parameters P10: Stress is [Persistent/Non-Persistent] PI 1: Destressing applies in the domain of [Word/Phrase],

128

Anne-Marie

Brousseau

According to Hayes (1995: 34), rules such as destressing (and stress shift) obey a universal pattern that tends to "make the strong constituents stronger and the weak constituents weaker". However, this pattern clashes with another universal principle, that of structure preservation, which tends to preserve the integrity of a structure in the course of derivation. Structure preservation has been independently shown to constrain markedness values of prosodic structures. For instance, Nespor and Vogel (1986) argue convincingly that resyllabification of structures in domains larger than words (non-persistent syllabification) is the marked option. Structure preservation thus identifies persistent stress as the default value for Ρ10. The Subset principle also identifies persistent stress as default since a grammar that allows non-persistent stress is more restrictive than a grammar that prevents it. The default value of PI 1 is determined based on frequency. Hayes (1995) notices that destressing occurs word-internally more often than phrase-internally. Halle and Vergnaud (1987) observe that individual word contours are generally not affected when words are grouped into phrases. To the best of my knowledge, there exist no well-defined set of parameters for tonal languages. However, many factors indicate that tone and stress systems show common properties determined by a shared set of suprasegmental principles (cf. Devonish 1989, RiveraCastillo 1998). First, there are numerous similarities between tone and stress systems. Second, languages with so-called pitch-accent systems show a clear interaction between stress and tone. Third, the hybrid system found in some Creoles (e.g. Djuka, Guyanese Creole, Papiamentu, Principense, Saramaccan) seem to have emerged from a reinterpretation of the stress system of the lexifier language based on the tonal system of substratum languages. This suggests that both types of system are ruled by the same general set of constraints and principles. Building on previous work by Goldsmith and Hyman on accentual systems, RiveraCastillo (1998) puts forth the basics of a unified accentual system. She suggests that tone and stress are manifestations of the same suprasegmental system; they correspond to suprasegmental features associated in various ways at different prosodic levels. Three types of languages can thus be distinguished depending on the prosodic domain relevant to the feature [Pitch] (i.e. the feature that is manifested in surface as tone): (3)

[Pitch] can be characterized as: 1. an edge phrasal level feature in intonational stress-accent languages, 2. a word level feature in pitch-accent languages, and 3. a head syllable level feature in tonal languages. (Rivera-Castillo 1998: 330)

Rivera-Castillo further suggests that the same typology could apply to stress, based on languages such as Mandarin (cf. Yip 1980), which exhibit stress distinctions at the phrase level only. Applied to stress, the statement in (3) would translate as follows: (4)

[Stress] can be characterized as: 1. an edge phrasal level feature in phrasal stress tone languages, 2. a word level feature in free (lexicalized) stress languages, and 3. a head syllable level feature in syllable-timed stress languages.

The accentual system of Haitian Creole

129

In line with this view, I will assume that languages all share a general accent system, where accent is to be understood as the way prominence is associated with given prosodic structures. This implies a unified set of principles and parameters that underlies all manifestations of accent, be it tone, stress or a combination thereof. The parameters needed to describe this unified accent system must minimally specify the following information. First, which features determine prominence: tone, stress or both? Second, what are the prosodic domains to which the prominent features are assigned: syllable, word or phrase? Third, what are the phonetic realizations of stress: pitch, intensity, length or a combination? For tone, the phonetic realization need not be specified in a parametric fashion; unlike stress, which shows various phonetic cues crosslinguistically, tone is universally manifested as pitch. What is required for tones is the specification of the various pitch contrasts (High, Low, Mid) and their possible combinations (register tones or not), as well as some cooccurrence constraints. The information specifically relevant to stress would be provided by the parameters listed in (1) above. The formulation of the actual parameters is a difficult task, which would require taking into account a huge sample of languages, sufficiently representative of the language accentual typology. In the following, I propose a first approximation of the set of parameters that will allow the comparison of Haitian with its contributing languages. The form of the parameters may not be the most adequate (should we allow, for instance, ternary values?) and the default values, if any, are tentative, hence the term pseudo-parameters. Furthermore, I will not consider the parameters specific to tones (pitch contrasts and cooccurrence), since they are not directly involved in the stress system of Haitian. (5)

Metrical pseudo-parameters PP1: Prominence is determined by [One/Two] feature(s) PP2: The feature Tone determines prominence [Yes/No] PP3: The feature Stress determines prominence [Yes/No] PP4: The feature Tone is assigned to the [Syllable/Word/Phrase] PP5: The feature Stress is assigned to the [Syllable/Word/Phrase] PP6: Stress is realized as [Pitch/Intensity/Vowel Length]

Given the Subset principle, the default value for PP1 is quite direct: languages that exhibit both tone and stress are more marked than languages exhibiting either one or the other. As for PP5 and PP6, I can see no principled reason to chose one value over the other as the default. The choice could only rely on an extensive calculation of frequency. The same situation holds for PP2 and PP3. 2 The default value of PP4 is based on indirect evidence but

2

However, the default values of PP2 and PP3 may be determined, at least tentatively, on indirect evidence. In cases of non-abrupt language change, accentual systems usually evolve from tone to stress, rather than from stress to tone. For instance, Carter (1983) claims that the change from a fully tonal to an intentional system, perhaps via pitch-accent, constitutes a typical developmental path for languages. This could suggest that, by default, Stress determines prominence. On the other hand, there is also evidence that change can proceed in the other direction. As an anonymous reader pointed out to me, the pitch-accent systems of some dialects of German and Dutch could only have evolved from an intonational system. And, again, I see no principled reason to assume that Stress is the default manifestation of prominence.

Anne-Marie Brousseau

130

intuitively, it seems correct. Pure tone languages, i.e. languages where accent is manifested solely as tone, tend to favor tone assignment at the level of the syllable. If we take pure tone languages as the basic tone system, it makes sense to identify the syllable as the default tone-bearing unit (PP4). In the following section, I characterize the accent systems of the three languages under comparison using the parameters and pseudo-parameters presented above.

4. T h e accentual s y s t e m s o f the three l a n g u a g e s

4.1. The data Due to the scarcity of historical data, and to the fact that early written texts provide no clues about the accentual phonology of the languages, the systems of Haitian Creole and of Fongbe were established on the basis of contemporary data. Haitian data was drawn from the available literature (d'Ans 1968, Cadely 1994, Hall 1953) and from my own analysis of the Lefebvre-Fournier corpus (1976). Gbe data relies on Capo's study of Gbe dialects (Capo 1991) and on the available literature addressing tones in Fongbe (Aboh 1999, Brousseau and Brillon 1990, Frechet 1990, Wieseman 1991). As we will see in § 4.2, heavy syllables in Haitian bear secondary stress. This feature is clearly manifested in the Lefebvre-Fournier corpus; it is also reported in Cadely's work, but not in the remaining literature dealing with stress in Haitian. 3 A possible explanation for this fact would be that the data showing secondary stress corresponds to a more acrolectal variety of Haitian, typical of speakers who also speak French. 4 If this were the case, we would expect the data to exhibit other phonological features typical of the acrolectal variety. According to Valdman (1991), phonological and morphological characteristics distinguish two main varieties of Haitian: a gallicized version of Haitian, the Haitian Creole of bilinguals, and a more basilectal/mesolectal variant, the Haitian Creole of monolinguals. The phonological characteristics are as follows (adapted from Valdman 1991: 79): (6)

Phonological variables a. front rounded vowels b. regressive nasalization of V c nasalization of final C d. postvocalic M e. s + C clusters

H C of YES NO NO YES YES -

bilinguals y, 0, oe e.g. [famij] e.g. [3äb] e.g. [pur] e.g. [spo]

H C of NO YES YES NO NO -

monolinguals i\u, e\o, ε\ο e.g. [fämij] e.g. [3am] e.g. [pu] e.g. [espo]

It is clear from the analyses and examples in Cadely (1994) that the variety examined is the Haitian of monolinguals (regressive nasalization, no postvocalic Irl, etc.). As for the 3

It is worth noting that most scholars who have looked into the phonology of Haitian have not mentioned stress. Among those who have, Cadely is the only one to provide a discussion of stress and to propose an analysis. 4 I thank Albert Valdman for pointing this out to me.

The accentual system of Haitian Creole

131

Lefebvre-Fournier corpus, 5 it displays most of the phonological characteristics typical of the monolingual variety. First, there is almost no occurrence of postvocalic Μ (variable d): words such as mache 'to walk', pale 'to talk', pou 'for', ankö 'again' (from French marcher, parier, pour, encore) abound in the corpus while forms such as premyermän ' (as) first' and estrordine 'extraordinary' are very scarce. Second, final consonant nasalization applies in all forms where the context is appropriate (variable c), for instance monn 'world', konprann 'to understand', prann 'to take' rather that mond, konprand, prand. Third, the occurrence of front rounded vowels is very limited (variable a). In most cases, the front round vowel in the French etymon is replaced by its front unrounded or back rounded counterpart, as shown in (7). Fourth, regressive vowel nasalization is almost systematic, as shown in (8). (7)

(8)

Front rounded vowels a. YES eu 'hour' msieu 'sir' kokelüch 'whooping cough'

Regressive nasalization of V a. NO koman 'how' remed 'medicine'

b. NO minout msie plou depi epi/epwi wit touye etc.

'minute' 'sir' 'more' 'since' 'and then' 'eight' 'to kill'

b. YES kanmyon kann können konmanse mwen menm moun etc.

'truck' 'sugar cane' 'to know' 'to start' 'myself 'person'

Note that, while there are many more examples such as those in (7b) and (8b), the forms in (7a) and (8a) are the only ones found in the excerpts. Finally, there is only one example of s+C cluster (variable e) in the excerpts analyzed, the word spritis 'spiritualism'. It seems to be realized without an epenthetic [ε], but since the word follows the vowel [e], it may be that the two adjacent vowels have merged, a common rapid-speech phenomenon. All in all, it is reasonable to consider that the data in Cadely (1994) and in the LefebvreFournier corpus are representative of the mesolectal/basilectal varieties of Haitian Creole. As for French, historical data is abundant enough to provide information on the accentual system of 17th-18th century French. But what varieties of French? In keeping with current 5

The corpus consists of a series of spontaneous conversations, each involving two or three speakers. In this preliminary study, I restricted the analysis to the speech of three speakers, two males (Speakers 7 and 8) and one female (Speaker A).

132

Anne-Marie

Brousseau

views on the French spoken by the settlers in America, I assume that the French who emigrated to Nouvelle-France and to the Antilles shared common social and geographical origins (cf. Brasseur 1997, Chaudenson 1979, 1994, Hull 1994). I assume further that these French emigrants constituted a fairly homogeneous speech community; they all spoke a mesolectal variety of French showing some regional peculiarities (mainly lie de France and western France), a local variety of the Parisian norm, referred to as Regional Standard French (cf the contributions by Chaudenson, Hull, Morin and Poirier in Mougeon and Beniak 1994). Thus, the analyses for French are based on historical data for 17th-18th century French and data on Quebec French, which still reflects the pronunciation of 17th-18th century French spoken in Central and Western France.

4.2. Haitian Haitian Creole has lexical stress. The last syllable of the word bears primary stress, and stress is maintained when words are combined into larger prosodic units (cf. d'Ans 1968, Cadely 1994, Hall 1953). Phonetically, stress is manifested primarily by higher pitch. Furthermore, Haitian has a quantity-sensitive system. Secondary stress is assigned to closed syllables, that is syllables with a branching rhyme (Cadely 1994). Thus, while most words display only one stressed syllable, words with heavy syllables may display two or more stressed syllables. The contrast is exemplified below (adapted from Cadely 1994). (9)

/zämi/ /depeje/ /mayasa/

(10) /batiste/ /manifestasyS/ /respekte/ /malfektez/ /guvenmä/ /esplike/

'friend' 'to hurry' 'twins' 'birth certificate' 'demonstration' 'to respect' 'burglar, lawbreaker' 'government' 'to explain'

The Lefebvre-Fournier Corpus does not include many words with heavy syllables in nonfinal position (that is, in a position that would clearly show secondary stress). Nonetheless, such syllables are always realized with secondary stress, which is phonetically realized as higher pitch, and is often accompanied by vowel lengthening. Here are some examples (where heavy rhymes are underlined). (11) a. Ii kapäb ^έη yon sistem poul kasel (Speaker 8) 'there is a way of breaking it' b. ou kwe an elektronik me yo palew, pa ^gzänp, de spiritis, yo palew de mistik (Speaker 8) 'some believe in electronics, but they talk to you, for instance, about spiritualism, they talk to you about mysticism'

The accentual system of Haitian

Creole

133

c. premyermän ... otomatikmän ou nan bin an (Speaker 7) ' f i r s t . . . automatically when you get used to it' d kote lap chire pimän yo, pa gen yon mounn ki ka respire (Speaker 7) 'in the area where he is shredding hot peppers, nobody can breathe' e. se ögmante koklüch la ap ögmante (Speaker A) 'the whooping cough was really getting worse' f. bagäy konsä, mpap esplikö (Speaker A) 'things like that, I can't explain' To account for the properties of lexical stress placement in Haitian, Cadely proposes the following algorithm (slightly adapted from Cadely 1994: 108): (12) a. Starting on the right, construct unlimited, right-dominant feet. b. Heavy syllables (i.e. (C)VC syllables) must not be dominated by a weak node. c. Construct a right-dominant word-tree. According to Cadely, secondary stress is maintained on every heavy syllable. For instance in words like /respekte/ or /malfektez/ (cf. (10)), no destressing applies to prevent stress clash between adjacent syllables. In the Lefebvre-Fournier Corpus, I found only one form showing two adjacent heavy syllables, the word estrordiner 'extraordinary'. It is pronounced /estrordiner/, with primary stress on the last syllable and secondary stress only on the first syllable. This form could be analyzed as the output of a destressing rule that deletes the weaker of two adjacent stresses, but such an analysis seems dubious for many reasons. First, destressing rules typically delete a secondary stress that is adjacent to the primary stress (cf. Halle and Vergnaud 1987, Hayes 1995, Pater 2000, Plag 1999). Clashes of secondary stresses are usually "well tolerated" (Pater 2000). The destressing rule of Haitian would thus apply in a very unexpected fashion: it would apply in the context of two adjacent secondary stresses (estrordine) but not in the context where a secondary stress precedes the primary stress (sistem, mistik, otomatikmän). Second, as mentioned above, only one form exhibits such a stress pattern. Third, this form shows another peculiarity: it is realized with a postvocalic Irl, a feature that is not typical of the mesolectal/basilectal variety of Haitian used in the corpus. The form estrordine would then appear as idiosyncratic. 6 Therefore, I will assume that Haitian has no destressing rule and that stress is persistent in all contexts. The parameters and pseudo-parameters needed to characterize the stress system of Haitian are as in (13) and (14) respectively: (13) Metrical parameters of Haitian PI: The word tree is strong on the Right. P2: Feet are Unbounded. P3: Feet are built from the Right. P4: Feet are strong on the Right. 6

Marked Unmarked Marked Marked

There is another theoretical possibility that could explain that the syllable is stressless: that the syllable is light, with a syllabic postvocalic Irl. This possibility is ruled out by the data, since forms such as premyermän show secondary stress.

Anne-Marie Brousseau

134 P5: P6: P8A: P10: PI 1:

Feet are quantity-sensitive. Feet are quantity-sensitive to the Rhyme. There is no extrametrical syllable. Stress is Persistent. Destressing applies in the domain of [Word/Phrase],

(14) Pseudo-parameters of Haitian PP1: Prominence is determined by one feature PP2: The feature Tone does not determine prominence PP3: The feature Stress determines prominence PP5: The feature Stress is assigned to the syllable PP6: Stress is realized as pitch and vowel length

Marked Marked Unmarked Unmarked N/A

Unmarked ? ? ? ?

4.3. French The stress system that characterizes modern French was already in place by the 17th century (cf Marchello-Nizia 1995 and the references therein, Thurot 1881-83). In modern French, lexical stress falls on the last syllable dominating a non-empty nucleus; stress is manifested phonetically by a lengthening of the vowel (cf. Dell 1984, Fonagy and Leon 1984, Leon 1992, Roca 1999, Schane 1968, Selkirk 1972). 7 At the phrasal level, however, a destressing rule deletes the stress on each individual word except for the last one. At the word level, stress is relevant only to determine the application of rules such as vowel truncation and vowel shift. It is not perceptible in phonetic units larger than the word. Only the last syllable of a phrase bears the stress, as illustrated below (where curly brackets delimit the phrasal units and underlining, the syllable bearing stress). (15) {I'enfant} {le premier enfant} {le tout premier enfant} {le tout premier enfant} {de sexe masculin}

'the 'the 'the 'the

child' first child' very first child' very first male child'

The following parameters and pseudo-parameters characterize the stress system of French. (16) Metrical parameters of French PI The word tree is strong on the Right. Feet are Unbounded. P2 Feet are built from the Right. P3 Feet are strong on the Right. P4 7

Marked Unmarked Marked Marked

The formulation of this rule is compatible with analyses that account for schwa ~ 0 alternations in terms of schwa insertion. On the other hand, within frameworks that posit schwa deletion to account for the alternation, the rule is formulated slightly differently. For instance, Schane (1968) proposes the following stress placement rule: the final syllable bears the stress unless its vowel is schwa, in which case the stress falls on the penultimate.

The accentual system of Haitian Creole P5: P8A: Ρ10: Ρ11:

135

Feet are not quantity-sensitive. There is no extrametrical syllable. Stress is Non-Persistent. Destressing applies in the domain of Phrase.

(17) Pseudo-parameters of French PP1: Prominence is determined by one feature PP2: The feature Tone does not determine prominence PP3: The feature Stress determines prominence PP5: The feature Stress is assigned to the word PP6: Stress is realized as vowel length

Unmarked Unmarked Marked Marked

Unmarked ? ? ? ?

4.4. Fongbe Fongbe and other Gbe languages are "true" tone languages (Devonish 1989, Hyman 1982): each syllable of a word is specified as High or Low (cf. Aboh 1999, Brousseau 1991, 1993, Capo 1991, Wieseman 1991). Fongbe has two phonological tones, Low and High, and three phonetic tones: Raising, Falling, and Mid. 8 Many properties of the tone system of Fongbe ascertain its truly tonal character. First, there exist numerous minimal pairs, distinguished solely in terms of their tones, as shown in (18). Second, tones are subject to assimilatory and dissimilatory processes (High tone spreading, Low tone absorption). Third, the nature of consonants influences the realization of tones (a High tone preceded by a voiced consonant is realized as Raising). All these properties are typical of genuine tone systems. (18) gbä xü gän

'to break' 'sea' 'iron'

gbä χύ gän

'to build' 'bone' 'chief

Spreading of the High tone is the main process that modifies lexical tones when words are concatenated in a sentence or in a compound word. The High tone spreads from left to right until the end of its tonal domain. Wherever it stops spreading, it forms a contour Falling tone with the Low tone in the last syllable of the domain, as shown in (19a,b). Spreading may dissociate a Low tone, thus creating a downstep, as shown in (19c). The sequence formed by the subject NP, Tense/Aspect markers and the verb constitutes such a tonal domain, as does the N P object of the verb. 9 Phrases where High tone spreading has applied have a general descending melody (downdrift).

8

The raising L-H tone is the realization of the High tone following a voiced consonant. The falling tone H-L is the realization of a Low tone after the spreading of a preceding High tone. Finally, the mid tone is a phonetic variant of the two complex tones (raising and falling) found in some dialects.

9

A derived word, affixed or compound, always constitutes a tonal domain. Within sentences, tonal domains correspond to phrasal units such as [NPsubject - INFL -Verb], [N-Determiner - Plural marker], [NPcomplement - N], [NPcomplement - P],

136

Anne-Marie

(19) a. {έ xo} {äson we} (s)he buy crab two '(S)he boughttwo crabs.' b. {e kö no

Brousseau

{e χδ} {äson we}

sä}

( s ) h e ANTERIOR HABITUAL sell

{äson}

{e kö no sä} {äson}

crab

'(S)he usually sold crabs.' c. {hwevi-sä-to} fish-sell-AGENTIVE 'fishmonger'

{hwevisä'to}

Lexical tones undergo changes that are imposed by their organization in larger metrical units, determined on the basis of grammatical constructions, in a way that parallels the stress structures in full-fledged intonation languages (cf. Aboh 1999, Brousseau and Brillon 1990, Frechet 1990, Wieseman 1991). In other words, Fongbe has a mixed system, comprised of lexical tones and phrasal stress (cf. Yip 1980 on Mandarin for a similar proposal). Placement of phrasal stress is determined by the algorithm in (20) and illustrated in (21). (20) a. Starting on the left, build unbounded left-dominant phrase-trees, b. Phrase-trees are dominant on the left. (21) s

s \

e

ko

Κ

w Β

w Β

s Β

w Η

no

sa

a

son

{e kö no sä} {äson} - » {e ko no sä} {äson} In the example above, each High tone of the first phrase has a higher pitch than the High tone to its right. Similarly, the Low tone of the second phrase has a higher pitch, as is always the case when a Low tone occupies the initial position in a phrase. This reflects the relational and culminative properties of the stress structure that dominates the tone structure. Leaving aside the options needed to account for the purely tonal system of Fongbe, the accentual system may be characterized by the following parameters and pseudo-parameters. Note that PI has been slightly modified to describe a system that has phrasal stress but no lexical stress. If the prosodic domain relevant for stress assignment is determined by PP5 (word or phrase), then the role of PI amounts to determining the dominance of this prosodic domain.

The accentual system of Haitian Creole

137

(22) Metrical parameters of Fongbe PI: The word/phrase tree is strong on the Left. P2: Feet are Unbounded. P3: Feet are built from the Left. P4: Feet are strong on the Left. P5: Feet are not quantity-sensitive. P8A: There is no extrametrical syllable. Ρ10: Stress is Persistent. PI 1: Destressing applies in the domain of [Word/Phrase].

Unmarked Unmarked Unmarked Unmarked Unmarked Unmarked Unmarked N/A

(23) Pseudo-parameters of Fongbe PP1: Prominence is determined by two features PP2: The feature Tone determines prominence PP3: The feature Stress determines prominence PP5: The feature Stress is assigned to the phrase PP6: Stress is realized as pitch

Marked ? ? ? ?

5. T h e stress system of Haitian: a c o m p r o m i s e in m a r k e d n e s s

Haitian shares with French the stress nature of its accentual system. However, the two languages show substantial differences. While there is only one stressed syllable per word in French (the last one), Haitian words may bear multiple stresses: primary stress on the last syllable and secondary stress on every heavy syllable. In addition, while stress is persistent in Haitian, lexical stress in French is maintained only on the last syllable of the phrase. Hence, in most cases, the French words the creators of Haitian could identify in the data to which they were exposed would bear no stress at all. The fact that an accentual system like that of Haitian has emerged from a nearly stressless system argues for the existence of guiding principles in the elaboration of the interlanguage. These guiding principles may be provided by Universal Grammar and by the grammar of Fongbe. How can the respective roles of the two be identified? Assuming that the creators of Haitian used the grammar of their first language to identify phonetic strings of French and structure them into phonological representations (Full Transfer), they relied on the parameters defining their accentual system, including those required for tones. The grammar of Fongbe generates syllables and words fully specified for tones, and phrases bearing stress. If a full tonal specification had been imposed on phonological representations based on the phonetic strings of French, the resulting forms would depart substantially from the actual stress system of Haitian. To quote Devonish (1989), it takes two to tango. The contact between a stress accent language (superstratum) and a tonal language (substratum) will yield a tonal accent language (i.e. a tone language with incomplete tonification), unless the superstratum does not have lexically significant stress. Tonal accent Creoles (Djuka, Guyanese, Saramaccan, Papiamentu, Principense) have all emerged from superstratum languages that have lexically significant stress (English, Spanish, Portu-

138

Anne-Marie

Brousseau

guese). In contrast, given the phrasal stress system of French, Haitian and other Frenchbased Creoles developed a pure stress system. This is not to deny any influence from LI in the development of the interlanguage, but if transfer were involved, they must have been highly constrained by other principles, namely Universal Grammar. As discussed in § 2, parameter settings of LI that are likely to be transferred in L2 are the unmarked (or less marked) ones. At the same time, these unmarked parameter settings of LI are easier to reset to the marked value of L2 since this can be done on the sole basis of positive evidence. How can the parameter settings of Haitian shed some light on this paradox? As shown in (24) and (25), the parameter setting of Haitian appears as a sort of compromise between those of its source languages, as is often the case for the prosodic system of interlanguages (cf. Alber and Plag 2001, Archibald 1998a, 1998b, Baptista 1989, Broselow and Finer 1991, Pater 1997). (24) Metrical parameters for the three languages PI: P2: P3: P4: P5: P6: P8A: P10: PI 1:

The phrase/word tree is strong on the Feet are Feet are built from the Feet are strong on the Feet are quantity-sensitive Feet are quantity-sensitive to the There is an extrametrical syllable Stress is Destressing applies in the domain of

Fongbe Left Unbounded Left Left No N/A No Persistent N/A

French Right Unbounded Right Right No N/A No Non-Persist. Phrase

Haitian Right Unbounded Right Right Yes Rhyme No Persistent N/A

(25) Pseudo-parameters for the three languages PP1: PP2: PP3: PP4: PP5: PP6:

Prominence is determined by Tone determines prominence Stress determines prominence Tone is assigned to the Stress is assigned to the Stress is realized as

Fongbe two features Yes Yes Syllable Phrase Pitch

French one feature No Yes N/A Word Length

Haitian one feature No Yes N/A Word Pitch(/Length)

Among the 15 settings of Haitian, four remained unchanged, six were reset to the value of French, three were reset to a value diverging from both source languages, and two (PI 1 and PP4) are vacuous. Of the four settings of Fongbe that remained unchanged in Haitian, three (P2, P8A and PP3) are straightforward: no resetting was needed since the parametric values are the same in Fongbe and in French. The fourth one (P10) has been transferred from Fongbe with its unmarked value. Presumably, the parameter could not have been reset because no positive evidence was available. According to Dresher and Kaye (1990), triggers appropriate to particular parameters must be found so as to reset the parameters adequately. In order to make a change to the grammar, there must be a perceived mismatch between the input and the (potential) output.

The accentual system of Haitian

Creole

139

Of the six parameters that were reset to the value of French, three (PI, P3 and P4) involved resetting from the unmarked to the marked option. The resetting has been achieved successfully on the basis of positive evidence. For directional parameters such as PI, P3 and P4, the appropriate cue is as follows (Dresher 1999: 39): "For each setting of direction and headedness, scan across the word. The presence of a stressed syllable in what should be a weak position rules out a setting". Scanning French words with the leftward settings of Fongbe would have revealed stressed syllables on an expected weak position (right edge), hence triggering the resetting of the parameters. Conversely, PP1 was reset from the marked value of Fongbe (two features determine prominence) to the unmarked value of French (one feature determines prominence). PP2 was also reset to the value of French, a value that could also be unmarked (see footnote 1). Changing a parameter to its unmarked value constitutes a much more difficult task since it requires some kind of negative evidence, namely the absence of any tonal distinction. This raises the question of determining what exactly constitutes negative evidence: does the absence of something necessarily amounts to negative evidence? If we take the Full Transfer/Full Access Mode Hypothesis seriously, the answer must be no. According to the hypothesis, the resetting of PP1 was possible only if the creators of Haitian were exposed to evidence showing that their LI setting was wrong. The fact that, in French, stress is realized only as length could have rendered the absence of any tonal prominence in the French data more "salient". The perceived mismatch between the input (tone and stress prominence) and the output (stress prominence only) would have triggered resetting of the parameter. Now this scenario raises another question (that I will address shortly): if the absence of tonal distinction constitute an appropriate trigger for resetting PP1, why is it that the absence of stress on many words (non-persistent stress) cannot constitute an appropriate trigger for resetting Ρ10? The last parameter sharing the value of French is PP5: it has been reset from Phrase to Word, as in French. It is difficult to identify the unmarked value in this case, but the very fact that resetting was successful suggests that positive evidence was available to trigger resetting. However, it is not obvious that such evidence could be easily perceived. Since stress is not persistent in French, most words are realized without it. Superficially, French appears as a system where stress is assigned to the Phrase, as is the case in Fongbe. Given the apparent similarity between French and Fongbe, the same option for PP5 (Phrase) would be expected in Haitian. The fact that Haitian has reset the value of PP5 to the correct, but opaque, option of French suggests that some other cues were perceived as relevant (see below). The most surprising result revealed by the comparison is the value of P5 (and P6) in Haitian. The interlanguage has developed a quantity-sensitive system, the marked option, from two quantity-insensitive systems, the unmarked option. This is completely unexpected, given what we know of the role of markedness values in second language acquisition. Situations where LI and L2 are both unmarked are the most straightforward: the unmarked value must be carried over to the interlanguage. The marked value of P5 in Haitian would be expected only if the parameter of Fongbe was reset to the value of French, on the basis of positive evidence. In other words, P5 would have been reset because the creators of Haitian found some clues in the French data indicating a quantity-sensitive system. In the development of an interlanguage, it is difficult for the system to know what aspect of the representation to change when an error in output has been identified. In the cue-based

Anne-Marie Brousseau

140

learning model of Dresher and Kaye (1990), cues have, among others, the following properties: 1) a cue is not an input but an abstract property or feature that can be derived from an input; 2) a cue must reflect a fundamental property of the parameter. With these properties in mind, we can devise a possible scenario for the unexpected value of P5 (as well as P10 and PP5). According to Archibald (1998b: 288), irregular rhythm constitutes the appropriate clue for resetting the quantity parameter from its unmarked value (quantity insensitive) to its marked value (quantity sensitive). For Dresher (1999: 31), evidence that words with the same number of syllable have different stress contours constitutes the cue to resetting to the marked value. Now, French prosody is characterized by an irregular rhythm, which results from the occurrence of long vowels. French forms include phonemic long vowels (compare fete and faite, bile and belle) and phonetic long vowels (when followed by a "lengthening consonant": /z/, /v/, /r/ or /V). Furthermore, vowel length is also a possible phonetic cue for stress and it is the only phonetic cue for stress in French. Finally, there are no long vowels in Fongbe (and in Haitian), neither phonemic or phonetic. Thus, the creators of Haitian would have interpreted the length of vowels of French as a manifestation of stress. With this interpretation, the contrasts in vowel length shown in words such as those in (26) would have been understood as a contrast in stress contours, hence as a fundamental cue for quantity-sensitive stress. (26) Short vowel /äbelir/ /megroe:r/ /episje/ /vilmä/ /kalmamä/

'to embellish' 'leanness' 'grocer' 'vilely' 'calmly'

Long vowel /äbe:te/ /fre:Joe:r/ /pa:tisje/ /vi:vmä/ /Ja:r33mä/

'to bother' 'freshness' 'baker' 'vividly' 'load'

If correct, this scenario implies that the creators of Haitian interpreted the French data relevant to quantitive-sensitivity only partially. Forms such as /vilmä/ or /kalmamä/, where heavy-rhyme syllable does not show perceivable stress (i.e. long vowel), were not seen as counter-evidence to quantity-sensitivity. This is consistent with Dresher's idea (Dresher 1999: 42-43) that the language learner proceeds in an incremental fashion ("Adjust parameter settings as each datum comes in.") rather than in batches ("Collect all data, then set parameters."). Interpreting vowel length as a cue for quantity-sensitive stress constitutes the kind of "false move" to which the incremental learner is exposed, a move than cannot be corrected because the learner cannot backtrack to undo the (re)setting that have already been set (to prevent infinite loops). Now this scenario provides an answer to the question raised earlier: why is it that the absence of stress could not constitute an appropriate trigger for resetting P10 while the absence of tonal distinctions could constitute a trigger for resetting PP1? In fact, such an absence was probably not detected by the creators of Haitian. For if they have interpreted vowel length in French as a manifestation of stress, they have perceived as stressed all words including a long vowel. Phrases of French were thus perceived as containing stressed words in any possible position, not only in the final position of the phrase, and no clue was available to suggest the marked value (non-persistent stress) of Ρ10.

The accentual system of Haitian Creole

141

The marked value of P5 requires a specification for P6. In this case, the process involves setting rather than resetting. The parameter was set to the only option available in Haitian: a heavy rhyme. Heavy nuclei are not part of the possible syllable structures of Haitian. Finally, PP6 shares the values of both Fongbe and French. The main phonetic realization of stress in Haitian is pitch. Both primary and secondary stress are signaled by a higher pitch; secondary stress is also signaled by vowel length in many cases. Thus, the Fongbe setting for PP6 was transferred in the interlanguage (pitch) and modified slightly to incorporate the value of French (length). This is possible because the three values of PP6 are not mutually exclusive. Languages may signal stress with pitch and length, pitch and intensity, etc. In the absence of well-defined markedness values for the various phonetic cues, I will not offer a better characterization of the way PP6 was reset. 10

6. C o n c l u s i o n

The distribution of unmarked values in the three languages is the following: 3 Ρ and 1 PP in French, 3 Ρ and 1 PP in Haitian, 7 Ρ and 1 PP in Fongbe (putting aside the parameters needed for the tonal component of the system). From a solely quantitative perspective, Haitian is not THE unmarked language with respect to the accentual system. Haitian is as unmarked as French is. Of course, such a quantitative comparison is somewhat misleading. Fongbe appears as the less marked of the three languages because the parameter set relevant to the tonal component of the accentual system is missing. Presumably, this set would include unmarked and marked parameters, while the corresponding sets in French and Haitian would be vacuous. Thus, an extensive comparison would characterize Fongbe as a more marked language, but still, Haitian would be as unmarked as French. This leads me to believe that markedness values prove to be useful in a comparison only as long as qualitative aspects are taken into account. From a qualitative perspective, a promising way to undertake the comparison is to look at the interaction between markedness values in the source languages. Research on second language acquisition has shown that different outputs may result in the interlanguage when there is a mismatch between the markedness values of LI and L2. In the development of an interlanguage, transfer is not likely to occur when LI is marked and L2 is unmarked. The same holds true for Haitian: the only two marked parameters of Fongbe, PP1 and PP2, have been reset to the unmarked value. On the other hand, when LI is unmarked and L2 is marked, the interlanguage can show one value or the other: either the unmarked value trans-

10

To better represent the fact that phonetic cues are not mutually exclusive, the pseudo-parameter PP6 likely needs to be split up in the following manner: PP6: Stress is realized as [One/Two/Three] phonetic cue(s) PP7: Phonetic cue for stress is pitch [Yes/No] PP8: Phonetic cue for stress is length [Yes/No] PP9: Phonetic cue for stress is intensity [Yes/No] The markedness values of the above pseudo-parameters remain to be identified, except for the 'new' PP6 whose default can directly be set on the basis of the Subset Principle.

142

Anne-Marie

Brousseau

ferred from LI remains, or the marked value o f L2 is adopted, following parameter resetting. Again, these two options are attested in Haitian. A n unmarked parameter o f Fongbe may be kept as is (e.g. P2, P8A, PP4 and maybe P10), or it may be reset to the marked value o f French (e.g. PI, P3, P4 and maybe P10). Of course, the determining factors that trigger or hinder parameter resetting in these cases have yet to be understood. These factors may include the presence/absence o f appropriate cues, the perceived similarity between LI and L2 ("it takes two to tango"), and the interaction of other parameters within the grammar. I have maintained that the Full Transfer/Full Access Mode Hypothesis has proven very successful in accounting for phonological aspects o f interlanguages. I hope to have shown that the "Full Substrate/Full U G Genesis Hypothesis" has proven promising in explaining the development of a Creole phonological grammar.

References

Aboh, Enoch Olade (1999): From the syntax of Gungbe to the grammar of Gbe. Sierre: Editions ä la Carte. Alber, Birgit and Ingo Plag (2001): Epenthesis, deletion and the emergence of the optimal syllable in Creole: the case of Sranan. - In: Lingua 111, 811-840. Altenberg, Evelyn P. and Robert M. Vago (1987): Theoretical implications of an error analysis of second language phonology production. - In: G. Ioup and S. H. Weinberger (eds.): Interlanguage phonology. The acquisition of a second language sound system, 148-164. Cambridge, MA: Newbury House. Anderson, Roger (ed.) (1983): Pidginization and creolization as language acquisition. Rowley: Newbury House. Ans, Andre Marcel d' (1968): Le Creole frangais d'Haiti. The Hague: Mouton. Archibald, John (1998a): Second language phonology. Amsterdam: John Benjamins. - (1998b): Metrical parameters and lexical dependency: acquiring L2 stress. - In: S. Flynn, G. Martohardjono and W. O'Neill (eds.): The generative study of second language acquisition, 280301. Mahwah, NJ: Lawrence Erlbaum Associates. Baptista, Barbara O. (1989): Strategies for the prediction of English word stress. - In: International Review of Applied Linguistics 27, 1-14. Brasseur, Patrick (1997): Creoles ä base lexicale frangaise et franfais marginaux d'Amerique du Nord: quelques points de comparaison. - In: M.-C. Hazael-Massieux and D. de Robillard (eds.): Contacts de langues, contacts de cultures, creolisation, 141-166. Paris: L'Harmattan. Broselow, Ellen (1983): Nonobvious transfer: on predicting epenthesis error. - In: S.M. Gass and L. Selinker (eds.): Language transfer in language learning, 269-281. Rowley, MA: Newbury House. Broselow, Ellen and Daniel Finer (1991): Parameter setting in second language phonology and syntax. - In: Second Language Research 7, 35-39. Brousseau, Anne-Marie (1991): Domaines et relations de gouvernement dans les processus tonals du fongbe. — In: Revue linguistique de Vincennes 20, 27-45. - (1993): L'interaction entre consonnes et tons en fongbe: pour une representation 'tonale' du voisement. - In: A. Kihm and Claire Lefebvre (eds.): Aspects de la grammaire du fongbe: etudes de phonologie, de syntaxe et de semantique, 7-29. Paris: Peeters.

The accentual

system of Haitian

Creole

143

Brousseau, Anne-Marie and Joelle Brillon (1990): Le systeme tonal du fongbe. - In: La genese du Creole ha'itien - Etudes syntaxiques, morphologiques et phonologiques, Rapport pour I'annee 1989-90, T o m e I, under the direction of C. Lefebvre and J. Lumsden, 521-555. Cadely, Jean-Robert (1994): Aspects de la phonologie du Creole ha'itien. Ph.D. Dissertation, Universite du Quebec ä Montreal. Capo, Hounkpati B.C. (1991): A comparative phonology ofGbe. Dordrecht: Foris. Carter, Hazel (1983): How to be a tone language. - In: L.D. Carrington (ed.): Studies in Caribbean language, 90-111. St. Augustine, Trinidad: Society for Caribbean Linguistics. Chaudenson, Robert (1994): Frangais d'Amerique du Nord et Creoles fran?ais: le frangais parle par les immigrants du XVIIe siecle. - In: Raymond Mougeon and Edouard Beniak (eds.) (1994): Les origines du frangais quebecois, 167-180. Quebec: Presses de l'Universite Laval. - (1979): Les Creoles frangais. Paris: F. Nathan. Clahsen, Harald (1988): Parametrized grammatical theory and language acquisition: A study of the acquisition of verb placement and inflection by children and adults. - In: S. Flynn and W. O'Neil (eds.): Linguistic theory in second language acquisition, 261-278. Dordrecht: Kluwer. Clahsen, Harald and Peter Muysken (1989): The UG paradox in L2 acquisition. - In: Second Language Research 5, 1-29. Dell, Francois (1984): L'accentuation dans les phrases en frangais. - In: F. Dell, J.-R. Vergnaud and D. Hirst (eds.): Forme sonore du langage, 65-122. Paris: Hermann. Devonish, Hubert (1989): Talking in tones: a study of tone in Afro-European languages. London: Karia Press. Dresher, B. Elan (1999): Charting the learning path: cues to parameter setting. - In: Linguistic Inquiry 30, 27-67. Dresher, B. Elan and Jonathan D. Kaye (1990): A computational learning model for metrical phonology. - In: Cognition 34, 137-195. Eckman, Fred Ross (1977): Markedness and the contrastive analysis hypothesis. - In: Language Learning 27, 315-330. Flynn, Suzanne (1987): L2 acquisition: resetting the parameters of universal grammar. Dordrecht: Reidel. Flynn, Suzanne, Samuel D. Epstein and Gita Martohardjono (1993): The full access hypothesis in SLA: some evidence from the acquisition of functional categories. Paper presented at AILA, Amsterdam. Fonagy, Ivan and Pierre Leon (1984): L 'accent en frangais contemporain. Ottawa: Didier. Frechet, Anne-Lise (1990): Le downdrift en Gungbe. - In: Cahiers d'etudes linguistiques 3, 6-28. Greenberg, Joseph Η (1966): Language universals, with special reference to feature hierarchies. The Hague: Mouton. Hall, Robert A. Jr (1953): Haitian Creole. Grammar, texts, vocabulary. Philadelphia: American Folklore Society. Halle, Morris and Jean-Roger Vergnaud (1987): An essay on stress. Cambridge, M A : The M I T Press. Hayes, Bruce (1995): Metrical stress theory. Principles and case studies. Chicago: The University of Chicago Press. Hull, Alexander (1994): Des origines du fran?ais dans le Nouveau Monde. - In: Raymond Mougeon and Edouard Beniak (eds.) (1994): Les origines du frangais quebecois, 183-198. Quebec: Presses de l'Universite Laval. Hyltenstam, Kenneth (1987): Markedness, language universals, language typology, and second language acquisition. - In: C. Pfaff (ed.): First and second language acquisition processes, 55-81. Cambridge, MA: Newbury House. Hyman, Larry (1982): Globality and the accentual analysis of Luganda tone. - In: Journal of Linguistic Research 2, 1-40.

144

Anne-Marie

Brousseau

Ioup, Georgette (1984): Is there a structural foreign accent? A comparison of syntactic and phonological errors in second language acquisition. - In: Language Learning 34, 1-17. Kellerman, Eric (1987): Aspects of transferability in second language acquisition. Nijmegen: Katholieke Universiteit te Nijmegen. Lefebvre, Claire and Robert Fournier (1976): Corpus du Creole hai'tien. Recordings on tapes. Leon, Pierre (1992): Phonetisme etprononciations du frangais. Paris: Nathan. Marchello-Nizia, Christiane (1995): L 'evolution du frangais: ordre des mots, demonstratifs, accent tonique. Paris: Armand Colin. Mougeon, Raymond and Edouard Biniak (eds.) (1994): Les origines du frangais quebecois. Quebec: Presses de PUniversitö Laval. Mufwene, Salikoko (1991): Pidgins, Creoles, typology and markedness. - In: F. Byrne and T. Huebner (eds.): Development and structure of Creoles languages, 123-144. Amsterdam/Philadelphia: John Benjamins. Nespor, Marina and Irene Vogel (1986): Prosodic phonology. Dordrecht: Foris. Pater, Joseph V. (1997): Metrical parameter missetting in second language acquisition. - In: S.J. Hannahs and M. Young-Scholten (eds.): Focus on phonological acquisition, 235-263. Amsterdam/Philadelphia: John Benjamins. -

(2000): Non-uniformity in English secondary stress: the role of ranked and lexically specific contraints. - In: Phonology 17, 237-274.

Plag, Ingo (1999): Morphological productivity. Structural constraints in English derivation. Berlin/New York: Mouton de Gruyter. Rivera-Castillo, Yolanda (1998): Tone and stress in Papiamentu: the contribution of constraint-based analysis to the problem of Creole genesis. - In: Journal of Pidgin and Creole Languages 13, 297335. Roca, Iggy M. (1999): Stress in the Romance languages. - In: H. van der Hülst (ed.): Word prosodic systems in the languages of Europe, 659-769. Berlin/New York: Mouton de Gruyter. Schane, Sanford (1968): French phonology and morphology. Cambridge: The M I T Press. Schwartz, Bonnie and R. Spouse (1996): L2 cognitive states and the full transfer/full access mode. In: Second Language Research 12, 40-72. Selkirk, Elisabeth (1972): The phrase phonology of English and French. M I T PhD Dissertation. Bloomington: Indiana University Linguistics Club. Singler, John Victor (1993): African influence upon Afro-American language varieties: a consideration of sociohistorical factors. - In: Salikoko M u f w e n e (ed.): Africanisms in Afro-American language varieties, 235-253. University of Georgia Press. Thurot, Charles (1881-83): De la prononciation frangaise depuis le commencement d'apres le temoignage des grammairiens. Geneve: Slatkine Reprints (1966). Valdman, Albert (1978): Le Creole: Statut et origine. Paris: Klincksieck. -

du XVIe

(1983): Creolization and second language acquisition - In: Roger Anderson (ed.):

siecle

Pidginization

and creolization as language acquisition, 212-234. Rowley, MA: Newbury House. - (1991): Decreolization and dialect contact in Haiti? - In: F. Byrne and T. Huebner (eds.): Development and structure of Creole languages, 78-90. Amsterdam/Philadelphia: John Benjamins. Wexler, Kenneth and Rita Manzini (1987): Parameters and learnability in binding theory. - In: T. R o e p e r a n d E. Williams (eds.): Parameter setting, 41-76. Dordrecht: Reidel. White, Lydia (1989): The principle of adjacency in second language acquisition: do learners observe the subset principle? - In: S.M. Gass and J. Schachter (eds.): Linguistic perspectives on second language acquisition, 134-158. Cambridge: Cambridge University Press. - (2000): Second language acquisition: from initial to final state. - In: J. Archibald (ed.): Second language acquisition and linguistic theory, 130-155. Oxford: Blackwell. Wieseman, Ursula (1991): Tone and intonational features in Fon. - In: Linguistique Africaine 7, 6590.

The accentual system of Haitian Creole

145

Wode, Henning (1986): Language transfer: a cognitive, functional and developmental view. - In: E. Kellerman and M. Sharwood Smith (eds.): Crosslinguistic influence in second language acquisition, 173-186. New York: Pergamon Institute of English. Yip, Moira J. W. (1980): The tonal phonology of Chinese. Ph.D. dissertation, MIT. Bloomington: Indiana University Linguistics Club. Youssef, Amani and Irene Mazurkewich (1998): The acquisition of English metrical parameters and syllable structure by adult native speakers of Egyptian Arabic (Cairene dialect). - In: S. Flynn, G. Martohardjono and W. O'Neill (eds.): The generative study of second language acquisition, 303333. Mahwah, NJ: Lawrence Erlbaum Associates.

Anne-Marie Brousseau, University of Toronto, NF 313, Victoria College, Toronto (ON), M5S 3K7, Canada, Email: [email protected]

David

Sutcliffe

African American English suprasegmentals: Α study of pitch patterns in the Black English of the United States

1. Introduction1

The similarity we observe across the range of Atlantic Creoles in terms of grammar and phonology is also evident on the suprasegmental level. The view taken in this chapter is that this similarity or unity extends from the far south of the area of distribution (Surinam) to the far north (the United States) and from the deepest Creoles (Saramaccan and its close kin) to semi-creoles and African American Englishes (AAE).2 This grouping takes in more than the Atlantic Creole area proper, and constitutes a wider hinterland of African American linguistic influence. The idea of a wider area of common African heritage and experience, reflected in language and language culture, is not new. There is a long tradition of studies in this vein, going back to Herskovits (1941), and followed by Abrahams (1976), Rickford and Rickford (1976), Alleyne (1980), Holloway (1990) and Labov (2001). What is more, this view can be readily supported by simple observation3 in addition to any amount of formally adduced evidence. Yet, as far as I know, there is no existing term for this commonalty, so to coin an acronym, let us call it the GLLAD (Greater Linguistic Legacy of the African Diaspora). To take this view is not to disregard or underestimate connections on the suprasegmental level with other Englishes in North America and elsewhere. AAE's use of stress-timing, for example, has been transmitted from the British Isles English of earlier centuries, along with most of its lexis. It is clear from simple impressionistic observation that AAE is basically stress-timed (as most other natively spoken Englishes are) and has word stress patterns which partially or largely correspond to the equivalent patterns in other Englishes. AAE derived most if not all of its lexical entries from this common stock of Englishes. In doing so, it copied the associated pitch and stress patterns of these words as they were carried 1

My thanks go to David Prendergast and to two anonymous reviewers for their helpful comments without which this article would not have reached its present form. Also, I would especially like to thank editor Ingo Plag for his patience, forbearance and guidance. Lastly, I must acknowledge my particular debt to Hazel Carter, whose work on Atlantic Creole suprasegmentals paved the way for this article.

2

Virginia, the Carolinas (and later, Georgia) formed the original heartland of African American speech in the United States. This speech was spread as far west as Texas during slavery, and then in the great migrations of the twentieth centuiy, to all the great urban centres of North America. Of course, on the east (far east) this huge linguistic arena is bound by the Krio and English Pidgin speaking areas across the ocean in West Africa.

3

For example, at a concert where performers are Black Americans and the audience is Caribbean. At such an event it is very clear that artists and public are able to interact within a common "Black" style, deriving above all from the common African diaspora heritage.

148

David Sutcliffe

over into the nascent AAE prosodic system. The question that arises here, however is: How different was that prosodic system, and (as a result) how much reanalysis could have been involved.

2. T h e d i a c h r o n i c a n d s y n c h r o n i c status o f A A E

As readers may be well aware, there has been a long drawn-out debate between the socalled Anglicists and Creolists over the origin of African American English, and my remarks here, as well as the overall evidence presented in this chapter, bear on this debate. The position taken by the main proponents of the Anglicist point of view finds its fullest expression in Poplack and Tagliamonte (2001). Their thesis, is that A A E descends directly from the English of the British Isles with little or no adstrate or substrate contribution from the Anglophone Creoles (they do not even consider the possibility of African influence). There are linguists on the other hand who have found evidence to support to some form of the Creole hypothesis. Most recently Wolfram and Thomas (2002) provide evidence not so much of direct Creole influence as creole-like substrate influence within a distinctly Africanised A A E which must have emerged almost from the very beginning. This view, taken also by Sutcliffe and Wright (2001), may be gaining ground at this time. 4 Interestingly, William Labov (1998) has argued for a third position in this debate which in a sense creates a synthesis out of disparate and seemingly contradictory findings. He argues that on the evidence available to him, contemporary AAE has two main subsystems operating within the grammar, an African-American component (containing creole-like structures) and an English Component (containing structures derived from English). This proposal makes considerable sense. It embodies what could be called the Chinese box principle·. the more one system dominates or replaces another the more the latter may remain as a Chinese box within the other (massive change from above). Nonetheless, the African-American component within AAE evidently includes more than Labov (1998) allows for. As we shall see, there is evidence to suggest that the notion can be extended to include the suprasegmental system, which in turn interfaces with syntax, morphology and lexis. Notice that Labov himself sees the African-American component as a product of the twentieth century evolution of AAE, not an inheritance from the nineteenth century. In other words, AAE has invented rather than inherited its creoleness. This view can be challenged - Wolfram and Thomas (2002) show that, in Hyde county, NC, at least, A A E was very approximately as divergent a century ago as it is now, on important parameters such as intonation and patterns of copula non-insertion. And in my own work I have found a large number of creole-like microswitches in the Ex-Slave Recordings, pointing to the existence of a plantation Creole or semi-creole in use in the South well before the Civil War (Sutcliffe 2001). Yet Labov's view is felicitous in some respects. It coin4

Edgar Schneider and associates, for example Kautsch and Schneider (2000), have also shown that the Gullah speaking area formerly extended much further inland than now And for the argument for substrate and adstrate influence from an earlier more extensively spoken creole on the North American Plantations see also Winford (1997, 1998).

African American English

suprasegmentals

149

cides with a principle that can be observed to operate on the suprasegmental level, and which could be called the re-emergence principle: When a substrate disposition or typological disposition (for example, to tonality) exists, this will have a strong tendency to emerge sooner or later - for example, under increased pressure to radicalise or reassert the racial identity associated with the language. 5

3. T h e w i d e r comparative picture: the Cline

The suprasegmental systems of the various languages and language varieties in the GLLAD can be seen to form a cline or continuum from most obviously tonal, with the highest degree of tonal specification, to least obviously tonal with the lowest degree of tonal specification. In the latter type of languages, which are invariably systems with only two contrastive tones, a maximum of one tone per lexical word (usually high, in most such languages) is specified underlyingly, and all other tones are assigned by low-level rules. All of these varieties, however, assign pitch patterns employed at the lexical level, to moras (tonebearing units) as a potentially separable feature from stress.

3.1. Southern Rim In the far south of the area, Saramaccan Creole is indisputably a tone language with two contrastive tones and a limited number of tonal distribution patterns (Rountree 1972). The most common patterns utilise one High tone per word, corresponding to the stressed syllable in the European lexifier (Portuguese, English or Dutch). In addition we find words of African or putative African origin which may display other patterns - in this case the Η tone, if there is one, is not necessarily a reflex of stress either in the original language, or indeed in the Saramaccan word in question. For some discussion of such patterns, whether European-derived or African-derived, see Good (2003).

3.2. Western and Eastern Caribbean In the Anglophone Creoles of the Western Caribbean, such as Jamaican, we find the following situation: While the most conservative Jamaican maroon varieties that we have evidence for (described in Bilby 1983) have not been analysed for suprasegmental patterns, in the slightly less extreme level of Jamaican Creole - the level extensively documented in Bailey (1966), Cassidy and Le Page (1980) - the most usual pattern is for low (L) rather than high (H) pitch marking on what would be the main stressed syllable in the equivalent 5

A possible instance of the re-emergence principle in action is to be found in Saramaccan (Surinam). Just over two centuries ago, according to Schumann (1788), Saramaccan had initial consonant clusters which have since disappeared as the syllable structure has reverted to CV(CV), the substrate pattern (in most West African languages) which has since "re-emerged."

150

David Sutcliffe

word in English (Sutcliffe 2001). This of course applies to words of English origin, but since the majority of the words in daily use are English-derived, this is a very common pattern. Yet it is important to note that even among English-derived words, there are some significant exceptions (Shelome Gooden, personal communication) where, conversely, the stressed syllable gets high tone marking. If, on the other hand, the source of the word is clearly African, partly African, or putatively African, we may find a different pattern, certainly in trisyllabic words like kötäkö (HLH, "basket"), or käsädä (HHL, "cassava" ) (Lawton, 1963).6 There is also some evidence of rarer patterns; for example there is a single instance in my data of LHH, in the word sueli (citation form, referring to an unidentified fresh water creature) recorded in the Bedford Survey (1973-1974). Jamaican has grammatical tone. For example, it uses unstressed low tone subject pronouns before verbs in the indicative, above all in the first clause in any sequence of clauses. This is ostensibly a case of grammatical tone, since in other syntactic slots, for example the equivalent of English embedded subjects, unstressed subject pronouns regularly take a high tone. As we shall see, this is a pattern which is found elsewhere, most notably in Gullah and United States AAE. Anglophone varieties in the Eastern Caribbean have similar and yet in part contrastive patterns. Here the stressed or stressed equivalent syllables carry Η tones on the surface, except in the case of a significant number of disyllables that have an underlying Η on the stressed syllable which however surfaces as L, (Devonish 1989, Holder 1998). The existence of these patterns gives rise to characteristic minimal pairs such as worka /LH/ ("seamstress", "needlewoman") versus workä /HL/ ("one who works"). These are primarily pitch differences rather stress differences and are perceived as such by native speakers (Roberts 1988). Indeed, as Roberts points out, in these cases the stress remains constant on the initial syllable. Turning to the morpho-syntactic level, we find that use of grammatical tone is extensive in Eastern Caribbean varieties, and Allsopp (1996) gives some examples of minimal pairs of this sort. For instance, where in means 'as soon as', as in sentences like "/« he hit de road, de police hol[d] [h]im", this particle typically takes high pitch and stress (Allsopp 1996:304); Roberts (1988:95) attests essentially the same pattern in Bajan, such that äs meaning 'as soon as', as in As you came, I went, is regularly accorded High pitch, whereas as meaning 'because' is accorded low tone. Similarly, in noun compounds, Allsop gives long-time (HL, 'a long time ago', with intensive meaning) as contrasting with long-time (LH, 'some time ago') with diffuse meaning (Allsoppl996:354). See also James (this volume) for similar contrasts in Tobagonian, and above all Gooden (this volume) for intensive versus diffusive meanings in Jamaican being signalled in the adjective by HL and LH pitch patterns respectively.

6

In marking contrastive pitches or tones I follow the standard convention o f marking Η with an acute accent (e) and L with a grave accent. Downstep is marked with a prefixed exclamation mark (!έ), and upstep with an upward-pointing arrow ( A e).

African American English

suprasegmentals

151

3.3. Northern Rim When we come to the northern anglophone varieties, principally AAE and Gullah (which probably need to be grouped together with Bahamian and Bermudan, along with a number of very minor Creoles) what we find is a basically similar picture, and the idea of a cline or continuum holds. That is, moving from the Southern rim to Western and Eastern Caribbean Creoles to the Northern varieties, there is a detectable decrease in the degree of tonal specification. This is perhaps more a question of linguistic history than of geography, and the north-south alignment of the cline may be little more than a coincidence. At all events, it remains an open question whether the African American English of the United States is tonal at all under the normally accepted definitions of the term. Until now, the default assumption has simply been that AAE is not tonal. Smitherman's (1986) remarks on its 'tonal semantics', meaning its notable use of pitch and rhythm for expressive purposes, are more an intuitive (if accurate) observation than a formal linguistic finding. Interestingly, however, Herskovits (1941), echoed by Loman (1975) traces its striking use of pitch to the original West African substratum, while commenting that AAE lacks a functioning tonal system as such. Geneva Smitherman's remarks about the way these striking suprasegmental features are incorporated into verbal art underline their importance as resources of the language which are consciously drawn upon.

4. Lexical pitch patterns in A A E

Almost all of the AAE speech data featured in this chapter were produced by nineteenth century-born African American speakers. The source for this material are the Ex-slave Recordings - audio recordings of African Americans born during the last years of slavery and recorded (in most cases under the Works Project Administration scheme) when they were aged ninety or more. Most of these interviews were transcribed and published for the first time in the Emergence of Black English (Bailey, Maynor and Cukor-Avila 1991), henceforth the EBE. The former slaves whose speech is quoted from below are: Laura Smalley (LS), Billy McRae (spelled McCrea in the EBE), 1 Joe McDonald (McD), Fountain Hughes (FH), African-born Charlie Smith (CS), Celia Black (CB), Isome Mosely (IM) and Harriet Smith (HS).

4.1. The initial stress HL pattern In the Northern varieties there is a tendency for the stressed syllable - or potentially stressed syllable - in words of general English derivation, to be assigned high relative pitch. In this respect, the northern varieties align with Eastern Caribbean varieties, and differ from Western Caribbean Jamaican. Thus, in AAE we find initially stressed disyllabic words like ράρά (HL), woman (HL), sister (HL), butter (HL), seven (HL), feather (HL), master (HL), 7

The 1860 census for Jasper TX gives his surname as McRae.

152

David Sutcliffe

breakfast (HL), dinner (HL), waggon (HL), punishment (HHL), as well as words with stress on subsequent syllables like commence (LH), manure (LH), M(iss)issippi (LHH) and so on. In all these cases it is the stressed syllable that is assigned high pitch or tone. This is the normal or default pattern, subject to surface perturbations (modifications), as we shall see.

4.2. The initial stress LH pattern However, in AAE (and in Gullah) there is also a shorter list of initially stressed two-syllable nouns where this pattern is reversed, so that it is the unstressed syllable which is assigned high tone. These LOW-HIGH words include mämä (LH), bäby (LH), money (LH) and husbänd (LH), as well as names like Charlie (LH) and Jasper (LH). For Gullah, Turner (1949:252) gives not only bebi (LH, 'baby') and huzbän, (LH, 'husband') with this pattern, but also bükrä (LH, 'whiteman'), and ökrä (LH, 'okra'), (Turner 1949: 252, adapted). In words of three or more syllables we may find such reversals too. For instance we have Löu(i)s(i)änä ( HLL) (EBE LS: 260) where the stress would suggest LHH, as in M(iss)issippi, above. For various reasons (having to do with the complexity of permutations in context) more work needs to be done to confirm the existence of these patterns beyond doubt, and then to confirm the complete inventory of words that have them, but to the extent that such patterns exist, it would seem that they are assigned directly by the lexicon rather than triggered by stress, such that stress and high pitch are assigned separately. This appears to be so in both AAE and Gullah. We have already mentioned that a number of other Atlantic Creoles besides Gullah have a list of words like mämä (LOW-HIGH pitch pattern, with L on the initial stressed syllable). In every case, this pattern represents a departure from a more extended pattern which has Η on the initial stressed syllable (compare seven, feather, mäster, briakfast, dinner, waggon). In every case, too, approximately the same lists of words are involved. Devonish (1989) notes this phenomenon for Guyanese, as we have said, and Carter (1989) similarly attests it for Guyanese, Krio, and Nigerian Pidgin. Roberts (1988) attests the same thing for Bajan, and Christine Corcorran (personal communication, 2001) confirms its existence for Krio. Finally, Holder (1999) provides an exhaustive list for Guyanese, and argues that the phenomenon is one of pitch accent - that is, lexical assignment of pitch which has the word rather than the syllable as its domain. Holden calls words like mämä, husbänd etc. accent two words. An unexpected finding that has emerged during the writing of this chapter is that this same pattern (and so presumably the associated principle of lexical pitch assignment) also occurs in the English of White Missippians and probably other White Southerners, as can be seen in figure 1. Perhaps not so surprising, on further reflection, in view of the intense, prolonged contact that there has been over centuries (Feagin 1997).

African American

English

153

suprasegmentals

Table 1: NP pitch patteras in 3 Northern varieties 8 Variety

LH* / HL*

Title + name / compound

Gullah (Sea Island Creole)

husbän,

buh Rabbit

Texas A A E

husband,

mämä /Charleston

Iridin

\hö

Yse,

work!brid'le mämä

/ράρά

Gälvöstön, Bellville Mississippi White English husband, mämä/ράρά

mister

Payne,

ünc Ίβ Saul /

fritter-tree, post! office mlsterWhitäkir Rdndd

/

f (given

Stelläname),

post!office (*) with stress on first syllable.

( t ) also Stella Rondo, in context.

However, Wolfram and Thomas (2002) also point to significant differences on this level between Black and White varieties, at least in the case of Hyde County, NC. Of course, very few of these LH words are from African sources, with the possible exception of mämä ('mother'; 'ma'am', address term) which may be of part African derivation. It will be readily appreciated that in contemporary and nineteenth century AAE data there are vanishingly few lexical items that are clearly or even arguably of African origin. Thus far, we have seen that AAE, like Gullah (Sea Islands Creole) and other African Diaspora varieties, apparently operates a two-pitch or tone suprasegmental system, and that while the language has stress (or word accent) these pitch patterns would seem to be lexically distributed independently from stress. This in itself is a substantial finding, with some potentially far-reaching implications. However, both the pitch patterns and the exponents of stress in AAE need to be understood in greater detail based on more extensive investigation.

5. Pitch j u n c t u r e and intonation

In this section we look at the way pitch patterns are realised in continuous discourse and interact with intonation

5.1. Discrete levels and downdrift Wolfram and Thomas (2002:151-157) in noting the unusual frequency of discrete high tones in their data from African American speakers (in Hyde county, NC) conclude that these speakers and their variety are non-aligned on this feature with respect to White speakers in the same locality, and conclude that this represents a long-standing difference. In the AAE data considered here, taken from nineteenth-century born speakers in the Ex-Slave

8

Source: Stoddard 1949 for Gullah, Ex-Slave Recordings (LS, CS) for AAE, Welty 1958 (short story: Why I live at the P.O.) for Mississippi White English.

154

David

Sutcliffe

Recordings, discrete HL (LH, etc.) patterns tend to be especially clear and audible, while bridging or levelling across pitches or tones is actually less common than in typical Caribbean Creole data. These discrete pitch levels are assigned to syllables or mora (moras being potentially tone-bearing units, which in AAE may consist of vowels, nasals or continuants). Discrete tonality of this sort is very salient in the following utterance: a comment by former slave Laura Smalley on the way African American mothers in her locality in the nineteen forties would leave their children with minders so that, as she observes, little had changed since plantation slavery. The pitches are shown in "musical" notation also, to make the point clearer. (1)

Dem mother hardly A ever take lit in he's hän an do di A säme släv ry Iti'me. (EBE LS: 148) 'Those mothers hardly ever take [baby] in her arms and do the same as in the time of slavery.' Dem möthir hardly Aeve> take lit in he's hän an do di Asame släv ry !ti'me.

\

This example shows overall downdrift, interrupted by several upsteps. Downdrift is an intonational process found in many languages including English and many of the tone languages of Africa. The way that downdrift operates in African tone languages with two tones - such as Igbo, Efik and many Bantu languages - as well as in characteristic Anglophone Atlantic Creoles, is that Η tones are progressively lowered towards the end of the utterance or tone group. More exactly, Η is automatically lowered by approximately a semitone when immediately following L. Low tones, too, may show a slight slope downwards until the last relative pitch of the utterance is reached, which (when a terminal downturn known as final cadence is applied, see below) will be a fall to the base-line or lowest pitch level normally used in speech. (Weimers 1973:83). This same process is seen in AAE, and is therefore qualitatively different from the downdrift that occurs in Englishes without an African substrate or adstrate connection. Quantitatively, too, there may be a difference: Wolfram and Thomas (2002:152-153) note that African Americans begin on a higher pitch in statements and W-Tj-questions and show greater range in overall falling intonation,

5.2. Downstep and Upstep In addition to lowering by downdrift, high tones may be lowered by downstep, producing a Η which is a step down from a preceding H. An instance is seen in (1) above, where the object pronoun it is realised on a downstep, /lit/. Downstepping on the object pronoun after

African American English suprasegmentals

155

Η on the verb is a regular pattern in AAE. Upstep also occurs in this system. This is where Η is upstepped or realised as extra Η after H. There is an instance of this, too, in sentence (1), on the initial vowel of Aeve>. Turning to sentence (2), below, an utterance from Virginian Fountain Hughes, we again find all these features. Here too, there are very neatly discrete contrastive pitches or tones, displaying overall downdrift, interrupted by some upstepping (shown in the transcript by an upward-pointing arrow). (2)

But someday you come over here, you !come in, duh call me up an let A me know how üh my voice. (EBE FH: 346-347)

The use of discrete pitches or tones in both (1) and (2) is not only striking, it is also paradigmatically creole-like in other respects (even though the segmental phonology in Fountain Hughes' speech in (2) sounds almost British). For example, the assignment of upstepped high pitch or tone to the embedded subject in Ιέί rmi kndw in example (2) exactly parallels a pattern in JC in which such embedded subjects after let or mek are invariably upstepped. We mentioned this earlier in connection with JC, and in a later section we shall be looking at these grammatically dictated patterns in AAE, in more detail.

5.3. Cadence rules Approximately as in Saramaccan, Jamaican, Gullah and other African Diaspora varieties, there is a final cadence effect in AAE. This is a downturn due to a rule which lowers or rounds down the final two syllables of an utterance, such that underlying LH (as in mämä, baby) becomes LL in final position, while HL (as in Bellville, feather, ράρα, woman) in final position remains HL after L, and becomes !HL after H. The final cadence rule operates in utterances which take the form of statements or Wh-questions: (3)

ί (düh) äsk üm, ί säy: Mämä, whäh deh, whäh !deh gö:n? /H!HL/ (EBE McR:49-50) Ί asked her, I said: Mama, where are they, where are they going?'

Thus the rules for final cadence pitch patterns in AAE may be stated succinctly as: HL-> HL LH

HL/L_# !HL /H_# LL/X_#

# marks boundary with final cadence, and X marks a preceding mora, regardless of pitch. This, then, is a set of intonational rules, but mediated through AAE's tonal-like pitch system. Once again, however, it results in an intonational pattern which approximates to intonational downturns in other varieties of English. The question of rule ordering is pertinent here. In the kind of suprasegmental system we are suggesting obtains for AAE (that is, approximately equivalent to what we find in Anglophone Caribbean Creoles) such intonational rules need to be applied last, after the lexical and grammatical "tone" rules. Thus, the order is: firstly, lexical pitch pattern assignment (the output, presumably, of the lexicon);

156

David Sutcliffe

secondly, grammatical pitch pattern assignment which perturbs or operates on the lexical patterns; thirdly, intonational rules which operate on that output according to sentence type: statement, Wh-question, yes-no question, exclamation, or imperative. Wh-questions show overall falling intonation, and begin with a fixed h i g h pitch o r tone on the initial question word. Loman in his study, found that wh-questions began on a high pitch (Loman 1975:228-229). As one might expect, the final cadence rule is not applied in yes-no questions and, in our Ex-slave data at least, is only variably applied in imperatives: (4)

An him seh, Joe? Seh sir? /LH/ {EBE McD 68) 'And he said, Joe? [I] said sir?'

(5)

Take out Ole John [horse] go in the house, feed the bäby. /LH/ (EBE McD:73)

Baby, of course is one of the LH words, and here it retains this pattern even in final position in an imperative sentence. Similarly, final cadence does not apply in vocatives. Thus, vocative use of mämä (i.e. as an address form) retains its basic LH pattern, and may raise the Η to extra Η - as in (3) mäm^ä (whah !deh go:n ), (McR:49-50). This pattern is retained even in final position in the tone group and final cadence is not applied, so that the illocutionary force of the vocative is similar to a yes-no question: (6)

He wänts tiih know, mämä /LH/, if you danced in you younger days. (AAE-speaking nurse assisting in CB interview, EBE CB:189)

(7)

Big building, mämä /LH/. (AAE-speaking nurse assisting in CB interview, EBE CB:200)

Here, the address term mämä, used by the nurse out of deference for her addressee, 115year old Celia Black, does not literally mean "mother." This LH lexical and vocative pattern becomes HH when in the focus.: (8)

1 nuse heär A mämä (HH) !tälk about üm. (LS: unpublished archive 4597A) Ί used to hear mother talk about them [native Americans].'

Like mämä are the other "accent two" words like bäby, hiisbänd, money, and names like Jäsper, Chärlie, Läurä, and so on. Disyllabic LH names and other nouns with stress on the second syllable (McRae, New York) probably pattern in the same way, but this remains to be confirmed.

African American English suprasegmentals

157

6. Grammatical system and pitch patterns

In AAE, preverbal particles and their following stem verbs carry fixed pitch patterns, and it would seem that a significant part of the language's verb morphology surfaces as relative pitch mapped onto the segmental shape of morphemes.

6.1. Pronominal pitch patterns In the Creole language typology to which AAE can in some respects be seen to belong, unstressed pronouns are analysable as pronoun clitics forming part of the verb phrase morphology. The notion of strong (non-clitic) and weak (clitic) pronominal forms in AAE - as discussed below - is in line with this view. What we find generally, probably universally, in the Atlantic Creoles is that unstressed initial subject pronouns take a low pitch morpheme. There are numerous examples of this in the illustrative AAE sentences already given. However, while initial unstressed subject pronouns are consistently assigned low pitch, a different pattern emerges wherever such pronouns are 'embedded' (object of V I , subject o f V 2 ) or are simply non-initial, or are the subjects of relative clauses. In all such cases, the pronoun is regularly assigned high relative pitch - upstepped, or extra high, after an immediately preceding high. (Sutcliffe 2001:142). This pattern is very clearly present in AAE: (9)

Dät düh pünishmdnt th6y got. (EBE McR:42) 'That's the punishment [REL] they received.'

When question words are relativised they trigger high pitch-marking on the following subject pronoun, if any, as in the following extract from the speech of Laura Smalley: (10) Nöw ί heard mämä say when she wühz üh girl, you know, she wühz brought fuhm Missippi, when sh£ wühz üh girl. (EBE LS 394-397) '...when [REL] she was a girl, &you know, she was brought from Mississippi, when [REL] she was a girl.' It will be seen that there is also a high on the subject pronoun you in the tag utterance you know. As the gloss indicates, with an ampersand sign (after Hyman 1971), this is a case of a non-initial subject pronoun, or linking pronoun, which takes a high pitch or tone replacive, a pattern we illustrate in more detail below. The you know tag regularly exhibits this pattern in AAE, with the following verb regularly taking low relative pitch.9 (11) An thäs way they fid him [Laura's stepfather] you know, punish him yöu know would'η give im nothing tuh 6at. An säy löok like he wühz movin älöng too Isl'öw, too !fäst with thät, you knöw, töo good, ΐόο göod with that !yöu know. CEBE LS:290-292). 9

Regularly, but perhaps not categorically, as Erik Thomas (personal communication, 2002) has pointed out.

158

David

Süteliffe

This compares closely with γύ nüo tags in Caribbean and British JC, which have the same relative pitch pattern for the same reason. Sebba and Tate (1986) note the tune assigned to JC yu nüo, but do not analyse it as triggered by the syntax in the way outlined here. Instead they comment on the fact that while it sounds contrastive (you know) there is no evidence that it is used contrastively in contexts in which it occurs. As we just saw, Η (or upstepped Η after H) is mapped not only onto the subject pronouns in relative clauses but onto pronouns in other non-initial subject slots, thus marking their VP as part of a sequence, and subsequent to a previous VP - expressing the idea of 'then' or 'and then.' Hyman (Op. Cit.) discusses similar or highly comparable VP structures in West African languages, referring to them as consecutivisation. He uses an ampersand sign - which we have adopted as our notation, above - to denote the morpho-semantics of such consecutive verbs in V2, V3 (etc.) position: &come, &say, etc. 1 0 (12) But see she'd A nürse this baby that d i d . . . Μ düh be hongri. (EBE LS: 131) 'But see she'd nurse the baby that was...[REL ~ &] it would be hungry' (13) Didn have but one an hit died. None but one an hit died. (EBE IM:78) Ί only had one [baby] &it died. Only one and &it died.'

6.2. Other VP pitch patterns There are other grammatically dictated pitch patterns in AAE that could be described. Above all, there is the area of serial verbs, and what we could call post-serial verbs (serial verbs filled out to appear less divergent from general English, see Sutcliffe 1992:142). Here, as elsewhere, correspondences with Caribbean Creole patterns are evident," although patterns are not always identical. Furthermore, preverbal particles take fixed pitch patterns in somewhat the same way as serial verbs. For example, morphemes like had düh ("hadda") invariably have LL, with Η (Η) on the verb. The same pattern is found with preverbal done, in this case agreeing with Gullah, but contrasting with Caribbean Creoles which have Η on preverbal don, (see James in this volume). While limits of space don't allow us to enter into this in further detail, we shall return to the subject of had düh and its pitch patterns in the next section. The section in question, which concludes this chapter, takes a look at the way significant pitch patterns assigned to syllables provide a fine source of insights into gram-

10

These verbs may even occur in VI position, for example, in a narrative.

" It is worth noting at this point that directional serials (i.e. strings of verbs of locomotion that have go, gone, or come in V 2 or post main-verb position) seem to pattern in A A E as in Caribbean Creoles. See, for instance, the Tobagonian tonal pattern mentioned by James (2001, this volume) for the closely related benefactives with gi(ve) in V 2 or V3 position: Bring it gi mi. ('Bring it for me'). Compare Our white folks they'd ride go to church. (EBE HS:338). Conversely, when the same locomotion verbs go, gone, or come occur as a preverbal auxiliary (in VI or V 2 position before the lexical verb V 2 or V3 in a serial verb string) they take low pitch or tone: Das we de go do. (EBE LS:325, the EBE main transcript has That the way they would do). Finally, see also Tarone (1973) for possible other patterns to explore.

African American English

suprasegmentals

159

matical structure and difficult-to-elucidate cruxes, particularly when in routines which are repeatedly reworked or recycled.

6.3. Repetition with reworking Repetition with reworking occurs very often in the Ex-slave Recordings, in part in accord with the dictates of black stylistics, as we have said. This affords us a number of fascinating sequences, where what seems to be happening is that the camouflaged form typical of modern AAE, and perhaps usual even in nineteenth century AAE when used with outsiders (including strangers with microphones!), is progressively stripped of its camouflage to reveal a more basilectal system. Harriet Smith, an elderly African-American woman, provides us with just such a series. She was aged around 90 when she was recorded by white interviewer John Henry Faulk in 1941. As with other series elsewhere in the corpus, the repetitions under scrutiny here allow patterns to emerge, and show reworking of a basically camouflaged form. They point to the pervasive occurrence of camouflage in AAE, an abounding phenomenon both in AAE's nineteenth century and contemporary versions. Harriet Smith's interview also contains a neat instance of the way in which relative pitch patterns, when correctly construed, make it easier to understand and parse correctly what is being said. The example in question is a particularly shocking sentence in the EBE transcript of this interview. Mrs. Smith is describing the way her husband was gunned down in cold blood, and left to die. White neighbours run to his aid, she says, but are too late to do anything. The EBE transcript at this point (lines 451-453) reads: (14) I don't hardly, how they done, but the white people, W.K. an them, was the first one got to him when he was killed. They had to shoot him, you know. (EBE HS: 451-453) The EBE transcript's version of the sentence is not only shocking (someone having to be shot because of their gunshot wounds?) but probably not correct. The EBE transcription tallies almost perfectly with the segmental phonology - but the suprasegmental phonology is another matter. The fact is that the pitch patterns in the recording, and the pitch patterns that would be assigned to a string like They had to shoot him, differ in detail. As I hear it (based on both the segmental and suprasegmental phonology) what Harriet Smith actually says at this point, is: (14a) I don't hardly [know] how they done, but the white people, W.K. an them, was the first one got to him when he was killed. They heard düh shootin, you know. {EBE HS: 451-453) '.. .They heard the shooting, you know.' If she had actually said "they had to shoot him", this would have had the pitch pattern: they had düh shoot him. In AAE, had duh ('hadda') invariably has LL, with Η (Η) on the verb. This is part of a story told by Harriet of racial envy leading to the murder of her gifted first husband. This man, James, is successful both in African American and White circles, an orator who draws people of both races from miles around when he speaks on political

160

David Sutcliffe

issues. This is intolerable to a group of poor whites in the locality, who resolve to kill him. The leader of this self-appointed posse is one Walter B., who by a bitter twist of fate had been 'nursed' by Harriet herself when he was a baby. In a surprisingly laconic way, she states and restates the identity of the killers and the place of the crime {where Ahe wühz killed ät, HS: 419 - note the upstepped hi after the relativised where).

7. Conclusion

AAE operates a suprasegmental system which is demonstrably similar to those found in other anglophone language varieties in the African Diaspora of the Americas. More exactly, these Diaspora varieties can be seen to form a cline from most to least tonal-like on which AAE may impressionistically be placed. AAE can, furthermore, be transcribed as a system operating High and Low relative pitches or tones. Nonetheless, it remains to be seen whether AAE qualifies as tone language, under any of the currently accepted definitions. If we accept the definition that a tonal system is one where suprasegmental pitch can be analysed as two or more contrastive tones, mapped onto individual syllables of the language, then AAE is tonal. However, if being tonal also means that any alteration to such mapping changes meaning or grammatical function (or leads to lack of meaning, or to grammatical ill-formedness) then the question is not so easily answered at this point. I think it is already clear from our evidence that what seems to be grammatical pitch marking (tied to function words and structures) not only exists in AAE but has a fairly extensive functional load. However, we can be less sure of the extent of lexical pitchmarking (fixed or predictable pitch marking in the lexicon) assigned independently from stress, and having as its domain the syllable or mora rather than the word. However it is easy to see where we need to go from here, using the knowledge gained so far as a springboard. In section 2.1, above, we saw that while stressed syllables in AAE are generally assigned high relative pitch, there is also a short list of words where this pattern appears to be reversed. The latter are LH words like mama, husband, baby money, etc) which conversely appear to have high pitch on the unstressed syllable. This pattern needs to be thoroughly investigated on a larger scale to establish its occurrence beyond doubt, and to determine all the possible perturbations of this LH pattern produced by permutations of 1) cadence, 2) function, 3) focus, 4) emphasis.12 This task will necessarily be undertaken using statistical methods, and a much more extensive data base. Interestingly, grammatical pitch-marking could be investigated in terms of quantifiable variables. For example, the tendency we have noted for an embedded subject pronoun or a non-initial subject pronoun to have Η or upstepped Η could easily be investigated in precisely this way, using standard F° measurement techniques and a statistical research design (taking our lead from Wolfram and Thomas (2002:155). Plainly there is a great deal more work to be done, both of a descriptive and theoretical nature, in analysing the AAE system at this level, and for example relating findings here to our overall knowl12

That is, to rule out the possibility that the LH pattern is triggered for other than purely lexical reasons.

African American English suprasegmentals

161

edge of AAE structure. Nonetheless our findings to date already suggest that African American English suprasegmentals constitute an important exponent of morphology and other grammar, and that further analysis of the suprasegmentals should provide an extremely handy tool with which to tease out the more African-like or creole-like aspects of this highly ambiguous system.

References

Abrahams, Roger D. (1976): Talking Black. Rowley MA: Newbury House. Alleyne, Mervyn C. (1980): Comparative Afro-American. Ann Arbor: Karoma. Allsopp, Richard (1996): Dictionary of Caribbean English Usage. Oxford: Oxford University Press. Bailey, Beryl L. Loftman (1966): Jamaican Creole Syntax: A Transformational Approach. Cambridge: Cambridge University Press. Bailey, Guy, Natalie Maynor and Patricia Cukor-Avila (1991): The Emergence of Black English: Texts and Commentary. Amsterdam / Philadelphia: John Benjamins. The Bedford Survey (1973-1974): A survey made by David Sutcliffe of first and second generation speech in Bedfordshire, reported in Sutcliffe 1978, and Sutcliffe 1982. Bilby, Kenneth (1983): How the older heads talk. A Jamaican Maroon spirit possession language and its relationship to the Creoles of Suriname and Sierra Leone. - In: New West Indian Guide 57 (12), 37-88. Carter, Hazel (1989): Three Creole Pitch Systems. - In: Isabelle Herik (ed.): Papers from the 18!h Conference of African Linguistics. Montreal: Universite de Quebec ä Montreal, April 1987, 27-44. Cassidy, Frederick and Robert LePage (1980): Dictionary of Jamaican English. Cambridge: Cambridge University Press. Devonish, Hubert (1989): Talking in Tones: A Study of Tone in Afro-European Languages. London and Barbados: Karia press, and Caribbean Academic Publications. Feagin, Crawford (1997): The African contribution to Southern States English. - In: Cynthia Bernstein, Thomas Nunnelly and Robin Sabino (eds.): Language Variety in the South. Tuscaloosa: University of Alabama Press, 59-71. Good, Jeff (2003): Tonal morphology in Creole: High tone raising in Saramaccan serial verb constructions. - In: Geert Booij and Jaap van Marie (eds.): Yearbook of Morphology 2002. Dordrecht: Kluwer. Gooden, Shelome (this volume): Prosodic Contrast in Jamaican Creole Reduplication. Herskovits, Melville Jean (1941): The Myth of the Negro Past. New York: Harper brothers. Holder, Maurice (1998): A proposal regarding prosodic features and levels of representation in Guyanese English. Paper given at the conference of the Society for Pidgin and Creole Linguistics, New York, January 1998. - (1999): Accent Tonal en Anglais Est-Caribeen: Une Breve Esquisse. Manuscript. - In: Holloway, Joseph (ed.) (1990): Africanisms in American Culture. Bloomington and Indianapolis: Indiana University Press. Holloway, Joseph E. (1990): Africanisms in American Culture. Bloomington and Indianapolis: Indiana University Press. Hyman, Larry (1971): Consecutivization in Fe' Fe'. - In: Journal of African Languages, 10. James, Winford (this volume): The role of tone and rhyme structure in the organisation of grammatical morphemes in Tobagonian,165-192.

162

David

Sutcliffe

Kautsch, Alexander and Edgar Schneider (2000): Differential creolization: Some evidence from earlier African American Vernacular English in South Carolina. - In: Ingrid Neumann-Holzschuh and Edgar W. Schneider (eds.): Degrees of Restructuring in Creole Languages. Amsterdam /Philadelphia: Benjamins, 247-74. Labov, William (2001): Foreword. - In: Shana Poplack and Sali Tagliamonte (eds): African

American

English in the Diaspora. Amsterdam/Philadelphia: Benjamins, xiv-xvii. - (1998): Co-existent Systems. - In: Salikoko S. Mufwene, John R. Rickford, Guy Bailey and John Baugh (eds.): African American English: Structure, History and Usage. New York, Routledge. Lawton, David (1963): Suprasegmental Phenomena in Jamaican Creole. Unpublished PhD. Dissertation, University of Michigan. Loman, Bengt (1975): Intonation of an African American speaker. Manuscript. Quoted in: Wolfram and Thomas (2002). Poplack, Shana and Sali Tagliamonte (2001): African American English in the Diaspora. Oxford/ Maiden MA: Blackwell. Rickford, John and Angela Rickford (1976): Cut-eye and suck-teeth. - In: Journal of American Folklore. Spring, 295-309. Roberts, Peter (1988): West Indians and their Language. Cambridge/New York: Cambridge University Press. Rountree, S. Catherine (1972): Saramaccan tone in relation to intonation and grammar. - In: Lingua, 29, 308-325. Sebba, Mark and Susan Tate (1986): You know what I mean? - In: Journal of Pragmatics, 163-172. Smitherman, Geneva (1986): Talkin' and Testifying The Language of Black America. Detroit: Wayne State University Press. Stoddard, Albert (1949): Animal Tales Told in the Gullah Dialect (record albums and mimeo transcriptions). Washington, D.C.: Library of Congress. Sutcliffe, David (1982): British Black English. Oxford: Blackwell. - (1992): System in Black English. Avon: Multingual Matters. - (2001): Voice of the Ancestors. - In: Sonja Laneheart, 129-168. - (2003): Eastern Caribbean suprasegmental systems. - In: Michael Aceto (ed.): Contact Languages of the Eastern Caribbean. Amsterdam / Philadelphia: John Benjamins. - Forthcoming. Review of Wolfram, Walt and Erik R. Thomas 2002 and Poplack and Tagliamonte 2001. To appear in: Language and Multicultural Education. Sutcliffe, David and Fiona Wright (2001): Unlikely though it be. Paper presented at the conference of the Society for Pidgin and Creole Languages, Washington, DC: LSA meeting, January 2001. Tarone, Elaine (1973): Aspects of intonation in Black English. - In: American Speech 48, 29-36. Turner, Lorenzo (1949): Africanisms in the Gullah Dialect. Ann Arbor: University of Michigan Press. Weimers, William E. (1973): African Language Structures. Berkeley: University of California Press. Welty, Eudora (1954): Why I live at the Post Office. Jackson: Southern Writers. Available at

Winford, Donald (1997): On the origins of African American Vernacular English: a creolist perspective. Part 1: The sociohistorical background. - In: Diachronica 14(2), 305-44. - (1998): On the origins of African American Vernacular English: a creolist perspective. Part 2: Linguistic features. - In: Diachronica 15(1). Wolfram, Walt and Erik R. Thomas (2002): The Development of African American English. Oxford/ Maiden MA: Blackwell.

David Sutcliffe, Universität Pompeu Fabra, Rambla 30-32, 08009 Barcelona, Spain, Email: [email protected]

Section 3: Morphophonology

Winford James The role of tone and rhyme structure in the organisation of grammatical morphemes in Tobagonian

1. I n t r o d u c t i o n 1

In Tobagonian Creole (TOB), certain grammatical morphemes and phrases behave in discourse in such a way as to suggest that their phonological structure plays a critical role in that behaviour. This paper examines morphemes, especially from the following subsystems: 1) auxiliaries (AUX) and negators (NEG), 2) determiners (DET), and 3) suffixes (SUF), 2 in an attempt to describe how tone (or pitch) and (expiratory) stress are related to their rhyme structure and how that relationship syntactically and semantically differentiates morphemes in those subsystems. 3 It will be shown that some monosyllabic morphemes associate only with (L)ow tone, some only with (H)igh tone, and others with either usually for morphemic contrast, and still others with either, depending on their location as suffixes in the syntax of the phrase or sentence. Furthermore, it will be shown that non-monosyllabic (and nonmonomorphemic) items have different tone patterns depending on their location in the phrase or sentence. More specifically, the paper demonstrates how tone, stress, rhyme structure, syntactic category, semantics, and syntax combine to organise morphemes in the subsystems, raising and attempting to answer a variety of questions relative to those subsystems. After treatment of the AUX/NEG, DET, and SUF morphemes, the paper takes the opportunity to tabulate, without illustration, other monosyllabic morphemes that associate with either high or low tone, as well as other non-monosyllabic ones that have a variable tone pattern. Finally, it summarizes the facts that were presented on the phonology-morphosyntax interface in discourse. The TOB data are spelt with the normal phonemic symbols, except e and ö which are respectively used for the tense vowel sounds in words like 'make' and 'own' that, unlike the case in RP, are not diphthongised. The rest of the paper is organised as follows. Section 2 provides an overview of the general accentual system of Caribbean English Creoles as a backdrop against which to better appreciate the TOB facts, noting, in the process, a virtual absence of focus on the role of tone and rhyme in the organisation of Creole grammatical (as distinct from lexical/content)

1

2

3

I must express my deep gratitude to the editor of this volume and two reviewers who reviewed the paper and made valuable suggestions for its improvement. In particular, I would like to commend the reviewer who provided 19 pages of commentary that reflected close reading of the text. The term 'suffix' is used loosely to refer to morphemes that routinely attach to specific morphemes in normal discourse, e.g., locative de to demonstrative dem. No special theoretical claim as to its grammatical status is intended by its use. Sometimes, the opportunity is taken to show that tone differentiates grammatical/morphosyntactic morphemes/structures from content ones, for example, maa" granny' v. mai Ί am...'.

166

Winford James

words. The bulk of the paper comes in section 3, which describes tone distinctions in the subsystems in relation to stress, rhyme structure, syntax, and semantics. And section 4 summarises those distinctions.

2. Overview of the Caribbean English Creole accentual system

As languages that have emerged essentially out of the contact of European and West African peoples in, characteristically, a master-slave relationship, Creoles are spoken with accentual systems that are audibly different from those of the older European languages from which they have drawn most of their lexical material and with which they have co-existed and interacted for centuries. TOB is an English-lexicon Creole which, if we date its existence from the early 1760s (from which time up to the political independence of Trinidad and Tobago in 1962 there was effective British colonisation of Tobago), 4 has been interacting with English for just under two and a half centuries now (James and Youssef, 2002). Because English is by consensus a stress system and some West African substrate languages are, also by consensus, tonal systems, Caribbean English speech (not excluding or isolating the Creole part of the speech!) has been described in the literature (which, at the foundation of Creole studies, was produced mostly by Europeans) as reflecting a mixture of tone and stress. At its most acrolectal, it is heard by and large as a stress system, and at its most basilectal as a tone system, though Caribbean acrolectal speech is also produced with tonal prosody and Caribbean basilectal speech also with syllable stress (see Solomon, 1993, e.g., for a discussion of the Trinidadian accent). But if we see Caribbean Creole speech broadly as speech that is different from SE speech/RP, then the accent we must be concerned with is that which is different from syllable stress. The nature of the accentual system of Caribbean English Creoles - that is, whether it is essentially characterised by tone/pitch or stress or a combination of both - has been a topic in Caribbean creolistics since the 1960s (see, e.g., Lawton, 1963 and Cassidy and Le Page, 1967/1980); and though over the years there have been different views, the dominant one is that tone is the main and distinguishing feature (e.g., Allsopp, 1972, 1996; D'Costa and Berry, 1971; Alleyne, 1980; Carter, 1987; Solomon, 1993; and Devonish, 1989, 2001), or, at the very least, that tone relates with stress in such a way as to make Caribbean English pronunciation distinctive from SE pronunciation (see, especially, Alleyne, 1980; Allsopp, 1996; and Devonish, 1989). Both Lawton (1963) and Cassidy and Le Page (1967/1980) thought that the Jamaican system was based on stress. The latter, in particular, held that much of Jamaican phonology was a reflex of RP phonology, including syllable-prominence, and that since tone plays a phonemic role in tone-languages, it cannot be involved in syllable-prominence (p. xliv). Allsopp is perhaps the most forceful proponent of tone. In his 1972 paper, he pointed to the existence of minimal tonal pairs in Guyanese Creole, e.g., high-pitch, long-vowel kyää 'can't' and low-pitch, long-vowel kyää 'can' (p. 6); and in his 1996 magnum opus, he observed that the most distinguishing feature of the Caribbean English accent is 'the phrasal 4

There were numerous short-lived efforts by different groups of Europeans before.

Tone and rhyme structure in grammatical

morphemes in

Tobagonian

167

intonation in which the separation of syllabic pitch and stress is a major factor of difference from spoken Standard English...' (p. xliv). Allsopp (1996, pp. xliv-xlv) distinguished Caribbean Creole and SE lexical patterns as follows (numbers representing high-low tone levels and the aigu representing stress on the syllable immediately preceding it):5

tea. eher foun.tain-pen ex.er.cise e.du.cat.ed

SE pattern IT Μ 13' 2 1/ ß'2 11 /3' 2 2 I/

CE pattern IV 21 / Γ 12/ IV 1 2 / IV 1 2 2/

D'Costa and Berry (1971) proposed that all nouns (and many other words) in Jamaican Creole have inherent tone and that, therefore, the language is a residual tone system. Alleyne (1980: 73) proposed that, in Jamaican and Guyanese, tone 'became less and less prominent as a distinctive feature' under English influence, but that 'it remains a characteristic feature of the phonetic shapes of morphemes and larger speech segments such as phrases and sentences.' He further observed that its primary lexical role 'is to differentiate homonymous members of pairs of proper nouns and common nouns such as baker, Baker...'. Carter (1983) did not think there were 'sufficient grounds for classifying CE as tonal' (p. 110), but four years later (1987, p. 220), she described Guyanese Creole as a tonal system since '[t]he association of stress and pitch is such that any syllable with primary stress has surface high pitch, though not all surface high pitches are analysed as having underlying H ' . Solomon (1993, p. 34) stated the matter directly: "It seems correct to say that the accentual system of Trinidadian, as of other English-lexicon dialects of the Caribbean, is based on pitch and not on stress. ...English stress is heard as pitch (and Trinidadian pitch accent is heard as stress by Standard English speakers); disyllables are either high-low... or low-high, the latter being the pattern characteristic of the oldest and most frequently used vocabulary items." (Emphasis added.)

Devonish (2001) described a Creole (Guyanese, for example) restructuring of English words that has resulted in tone-shifting or tone-conversion, a state of affairs in which the Η tone on the stressed first syllable of certain multisyllabic English words (e.g., wätaa' water' and häaspital 'hospital') is replaced by L tone, with Η tone shifting to another syllable, usually the following one. Devonish looked at the broad Guyanese lexicon and found a complex situation in which, for example, 1) word-initial stress is reinterpreted as underlying Η tone (UHT) in an already-tonal system, e.g., ändastäan 'understand' (pp. 85, 91); 2) there are monosyllabic items with UHT, e.g., mii Ί , me, m y ' (emphatic), da 'that', gi " g i v e ' (pp.88-89); 3) there are items with 'sole' U H T on the first syllable, e.g., farin 'foreign', komfatebl 'comfortable' (p. 89); 4) there are monosyllabic items without UHT, e.g., nil Ί , m e ' (unemphatic), fü 'for, to' (p. 90); and 5) there is the addition of a second U H T on disyllabic items, resulting in minimal pairs such as päk'it 'packet' - pakit'pocket' and torkii 'Turkey' - torkii 'turkey' (p. 94). Devonish (2001: 186) observed that it is normal in 'Anglo-West African varieties' 'for words borrowed from English to have the location of

5

In other words, on the items the stress is on the same syllable in both languages, but the tone levels over syllables are different in the languages.

168

Winford James

the first High tone...corresponding to the location of word stress in the English cognate', but that there is a large number of exceptions which manifest tone-shifting, which he regarded as an 'innovation' (pp. 198-203). He went further in that work to posit an HL melody on the first and second syllables of tone-shifted items such as bölii [belli] 'belly' and bäaskit [bääskit] 'basket'. 6 Until now, there has been no work on tone in TOB and, in any case, in the literature on Caribbean tone the focus has been on lexical tone, with hardly any isolation of grammatical tone. It is evident from Devonish (1989), for example, that tone plays a role in the organisation of grammatical morphemes (cf. rni vs. mii above), but there is no special emphasis on this. One researcher who has given it more than passing mention, however, is Sutcliffe (1986 and this volume). Sutcliffe (1986) looked at how tone in Jamaican Creole expresses grammatical relations (in different clauses and constructions) within syntax and discourse, and observed in part that 'preverbal particles, modal verbs, conjunctions, prepositions and other particles generally carry a fixed tone' which 'is usually if not always a fixed high tone' (p. 2). Sutcliffe (this volume) is concerned, however, with identifying 'principle aspects of the [19th century ex-slave] African American English suprasegmental system' and finds, inter alia, that Η and L tones are differentially associated with different grammatical categories such as preverbal particles, the copula, and pronouns. In view, therefore, of the paucity of attention in the literature to the role of tone in Creole grammar, my paper will hopefully bring to light observations and insights that will advance our understanding of the complexity of systems that have traditionally been seen as simple by comparison with older, established, and more studied languages such as English, one of their partners in social interaction.

3. Tone distinctions in the selected subsystems

3.1. Non-emphatic vs. emphatic elements in the AUX/NEG system The subsystem of auxiliaries/negators contains the preverbal markers imperfective a, 7 future/modal go, (remote) past bin, and (present) habitual doz (including their reduced forms), and the positive modal k(y)3, all of which bear only low tone. It also contains completive don, passive ge, emphasiser duu, and all morphemes of negation, including negative modals (e.g., no, en/e, k(y)aan /k(y)ä(ä), kun/kü), which all bear only high tone. The sentences in (l)-(2) illustrate this, with L tone also represented by the grave accent (') and Η tone by the acute accent ('):

6

7

According to Devonish (2001, p. 188), 'the first syllable receives an HL melody assigned by rule. In addition, however, a lexical HL melody is assigned to the second syllable.' The L-tone form ä also serves a pre-NP copula, as in shi ä tiicha. 'She's a teacher' and ä shii. 'It's she.'

Tone and rhyme structure in grammatical morphemes in Tobagonian (1)

a. mi ä tiich. 'I'm teaching.' b. mi go> ό tiich di klaas. Ί will teach the class.' c. mi bin> in/bl/bi>l> ι tiich di klaas. Ί (had) taught the class.' d. mi döz> öz/dö > ö tiich. Ί teach.' e. shi k(y)ä kuk. 'She can cook.' f. shi kü kuk. 'She can cook.'

(2)

a. mi don tiich. 'I've finished teaching.' b. wen mi düu tiich di klaas + MAIN CLAUSE. 'When I DID 8 teach the class....' c. hi g e nak dong yeside. 'He was knocked down yesterday.' d. mi no nö. Ί don't know.' e. mi en> 4 nö. Ί don't know.' f. shi k(y)ään(t)> k(y)ä(ä) kuk. 'She can't cook.' g. shi k(y)ä kuk. 'She can't cook.' h. shi kun> kfl kuk. 'She couldn't cook.'

169

As a result of the reductions in (1), regressive assimilation occurs in which there are phonological fusions of two adjacent L-tone syntactic categories - V(owel)-rhyme pronominal subject and V-rhyme preverbal marker, creating monosyllabic forms such as m'ää

g SIL 0.055

Di

r

i

η

g r

i

η mango

greengreen

TW^y a. The green-all-over (spotted) mango. (SG)

SIL

Prosodic

201

contrast

b. Green-all-over (spotted) things. (PR)

Disyllabic base words. The pattern for disyllabic unreduplicated words is also HL, i.e. Η on the first syllable and L on the second syllable (see figure 5). Figure 5: F 0 pattern for monosyllabic unreduplicated word (SG) 0.004913

0-

-0.0034790.05 Time (s)

250 200-

,50

· ·

I ' I

100·

f g SIL

I

0

m gummy

0.05

I SIL 0.7

Time (s)

Gummy

202

Shelome Gooden

Time (s)

* ·

1 1

ε

y

0 SIL

yellow

SIL 0.055

Time (s)

Yellow As shown in figure 6, the pitch pattern for intensives is Η on the first syllable, which remains high over the next two syllables and L on the final syllable (Η Η Η L). Figure 6: F 0 pattern for intensive reduplicated words with disyllabic bases 0.009033

-0.0078744— 0.09 Time (s)

yellowyellow

mango

Time (s)

a. The mango is very yellow. (SG)

203

Prosodic contrast

i" ... Η

y 3L

4. Η >

L

Η

ε

1 ο

ydlcwjellow

rrangp

3L

Tine®

b. Very yellow mango (JR)

c. O, the mango is very yellow. (PR) When we compare the distributive words (figure 7) we see that the high is maintained only up to the second syllable, after which the pitch is low (Η Η L L), i.e. Η on the reduplicant and L on the base.

204

Shelome

Gooden

Figure 7: F0 pattern for distributive reduplicated words with disyllabic bases

0.09

1.6 Time (s)

a. The mango is yellow all over. (SG)

•

•

Η

y 3L

ε

....

. · " " · ·

L I

Ο y ε

1

ο mango

D

3L

TirrB (s)

b. The yellow-all-over (spotted) mango. (PR) The difference between the two patterns is highlighted in figure 8,10 which shows the F0 trace of both the intensive (solid line) and distributive (dotted line) word. 10

The reduplicated words were extracted from figures (6a) and (7a) using the extraction method in PRAAT 4.026. This method preserves the start and end time of each reduplicated word.

Prosodic contrast

205

Figure 8: F 0 pattern for intensive and distributive reduplicated words with disyllabic bases 250n

100-1 0.795125

1.45243

Time (s) Very yellow (Η Η Η L)/Yellow all over (Η Η L L) Summary and Discussion. Based on the observations made above, the general pitch pattern for distributive and intensive reduplications might be summarized as shown in table 1; the suggested phonological patterns are shaded. The subscripts indicate which portion of the reduplicated word the 'tone' is associated with, reduplicant (R) or base (B). For both intensives and distributives, differences can be captured with reference to the base. In words with a monosyllabic base, both intensives and distributives have a Η on the reduplicant but !H and L respectively on the base. Similarly, for words with a disyllabic base, the intensive and distributive reduplicant have a Η pitch on both syllables but the two syllables of the reduplicant have a HL pitch and a LL pitch respectively. Table 1: Summary of pitch patterns: Phonetic

Intensive

Η μ !H[B]

Disyllabic base Η H[R] Η L[B]

Distributive

H[R] l[B]

Η H[R] LL[bj

Monosyllabic base

Phonological properties'. As noted earlier although there is no one-to-one correspondence between the phonetic patterns and phonological properties of the words, suggestions are made for a preliminary phonological analysis. Recall that both monosyllabic and disyllabic unreduplicated words have a HL pattern in citation form. We might argue that this HL pattern is the underlying pattern and that under reduplication it is realized differently depending on the semantic property of the output form. In the case of monosyllabic based distributives, the HL pattern is maintained but is realized over two syllables instead, the Η on the reduplicant and the L on the base. In monosyllabic base intensives on the other hand, only the Η is realized and on the reduplicant. Notice however, that in order to account for the presence of the !H on the base, one could argue that downstep is triggered by the presence of an underlying (L) on the reduplicant. Positing this underlying L allows us to account for the difference between an intensive with a surface Η !H pattern and a distributive with a surface HL pattern.

206

Shelome

Gooden

Given the pattern for disyllabic base words, the HL pattern seen in the unreduplicated word is again maintained in distributive words. That is, the initial Η is realized over both syllables of the reduplicant and the L is realized over both syllables on the base (H L). In the intensives, the Η of the HL pattern is realized over both syllables of the reduplicant as well as the first syllable of the base and the L is realized only over the last syllable (H HL). 11 The working hypothesis then is that reduplicated words with an intensive meaning have two prominent syllables while words with a distributive meaning have only one. This phonological finding is captured in table 2. Table 2: Summary of pitch patterns: Phonological

Intensive Distributive

Monosyllabic base H,R1 !ΗΓΒΙ HiRi L[B]

Disyllabic base H|ri HLfBi HfR! L iB1

6. Conclusion

This paper investigated a possible prosodic difference between two segmentally identical reduplications in Jamaican Creole, distributive and intensive. An analysis of the F 0 contours showed that there is a pitch difference between distributive and intensive reduplication and a preliminary phonological analysis was proposed. However, more needs to be learned about these differences. For example, what other acoustic cues to meaning are there and are these cues maintained in all contexts? It is possible that other acoustic properties such as vowel quality, vowel duration and intensity are cues to the differences in meaning as well, especially since it is not clear whether the prosodic differences are due to differences in tone or differences in stress. For example, cross-linguistic studies on the acoustic cues to stress contrasts in segmentally identical words show that the most consistent and significant acoustic correlates of stress are fundamental frequency and duration (Fry, 1955, 1965; Lehiste, 1961; Bruce, 1977). The results are more varied for other cues like amplitude (intensity) and spectral characteristics of the vowel. These studies show that there is no single acoustic property that is the most important one for the perception of prominence in utterances. The impression of prominence is a composite phenomenon that is best understood by reference to several parameters. An important area for further research is to discover the full extent of the acoustic differences between the two reduplications and further to find out which of the acoustic properties are salient for listeners (Gooden, in progress). This will

11

An alternative analysis for disyllabic base intensives is Η !H as in monosyllabic base words. If this approach is taken, then w e are able to refer to the presence of an underlying L in the reduplicant of both monosyllabic and disyllabic base words and distinguish them for distributives which have no underlying L. This analysis is similar in part to H. Devonish's suggestion (1998, email communication) that disyllabic base distributives have an initial HL tone on the first mora e.g. likiliki : lik (HL) i (H). Given the data thus far, the Η HL analysis is preferable since this is the dominant pattern observed.

Prosodic

contrast

207

certainly provide additional empirical support for the distinction between the reduplicated forms discussed here.

References

Adamson, Lillian and Norval Smith (1999): Derivational Reduplication in Sranan. Paper presented at the 4th Westminster Creolistics Workshop on Reduplication in Contact Languages April 8th-10th. London. Alderete, John (1993): The Prosodic Morphology of Jamaican Creole Iteratives. - In: Elena Benedicta (ed): University of Massachusetts Occasional Papers 20. The U M O P in Indigenous Languages. Amherst: Graduate Linguistic Student Association, 29-50. (Reprinted in 2001 in ROA) Alleyne, Mervyn (1987): Predicate structures in Saramaccan. - In: M. Alleyne (ed): Studies in Saramaccan language structure. Caribbean Culture Studies 2. Amsterdam/Jamaica: University of Amsterdam and University of the West Indies, 71-87. Anagbogu, Philip N. (1995): The Semantics of Reduplication in Igbo - In: Journal Languages 25:1, 43-52.

of West

African

Baker, Philip (1999): Notes on Reduplication in Mauritian Creole. Paper presented at the 4 lh Westminster Creolistics Workshop on Reduplication in Contact Languages April 8 th -10 th . London. Bakker, Peter (1987): Reduplications in Saramaccan. - In: M. Alleyne (ed): Studies in Saramaccan language structure. Caribbean Culture Studies 2, Amsterdam/Jamaica: University of Amsterdam and University of the West Indies, 17-40. Carter, Hazel (1987): Suprasegmentals in Guyanese. Some African Comparisons. - In: Glenn Gilbert (ed.): Pidgin and Creole Languages: Essays in Memory of John E. Reinecke. University of Hawaii Press: Honululu. Cassidy, Frederick (1957): Iteration as a word-forming device in Jamaican folk speech. - In: American Speech 32, 49-53. Cassidy, Frederick and Robert B. LePage (eds.) (1967): Dictionary N e w York: Cambridge University Press. Comhaire-Sylvain, Suzanne (1936): Le Creole ha'itien; morphologie Imprimerie De Meester: Port-au-Prince (Haiti) L'auteur.

of Jamaican

English. London and

et syntaxe. Weteren (Belgique)

DeCamp, David (1974): Neutralizations, iteratives and ideophones: the locus of language in Jamaica. - In: DeCamp, David & Hancock, Ian (eds.): Pidgins and Creoles: current trends and prospects. Georgetown University Press, 46-50 Devonish, Hubert (2000): Reduplication as Lexical and Syntactic Aspect Marking: The Case of Guyanese Creole. Unpublished manuscript: University of the West Indies, Mona. Fry, Dennis (1955): Duration and Intensity as physical Correlates of Linguistic Stress. - In: Journal of the Acoustic Society of America 45:4, 765-768. -

(1965). The Dependence of Stress Judgments on Vowel Formant Structure. - In: Proceedings of the Fifth International Congress of Phonetic Science 306-311. Gooden, Shelome (2001a): Is there tone in Jamaican Creole Reduplication? Paper presented at the SPCL Meeting. Washington DC. January 4 th -5 th . - (2001b): A Case for Distributive Reduplication in Jamaican Creole. Paper presented at the Montreal-Ottawa-Toronto Workshop on Phonology. University of Ottawa. February 2 nd - 4 th . -

(2002): Reduplication in Jamaican Creole: Semantic Functions and Prosodic Constraints. - In: S. Kouwenberg (ed): Twice as Meaningful: Reduplication in Pidgin and Creoles. Westminster Creolistics Series Volume 8. London: Battlebridge Publications.

208

Shelome

Gooden

-

(in progress): The Phonology and Phonetics of Jamaican Creole Reduplication. Ph.D Dissertation. The Ohio State University. Huttar, Mary L. and George L. Huttar (1997): Reduplication in Ndyuka. - In: Arthur K. Spears and Donald Winford (eds.): The Structure and Status of Pidgins and Creoles. CLL Vol. 19. Amsterdam and Philadelphia: John Benjamins Publishing Company, 395-414. Kouwenberg, Silvia and Dartene LaCharitö (1998): Inflectional versus Non-iconic Derivational Reduplication in Creole Languages. The Creolist Archives. - (2000): Iconicity in Caribbean Creole Reduplication: Inflection Versus Derivation. Manuscript. University of the West Indies and Universite Laval. - (2001): The Iconic Interpretations of Reduplication. Issues in the Study of Reduplication in Caribbean Creole Languages. - In: European Journal of English Studies. Vol. 5:1, 59-80. Lawton, David (1963): Suprasegmental Phenomena in Jamaican Creole. PhD Thesis. Ann Arbor. Department of English. Lehiste, Ilse (1961): Some Acoustic Correlates of Accent in Serbo-Croatian. - In: Phonetica 7, 114147. Meade, Rocky R. (1996): On the Phonology and Orthography of Jamaican Creole. - In: Journal of Pidgin and Creole Languages 11:2, 335-341. Migge, Bettina (2002): The origin of predicate reduplication in the Surinam Eastern Maroon Creole(s). - In: Silvia Kouwenberg (ed.): Twice as Meaningful: Reduplication in Contact Languages. Westminster Creolistics Series Volume 8. London: Battlebridge Publications. McCarthy, John (1985): Phonological features and Morphological Structure. - In: John F. Richardson, Mitchell Marks and Amy Chuckerman (eds.): Papers from the Parasession on the Interplay of Phonology, Morphology and Syntax. Chicago Linguistic Society, 134-161. Sebba, Mark (1981): Derivational Regularities in a Creole Lexicon: the case of Sranan. - In: Linguistics 19, 101-117. Winford, Donald (2000): Property Items and Predication in Sranan. - In: Journal of Pidgin and Creole Languages 12:2, 237-301.

Shelome Gooden, Department of Linguistics, The Ohio State University, 222 Oxley Hall, 1712 Neil Avenue, Columbus, OH 43210, USA, Email: [email protected]

Thomas Β. Klein

Syllable structure and lexical markedness in Creole morphophonology: Determiner allomorphy in Haitian and elsewhere

1. Introduction 1

Creole languages are often claimed to have no phonological or morphological processes of noteworthy complexity. By the same token, the phonology and morphology of creole languages is alleged to be simple or unmarked in some significant sense. In contrast to the prevailing view, patterns of considerable complexity and markedness may be found in linguistic descriptions of Creole languages. One such pattern is the well-documented morphophonology of the postposed definite determiner (DET) in French-lexified Antillean Creoles such as Haitian and St. Lucian. In these languages, DET appears as a CV or V allomorph depending on the phonology of the preceding stem. Interestingly, the CV allomorph is chosen with stems ending in consonants or glides, e.g. pitit la 'the child' (cf. pitit 'child'), whereas the V allomorph appears with stems ending in vowels as in ru a 'the wheel' (cf. ru 'wheel') (data from Valdman 1978: 80 ff.). In essence, the allomorphy of DET in Antillean Creoles shows a typologically unusual preference for consonant clusters and vowel hiatus in external sandhi environments. Syllable structure markedness would lead us to expect the opposite pattern, namely a preference for CV structures and, hence, the avoidance of consonant clusters or vowel hiatus. Thus, we would expect ungrammatical forms such as * petit a and *ru la. The allomorphy of the English indefinite determiner, for instance, follows markedness expectations much more neatly. The VC allomorph appears essentially before vowel-initial stems, e.g. an apple, whereas the V allomorph appears before consonant-initial stems as in a book. If the allomorphy of English a/an behaved analogously to Antillean DET, English would be expected to license ungrammatical forms such as *a apple and *an book. Creole languages are often claimed to be characterised by a strong tendency towards CV sequences, that is, open syllables with onsets of single consonants, although French-lexified Creoles are sometimes said to be excluded from this supposed trend (Holm 1988, 2000, McWhorter 2000). In Romaine's work, for instance, one reads: "Creoles [ . . . ] have no initial or final consonant clusters. They have a simple syllable structure which consists o f alternating consonants and vowels, e.g. C V C V . " (Romaine 1988: 63)

1

Earlier versions of this material have been presented at the 2001 meeting of the Society for Pidgin and Creole Linguistics in Coimbra, Portugal and at the 2001 International Workshop on the Phonology and Morphology of Creole Languages in Siegen, Germany. Special thanks to Ingo Plag for discussion and encouragement, to Emmanuel Nikiema and Ingo Plag for supplying materials and to Birgit Alber, Michel DeGraff, Viviane Deprez and an anonymous reviewer for useful comments. All responsibility for errors lies with the author.

Thomas Β. Klein

210

However, recent detailed investigations of the syllable structure of individual Creole languages such as Aceto (1996) for Saramaccan, Alber & Plag (2001) for Sranan and Sabino (1993) for Negerhollands have shown that the syllable structure of creole languages may exhibit onset clusters and, hence, appears to be more complex than generally assumed. Thus, the present paper may be understood as a contribution towards the better understanding of the occurrence of vowel and consonant clusters in creole sound structure. It is considered a truism by many language scientists that creole languages exhibit no morphology to speak of, in particular as far as inflection is concerned. It is surprising to such views that there should be a significant set of creole languages in which phonologically conditioned allomorphy may be found. Thus, the data from Antillean DET are striking pieces of evidence against traditional views of the supposed simplicity and unmarkedness of creole language phonology and morphology. Instead, DET allomorphy in French-lexified Antillean Creoles is an intricate pattern that deserves detailed formal analysis. The aim of this paper is to examine the morphophonology of D E T in French-lexified Antillean Creoles with particular reference to syllable-related effects. Much of the data is drawn from discussions of Haitian Creole, but data from other Creoles are also prominent. A critique of earlier formal analyses of DET allomorphy is presented and it is shown how this pattern presents a challenge to conventional Optimality Theory (OT). An alternative analysis invoking Lexical Representation as Pure Markedness (LRPM) is presented. Finally, implications of this allomorphy for current notions of creole simplicity as in McWhorter (2001a, b) are discussed.

2. A t y p o l o g y o f d e f i n i t e d e t e r m i n e r a l l o m o r p h y

The allomorphy of DET in Haitian is richly documented in sources such as Valdman (1978), Bernabe (1987), Cadely (1995, 2002) and Nikiema (1999). Analogous facts are found in St. Lucian French Creole (Carrington 1984, Bhatt & Nikiema 2000a,b) and Dominican French Creole (Taylor 1977). The shape of DET in these languages varies under the influence of three different morphophonological and phonological processes: V/CV alternation, glide insertion and nasalisation. (1)

Consonant- and glide-final stems a. /malad/ 'sick' [malad+la] b. //at/ 'cat' [Jat+la] 'book' c. /liv/ [liv+la] 'thing' d. /bagaj/ [bagaj+la] e. /kaw/ 'crow' [kaw+la]

'the 'the 'the 'the 'the

sick (person)' cat' book' thing' crow'

The data in (1) show that DET appears as the CV allomorph la after stems ending in a consonant as in ( l a ) - ( l c ) or a glide as in (Id) and (le).

Syllable structure and lexical markedness in Creole

(2)

Vowel-final stems a. /papa/ 'father' b. /bujwa/ 'kettle' c. /papje/ 'paper' d. /lapli/ 'rain' e. /bato/ 'boat' f. /tu/ 'hole'

[papa+a] [bujwa+a] [papje+ja] [lapli+ja] [bato+wa] [tu+wa]

'the 'the 'the 'the 'the 'the

211

father' kettle' paper' rain' boat' hole'

The data in (2) show that DET appears as the (G)V allomorph after stems ending in vowels. The glide [j] is inserted after front vowels, as in (2c) and (2d), whereas the glide [w] occurs after back rounded vowels as in (2e) and (2f). No glide appears after stem-final [a], as shown in (2a) and (2b). Instead, DET appears as the V allomorph [a] which results in vowel hiatus. 2 There appears to be interesting phonological variation with respect to glide insertion. Valdman (1978) has reported that glide insertion depends on whether the final vowel is tense/close or lax/open. (3)

Glide insertion (Valdman 1978: 75) a. /ru+a/ [ru w a] 'the wheel' b. /po+a/ 'the skin' [po w a] c. /diri+a/ [diriJa] 'the rice' d. /pje+a/ [pje J a] 'the foot' e. /papa+a/ [papaa] 'the father' f. /boko+a/ [bokoa] 'the sorcerer' [vea] 'the glass' g· /νε+a/

Valdman's data in (3a) - (3d) show that stem-final tense vowels license glide insertion with postposed DET, whereas stem-final lax vowels do not appear with inserted glides, as shown in (3e) - (3g). This particular distribution of inserted glides does not appear to be observed in all varieties. Thus, Nikiema (1999: 71) transcribes lax non-low vowels as subject to glide insertion as in his example [lapejä] 'the rabbit' (cf. / lap ε/ 'rabbit'). A detailed investigation of this variation is beyond the scope of this chapter. However, two important points can be made for the present purposes. First, the data in (3f) and (3g) are a good indication that phonetic vowel hiatus may result from postposed DET. Secondly, the distribution of glide insertion in Valdman's data is evidence for another Creole pattern of interesting complexity that comes as a surprise to proponents of the view that creole languages are in some significant sense radically simple. Note that the inserted glide in (2) and (3) above breaks up vowel sequences if the first vowel is non-low (or tense, in the case of Valdman's data). I make the uncontroversial assumption that the constituent dominating the glide in these contexts is the syllable onset. 2

It is not entirely clear from the available descriptions if a true vowel cluster or perhaps a long vowel surfaces in the a+a environment. Note that phonetically long vowels may occur in some varieties of French-lexified Antillean Creoles as in St. Lucian [la:pe] 'rabbit' (Carrington 1984: 19). The phonetic appearance of vowel hiatus is not crucial for the purposes of this paper, however. The fundamental fact is that there are two adjacent vowel slots across the morpheme boundary.

Thomas Β. Klein

212

Overall, then, there are conflicting trends as to markedness in the morphophonology of DET. On the one hand, marked V and C clusters occur. On the other hand, homorganic glides are inserted to create unmarked onsets. Thus, apart from homorganic glide insertion, the phonological structures in the DET phrase show a reversal in comparison to ordinary markedness. Given that a V allomorph of DET is available, the fact that it is not used with consonant- and glide-final stems to create a CV structure runs counter to markedness expectations. Analogously, the fact that the CV allomorph of DET is not used with vowelfinal stems to create an unmarked CVCV sequence is the opposite of what is predicted under markedness. The account of these anti-markedness effects is laid out further below. Nasalisation produces three additional surface allomorphs of DET, [lä], [na] and [nä], after local nasal vowels and consonants (see, in particular, Bhatt & Nikiema 2000a, b and Cadely 2002 for extensive discussion). Whereas nasalisation is outside the scope of this paper and distinct from the present focus on the cluster-related effects of DET morphophonology, I wish to note that this nasalisation is of considerable complexity in that the consonant, the vowel or both may be nasalised. Phonological processes of such complexity are unexpected in languages that are supposedly very simple and straightforward in their grammatical structure. Given that French-lexified Antillean Creoles allow word-initial obstruent plus liquid clusters as in glas'ice' and klu'nail' (see Valdman 1978: 57f. and Carrington 1984: 30f. for discussion of permissible syllable types), the question arises as to how sequences of this kind are syllabified across word boundaries. As far as the syllabification of DET and the preceding word is concerned, I follow the unanimous view in the previous literature (Tinelli 1981, Nikiema 1999, Bhatt & Nikiema 2000a, Cadely 2002) that the morpheme boundary in these contexts coincides with a syllable boundary. That is, the allomorphs of DET constitute their own syllable and do not include material from the preceding morphological unit. This does not mean that there is never any resyllabification in the languages at hand. Thus, Cadely (1995, 2002) has argued that vowel elision with Haitian clitics is accompanied by resyllabification into the clitic host. Assuming traditional syllabic constituency, the syllabification of DET and its preceding morphological unit may be represented as in (4) (see also Cadely 2002). 3 (4)

Syllable structure representations a. σ σ b. σ σ /I /I po+wa

σ

/ ι / ι I papa+a

c.

σ

σ

/ ι \ /I Jat+la

The representations in (4) exemplify the conflicting trends in the syllable structures licensed in the morphophonology of DET. On the one hand, marked vowel hiatus and closed syllables occur via heterosyllabic V and C clusters as in (4b) and (4c), respectively. On the other hand, homorganic glides are inserted to create unmarked onsets as in (4a). 3

Nikiema (1999) and Bhatt & Nikiema (2000a) have presented an analysis of Haitian and St. Lucian syllable structure in terms of Government Phonology (see below). However, the fundamental ideas that DET makes up a syllable distinct from the preceding morphological unit and that the inserted glide occupies the syllable onset are shared between their approach and the present one.

Syllable structure and lexical markedness in Creole

213

The definite determiner does not alternate in all French-lexified Antillean Creoles. Consider the DET typology presented in table 1. Table 1: Forms of DET in French-lexified Antillean Creoles (Bernabe 1987: 14) Guyanese After vowel After consonant or glide

a a

Guadeloupean; Ν Dominican la la

Haitian; S Dominican; St. Lucian etc. a la

Table 1 shows that the definite determiner in the French-lexified Creole of Guyana is nonalternating a, whereas Guadeloupean and Northern dialects of Dominican have non-alternating la. All other French-lexified Antillean Creole varieties exhibit the alternation under discussion. Note that any formal analysis of DET allomorphy should be able to account for this typology straightforwardly. This section has shown that the allomorphy of DET in Haitian, St. Lucian and other French-lexified Antillean Creoles is of considerable complexity. The V allomorph occurs after stem-final vowels, whereas the CV allomorph is found after stem-final consonants, thus displaying an anti-markedness effect. Homorganic glide insertion takes place to provide a syllable onset after front and back rounded vowels or, in some varieties, only after non-low tense vowels. Not all Antillean French-lexified Creoles show this alternation, however. Following the earlier literature, DET and the preceding stem are assumed to be in distinct syllables. In the next section previous formal analyses of the DET allomorphy are reviewed and critiqued.

3. Critique of previous analyses

It is clear from the preceding description of the allomorphy of DET that markedness and the licensing of vowel and consonant sequences are indispensable parts of the understanding of this pattern. Previous analyses may be evaluated in terms of how well they capture this connection. Most of the previous analyses of DET allomorphy have interpreted the observed V/CV change as some kind of phonological /^-alternation. In the literature employing linear rules (Fournier 1978, Valdman 1978, Tinelli 1981), it has been proposed that DET has the underlying representation /la/ which is then subject to a rule of /-deletion to derive the V allomorph in the correct environment. Fournier (1978) proposes the rule in (5); a similar, but more complex rule is found in Tinelli (1981: 69). (5)

Definite article truncation (obligatory) (Fournier 1978: 103) C —> 0 / [+syll] # [D(

214

Thomas Β. Klein

The rule in (5) states that a consonant is deleted at the left edge of the determiner after a vowel-final word. First, it is surprising to find C in the input to this rule, given that only 1 is supposed to be subject to it. Secondly, this rule is merely a technical restatement of the facts without providing any real insight into the phenomenon. In particular, there is no link to markedness or syllable structure which are so decisive in understanding the allomorphy of DET. It has been pointed out in Valdman (1978) that the account of DET allomorphy as a 1/0alternation is complicated by the fact that the / in the homophonous adverbial la 'there' never deletes even though the phonological environment may be identical otherwise (see Valdman 1978: 84). The comparison with the adverbial la is important because it shows that DET allomorphy is subject to a degree of lexical idiosyncrasy which must be captured in any viable account of the alternation. Obviously, the linear rule deleting C does not affect the adverbial la because the rule only applies to determiners. However, it seems fairly clear that the presence versus absence of 1 in the determiner la is not due to the syntactic category of this item, but instead has to do with more intricate lexical information. Consequently, linear rules such as (5) are not suited to capture the lexical boundedness of DET allomorphy and its connection to markedness. Syllable structure is essential in the Government Phonology (GP) (Kaye, Lowenstamm and Vergnaud 1990, among others) analysis of DET allomorphy presented in Nikiema (1999). In this analysis, the underlying representation of alternating DET includes / as a floating consonant, as shown in (6). (6)

Underlying representation of alternating DET (Nikiema 1999: 84) Ο R I I I N XX 1 a

The consonant in (6) is floating in the sense that is not associated to the timing slot x. Given this representation, the central idea of Nikiema's analysis is that / is realised if proper government is not possible, but that it is not realised if proper government obtains. The reader is referred to Nikiema (1999), Bhatt & Nikiema (2000a), Nikiema & Bhatt (this volume) for full details. To summarise their analysis, forms with surface hiatus such as [papaa] are licensed because the empty onset of DET is properly governed by the following vowel. Word-final consonants in words like [Jat] are not in coda position, but instead are onsets of syllables with an empty nucleus. Attachment of DET to such a word creates a sequence of two empty positions that cannot both be properly governed. The floating consonant is anchored and, hence, pronounced at the surface to ensure proper government of the stem-final empty nucleus. Anchoring of the floating consonant results in the observed form [Jatla]. Note that this analysis may capture the non-alternating nature of adverbial la easily by saying that the segmental melodies in the representation of this element are associated to their timing slots at all levels of representation. Whereas this analysis assigns an essential

Syllable structure and lexical markedness in Creole

215

role to syllable structure and improves considerably on the linear rule analysis by capturing the lexical idiosyncrasy, the connection to markedness is not explained. Thus, it is unclear how the GP account of Haitian and St. Lucian DET squares with the less marked and more common constellation in which consonant clusters and vowel hiatus are avoided. 4 The framework of Optimality Theory (OT) (McCarthy & Prince 1993 et seq., Prince & Smolensky 1993) appears in principle well suited to account for allomorphy involving cluster constraints and markedness because both notions play a prominent role in the theory. However, as is argued in the next section, no viable account of DET allomorphy is available in conventional OT. Instead, Lexical Representation as Pure Markedness (LRPM) as developed in Klein (2000) based on Golston (1996) is invoked in order to present a viable and insightful analysis of DET allomorphy.

4. T h e D E T a l l o m o r p h y c h a l l e n g e t o O T

Syllable structure in OT flows from the conflict between faithfulness and structural wellformedness (a.k.a. markedness) constraints. The basic syllable structure constraints ONSET and NOCODA in (7) and (8) demand unmarked syllables, that is, syllables with onsets, but without codas (Prince & Smolensky 1993; see Kager 1999 for discussion). (7)

ONSET

Syllables must have onsets. (8)

NOCODA

Syllables are open. Faithfulness constraints such as MAXlO are ranked high in languages in which more complex syllabic structures such as closed syllables are allowed. (9)

M a x I O (McCarthy & Prince 1995) Every input element has a correspondent in the output.

A basic consequence of this system is that marked structures may be permitted, but are never required to the exclusion of unmarked structures. Thus, this system predicts that no language as a whole requires codas but excludes onsets (see Hammond 1997 for further discussion). The system further predicts that unmarked structures may be preferred in processes of prosodic morphology such as reduplication even though the language as a whole allows marked structures. This effect is dubbed the emergence of the unmarked (TETU; McCarthy & Prince 1994). For example, under TETU a language may have exclusively CV reduplicative affixes, even though codas commonly occur elsewhere. 4

Cadely's (2002) analysis seems to represent no improvement over the GP analyses. Using floating I similar to the representation in (6), he simply stipulates (p. 460) that "in Haitian Creole a floating consonant preceded by another consonant generally associates to a timing unit." (my translation)

216

Thomas Β. Klein

The anti-markedness effect observed in the allomorphy of DET in French-lexified Antillean Creoles presents an immediate challenge to the conventional O T approach just outlined. Recall that D E T allomorphy prefers consonant sequences and vowel hiatus, even though less marked options are in principle available. This is in sharp contrast to the conventional O T system where marked structures are allowed, but not required in lieu of unmarked structures. On the other hand, the O T approach cannot be dismissed outright, given that glide insertion supplying onsets is fully compatible with the OT predictions, as in Rosenthall's (1994) analysis, for instance. The challenge to O T may be understood better by comparing the Antillean facts to French elision and liaison which are close to the mirror image of D E T allomorphy in the Creoles. Consider the data in (10). (10) Elision and liaison in French a. l'ami 'the friend' b. /lez#ami/ 'the friends'

le camarade /le#kamarad/

'the comrade' 'the comrades'

Note that vowels in the French determiners are elided before vowel-initial stems, as shown in (10a), thus satisfying the ONSET constraint. On the other hand, final consonants inherent in the French determiners surface before vowel-initial stems, but are suppressed before consonant-initial stems, as shown in (10b), satisfying ONSET and avoiding violation of the NOCODA constraint. In other words, elision and liaison avoid vowel hiatus and consonant clusters. This is in sharp contrast to Antillean DET allomorphy where vowel hiatus and consonant clusters are preferred. The conventional OT account works well for French elision and liaison, as shown in Tranel (1994, 1995). But note that it is impossible under this account to arrive at the structures needed for Antillean DET allomorphy. Consider the tableaux in (11) and (12) where the sad face symbol ® indicates a candidate that is incorrectly optimal.

(Π)

Input: papa, DET ® a. papa+la b. papa+a

ONSET

NOCODA

*I

Given that a and la are in principle available as phonological representations of Antillean DET, the syllable structure constraints ONSET and NOCODA optimise the ungrammatical CV candidate, as shown in (11). Input: bagaj, DET a. bagaj+la © b. bagaj+a

ONSET

NOCODA *!

Just like for vowel-final stems, the incorrect result obtains for stems ending in consonants and glides, as exemplified in (12). Note that no ranking of the constraints ONSET and NOCODA or other commonly invoked syllable structure constraints is able to render vowel hiatus and consonant clusters as in ( l i b ) and (12a), respectively, optimal at the expense of the less marked CV-type structures in (11a) and (12b) (see Lapointe 2001 for related dis-

Syllable structure and lexical markedness in Creole

217

cussion). Consequently, an alternative approach must be sought. As demonstrated in the next section, such an alternative is available through the Lexical Representation as Pure Markedness (LRPM) approach developed in Klein (2000).

5. Lexical constraint violations, homorganic glide insertion and the distribution of D E T allomorphs

The preceding sections have highlighted a number of significant issues that need to be addressed in any viable formal analysis of Antillean DET allomorphy. First, the analysis should give a clear account of the anti-markedness effect. Secondly, it should be obvious from the analysis how DET induces the observed allomorphy, but not homophonous items such as the location adverbial la. Thirdly, there should be a transparent formal connection between the non-alternating and alternating forms of DET in the various French-lexified Creoles. Finally, the insertion of the glides [j] and [w] after non-low vowels to supply an onset for DET needs to be accounted for. The following proposal for an OT analysis of DET allomorphy is designed to address these issues. The analytical framework is also shown to be of use beyond the account of DET allomorphy and argued to have significant advantages over competing models such as morpholexical listing in OT (Lapointe 2001) and Sympathy Theory (McCarthy 1999).

5.1. Input allomorphy and Lexical Representation as Pure Markedness (LRPM) Recall that there are two non-alternating phonological surface forms of DET, a and la, which are also prominent in the alternating Creoles. Thus, it makes sense to propose two input allomorphs, roughly /a/ and /la/, for the Creoles with non-alternating and alternating DET. Under this proposal, the non-alternating Creoles have /a/ or /la/, as appropriate, whereas the alternating Creoles have both. The task of the grammar, then, is to produce the correct distribution of these allomorphs. Analyses based on linear rules (Fournier 1978, Valdman, 1978, Tinelli 1981) did not take the non-alternating patterns into account and are, thus, at an empirical disadvantage. In contrast, Nikiema (1999) has proposed the inputs in (13) to account for the DET typology.

218

Thomas Β. Klein

(13) DET inputs according to Nikiema (1999: 84) a. [la] b. [la]/[a] c. Ο

R

Ο

Ν χ

χ a

χ

R

[a] Ο

R

Ν

Ν

χ

χ

a

a

The plain input in (13a) is a straightforward representation of non-alternating la. The input for alternating DET in (13b) has been discussed above. (13c) is proposed to underlie nonalternating a. This form includes the melody for /, but is distinct from (13b) through the absence of the timing slot x. Note that this input is highly counter-intuitive given that the floating /-melody never surfaces anywhere in the varieties with non-alternating a and, thus, introduces unwarranted redundancy. Consequently, the simple /a/ input proposed below is preferable for non-alternating a. I propose that the input allomorphs /a/ and /la/ are enhanced through a lexical constraint violation in the alternating Creoles. This lexical constraint violation flows from the model of Lexical Representation as Pure Markedness (LRPM) proposed in Klein (2000). This model has been developed to account for the phonology - morphology interactions of umlaut in German and the Austronesian language Chamorro. LRPM is based on the two-level OT model in which inputs are mapped to outputs (McCarthy & Prince 1995) and on Golston's (1996) idea that phonological and morphological structure may be represented through constraint violations. In this model each constraint violation mark in the representation encodes linguistic markedness. Morphemes may carry constraint violations as part of the input representation so that they are distinguished in part by the presence versus absence of such violations. The lexical constraint violations or desiderata are kept track of in the output through faithfulness constraints such as MAXIO. The maintenance of an input violation mark in the output satisfies faithfulness, whereas the cancellation of an input violation in the output is a violation of faithfulness. Candidates with a cancelled lexical constraint violation may be suboptimal because of the concomitant faithfulness violation. A given constraint may only be violated once in a given input representation. Furthermore, no constraint ranking may be determined through any input representation. This limits the power of potential input representations considerably. Note that markedness is represented directly in this model. Any lexical constraint violation encodes markedness that a representation without it lacks. I follow Rubach & Booij (2001) and Tranel (1994) in assuming that allomorphs are freely available in the input and that the correct surface allomorph is chosen through the constraint hierarchy. Note, however, that I do not endorse the idea of Richness of the Base (Smolensky 1996) which does not allow any specific property to be stated at the level of input representations. Thus, the LRPM model may be seen as a development of other OT approaches in which representationally specified inputs have been proposed. Examples of specific properties assigned to input representations in earlier OT work include the underspecified 'empty' morpheme RED in standard OT treatments of reduplication (McCarthy &

Syllable structure and lexical markedness in Creole

219

Prince 1995 et seq.), inputs of varying degrees of syllabification (Sprouse 1997), latent input segments (Zoll 1998) and catalectic input moras (Stiebels and Wunderlich 1999). The advantage of LRPM is that inputs in this model utilise elements that make up the fabric of OT grammar elsewhere, namely, constraints and constraint violations. I propose that /la/ in the alternating Creoles, but not /a/, carries a lexical constraint violation of the constraint STEM-FINAL-NOCODA as part of the input. (14) STEM-FINAL-NOCODA: 'Stems must end in an open syllable.' Align the right edge of the stem with the right edge of a syllable nucleus. STEM-FINAL-NOCODA is a Generalised Alignment (GA) constraint (McCarthy & Prince 1993a) and is a member of the family of NOCODA constraints (Ito & Mester 1999). An analogous constraint has been invoked in Hume (1998) to account for phrase-final metathesis in Leti (e.g., /urun/ -> [urnu] 'breadfruit'). 1 propose that the desideratum for alternating /la/ consists of its phonological form and a lexical violation of the constraint STEM-FINAL-NOCODA as shown in (15). (15) Desideratum for alternating [la] /la/

STEM-FINAL-NOCODA

Note that the lexical violation of STEM-FINAL-NOCODA can only be respected through the presence of a stem-final coda in the output. Resyllabification from the stem into DET is prevented through the GA constraint R-ALIGN-STEM-SYLL given in (16). (16) R-ALIGN-STEM-SYLL: 'Stems must end in a syllable.' Align the right edge of the stem with the right edge of a syllable. Given LRPM, the typology of DET in French-lexified Antillean Creoles is represented through the inputs in table 2. Table 2: Typology of DET inputs Alternants [a] [la] [la]/[a]

Input /a/ /la/ /la/ with lexical constraint violation·, /a/

Creoles Guyanese Guadeloupean; Ν Dominican Haitian; S Dominican; St. Lucian etc.

According to table 2, the alternating varieties of French-lexified Creoles have two input allomorphs for DET (cf. Carrington 1984: 55). In the present proposal, the markedness of the morphophonology of alternating DET is expressed through the presence of a lexical constraint violation on /la/, whereas /a/ lacks this part of the lexical representation. Nonalternating /la/ in varieties such as Guadeloupean is represented in the input only through

220

Thomas Β. Klein

its phonological form and, thus, is predicted not to show marked behaviour like in the alternating Creoles. Similarly, homophonous items like the adverbial la have no lexical constraint violation and, hence, are predicted to show no morphophonology out of the ordinary in any of the Creoles. Overall, there is a close match between the number of DET inputs and its surface typology. The presence of only one DET input corresponds to a single DET allomorph at the surface as in Guyanese, Guadeloupean and Northern Dominican, whereas two DET inputs correspond to two surface alternants as in Haitian, Southern Dominican and St. Lucian.

5.2. The distribution of V and CV allomorphs The input representations and constraints introduced to this point suffice to arrive at the correct distribution of DET in environments with consonant sequences and vowel hiatus. The tableau in (17) shows how the preference for vowel hiatus is obtained in the LRPM system. Input 1: /papa+la/

STEM-FINALNOCODA *

Input 2: /papa+a/ Candidates:

R-ALIGN-

MAXIO

ONSET

STEM-SYLL

a. [pa.pa.+la][ nDutl b. [pa.pa.+a] InDUt2

STEM-FINALNOCODA

*! *

Both DET allomorphs are available in the input, as shown in the upper half of the tableau in (17). /la/ carries the lexical violation of STEM-FINAL-NOCODA whereas /a/ carries no such violation. The constraint hierarchy evaluates forms based on both input representations of DET as shown in the lower half of tableau (17). Candidate (17a) makes perfect sense from the point of view of syllable markedness because it employs CV syllables at the juncture between the stem and DET, and yet this form must be ruled out in the alternating Creoles. In the present system, (17a) is suboptimal and, hence, correctly ungrammatical because the lexical constraint violation from the /la/ input is not respected in the output. The stem in (17a) ends in an open syllable which means that the input violation of STEM-FINALNOCODA has been cancelled. This cancellation, indicated by putting the violation mark between angled brackets, represents a departure from the input and, consequently, is a violation of the faithfulness constraint MAXIO. Given that MAXIO is ranked fairly high, this faithfulness violation renders the form suboptimal in comparison to the vowel hiatus. The form with vowel hiatus from the /a/ input in (17b) is optimal even though it violates the syllable structure constraint ONSET, /a/ has no input constraint violation and, hence, passes the higher-ranked constraint MAXIO. AS a result, vowel hiatus is correctly optimal in comparison to CV syllables at the juncture of DET and the stem. The same system of representations and constraints is able to correctly favour consonant clusters with /la/ in the alternating Creoles.

Syllable structure and lexical markedness in Creole

(18)

221

Input 1: /liv+la/

STEM-FINALNOCODA *

Input 2: /liv+a/ Candidates:

R-ALIGN-STEM-SYLL

MAXLO

ONSET

STEM-FINAL-

NOCODA β"

*

a. [liv.+la]i n p Ut i

b. [li.V+laJinputi c. [li.v+a] InDUt2

*

*!

*

*!

d. [liv.+a]InDU,2

*

*!

The choice of the V allomorph as the exponent of DET is correctly suboptimal in the alternating Creoles after stem-final consonant or glide. Alignment of the end of the stem with a syllable is suboptimal in this environment because the V allomorph of DET induces a fatal violation of ONSET as shown in (18d). Resyllabification to form a CV syllable, as seen in (18c), is suboptimal because it causes the high-ranking constraint R-ALIGN-STEM-SYLL to be fatally violated. Analogously, resyllabification with the CV allomorph of DET is also suboptimal as shown in (18b). The heterosyllabic consonant cluster with the CV allomorph of D E T correctly emerges as optimal as seen in (18a) because this form passes the two high-ranking syllable structure constraints and respects the lexical constraint violation. Note that the NOCODA constraint adduced in (8) above is ranked low so that it plays no role in the morphophonology of DET. The tableaux in (19) and (21) are presented to give a sense of the morphophonology of D E T in the non-alternating Creoles. (19) D E T as non-alternating CV (Guadeloupean, Northern Dominican) Input: /papa+la/

R-ALIGN-STEM-

MAXIO

ONSET

STEMFINAL-

SYLL

NOCODA

a. [pa.pa.+la] b. [pa.pa.+a]

*!

*

In the non-alternating Creoles, only one form of D E T is available in the input. None of these inputs contains a lexical constraint violation. Thus, in the Creoles preferring CV throughout, a V allomorph of DET can only result through the deletion of input /l/ which results in a fatal violation of the high-ranking faithfulness constraint MAXLO, as shown through the comparison of (19a) and (19b). Analogously, Creoles with V throughout could exhibit ungrammatical ""[la] only through the insertion of [1]. This insertion may easily be ruled out by ranking an appropriate faithfulness constraint from the DEPLO family such as DEPIO-C in (20) sufficiently high. (20) DEPIO-C: 'NO consonant insertion' (cf. McCarthy & Prince 1995, among others) Output consonants must have input correspondents.

222

Thomas B. Klein

(21) DET as non-alternating V (Guyanese) Input: /papa+a/

R-ALIGN-

MAXIO

; DEPIO-C

ONSET

a. [pa.pa.+la] b. [pa.pa.+a]

STEM-FINALNOCODA

STEM-SYLL

:

*! *

In the Creoles preferring V throughout, a CV allomorph of DET can only result through the insertion of [1], However, this insertion is correctly ruled out through the ranking of DEPIOC, the constraint prohibiting insertion of consonants, above the constraint demanding onsets. Checking the analytical objectives outlined at the beginning of this section, the analysis in this section has accounted for the anti-markedness effect as an outcome of marking DET through a lexical constraint violation in the input. Homophonous forms such as adverbial la and non-alternating forms of DET simply lack this part of the input representation and have been shown to surface correctly. To complete the analysis, the next section demonstrates how glide insertion supplies homorganic onsets.

5.3. Agreement and onset-driven glide insertion It has been noted above that the allomorphy of DET in French-lexified Antillean Creoles exhibits conflicting trends in terms of syllable structure. On the one hand, heterosyllabic consonant clusters and vowel hiatus are preferred in certain environments, on the other hand glide insertion supplies onsets with stem-final front and round back vowels to result in unmarked CV syllables. The analytical challenge is to demonstrate why glide insertion takes place with front and round back vowels, but not with stem-final /a/. The fundamental idea that I would like to invoke is that glide insertion is demanded by the ONSET constraint as argued in Rosenthall (1994), but that it is also counteracted by agreement constraints. These agreement constraints demand a match between distinctive features of the stem-final vowel and the inserted glide. Homorganic glide insertion after stem-final /a/ is not possible in this approach because there is no glide available to meet the agreement requirements in this particular environment. I invoke the agreement constraints in (22a) and (22b) based on the syntagmatic constraints proposed in Pulleyblank (1997). (22) Featural agreement constraints a. AGR-FRONT: A vowel and a following glide must agree in [FRONT], b. AGR-ROUND: A vowel and a following glide must agree in [ROUND]. The demand for onsets outranks the avoidance of inserted segments expressed through the constraint DEPIO in (23). (23) DEPIO: 'No insertion' (McCarthy & Prince 1995) Output segments must have input correspondents.

223

Syllable structure and lexical markedness in Creole

The tableaux accounting for homorganic glide insertion in (24) - (26) are simplified in that only the relevant constraints and the /a/ input for DET are shown. (24)

Input: /papje+a/ a. [pa.pje.+ja] b. [pa.pje.+a] c. [pa.pje.+wa]

AGR-FRONT i

AGR-ROUND

ONSET

DEPIO *

#!

!

*!

*

*

The vowel-glide sequence at the DET-stem juncture in (24c) disagrees in frontness and roundness and is, hence, correctly suboptimal. There is no featural disagreement in the hiatus form in (24b), but there is also no onset for DET. The insertion of the homorganic glide [j] in the onset of DET, which agrees in the feature [FRONT] with the stem-final front vowel, is correctly optimal even though material has been inserted, as shown in (24a). Note that forms such as (24a) pass AGR-ROUND trivially because there is no [ROUND] feature in the adjacent vocoids. The account of the insertion of [w] after round stem-final vowels is analogous to the insertion of [j] above. The homorganic glide is the best option given the agreement constraints and the demand for a syllable onset, as shown in the tableau in (25). (25)

Input: /bato+a/ a. [ba.to.+ja]

AGR-FRONT

*|

· AGR-ROUND

J

*

ONSET

*

b. [ba.to.+a] c. [ba.to.+wa]

(26)

Input: /papa+a/ a. [pa.pa.+a] b. [pa.pa.+ja] c. [pa.pa.+wa]

DEPIO

*

AGR-FRONT

J AGR-ROUND

ONSET

DEPIO

* *

;

*i

*

Glide insertion with a-final stems is worse than vowel hiatus because no glide agrees with a in terms of [FRONT] or [ROUND], Consequently, [j]-insertion introduces disagreement with respect to the feature [FRONT], whereas the insertion of [w] disagrees in terms of [ROUND], as shown in (26b) and (26c), respectively. Vowel hiatus with stem-final a as in (26a) is optimal because it avoids violating the agreement constraints, even though it is at odds with the lower-ranking ONSET constraint. In sum, glide insertion is driven by the demand for onsets, but is suspended where homorganic glides are not available.

5.4. Wider functionality Given the analysis of DET allomorphy proposed in this section, the resulting constraint hierarchy is as in (27).

224

Thomas Β. Klein

(27) Partial constraint hierarchy for French-lexified Antillean Creoles R-ALIGN-STEM-SYLL, AGR-FRONT, AGR-ROUND » M A X I O , D E P I O - C » ONSET » DEPIO, STEM-FINAL-NOCODA » NOCODA

Note that the constraint hierarchy in (27) is identical in alternating and non-alternating Creoles. The typology of the allomorphs of DET results from this constraint hierarchy together with the input representations summarised in table 2. Given the number of constraints employed in the present analysis, one may wonder about the utility of the constraints in (27) beyond the present account of DET allomorphy. Note that the alignment c o n s t r a i n t R-ALIGN-STEM-SYLL, t h e f a i t h f u l n e s s constraints MAXIO, DEPIO and DEPIO-C

and the syllable structure constraints ONSET and NOCODA are all very basic and are commonly invoked in OT analyses of a great variety of languages. Given that homorganic glide insertion is quite widespread in the languages of the world (see Rosenthall 1994 for discussion), the approach invoking agreement constraints proposed in this paper has considerable potential for cross-linguistic utility. Thus, there are other cases of homorganic glide insertion matching the Antillean facts very closely. In Dutch, for instance, homorganic glides are inserted in vowel hiatus contexts except after /a/ (Booij 1995: 6 5 f f ) . Finally, the constraint STEM-FINAL-NOCODA is a straightforward member of the NOCODA constraint family (Itö & Mester 1999) and, furthermore, akin to Hume's (1998: 164) constraint for Leti phrasefinal metathesis. Interestingly, versions of the anti-markedness effect found with Antillean French-lexified DET allomorphy are observed in the morphophonology of other Creole and non-creole languages, lending further utility to the LRPM analysis proposed in this paper. According to Huttar (1996), the appearance of the transitive suffix -mi in the Atlantic English-lexified Creole language Ndyuka depends on a complex set of phonological conditions. Verbs after which this suffix may occur are nearly always monosyllabic and must have a final high tone. In addition, -mi is found before constituents of the object phrase if they are monosyllabic and vowel-initial, but not if they are polysyllabic or consonant-initial. Consider the minimal pair in (28), taken from the data in Huttar (1996). (28) Morphophonology of -mi in Ndyuka (fon 'hit') a. a fon den pikln 'he hit the children' b. a fo+mi en pikin 'he hit his child' The comparison between (28a) and (28b) shows that -mi in Ndyuka displays an anti-markedness effect similar to DET in Antillean French-lexified Creoles, -/π/does not occur where a CV structure would have been obtained at the juncture with the consonant-initial object phrase, as in (28a), but it does occur in the vowel hiatus environment in (28b). This minimal pair shows that an account invoking LRPM is likely to play a central role in a viable OT analysis of the transitive suffix -mi in Ndyuka. Lapointe (2001) has reported facts from Korean which are strikingly similar to the allomorphy of Antillean DET although he is not aware of these data. On the one hand, as expected under conventional OT, the Korean topic-focus marker -unt-nun and the accusative marker -uil-lui appear in their C-initial forms when attached to V-final stems and in their V-initial forms when attached to C-final stems (cf. Kim-«/ 'Kim-Acc' and ai-lul 'childAcc'). On the other hand, the Korean nominal conjunctive particle - wal-kwa behaves un-

Syllable structure and lexical markedness in Creole

225

expectedly. Under conventional OT, we would expect that the C-initial allomorph -kwa attaches to V-final stems and that the glide-initial - wa attaches to C-final nouns. However, we find just the opposite, e.g., papvkwa 'rice and' and ai-wa 'child and'. As a solution, Lapointe (2001) proposes to employ pre-OT morpholexical listing as exemplified in (29). (29) Morpholexical listing in OT (Lapointe 2001: 273) Korean nominal conjunctive suffix: + /kwa/ /[+cons] + /wa/ /[-cons] According to Lapointe (2001), morpholexical restrictions such as (29) function to bleed the inputs available to GEN. In his model, only forms meeting the lexically and morphologically listed restrictions are used as inputs to GEN. Lapointe's model seems retrograde because it mixes aspects of rule-based phonology and morphology with an OT-type grammar, resulting in an exponential increase in the power of the model. The LRPM model advocated in this chapter could easily accommodate the Korean data and avoids the drastic step of mixing rule-based phonology and morphology with OT. Instead, it addresses the issue of anti-markedness effects employing essential OT techniques, that is, only general and crosslinguistically valid constraints, constraint violations and faithfulness. Other competing approaches such as the Sympathy model (McCarthy 1999) also increase the power of the OT grammar much more fundamentally. The set of Sympathy constraints is added to the set of conventional constraints, resulting in a significant inflation of the factorial typology. In contrast, only standard constraints are used to encode input representations in LRPM and their number does not increase in comparison to the conventional model. In sum, the constraints and the LRPM approach invoked in the present analysis are not specific to DET allomorphy, but instead can easily be shown to be useful in other languages. Given the analysis of creole morphophonology presented in this section, it is instructive to return to the theme of the supposed simplicity of creole languages.

6. Complex morphophonology in simple Creoles?

Recently, McWhorter (2001a, b) has made a much noted argument designed to demonstrate the supposed simplicity of creole grammars. McWhorter (2001a) includes the following claims. "[C]reoles display less complexity than the rest of the world's natural grammars, (p. 133) [T]he

least complex grammars in the world are all Creoles. [...] [I]f all of the world's languages could be ranked on a scale of complexity, there would be a delineable subset beginning at the "simplicity" end and continuing towards the "complexity" one all of which were Creoles. [...] [I]n the final analysis, there would be a healthy band of languages beginning at the "simplicity" pole which would all be Creoles." (p. 162) McWhorter (2001a) makes two caveats to his hypothesis that "the world's simplest grammars are Creole grammars" that are important in the context of the present paper. First,

226

Thomas Β. Klein

according to McWhorter (2001a: fn. 20), French-lexified Creoles which have heavily borrowed from the French lexicon are to be excluded from the list of Creoles with simple grammars. Unfortunately, McWhorter (2001a) does not provide an explanation for the special status of French-lexified Creoles in terms of their lack of simplicity. Secondly, McWhorter (2001b: 391) admits that the phonology of creole languages had been addressed superficially and preliminarily in McWhorter (2001a) and that non-creole languages may easily have less complex phonologies than Creoles. Thus, McWhorter's (2001a, b) claims concerning the supposed simplicity of creole grammars appear to be compromised as far as Creole phonology is concerned. A fortiori, the results presented in the current chapter represent evidence against McWhorter's simplicity hypothesis. Thus, this paper, relying to a considerable degree on earlier research, has demonstrated that the morphophonology of the definite determiner in Antillean French-lexified Creoles is of considerable complexity. There are at least three intricate processes involved in arriving at the surface shape of DET, V/CV alternation, homorganic glide insertion and nasalisation, and they may exhibit conflicting trends and anti-markedness effects. Furthermore, the complex morphophonology of -mi in Ndyuka summarised above suggests that English-lexified creole languages that are closely related to McWhorter's prime example of Saramaccan have complex morphophonologies as well.

7. Conclusion

The phonology and morphology of creole languages has been one of the least studied areas in linguistics to date. This volume has been designed in part to redress this imbalance. The present paper has shown that the detailed investigation of the morphophonology of creole languages brings to light intricate patterns of interest and significant consequence to descriptive linguistics, typology, phonological and morphological theory and to general claims concerning the structure of creole languages. I have demonstrated in this paper that the morphophonology of the definite determiner in Creoles such as Haitian presents a challenge to conventional Optimality Theory and I have proposed to resolve this challenge by invoking the Lexical Representation as Pure Markedness model developed in Klein (2000). The preference for consonant sequences and vowel hiatus instead of less marked CV structures in the allomorphy of the creole definite determiner is an anti-markedness effect that represents an interesting issue for any formal theory of phonology and morphology. This anti-markedness is counteracted in part by the creation of unmarked syllable onsets via glide insertion. The analytical challenge has been to work out a model that is able to balance these conflicting trends correctly. It is hoped that the present paper contributes to showcasing creole phonology and morphology as a legitimate and rewarding area of inquiry for creolists and theoretical linguists alike.

Syllable structure and lexical markedness in Creole

227

References Aceto, Michael (1996): Early Saramaccan syllable structure: An analysis of Schumann's 1778 manuscript. - In: Journal of Pidgin and Creole Languages 11, 23-44. Alber, Birgit and Ingo Plag (2001): Epenthesis, deletion and the emergence of the optimal syllable in Creole: The case of Sranan. - In: Lingua 111,811-840. Archangeli, Diana and D. Terence Langendoen (eds.) (1997): Optimality Theory. An overview. Oxford: Blackwell. Bernabe, Jean (1987): Grammaire Creole. Fondas Kreyol-la. Elements de base des Creoles de la zone americano-caraibe. Paris: L'Harmattan. Bhatt, Parth and Emmanuel Nikiema (2000a): Nasality and the definite determiner in St. Lucia Creole. - In: Proceedings of the 13th biennial conference of the Society of Caribbean Linguistics, Jamaica, 61-73. - (2000b): Le Statut de la nasalite en crdole de Sainte-Lucie. - In : Revue quebecoise de linguistique 28, 23-45. Booij, Geert (1995): The phonology of Dutch. Oxford University Press. Cadely, Jean Robert (1995): Elision et agglutination en Creole hai'tien: Le cas des pronoms personnels. - In: Etudes Creoles 18, 9-38. - (2002): Le Statut des voyelles nasales en Creole hai'tien. - In: Lingua 112, 435-464. Carrington, Lawrence (1984): St. Lucian Creole. A descriptive analysis of its phonology and morphosyntax. Hamburg: Buske. Fournier, Robert (1978): De quelques anomalies dans le traitement de l'article defini par H. Tinelli (1970): Generative phonology of Haitian Creole. - In: Amsterdam Creole Studies 2, 101-113. Golston, Chris (1996): Direct Optimality Theory: Representation as Pure Markedness. - In: Language 72, 713-748. Hammond, Michael (1997): Optimality Theory and prosody. - In: D. Archangeli and T. Langendoen (eds.), 33-58. Holm, John (1988): Pidgins and Creoles. Volume I: Theory and structure. Cambridge: Cambridge University Press. - (2000): An Introduction to Pidgins and Creoles. Cambridge: Cambridge University Press. Hume, Elizabeth (1998): Metathesis in phonological theory: The case of Leti. - In: Lingua 104, 147186. Huttar, George (1996): Epenthetic -mi in Ndyuka: A transitive marker? - In: SIL Electronic Working Papers 1996. Itö, Junko and R. Armin Mester (1999): Realignment. - In: R. Kager et al. (eds.): The prosody-morphology interface, 188-217. Cambridge: Cambridge University Press. Kager, Ren6 (1999): Optimality Theory. Cambridge: Cambridge University Press. Kaye, Jonathan, Jean Lowenstamm and Jean-Roger Vergnaud (1990): Constituent structure and government phonology. - In: Phonology 7, 193-231. Klein, Thomas (2000): Umlaut in Optimality Theory. A comparative analysis of German and Chamorro. Tiibingen: Niemeyer. Lapointe, Steven (2001): Stem selection and OT. - In: G. Booij & J. van Marie (eds.): Yearbook of morphology 1999, 263-297. Dordrecht: Kluwer. McCarthy, John (1999): Sympathy and phonological opacity. - In: Phonology 16, 331-399. McCarthy, John and Alan Prince (1993a): Generalized Alignment. - In: G. Booij & J. van Marie (eds.): Yearbook of morphology 1993, 79-153. Dordrecht: Kluwer. - (1993b): Prosodic morphology: Constraint interaction and satisfaction. - In: Rutgers Optimality Archive #482.

228

Thomas Β. Klein

-

(1994): The emergence of the unmarked: Optimality in prosodic morphology. - In: M. Gonzalez (ed.): Proceedings of the North East Linguistic Society 24, 333-379. University of Massachusetts, Amherst: GLSA. - (1995): Faithfulness and reduplicative identity. - In: J. Beckman et al. (eds.): Papers in Optimality Theory, 249-384. University of Massachusetts, Amherst: GLSA. McWhorter, John (2000): The missing Spanish Creoles. Recovering the birth of plantation contact languages. Berkeley: University of California Press. -

( 2 0 0 1 a ) : T h e world's simplest grammars are Creole grammars. - In: Linguistic

Typology

5, 125-

166. - (2001b): What people ask David Gil and why: Rejoinder to the replies. - In: Linguistic Typology 5, 388-412. Nikiema, Emmanuel (1999): De la variation du diterminant /la/ dans les Creoles ha'itien et St-Lucien. -In-.Lingua 107, 69-93. Nikiema, Emmanuel and Parth Bhatt (this volume): Two types of R deletion in Haitian, 43-69. Prince, Alan and Paul Smolensky (1993): Optimality Theory: Constraint interaction in generative grammar. Technical report no. 2, Rutgers University Center for Cognitive Science. Pulleyblank, Douglas (1997): Optimality Theory and features. - In: D. Archangeli & T. Langendoen (eds.), 59-101. Romaine, Suzanne (1988): Pidgin and creole languages. London: Longman. Rosenthall, Samuel (1994): Vowel/glide alternation in a theory of constraint interaction. Ph.D. dissertation. University of Massachusetts, Amherst. - In: Rutgers Optimality Archive #126. Rubach, Jerzy, and Geert Booij (2001): Allomorphy in Optimality Theory: Polish intonation. - In: Language 77, 26-60. Sabino, Robin (1993): On onsets: Explaining Negerhollands initial clusters. - In: F. Byme & J. Holm (eds.): Atlantic meets Pacific. A global view of pidginization and creolization, 37-44. Amsterdam: Benjamins. Smolensky, Paul (1996): The initial state and 'richness of the base' in Optimality Theory. Ms., Johns Hopkins University, Baltimore. - In: Rutgers Optimality Archive #154. Sprouse, Ronald (1997): A case for enriched inputs. Ms., University of California, Berkeley. - In: Rutgers Optimality Archive #193. Stiebels, Barbara and Dieter Wunderlich (1999): Second stems in Hungarian nouns. - In: The Linguistic Review 16, 253-294. Taylor, Douglas (1977): Languages of the West Indies. Baltimore: The Johns Hopkins University Press. Tinelli, Henri (1981): Creole phonology. The Hague: Mouton. Tranel, Bernard (1994): French liaison and elision revisited: A unified account within Optimality Theory. - In: Rutgers Optimality Archive #15. - (1995): Exceptionality in Optimality Theory andfinal consonants in French. - In: Rutgers Optimality Archive #61. Valdman, Albert (1978): Le creole: Structure, Statut et origine. Paris: Klincksieck. Zoll, Cheryl (1998): Parsing below the segment in a constraint-based framework. Stanford: CSLI.

Thomas Klein, Georgia Southern University, Department of Writing and Linguistics, P.O. Box 8026, Statesbor ο, GA 30458, USA, Email: [email protected]

Section 4: Derivational morphology

Margot van den Berg Early 18 th century Sranan -man

1. Introduction1

While Sranan is relatively well documented in its early stages in comparison with other Creole languages, many of its early sources derive from white authors. In addition to that, virtually no records are available for the pre-1765 period. It is important therefore, to supplement the corpus of early Sranan texts both with regard to type of source and with regard to year of writing. Early court records may be particularly useful in both regards, since (parts of) the testimonies of blacks are occasionally reproduced in Sranan. The great majority of these records are contained in the archives of the Hof van Politie en Criminele Jusititie (Court of Police and Criminal Justice), while some are also stored in the archives of the Societeit van Suriname (Society of Suriname). Both archives are stored in the Algemeen Rijksarchief (State Archives) in The Hague. These records, covering the 100-year period between 1667, when Suriname came into Dutch hands, and 1767, consist mainly of depositions, statements and reports of examinations in which some 500 Sranan words (tokens) and 54 Sranan passages were found (Van den Berg 2000, Van den Berg and Arends 2001). Table 1: Chronological distribution of Sranan words found in Court Records (1667-1767) Year

Code

Inventory Number

Number of Sranan words (tokens)

1707- 1730

1.05.10.02 1.05.04.01 1.05.10.02 1.05.10.02 1.05.04.03 1.05.10.02 1.05.04.06 1.05.10.02 1.05.04.06

784, 786, 787 231,234, 239 787, 787 (II), 788, 789, 790, 791, 792, 793 795, 797, 798, 801, 929, 937, 938 272, 274,275, 285 803, 932, 938, 939, 942, 947, 948 296,307,308, 309,310 788, 805, 806, 808, 809, 811, 812, 813 312,313,315,317, 320,330,331

6 36 52 128 9 133 20 88 91 Total

1731 - 1740 1741 - 1750 1751 - 1760 1761 - 1767

1

564

I would like to thank Adrienne Bruyn, Enoch Aboh, Michel DeGraff, Jacques Arends, Lilian Adamson, Bettina Migge, Ingo Plag, the anonymous reviewers and the participants of the Siegen workshop for their comments on earlier versions of this paper. Many thanks to Norval Smith. All faults are mine, of course.

Margot van den Berg

232

One reason to assume that court records contain accurate reflections of the language as it was actually spoken by the blacks, is the fact that, by their very nature, they are (intended as) verbatim accounts of what was actually said during interrogations. And even if the ideal of 100% verbatim reporting was not always achieved, the data may be assumed to reflect natural speech reasonably well. A further indication of the reliability of these records can be found in some linguistic characteristics of the textual data itself. For example, in the court records the word kaba rather than arede is used consistently to express the notion of 'already' (1). In his (1783:50) Sranan dictionary, Schumann comments on the use of arede with the remark da Bakkratongo ('that is White Sranan'), proceeding with Ningre takki "kaba" (blacks say "kaba"). (1)

mi doe langa hem caba (CR 1.05.10.02, 798, f-v°; 1745) I do with him already 'I'm finished with him already.'

Another indication of the unabridged nature of these records may be found in the repeated use of a vulgar term of abuse such as you mama pima ('your mama's cunt'). Finally, on a phonological level, the occasional use of epenthetic vowels in words such as bakara 'white person' and masara 'master' (next to bakra and masra) may also be indicative of the authenticity of the language, since this type of epenthesis appears to have been typical of blacks, especially plantation slaves (cf. Focke 1855: XII; see also Smith, this volume, and Lappe and Plag, this volume). Another type of document that may serve to supplement the corpus of early Sranan texts both with regard to authorship and with regard to the year of writing is the recently discovered 18th century Sranan version of the Saramaka Peace Treaty (1762). While the Dutch version of the treaty has been known ever since it was published in Hartsinck (1770: 802809), the Sranan version remained unknown until a transcription was published in November 2000 by Wim Hoogbergen and Thomas Polime. They claim that this is the text as it was actually read to the Saramaccans by the Dutch authorities on 18 or 19 September 1762. Apart from the text of the fifteen clauses of the treaty, the document also contains the Saramaccans' response (in Sranan) to each clause. The text, which contains approximately 1,900 words, is located in the Oud-Archief Suriname / Hof van Politie en Criminele Justitie (ARA, code 1.05.10.02, number 66), as an appendix to the minutes of the Court of Police of December 27, 1762. There is no doubt about the authorship of the text. In the preamble, Louis Nepveu, the leader of the Dutch delegation, is identified as such: (2)

A poti alle dissi santie deja na inni gi dem foe hakisi effi dem wandi holli dati alle, he put all this thing here PREP in give they for ask if they want hold that all 'He put all these things in here for them in order to ask if they are willing to hold to all that.' (1762, SPT 1.05.0.02, 66, f 177 v°)

Louis Nepveu was a brother of the later Governor Jan (or Jean) Nepveu, the author of 'Annotations' on Herlein's 1718 book (see Arends and Perl 1995). If Louis, like brother, came to Suriname early in life (their Huguenot parents had been living there some time when Jan arrived in 1734 (Gordijn et al. 1977:424)), it may tentatively be

the his for as-

Early 18th century Sranan -man

233

sumed that he had a good knowledge of Sranan. As to the Saramaccans' responses, I assume that these were also recorded by Nepveu. It seems, therefore, that both the treaty and the Saramaccans' response are second language varieties of Sranan; one from a speaker of Dutch (and, perhaps, French), the other from speakers of Saramaccan. As to the latter, however, it is known that the Saramaka Maroons, due to their frequent interactions with slaves, had a reasonable knowledge of Sranan (after all, if this were not the case, the treaty would not have been read to them in Sranan).2 In both texts, epenthesis is encountered: garanman 'chief (1762, SPT 1.05.10.02, 66, f 180 r°) occurs as well as granman (1762, SPT 1.05.10.02, 66, f 177 v°); coertoe 'court' (1762, SPT 1.05.10.02, 66, f 177 v°) occurs alongside couroetoe (1762, SPT 1.05.10.02, 66, f 180 r°) (see Smith, this volume). Since epenthesis could be indicative of the authenticity of the language, it supports the claim that this document is the text as it was actually read to the Saramaccans. The inconsistency seen in epenthesis and the spelling in general may reassure us as far as the lack of influence of any standard spelling. In this paper some data from the court records and the Saramaka Peace Treaty will be examined from the perspective of word formation. Not surprisingly, compounds such as [noun-noun]N, [adjective-noun]N and [verb-noun]N (i.e. concatenations of lexical items) are encountered most frequently - compounding is, after all, the unmarked way to extend the lexicon (Smith 2001: 52). Table 2 is a selection of the compounds encountered in the court records (1667 - 1767) and the Saramaka Peace Treaty (1762).

2

Even today Sranan functions as a lingua franca in Surinam (Adamson 2001), although its use seems to have decreased since the 1950s. In those days, as much as 90% of the entire population of Surinam could express themselves in (a variety of) Sranan, and understand Sranan, while only 50% of the population could speak and understand Dutch (Adamson 2001: 295).

234

Margot van den Berg

Table 2: A selection of compounds in Court Records and the Saramaka Peace Treaty [noun-noun] year, source

[adjective-noun] year, source

[verb-noun] year, source

koebiston 'limestone in balancing organ of a koebi (sp. fish)'

bitta wirriwirri 'bitter leaves' plant

kweekooko 'cultivated okra'

1738, CR 1.05.10.02, 792, f 2 v°

1745, CR 1.05.10.02, 798, f - v °

1735, CR 1.05.10.02, 798, f 17 v°

papamonie 'cowry shell'

astranti moffo 'big mouth, cheeky'

ronnewee nengre 'runaway bl.'

1742, CR 1.05.10.02, 794, f - v °

1762, SPT 1.05.10.02,66, f 181v°

1762, SPT 1.05.10.02, 66, f 180 v°

wataamamas 'water goddess'

reijtie pikien '(natural) child'

crabbejas 'yaws' (disease)

1742, CR 1.05.10.02, 795, f54 v°

1762, SPT 1.05.10.02, 66,fl81v°

1762, CR 1.05.10.02, 8 0 6 , f - v °

agoema wieri wieri 'a. leaves' 1752 CR 1.05.10.02, 8 0 3 , f l 0 9 v °

Saramaca vrieman 'SA. freeman' 1762, SPT 1.05.10.02, 66, f 181 v°

Saranam condre 'Surinam' 1762, SPT 1.05.10.02, 66, f 181 v°

maka klossi 'type of linen' 1766, CR 1.05.10.02, 812, f - v °

Most compounds are endocentric and right-headed, i.e. the left member of the compound modifies the right member. A clear example is presented by Schumann (1783: 78). While bom calabasi is the word for the fruit of a calabash tree, the word for the tree bearing calabash fruit is calabasi bom. Semantically speaking, the whole compound can be interpreted as a hyponym of the head. Since the word order of the compound follows the word order of the syntactic phrase, which, in turn, follows the word order in the lexifier, the lexifier and the creole share the position of the head in the compound (Lefebvre 1998). However, in the data from the court records and in Schumann's (1783) dictionary (Bruyn 1989) a few exceptions to this pattern occur (table 3):

Early 18fh century Sranan -man

235

Table 3: Exceptional compounds: 'Left-headed' compounds and 'Mobile Heads' 'Left-headed' compounds watra muffe 'saliva' Schumann 1783: 299 flamm faija 'fire flame' Schumann 1783: 100 'Mobile Heads' boombamba 'sp. tree' (Ocotea) 1738, CR 1.05.10.02, 792, f l v° houtbamba 'wood of sp. tree' (Ocotea) 1732, CR 1.05.10.02, 787 (II), f 3 1 v° horrowatra 'fountain' Schumann 1783: 133 (preferred)

bambaboom 'tree' (Van Donselaar, 1989:59) (20 th century Surinam-Dutch) bambahout 'wood of sp. tree' (Ocotea) 1732, CR 1.05.10.02, 787 (II), f 2 5 v° watra horro 'fountain' Schumann 1783: 133

hai watra 'drop of water, tear' Schumann 1783: 198 sakka janjam 'stomach' Schumann 1783: 238

watra ai 'tear' Snijders 2000: 172 janjam-sakka 'stomach' Schumann 1783: 59

The words bambahout and houtbamba are both found in the interrogation of the slaves Lalande and Mainby. Lalande is charged with poisoning and Mainby is a witness. The compound houtbamba occurs in the minutes of the questioning of Mainby, while bambahout is encountered in those of Lalande. The latter are difficult to transcribe due to the state of the paper. The former, however, are in a better state: (3)

Segt van Lalande verstaen te hebben dat hetselve houtbamba genaamt werd, en op hooge gronden groeyt. (Mainby, 1732, CR 1.05.10.02, 787 (II), f 3 1 v°) 'He says that Lalande told him that it is called houtbamba and grows on high ground.'

These examples show that there is at least some variation in the structure of compounds which, as far as I know, cannot be accounted for in terms of relexification 4 (Lefebvre

3

Both compounds, aywatra

and watra ai, occur in 20 th century Sranan. The first is reported by

Michael Ietswaart and Vinije Haabo (1999) in their language manual Sranantongo. voor reizigers

en thuisblijvers.

Surinaams

(Amsterdam: Jan Mets). The second is listed by Ronald Snijders

(2000). The authors do not mention both. 4

Under relexification as outlined by Lefebvre (1998), the word order in Haitian compounds is explained as the result of the word order in Haitian syntactic phrases, which, in turn, follows the order in

French syntactic phrases (French is the lexifier of Haitian Creole). Thus, Haitian and

French compounds have the position of the head in common, since it is determined by the position of the syntactic head in phrases (Lefebvre 1998: 342).The examples in table 3 show that the head of several Sranan compounds can occur on the lefthand side (i.e. opposite to the position of the

236

Margot van den Berg

(1998), Muysken (1981), Muysken and Smith (1990)), nor can these examples be explained in terms of substrate influence or convergence/divergence: Both the lexifier's and the substrates' compounds are right-headed. In the remainder of this paper the status of the morpheme -man in Sranan words such as gadomcm 'priest/oracle' and voevoereman 'thief will be discussed: Is it grammatical or lexical? In contemporary Sranan it is a grammatical morpheme, i.e. a derivational affix that creates animate nouns from verbs or other nouns. It denotes an agent (someone who X's) or a possessed-possessor relation (someone who has/is X; someone who is related to X). Thus, words such as gadoman, voevoereman and skrifiman 'clerk' are the results of derivation in contemporary Sranan. What were they in the 18th century: compounds or derivations? According to Arends (1995b: 42), these derivations are alternatives for circumlocutions such as wan somma dissi loeke schribi zanti a person who looks written things 'a clerk' (i.e. skrifiman). Derivations as alternatives for circumlocutions would then become productive during the 18th century: Schumann (1783) and Wullschlägel (1856) have a preference for compounding with man 'man' to express the concepts that are described in terms of a circumlocution by Van Dyk (cl765): "It thus sees that the only productive way of making new nouns to refer to persons, namely by compounding (or suffixing) with man, is used only in a limited number of cases in Van Dyk. Since Schumann this device has become more fully productive" (Arends 1995b: 42). Here, the distinction between compounding and suffixation may not be very clear, but they are connected through grammaticalization in the following way: "The fact that man is used to refer to females as well as to males (e.g. helpiman lit. 'help-man' for 'midwife', Schumann p. 128) is the result of grammaticalisation process, which made possible the use of man as a gender-neutral agentive suffix" (Arends 1995b: 42). Based on an examination of the available 18th century data, in particular those from the court records and the Saramaka Peace Treaty, I will propose that the emergence of the gender-neutral suffix -man is not the result of the morphological change outlined above. There are no 18th century data available that record the stage in which the compounded genderspecific noun man grammaticalized into a gender-neutral affix morpheme -man in an unambiguous manner (section 2). In section 3, a more plausible explanation for the emergence of the affix -man is offered in terms of relexification, following Smith and Veenstra (1998). Grammaticalization theory holds that all linguistic expressions may be arranged on a grammaticalization cline from open-class lexical expressions to closed-class grammatical markers (Hopper and Traugott 1993, Bybee et al 1995). Grammaticalization "consists in the increase of the range of a morpheme advancing from a lexical to a grammatical or from a less grammatical to a more grammatical status, e.g. from a derivative formant to an inflectional one" (Kuryfowicz 1965: 69 [1975:52]). This development can be represented by the following cline or pathway, though keep in mind that it is a gross simplification: lexical item > clitic > derivative formant > inflectional formant

head in the English or Dutch compound/phrase), and that the head of several (other) compound can occur in either position.

Early 18lh century Sranan -man

237

The development of a 'clitic' 5 into an inflectional formant is also known as morphologization or univerbation, although the term morphologization is also used for the development of morphophonemic alternations (Hopper and Traugott 1993: 130 footnote). For reasons of clarity I prefer to stick with the more familiar term grammaticalization to denote the kind of clause internal morphological change discussed here, i.e. from the lexical item man 'man' via a reduced form -man in a compound to a derivative formative. Veenstra (2001) describes the development of the Saramaccan agentive affix -ma as follows: "Older form ma(n) was not an affix, but a word(/stem), and formed compounds (...) Later on, it has been grammaticalized into an affix" (Veenstra 2001).

2. Three patterns with -man

In the data from the court records and the Saramaka Peace Treaty, there are several Sranan nouns ending in : inherited words, such as paiman 'payment, reward', and kajeman 'caiman'; nouns ending in which may be inherited but could also be innovations (table 4) and nouns ending in that are clear-cut innovations i.e. independent instances of word formation (table 5).

2.1. Inherited words The first category of nouns ending in consists of inherited words such paiman 'payment, reward, debt' (1762, SPT 1.05.10.02, 66, f 181 v°), kajeman 'caiman' (1757, CR 1.05.10.02, 942, f 24 v°) and kaeijman 'caiman' (1757, CR 1.05.10.02, 942, f 27 v°). Here, the form should not be regarded as a morpheme -man, for these words are inherited (in toto). Although it is uncertain where and when the word kajemanlkaeijman 'caiman' originated, it is clear that its formation did not take place in Surinam. The form paiman probably originated in the English form payment (Focke 1855).

2.2. Inherited or innovative? The nouns in this category may be inherited or they may represent clear-cut innovations. Contrary to inherited words such as paiman, they denote persons, but, since corresponding forms are encountered in the lexifier (table 4), it is not clear to what extent their occurrence in the language can be contributed to the process of innovative word formation.

5

The term 'clitic' refers to a reduced lexical item in a compound.

238

Margot van den Berg

Table 4: Sranan words of ambiguous origin in Court Records and the Saramaka Peace Treaty bossimam 1761, CR contreman 1745, CR contriman 1745, CR

'bushman' 1.05.04.06, 313, f318 v',0 'countryman' 1.05.10.02, 798, f - v° 'countryman' 1.05.10.02, 798, f - v °

friman 'freeman' 1762, SPT 1.05.10.02, 66, f 182 v° wie man 'freeman' 1762, SPT 1.05.10.02, 66, f 181 v° heddeman 'head man, chief 1762, SPT 1.05.10.02, 66, f 178 v° hedie man 'head man, chief 1762, SPT 1.05.10.02, 66, f 180 r"

Although the corresponding forms in Sranan and English may look quite similar, there are differences in semantics. Note, for example, that contriman not only refers to someone from the same country (cf. English countryman). According to Schumann (1783: 86) kondre can also refer to the world, a city, a village or a plantation - 1 assume that example (4) illustrates the latter. It is a Dutch sentence featuring a Sranan borrowing, taken from the minutes of the interrogation of the slave Primo in 1745. Primo is accused of poisoning several slaves of his plantation, among them Coridon. Example (5) is taken from the Saramaka Peace Treaty (1762). It illustrates the use of condre meaning village. These additional meanings of Sranan contreman are not shared by English countryman. (4)

Gij kent de neger Coridon van de Heer D. Ja dat is mijn contreman. 'Do you know the black Coridon of Master D.? Yes, he is my contreman. , (Primo, 1745, CR 1.05.10.02,798, f - v°; underlining mine, MB)

(5)

Dem sa moesoe sorie alle dem condre na Bacara. They FUT must show all their villages to white man 'They will have to show all their villages to the whitemen' (1762, SPT 1.05.10.02, 66, f 180 v°; underlining mine, MB)

2.3. Innovation Nouns ending in of the third category are presented in table 5. They are true innovations; they do not occur in the lexifiers, nor in the substrate languages, and, thus, reflect the autonomous character of the language. Since word forms that correspond to adjectives in the European lexifiers have been shown to act as verbs in a number of Creole languages, 6 1 have momentarily excluded forms that could be analyzed as [adjective]-/wa«.

6

See for example Voorhoeve (1957), Sebba (1986) and Winford (1996, 1999). Seuren (1986) demonstrates that these forms are adjectives.

Early I8lh century Srartan -man

239

Table 5: Innovations in word formation in the Court Records (1667-1767) [noun]-ma/t year, source

[verb]-#«a/i year, source

gadoman 'oracle' 1752, CR 1.05.10.02, 803, f 51 r° wieri wieri man 'witch' 1745, CR 1.05.10.02, 798, f - v°

wiesieman '(black) magician' 1760, CR 1.05.10.02, 932, f - v° voevoereman 'thief 1763, CR 1.05.10.02, 808, f - v ° loeckeman 'seer, witch' 1745, CR 1.05.10.02, 798, f - v°

The categorization of wiesieman 'witch' (1760, CR 1.05.10.02, 932, f - v°) in the second column of table 5 may be less straightforward than in the case of voevoereman or loeckeman (lukuman). Here, I have analyzed wiesie as a verb, since it also occurs as a verb in the court records: (6)

Joe wiesie dat soekre. (1761, 1.05.10.02, 805, f - v°) you poison that sugar 'You poisoned that sugar.'

Schumann (1783: 204) glosses wissiman as someone who mixes poisonous potions and/or someone who gives them to others ('ein Giftmischer; der andern Gift beybringt'); wissi can be a verb 'to poison' or a noun 'poison'. In other words, wiesieman can also be analysed as [noun]-mart literally meaning 'poison-person'. Compare, for example, wieriwieriman (1745, CR 1.05.10.02, 798, f - v°), which literally means 'herb person'. 7 There are two reasons why the analysis of wiesiman in terms of [verb]-ma« is favored over an analysis in terms of [noun]-maw. One reason to prefer this analysis of wiesiman is the productivity of the [verb]-maw pattern in 18th century Sranan, already noticed by Bruyn (1989). There are other, even older examples exhibiting this pattern (Bruyn 1995: 168-169 footnote). For example, Van Dyk (cl765:15, reprinted in Arends and Perl 1995) lists wan Zeri man 'merchant' and Scrifiman 'Bookkeeper'. In the eighth dialogue of his language manual, the director of the plantation tells his daughter, Suzanna, to stop annoying him with her questions: (7)

Go we joe wan pikien hakki siman. go away you one little ask-man 'Go away, you little inquisitor' 'Gaä weg jou klyne Vraagal' (Van Dyk, cl765: 39, reprinted in Arends and Perl 1995: 152, underlining mine, MB)

Other support for the analysis of wiesiman in terms of [verb]-man is found in the etymology of the verb wiesie. It is derived from English witch (Smith 1987: 241); the /tS/ is regularly replaced by /s/ (wisi). The 17th century English word witch can be a noun as well as a 7

No verb wieriwieri is attested in the 18th century sources or in contemporary Sranan.

240

Margot van den Berg

verb ( O x f o r d English Dictionary, 1989). The intransitive use of the verb witch means 'to practice witchcraft; to use sorcery or enchantment'; 8 the transitive use of witch means 'to affect (a person) with witchcraft or sorcery; to put a spell upon' 9 (see example 6), co-occurring with the verb bewitch. The Oxford English Dictionary Online mentions the existence of witchman 'wizard' and witchwoman 'sorceress' in folk contexts. It is possible that it derived form English witchman, although considering the rarity of its occurrence, it doesn't seem very likely. In short, I propose that wiesie entered the language as a verb derived from English to witch and used in contexts such as example (6), and that wiesieman is derived from this verb by combining a nominalizing agentive morpheme -man with (the stem of) the verb. The other [verb]-/waw examples in table 5 are less ambiguous in their derivation than wiesieman·, they are clearly derived from the verbs luku and fufuru, as examples (8) and (9) show for luku. Neither luku 'to see' (< English to look) nor voevoere 'to steal' (< Kalahari furu 'to steal' 10 ) occur as nouns in the 18th century sources. They are verbs. (8)

Jo wantje loeke mie jary. (Herlein 1718: 121) you want look my garden 'Would you like to see my garden?'

(9)

mi loekoe hem (Nepveu 1770: 282; reprinted in Arends and Perl 1995: 90) I take-care-of him 'I'm taking care of him.' ('Ik pas hem op, ik zie na hem')

Example (10) is evidence of the fact that already in the 1740s the Dutch in Surinam considered words such as loekeman and loekoe semantically specialized; they borrowed them. (10) Bent gij niet een loecke man? Jae. Wat loeckt gij aan? Dat wanneer jemand zulx is om te sien wat off haar ook schorte." (1745, CR 1.05.10.02, 798, f - v°) 'Are you not a seer? Yes. What do you 'look for'? I 'look for' clues why somebody is not feeling well.'

g

9

10

1623 T. SCOTT Projector 30 Hath not Iesabell painted, and whored, and plotted, and witched, and waded through blood? (Oxford English Dictionary, 1989) 1647 J. MARCH Actions for Slaunder 15 Thou art a Witch..and diddest procure Mother Bale to witch the Cattell of J. S. 1693 I. MATHER Cases Consc. 26 See if you can Witch them into a Fit,..and..Witch them well again 1611 B E A U M . and FL. King and No K. I l l , With her eyes She witches people (Oxford English Dictionary, 1989) Smith (p.c.) makes a case for the derivation of Sranan fufuru from Eastern Ijo furu via reduplication fur(u)furu > fufuru. Note that Eastern Ijo is spoken in the Niger-Delta, and that although it is not impossible for Eastern Ijo speakers to have been shipped to Surinam, there is little evidence supporting the existence of a large or influential group of Kalahari speakers among the slave population of Surinam. On the other hand, the historian Postma has mentioned to Arends (p.c.) that (unpublished) data on 3 shipments f r o m the Niger-Delta exists for the 1660-1675 period. Another Eastern Ijo lexical item is fom 'to beat', which also occurs in Sranan. This is, however, widespread among the Anglo-Caribbean Creoles in contrast to fufuru.

Early 18lh century Sranan -man

241

3. The morpheme -man: grammatical or lexical?

Is man in the innovative word formation process described above a grammatical (derivational) morpheme involved in derivation or a lexical item (noun) involved in compounding? Smith and Veenstra (1998) asked a similar question with regard to Saramaccan, another Creole language spoken in Suriname and closely related to Sranan. What is the status of the Saramaccan morpheme -ma in words such as sipima ship-ma 'sailor'; hondima hunt-ma 'hunter'; idi-boto-ma head-boat-zwa 'pilot'; bebe-dadn-ma drink-rum-/wa 'drunkard' etc. (Smith and Veenstra 1998). They argue in favor of a grammatical status of -ma based on the following: a) in contemporary Saramaccan, there is a clear distinction between the bound affix ('-ma*) and the free word/stem (mdnu 'husband'), although they both originate from the same form (i.e. "); b) there is tonal polarity, a sort of tone-sandhi inducing reversal of tone on the affix with respect to the stem i.e. the tone on '/wa' is dependent on that of the preceding element; c) argument demotion is possible with affixes (deliverer of pizzas) whereas it is not with compounded words (*deliveryman of pizzas). However, these criteria are applicable to 18th century Sranan only up to a certain extent. The differences between the suffix -man and the noun man in 18th century Sranan (or contemporary Sranan) may not be as striking as in the Saramaccan case - but there are differences between the free form and the bound form that indicate a loss of lexical autonomy (see below). Further, although Plantation Sranan (the variety of Sranan spoken on the plantations in the 18th century) must have been a tone language,12 it is very difficult to show tonal polarity based on data from sources written in the 18th century, an age without a spelling standard or conventions for the use of punctuation marks etc. Therefore, it is not clear whether Schumann (1783) uses accents on the vowels to indicate vowel length or high tone, for example. It is equally difficult to show whether argument demotion is possible or not - especially since synthetic compounds of the VN-man type do not occur in the court records or the Saramaka Peace Treaty; nor in Herlein (1718), Van Dyk (cl765), Nepveu (1770) and Schumann (1783). The only construction in these sources that vaguely resembles a synthetic compound is boonjamman 'someone who suffers from gout or rheumatism' (Schumann 1783: 20), although the ordering of the constituents boon 'bone' (noun) and jam 'to eat/enjoy/hurt' (verb) is opposite to what one would expect if it had been a synthetic compound:13 boon would have been the object of the verb jam and, thus, it would have followed the verb (compare with (11)): *jamboonman eat/suffer/enjoy-bone-man 'boneeater/bone-sufferer'. Since boonjam is glossed separately as 'gout, rheumatism' (Schumann 1783: 20), I assume that boonjamman is similar to the [noun]-wa« constructions listed in

11

12

13

Ultimately, this form derives from English man via Caribbean Slave Pidgin English, or via Plantation Sranan. That Plantation Sranan was a tone language follows from the fact that the Marronlects of Surinam (Ndyuka, Aluku, Paramaka and Kwinti), which are descended from Plantation Sranan, are tonal. This can only be explained if the source of all four languages, i.e. Plantation Sranan, is tonal (Smith, p.c.). Contrary to English synthetic compounds, in which the internal object of the verb occurs in front of the (derived) verb: he drives a truck «-»truck driver.

242

Margot van den Berg

the first column of table 5; i.e. [[[boori\\jam]][mari\]. Note that in contemporary Sranan, synthetic compounds ending in a derivational morpheme -man do occur, suggesting that either the Creoles possess a shared innovation or, more likely, a shared retention: (11) a. figi-strati-man sweep street man 'street sweeper'

b. meki-kuku-man (Adamson, p.c.) make cake man 'baker of bread, cake and pastries'

Apparently we can make a case for -man being a derivational morpheme in contemporary Sranan: When attached to a verbal stem it functions as nominalizing gender-neutral agentive affix denoting animate beings. Attached to a nominal (N) stem, it functions as an affix forming an animate noun ('person/bird/insect/snake/fish/crab who is/has/does something with N'): balataman rubber-AFF 'rubber-collector', temreman do joinerwork-AFF 'carpenter/woodpecker', oloman hole-AFF 'bee', alataman rat-AFF 'snake' (mastrigodryas boddaertii), barbaman beard-AFF 'fish' (Ariidae), odiodibotoman greet-boot-AFF 'crab' (Ocypodidae, O. Decapoda). Did its status change from lexical to grammatical (grammaticalization), or did it emerge as a grammatical morpheme in the onset of the language? According to Arends (1995b) the morpheme -man grammaticalized during the 18th century, functioning as a productive gender-neutral affix in the latter half of the 18th century (see above). Schumann (1783) lists several constructions with -man, among them the [verb]-man combinations represented in (12). They are evidence of the grammaticalized status of the morpheme -man (Arends 1995b, Bruyn 1989). Here, both male and female agents are explicitly mentioned by Schumann: (12) helpiman 'helper, Saviour; midwife' ('Helfer, Heiland; it. Wehmutter' Sehl783: 227) naiman 'cutter, seamstress' ('ein Scheider; Neherin' Schl783: 203) suttaman 'adulterer; adultress' ('ein Ehebrecher, Ehebrecherin' Schl783: 169) How could the cliticized noun man grammaticalize into an affix? The words could start out as gender-specific compounds marked for the feature [+ male] (helpiman 'help-man'). Then, its gender specification could be distorted [+ male, -female] in certain contexts such as for example when one needs a midwife (helpiman 'midwife'). The distortion stimulates a reanalysis of the morpheme -man as a gender-neutral person reference marker {helpiman 'helper') and eventually forming gender-neutral animate nouns. This is in accordance with the animacy hierarchy, which holds that human nouns are more likely to be included in linguistic rules than animates in general, and animates are more likely to be included than inanimates (Hopper and Traugott 1993). Let us return to the available 18"' century Sranan data and examine whether they support this scenario. Other mid-18th century examples of [verb]-man are kryman 'wailer' and zingiman 'singer' (Van Dyk cl765: 111 and 108, reprinted in Arends and Perl 1995: 237 and 234, respectively). These examples are taken from Van Dyk's play in which a bad plantation-manager is driven away by the owner after he has raped and killed one of the female slaves. The slaves then organize a party to celebrate; they dance but no one sings. The black overseer, then, complains about the absence of a zingiman. According to Arends

Early 18lh century Sranan -man

243

(1995: 242), the deceased woman was the (lead-)singer or the trokiman,u called at her funeral:

as she is also

(13) trokkeman Iassi ö! lead-singer to lose oh O u r trokiman has died, oh!' (Arends 1995: 237) 'onze Voorzangerheeft het afgeleid, δ! (Van Dyk cl765: 111) Further, recall example (7), in which hakkisiman refers to Suzanna, the daughter of the director in the dialogue. Apparently, the morpheme -man in [verb]-man could refer to both men and women at least as early as the 1760s; it could be used to form an innovative construction such as hakkisiman, which does not resemble the lexifier or superstrate, nor the substrate language(s). 15 Several 18th century [verb]-man examples do not explicitly select a male or a female referent (14) and Schumann's translation does not give us a clue whether these words refer to males or females. Thus, they could be regarded as gender-neutral, selecting a male or female referent based on the context. Here, one cannot tell from the form if it refers to male or a female: If it is a female referent, the derivational morpheme will still be -man: (14)

gakuman konkroman koliman

[stutter] v -person [trick] v -person [cheat] v -person

'stutterer' 'trickster' 'cheat'

(Sehl783:49) (Sehl783:87) (Sehl783:84)

*gakuuman *konkrouman *koliuman

In general, postposed uman is rare. Only a few [noun]-uman and [adjective]-M/naw constructions are encountered in the sources: Hossi Homan house-woman 'maid' (Van Dyk cl765: 11); Koekroehoman kitchen-woman 'female cook' (Van Dyk cl765: 57) and pikin uman young woman 'girl' (not: daughter) (Schumann 1783: 135). The pattern [verb]-uman does not occur in Herlein (1718), the Saramaka Peace Treaty (1762), Van Dyk (cl765), Nepveu (1770), Schumann (1783) or the court records at all. Thus, I see no reason for not regarding these constructions as compounds. In contrast with the postposed noun uman, preposed uman- occurs more frequently in the sources (15). It indicates female gender, similar to preposed uman- in Eastern Maroon Creole, 16 like Saramaccan related to Sranan. In Eastern Maroon Creole the morpheme uman- is

14

Arends (1995: 241, endnote) "The troki man is the woman who accompanies dancing with solo singing (Lichtveld and Voorhoeve 1980: 247). In the present context it refers to the slaves' habit o f singing during work".

15

Note, that even if you consider Van Dyk not a reliable source (Lichtveld and Voorhoeve 1980) and, thus, refuse to take the word hakkisiman as a real 18th century Sranan word, a fruit o f the imagination of the author, the productivity of the pattern \\erb]-man with the morpheme -man specified for the features [± male, ± female] is not falsified.

16

Eastern Maroon Creole is a term covering the language varieties spoken by the members of four Eastern Maroon communities, the Aluku-Boni, the Paamaka, the Okanisi and the Kwinti (Migge 2001). They descend from Plantation Sranan, a language variety spoken on the plantations in the 18 th century which is assumed to be represented in Van Dyk ( c l 7 6 5 ) .

244

Margot van den Berg

preposed to refer to female referent if it is a personal noun with a predominantly male interpretation: uman sikoutu 'female police officer' (Migge 2001). umanpikin uman doksi uman kau uman vool umansanni uman-futuboi

woman-young woman-duck woman-cow woman-poultry woman-thing woman-servant

'daughter' 'duck' 'cow' 'chicken' 'female genitals' 'maid-servant'

(Sehl783: (Sehl783: (Sehl 783: (Schl783: (Sehl783: (Sehl783:

136) 136) 136) 136) 190) 48)

* pikin uman *doksi uman *kau uman *vool uman *sanni uman *futuboi-uman

mann-Ningre mann-kau mann-doksi mannpikin mann-futuboi

man-negro man-cow man-duck man-young man-servant

'blackman' 'bull' 'drake' 'boy, son' 'servant'

(Sehl783: 107) (Sehl 783: 136) (Schl783: 136) (Sehl783: 107) (Sehl783: 48).

*ningre-mann *kau-mann *doksi-mann *pikinmann *futuboi-mann

The preposed morpheme man- indicates male gender [+ male] (16), it is frequently written by Schumann (1783) as mann- with , probably to distinguish it from the lexical item man (Bruyn 1989). This conceivably indicates some phonological difference in addition to the morphological difference already noted. Another example of the difference between the postposed element -man and the preposed man- is (17). While the word for 'black woman' is formed by compounding the word for 'Black', Ningre, and the word for 'woman', uman, the word for 'black man' is formed by preposing man- in front of the word for 'Black' (* ningreman',11 *uman ningre). This is also true for Karboegerw (* kabugru-man; *umankabugru). (17) man-Ningre, Ningre uman (Sehl783: 122) man-negro, negro-woman'Black' man-kabügru, kabügru-uman (Sehl783: 77) man-karboekel, karboekel-woman This alternation between preposed man- and postposed uman in (17) does not make sense if you regard the postposed element -man as specified for the feature [+male]; why would Schumann (1783) distinguish between the preposed morpheme man and the postposed morpheme uman in (17)? The answer to this question brings us back to the status of the morpheme -man' and the scenario outlined above for its emergence. If -man is indeed a gender-neutral affix meaning 'person' i.e. [± male, ± female], then the alternation between the preposed element man{ri)- and the postposed element -man makes sense: Ningre-man could refer to both male and female blacks; while mann-Ningre selects the black man from the set of black people. (Note that in contemporary Sranan, man-ningre is used sometimes to refer to both black men and white men (Adamson, p.c.). Although the right member of this compound has lost its original meaning ('Black'), man still refers to the male gender.).

17

18

The form ningreman is used frequently by Schumann (1783), alternating with ningre 'black person' or 'black people'. The Karboegers (or Muraato) are a group of former slaves who ran away from the plantations and settled in the hinterland of Suriname (Smith 2001).

Early 18th century Sranan -man

245

The identity of the words in (15) and (16) as compounds is guaranteed by their status in contemporary Sranan. Since Eastern Maroon Creole is a direct descendant of 18th century Plantation Sranan, while Saramaccan split off Sranan by the last decade of the 17th century (Smith 1998, McWorther 1998), the word formation process in Eastern Maroon Creole may help us to deepen our understanding of what happened in 18th century Sranan. In contemporary Eastern Maroon Creole, the free morpheme man may refer to both male and generic referents, but not to overtly female-marked referents (Migge 2001). Personal nouns formed by derivation with the bound morpheme man, such as singiman 'singer', select male and generic referents. Moreover, they denote overtly female-marked referents: "In addition, a great number of them may also be used to refer to overtly female-marked referents (...) Their overtly female-marked counterparts (cf. biibiuman and singiuman) are relatively uncommon even though, for example, most of the active Christians among the Eastern Maroons are women" (Migge 2001: 96-97). According to Migge (2001), word formation with the gender-neutral bound morpheme -man [± male, ± female] can only take place with stems that are derived from verbs and nouns associated with activities that may be carried out by both men and women. Migge (p.c.) points out that, as in all languages, most personal expressions in Eastern Maroon Creole have a male-dominant reference, but that the visibility of the form may have been even higher since slave groups often consisted of large groups of men, at least initially. Note that in Eastern Maroon Creole, like in Saramaccan and in 18th century Sranan, the free morpheme man/manu/man means 'man' while the bound morpheme -manl-mal-man forms a gender-neutral animate noun. This pattern fits in neatly with the West African Gbe languages Xwlagbe, Gengbe, Fongbe and Gungbe (Da Cruz 1998). In the Gbe languages, a limited set of derivational morphemes is used to form agents and possessor nouns, diminutives, nouns of geographical reference and ordinal numerals. This last derivational morpheme exhibits some variation across the Gbe languages (table 6). Table 6: Derivational morphemes in Gbe and their most probable source (Da Cruz 1998) [X]-x

function

source

X-to

- agent of Χ (X is an activity) - possessor of Χ (X is a possession) - referent has a qualifying property X - referent originates from Χ (X is a place) - ordinal of number X (Gungbe) possessor of Χ (X is a possession or a qualifying property) diminutive of X referent originates from Χ (X is a place) ordinal of number X (Fongbe, Xwlagbe) ordinal of number X (Gengbe)

to 'father'

X-nd X-vi X-nu X-go X-gon

no 'mother' vi 'child' -

246

Margot van den Berg

The gender-neutral derivational morpheme -to is used to denote an agent (someone who V's) or an entity characterized by a possessed-possessor relation (someone/something who has/is/does N) in all Gbe languages (18). (18) äjo eci tünfele gbe

'to steal' 'money' 'to forge' 'world, life'

äjö.to eci.to fele tun.to gbe.to

'thief 'rich person' 'blacksmith' 'person'

(Xwlagbe, Da Cruz 1998: 153) (Gengbe, Da Cruz 1998: 154) (Fongbe, Da Cruz 1998: 154) (Gungbe, Da Cruz 1998: 154)

It has a wider distribution than the derivational morpheme -no, although -no is sometimes in competition with -to'(Da Cruz 1998: 158). For example, in Gengbe äxweno can mean: a) the wife of the landlord or b) the female owner of the house, i.e. the landlady. The Gengbe word äxweto 'landlord, landlady' can also refer to a landlady. In general, both derivational morphemes are gender-neutral and they cannot be combined with adjectives. Both forms no and to can co-occur as free morphemes meaning 'mother' [+ female] (19a), and 'father' [+ male] (19b) respectively: (19) a. kofi sin nä wä. Kofi of mother has come 'Kofi's mother has come' b. ayäbä sin to wä Ayäbä of father has come 'Ayäbä's father has come'

(Gungbe, Da Cruz 1998: 157)

(Gungbe, Da Cruz 1998: 156)

This kind of derivation is present in 'Old Gbe' (Aboh 2000). Since most of the earliest slaves were imported from the Slave Coast 19 (1650 - 1720, Arends 1995a), where mostly Gbe is spoken, the majority of the newly imported slaves in Surinam would have used the nouns to 'father' and no 'mother' alongside the derivational morphemes -to and -no in their mother tongue, especially since "[d]uring the first fifty years of colonization the entire black population was outnumbered by new arrivals from Africa every three to five years" (Arends 1995a: 269). Is there any evidence that the Gbe process of derivation has influenced Sranan? The following example from Schumann (1783) provides clear evidence for this influence: While the Gbe word for the events and actions that can best be subsumed under the label witchcraft or sorcery (with both negative and positive undertones) is äze and someone who invokes äze is denoted by äzeto, 18th century Sranan has aseh and asih-man, respectively. Schumann describes aseh as follows:

19

Slave Coast: Eastern Togo, Benin and the western part of Nigeria.

Early 18fh century Sranan -man

247

"Die Hexery Zauberkraft; eigentlich just derselbe Aberglaube, was in Europa die Hexenhistorien sind, zum Beispiel dass sich ein Mensch in eine Katze oder sonst ein Thier, oder in einen brennenden Baum verwandelt, und in der Gestalt bey Nacht herum geht und Schaden thut, und so weiter hernach wieder sein vorige Gestalt annimt (...) asehman, eine vermeinte Hexe, Mann oder Frau (solche Leute, von denen die Neger glauben, dass sich sich des Nachts in Feuer oder in sonst eine Gestalt verwandeln, um die Leute zu beunruhigen, oder auch zu beschädigen." (Schumann 1783:7-8). [magical witchcraft; in a way similar to the superstition what is known in Europe as tales of the supernatural, for example when a human changes into a cat or another animal, or a burning tree, and in that form causes damage at night time until he reverts to his former shape (...) asehman, an alleged witch, man or woman (people of which the blacks think that at night time they turn into fire or another form to disturb people or hurt them]

In short, the Sranan word for witchcraft is inherited from Gbe, while the Gbe grammatical morpheme -to is relexified by -man, without implications as to gender. Further research should establish to which extent words such as guduman goods-person 'wealthy person' and voevoereman steal-person 'thief are caiques of Gbe (eci.to money-person 'wealthy person' (Gengbe) and äjö.tdsteal-person ' t h i e f , respectively). 20

4. C o n c l u s i o n

In conclusion, I have presented several examples which show that the suffix -man is a derivational morpheme in [\ex\i]-man constructions, which have been recorded as early as 1745 (loeckeman 'seer, witch' 1745, CR 1.05.10.02, 798, f - v°); the earliest [verb]-wa« construction referring to a woman is recorded in (c)1765. This bound derivational morpheme meaning 'person' co-exists with a free lexical item man that can occur as a noun ('man'), a verb ('to be able to, can', see Van den Berg and Arends 2001) and as prefixed noun indicating male gender. The mid-18 th century data presented in this paper cast doubt the hypothesis that the compounded noun man grammaticalized during the 18th century: if the noun man in this construction grammaticalized, then this process must have been completed before the 1760s in order to account for the data presented here. Or even earlier, since some of these mid-18 th century data are taken from Van Dyk (cl765), a source generally regarded as representing an older, or more conservative, variety of Sranan. 21 The date of completion can be pushed back even further into the 17th century, since the distinction between the bound genderneutral derivational morpheme and the free gender-specific lexical item is present in Saramaccan. It could have survived the split with Sranan and subsequently have developed into the contemporary forms manu 'man' and '-ma'. As the gender-neutral bound derivational morphemes occur in combination with genderspecific free lexical items in Gbe, a parallel occurrence of bound gender-neutral deriva20

21

An introductory study of the structural relationship between Gbe and Sranan in word formation is Aboh and Van den Berg (2001). See Bruyn (1995: 39-44) and Arends and Perl (1995) for a discussion of Van Dyk (cl765).

248

Margot van den Berg

tional morphemes and free lexical items is found in Sranan. We could conclude that the free lexical item -man did not grammaticalize during the 18th century, but already functioned as a derivational morpheme in the earliest stages of the language. Why did the slaves of the Surinamese plantations select the form to realize both functions? Migge may have a point that the selection of this particular form may be related to the demographics of the slave population. The ratio of males to females among slaves deported from Africa was roughly 6-to-4 (Arends 1995a: 257). Although most languages have a male-dominant reference, the frequency of occurrence of such forms may have been even higher in Suriname. The slaves could have encountered words such as countryman in Caribbean Slave Pidgin English (Smith 1999) or in the language varieties of their overseers 22 and reanalyzed it according to their native patterns. This development can be classified an instance of grammaticalization, albeit the kind of grammaticalization identified as 'instantaneous grammaticalization' by Bruyn (1996). This kind of grammaticalization differs from 'ordinary' grammaticalization (described in section 1) in that the developments under investigation can take place within a short time span in creolization due to relexification, while they proceed gradually under ordinary grammaticalization. In this paper, then, I have examined the word formation processes of compounding and derivation involving the gender-neutral morpheme -man, giving animate nouns in the Court Records (1667-1767) and the Saramaka Peace Treaty (1762). I have said nothing about, for example, other possible derivational morphemes such as -fasi 'manner' or -wan '(some)one' and reduplication, but these documents contain interesting data on these formations as well. In view of the theoretical relevance of creole morphology, these records may prove to be especially valuable, both because of their early date and because of their reliability as a source of authentic Sranan.

Archival Sources

ARA = Algemeen Rijksarchief Overgekomen brieven en papieren uit het archief van de Societeit van Suriname, 1683-1715, (ARA, code 1.05.04.01, inventory numbers 212-240) Overgekomen brieven en papieren uit het archief van de Societeit van Suriname, 1751-1767 (ARA, code 1.05.04.06, inventory numbers 286-335) Inventaris van de archieven van de Raad van Politie (1669-1680) en de Raad van Politie en Justitie (1680-1683) en het Oud-Archief van het Hof van Politie en Criminele Justitie in Suriname (ARA, code 1.05.10.02, inventory numbers 781-948). Oud-Archief Suriname / Hof van Politie en Criminele Justitie (ARA, code 1.05.10.02, inventory number 66) Oxford English Dictionary. 2nd ed. 1989 (ed. J. A. Simpson and E. S. C. Weiner), Additions 1993-7 (ed. John Simpson and Edmund Weiner; Michael Proffitt), and 3rd ed. (in progress) Mar. 2000(ed. John Simpson). OED Online. Oxford University Press,

22

Compounding with the gender-specific English lexical item man is productive in contemporary and 18th century English (Oxford English Dictionary 1989).

Early 18"1 century Sranan -man

249

References

Aboh, Enoch (2000): Old Gbe: Notes from La Doctrina Christina (1658) and La Grammaire Abregee (1730). - Paper presented at the Trans-Atlantic Sprachbund meeting, Sept. 2000, Amsterdam. Aboh, Enoch and Margot van den Berg (2002): The structure of word formation in 18th century Sranan, Gungbe and Old Gbe'. - Paper presented at the Biennial Conference of the Society for Caribbean Linguistics, University of the West Indies, St. Augustine, Trinidad. Adamson, Lilian (2001): Sranan. - In: Guus Extra and Jan Jaap de Ruiter Babylon aan de Noordzee: Nieuwe Talen in Nederland. Arends, Jacques (1995a): Demographic factors in the formation of Sranan. - In: Jacques Arends (ed.) The early stages of creolization, 233-85. Amsterdam: Benjamins. - (1995b): Part 1 Sranan texts. - In: Jaques Arends and Matthias Perl (1995): Early Suriname Creole texts. A collection of 18/h century Sranan and Saramaccan documents. — Frankfurt/Madrid: Iberoamericana. Arends, Jacques and Matthias Perl (1995): Early Suriname Creole Texts: A Collection of 18th-century Sranan and Saramaccan documents. Frankfurt/Madrid: Vervuert. Bruyn, Adrienne (1989): De nominale constituent in het 18de-eeuws Sranan. Een beschrijving van grammaticale elementen en morfosyntactische processen. Thesis, University of Amsterdam. Joan Bybee, R. Perkins and W. Pagliuca (1994): The evolution of grammar: tense, aspect and modality in the languages of the world. Chicago: The University of Chicago. - (1995): Grammaticalizaion in Creoles: The development of determiners and relative clauses in Sranan. Amsterdam: IFOTT (Studies in language and language use; 21) Dissertation, University of Amsterdam. - (1996). On identifying instances of grammaticalization in Creole languages. - In: Philip Baker and Anand Syea (eds.): Changing meanings, changing functions. Papers relating to grammaticalization in contact languages, 29-46. London: University of Westminster Press. Bybee, Joan, Revere Perkins and William Pagliuca (1994): The evolution of grammar. Chicago: The University of Chicago Press. Da Cruz, Maxime (1998): La Derivation lexicale en Gbe. - In: GbeGbo Studies. Scientific Review of ofLabo Gbe (Int.), Vol. 2, 152-175 Garome: Benin. Focke, Hendrik (1855): Neger-Engelsch Woordenboek. Leiden: P.H. van den Heuvell. Gordijn, W., C.F.A. Bruijning and J. Voorhoeve (eds.) (1977): Encyclopedie van Suriname. Amsterdam: Elsevier. Hartsinck, Jan Jacob (1770): Beschryving van Guiana, of De wilde kust in Zuid-America, betreffende ...de bezittingen der Spanjaarden, Franschen en Portugeezen en voornaamelyk de volksplantingen der Nederlanden, als Essequebo, Demerary, Berbice, Suriname ... Waarby komt, eene verhandeling over den aart en de gewoontes der neger-slaven. Amsterdam: Gerrit Tielenburg. Herlein, J. (1718): Beschrijving van de volksplantinge Zuriname. Leeuwarden: Meindert Injema. Hopper, Paul J. and Elizabeth Closs Traugott (1993): Grammaticalization. Cambridge: Cambridge University Press. Kramp, Andre (1983): Early Creole lexicography - a study of C.L. Schumann's manuscript dictionary of Sranan. Dissertation. University of Leiden. Kurylowicz, Jerzy (1965): The evolution of grammatical categories. - In: Diogenes 51, 55-71 [Reprinted in Kurylowicz, Jerzy (1975): Esquisses Linguistiques, Vol II: 38-54], Lappe, Sabine and Ingo Plag (this volume) Rules vs. analogy: modeling variation in word-final epenthesis in Sranan, 71-90. Lefebvre, Claire (1998): Creole genesis and the acquistion of grammar. - Cambridge: University Press.

250

Margot van den Berg

Lichtfeld, Ursy M. and Jan Voorhoeve (1980): Suriname: Spiegel der vaderlandse kooplieden. 2nd revised edition. Den Haag: Nijhoff. McWorther, John (1998): Identifying the Creole prototype: Vindicating a typological class. - In: Language 74, 788-818. Migge, Bettina (2001): Communicating gender in the Eastern Maroon Creole of Suriname. - In: Marlis Hellinger and Hadumod Bussmann (eds.): Gender across Languages: The linguistic representation of women and men. Vol. 1 (Published as Vol. 9 of the series Impact: Studies in Language and Society). Muysken, Pieter (1981): Halfway between Spanish and Quechua: The case for relexification. - In: A. Highfield and Albert Valdman (1981): Historicity and variation in Creole studies. Ann Arbor: Karoma Muysken, Pieter and Norval Smith (1990): Question words in pidgin and Creole languages. - In: Linguistics 28, 883-903. Nepveu, Jan (1770): Annotatien op de Surinaamsche Beschrijvinge van A 0 1718. MS published in Jacques Arends and Mattias Perl (1995): Early Suriname Creole Texts, 77-91. Frankfurt/Madrid: Vervuert. Schumann, Christian (1783): Neger-Englisches Wörterbuch. - MS published in Andre Kramp (1983): Early creole lexicography: A study of C.L. Schumann's manuscript dictionary of Sranan, 44-305. Diss, University of Leiden. Sebba, Mark (1986) Adjectives and copulas in Sranan Tongo. - In: Journal of Pidgin and Creole Languages 1: 1, 109-121. Seuren, Pieter (1986): Adjectives as adjectives in Sranan: A reply to Sebba. - In: Journal of Pidgin and Creole Languages 1:1, 123-134. Smith, Norval (1987): The genesis of the creole languages of Surinam. Diss. University of Amsterdam. - (1998): The vowel system of 18th century St Kitts Creole ... evidence for the history of the English Creoles? - In: Philip Baker and Adrienne Bruyn (eds.): St Kitts and the Atlantic Creoles : the texts of Samuel Augustus Mathews in perspective. Westminster creolistic series 4, 145 - 172. London: University of Westminster Press. -

(2001): The creole languages of Surinam - past and present. - In: J.T.G. Arends and E. Carlin (eds.): Language atlas of Surinam, 2001. Leiden: KITLV. - (this volume) New evidence from the past: To epenthesize or not epenthesize, that is the question, 91-107. Smith, Norval and Tonjes Veenstra (1998): Synthetic Compounds in a Radical Creole: abrupt versus gradual change. - Paper presented at the Creole Conference, June 1998, Regensburg Snijders, Ronald (2000, 5 th edition): Surinaams van de straat/ Sranantongo fu strati. - Amsterdam: Prometheus Van Dyk, Pieter, n.d. (cl765): Nieuwe en nooit bevoorens geziene onderwijzinge in het Baster Engels, of Neeger Engels ... Amsterdam: Wed. Jacobus van Egmont - Republished in Jacques Arends and Mathias Perl (1995): Early Suriname Creole Texts, 93-239. Frankfurt/Madrid: Vervuert. Van den Berg, Margot (2000): "Mi no sal tron tongo". Early Sranan in court records, 1667 - 1767. Thesis, University of Nijmegen Van den Berg, Margot and Jacques Arends (2001): Court records as a source of authentic early Sranan. - Paper presented at SPCL, Jan. 2001, Washington D.C. Van Donselaar, Jan (1989, 2 nd edition): Woordenboek van het Surinaams-Nederlands. Muiderberg: Coutinho. Veenstra, Tonjes (2001): Properties of the third kind in Language Contact. - Paper presented at the ACLC-vrijdagmiddaglezing, May 2001, Amsterdam. Voorhoeve, Jan (1957): The verbal system in Sranan. - In: Lingua, 6, 374-396.

Early 18fh century Sranan -man

251

Winford, Donald (1996): Verbs, adjectives and categorical shift in Caribbean English Creoles. - In: P. Christie (ed.) Caribbean language: Issues old and new. Mona: Jamaica: University of the West Indies Press, 12-26. - (1997): Property items and predication in Sranan. - In: Journal of Pidgin and Creole Languages 12:2, 237-301.

Margot van den Berg, Department of Theoretical Linguistics, Amsterdam Center for Language and Communication, University of Amsterdam, Spuistraat 210, room 502, 1012 VT Amsterdam, The Netherlands, Email: [email protected]

Patrick Steinkrüger Morphological processes of word formation in Chabacano (Philippine Spanish Creole)

1. Introduction 1

This contribution is concerned with word formation in the Spanish-based Creole on the Philippines, commonly known as Chabacano. The study of Chabacano dates back to Hugo Schuchardt and his Kreolische Studien (1883). Geolinguistically, Chabacano can be divided into two main varieties: In the north, there are the dialects of Manila Bay Creole (MBC) spoken in the towns of Ermita, Cavite and Ternate, whereas in the very south, Southern Mindanao Creole (SMC) is located. SMC is mainly spoken in the city of Zamboanga and its surroundings. The highest vitality of Chabacano can be observed in the city of Zamboanga in the extreme western part of the island Mindanao (with nearly three hundred thousand speakers). A subvariety of Zamboangueno is also spoken in the town of Cotabato, whereas nearly nothing is known about the vitality of Chabacano in Davao. The genesis and history of Zamboangueno is somewhat complex due to different contact situations in time and space.2 In this context the most complete investigation has been done since the 1980s by John M. Lipski (1992, 2001). It is worth mentioning that Zamboanguefio does not fit neatly the classification of plantation, fort and maritime Creoles,3 but several scenarios could be observed. During the creolization process the influx of Spanish and Mexican speakers played certainly an significant role, though perhaps less significant than in the Americas. Slavery played a marginal role but it is perhaps an overstatement that the genesis of Chabacano be characterized by a "complete disassociation from slavery" (McWhorter 1995:228-9). For a detailed treatment of the genesis and development of Chabacano see Lipski (2001). Nor the dichotomy of exogene and endogene Creoles, formulated by Robert Chaudenson, does not exactly grasp the situation of Zamboanguefio. More generally speaking, I think it is not consistent to define Creoles by certain linguistic features and on the other hand to limit their existence to a certain region with a special sociohistorical scenario. Consequently, creolization should be regarded as a global phenomena and whether Creoles can be defined typologically must be shown by more data (especially from non-European based Creoles).

1

2 3

Thanks to all the participants of the Workshop 'The Phonology and Morphology of Creole Languages' for their comments, especially to two anonymous reviewers and to the editor of this volume. For the history of Chabacano see also Whinnom (1956), Casado/Qilis (1992) and Lipski (1992). Mühlhäusler (1987:30, Fn.3) argues against this classification as follows: "Bickerton's [...] distinction between plantation, fort and maritime contact situations would seem to account for only some of the types of contact in the Pacific. Given his general neglect of sociohistorical detail this is not suprising.".

254

Patrick

Steinkrüger

Let me list some characteristics of the language, especially concerning the lexicon and the morphological inventory: Some typological features are due to Austronesian contact (cf. Lorenzino 1993), for example the basic word order in Chabacano is VSO. Apart from some Indo-Portuguese Creoles where we find SOV and OVS basic word order, this is a marked case in comparison to other European-based Creoles (cf. Bartens 1995:149). The lexicon principally consists of Spanish vocabulary, very few words are taken from Portuguese. The major input from Philippine languages comes from Ilonggo (the proper name is Hiligaynon), a Western Visayan language used formerly as a lingua franca in the region (Frake 1971, Lipski 1992). Besides Spanish and Hiligaynon, national, regional and different local contact languages have influenced the lexicon of Zamboanguefio. Here I should mention Cebuano, the major Western Visayan Language on the Southern Philippines and Tagalog (Pilipino) which only recently has influenced the Philippine languages all over the Islands. The local languages Tausug, Subanon and Yakan, inspite of the close geographical contact with Chabacano, have very little influence on the Chabacano lexicon. In her analysis of a basic vocabulary of more than 6000 entries, Riego de Dios (1976/89:188) states that 83 % of the items are of Spanish, ca. 15% of Philippine and 2,5% of English origin. Nearly 30 years after her analysis, it can be assumed that the English influence has increased and the Spanish has diminished. As far as I have found out, all productive morphemes in Zamboanguefio can be identified as Spanish, Hiligaynon and Cebuano, the latter two being Visayan languages. The form of the morphemes is often identical with those of the source languages but in many cases they have different functions. The productivity and diversity of derivational morphemes in Chabacano stands in contrast to the alleged simplicity of Creole morphology, claimed already by Bloomfield (1933), and recently revived by McWhorter (1998, 2001). 4

2. D a t a sources and procedure

This paper highlights morphological processes of word formation in the Spanish-based Creole spoken in Zamboanga and Cotabato. From both varieties I have used data for my analysis. The spelling of Chabacano words is generally adapted as found in the sources. 5 That means, the spelling of lexical items is adapted without modifications and stress indication is marked by an acute if indicated in the sources. 6

4

5

6

For critical remarks concerning word formation in Creoles see Degraff (2001), Plag (2000), Braun/Plag (in press). On the supposed simplicity of Creoles in general see the replies to McWhorter (2001) in the same issue of Linguistic Typology, e.g. by Wurzel (2001). I have taken my data mainly from the following sources (incl. some texts on the internet in Chabacano): Forman (1972), Hiligaynon-Site, Meaning of Affixes: http://www.copewithcytokines.de/ NAGATUON/cope.cgi, Languages of the Southern Gateway (1979), Moliner (1999), Philippine On-Line Dictionary: http://www.bohol.ph/diksyunaryo.php, Riego de Dios (1976), Slaby/Grossmann (1975). Only Forman (1972) and Riego de Dios (1976) mark stress, the first with accents and the latter with capitals.

Word formation in Philippine Spanish Creole

255

The main aim of this paper is to classify the morphological processes of word formation in Chabacano and to determine the etymology of the derivational morphemes. Another goal consists in pointing out the main differences in comparison to Spanish and to demonstrate that Chabacano is not simply an intertwined language (as a superficial glance could suggest), where a Spanish lexicon interacts with a grammar of essentially Philippine morphosyntactic rules. Another reason to choose Chabacano is that, contrary to general assumptions about Creoles, its morphology (in this case word formation) shows rather complex structures. There is a single chapter on derivation in Chabacano in the only available grammatical sketch of this Creole (Forman 1972), which also includes a useful wordlist. 7 This structuralist analysis of word formation in Chabacano was the starting point of my own analysis. Forman's grammar neglects two aspects that are important in the context of this paper: the origin of derivational morphemes and the contrast of derivates with their Spanish equivalents. 8 This paper does not only attempt to present a more complete systematization of derivational affixes, but also to include additional data.

3. W o r d f o r m a t i o n in C h a b a c a n o

In Chabacano all principal processes of word formation can be found, namely (la) derivation with affixes, including ( l b ) zero-derivation (conversion), (2) compounding, and (3) reduplication respectively. The following examples may illustrate this: (1)

a. kwento

—>

'story'

man-kwento VERB-story 9 'tell a story'

b. mal 'evil' (N + Adj) (2)

7

8

9

pwerko + monte 'pig' 'mountain'

pwerko-y-monte 'wild pig'

Also Riego de Dios (1976:27-29; 35-37; 39-42) contains a sketch of word formation in Chabacano. Forman himself admits this weak point in his description: "While I have had some undergraduate exposure to Latin, I really know next to nothing about the modern Romance languages or the field of Romance linguistics" (Forman 1972:12-3). I have used the following abbreviations in the text: Languages: Ceb - Cebuano, Chab - Chabacano, Engl - English, Hil - Hiligaynon (= Ilonggo), Phil - Philippine language(s), SMC - Southern Mindanao Creole/Chabacano, Span - Spanish, Tag - Tagalog (= Pilipino) Linguistic Categories: Adj - adjective, ASP - aspect marker, CAUS - causitive affix, V - verb, VERB - verbalizing affix, Ν - noun.

256 (3)

Patrick

birä -> 'return'

Steinkrüger

bira-birä 'keep turning'

It is particularly interesting, that derivational prefixes are taken from Austronesian (especially from Hiligaynon) and are added in an agglutinating manner. In contrast, nearly all suffixes are taken from Spanish. This is shown in the following examples: (4)

a. höben 'young'

->

ta- m- pa- höben ASP.VERB.CAUS.young 'act youthfull'

b. päharo 'bird'

paharyä 'to hunt birds'

Besides the productive processes of word formation in Chabacano mentioned above, let me mention that in a very limited group of Spanish lexemes, the noun-verb-distinction is morpho-phonolocically marked by a different stress pattern. While nouns have penultimate stress, verbs are characterized by final stress. As this is only a fossilized structure of Spanish and no longer productive, I don't consider this as a productive morphological process in Chabacano. The examples in (5) illustrate such cases: (5)

diida 'doubt' länsa 'lance' piirga 'purgative' sena 'supper' desgräsya 'accident' pläncha 'iron' giya 'guide'

-> ->

—>

->

dudä 'to doubt' lansä 'to aim or peirce with a lance' purgä 'to drink a purgative' send 'eat supper' desgrasyä 'happen by accident' planchä 'to iron' giyä 'to guide, to lead'

In the following sections I will represent the most productive morphological processes of word formation in Chabacano, namely derivation by affixation, reduplication and finally compounding.

4. Derivational Affixes

There is a relatively rich stock of derivational suffixes in Chabacano which contradicts the claimed poverty of Creole morphology in general. Nevertheless we should admit that the stock of affixes compared to that of both donor languages, namely Spanish and Hiligaynon respectively, is reduced in number (for Spanish compare the large list of productive affixes in Rainer 1993:299ff and for Ilonggo the URL quoted in footnote 5). We could classify the derivational affixes of Chabacano following different criteria, namely their etymology (Spanish or Austronesian), their position (prefix or suffix respectively) and finally their function (in the donor and target language).

Word formation in Philippine Spanish Creole

257

In tables 1 and 2 I have tried to present all productive affixes in Chabacano by listing their function and meaning, their origin and quoting examples involving the morpheme in question. Table 1: Synopsis of derivational prefixes in Chabacano form

etymology

ika-

< Hil ika-

ma-

< Hil ma-

maka-

< Hil maka- (?)

man-

10

< Hil mang-

mag-

not clear

pa-

< Hil -pa and -pag

paka-

< Hil pagka-

10

11

function/meaning in Chabacano prefix for ordinal numbers

function in donor language prefix for ordinal numbers

prefix forming adjectives generally from nouns causative prefix for nouns, verbs and adjectives

adjective formative for many roots prefix signalling inherent ability specified by the root or quality named by the root a noun verbalizer amigo 'friend' —> man-amigo 'to be(come) friend' ermäno ,brother' magermärto —> 'referring to or among brothers' pa-alisto verbal noun 'test each other's skill' formative affix for many roots11 man-pa-balabak 'get set askew'

prefix for transitive verbs: + Ν or Adj -> V prefix for collectives

prefix forming causative verbs or directional nouns +Adj - » reciprocal V +V causativity +N -> directionality noun-forming prefix for adjectives

prefix employed to form abstract nouns from any expression whatever

examples dos 'two' —>· ika-dos 'second', sinko 'five' -»ika-sinko 'fifth', etc. pyedra 'stone' —> ma-pyedra 'stony, full of stones' bungul'deaf —> maka-bungul 'deafening'

alisto 'alert; active' —> paka-alisto 'talent'

Forman (1972:122) claims that man- is a prefix for collective nouns (e.g. man-ermdno 'a group of brothers'), but I can't confirm this with other data. Perhaps there is confusion with mag-. Some authors like Forman claim a Spanish origin (< Span para).

Patrick Steinkrüger

258 form

etymology

ta-

< Span

estä

function/meaning in Chabacano verbal prefix expressing imperfective aspect; derivational function in certain combinations

function in donor language

examples

auxiliary (3.pers.sg.pres. ind.); in combination with the gerund indicates imperfective aspect

hoven 'young'

—> tampahoven 'act youthfull'

(ta-man-pa- —> tampa-)

Table 2: Synopsis of derivational suffixes in Chabacano form

etymology

-m

< Span -ar,

-ϊ,έ -a,-ο

-ir, -er < Span

-a, -o

function/meaning in Chabacano suffixes for verbalization gender distinction for nouns 12

function in donor language

examples

bibo 'alive' - » bibi 'to live' nyita 'niece' vs. noun suffix for gender distinction nyeto 'nephew' infinitive marker

-da

< Span -da

nominalization: manner, act place, product etc. of X-ing

suffix of nomina actionis

ensenyada 'teachings', abläda 'manner of talking', entendida 'manner of understanding or comprehension'

-do, -w

< Span

suffix forming adjectives

masculine past participle suffix

kurta 'to cut' —» kurtdw

do13

'cut'; manteka 'lard; cooking

fat' —> mantekddo -dor, -dora

12

13

14

< Span

-dor, -dora

suffix for nomina agentis

nomina agentis

'with ample lard' nada , to swim' —> nadador 'sailor' 1 4

Also existing in Visayan languages (e.g. Cebuano) for words borrowed from Spanish (distinguishing sex). The deletion of intervocalic d in this participle suffix is very frequent in Spanish. It seems that the spelling o f this d in Chabacano is due to orientation towards written Spanish. The derivates are semantically not identical with the Spanish forms.

Word formation in Philippine Spanish Creole

form

etymology

-esa

< Span -esa

-han, -an

< Hil -(h)aniS

-hin, -in

< Hil -gin (?)

-ista

< Span -ista (same function) < Span -cion < Span -tivo

+ Ν - » 'one who does a regular occupation etc.' nominalizing suffix suffix forming adjectives

< Span -ero, -era

suffix for nomina agentis

-syon -tibo

-ero, -era

function/meaning in Chabacano X (Adj) + -esa -> quality of X

259 function in donor language nomina qualitatis

1. man- +V/Adj + locative suffix han reciprocal V 2. 'the place where X happens, the Xing place' (N) adjective marker signals a passive past tense construction

nomen agentis

nomen actionis deverbal suffix forming adjectives nomina agentis

examples riko 'rich' rikesa 'wealth';pobre 'poor' -> pobresa 'poverty' 1. kwento 'story' —> man-kwento-han 'to tell each other' 2. tubii 'sugarcane' -> tubu-han 'canefield' myedo 'fear' myedu-hin 'one easily frightened'; dsrna 'asthma' —» asma-hin 'asthmatic' (syn. asmätiko) cabakanista 'Chabacano expert' dibidi 'divide' -> dibisyon 'division' pensa 'think' —> pensatibo 'pensive' kaskäs (Tag) —» kaskasero 'speed maniac, reckless or fast driver' 16

As for the productivity and combination of these affixes, it is appropriate to add some short remarks. A raw system of combination can be formulated as follows: affixes from Phil origin can be combined with every root, but Spanish affixes only with roots from Spanish origin (one exception is -ero; see the example at the bottom of table 2). There are also some yet unclear restrictions with regard to the morphological integration of roots from English origin. It is generally doubtful, whether the Spanish suffixes are really productive or just a part of a lexicalized form. This holds at least for -esa, -ista, -syon, -tibo, and -a,-o. Among the verbal endings, the suffix -ά could be the only productive one, if at all. On the other hand, the suffixes -do/-w and -da seem to be very productive in SMC. To demonstrate that word formation in Chabacano with affixes originating form Spanish is not a only a matter of wholesale borrowing let us look at the suffix -(a, i)da. Historically it derives from the Spanish inflectional past participle female -da. Many Spanish nominalizations are formed with -da (e.g. correr 'to run' corrida 'race', beber 'to drink' 15 16

See Spitz (2001:23). Does also exist in Tagalog with identic meaning. The Cebuano synonym is tigpatulin.

Patrick Steinkrüger

260

-> bebida 'drink'). 1 7 In Chabacano -da is a mere derivational suffix added to a verbal stem in many but not all cases meaning 'act or manner of X-ing', a meaning which is already found in Spanish. So you have in Chabacano many forms which in Spanish are not formed with -da: (6)

Spanish pesca bajada compra pregunta

Chabacano peskäda abahäda kompräda preguntäda

'fishing' 'descent, way down' 'shopping' 'question, query'

From a synchronic point of view, -da in Chabacano is not transparent in many cases, a situation similar to that in Spanish. Compare the following examples: (7)

marihäda apuntäda kerida eskribida

'wave' 'heading' 'mistress' 'writings'

(< Span marejada) ( tampa-. Compare the following examples: (8)

gwäpa riko kyeto durmi

'pretty 'rich 'quiet' 'sleep'

->

man-pa-gwäpa 'beautify; act pretty' 'ta-m(an)-pa-riko 'pretend to be rich' ta-m-pa-kyeto-kyeto 'pretend to be quiet' ta-m-pa-durmi-durmi 'pretend to sleep'

To summarize, it is important to mention that the source of the root determines the restrictions on possible combinations and that the productivity of Spanish affixes is in many cases not certain (in contrast to the analysis of Forman 1972 and Riego de Dios 1976). More data are needed for a more detailed analysis of this problem.

4. Reduplication

Apart from affixation reduplication is the most productive morphological process in Chabacano. In this paper we deal only with cases of full repetition of morphemes, partial reduplications have not been observed. Diachronically we distinguish the repetition of words from Spanish origin and of Philippine origin. All tokens of the latter kinds in my 17

For the function of -da in Spanish cf. Rainer (1993:437ft)·

Wordformation in Philippine Spanish Creole

261

data could be identified as lexicalized unities. Very few words of English origin can be observed in their reduplicated form. As to the ad- and substrate languages of Chabacano, reduplication is a very productive morphological process in Austronesian languages (although their function is manifold). Reduplication (repetition of the whole lexical item) is also known in Spanish, although it is not very productive. Normally it intensifies adjectives (e.g. grande 'big'—» grande-grande 'very big') and with nouns it can indicate their authenticity (e.g. cafe 'coffee' —> cafe-cafi 'real coffee'). Considering the few data I have gathered it is not possible to clear in detail the function and productivity of reduplication in Chabacano; the statements here should be considered as a preliminary analysis. Nevertheless the four main groups of reduplicated forms with their functions respectively could be classified as follows: • entire loanwords from Philippine languages; cf. 4.1. • derivation of nouns (semantically often by metonymy); 18 cf. 4.2. • intensified or diminished qualities, quantities or actions (borderline to inflection); 19 cf. 4.3. and 4.5. • indefiniteness of interrogatives; cf. 4.4. • onomatopoetica, e.g. sirik-sirik 'cicada' (cf. Hil sirum-sirum, siriritan, kikik)

4.1. Reduplicated words from Philippine languages Some Philippine source reduplications do not occur in a non-duplicate form: duha-duha 'hesitate, hesitant', utay-utay 'slowly by stages', kyaw-kyaw (Phil) 'fuss', bitaybitay (Adj) 'hanging without purpose', hinay-hinäy ,slowly', ungul-ungul 'to grumble', pang-pang (Tag) 'riverbank', halo-hälo 'a dessert', gaya-gaya (Tag) ,to ape, to imitate', galung-galung (Ceb) 'a ring attached through the nose of domesticated farm animals, such as cows or carabaos', lapu-lapu (Tag) 'rock bass, spotted grouper', lawa-lawa (Tag) 'spider web', lutaw-lutaw (Ceb)'floating about', daliq-ddliq 'quick(ly)'.

4.2. Nouns The reduplication of nouns in Chabacano can form new nouns or adverbs:

(9)

18

19

päya hente käsa pärti tyempo

'coconut shell' 'people' 'house' 'party' 'time' (< Span tiempo)

- >

- > - >

- >

paya-päya 'kneecap' (like a paya) hente-hente 'scarecrow, effigy of a person' kasa-käsa 'playhouse' parti-pärti 'something like a party' tyempo-työmpo 'from time to time; timely'

For quantitative reasons of tokens it would not be convenient to formulate a more precise statement about the derivational function of reduplication (e.g. restriction to semantical classes etc.). The intensification and similar functions as repetition of verbal constructions belong crosslinguistically to the inflectional category of aktionsart.

Patrick Steinkrüger

262 4.3. Adjectives

There are adjectives which are formed by reduplicated adjectives, nouns or verbs. Reduplication of adjectives indicates approximation of the respective quality: (10) loko grande diiro pronto apurä dyutay berde bibo largo

'crazy' (< Span 'big' (< Span 'hard' (< Span 'quick' (< Span 'to hurry'+ -w 'small' (Hil) 'green' (< Span 'alive' (< Span 'long'

loco) id.) duro) id.)

loko-loko 'crazy' -> grande-grande 'very big' -> duro-duro 'somewhat hard' pronto-pronto 'very quickly' -»• apuräw 'quick' apura-apurä-w 'very quickly' dyutay-dyutay 'very small' verde) berde-berde .greenish; very green' vivo) bibo-bibo 'alive' (diference unclear) largo-largo 'longish'

Some adjectives are derived from verbs: (11) tyene 'there is, to have' (< Span tiene 'she/he has') —> tyene-tyene (Adj) 'well to do' dura 'to take a long time' (< Span durar) —> dura-durä (Adj) 'long' Different distributive functions are expressed by reduplication: (12) unu-iinu 'one-by-one' doble-doble 'in doubles, doubly' dos-dos 'by two's; by pairs' solo-solo 'alone, singly' huntu-huntu 'together as a group' Adjectives of likelihood can be formed by reduplicated nouns: (13) pato 'duck' (Span) pato-päto 'ducklike, especially referring to manner of walking' doktor 'doctor' - » doktor-doktor 'playing at doctor; fake doctor' sänto 'saint' (Span) -» santo-santo 'acting like a saint, saintly' ekis 'X' (Span) -» ekis-ekis 'X-formed' lenggwa 'tongue'(Span) —» lenggwa-lenggwa 'tongue-like forms or shapes' A lot of other adjectives are formed by reduplicated nouns: (14) dya 'day' diya-diya 'daily; everyday; day after day' (syn. dyaryo < Span diario) swerte 'chance' (Span) —> swerte-swerte 'by more luck or chance' pila 'line'(< Span fila) pila-pila 'line by line; in line' pitik 'dot' (Phil) pitik-pitik 'dotted' kanto 'edge' (< Span canto) -» kanto-kanto N/Adj 'at the very edge'

Word formation in Philippine Spanish Creole

263

4.4. Interrogatives Reduplicated interrogatives express indefiniteness: (15) kosa kyen donde onde

'what' - > 'who' 'where' - » 'where' —>

kosa-kosa 'what all, whatsoever, all sorts of things' kyen-kyen 'whoever' donde-donde 'whereabouts; everywhere' onde-onde 'wheresoever'

4.5. Verbal Reduplication A lot of verbs are formed by reduplication of other verbs but also nouns or adjectives. Especially with verbs from Spanish origin, continuity of the action is indicated. Non-verbal roots are combined in their reduplicated form with a verbalizing prefix. For the combination with verbalizing prefixes see the end of section 3. Compare the following examples: (16) bisyä rabyä rebahä tintin kurtä kaminä parti llorä kantä lyibä kisäp bulak kähig

'to watch' 'to be angry' 'to bargain' 'to stand on tiptoe' 'to cut' 'to walk' 'to devide' 'to cry' 'to sing' 'to carry, bring' 'winking, blinking' (N) 'blind'

-> —> —> —» -» -> -> -> —> ->

'rake'

—•

praktis 'practice' (N)

->

bisya-bisya 'look around continually' rabya-rabyä 'angered (much or little?)' rebaha-rebahä 'to bargain (a bit)' tintin-tintin 'to hop 'a children's w o r d " kurta-kurtä 'to cut or chop into (small) pieces' kamina-kaminä 'to stroll' parti-parti 'to devide into several pieces' llora-llorä 'to pretend to cry' kanta-kantä 'sing continously' lyiba-lyibä 'carry continually' man-kisap-kisäp 'wink or blink continually' manpa-bulak-bulak 'pretend to be blind' VERB.blind 2 man-kahig-kahig 'scratch leisurely' VERB.rake 2 man-praktis-präktis 'practice continually' VERB.practice 2

5. C o m p o u n d i n g

The combination of more then one lexical item is also very common in Chabacano. This morphological process can be observed chiefly with nouns. In all cases in my data compound stress is assigned to the rightmost member of the compound. In some cases it is somewhat difficult to say whether a compound is a wholesale copy of the Spanish (i.e. a lexicalization of a binominal syntagma) or whether it is the result of a productive process.

264

Patrick

Steinkrüger

Nominal compounding works either with a zero-marker or with the Spanish linkers de, del (< Span de + el, 'from' plus masculine singular article) andj> (< Spany 'and'). 20

5.1. Lexicalised compounds adopted directly from Spanish or English Some nouns or nominalized syntagmas cannot be regarded as reanalysed as compounds by Chabacano native speakers because there were taken from Spanish or English as a single lexical unit. No semantic shift has taken place. The following compounds belong to this group: binagre (< Span vinagre) 'vinegar', bagamundu (< Span vagamundo) 'vagabond; wanderer', aeroplano (< Span id.) 'plane', portamoneda (< Span id.) 'purse', lentehwelas (< Span lentejuelas) 'sequins', noche bwena (< Span noche buena) 'Christmas', luna-llena (< Span luna llena) 'full moon', dama-hwana (< Span damajuana) 'demijohn', iday-vuelta (< Span iday vuelta) 'back and forth', didiya (< Span de dia 'in the day'), kubre-kama 'bed cover' (< Span cobrecama). Finally an example from English: blakbord 'blackboard'.

5.2. Compound words without linking element Some compound words in Chabacano, consisting of two nouns, lack a linking element, contrary to Spanish. A few examples may illustrate this: (17) karnebaca karne pwerko corta pelo

'beef 'pork' 'hair cut'

(cf. Span came de vaca) (cf. Span came depuerco) (cf. Span corte del cabello)

Other compound constructions of two nouns serve to mark gender distinction whereas this is marked by suffixes in Spanish. This process is typical of many other Creoles. Compare the following example: (18) bata mujer (< Phil bata 'child' + Span mujer 'woman') 'girl' bata hombre (< Phil bata 'child' + Span hombre 'man') 'boy'

(cf. Span muchacha) (cf. Span muchacho)

A lot of analytical verb constructions (V+N) contrast with their monolexical equivalents in Spanish or even in Philippine languages. They are lexicalised syntagmas and can therefore treated as complex verbs. Compare the following examples: (19) tirapusil pone iniksyon dale presta maska biiyu

20

'to 'to 'to 'to

shoot with a gun inject' lend' chew betel nut'

(cf. Span fusilar) (cf. Span inyectar) (cf. Span prestar) (cf. Tag man-nganga)

I follow here the tradition of Romance linguistics to take a binominal construction with a preposition (ligature prepositionelie) as a compound.

Wordformation in Philippine Spanish Creole

265

As in other Creole languages, some interrogatives are bimorphemic, i.e. the combination of an interrogative element and a noun. They often contrast to their equivalents in the lexifier languages; for example: kemodo (< Span que modo 'what manner') 'how' (= Span como) and with the same meaning kemanera (< Span que manera 'what manner'); also kelaya (< Span que laya 'what quality, kind').

5.3. Compounds with linking element In Chabacano as in Spanish, a lot of compound nouns (N+N constructions) are found with a linking element. The most common are de and del, whereas y plays a marginal role in compounding. Compare the following examples: (20) suput de plastic 'plastic bag' < Tag supot 'bag' + Engl plastic (cf. Span bolsa de plästico) agua de olor 'perfume; lotion' < Span agua 'water' + Span olor 'scent' (cf. Span perfume) anäk de apwera 'illegitimate child' < Phil anak 'child' + Span afuera 'outside' (cf. Span hijo ilegitimo, nacido fuera de matrimonio) dedo de pies