Synchronic and Diachronic Perspectives on Contact Languages (Creole Language Library, Volume 32)

Synchronic and Diachronic Perspectives on Contact Languages

Creole Language Library (CLL) A book series presenting descriptive and theoretical studies designed to add significantly to the data available on pidgin and creole languages.

Editors Miriam Meyerhoff

University of Edinburgh

Umberto Ansaldo

University of Amsterdam

Editorial Advisory Board Mervyn C. Alleyne

Pieter Muysken

Marlyse Baptista

Peter Mühlhäusler

Nicholas Faraclas

Shobha Satyanath

George L. Huttar

John Victor Singler

John Holm

Norval Smith

Sylvia Kouwenberg

Sarah G. Thomason

Susanne Michaelis

Tonjes Veenstra

Kingston, Jamaica

Athens, Georgia, USA Rio Piedras, Puerto Rico Dallas

Coimbra

Kingston, Jamaica Leipzig

Nijmegen Adelaide

University of Delhi New York

Amsterdam Ann Arbor Berlin

Salikoko S. Mufwene Chicago

Volume 32 Synchronic and Diachronic Perspectives on Contact Languages Edited by Magnus Huber and Viveka Velupillai

Synchronic and Diachronic Perspectives on Contact Languages Edited by

Magnus Huber University of Giessen

Viveka Velupillai

John Benjamins Publishing Company Amsterdam / Philadelphia

8

TM

The paper used in this publication meets the minimum requirements of American National Standard for Information Sciences – Permanence of Paper for Printed Library Materials, ansi z39.48-1984.

Library of Congress Cataloging-in-Publication Data Synchronic and diachronic perspectives on contact languages / edited by Magnus Huber, Viveka Velupillai. p. cm. (Creole Language Library, issn 0920-9026 ; v. 32) Includes bibliographical references and index. 1. Creole dialects. 2. Languages, Mixed. I. Huber, Magnus. II. Velupillai, Viveka, 1974PM7831.S96    2007 417'.22--dc22 2007028073 isbn 978 90 272 5254 8 (Hb; alk. paper)

© 2007 – John Benjamins B.V. No part of this book may be reproduced in any form, by print, photoprint, microfilm, or any other means, without written permission from the publisher. John Benjamins Publishing Co. · P.O. Box 36224 · 1020 me Amsterdam · The Netherlands John Benjamins North America · P.O. Box 27519 · Philadelphia pa 19118-0519 · usa

We dedicate this volume to Aunt Lischka and Uncle Lois in gratitude for their tireless care and devotion to our beloved Mama

Dieses Buch widmen wir Lischka und Lois, in Dankbarkeit für all die Jahre unermüdlicher und liebevoller Fürsorge für unsere “Mama”

Table of contents Preface Part I . Maintenance or assimilation? Phonological variation and change in the realization of /t/ by British Barbadians Michelle C. Braña-Straw

  

.

Universal and substrate influence on the phonotactics and syllable structure of Krio Malcolm Awadajin Finney



.

Tone on quantifiers in Saramaccan as a transferred feature from Kikongo Marvin Kramer



.

Morphophonological properties of pitch accents in Jamaican Creole reduplication Shelome Gooden



.

Effort reduction and the grammar: Liquid phonology in Haitian and St. Lucian Eric Russell Webb



Part II . Reflexivity in Capeverdean: Predicate properties and sentence structure Maria Alexandra Fiéis and Fernanda Pratas

 

.

An additional pronoun and hierarchies in Lower Columbia Chinúk Wawa David D. Robertson



.

Three irregular verbs in Gullah David B. Frank



.

Afro-Bolivian Spanish: The survival of a true creole prototype John M. Lipski



.

Copula patterns in Hawai‘i Creole: Creole origin and decreolization Aya Inoue



 Table of contents

Part III . On the properties of Papiamentu pa: Synchronic and diachronic perspectives Claire Lefebvre and Isabelle Therrien

 

.

No exception to the rule: The tense-modality-aspect system of Papiamentu reconsidered Nicholas Faraclas, Yolanda Rivera-Castillo and Don E. Walicek



.

A look at so in Mauritian Creole: From possessive pronoun to emphatic determiner Diana Guillemin



.

Chinese Spanish in 19th-century Cuba: Documenting sociohistorical context Don E. Walicek



.

Comparative perspectives on the origins, development and structure of Amazonian (Karipúna) French Creole Jo-Anne S. Ferreira and Mervyn C. Alleyne



Index



Preface The present volume is a selection of papers presented at the Society for Pidgin and Creole Linguistics meetings in Curaçao (August 2004), Oakland (January 2005) and Albuquerque (January 2006). A total of twenty-five papers were submitted to this volume. Each article was peer reviewed by two, in some cases three, external specialists. We accepted fifteen submissions, all of which are much revised and extended versions of the conference papers, and all of which we consider valuable contributions to the field Contact Linguistics. The articles in this volume do not only present a wealth of data, something which in itself could certainly be considered the backbone of all linguistic research. They also address a number of issues relevant far beyond the study of pidgin and creole languages. The present volume thus brings together a beneficial cross-fertilization of approaches to language studies, such as those discussing the role of identity when explaining phonological developments (see the article by Michelle C. Braña-Straw, on British Barbadian English, but also Marvin Kramer’s discussion on tone in Saramaccan); when tracing the demographic patterns of plantation development (in Don E. Walicek’s article about the intermediary role of Chinese Spanish in Cuba); or when analysing geographically hidden varieties (as John M. Lipski does in his article about Afro-Bolivian Spanish). Although the respective contributions to this volume are quite diverse in nature and topic, they nonetheless have some overlapping interests. Thus, while a number of authors focus more or less on synchronic descriptions of a variety (Michelle C. Braña-Straw, Shelome Gooden, Maria Alexandra Fiéis and Fernanda Pratas, David D. Robertson, David B. Frank, John M. Lipski, Aya Inoue, Claire Lefebvre and Isabelle Therrien), the approach they choose also furthers the discussion on related issues, for instance sociolinguistics, as mentioned above; functional linguistics (see, for example, the article on animacy hierarchies in Chinúk Wawa by David D. Robertson, or the various semantic properties in Papiamentu pa by Claire Lefebvre and Isabelle Therrien); Optimality Theory (such as Shelome Gooden’s paper on pitch accents in Jamaican Creole reduplication, as well as Eric Russell Webb’s discussion on liquids in Haitian and St. Lucian). Conversely, those who set out to give diachronic analyses of their data (Eric Russell Webb, Malcolm Awadajin Finney, Marvin Kramer, Claire Lefebvre and Isabelle Therrien, Nicholas Faraclas, Yolanda Rivera-Castillo and David E. Walicek, Diana Guillemin, Jo-Anne Ferreira and Mervyn C. Alleyne), also contribute significantly to the general discussion of syntax (for instance, the article by Nicholas Faraclas,



Preface

Yolanda Rivera-Castillo and David E. Walicek on the origin(s) of the Papiamentu tense-mood-aspect system); the issue of grammaticalization (see Diana Guillemin’s analysis of the determiner so in Mauritian Creole, as well as the discussion by Malcolm Awadajin Finney on the phonotactics and syllable structure of Krio). While there are no doubt several ways to group the contributions of this volume, we have opted for the following thematic subsections: Part I:

Papers dealing with phonetics and phonology, both from a synchronic and a diachronic perspective: Braña-Straw, Finney, Kramer, Gooden, Russell Webb Part II: Papers with a more synchronic perspective on their language(s): Fiéis/Pratas, Robertson, Frank, Lipski, Inoue Part III: Papers with a more diachronic perspective on their language(s): Lefebvre/Therrien, Faraclas/Rivera-Castillo/Walicek, Guillemin, Walicek, Ferreira/Alleyne As already mentioned, most of the articles cover, or are in various ways significant for, more than one linguistic subgrouping. However, we find that the predominantly phonetic/phonological approach tie the first five chapters in this volume together into a coherent section. By examining a stereotypical feature of British Barbadian speech, the variable /t/, Braña-Straw discusses the assumption that the speech of all AfroCaribbeans, irrespective of island of origin, in Britain tends to drift towards Jamaican Creole. She effectively shows how linguistic strategies of simplification, leading to maintenance and assimilation of the salient feature, are being employed as acts of identity by British Barbadians. In Chapter 2, Finney analyses the phonotactics and syllable structure of Krio in an effort to determine whether what can be found is mainly attributable to first language acquisition or whether the features are influenced by the West African substrate languages. By showing that Krio essentially exhibits phonologically unmarked features – which strengthens a universalist explanation of its origins – as well as phonologically rather marked features – which in turn supports a substratist view – Finney establishes that Krio, like so many other languages (whether creole or not) does not readily conform to any of the two explanations. In the same vein, Kramer discusses the less than straightforward shaping of Saramaccan: while the main substratal influence on the English-lexified creoles of Suriname is generally agreed to be Fongbe, the tonal characteristics of quantifiers in Saramaccan are better explained if Kikongo is seen as the source. Part I of the volume concludes with two articles devoted to the Optimality Theory framework. Gooden takes a synchronic perspective on Jamaican Creole, and shows how stress is the main distinguishing factor between segmentally identical reduplications. Russell Webb uses Optimality Theory as a diachronic tool to explain the surface variability of liquids in Haitian and St. Lucian and demonstrates how the respective outcomes both involve articulatory effort reduction. The five papers of the second section in this volume share predominantly synchronic analyses of their data. Fiéis and Pratas use a Principles and Parameters framework to

Preface

show that Capeverdean does not have a functional projection for reflexive contexts, even in situations when the reading is obligatorily reflexive. In comparing with European Portuguese strategies in the equivalent domain the authors propose an internally motivated strategy in Capeverdean, and thus by extension contribute to the diachronic discussion of that language. Similarly, Robertson’s in-depth analysis of 3.singular in Chinúk Wawa not only furthers our understanding of the history and development of Salish languages, but also adds significant linguistic material to general typological issues of animacy and markedness. Another valuable contribution to the linguistic community in general is the analysis of Gullah verbs by Frank. He shows that this classic language in creolistics debates has three irregular verbs, all of which in turn are related to the most common tense-moodaspect markers, a pattern that has been observed in numerous non-creole languages as well. Latching on to the debate of creole prototypes and decreolization, Lipski points to the data of Afro-Bolivian Spanish, the now rapidly vanishing variety spoken in remote areas of Bolivia by descendants of Africans who, for various reasons, arrived in the 1500s. Using a multivariate analysis of copula absence in Hawai῾i Creole (HC), Inoue shows that there is little or no basis to assume that decreolization processes have taken place on the O῾ahu island during the last 30 years. She also shows that the copula patterns of HC are similar to copula patterns in other English Based Creoles, notably those in the Caribbean and African American Vernacular English, and as such gives a non-Atlantic case of a feature often assumed to be due to African substrate languages. We consider the article by Lefebvre and Therrien an appropriate bridge between the more synchronically oriented (Part II) and the more diachronically oriented (Part III) sections of the volume. They first give an in-depth analysis of the properties of Papiamentu pa, drawing on a wealth of data pulled together from various sources. The authors then proceed to demonstrate why they consider the historical origin of pa a relexified form of Portuguese para, with its functions derived from the West African substrate languages. Papiamentu and its West African input languages are also discussed by Faraclas, Rivera-Castillo and Walicek, who focus on the tense-moodaspect system and show that those features of Papiamentu that have been considered ‘anomalous’ actually display patterns of change typically found in languages of southern Nigeria, which are typologically and genetically relevant in the study of African substrate input on Atlantic Creoles. Likewise, in tracing the development of the Mauritian Creole (MC) determiner system, Guillemin draws on cross-linguistic patterns of grammaticalization and argues that so, from French son – nowadays only used in MC as an emphatic determiner in both singular and plural NPs – is a legacy of its use in genitive constructions as a 3.singular possessive pronoun. Walicek bases his discussion of Chinese Spanish in Cuba on extensive archival research, and provides the field of pidgin and creole studies with a detailed picture of the sociodemographic context of the plantation system on the island. He shows that, with the arrival of the large number of Chinese indentured labourers, those who became bilingual in Spanish essentially formed an intermediary group between the



 Preface

Asian labourers and the African-born slaves of the plantations, thereby adding another dimension in the linguistic development of the contact language. We conclude the volume with the contribution by Ferreira and Alleyne, who use a historical linguistic comparative method to trace the origins and development of two separate varieties of Amazonian French Creole spoken in the very North-East of Brazil: Karipúna French Creole and Galibi-Marwono French Creole. The authors make use of both contemporary data obtained in field-work as well as data from archival secondary sources and establish that there has been a history of contact between the people on the mainland and speakers of Antillean French creoles in the South-Eastern Caribbean islands since the 16th century. Our main thanks goes to Bettina Migge for having got the whole process of this volume started. We would also like to thank the peer reviewers for their conscientious work, as well as, of course, all the contributing authors for all their efforts in advancing not only the field of pidgin and creole studies, but also the broader field of linguistics in general. Magnus Huber University of Giessen

Viveka Velupillai

 

 

Maintenance or assimilation? Phonological variation and change in the realization of /t/ by British Barbadians* Michelle C. Braña-Straw University of Essex

Studies from the 1970s and 1980s of the linguistic behaviour of Afro-Caribbeans in Britain argued that first and second generation Jamaicans were maintaining Jamaican Creole features and that Afro-Caribbeans were also acquiring Jamaican Creole regardless of island of origin, even in areas where Jamaicans were not the most dominant subgroup. Much of the discussion of this time has concentrated on morphosyntactic features. This research compares the speech of Barbadians to their Anglo1 counterparts in an area not completely dominated by Jamaicans (Barbadians are almost as numerous) and focuses particularly on the linguistic behaviour of one Barbadian speaker (Edward) who arrived at age 9, in the critical period for acquisition. In addition, this paper extends earlier research by considering changes to the phonology of Barbadians and examines the realization of the variable /t/, a feature which is commonly found in Barbados and Britain, but not in Jamaica. Furthermore, this feature is stereotypical of Barbadian speech to speakers of other Caribbean varieties in the Caribbean, and of vernacular varieties of British speech. An examination of the changes to this salient linguistic variable reveals a complex set of interactions between linguistic, psychological and social factors leading to both maintenance and assimilation amongst the Barbadians. Edward uniquely employs linguistic strategies of simplification – the reduction in number of linguistic forms, and overgeneralization – which in this study refers to increased rates of usage of a variant and extension of a variant to

* This research is supported by the Economic and Social Research Council (grant no. PTA030-2002-01347). I should like to thank Peter Patrick for his comments on previous versions of this paper and to the anonymous reviewers for their helpful suggestions. Any remaining errors are my own responsibility. . For the purpose of this research, Anglos are defined as those speakers who self-identify as racially ‘white’ and who have been born and raised in Britain, along with their parents and grandparents.



Michelle C. Braña-Straw

new environments. Edward’s intermediary status between the two communities allows him to employ these two linguistic strategies to affiliate to Anglo norms in certain environments and to Barbadians norms in others.

.

Introduction

The contact literature largely agrees that, the pressure for minority linguistic communities to assimilate to the majority community is overwhelming. Horvath (1991) for example, comments that minority immigrant groups must make many kinds of changes, both social and linguistic, in order to join the host community. Racial and ethnic difference between communities in contact is often a complicating factor that could lead to a stable bilingual situation in minority ethnic communities. The early BrACE literature, which had studied first and second generation Afro-Caribbean communities in Britain, concluded that such a situation existed for Afro-Caribbeans who were maintaining Creole features, whilst acquiring the English of local communities, and from which they were indistinguishable to all but very local people (Hewitt 1986: 139; Sebba 1993: 58–64). The latter assertion has yet to be adequately tested (but see Wells, 1973 who reports on first generation Jamaicans’ acquisition of British English phonology). Modood et al. (1997) conducted a sociological survey of ethnic minority groups in Britain and they report that there was a resurgence of Creole in second generation Afro-Caribbeans linked to ethnic identity. Their finding supports the claims of the BrACE literature. They also report, however, that the meaning of ethnic identity for Afro-Caribbeans had become politicised and highly unstable. Thus individuals developed a range of responses to the host community and asserted a variety of identities that might or might not include racial/ethnic identities. In this unstable situation, Afro-Caribbeans in general, and Ipswich Barbadians in particular, must negotiate their position within mainstream society. It is likely that the individual’s response to a range of social and interpersonal factors will determine their linguistic behaviour. This paper presents a sociophonetic study of the language change of first and second generation Barbadian Creole speakers who migrated to Ipswich, in Suffolk, Eastern England. It extends the previous study reported in Straw and Patrick (2007) and Straw (2006) and employs fine-grained acoustic analysis to discover complex social patterns which differentiate the Barbadians from the Anglos. I report on the (t) variable: (1) what, want – word-final, and (2) better, country – word-medial, with particular focus on one speaker (Edward) who migrated from Barbados to Ipswich, aged 9. Glottalisation of /t/ is a shared feature for both groups. In England, it is stereotypical of vernacular British speech, particularly intervocalically, e.g., water, but is not associated with any one variety. In Barbados, it is recognized by Afro-Caribbeans as stereotypical of Barbadian speech. Wells (1982) suggests that in Barbados, /t/ is glottalised wordfinally and word-medially but not in intervocalic environments i.e., what and country, but not better, while British varieties also glottalise in intervocalic environments.

Chapter 1. Maintenance or assimilation?

The Barbadians in this study have lived in Ipswich for at least 25 years and arrived at different ages (adults and children). Research reported in (Straw and Patrick 2007), discusses the language behaviour of the Barbadians as a group. Edward is selected for further investigation, because he was at a critical age for acquisition when he arrived in Britain (aged 9), whereas the other Barbadians came as adults or teenagers. This study has three main research questions: 1. 2. 3.

.

How distinct are the Anglo and Barbadian patterns for (t)? Has Edward wholly assimilated to local Ipswich Anglo patterns, or has he retained Barbadian patterns for this variable? What role do social and individual factors, such as age of arrival and social networks play in the acquisition of Edward’s patterns?

Research issues relating to British Afro-Caribbean English speakers

Initial studies of speakers of Afro-Caribbean background in Britain, conducted during the 1970s and 1980s, have mostly focused on morphological and syntactic Creole features of adolescents of Jamaican background (but see Sutcliffe 1982 Chapter 4; Wells 1973), reflecting the ‘lack of published work on the phonetics and phonology of the English spoken in ethnic subcommunities’ in general (Foulkes and Docherty 1999: 16). Jamaicans accounted for just over half of the immigrants to Britain during the 1950s and 1960s. Most research concentrated on Creole features in areas where Jamaicans were the greatest Afro-Caribbean group and studies in the 1970s and 1980s reported that British Afro-Caribbeans were acquiring Jamaican Creole features irrespective of their island of origin. For example, Sutcliffe (1982: 138) reports the acquisition of Jamaican Creole features in Bedford by non-Jamaicans where Jamaicans outnumbered any other Afro-Caribbean group. However, he mentions some continuing influence from the Eastern Caribbean. More unusually, Tate (cited in Sebba 1993: 138–9) found that Jamaican features were acquired by Afro-Caribbeans in Bradford, where Dominicans (not Jamaicans) were the dominant group. A strong theme running through the literature at the time is the notion that island differences were apparent in first generation AfroCaribbean speakers but that Jamaican Creole came to represent a unifying language of resistance to mainstream society for second generation Afro-Caribbeans, leading to a British Black identity. The early work in Britain was prompted by educationalists concerned about the poor treatment of Afro-Caribbeans in schools, in contrast to other non-native English speakers whose language background was taken into account (Rosen and Burgess 1980; Wight and Norris 1970). In their efforts to contrastively describe the systematic structural properties of Creole features, the emphasis was on maintenance of those Creole features. Although it was assumed that the English spoken by British born Afro-Caribbeans was virtually indistinguishable from that of their white counterparts, and that their primary dialect was a local British English vernacular, the English of British Afro-Caribbeans remains to be fully explored (but see Sebba 1993





Michelle C. Braña-Straw

Chapter 5; Wells 1973). Most of the early research concentrated on the maintenance of morphosyntactic creole features. More recent work on British Afro-Caribbean English (BrACE) indicates a complex set of responses in the linguistic behaviour of Afro-Caribbean speakers in contact with native varieties, involving maintenance and processes of assimilation resulting in a high degree of variation both at the level of the community and the individual. For example, Knight, Patrick and Straw (2002) describe the language behaviour of two British-born brothers with Jamaican parents. Young men at the time of the study, both sounded very local, but the younger brother had acquired Jamaican features and constraints which the older brother had not. Initial work on Barbadian adults, resident in Ipswich for at least 25 years found that for /t/-glottalling, Barbadians seemed to use non-local patterns in certain contexts but adapt to local Ipswich patterns in others (Straw and Patrick 2007). This paper extends the previous study reported in Straw and Patrick (2007) and Straw (2006). I report on the variable (t) in word-final (what, want) and word-medial (better, country) positions. Principal variants include released [t] and glottal consonants, e.g., [‘, t‘, ‘t]. The principal research objectives are: 1. 2. 3. 4.

.

To answer the call for phonological studies of Creole speakers and of Barbadian Creole speakers in particular (Burrowes 1983: 40; Van Herk 2003: 259); To extend the research of British Black English with its emphasis on the morphology and syntax of Jamaicans in Britain (Edwards 1986; Sebba 1993; Sutcliffe 1982); To illustrate it with particular focus on one speaker, who arrived at a critical age for the acquisition of a new dialect; To explore the crucial role of social and individual factors in second dialect acquisition, employing a dialect contact and acquisition frame-work, to examine the range of variation for (t) for Barbadian immigrants to Ipswich.

Overview of the study

I examine data from first and second generation Barbadian immigrants (male and female) living in Ipswich, Suffolk. First generation Barbadians arrived in Ipswich as adults, whilst second generation speakers came as children. All the participants have lived in the Ipswich for over 25 years and their ages range from 40 to 75 years old. Historically, Suffolk people were amongst the first settlers to Barbados, which was one of the earliest Caribbean islands settled by the British in 1625 (Niles 1980: 4). The perception of Barbadian English in Ipswich as ‘local and yet distinctive’ may be due to features shared with the Suffolk dialect, the most salient of which is the glottal stop replacement of /p, t, k/. Among Caribbean English varieties, glottal replacement as a stereotypical feature of Barbadian has often been noted, but rarely studied (Blake 1997: 34; Le Page and Tabouret-Keller 1985: 93; Roberts 1988: 92; Wells 1982: 584; but see Blake 1994). This reflects a general lack of phonological studies in Caribbean Englishes.

Chapter 1. Maintenance or assimilation?

.

Developmental factors – Critical period for acquisition

For an individual acquiring a second dialect, the degree of success depends on a number of factors: developmental, linguistic and psychosocial. Age of first exposure has the strongest effect on the success of acquisition: the younger the child, the greater the success. Studies seem to confirm that children can acquire a second dialect to a nativelike standard, before the age of 8, but this ability is lost after the age of 14. Between the ages of 8 to 14, acquisition depends on a number of psychosocial factors (Chambers 1992: 689; Kerswill 1996: 192; Payne 1980: 155–6; Trudgill 1986: 33).

.

Linguistic factors: Salience

Trudgill (1986: 14, 20) predicts that salient features will be selected for acquisition because they are high in the speaker’s consciousness. However other important factors could impede the learning of a feature, for instance where the feature is too salient and negatively stereotyped, or where it involves learning new rules and constraints. In Britain, word-medial and word-final environments e.g., better, bet, when /t/ is realized as [t] or [t‘], it is associated with careful, formal and ‘posh’ speech, whereas when /t/ is realized by glottal variants (see Section 4 for description), it is associated with careless, informal, rough or lower class speech. Studies have found that glottal stops are most frequent word-finally before a consonant – the least salient environment, but least frequent within a word and between two vowels (intervocalically) – a stereotype of ‘careless’ or working class speech (Tollfree 1999: 171; Trudgill 1986: 20; Williams and Kerswill 1999: 159). Applying Trudgill’s (1986: 10–1) salience principle, I predict that acquisition will occur in salient environments, i.e.,: ●

● ● ●

.

word-finally: before a pause – e.g., you what? before a vowel – e.g., get a life! but not before a consonant – e.g., Have you got the paper? – where it is not salient word-medially – e.g., country, factory but not intervocalically – e.g., water, butter – where it is stigmatized.

Social factors: Networks and peer group orientation

Kerswill and Williams (2000) suggest that a child’s adoption of features depends on his or her peer group orientation and social networks. They describe children’s acquisition of a complex phonological feature in Milton Keynes, the Roland/rolling distinction. Monomorphemic words e.g., Roland, solar, cola, solo, etc. have a front allophone [ә~] while dimorphemic words e.g., rolling [roll + ing] have a back allophone [ɒ~]. Generally children born in Milton Keynes had acquired the distinction. However, they compared three brothers, who were born in Milton Keynes and who had Asian parents. The authors found that the twin boys, aged 9 had mainly Asian friends and did not





Michelle C. Braña-Straw

acquire the distinction, whereas their brother, age 14 had mainly British friends and had acquired the distinction. Social networks and peer group orientation were the decisive factor as to whether the boys acquired the distinction or not. Age was not at issue, as the boys had all been born in Milton Keynes. Rather than focus simply on synchronic social factors, in the present case, this study explores the current social networks and orientation of Edward, now aged 50, against the sociohistorical backdrop of the Barbadian experience in Ipswich. I argue that ongoing language change reflects changes in social networks throughout an individual’s lifetime.

.

Sociolinguistic setting

The typical pattern of settlement for Afro-Caribbeans was through family networks and many have extended families living in Ipswich. Not uncommonly, these networks have extended into the Anglo population through mixed marriages or relationships, and it is not unusual to see mixed parentage 3rd and 4th generation Afro-Caribbeans and Anglos.2 Although migration largely ended at the end of the 1960s, many AfroCaribbeans maintain family contacts with their Caribbean country of origin, and relatives living in the US, Canada and other places, through visits and extended stays, as well as contacts with family in other British towns, most notably London.

.

Ipswich

Ipswich is the administrative centre for the rural county of Suffolk, located in the Eastern part of England. It has a relatively small population in British terms, of about 100,000 of which about 10% are from ethnic minorities, with just 2.7% being AfroCaribbean (Census 2001). However, despite their small numbers, Afro-Caribbeans form a cohesive group and are heavily represented in certain occupations, most notably health and social welfare, factory work or local government. In Ipswich today, a range of services are available specifically catering for AfroCaribbeans (for example, export shipping, hairdressers and hair products, social clubs, music and radio, Caribbean foods, black churches) which draw other Afro-Caribbeans from neighbouring towns. I myself am not unusual in travelling the 30km from my home to visit my regular hairdresser in Ipswich. My own parentage is racially mixed, Jamaican and Spanish. Although I was not brought up in Ipswich, I had worked in the town for several years before beginning the fieldwork and regularly participated in sociocultural events. As a musician, I already had access to social networks amongst local performers. . Interestingly, during the two years of fieldwork, I did not meet or hear of any mixed relationships between Afro-Caribbeans and other ethnic groups.

Chapter 1. Maintenance or assimilation?

The 1991 Census reports that the two largest groups settling in Ipswich during the 1950s and 1960s were Jamaicans (34%) and Barbadians (21%). Together, they made up over half of the Afro-Caribbean population of Ipswich. The remaining third came from other, mainly Anglophone, Caribbean countries. Compare this with national patterns of immigration in Britain during the same period, where 54% of AfroCaribbeans settling in Britain were Jamaicans and just 8% percent were reported to have immigrated from Barbados (Peach 1996: 28). By comparison, Barbadians represented a substantial Afro-Caribbean presence in Ipswich which remains the case today.3 Currently, it is difficult to gain an accurate picture of the Afro-Caribbean population in Ipswich due to changes in census reporting of ethnic groups, but figures do give an idea of comparative size of Afro-Caribbean subgroups compared with the national average.

.

The position of Barbadians in Ipswich

Barbadians were active in the early days of settlement in obtaining political representation and recognition for the rights and needs of the Afro-Caribbean community in Ipswich, culminating in the 1990s when the only two mayors from ethnic minorities were both Barbadian. They were instrumental in establishing an association to advocate on behalf of Afro-Caribbeans that still exists today. From the outset, Barbadians have worked closely with mainstream organizations in education, local government, the police and housing. The Barbadians in this study have been involved in this association at some level, either in its organization or through related community work. Both Gary and Edward have been chairmen of the association and many of the tensions between different island groups have been played out here (see below). The first and second generation Barbadians in this study identify strongly with their island affiliation, self-identifying as Barbadians or ‘Bajans’ but also being identified as Barbadians by other Afro-Caribbeans. Ipswich Afro-Caribbeans often comment that Barbadians and Jamaicans are the most distinctive Afro-Caribbean sub-groups, and sometimes polarized in terms of language, culture, temperament and attitudes to mainstream society. It seems that Barbadian and Jamaican stereotypes have travelled from the Caribbean (LePage and Tabouret-Keller 1985; Roberts 1988) and are being played out in diaspora communities in Ipswich and elsewhere in Britain. For example, the London Barbadians in

. The ethnographic fieldwork I conducted during 2000–2001 suggested that there remained a substantial established Barbadian community in Ipswich as compared to the Jamaican community. Whilst in- and out- migration has occurred since the first generation of immigrants, Barbadians have extended family networks living in Ipswich extending to three or four generations. Accurate numbers are not possible to obtain because subsequent generations will have been born in Britain. Country of origin refers only to first generation adult and second generation child immigrants and does not include British-born people of Barbadian heritage.





Michelle C. Braña-Straw

Western’s (1992) oral history collection displayed similar attitudes to Jamaicans that I found for Ipswich Barbadians. Below is a selection of the most often-heard comments which I collected from sociolinguistic interviews and participant observation. Table 1. Often-heard comments about Barbadians and Jamaicans Barbadians

Jamaicans

Not open to change Always want their own way Too close to whites Sound different from other Afro-Caribbeans Have a twang, dialect Small island Rude

Rough Arrogant Sound different from other Afro-Caribbeans Have a patois Big island Rude

Some of the more negative comments came out of general conversations or during the course of jokes in which I was present. In general, Anglos do not distinguish between different Afro-Caribbean groups unless they are in close contact with them, as evidenced by Margaret’s and Edward’s observations: Margaret, (Retired Barbadian): File location: Bessy2: 283 because of our visibility and our nationality, people would point us out, you know, “oh una Jamaic-” everybody was a Jamaican or American, because although we supposed to be so illiterate, they didn’t know anything about us, but we quite a bit…(about them). Edward, (Mid age Barbadian): File location: Bobby13_50 way way back when I was at school, there was a guy come over from Jamaica, and he was speaking, Jamaican patois ((_)) and um, I was there at school, I was about…, I think I must have been, I dunno, in the second year or something, and so was only young myself, and they said: “oh we get {Edward}, he can understand what they’re saying”, cause I went and thought to myself, and I couldn’t really work out, I knew some of what he was saying, but I couldn’t really work,

Chapter 1. Maintenance or assimilation?

so that was the first t- saying well: “oh crumbs! What do they… something else, why get me to do it?”, you know {laughter}

Both Margaret and Edward resist their identification by Anglos as Jamaicans. As mentioned earlier in this section, they retain their island affiliation which is partially expressed as difference from Jamaicans. This is unsurprising since the Jamaican presence in Britain far outweighs that of any other subgroup. A strong island affiliation is likely to lead to maintenance of Barbadian features rather than the acquisition of Jamaican Creole features as suggested in the BrACE literature (e.g., Sutcliffe 1982). First generation Barbadians tended to marry other Barbadians which is the case for the Barbadians in this study as well as the London Barbadians in Western (1992). With the second generation, intermarriage with the Anglo community became more common. Edward married a local Suffolk-born Anglo woman. Intermarriage is likely to lead to assimilation towards Anglo Ipswich features. It would be unsurprising then, if Edward shows more evidence of acquisition of Anglo Ipswich patterns than first generation Barbadians given his earlier age of arrival and his strong Anglo family networks. Table 2. Characteristics of speakers Ethnicity

Sex

Pseudonym

Age

Age Arrived in Ipswich – background

Anglo Anglo Anglo Anglo Barbadian Barbadian Barbadian Barbadian

F F M M F F M M

Betty Mary Keith Patrick Margaret Michelle Gary Edward

Retired Mid-aged Retired Mid-aged Retired Mid-aged Retired Mid-aged

born in Ipswich born in Ipswich born in Ipswich born in Ipswich Adult – rural St Michael 14 – rural St Joseph Adult – urban Bridgetown 9 – rural St John

Table 2 shows the characteristics of the 8 speakers in this study. Gary, Margaret and Edward are key members of the ‘Bajan’ community and the wider Afro-Caribbean community. The focus of the study will be on Edward, a 50 year old electrician, who came from rural St John to Ipswich at the age of 9, within the critical age for acquisition of a new dialect. His father was a cane cutter and had worked previously in America. His parents moved directly to Ipswich in the 1960s and he and his other siblings were cared for by their eldest sister in Barbados for several years before they all joined their parents in Ipswich. Like the other Barbadians in this study, he has an extensive Barbadian family network in Ipswich, including mother and father (now deceased), aunts and uncles, brothers and sisters, cousins, nieces and nephews, but unlike the other Barbadians in the sample, he also has extended family networks through his Anglo partner and his own children are mixed parentage. Edward has strong links to both Barbadian and Anglo communities in Ipswich. We can therefore expect his language to depart from those of his Barbadian counterparts whose family networks are within the Afro-Caribbean community.





Michelle C. Braña-Straw

.

Ethnicity, social position and language outcome

The site for this research differs from the UK patterns of settlement, in that Barbadians are almost as numerous as Jamaicans. Whilst an Afro-Caribbean community has emerged in Ipswich, island affiliations still persist into the second-generation. Barbadians and Jamaicans are the most prominent groups and differences between these groups are often emphasized. In the UK, Jamaican stereotypes involve resistance to White norms and the use of Jamaican Creole has become associated with a language of resistance to mainstream society. Barbadian stereotypes involve acquiescing to Whites. They might, therefore, be more predisposed to adopt local patterns of speech. Barbadians in Ipswich have a history of working with mainstream institutions, particularly in local government and show similar antagonism to Jamaicans as has been documented in the Caribbean and London. There is likely to be a degree of ambivalence to adopting mainstream society’s speech patterns because Barbadians still want to retain a Barbadian island identity. Of the Barbadians in the sample, Edward has the most opportunity and perhaps, motivation to adopt Anglo Ipswich patterns of speech. He arrived as a child and arguably, is not subject to the same level of maturational constraints as the adults. He has close family networks in the Anglo community and therefore has access to their speech patterns and motivation to acquire them, yet he has a strong sense of Barbadian identity and strong networks in the Barbadian community. Communities with strong social networks have been found to exert pressure on individuals to maintain community linguistic norms (Milroy 1987). Edward’s response to these pressures will determine the degree to which he orients to Anglo or Barbadian speech patterns for (t).

.

Methodology

The chosen sample of 8 speakers consists of two first generation Barbadians who arrived in Ipswich as adults, and two second generation Barbadians who arrived as children. The Anglos have been selected to match Barbadians in age, socio-economic indicators and because they lived in the same, or similar working class neighbourhoods. An overall, comparison of Barbadian and Anglo patterns allows us to examine (1) whether Barbadians match Anglo patterns, in which case we can infer that acquisition is taking place and (2) the degree to which they seem to be converging to, or diverging from Anglo patterns, by comparing the language of first generation Barbadians to that of second generation Barbadians. Table 3. Number of tokens analysed per speaker Word-final preconsonantal 10

Word-final prevocalic

Word-final prepausal

Word-medial intervocalic

Word-medial other

10

10

15

15

Chapter 1. Maintenance or assimilation?

Table 3 shows the number of tokens per environment, analysed for each speaker. The total number of tokens analysed per speaker is 60, making a total token count of 480 (compared with Docherty and Foulkes’ (1999: 55) 549 glottal tokens analysed for Newcastle). I used fine-grained acoustic analysis of conversational recordings to identify the acoustic features of the (t) variable for each speaker. The analysis is described in more detail in Straw and Patrick (2007). Word-finally, the dominant variants were glottal variants which I distinguish as: ● ●

glottal stops (GS) – involving a period of absence of vocal chord vibration; other glottalisation (GZ) – involving a range of phonetic features which sometimes contribute to the impression of a glottal stop, but where the vocal chords continue to vibrate.

Both of these glottal variants involve the complete replacement of [t]. I was unable to hear the difference between the two types of glottal variants with any reliability. Similarly, Docherty and Foulkes (1999) have reported for their Newcastle study, that even though they too, had difficulty distinguishing glottal variants, acoustic analysis revealed more than one type of glottal variant with differing sociolinguistic patterns. In the Ipswich data, there were other variants such as [t] and /t/ deletion but as they accounted for less than 10% of variants, they occurred too infrequently to analyse. For word-medial positions, four main variants were identified as important in analyzing speakers’ patterns.

●

Glottal stops (GS); Other glottalised variants (GZ); Glottally reinforced, [t‘] (glottalisation co-occurs with a released [t] burst); Non-reinforced, [t].

.

Results

● ● ●

Figure 1 below shows the overall glottal patterns in word-final position divided by environment; preconsonantal – PreC, prevocalic – PreV and prepausal – PreP. For the Anglos, the highest rate of glottals occurs in prevocalic environment. This effect is more prominent for males than females. For the other environments there is no clear order. For the Barbadians, Edward and the other male (Gary) seem to have this pattern, but the females do not. However, we see that rates for Edward, in both prevocalic (PreV) and prepausal (PreP), are nearly categorical at 90% and exceed those of any other speaker, Anglo or Barbadian. For preconsonantal (PreC), he has one of the lowest rates, along with the older Anglo Keith.



Michelle C. Braña-Straw Barbadians

Anglos

100

% GS+GS

80 PreC

60

PreP 40

PreV

20 Mary

Betty

Patrick

Keith

Michelle

Margaret

Edward

Gary

0

Figure 1. Glottals, all types – Word-final e.g., want, get

I then compared in more detail, the patterns in Edward’s speech to the other speakers by separating out glottal stops (GS) from other glottal variants (GZ). Figure 2 shows that Edward mostly pronounces glottal stops prepausally (PreP). The order of environments that he exhibits: i.e., high prepausal, lower rates for prevocalic (PreV) and preconsonantal (PreC) environments, although much more exaggerated by Edward, is also found amongst the female Barbadian speakers. This pattern is not consistently found in Anglo speakers, but we do find it for Patrick who belongs to the same age/gender group as Edward. Anglos

Barbadians

90 80 70 60 50 40 30 20 10 0

Mary

Betty

Patrick

Keith

Michelle

Margaret

Edward

PreC PreP PreV

Gary

% Glottal Stops (GS)



Figure 2. Glottal stops (GS) only – Word-final e.g., want, get

In Figure 3, we see that most of the Barbadians’ high prevocalic (PreV) rate is made up of GZ variants. Edward is typical of the Barbadians in this respect, but has the highest rates of any speaker at 80%. The order of environments: prevocalic followed by prepausal (PreP) and finally preconsonantal (PreC) is shared by the older Barbadian male, Gary, as well as the male Anglos.

Chapter 1. Maintenance or assimilation? Anglos

Mary

Betty

Patrick

Keith

Michelle

Margaret

Edward

PreC PreP PreV

Gary

% Glottalization (GZ)

Barbadians 90 80 70 60 50 40 30 20 10 0

Figure 3. Word-final – other glottalisation (GZ)

Next, I examined word-medial environments – Figures 4 and 5. Acoustic analysis again revealed the same four principal variants: Glottal stops (GS); Other glottalised variants (GZ); Glottally reinforced, [t‘]; Non-reinforced, [t].

●

Barbadians

Gary

Mary

Patrick

Betty

GS & GZ [t] & [‘]

Keith

% Variants

Anglos 100 90 80 70 60 50 40 30 20 10 0

Michelle

●

Edward

●

Margaret

●

Figure 4. Glottal variants compared with (t) variants – intervocalic butter, water

Figure 4 compares glottal variants and (t) variants. While other comparisons are possible (e.g., grouping reinforced [t‘] with other glottal variants), the comparison in Figure 4 showed the most striking difference between Barbadian and Anglo patterns. We see that Edward, like the other Barbadians, almost exclusively favours (t) variants intervocalically, unlike the Anglos who favour glottal variants. The black bars for Margaret and Edward show that they have acquired some glottal variants, however.



Michelle C. Braña-Straw

Though not distinguished here in the graph, this was made up exclusively of GZ glottalisation, which was also the most frequent glottal variant for Anglos. Anglos

Barbadians

60 50 % Variants

40

GS GZ

30 20 10 Michelle

Margaret

Edward

Gary

Mary

Betty

Patrick

0 Keith



Figure 5. Glottalised variants (GZ) – word-medial (other) country, little, button

The other striking finding is only found if we compare the different types of glottal variants word-medially, in Figure 5. The Anglos have both glottal stops (GS) and GZ glottalisation but do not seem to favour one variant over another. Edward and the other Barbadians have a negligible level of glottal stops (GS) – less than 10%. The Barbadians do not have GZ glottalisation either, except for Edward, who has it at rates that far exceed those of the Anglo speakers. Finally, when /t/ variants are compared, Anglos and Barbadians seem to have similar rates for both reinforced and non-reinforced [t].

.

Summary of findings

Developmental factors indicate that Edward should be the most successful in acquiring Anglo patterns, and this seems to be the case. He shows most similarity with the Anglos across a range of environments, also to be expected given that he has strong family networks in the Anglo community. Edward has not, however, completely adopted Anglo patterns. Rather he employs the linguistic strategies of overgeneralisation and simplification in salient environments. I suggest that what appears to be ‘incomplete’ learning is, in fact, socially motivated variation to show his affiliation with each group of speakers. Edward appears to identify GZ glottalisation with the Anglos by aligning himself with them in the following salient environments: 1. 2.

word-final prevocalic word-medial (other)

Although Edward’s patterns resemble those of the Anglos in these positions, the latter use both glottal variants (GS and GZ), whereas Edward uses GZ glottalisation almost

Chapter 1. Maintenance or assimilation?

all of the time in prevocalic, and exclusively, for word-medial (other). His behaviour exemplifies the linguistic process of simplification, i.e., reduction in the number of forms. In both cases, he far exceeds the rates of use for the Anglos, i.e., overgeneralisation. Additionally, a further breakdown of the word-medial environment reveals that Edward employs GZ glottalisation in the same positions as the Anglos, but he also extends his use to other positions where GZ glottalisation is normally blocked (e.g., particular, estate where the following syllable is stressed; Tollfree 1999: 172). This again is an example of overgeneralisation. On the other hand, he appears to identify the glottal stop as Barbadian and aligns himself with Barbadians in word-final prepausal position, which is also salient. Again, he exhibits the processes of simplification and over-generalisation by using glottal stops most of the time and more than any other speaker in this position. Barbadian use of glottalisation is often assumed by Anglos to be a Suffolk feature, but my initial impression on first hearing Barbadians in Ipswich, was that their patterns of glottal variation sounded more distinctive than the Suffolk patterns. It was therefore a surprise that Barbadians did not glottal in intervocalic position, the most salient position for glottal variation. Barbadians’ use of glottal stops prepausally may contribute to its distinctiveness. Most strikingly, the only other occasion that Edward aligns himself with the Barbadians, is in intervocalic environment. Recall that this environment is marked for Anglos as stereotypical of vernacular speech (Trudgill 1986: 20; Wells 1982: 322–8). Here, Edward has not acquired the glottal patterns of the Anglos but employs the more standard (t) variants of the Barbadians. Interestingly, he does not display any evidence of overgeneralisation or simplification in this position.

.

Discussion

Edward’s linguistic patterns reflect his social position in the Ipswich community today. The dialect acquisition literature has focused mainly on studies of children and adolescents, showing that peer group orientation and social networks are crucial to acquisition. But there is evidence to suggest that acquisition need not be a one-way, one-time process and, that changes in social networks over a person’s lifetime influence changes in their speech. Auer et al. (1997: 4–5) in their study of the acquisition of standard German by speakers of the Upper Saxonian vernacular, found that one exceptional speaker did not have a linear pattern of change. He made good progress in the first year towards the new dialect, but in the second year reversed the direction of change and in some ways surpassed his old ways of speaking. In Britain, codeswitching between British English and Jamaican Patois has been observed to occur relatively late, in adolescence onwards, by British-born Afro-Caribbeans along with their Asian and White peers (Edwards 1986; Hewitt 1986; Rampton 1995; Sebba 1993; Sutcliffe 1982).





Michelle C. Braña-Straw

After initial exposure to a new variety, ongoing changes during various stages of life and social networks will give the individual access to different models of language behaviour. I suggest that Edward has access to several models in Ipswich including input from Barbadian and Jamaican Creoles. The literature claims that Jamaican Creole is the main way in which Afro-Caribbeans in Britain can index their ethnic identity, and that use of Jamaican Creole features has become symbolic of a resistance to mainstream culture and institutions (Edwards 1986: 122, 132). However, this is incompatible with the ‘Ipswich Barbadian’ (and Edward’s) cooperative approach to mainstream institutions. However, even within the Barbadian community there is a varied response to Jamaican Creole. Edward reports that, while he does not speak Jamaican patois some of his brothers do, as shown in the following excerpt: Edward: File Location Bobby13_23 [Interviewer] Do you speak it? [Edward] patois? No, I can understand bits of it, but um, I don’t, no [Interviewer] yeah [Edward] I’ve got a few brothers who can {laughter} [Interviewer] {laughter} [Edward] you know, so um, cause I think, at the time, well now it seems to be the norm I suppose, um um I’ve known some white guys who can, who er can quite well understand it, and do it as well so may- you know, at the time, when I was growing up, that was something you didn’t, you know, you heard it and sort of…

Crucially, his willingness to be involved within mainstream institutions and organisations targeting Afro-Caribbeans specifically, or ethnic minorities generally, follows the tradition of his parent’s generation in which Barbadians were major innovators in bridging relations between the two communities. Edward is involved in extensive voluntary work helping ethnic minorities: he was formerly Chairman of the Caribbean Society and has been extensively involved in community education and race relations. His work with victim support (a charity working with the police supporting victims of crime) specializes in racially motivated crimes. He is also involved in his local residents’ association which lobbies local authorities on issues relevant to the community, such as housing, planning, schools, policing, etc. This is a residential area not traditionally

Chapter 1. Maintenance or assimilation?

settled by Afro-Caribbeans, but where they now reside in greater numbers, and an area considered to contain pockets of racism. Edward is an outgoing, confident individual who enjoyed school. Most of his classmates would have been white. However, hostility in the early days of immigration would have isolated him from many of the networks in which his classmates operated. This may have been in subtle ways, as we have seen, for instance by assuming he could understand and translate Jamaican patois. Or it may have been in more overt ways – racial abuse from other children, and police harassment, were commonplace experiences for Afro-Caribbeans, and Edward was no exception. The linguistic and social models from the Barbadian community, along with the early school experiences already discussed, would have been important in Edward’s separation of the Barbadian and Anglo patterns in his English. This separation would have been further consolidated by extensive networks in both communities in adulthood, reflecting his current position as an intermediary. His role as intermediary is facilitated by having family ties in both communities: Edward’s current partner and former spouse are both Suffolk Anglos and he has children by his ex-wife, with whom he maintains good contacts. They would have provided him with access to Anglo speech patterns, albeit in his adulthood. He also has a large nuclear and extended Barbadian family living in Ipswich which would have provided him with continued access to a Barbadian linguistic model. File location: Bobby12: 262 [interviewer] do do people generally think you have an Ipswich accent? or… [Edward] er yes they they well they do ununtil I speak to other West Indians and they say “you’re Bajan aint you”, {laughter} you know so, and my brother was saying that, saying the other day, saying well, he’d thought he’d lost his, but um you jus- you just don’t realize it’s there, until someone else remark on it. I mean I had a colleague at work, um he he’s a white guy, and he was saying to me um, “you see {Edward},” he said, “you know when you speak to me, you speak to me and you speak to other black guys, you your ((twang does change)).” I said, “well, I didn’t realize that, but I suppose,





Michelle C. Braña-Straw

I I do cause, I I’ve known ((Italians)), Yugoslavs and whatever, and when they speak to me, well when they speak to another ((Italian)), you notice the difference,” I said, “well that’s obviously, um what some of you do, jus- um click into it, you don’t realize you um you’re doing it, you know so,” I suppose it’s understandable, most well anyone who speaks another language, I suppose really you tend t- tend to drift into one and out without realizing you’re doing it

In this excerpt, Edward discusses his linguistic behaviour in terms of code-switching, likening it to the behaviour of second language learners, (Yugoslavs and Italians). Although this research cannot test for actual code-switching behaviour, Edward’s changing linguistic behaviour has been observed by his White colleagues at work. This research does show for the variable (t), that Edward employs Barbadian and Anglo patterns of variation, simultaneously within his English, irrespective of whether he code-switches or not.

.

Conclusion

Faced with the choices of assimilation or maintenance, Edward, a second generation immigrant, reports that he attempts to do both, responding to his audience by switching between a Barbadian accent and an Ipswich accent in much the same way as bilinguals do. His assertion is beyond the scope of this study. However, I have been able to test whether his English is ‘indistinguishable’ from local Anglos as has been claimed by the BrACE literature (see Section 1.1) and found that for the (t) variable, Edward exhibits elements of a mixed system, employing the processes of simplification and overgeneralization to accentuate Barbadian patterns in some environments (e.g., word-final prevocalic) and accentuating Anglo patterns in other environments, i.e., prepausally and in the stigmatized environment, intervocalic. To describe his linguistic behaviour in terms of maintenance and assimilation is to ignore Edward’s creative response to his social position in Ipswich. He has negotiated his position within the two communities; by accentuating Barbadians norms, he retains an ethnic identity based on an island Barbadian affiliation; by accentuating Ipswich norms he achieves a local identity (an Ipswich accent) associated with White mainstream society, without having to completely adopt White norms.

Chapter 1. Maintenance or assimilation?

His language behaviour results from having strong ongoing social networks in both Barbadian and Anglo communities, thereby allowing him to fulfil his role in Ipswich society as an ‘intermediary’ following the tradition of first generation Barbadians. Edward’s language behaviour may contribute to a new model for subsequent generations of Afro-Caribbeans, an alternative to the ‘language of resistance’ provided by Jamaican Creole (Edwards 1986: 122, 132) – which nevertheless can potentially index ethnic identity, but allows for integration into mainstream Anglo society.

References Auer, P., Barden, B. and Groshopf, B. 1997. Long-term linguistic accommodation: A single case analysis of inner-German West-East migration and its wider implications. Paper presented at the workshop Migration as a factor in connection with convergence and divergence processes of dialects in Europe, Heidelberg, October. (European Science Foundation Network on The convergence and divergence of dialects in a changing Europe.) Blake, R. 1994. On the constraints placed by phonological models on the description of language structures: The case of the glottal stop in Barbadian Creole English. Paper presented to the Society for Caribbean Linguistics. August 26, University of Guyana, Georgetown, Guyana. Blake, R. 1997. All o’ we is one? Race, class and language in a Barbados community. PhD thesis, Stanford University. Burrowes, A. (in collaboration with R. Allsopp). 1983. Barbadian Creole: A note on its social history and structure. In Carrington, L., Craig, D. and Dandaré, R.T. (eds). Studies in Caribbean Language. St. Augustine, Trinidad: Society for Caribbean Linguistics. 38–45. Census 1991. County Report Suffolk. (Pt1) Table 7. OPCS. Census 2001. Population and migration: Ethnic group 2001. Online report, Table KS06P. http:// neighbourhood.statistics.gov.uk. Accessed 6 March 2004. Chambers, J. K. 1992. Dialect acquisition. Language 68.4. 673–705. Docherty, G.J. and Foulkes, P. 1999. Derby and Newcastle: instrumental phonetics and variationist studies. In Foulkes, P. and Docherty, G. (eds). Urban voices: Accent studies in the British Isles. London: Arnold. 46–71. Edwards, V. 1986. Language in a Black community. Clevedon: Multilingual Matters. Foulkes, P. and Docherty, G.J. 1999. Urban voices – Overview. In Foulkes, P. and Docherty, G. (eds). Urban voices: Accent studies in the British Isles. London: Arnold. 1–24. Hewitt, R. 1986. White talk, black talk: inter-racial friendship and communication amongst adolescents. Cambridge: Cambridge University Press. Horvath, B. 1991. Finding a place in Sydney: migrants and language. In Romaine, S. (ed.). Language in Australia. Cambridge: Cambridge University Press. Kerswill, P. 1996. Children, adolescents, and language change. Language Variation and Change 8. 177–202. Kerswill, P.E. and Williams, A. 2000. Creating a new town koine: children and language change in Milton Keynes. Language in Society 29. 65–115. Knight, P., Patrick, P.L. and Straw, M. 2002. Difficult acts: Aspects of linguistic identity in urban Afro-Caribbean English communities. Paper presented at the Sociolinguistic Symposium 14. April 5, Ghent.





Michelle C. Braña-Straw Le Page, R.B. and Tabouret-Keller, A. 1985. Acts of identity: Creole-based approaches to language and ethnicity. Cambridge: Cambridge University Press. Milroy, L. 1987. Language and social networks, 2nd ed. Oxford: Blackwell. Modood, T., Berthoud, R., Lakey, J., Nazroo, J., Smith, P., Virdee, S. and Beishon, S. 1997. Ethnic minorities in Britain: Diversity and disadvantage. London: Policy Studies Institute. Niles, N.A. 1980. Provincial English dialects and Barbadian English. PhD thesis, University of Michigan. Payne, A. 1980. Factors controlling the acquisition of the Philadelphia dialect by out-of-state children. In Labov, W. (ed.). Locating language in time and space. New York: Academic Press. 143–78. Peach, C. 1996. Ethnicity in the 1991 Census: Volume 2. London: HMSO. Rampton, B. 1995. Crossing: Language and ethnicity among adolescents. London: Longman. Roberts, P.A. 1988. West Indians and their language. Cambridge: Cambridge University Press. Rosen, H. and Burgess, T. 1980. Languages and dialects of London school children. London: Ward Lock Educational. Sebba, M. 1993. London Jamaican: Language systems in interaction. London: Longman. Straw, M. 2006. Dialect acquisition and ethnic boundary maintenance: Barbadians in Ipswich. PhD thesis, University of Essex. Straw, M. and Patrick, P.L. 2007. Dialect acquisition of glottal variation in /t/: Barbadians in Ipswich. Language Sciences 29. 385–407. Sutcliffe, D. 1982. British Black English. Oxford: Blackwell. Tollfree, L. 1999. South-East London English: Discrete versus continual modelling of consonantal reduction. In Foulkes, P. and Docherty, G. (eds). Urban voices: Accent studies in the British Isles. London: Arnold. 163–84. Trudgill, P. 1986. Dialects in contact. Oxford: Blackwell. Van Herk, G. 2003. Barbadian lects. Beyond meso. In Aceto, M. and Williams, J. P. (eds). Contact Englishes of the Eastern Caribbean. Amsterdam: John Benjamins. 241–64. Wells, J.C. 1973. Jamaican pronunciation in London. Oxford: Blackwell. Wells, J.C. 1982. Accents of English. (3 volumes.) Cambridge: Cambridge University Press. Western, J. 1992. A passage to England: Barbadian Londoners speak of home. Minneapolis: University of Minnesota Press. Wight, J. and Norris, R.A. 1970. Teaching English to West Indian children: The research stage of the project. Schools Council Working Paper 29. London: Evans/Methuen Educational. Williams, A. and Kerswill, P. 1999. Dialect levelling: Change and continuity in Milton Keynes, Reading and Hull. In Foulkes, P. and Docherty, G. (eds). Urban voices: Accent studies in the British Isles. London: Arnold. 141–62.

 

Universal and substrate influence on the phonotactics and syllable structure of Krio Malcolm Awadajin Finney California State University Long Beach

This paper evaluates the extent to which the phonotactics and syllable structure of Krio demonstrate universally unmarked properties (based on presence or frequency of use across languages) evident in first language (L1) acquisition or are influenced by properties of West African substrate languages. Some researchers maintain that basic, universal, and unmarked principles, in the form of phonological processes evident in language acquisition (e.g., substitution, deletion, consonant cluster simplification), constrain the type, number, and sequence of segments that are allowed in onsets and codas of syllables in creoles. Others affirm moderate to strong influence of the phonological properties of substrate languages on creoles. Yet a third group suggests an interaction of universal and substrate influence in the development of the phonological properties of creoles. I acknowledge presence of some phonologically unmarked properties, which are also observed in L1 acquisition (including vowel and consonant inventories, phonological processes such as ‘stopping’ and ‘fronting’, and the structure of syllables), though such properties are not necessarily maximally unmarked and are not necessarily applied in Krio in scope and regularity as they are in L1 acquisition. I further present data to support the influence of the relatively marked phonological properties of substrate languages (including vowel nasalization and palatalization). Krio syllables, I conclude, exhibit neither maximally marked nor maximally unmarked phonological properties. Krio properties are somewhere in between.

.

Introduction

A well-discussed but still contentious issue in creolization is the extent to which the linguistic properties of creoles provide evidence in support of one or more of the different, and often polar, views of creole genesis. The Universalist account of creolization (Bickerton 1975, 1977, 1988, 1999) argues for the development of a prototypical creole grammar primarily through the adoption of universally unmarked properties of grammar by creole children. This position holds that creoles generally display properties that are typical of a child’s linguistic output during early stages of first



Malcolm Awadajin Finney

language acquisition. Though input is received from a variety of linguistic sources, creole children’s innate capacity for language supposedly predisposes them to adopt simple universally unmarked linguistic properties. Bickerton proposes that creole children, unlike child speakers of other languages, do not have a consistent adult model against which to check their hypotheses and are therefore guided by Universal Grammar (or a Language Acquisition Device) to adopt these universally unmarked properties. The universalist proposal primarily draws support from the morphosyntactic properties of creoles though their arguments could be extended into phonology in attempting to account for the segmental and phonotactic properties of lexical items in creoles, some of which may have undergone phonological processes that have resulted in the modification of both the quality and quantity of segments in the syllables of these lexical items. The Substrate account of creolization (Alleyne 1980; Holm 1988), also drawing support from both morphosyntactic and phonological properties, maintains that West African substrate languages have predominantly influenced linguistic properties of creoles. They maintain that adult native speakers of substrate languages, because of minimal competence in superstrate European languages, were compelled to modify lexical items originating from lexifier languages by substituting segments and structures that are not present or allowed in substrate languages with those that are allowed. What superficially may seem like universally unmarked phonological properties of creoles could in reality be the result of the transfer of substrate properties into the creoles. A more moderate and compromise version of this position maintains that borrowings of both superstrate and substrate origins are present in creoles, but these borrowings are generally constrained by universally unmarked properties of language (Field 2004; Holm 1988; Mufwene 1986, 2001; Mühlhäuser 1980, 1986; Siegel 1999; Uffman 2003). Examples and arguments presented in this paper argue for the phonotactics and syllable structure reflecting both general unmarked phonological properties and more specific phonological properties likely borrowed from West African substrate languages. Such properties are neither maximally marked nor maximally unmarked.

.

The history of Krio

Circumstances leading to the emergence of Krio are highly debatable. One view argues that Krio emerged from Creoles of the Americas and in effect shares some linguistic similarities with other Atlantic Creoles. Huber (1999, 2000) proposes that Krio emerged from varieties of creoles used by settlers – mostly freed slaves primarily from four areas – who were resettled in the Sierra Leone peninsula, including Freetown, between 1787 and 1850 (Huber 1999: 59–65; Huber 2000: 276–7). The Original Settlers (The Black Poor) from England, numbering about 328 settlers, arrived from England in 1787. However, there are no records of contributions of the Black Poor to the development of present-day Krio.

Chapter 2. The phonotactics and syllable structure of Krio

The Nova Scotians were originally slaves in America that were promised freedom and much better living conditions in return for support for the British during the American war. After the war, about 3,000 slaves were relocated, in 1783, in Nova Scotia, Canada (a British colony), where they continued to endure economic hardship and epidemics (Huber 1999, 2000). In 1787, in response to their protests, the British relocated them in a new colony – the Sierra Leone Peninsula – in West Africa (where most slaves in the New World were originally from prior to enslavement). In 1792, close to 12,000 freed slaves were shipped to Freetown from Nova Scotia (Huber 1999: 61–2). About 556 Jamaican Maroons, deported to Nova Scotia in 1796 after an unsuccessful revolt, were also transferred to the Sierra Leone peninsula in 1800. Creoles from the West Indies, particularly the variety brought by the Jamaican Maroon settlers, are proposed by Huber (1999) to have had significant input into what has now evolved into present-day Krio. Huber outlines a number of similarities between present-day Jamaican Creole and Krio in a number of linguistic categories including copula space, verb phrase, noun phrase, and lexical and functional items (Huber 1999: 108–14). According to Huber (1999, 2000), after the British declared slavery illegal for their subjects, their fleet patrolled the West African coast, intercepted slave ships and recaptured slaves, and released and resettled them in the Sierra Leone peninsula. They were generally referred to as Liberated Africans (or Recaptives). These were by far the largest group and were resettled in the Sierra Leone peninsula over a period that stretched from 1808 (when Sierra Leone was declared a crown colony) to 1863. Huber (1999: 63) estimates the number of Liberated Africans resettled in the Sierra Leone peninsula during this period at about 60,000, though only about 37,000 were alive in 1840. In 1860, the Liberated Africans and their descendants totalled 38,375. The influence of the Liberated Africans on the linguistic evolution of Krio is a debated issue. They comprised mainly of speakers of West African Kwa languages, including Yoruba (the most prominent), Igbo, Akan, and Gbe. Huber acknowledges their possible influence when he states that finally another major group to be considered in the development of Krio is that of the Liberated Africans, who by about 1812 outnumbered the Nova Scotians and Maroons. Judging from their enormous numerical increase over the following decades, the Recaptives could very well have dominated the linguistic scene in 19th century Sierra Leone by swamping any other variety that may have developed in the years prior to their arrival (Huber 2000: 278).

According to Huber (1999, 2000), they originally spoke African languages only, but with improved economic status and more interaction with the Maroons and Nova Scotians, a new variety of creole emerged that was described in transcripts written by British colonists as a “barbarous”, “defective”, “gibberish” and “jargon” form of English (Huber 2000: 282, 285). This variety incorporated properties of the creoles used by the Maroons and Nova Scotians, and it is reasonable to assume that the cross-linguistic





Malcolm Awadajin Finney

influence was bi-directional: that is, the creoles used in Freetown were in turn influenced by the variety developed by the Liberated Africans. The language and traditions of settlers of Yoruba origins have had a strong influence on the language, social life and customs of Krio speakers in Freetown. The influence of the languages (including Yoruba) of the Liberated Africans on the grammatical development of Krio should therefore not be underestimated and should be considered at least a contributory factor in the development of the linguistic properties of Krio (see Bradshaw 1966 for an exhaustive list of Krio lexical items borrowed or derived from Yoruba). A contrary view of the origin of Krio is held by Hancock (1986, 1987), who maintains that the original “core” creole emerged along the Upper Guinea Coast of West Africa in the 1600s, long before the trans-Atlantic slave trade. There is evidence of British settlement in that area and there are written reports of interaction, including intermarriages, between Europeans and Africans during this period. Products of the intermarriages – referred to as Mulattos – became the first creole speakers. Creoles in the Americas partly originated from this original creole (Guinea Coast Creole English (GCCE)), which was transmitted to the Americas by slaves transported by English and Dutch traders. Hancock suggests that the grammar of GCCE continued to be influenced by the properties of West African languages as a result of its extensive use by second language speakers in the region. According to Hancock, present-day Krio is an offshoot of GCCE. Eyewitness recorded transcripts of GCCE in the 17th and 18th centuries illustrate similar grammatical properties and lexical items between modern Krio and GCCE. The presence of these properties and items in present day Krio, Hancock maintains, is evidence that the emergence of Krio pre-dates the resettlement of freed slaves in Sierra Leone.

.

Krio data sources

As a native Krio speaker, who was born and raised until adulthood in Freetown, I provided the Krio examples used in this paper based on my intuitions and those of other native Krio speakers. Other examples were obtained from Bradshaw (1966), Fyle and Jones (1980), and Jones (1971). These examples, in my opinion, are reflective of properties of the Basilect and Mesolect varieties used by native Krio speakers residing in Freetown. A limitation of this paper, as is the case for research in creole studies, is that it is based on current phonological properties of Krio, which may not be an accurate representation of the properties that the language exhibited during its formative years. Some recently borrowed lexical items are integrated into Krio phonotactics while others remain phonologically unassimilated. Most of the examples presented in this paper are not proposed as absolutes. Creoles generally exhibit different lects (varieties) – Basilect, Mesolect, and Acrolect. Choice of

Chapter 2. The phonotactics and syllable structure of Krio

lect is generally conditioned by the social, economic, and educational status of the speaker. These lects vary in the extent to which phonological properties of borrowings from superstrate languages diverge from the properties of the superstrate languages, from significantly for the Basilect variety to minimally for the Acrolect variety. As a result, multiple pronunciation forms may co-exist in different lects for the same lexical item in Krio. Most of the examples provided are present in the Basilect and sometimes Mesolect varieties but not necessarily in the Acrolect variety. A point by Jones (1971) worthy of note is that the lack of consistency of the application of phonological processes is due to continued contact between Krio and English – its superstrate language. The continued prestige status of English continues to influence the Acrolect and Mesolect varieties of Krio.

.

Universal and substrate influence on the segmental features of syllables in Krio

The Universalist account of creolization emphasizes the contribution of children to the development of prototypical linguistic properties of creoles. Proponents of this view, particularly Bickerton (1975, 1977, 1988, 1999), contend that creoles generally display universally unmarked linguistic properties that are typical of a child’s linguistic output during early stages of first language acquisition. According to Bickerton’s Language Bioprogram Hypothesis, pidgins were converted into creoles almost exclusively by children whose primary language – the pidgin – could not be used adequately to satisfy their linguistic needs. As a result, the linguistic properties of creoles became more complex and elaborate primarily through the invocation of universally unmarked principles of language by emerging native creole speakers. Restructuring of the creole is triggered by the innate capacity for language by children, with continued input from the lexifier language and no significant input from the substrate languages of the parents. Uffman (2003: 3) argues for the process of creolization as “retention of the unmarked”. That is, the probability of a substrate feature adopted by a creole is dependent on the relative markedness of the feature: less marked properties are more likely to be adopted by a creole grammar than more marked ones. He proposes a formal account of substrate leveling that: “those substrate subgrammars are systematically retained that are relatively unmarked compared to competing subgrammars” (ibid.). The theory of Markedness has been used over the years to predict directionality of difficulty in the acquisition of linguistic properties in both first and second language acquisition, with unmarked properties predicted to be preferred and acquired earlier by language learners (Broselow et al. 1998; Comrie 1989; Croft 1990; Eckman 1977, 1981a, b; Rutherford 1984). The implication, on the basis of this approach, is that the phonological properties of creoles should include a predominance of unmarked phonological segments including vowel and consonant systems that include segments that are universal





Malcolm Awadajin Finney

or very common cross-linguistically. There should be a general absence of the less common segments.

.

Effects of universals on the vowel system of Krio

An overwhelming number of creoles, including Krio, exhibit a 7-pure-cardinal-vowel system – /i, e, ε, a, , o, u/ (Alleyne 1980; Hall 1966; Holm 1988; Romaine 1988). Some creoles exhibit a 5-vowel system with /ε/ and // absent or are realized as allophones of /e/ and /o/ respectively (Holm 1988). These vowels are proposed to be common across languages of the world. Less common vowels, including the vowels /i/, // and // are proposed to be generally absent in creoles. Hall (1966) further argues that the comparatively less common front and back rounded vowels – /y/ and /œ/ – lose their frontness or roundness in French-based creoles (including Haiti) in which /y/ is replaced by the more common /i/ or /u/ and /œ/ is replaced by the more common /e/ or /o/ in lexical items of French origin. Universalists generally interpret this as an indication of the adoption of more universal vowels (present in the repertoires of a majority of the world’s languages). They argue that the lexifier languages of creoles generally contain a lot more vowels in their repertoire but creoles have adopted the ones that are common across languages, and that there is a general absence of vowels that are less typologically common. Another evidence of the influence of universals on the development of vowels in creoles is proposed to be the simplification process of vowel substitution – a monophthongization of English diphthongs (Alleyne 1980; Holm 1988). Holm identifies São Tomé Portuguese Creole and Surinamese creoles as examples of creoles in which this process is evident. This process is also applied to a number of lexical items of English origin in Krio: [aj] → [ε] or [e] (1) a. /nεt/ c. /rεs/ e. /bεt/ g. /wεf/

‘night’ ‘rice’ ‘bite’ ‘wife’

[aw] → [o] or [] (2) a. /grn/ ‘ground’ c. /ms/ ‘how much’ e. /os/ ‘house’ [j] → [wε] (3) a. /bwεl/ ‘boil’

.

b. d. f. h.

/fεt/ /blεn/ /lεk/ /tεm/

‘fight’ ‘blind’ ‘like’ ‘time’

b. d.

/fdm/ ‘fall down’ /mt/ ‘mouth’

b.

/pwεl/ ‘spoil’

Effects of universals on the consonant system of Krio

For the consonant systems of creoles, the expectation, based on the Universalist account, is for the consonant systems of creoles to comprise of consonants such as stops, nasals, and glides, and to lack less common consonants such as some fricatives and affricates (/v, θ, ð, ∫, , t∫, d/), which should then be substituted with the universal

Chapter 2. The phonotactics and syllable structure of Krio

or more cross-linguistically common sounds, as is the case in L1 acquisition (Clarke 2003; Hulit and Howard 2002; Levelt and van de Vijver 2004). The phonemes /θ, ð/ are practically absent in Krio and are generally substituted with /t, d/ in lexical items of English origin. This substitution process, referred to as “stopping”, is proposed to be one of the universal or common processes (a fricative (or affricate) replaced by a stop) and is by far the most regular, predictable, and common process in Krio. It applies almost without exception in every position (initially, medially, and finally) to lexical items of English origin. For example: [θ/ð] → [t/d] (4) a. /tit/ ‘teeth’ c. /wit/ ‘with’ e. /fada/ ‘Catholic priest’

b. d. f.

/tεki/ ‘thanks’ /dεm/ ‘them’ /brda/ ‘brother’

Another “stopping” substitution process observed in L1 acquisition and evident, though not as common as the process in (4), in Krio is the substitution of /v/ with /b/. This process is evident in some lexical items of English origin in word-medial and word-final (5) but not word-initial (6) positions. For example: (5)

a. c. e.

/drεb/ ‘drive away’ /dεbul/ ‘devil’ /oba/ ‘over’

b. d. f.

/ebi/ /∫ub/ /kba/

(6)

a.

/vεks/

b.

/vmit/ ‘vomit’

‘angry, vex’

‘heavy’ ‘shove’ ‘cover’

Another common consonant substitution process in L1 acquisition is “fronting”, whereby back consonants (with places of articulation further back in the oral cavity) are substituted with front consonants (with places of articulation closer to the front of the oral cavity). This process involves a wide range of consonants, including /∫, , t∫, d/, in L1 acquisition. These consonants are present in Krio though the fricative /∫/ and affricate /t∫/ are sometimes substituted with /s/, as is the case in L1 acquisition (Clarke 2003; Hulit and Howard 2002). This process generally occurs in only word-medial and word-final positions (7) but not in word-initial position (8), with the possible exception of [sus] (shoe(s)). For example: [∫] or [t∫] → [s]: (7) a. /asis/ ‘ash(es)’ c. /was/ ‘wash’ e. /ms/ ‘how much’

b. /mas/ ‘tread on/mash’ d. /kalbas/ ‘calabash/gourd’ f. /tums/ ‘too much’

As mentioned above, this process does not apply word-initially. Further, there are lexical items of English origin in which the [∫] is maintained in word-final position. For example: (8)

a. c. e.

/∫ap/ ‘store; sharp’ /∫t/ ‘shirt; short’ /pu∫/ ‘push’

b. d. f.

/∫ub/ ‘shove’ /∫imi/ ‘chemise’ /fi∫/ ‘fish’





Malcolm Awadajin Finney

Thus, only a few phonological processes observed in L1 acquisition are also evident in lexical items of English origin in Krio. There are other processes that are not observed. Additionally, the processes that are observed vary in degrees of regularity, predictability, and frequency. With the exception of the “stopping” process of [θ/ð] → [t/d], the others generally have exceptions and are not usually applicable in word-initial position. Additionally, the affricates /t∫, d/, generally absent in the repertoire of children in early L1 development (Clarke 2003; Hulit and Howard 2002) and considered marked cross-linguistically (Haspelmath et al. 2005), do not generally undergo any substitution process in Krio. Affricates in lexical items of both English and non-English origins for the most part remain unchanged. For example: (9)

a. c. e.

/dam/ ‘bump into’ /t∫int∫/ ‘bedbug’ /ket∫/ ‘catch’

b. d. f.

/bint∫/ ‘beans’ /domp/ ‘jump’ /t∫ak/ ‘drunk’

Thus, Krio (and possibly other creoles) may exhibit some phonological properties of L1 acquisition generally considered unmarked; however, their occurrences are not as frequent or regular as is the case in L1 acquisition. Additionally, some phonological features that are considered marked and further generally absent in L1 acquisition are present in Krio. That is, Krio (and possibly other creoles) may exhibit some universally unmarked properties but its phonological repertoire cannot be considered maximally unmarked.

.

Influence of West African substrate languages on the segmental properties of Krio

One of the criticisms against the universalist account of creolization identified by Baker (2000: 42) is that “more rigorous comparisons with relevant non-European languages have produced more compelling evidence of their influence on particular creoles”, a position that is adopted by proponents of the substrate account of creolization. Much of the evidence linking the development and evolution of the linguistic properties of creoles with those of West African substrate languages (most of them belonging to the Kwa language sub-group) has been in morpho-syntax. Proponents (including Corne 1987; DeGraff 2001; Lefebvre 1993, 1996; Lumsden 1999) maintain that the apparent universal properties of creoles are superficial and that the underlying creole grammar (Atlantic varieties) exhibits structural properties resembling those of West African substrate languages. They argue that a major aspect of creolization is that it is a process of second language acquisition. Adult native speakers of substrate languages, because of minimal grammatical competence in superstrate European languages, were compelled to borrow grammatical structures from their primary (substrate) languages into which they superimposed lexical items derived from superstrate languages, in their attempts to communicate with speakers of superstrate languages.

Chapter 2. The phonotactics and syllable structure of Krio

Other substrate proponents (Alleyne 1980; Holm 1988; Spencer 1971) have attempted to extend this proposal to account for the development of some phonological properties of Atlantic creoles. A comparison of the consonant phonemic inventory of Krio (as well as other creoles) and substrate languages reveals obvious direct influence of substrate languages, most notably the presence of the co-articulated stops /kp/ and /gb/ – features of Kwa and Mande languages – as in the following Krio examples: (10)

a. b. c. d. e. f. g. h. i. j.

/kpata-kpata/ /agbo/ /gbagbati/ /akpl/ /agbada/ /agbado/ /igbak/ /kpεkpεkpε/ /lakpalakpa/ /∫akpa/

‘completely-finished’ ‘a medicinal herb’ ‘a show of force’ ‘frog’ ‘embroidered gown’ ‘a species of snake’ ‘large wooden spoon’ ‘trifles, odds and ends’ ‘ringworm’ ‘type of sauce’

Another possible evidence of substrate influence on the phonological properties of Krio is regressive vowel nasalization – a process whereby a vowel followed by a nasal consonant /n, m/ is pronounced as a nasal vowel and the nasal consonant is deleted. Spencer (1971) identifies this process as a common occurrence in unstressed syllables (such as the progressive or gerundive English suffix -ing) in West African English. It is also a common process in other creoles, including Saramaccan, Sranan, Ndjuka, and Jamaican (Alleyne 1980) and has been attributed to the influence of substrate languages. It is not applied in superstrate languages (including English) but is evident in lexical items of both African and English origins in Krio and in some substrate languages, including Yoruba, Isoko, Ewe, Fante, Ibo, Kongo, and Ijo (Dunstan 1969; Holm 1988). Alleyne (1980) further mentions that for lexical items of English origin ending in /n/ or /ŋ/, the preceding vowel may be nasalized and the nasal consonant dropped. Fyle and Jones (1980: xix) provides examples of the occurrence of this process in English and African borrowings in Krio (11)

a. b. c. d. e. f. g.

/bõ∫eka/ (boneshaker) /lõsm/ /dõs/ (Jones) /kõko/ /rõ∫o/ /abidũ/ (Abiodun) /dãs/

‘a rickety bus’ ‘lonesome’ ‘a family name’ ‘small hut’ ‘a kind of devil or spirit’ ‘personal name’ ‘dance’

Both Spencer (1971) and Fyle and Jones (1980) claim that oral and nasal vowels are in phonemic contrast in Krio (but not in English, Krio’s lexifier language) as well





Malcolm Awadajin Finney

as in West African languages, from which this feature is proposed to be borrowed. For example, the de-nasalization of the vowel in (11d) results in a completely different lexical item – /koko/ (a bump or swelling on the forehead). Additionally, the process of palatalization, observed in some creoles, is proposed to be evidence of the phonological influence of West African substrate languages, in which such a process is very productive (Alleyne 1980; Holm 1988). This process is applied in Twi, Ewe, and Yoruba, in which [s, z, t] are palatalized to their allophonic variants [∫, , t∫] before front high vowels. Holm (1988) claims that a similar process occurs in creoles such São Tomé Portuguese Creole, Papiamentu, and Negerhollands Dutch Creole, as well as in some Caribbean varieties of French Creole. Palatalization is not a very productive process in Krio though it is evident in a few lexical items of English origin in which [t] or [s], or a combination, is palatalized and realized as [t∫] after the nasal [n]. Consider, for example: (12)

a. c.

/fεnt∫/ /bint∫/

‘fence’ ‘beans’

b. d.

/rint∫/ ‘rinse’ /ant∫/ ‘ants’

Another phonological process evident in Krio and other creoles that is attributed to the influence of West African substrate languages is vowel harmony – a process whereby vowels in syllables in a sequence are made identical. In some cases, the whole syllable is duplicated. Alleyne (1980) identifies this as a common process among Niger-Congo languages including Twi and Yoruba (West African substrate languages).This is also evident in a number of creoles including Jamaican, Ndjuka, Sranan, and Saramaccan (Alleyne 1980). It is also very common in Krio in a number of lexical items of both English and West African origins. For example: (13)

a. c. e. g.

/bεlε/ /arata/ /ptpt/ /krkr/

‘belly’ ‘rat, mouse’ ‘mud’ ‘skin disease’

b. d. f. h.

/broko/ /tranga/ /sansan/ /pεtεtε/

‘broken’ ‘strong’ ‘sand’ ‘potato’

Advocates of the phonological influence of West African substrate languages on Atlantic creoles (including Krio) have further argued that the apparent universal nature of the inventory of segments and phonological processes observed in Atlantic creoles in reality reflect the phonological properties of West African substrate languages, primarily as a result of direct transfer of these properties into the pidgins (used initially as a second language) that eventually evolved into creoles. In defense of this position, Alleyne (1980: 140) argues that: ‘Substratum’ influences always have to be substantiated a fortiori . . . rather than merely asserted, and their validity weighed against other explanatory factors. But they cannot be automatically invalidated by the existence of similar forms elsewhere which are subject to other historical causal factors.

Chapter 2. The phonotactics and syllable structure of Krio

Spencer (1971) notes that the varieties of English used across English-speaking West Africa exhibit a number of similar phonological properties that are also shared by a number of West African languages. These properties are generally used to identify English speakers of West African linguistic backgrounds. This is evident in the phonemic substitution of consonants and under-differentiation of vowels (evident in a reduced set of contrasting vowels). Spencer further proposes that some of these properties have made their way into Krio. Merging of vowels in West African varieties of English is argued to be the effects of negative transfer from West African languages, which generally lack some pure vowels present in the inventories of British and American English (Fyle and Jones 1980; Jones 1971; Sey 1973; Simo Bobda 2000; Spencer 1971). According to Spencer (1971), vowel merging in Krio is the resulting effect of the general absence of these vowels in the English used by West Africans. Examples of merging in Krio include the following: (14)

i/i → /i/ beat/bit; bead/bid; sheep/ship are homophones u/ → /u/ pool/pull; fool/full are homophones / → // cop/cup; rob/rub are homophones

Substrate advocates maintain that the merging of vowel phonemes in creoles is likely the result of the influence of West African languages. The typical 7 (or 5) pure vowel inventory exhibited by creoles in general, and Krio in particular, is thus argued to be the result of the presence of a similar, if not identical, inventory in substrate languages (Alleyne 1980; Fyle and Jones 1980; Holm 1988; Jones 1971; Spencer 1971). The seven-vowel system exhibited by creoles is identified by Holm (1988) as the predominating vowel pattern in the Kwa (e.g., Yoruba, Bini, Ewe) and Mande (e.g., Bambara, Susu) languages of West Africa. Other languages belonging to these language subgroups are proposed to exhibit the five-vowel system with /e/ and /ε/ as well as /o/ and // existing in an allophonic relationship, as is the case in some creoles (Holm 1988). Substrate advocates have further argued that some of the proposed unmarked and universal phonological processes evident in Atlantic creoles (including Krio) may in reality be the result of direct transfer of these processes from the West African substrate languages into the creoles. For example, the general absence of the fricatives [θ/ð] and the use of [t/d] as substitutes in Atlantic creoles (including Krio) have been attributed to the influence of West African languages, in which these fricatives are absent. In lexical items of English origin in these languages, the stops [t/d] are generally used as corresponding substitutes (Dunstan 1969; Simo Bobda 2000) as is the case in Krio and other Atlantic creoles. Alleyne also identifies [v → b] as a phonological process that is common in a number of English loanwords in Yoruba and Twi. Dunstan (1969) and Alleyne (1980) further identify some Nigerian languages (including Yoruba) in which /s/ is substituted for /∫/ or both forms are used in alternation as allophones. A closer look at the examples of “fronting” in Krio (reproduced as (15))





Malcolm Awadajin Finney

reveals a possible alternation of [s] and [∫] or [t∫] in some lexical items of English origin containing [∫] or [t∫]: it is realized as [s] after low and mid vowels [a, ] but retains its original phonemic quality in initial position and after high vowels [i, u] (though it didn’t apply to the second /s/ in (15a)): (15)

a. c. e.

/asis/ ‘ash(es)’ /was/ ‘wash’ /ms/ ‘how much’

b. d. f.

/mas/ ‘tread on/mash’ /kalbas/ ‘calabash/gourd’ /tums/ ‘too much’

Attributing the absence of the above segments or the application of the phonological processes in Krio (and other Atlantic creoles) unequivocally to the influence of substrate languages could be problematic, as these tendencies are also evident in Pacific creoles (which are not claimed to have been influenced by West African substrate languages) and some non-standard varieties of English. Additionally, they are observed in some Indo-European languages. So there is a problem of identifying the real source(s) of these processes and we cannot conclude with certainty that the West African substrate languages were the primary sources. Singler (1996) further argues that similarities between features of creoles and substrate languages do not necessarily constitute evidence for substrate influence. Creole structures that could be accounted for by universal principles, in spite of similarities to substratal structures, should not be used as conclusive evidence of substratal influence. To enhance the validity of such influence, Singler (1996: 218) proposes three criteria that the features in question have to satisfy: 1. 2. 3.

They are not shared with the lexifier language. They are nontrivial. They are linguistically marked.

Thus, the proposal that vowel merging and some “stopping” and “fronting” phonological processes in creoles were possibly the direct effects of transfer from substrate languages is ruled out by Singler, as these phenomena are not linguistically marked.

.

The compromise account of the development of creoles

The compromise account of the development and evolution of the linguistic properties of creoles, unlike Singler’s (1996) proposal, focuses on the unmarked linguistic properties present in both creoles and substrate languages. It argues for an interaction of multiple factors – influence of both superstrate and substrate languages as well as universals of first and second language acquisition – in the development of the phonological systems of creoles (Alleyne 1980; Field 2004; Holm 1988; Mufwene 2001; Siegel 1999). As Holm (1988) states, there was influence from “both superstrate and substrate languages, universals of adult second language acquisition, borrowing from

Chapter 2. The phonotactics and syllable structure of Krio

‘adstrate’ languages, creole internal innovations, and the convergence of all or some of these” (ibid.: 105). Mufwene’s Complementary Hypothesis attempts to accommodate both the universalist and substrate positions on creolization. He proposes that these different positions are not mutually exclusive, but that they rather complement one another. As Mufwene suggests: Our position should not be based on the typically simplistic hypotheses which pervade the literature, in particular: baby talk, foreigner talk, exclusive or dominant substrate influence, language bioprogram, imperfect second-language learning, or exclusive or dominant superstrate influence (2001: 128).

The sentiment shared by these scholars (Holm 1988; Mufwene 1986, 2001; Mühlhäuser, 1980, 1986; Siegel 1999), nevertheless, is that the influence of substrate languages is mostly evident in properties for which substrate properties and universal tendencies converge. These are properties that are likely to be phonologically simple and unmarked. The Compromise position accounts for some (but not all) of the phonological inventory and processes (including the 7 cardinal vowel system and some phonological processes) observed in Krio. These proposed universal properties are also phonological properties of a number of West African languages. Unfortunately, there is evidence of segments in Krio (and other creoles), the co-articulated stops in particular, as well as other phonological processes that this position cannot account for. A number of Krio lexical items of African origin contain the co-articulated stops /kp/ and /gb/ (features of Kwa and Mande languages), which are certainly not universal or common cross-linguistically nor could they be argued to be unmarked, as their production involves the simultaneous articulation of two segments. This position additionally cannot account for the processes of palatalization and regressive vowel assimilation evident in Krio and other West African languages but not common across languages. Ironically, the process of palatalization involves a change from relatively unmarked and typologically more common phonemes to a relatively marked and typologically less common one.

. .

Universal and substrate influence on the syllable structure of Krio The theory of markedness and syllable structure

The theory of Markedness is based on the premise that humans are born with the same basic perceptual abilities and tendencies, and that children in language acquisition are predisposed to produce phonological phenomena that are relatively simple, evident in the widespread, almost universal, presence of such phenomena cross-linguistically. Unmarked phenomena are inherent in the initial assumptions that children make about the properties of language. There is thus an initial preference for the most basic





Malcolm Awadajin Finney Syllable Onset

Rhyme

(C)(C) Nucleus V

Coda (C)(C)

Figure 1. Traditional syllable structure

and simple syllable structure: CV. As a result, children apply a number of phonological processes during which they substitute phonemes that are less common crosslinguistically and are generally produced late by children with phonemes that are universal or very common across languages. They further simplify syllables to a more universal CV structure (Clarke 2003; Hulit and Howard 2002). Language acquisition thus involves a process of ordering and acquiring rules in a proposed sequence ranging from the least marked to the most marked. The traditional breakdown of a syllable is illustrated in Figure 1. Onset Clusters are proposed to conform to the Sonority Hierarchy:  →  →  →  →  → () That is, the initial member of the onset cluster should be the least sonorous. The reverse is true for a Coda: the final member is the least sonorous. A three-consonant onset in English invariably contains the phoneme /s/, which is proposed to occupy an extra syllabic position. Thus syllables could fall within an inventory ranging from the least marked (CV) to the most marked (CCVCC) based on the extent to which they adhere to the constraints of Markedness. Structurally unmarked languages are proposed to allow only the least marked syllable structure (CV) while more marked structures will allow increasing numbers of consonants in the onsets and codas of syllables depending on the degree of the structurally marked status of the language. The theory of Markedness has been used to account for not only the inventory of syllable structures across languages but also for the process of language acquisition (Archibald 1998; Davidson et al. 2004; Gnanadesikan 2004; Kager et al. 2004; Levelt and van de Vijver 2004; Meade 2004; Shinohara 2004). The child in L1 acquisition, guided by Universal Grammar, is proposed to initially produce syllable structures that are maximally unmarked (i.e., CV). Cluster reduction in syllables with consonant clusters is proposed to initially eliminate the coda as well as the cluster in the onset by deleting the more sonorous member(s) of the onset. In effect, complex onsets comprising of a stop-liquid or a fricative-liquid cluster are argued to be reduced to a single stop consonant while the coda is eliminated. More complex (and marked) syllable structures emerge later in the child’s linguistic development.

Chapter 2. The phonotactics and syllable structure of Krio

.

The theory of markedness and the process of creolization

The processes of elimination of the coda and (stop-fricative or stop-liquid) cluster reduction in the onset, resulting in the creation of a CV structure, are proposed to be exhibited by some creoles (Aceto 1996; Smith 2003) and is sometimes used as evidence of the “emergence of the unmarked” in creoles. Thus, the syllable structure of creoles is argued to be maximally unmarked.

.

Markedness, complex onsets, and codas in Krio

Lexical items of English origin containing consonant clusters in both the onset and coda generally undergo a process of simplification in Krio during which some of the segments in the cluster are deleted, but the deletion process does not always confirm to the Markedness proposal of cluster reduction. Thus, syllables are not necessarily modified to a maximally unmarked structure (CV). The Markedness constraints of no coda and no complex onset are generally violated in lexical items of English origin. In Krio, the extra syllabic /s/ in an /s/-initial cluster in lexical items of English origin is deleted eliminating the cluster in onsets containing two segments but reducing a cluster containing three segments to two (which is still a complex onset). That is, the ‘no onset cluster’ constraint applies only to two-segment clusters containing an extra syllabic /s/. For example: (16)

a. c. e. g. i.

/ton/ /pun/ /tinap/, /tanap/ /krep/ /trŋ/ or /traŋga/

‘stone’ (Basilect) ‘spoon’ (Basilect) ‘stand up’ ‘scrape’ ‘strong’

b. d. f. h.

/tap/ /pit/ /prεd/ /trit/

‘stop’ ‘spit’ ‘spread’ ‘street’

There are exceptions to this though. The vowels of some lexical items originating in English have undergone minor pronunciation changes while the pronunciations of consonants in the onset clusters have remained unchanged. They include words like: shrimp, shrink, square, scared, spanner, spider, speed, skin, school, etc. Nevertheless, cluster reduction in onsets in Krio (and possibly other creoles), when applied, is apparently triggered by the presence of an extra syllabic /s/, which is deleted in the process regardless of the number of segments in the onset. This is supported by the fact that onset clusters with no extra syllabic /s/ do not undergo any reduction process, as is evident in the following examples: (17)

a. c. e. g. i.

/ple/ /frk/ /drεb/ /broko/ /kraj/

‘play’ ‘frock’ ‘drive away’ ‘broken’ ‘cry’

b. d. f. h. j.

/pre/ /bred/ /blεn/ /klem/ /blo/

‘pray’ ‘bread’ ‘blind’ ‘climb’ ‘blow’

The markedness proposal that complex onsets comprising of stop-liquid or fricativeliquid clusters are reduced to a single stop consonant (an indication of the emergence/ retention of the unmarked) is not borne out in Krio.





Malcolm Awadajin Finney

Additionally, the ‘no coda’ markedness constraint is not adhered to in lexical items of English origin in Krio. Codas with single segments generally remain unchanged while those with two segments are consistently reduced to a single segment. For example: (18)

a. c. e. g. i.

/bεn/ /sεf/ /mεn/ /grn/ /bs/

‘bend’ ‘self ’ ‘mind’ ‘ground’ ‘burst’

b. d. f. h. j.

/εp/ /lεn/ /blεn/ /fs/ /trs/

‘help’ ‘lend’ ‘blind’ ‘first’ ‘trust’

There are isolated lexical items of English origin, though, (/gi/ ‘give’ and /gε/ ‘get’) which obey the ‘no coda’ constraint. As the above examples demonstrate, the syllable structure of Krio does not adhere to either the ‘no complex onset’ or the ‘no coda’ markedness constraints in lexical items of English origin, which in effect rules out Krio as a prototypically or maximally unmarked language in terms of syllable structure. Nevertheless, the fact that the number of segments is reduced in a complex onset when it contains an extra syllabic /s/ and in a coda when it contains a cluster also rules it out as a prototypically marked language. The evidence from modification of syllable structure in lexical items of English origin indicates that it is somewhere in between.

.

Substrate influence on the syllable structure of Krio

The argument that proposals of substrate influence should not be invalidated by the presence of similar phenomena elsewhere is extended by advocates (Alleyne 1980; Holm 1988; Spencer 1971) in accounting for the syllable structures exhibited by creoles (including Krio). The restricted range of syllable structures exhibited by creoles is proposed to have been influenced by the similarly restricted range allowed in substrate languages, which were the native languages of the speakers of pidgins that preceded most of the creoles. This view asserts that the restricted range of syllable structures in these pidgins was the direct result of transfer of the syllable structures of the primary (substrate) languages of its speakers. Many African languages are proposed to have restrictions that they impose on the presence of codas and onset consonant clusters (Dunstan 1969; Sey 1973; Simo Bobda 2000). The basic CV syllable structure, common in substrate languages, is thus argued to have been transferred into pidgins and subsequent creoles. Since substrate languages tend to exhibit a simple CV syllable structure, this has generally been maintained in lexical items of West African origin in Krio regardless of the number of syllables in the word. For example: Two syllables (19) a. /kab/ b. /kãga/ c. /koko/

‘welcome’ ‘magic’ ‘a bump on the body’

Chapter 2. The phonotactics and syllable structure of Krio

.

Three syllables (20) a. /arara/ b. /emina/ c. /okuru/ d. /rakpala/ e. /wahala/

‘nothing at all’ ‘plant with edible yam-like tuber’ ‘rabies’ ‘engage in rough and tumble’ ‘trouble’

Four syllables (21) a. /alafia/ b. /alakori/ c. /gbogbogia/ d. /kakatua/ e. /mlanke/ f. /sankamarũ/ g. /sumãgama/ h. /warawere/

‘peace of mind’ ‘good for nothing person’ ‘thick grass’ ‘the ring leader’ ‘handcart’ ‘a poor imitation of jewellery’ ‘incest’ ‘suddenly; in a flash’

Discussion and concluding remarks

As examples in this paper demonstrate, there are similarities between the vowel and consonant inventories in Krio (and other creoles) and vowel and consonant sounds that are proposed to reflect a default universal inventory because of the presence of these sounds in a majority of languages of the world (Haspelmath et al. 2005). Nevertheless, it will be simplistic and even erroneous to conclude that the Krio vowel and consonant systems necessarily exhibit universally maximally unmarked segments. The restricted representation of vowels and consonants is also evident in a number of substrate languages leaving open the possibility of transfer of these features from these languages into creoles (including Krio) during the pidginization phase. Additionally, the consonant inventory of Krio includes the co-articulated stops, which are not only common in a number of lexical items of West African origin but are also rare cross-linguistically, making them typologically marked. The similarities between the phonological processes (‘stopping’ and ‘fronting’) evident in Krio and in the outputs of children in L1 acquisition are inadequate as evidence of universally unmarked phonological processes in Krio. In L1 acquisition, the processes involve a much wider range of consonant segments that are replaced and they are generally applied in every position of the word (initial, medial and final) with regularity. The processes in Krio involve only a few segments, a couple of which (/θ, /) are substituted with regularity in all positions of the word and another handful (v, ∫) that are substituted in only some lexical items and in specific positions of the word. That is, the processes are applied with varying degrees of regularity, predictability, and frequency. For Krio to be considered maximally unmarked, these processes have to include a wider range of sounds and should be applied with greater regularity, as is the case in L1 acquisition.





Malcolm Awadajin Finney

The possible unmarked status of Krio is further undermined by the evidence of other phonological processes (including palatalization – which substitutes an unmarked segment with a marked one – and vowel nasalization) that are absent in L1 acquisition and not evident in a large number of languages. The presence of these processes, in addition to the presence of vowel harmony in both Krio and other West African languages, provides strong evidence of the influence of language transfer in the process of creolization. In terms of syllable structure, the lexical items of West African origin in Krio have generally retained the syllable configurations (generally CV) that they had in the languages from which they originated. The CV syllable structure is however not imposed on lexical items of English origin in Krio. A number of these lexical items have retained their complex onsets and reduction in onsets invariably involves the deletion of an extra syllabic /s/. The segments in the codas are reduced only when there is a cluster. A single segment cluster generally remains unchanged contrary to the ‘no coda’ markedness constraint. Thus, the syllable structure of lexical items of English origin manifested in Krio, like the segment inventory, does not conform to the maximally unmarked structure – CV. In effect, the Krio data presented do demonstrate an interaction of universal and substrate influences in both the type and number of segments allowed in syllables in Krio though there is no conclusive evidence to support a stronger influence of one over the other.

References Aceto, M. 1996. Early Saramaccan syllable structure: An analysis of complex onsets from Schumann’s 1778 Manuscript. Journal of Pidgin and Creole Languages 11.1. 23–44. Alleyne, M. 1980. Comparative Afro-American. Ann Arbor: Karoma Publishers. Archibald, J. 1998. Second language phonology. Amsterdam: John Benjamins. Baker, P. 2000. Theories of creolization and the degree and nature of restructuring. In Neumann-Holzschuh, I. and Schneider, E.W. (eds). Degrees of restructuring in creole languages. Amsterdam: John Benjamins. 41–63. Bickerton, D. 1975. Dynamics of a creole system. Cambridge: Cambridge University Press. Bickerton, D. 1977. Pidginization and creolization: Language acquisition and language universals. In Valdman, A., Reinecke, J. and Hancock, I.F. (eds). Pidgin and creole linguistics. Bloomington: Indiana University Press. 49–69. Bickerton, D. 1988. Creole language and the bioprogram. In Newmeyer, F. (ed.). Linguistics: The Cambridge survey. Volume 2: Linguistic theory: Extensions and implications. Cambridge: Cambridge University Press. 268–84. Bickerton, D. 1999. How to acquire language without positive evidence: What acquisitionists can learn from creoles. In DeGraff, M. (ed.). Language creation and language change. Cambridge, MA: MIT Press. 49–74. Bradshaw, A.T. von. 1966. A list of Yoruba words in Krio. Sierra Leone Language Review 5. 61–71.

Chapter 2. The phonotactics and syllable structure of Krio Broselow, E., Chen, S.-I. and Wang, C. 1998. The emergence of the unmarked in second language phonology. Studies in Second Language Acquisition 20. 261–80. Clarke, E.V. 2003. First language acquisition. Cambridge: Cambridge University Press. Comrie, B. 1989. Language universals and linguistic typology. Chicago: University of Chicago Press. Corne, C. 1987. Verb fronting in Creole: Transmission or bioprogram? In Gilbert, G.G. (ed.). Pidgin and creole languages: Essays in memory of John E. Reinecke. Honolulu: University of Hawaii Press. 93–112. Croft, W. 1990. Typology and universals. Cambridge: Cambridge University Press. Davidson, L., Jusczyk, P. and Smolensky, P. 2004. The initial and final states: Theoretical implications and experimental explorations of Richness of the Base. In Kager, R., Pater, J. and Zonneveld, W. (eds). Constraints in phonological acquisition, Cambridge: Cambridge University Press. 321–68. DeGraff, M. 2001. Morphology in creole genesis: Linguistics and ideology. In Kenstowicz, M. (ed.). Ken Hale: A life in language. Cambridge: MIT Press. 53–121. Dunstan, E. (ed.). 1969. Twelve Nigerian Languages. New York: Africana Publishing Corporation. Eckman, F. 1977. Markedness and the Contrastive Analysis Hypothesis. Language Learning 22. 315–30. Eckman, F. 1981a. On predicting phonological difficulty in second language acquisition. Studies in Second Language Acquisition 4. 18–30. Eckman, F. 1981b. On the naturalness of interlanguage phonological rules. Language Learning 31. 195–216. Field, F. 2004. Second language acquisition in creole genesis. In Escure, G. and Schwegler, A. (eds). Creoles, contact and language change: Linguistics and social implications. Amsterdam: John Benjamins. 127–60. Fyle, C. and Jones, E. 1980. A Krio–English dictionary. Oxford: Oxford University Press. Gnanadesikan, A. 2004. Markedness and faithfulness constraints in child phonology. In Kager, R., Pater, J. and Zonneveld, W. (eds). Constraints in phonological acquisition. Cambridge: Cambridge University Press. 73–108. Hall, R. 1966. Pidgin and creole languages. Ithaca: Cornell University Press. Hancock, I. 1986. The domestic hypothesis, diffusion and componentiality: An account of Atlantic Anglophone creole origins. In Muysken, P. and Smith, N. (eds). Substrata versus universals in creole genesis. Amsterdam: John Benjamins. 71–102. Hancock, I. 1987. A preliminary classification of the Anglophone Atlantic creoles with syntactic data from thirty-three representative dialects. In Gilbert, G.G. (ed.). Pidgin and creole languages: Essays in memory of John E. Reinecke. Honolulu: University of Hawaii Press. 264–333. Haspelmath, M., Dryer, M., Gil, D. and Comrie, B. 2005. The world atlas of language structures. Oxford: Oxford University Press. Holm, J. 1988. Pidgins and creoles. Volume 1. Cambridge: Cambridge University Press. Huber, M. 1999. Ghanaian Pidgin English in its West African context: A sociohistorical and structural analysis. Amsterdam: John Benjamins. Huber, M. 2000. Restructuring in vitro? Evidence from early Krio. In Neumann-Holzschuh, I. and Schneider, E.W. (eds). Degrees of restructuring in creole languages. Amsterdam: John Benjamins. 275–307. Hulit, L. and Howard, M. 2002. Born to talk: An introduction to speech and language development (3rd edition). Allyn and Bacon, Boston.





Malcolm Awadajin Finney Jones, E. 1971. Krio: An English-based language of Sierra Leone. In Spencer, J. (ed.). The English language in West Africa. Harlow: Longman. 66–94. Kager, R., Pater, J. and Zonneveld, W. 2004. Introduction: Constraints in phonological acquisition. In Kager, R., Pater, J. and Zonneveld, W. (eds). Constraints in phonological acquisition. Cambridge: Cambridge University Press. 1–53. Lefebvre, C. 1993. The role of relexification and syntactic reanalysis in Haitian Creole: Methodological aspects of a research program. In Mufwene, S.S. (ed.). Africanisms in Afro-American language varieties. Athens: University of Georgia Press. 254–79. Lefebvre, C. 1996. The tense, mood, and aspect system of Haitian Creole and the problem of transmission of grammar in creole genesis. Journal of Pidgin and Creole Languages 11. 231–311. Levelt, C. and van de Vijver, R. 2004. Syllable types in cross-linguistic and developmental grammars. In Kager, R., Pater, J. and Zonneveld, W. (eds). Constraints in phonological acquisition, Cambridge: Cambridge University Press. 204–18. Lumsden, J. 1999. Language acquisition and creolization. In DeGraff, M. (ed.). Language creation and language change. Cambridge, MA: MIT Press. 129–57. Meade, R. 2004. OT and the acquisition of Jamaican syllable structure. In Escure, G. and Schwegler, A. (eds). Creoles, contact and language change: Linguistics and social implications. Amsterdam: John Benjamins. 161–88. Mufwene, S.S. 1986. The universalist and substrate hypotheses complement one another. In Muysken, P. and Smith, N. (eds). Substrata versus universals in creole genesis. Amsterdam: John Benjamins. 129–62. Mufwene, S.S. 2001. The ecology of language evolution. Cambridge: Cambridge University Press. Mühlhäuser, P. 1980. Structural expansion and the process of creolization. In Valdman, A. and Highfield, A. (eds). Theoretical orientations in creole studies. New York: Academic Press. 19–55. Mühlhäuser, P. 1986. Pidgin and creole linguistics. Oxford: Basil Blackwell. Romaine, S. 1988. Pidgin and creole languages. London: Longman. Rutherford, W. (ed.). 1984. Language universals and second language acquisition. Amsterdam: John Benjamins. Sey, K. 1973. Ghanaian English: An explorative survey. London: Macmillan. Shinohara, S. 2004. Emergence of universal grammar in foreign word adaptations. In Kager, R., Pater, J. and Zonneveld, W. (eds). Constraints in phonological acquisition. Cambridge: Cambridge University Press. 292–320. Siegel, J. 1999. Transfer constraints and substrate influence in Melanesian Pidgin. Journal of Pidgin and Creole Languages 14. 1–44. Singler, J.V. 1996. Theories of creole genesis, sociohistorical considerations, and the evaluation of evidence: The case of Haitian Creole and the Relexification Hypothesis. Journal of Pidgin and Creole Languages 11. 185–230. Simo Bobda, A. 2000. Some phonological features across African accents of English. English Studies 81.3. 249–66. Smith, N. 2003. New evidence from the past: To epenthesize or not to epenthesize? That is the question. In Plag, I. (ed.). Phonology and morphology of creole languages. Tübingen: Max Niemeyer Verlag. 91–107. Spencer, J. 1971. West Africa and the English language. In Spencer, J. (ed.). The English language in West Africa. London: Longman. 1–34. Uffmann, C. 2003. Markedness, faithfulness and creolization: The retention of the unmarked. In Plag, I. (ed.). Phonology and morphology of creole languages. Tübingen: Max Niemeyer Verlag. 3–23.

 

Tone on quantifiers in Saramaccan as a transferred feature from Kikongo Marvin Kramer Quantifiers in Saramaccan have tonal characteristics that cannot be accounted for by the leftward tone spread rules that apply elsewhere, but can be by a rightward tone spread analysis. Tone spread in Kikongo is rightward. High tone spread on certain non-universal quantifiers maximizes the number of high tones in those phrases, while blocking high tone spread on certain universal quantifiers maximizes low tones in those phrases in Saramaccan. The same effects are achieved, but by a different mechanism, in quantifier phrases in Kikongo. Quantifiers in Kikongo affect the tones on items even when they are not adjacent. Quantifier phrases in Saramaccan block high tone spread in serial verb constructions, although the quantifiers are not necessarily adjacent to the affected items. The similarities in the effects in quantifier phrases in both languages would point to Kikongo as the source, but the differences in achieving those effects in Saramaccan represent compromises in the Kikongo input that would suggest transfer rather than borrowing.

.

Introduction

There is general agreement among researchers that Fongbe was the primary substratal influence on the Surinamese English-lexified creoles (Migge 1998; Smith 2001). This influence included grammatical features as well as lexical items. It is recognized that Kikongo was also a substrate language, but that its contribution was limited to lexical items and ideophones. In Saramaccan, one of the Surinamese creoles, there are tonal characteristics of non-universal quantifiers that cannot be attributed to tones in Fongbe, and in fact override tone rules in Saramaccan from Fongbe. These tone rules, however, resemble tonal characteristics of non-universal quantifiers in Kikongo. To a lesser extent, tone rules relating to universal quantifiers in Saramaccan resemble tonal characteristics of universal quantifiers in Kikongo. It is argued in this paper that the tonal characteristics of non-universal and universal quantifiers in Saramaccan are a transferred feature from Kikongo. As such, these tonal characteristics would be the



Marvin Kramer

first recognized influence of Kikongo on Saramaccan beyond the lexicon and lexical phonology.1 Suprasegmental features in Saramaccan have been described as tones or as tones and intonation (Good 2003, 2004; Kramer 2005; Rountree 1972; Voorhoeve 1961). There are basically two types of tone spread rules in Saramaccan, one for certain serial verb constructions (SVCs), and default rules for the other SVCs and for non-serial constructions. Migge argues that SVCs in the Surinamese creoles transferred from Fongbe. The tone spread rules for SVCs in Saramaccan resemble the general tone spread rules for Fongbe, as there are no special tone spread rules for SVCs in Fongbe. But in Saramaccan, tone spread rules that apply to certain SVCs override the default tone spread rules. Kramer (2005) argues that the default rules are a reinterpretation of the Fongbe tone spread rules, and that SVCs transferred with the Fongbe rules at a time after the default tone spread rules had been established. With this argument, the creolization of Saramaccan would have been a gradual process (Arends 1993). Now in addition, there are the tone spread rules for non-universal and universal quantifiers in Saramaccan that override both the SVC rules and the default rules. It will be argued in this paper that the tone spread rules for quantifiers transferred after the rules for SVCs and the default rules. Examples from Kikongo are from Odden (1994), and from Fongbe are from Lefebvre and Brousseau (2002). Examples from Saramaccan are from my own fieldwork.

.

Saramaccan high tone spread rules

Saramaccan has high and low phonemic tones, but tone spread only involves high tone. Tones are specified high or low lexically, but there are syllables that are unspecified for tone. High tone may spread onto these unspecified syllables, which otherwise would surface with low tone. Good (2004) argues that the lexicon in Saramaccan is split, with the majority of words having one specified high tone, reflecting the stressed syllable in English, Portuguese or Dutch, with the other syllables unspecified. But in the other type of word, every syllable is specified, either high or low. Since high tone may spread only to unspecified syllables, words where all the syllables are specified do not participate in high tone spread.

.

Default high tone spread rules

The default high tone spread rules are described in Rountree (1972). There is no low tone (L) spread. High tone (H) spreads to tone bearing units (TBUs) unspecified for tone in a domain that includes two TBUs with H that bracket TBUs unspecified for tone. Since every word has at least one TBU specified for tone,2 the domain for H spread can .

Smith (2004) argues that Kikongo influenced nasalization in lexifier items.

. It is easier to analyze surface L syllables in many function words as specified L tones rather than stipulate that they block tone spread.

Chapter 3. Tone on quantifiers in Saramaccan

involve only two words. TBUs specified for tone are marked with diacritics, e.g., á for H and à for L, while unspecified TBUs are unmarked, as in (1), with the contrasting tonal inflection mì ‘I’ in (1a) and mí ‘my’ in (1b): (1)

a.

mì woóko > mì wòókò I work ‘I worked’

b.

mí taáta > mí táátà my father ‘my father’

Most minimal pair contrasts involve TBUs that are specified and those unspecified for tone, that surface as low unless tone raising rules apply: (2)

a.

jàà > jàà Low-Low ‘to sow’

b.

jáa > jáà High-unspecified ‘year’

In (3) there are the domains {dí mujέ}ε ‘the woman’, and mu{jέε kulé} ‘woman run’, but not for kulé and the adverb taáánǵtáánga ‘hard’, as tone spread is blocked between verbs and adverbs (Rountree 1972: 322). There is change for {dí mujέ}ε and for mu{jέε kulé}, where the intervening TBUs surface with H. TBUs with changed tones are underlined: (3)

dí mujέε kulé taángátánga > dí mújέέ kúlé tàángátáángà the woman run strong-strong ‘the woman ran hard.’

Devonish (1989) argues for leftward H spread, based on intonational sentence final lowering as described in Rountree (1972: 325). For mu{jέε kulé} in (4), the sentence final H of kulé is effectively unlinked, and there is no H spread: (4)

dí mujέε kulé]s > dí mújέὲ kùlè]s the woman run ‘the woman ran.’

Since intonational rules follow tone spread rules, the final lowering occurs after H spread, represented by H > L, thus lowering the unspecified TBUs as well, in (5a). Had H spread been rightward, these unspecified TBUs would surface with H. Non-final H spread for mu{jέε kulé} is in (5b): (5)

a.

H H H > L

di muj kule]s > dí mújέὲ kùlè the woman run ‘the woman ran’





Marvin Kramer

b.

H H H

di muj kule taánátáána > the woman run strong.strong ‘the woman ran hard’

dí mújέέ kúlé tàánátáánà

The domain for leftward H spread includes TBUs with unspecified tone bracketed by TBUs with specified H, as in (6) where H spreads to the unspecified a of pampú. These rules are described in Rountree (1972): (6)

dí pampú lépi kàà > dí pámpú lépì kàà the pumpkin ripe already ‘the pumpkin is already ripe’

H spreads between subject and verb, as it does in Fongbe (Wiesemann 1991). In both Fongbe and Saramaccan predicative property items are intransitive stative verbs: (7)

dí foló donú àwáà > dí fóló dónú àwáà the flower yellow now ‘the flower is yellow now’

H spread is blocked at left edge of prenominal property items, but not at the right edge: (8)

dí donú foló àkí > dí dònú fóló àkí the yellow flower here ‘this yellow flower’

H spread is blocked at right edge of verbs, as it is in Fongbe (Wiesemann): (9)

mi pεέ tutú a dí tɔɔ́u > mì pɛ̀έ tùtú à dí tɔ́ɔ́ù I play horn at the wedding ‘I played horn at the wedding’

These rules from Rountree (1972) have been analyzed by Devonish (1989) as leftward H spread, as mentioned, and by Good (2003) as head-leftward, H spreading leftward from the head of a constituent to the preceding word. Thus H spreads leftwards from a verb to the preceding subject, or from a noun to the preceding modifier. The spread of H is blocked by a specified TBU, so that H spreads only to one word, since every Saramaccan word has at least one TBU specified for tone. Therefore the spread of H from a noun, for example, would not extend beyond the prenominal property item to a preceding modifier or determiner, explaining the H spread block in (8). But this rule does not apply to some non-universal quantifi ers, such as óto ‘other’. Thus in (10a), the unspecified TBU of óto surfaces as H, even

Chapter 3. Tone on quantifiers in Saramaccan

though the following TBU is specified L, so that H could not have spread leftward from the following specified L of sɛ̀mbɛ̀ ‘person’. In (10b), on the other hand, the unspecified TBU of hánso ‘handsome’ surfaces as L, by the head-leftward rule. Another exception to the head-leftward rule is in (10c), where the unspecified TBUs of both óto and donú ‘yellow’ surface as H, in contrast to (10d) where, by the head-leftward rule, leftward H spread does not extend from the specified H of foló ‘flower’ beyond the specified H of donú. In contrast to (10c), H spread does not occur before other non-universal quantifiers, such as síkísi ‘six’ in (10e), so óto surfaces in its citation form. In (10) and following examples, quantifiers are in bold type, surface Hs on unspecified TBUs are underlined: (10)

a.

b.

c.

d.

e.

dí óto sὲmbὲ > dí ótó sὲmbὲ the other person ‘the other person’ dí hánso sὲmbὲ > dí hánsò sὲmbὲ the handsome person ‘the handsome person’ dí óto donú foló àkí > dí ótó dónú fóló àkí the other yellow flower here ‘this other yellow flower’ dí ɡãã́ donú foló àkí > dí ɡã̀ã́ dònú fóló àkí the big yellow flower here ‘this big yellow flower’ dí óto síkísi sὲmbὲ > dí ótò síkísì sὲmbὲ the other six person ‘the other six people’

The default head-leftward H spread rule does not account for the surface H on unspecified TBUs in (10a) and (10c).

.

Rightward high tone spread on non-universal quantifiers

A rightward H spread analysis can account for the surface H on óto in (10a) and (10c). The rightward H spread, unlike leftward spread, does not require a target H at the opposite edge of the domain, as in (11a) and (11b). Rightward H spread applies only to óto and the following unspecified TBUs and is blocked by a specified TBU, in (11c), while the default leftward spread occurs on the noun. There is, of course, no leftward spread in (11d) for the final TBU of taánga ‘strong’: (11)

a.

H

L

dí oto smb > dí ótó sὲmbὲ the other person ‘the other person’





Marvin Kramer

b.

H

L

H

dí oto adja womi > dí ótóàdjà wómì the other husky man ‘the other husky man’

c.

H HH

dí oto donu folo dέ > dí ótó dónú fóló dέ the other yellow flower there ‘the other yellow flower there’

d.

H

H

L

dí oto taana smb > dí ótó táánà sὲmbὲ the other strong person ‘the other strong person’

Rightward H spread rules have been proposed for SVCs (Kramer 2005), but a rightward H spread rule in non-serial constructions, applying only to non-universal quantifiers, would appear to be somewhat ad hoc.

.

Non-universal quantifiers in Kikongo and Saramaccan

The surface H on óto and other non-universal quantifiers is somewhat anomalous in Saramaccan, in that it does not occur on any other modifier. There is, however, a parallel tonal anomaly with non-universal quantifiers in the substrate language Kikongo. These similarities in tone on non-universal quantifiers suggest Kikongo as the source.

.

Non-universal quantifiers in Kikongo

In Kikongo, phrases with non-universal quantifiers block normal tone lowering of the noun, termed argument effacement in Odden (1994), represented by subscript Ne. In (12a), the noun bikúní ‘farmers’ surfaces as bíkùnì, with lowered tones on the final two TBUs due to argument effacement that accompanies this and most other constructions. One of the rare contexts where argument effacement is blocked is in phrases with non-universal quantifiers (ibid.: 192), as in (12b). High tones in Kikongo are marked á, and low tones are unmarked: (12)

a.

mweení bikúníe byámbóte > mwení bíkunie byámbóte I.saw farmers good argument effacement ‘I saw good farmers’

Chapter 3. Tone on quantifiers in Saramaccan

b.

tubweení bikúní byánkáka > tubweení bikúní byánkáka we.saw farmers other no argument effacement ‘we saw other farmers’

Blocking argument effacement maximizes the number of high tone TBUs in phrases with non-universal quantifiers. The rightward spread of H on óto in Saramaccan has the same effect, and it also has this effect when it precedes an adjective and, therefore, is not adjacent to the noun, as in (10c). Interestingly in Kikongo, the non-universal quantifier blocks argument effacement even when not adjacent to the noun. The effect of maximizing H is similar to Saramaccan, but in Kikongo it is the noun that surfaces with H, not the non-universal quantifier. In (13b), milooní ‘teachers’ surfaces without effacement when myamííni ‘many’ is non-adjacent, just as it does when this conditioning item is adjacent (Odden 1994: 193), in (13a): (13)

a.

b.

tubweení milooní myamííni myákinshasa > tubweení milooní myamííni myákinshasa we.saw teachers many of Kinshasa ‘we saw many teachers of Kinshasa’ tubweení milooní myákinshasa myamííni > tubweení milooní myákinshasa myamííni we.saw teachers of Kinshasa many ‘we saw many teachers of Kinshasa’

Odden (1994: 198) argues that because of the variation in word order, seen in (13), the phonological rule of argument effacement must have direct access to the semantic level of representation. While there is no argument for such access in Saramaccan, in Kikongo it would make the semantics and the tone rules of non-universal quantifiers particularly salient, and a likely source for the transfer of the exceptional phonology of non-universal quantifiers in Saramaccan.

.

Non-universal quantifiers in Saramaccan

There are two types of non-universal quantifier in Saramaccan, distinguished by their tonal characteristics. One type, termed isolating by semantics, in (i) below, includes óto and exhibits the tonal phenomena of óto, analyzed as rightward H spread. The other type, termed associating by semantics, in (ii) below, does not have rightward H spread: i.

isolating:

ii. associating:

óto ‘other’, báka ‘last’; ordinals fósu ‘first’, u féífi ‘fifth’, etc. mɔ̃́ ɔ̃ ‘more’, híla/hía ‘many’; cardinals féífi ‘five’, etc.

The head-leftward H spread rule applies to associating non-universal quantifiers in contexts where rightward H spread applies to isolating non-universal quantifi ers,





Marvin Kramer

in (14a, b). But adjacent associating non-universal quantifiers have H spread, in (14c): (14)

a.

b.

c.

híla sɛ̀mbɛ̀ > hílà sɛ̀mbɛ̀ many person ‘many people’ déé síkísi donú foló àkí > déé síkísì dònú fóló àkí the. six yellow flower here ‘these six yellow flowers’ síkísi m̃́  foló > síkísí m̃́̃́ fóló six more flower ‘six more flowers’

As there is H spread outside the head-leftward rule, as seen in (14c), associating nonuniversal quantifiers in Saramaccan also maximize the possible number of TBUs with H in phrases, like isolating non-universal quantifiers in Saramaccan and all nonuniversal quantifiers in Kikongo.

.

Serial verb constructions and non-universal quantifiers in Saramaccan

Although the two types of non-universal quantifier in Saramaccan have different H spread rules for quantifier phrases, they both have the effect of blocking H spread rules for SVCs. This blocking as such cannot have transferred from Kikongo, because Kikongo is not a verb serializing language (Carter and Makoondekwa 1987), but may be the result of a conflict between Fongbe SVC rules and Kikongo quantifier rules. The effect on non-adjacent items in a quantifier phrase, however, in this case in the SVC, could be seen as a Kikongo feature. Verb serialization is characteristic of the Gbe languages, with the SVCs of Fongbe in particular being the primary source in the transfer of SVCs into the creole languages of Suriname (Migge 1998). There is H spread in Fongbe, and this spread is rightward (Brousseau 1991). In Fongbe there are no special H spread rules for SVCs, but in Saramaccan there are. However, certain SVCs can be analyzed as having rightward H spread (Kramer 2005), while non-serial constructions and the other SVCs have headleftward H spread. In Saramaccan, SVCs with non-universal quantifiers, which by the phonosyntactic rules for SVCs would be expected to have rightward H spread, have head-leftward H spread instead.

.

Rightward H spread in serial verb constructions in Saramaccan

Under certain conditions SVCs in Saramaccan have rightward H spread, while SVCs that do not meet these phonosyntactic conditions have the default head-leftward H spread (Kramer 2005). Where there can be rightward H spread to the left edge of the second verb (V2), there is leftward H spread on V2. There are two contexts for

Chapter 3. Tone on quantifiers in Saramaccan

rightward H spread in SVCs. One is an SVC where two serial verbs are juxtaposed. The other is an SVC where the TBUs of the constituent between serial verbs can, with head-leftward rules, surface with H, with the possibility of surface L at the edges of the constituent. This intervening constituent, then, would mimic the surface tones of a tonal domain in Fongbe, which would be H on all TBUs with the possible exception of L at the domain edges (Brousseau 1991). SVCs that do not meet these conditions have head-leftward H spread. There are, then, two types of leftward H spread; the leftward H spread from V2 without a target H at the domain edge in SVCs that have rightward H spread, and the head-leftward H spread that requires a target H at the domain edge.3 For SVCs with adjacent verbs there is tone raising between the verbs, in (15a), as there is in Fongbe (Wiesemann 1991), even though in both Saramaccan and Fongbe there is a domain boundary at the right edge of verbs in non-serial constructions. This domain boundary for adjacent verbs in Saramaccan marks the limits of rightward H spread up to V2, and leftward H spread on V2. As a convenient reference for this study, the juncture to the right of an item with rightward spread will be termed a sandhi juncture, marked by +, as in Rountree (1972), in (15b). Where the final TBU of V1 is L, there is no leftward H spread on V2 because there is no rightward H spread on V1, in (15c), and so there is no + mark in (15d). In (15e), rightward H spread can be seen on V1, when V2 has final lowering the H and leftward H spread are lowered, the H in (15f): (15)

a.

a hópo kulé ɡó a dɔ́ɔ > à hópó kúlé ɡó à dɔ́ɔ̀ he get.up run go at door ‘he got up and ran outside’

b.

H

H

a hopo + kule > à hópó kúlé ó à d 

c.

a lὲɡὲdὲ butá dí táki dá mi > à lὲɡὲdὲ bùtá dí tákí dá mí. . . he lie put the talk give me ‘he lied, put the talk to me’

d.

L H

a ld buta > a lὲὲdὲ bùtá dí tákí dá mi

e.

a hópo kulé]s > à hópó kùlè]s 3. rise run ‘he got up and ran.’

. An alternative analysis would allow rightward H spread to V2 to be the H target at the domain left edge, so that there would only be the default head-leftward H spread from V2.





Marvin Kramer

f.

H

H >L

a hopo + kule]s > à hópó kùlè

It is the SVC with non-adjacent verbs that is significant for non-universal quantifiers, because when one occurs in the intervening constituent it blocks rightward H spread. For non-adjacent verbs in an SVC, the phonosyntactic structure of the intervening constituent determines which of the H spread rules applies. As with adjacent verbs, if rightward H spread can occur, then leftward H spread applies to V2. If it cannot occur, then the head-leftward H spread rule applies. The syntactic constraints on headleftward H spread, however, in part determine whether rightward H spread can occur. In (16a), for example, there can be leftward H spread from a noun to a determiner, then there can be rightward H spread from the determiner, a sandhi juncture marked with + in (16b). The intervening constituent is marked with [ ] brackets. If, however, there can be no rightward H spread, as in (16c) where there is no determiner, then there is no sandhi juncture and no rightward H spread on the SVC. The H spread on butá ‘put’ in (16d) is the default head-leftward H spread: (16)

a.

a féni dí foló butá a dí táfa > à féní dí fóló bútá à dí táfà he find the flower put on the table ‘he found the flower and put it on the table’

b.

H

H

H

H

a feni [di + folo] buta... > à féní dí fóló bútá...

c.

a féni foló butá a dí táfa > à fénì fòló bútá à dí táfà he find flowers put on the table ‘he found flowers and put them on the table’

d.

H

H

H

a feni [folo] buta... > à fènì fòló bútá...

It is the possibility of rightward H spread that determines a sandhi juncture, as can be seen when a determiner is followed by a noun with TBUs specified as L, in (17a). The leftward H spread on butá is not the head-leftward rule, in (17b), since there is no H target at the left edge of the domain. Rather, it is the V2 leftward H spread conditioned by sandhi juncture. In (17c), like (16c), there is no sandhi juncture, and no head-leftward H spread on butá because there is no H target at the left edge of the domain: (17)

a.

a hópo dí lòɡòsò butá a téla > à hópó dí lòɡòsò bútá à télà he lift the turtle put on shore ‘he lifted the turtle and put it on shore’

Chapter 3. Tone on quantifiers in Saramaccan

b.

H

H

L

H

a hopo [di + looso] buta... > à hópó dí lòòsò bútá...

c.

a hópo lòɡòsò butá a téla > à hópò lòɡòsò bùtá à télà he lift turtle put on shore ‘he lifted turtles and put them on shore’

d.

H

L

H

a hopo [looso] buta... > à hópò lòòsò bùtá...

The block on leftward H spread from prenominal property items, determined by the head-leftward rule, also serves to block the possibility of rightward H spread in an SVC. In (18a), there is no rightward H spread on the SVC because not all the junctures in the intervening constituent are sandhi junctures, [dí donú + àkàtà]. In (18b), on the other hand, the sandhi juncture in the intervening constituent [lánɡa + sὲmbὲ] conditions rightward H spread on the SVC, as seen in the surface H on the second TBU of lánɡa ‘tall’. But in (18c), there can be no rightward H spread from ɡàdjà ‘husky’, so there is no sandhi juncture, [ɡàdjà wómi], and so no rightward H spread on the SVC: (18)

a.

b.

c.

a wási dí donú àkàtà jabí a sitónu > à wásì dí dònú àkàtà jàbí à sìtónù she wash the yellow head pad open on stone ‘she washed the yellow head pad and opened it out on a stone’ de féni lánɡa sὲmbε butá a fési > dè féní lánɡá sὲmbὲ bútá à fésì they find tall person put at face ‘they found tall people and put them in front’ (for soccer) de féni ɡàdjà wómi butá a fési > dè fénì ɡàdjà wómí bútá à fésì they find husky man put at face ‘they found husky men and put them in front’ (for soccer)

When two items in an intervening constituent have adjacent TBUs with specified H, there is a de facto sandhi juncture even if H spread does not occur under the headleftward rule, as in dí(+)lánɡa ‘the tall’ in (19). This condition is determined lexically rather than through H spread, as there is no H spread for dí donú ‘the yellow’ in (18a). But in (19), there is in effect a sandhi juncture at each juncture, [dí(+)lánɡa + sὲmbὲ], which allows rightward H spread on the SVC: (19)

de súti dí lánɡa sὲmbε kíi > dè sútí dí lánɡá sὲmbὲ kíì they shoot the tall person kill ‘they shot and killed the tall person’





Marvin Kramer

Adjectives of nationality, which are property items that do not occur as stative verbs, allow sandhi at their left edge with other property items and numerals (Rountree 1972), with rightward H spread on the SVC: (20)

de féni síkísi holánsi sὲmbὲ butá a fési > dè féní síkísí hólánsí sὲmbὲ bútá à fésì they find six Dutch person put at face ‘they found six Dutch people and put them in front’ (for soccer)

The determining factor in rightward H spread for SVCs with non-adjacent verbs is the continuity of sandhi junctures in the intervening constituent. Rightward H spread can occur, then, even if one of the verbs has TBUs specified L, as in (21) with gìlì ‘scrub’: (21)

a wásí dí koósu gìlì > à wásí dí kóósù gìlì she wash the dress scrub ‘she washed and scrubbed the dress’

The sandhi junctures between adjacent verbs in SVCs that allow high tone spread are marked with +. The sandhi junctures between items in an intervening constituent in SVCs with non-adjacent verbs are marked with [. . . + . . .]. There are two conditions where sandhi junctures do not occur: 1. 2.

where the TBU to the left is specified with L, so that rightward H spread is not possible where the TBU to the right blocks head-leftward H spread, and has surface L.

Sandhi juncture between the verbs of an SVC is necessary for rightward H spread. But for SVCs with non-adjacent verbs, rightward H spread on V1or leftward H spread on V2 is not necessary for rightward H spread on the intervening constituent, as either of these verbs may have a TBU that is specified L, such as gìlì, in (21). The surface forms of adjacent verbs therefore may not always be the same as the surface forms of those verbs when non-adjacent: (22)

a. b.

. . .gìlì jabí. . . > . . .gìlì jàbí. . . . . .gìlì [. . . + . . .] jabí. . . > . . .gìlì [. . . + . . .] jábí. . .

The surface tones in an intervening constituent with rightward H spread parallel the surface tones in a Fongbe tonal domain. In Fongbe, H, marked á, spreads rightwards from the leftmost H of a tonal domain to its right boundary (Lefebvre and Brousseau 2002: 22). In the rightward spread Ls, unmarked, are unlinked, including TBUs with a rising LH, marked ǎ. This LH occurs only after voiced consonants, but in H spread the TBU surfaces as H. In (23a), ɡbaˇ ‘build’ has LH, which surfaces as ɡbá when H spreads. Monosyllabic objects like xwé ‘house’ are considered stray syllables, and so are included in the domain that would otherwise border at the right edge of

Chapter 3. Tone on quantifiers in Saramaccan

the verb. In (23b), H spreads to the stray syllable4 tè ‘yam’, which surfaces as tê with HL, marked â, the surface tone for domain final L. In (23c) H spreads, unlinking the L of nɔ̀ and sà ‘sell’: (23)

a.

b.

c.

é ɡbǎ xwé > é ɡbá xwé (s)he build house ‘(s)he built some houses’ à só tè > à só tê you mash yam ‘you mashed yams’ é nɔ̀ sà tè > é nɔ̀ sà tê (s)he  sell yam ‘(s)he usually sells yams’

Fongbe (Lefebvre and Brousseau 2002: 24) Fongbe (ibid.: 23) Fongbe (ibid.: 23)

The surface tones in (23a) are all H, which is also acceptable for intervening constituents in rightward H spread SVCs in Saramaccan. In (23b) both the left edge and the right edge of the domain have an L, which is similarly acceptable in Saramaccan. And in (23c), there is rightward H spread over several words, again acceptable in Saramaccan intervening constituents with rightward H spread. For both the Fongbe tonal domain and these Saramaccan intervening constituents the surface tones are all H, with the possibility of L at either border. This parallel suggests that the intervening constituents that have the Fongbe rightward H spread also have the Fongbe tonal domain surface structure, as argued in Kramer (2005).

. Non-universal quantifiers in serial verb constructions The rightward H spread rules for SVCs do not apply when non-universal quantifiers are in the intervening constituent. Where there would be a [. . . + . . .] sandhi juncture and SVC rightward H spread, as in (24a), in (24b) there is no rightward H spread on the SVC, even though the isolating non-universal quantifier’s rightward H spread gives a surface resemblance to a sandhi juncture: (24)

a.

b.

a féni hánso sὲmbὲ butá a fési > à féní hánsó sὲmbὲ bútá à fésì he find handsome person put at face ‘he found handsome people and put them in front’ a féni óto sὲmbὲ butá a fési > à fénì ótó sὲmbὲ bùtá à fésì he find other person put at face ‘he found other people and put them in front’

When the number of TBUs with H in the intervening constituent is extended with a de facto sandhi juncture dí(+)lánɡa ‘the tall’ in (19), copied as (25a), there is rightward

.

‘Stray syllable adjunction’ in Hayes (1980).





Marvin Kramer

H spread. But there is no H spread on SVC verbs with a non-universal quantifier, in (25b), even though it has the same tonal structure as (25a). In (25c) there is similarly no H spread on SVC verbs, which would otherwise be licensed by the intervening constituent: (25)

a.

b.

c.

de súti dí lánɡa sὲmbὲ kíi > dè sútí dí lánɡá sὲmbὲ kíì they shoot the tall person kill ‘they shot and killed the tall person’ a téi dí báka àkàtà butá a táfa líba > à téì dí báká àkàtà bùtá à táfá líbà she take the last headpad put at table above ‘she took the last headpad and put it on the table’ a féni dí óto lánɡa sὲmbὲ butá a fési > à fénì dí ótó lánɡà sὲmbὲ bùtá à fésì he find the other tall person put at face ‘he found the other tall person and put him in front’

Associating non-universal quantifiers, unlike isolating, do not have rightward H spread, as seen in (14a), híla sὲmbὲ ‘many people’ > hílà sὲmbὲ. In this way they resemble prenominal property items, like hánso sὲmbὲ ‘handsome people’ > hánsò sὲmbὲ. But unlike the intervening constituent hánso sὲmbὲ > hánsó sὲmbὲ with SVC rightward H spread, in (24a), associating non-universal quantifiers in the intervening constituent do not condition rightward H spread on SVCs, seen in (26a). In this way they resemble isolating non-universal quantifiers. The de facto sandhi juncture as in dí(+)lánɡa in (25a) does not exist for tú lánɡa ‘two tall’ in (26b). The H spread for adjacent associating non-universal quantifiers would also create a de facto sandhi juncture, but does not condition rightward H spread on SVCs, in (26c): (26)

a.

b.

c.

.

a féni féífi sὲmbὲ butá a fési > à fénì féífì sὲmbὲ bùtá à fésì he find five person put at face ‘he found five people and put them in front’ a féni tú lánɡa sὲmbὲ butá a fési > à fénì tú lánɡà sὲmbὲ bùtá à fésì he find two tall person put at face ‘he found two tall people and put them in front’ a féni síkísi mɔ̃́ɔ foló butá a táfa líba > à fénì síkísí m  fóló bùtá à táfá líbà he find six more flower put on table above ‘he found six more flowers and put them on top of the table’

Non-universal quantifiers and SVCs with adjectives of nationality

Adjectives of nationality, such as holánsi ‘Dutch’, are prenominal property items that could be termed adjectives, in that unlike most other property items they do not also function as intransitive stative verbs. They have the nominal characteristic of H spread with a preceding property item (Rountree 1972: 319), but unlike nouns they do not have H spread with a preceding determiner.

Chapter 3. Tone on quantifiers in Saramaccan

Isolating non-universal quantifiers have H spread when they are followed by an adjective of nationality, in (27a), but they block rightward H spread in SVCs with adjectives of nationality, as with other SVCs, in (27b) and (27c): (27)

a.

b.

c.

dí fósu saamáka sὲmbὲ > dí fósú sá ámákà sὲmbὲ the first Saramaccan person ‘the first Saramaccan person’ a féni óto holánsi sὲmbὲ butá a fési > à fénì ótó hólánsì sὲmbὲ bùtá à fésì he find other Dutch person put at face ‘he found other Dutch people and put them in front’ a féni óto lánɡa saamáka sὲmbὲ butá a fési > à fénì ótó lánɡá sáámákà sὲmbὲ bùtá à fésì he find other tall Saramaccan people put at face ‘he found other tall Saramaccan people and put them in front’

Associating non-universal quantifiers, on the other hand, do condition H spread on SVCs with an adjective of nationality, as in (20) with síkísí hólánsí sὲmbὲ ‘six Dutch people’, and in (28): (28)

a féni m  holánsi sὲmbὲ butá a fési > à féní m  hólánsí sὲmbὲ búta à fésì he find more Dutch person put at face ‘he found more Dutch people and put them in front’

Isolating and associating non-universal quantifiers do not have H spread with each other, so an SVC with óto féífi ‘other five’ and an adjective of nationality would not have H spread, in (29), because of the surface L within the intervening constituent: (29)

a féni déé óto féífi holánsi sὲmbὲ butá a fési > à fénì déé ótò féífí hólánsì sὲmbὲ bùtá à fésì he find the. other five Dutch person put at face ‘he found the other five Dutch people and put them in front’

The different effects of the two types of non-universal quantifier on SVCs with adjectives of nationality may relate to the difference in their H spread; isolating have rightward H spread, and associating do not. Adjectives of nationality in Saramaccan, as mentioned, have the nominal characteristic of H spread with a preceding property item or non-universal quantifier, although they do not have the nominal characteristic of H spread with a preceding determiner. Adjectives of nationality may treat associating nonuniversal quantifiers as property items and spread H leftward, allowing rightward SVC spread, while treating isolating differently because they have rightward H spread.

.

Universal quantifiers

The argument for the transfer of universal quantifier tonal characteristics is not as strong as for the non-universal quantifiers, but the case for a phonology-semantics





Marvin Kramer

interface for universal quantifiers in Kikongo is stronger than for non-universal quantifiers.

. Universal quantifiers in serial verb constructions In Saramaccan the universal quantifier híi ‘all’ has no rightward H spread before nouns, in (30a), like associating non-universal quantifiers. Yet it is unlike them in blocking H spread to all modifiers, in (30c–e), including adjectives of nationality, in (30d), and to determiners, in (30b). Universal quantifiers are like non-universal quantifiers in blocking H spread on SVCs when in the intervening constituent, as illustrated by the postnominal universal quantifier túu ‘all’, in (30f). In the examples in (30), then, V1 has no tone raising. Universal as well as non-universal quantifiers are in bold type: (30)

a.

b.

c.

d.

e.

f.

a féni híi sὲmbὲ butá a fési > à fénì híì sὲmbὲ bùtá à fésì he find all person put at face ‘he found everybody and put them in front’ a féni híi déé sὲmbὲ butá a fési > à fénì híì déé sὲmbὲ bùtá à fésì he find all the. person put at face ‘he found all the people and put them in front’ a féni híi féífi sὲmbὲ butá a fési > à fénì híì féífì sὲmbὲ bùtá à fésì he find all five person put at face ‘he found all five people and put them in front’ a téi híi holánsi sὲmbὲ butá a fési > à téì híì hòlánsì sὲmbὲ bùtá à fésì he take all Dutch person put at face ‘he took every Dutch person and put him in front’ a féni híi féífi holánsi sὲmbὲ butá a fési > à fénì híì féífí hólánsì sὲmbὲ bùtá à fésì he find all five Dutch person put (at face) ‘he found all five Dutch people and put them in front’ a féni déé mujέε dέ túu butá a fési > à fénì déé mújέέ dέ túú bútá à fésì he find the. woman there all put at face ‘he found all the women and put them in front’

It might appear that the lack of H spread in (30b–e) where a universal quantifier precedes a modifier or determiner, to be the reason for the absence of H spread on the SVC. However, in (30f) the surface H on all the TBUs of the intervening constituent would otherwise condition rightward H spread on the SVC. An argument for the transfer of Kikongo tonal characteristics to universal quantifiers in Saramaccan is more tenuous than for non-universal quantifiers. The universal quantifiers híi and túu generally fall under the head-leftward rule. In fact, híi does not deviate from the head-leftward rule for an exception to that rule, namely H spread from determiners of nationality. In this, then, híi does achieve a lowering, like universal quantifiers in Kikongo, in (31) below. The effect of Saramaccan universal quantifiers on SVCs, however, would be the strongest claim to transfer from Kikongo.

Chapter 3. Tone on quantifiers in Saramaccan

.

Universal quantifiers in Kikongo

In Kikongo the semantics of a universal quantifier in a phrase triggers effacement of the noun, even if, as with non-universal quantifiers, the universal quantifier word itself is not adjacent to the noun. There is effacement even if the semantics of a universal quantifier phrase has no actual universal quantifier word, as in (31a–b); the associative prefix (AP) with a numeral denotes ‘all (number)’. Odden (1994: 177–8) argues that for universal quantifiers, as with non-universal quantifiers, the syntactic variation (31e–f) or even lack of universal quantifiers (31a–b) means phonological rules must reach beyond the syntactic level and have direct access to the semantic level of representation. In (31a) the semantics of ‘all’ is conveyed by the AP and a numeral, without a universal quantifier, conditioning argument effacement. In (31b) the AP and numeral are not adjacent to the noun, yet still condition argument effacement. In (31c) there is no argument effacement, but the substituting a universal quantifier in (31d) results in argument effacement. In (31e) the universal quantifier is adjacent to the noun, but in (31f) it is not, yet there is still argument effacement. The examples in (31) are from Odden (1994: 177–8): (31)

a.

b.

c.

d.

e.

f.

bikúníe byábítatú byámbóte > bikunie byábítatú byámbóte farmer .three .good argument effacement ‘all three good farmers’ bikúníe byámbóte byábítatú > bikunie byámbóte byábítatú farmer .good .three argument effacement ‘all three good farmers’ bikúní byánáni > bikúní byánáni farmer .whose no argument effacement ‘whose farmers?’ bikúníe byábínsó > bikunie byábínsó farmer .all argument effacement ‘all the farmers’ byábánduundu > bikunie byábínsó byábánduundu bikúníe byábínsó farmer .all .Banduundu argument effacement ‘all farmers of Banduundu’ bikúníe byábánduundu byábínsó > bikunie byábánduundu byábínsó farmer .Banduundu .all argument effacement ‘all farmers of Banduundu’

Unlike non-universal quantifiers in Kikongo that maximize phrasal H, universal quantifiers maximize phrasal L. But as with non-universal quantifiers in Kikongo, the phonology-semantics interface could arguably give the phrasal tonal morphology of universal quantifiers a greater salience than the local tonal morphology elsewhere, increasing the likelihood of transfer.5 If the tonal characteristics of híi and túu in . See arguments in Mufwene (1991: 133) on the salience and transferability of stressed over unstressed items.



 Marvin Kramer

Saramaccan are transferred features from Kikongo, the blocking of H spread, especially from adjectives of nationality, would be a strategy for maximum phrasal tone lowering, similar to argument effacement in Kikongo. As with non-universal quantifiers in Saramaccan, the presence of universal quantifiers in intervening constituents blocks rightward H spread rules in SVCs.

.

Transfer of serial verb constructions and quantifiers

Migge (1998) argues convincingly that the SVCs in the creoles of Suriname transferred from the Gbe languages, principally Fongbe. In Saramaccan, the H spread rules for SVCs and their intervening constituents parallel Fongbe H spread rules, also suggesting transfer from Fongbe, and these H spread rules for SVCs are different from the default H spread rules, suggesting as well that the transfer of SVCs was independent of the incorporation of the default rules (Kramer 2005). The H spread rules for SVCs override the default rules, and similarly, the H spread rules for quantifiers are different from the default rules, and override both the default rules and the SVC rules. This suggests that these rules were incorporated independently, allowing the possibility that they may have entered the language at different times.

.

The default head-leftward H spread rule

There is no known source for a transfer of the head-leftward H spread rule, leaving the possibility that it developed in Suriname. There is a possible scenario for the innovation of leftward spread in Saramaccan from rightward spread transferred from Fongbe. In Fongbe, as in Saramaccan, H spread occurs at the left edge of a verb, but is blocked at the right edge of a verb, which is termed a disjunctive border in Fongbe by Wiesemann (1991). A rightward spread in Saramaccan to this disjunctive border could be reinterpreted as leftward spread from the disjunctive border, but only when there is rightward spread from a TBU specified as H. As the head-leftward rule requires a TBU with H at the left border of the domain, the change in direction would then always include this required TBU with H. In (32a) below, rightward H spread parallels Fongbe with domain-final H at the right border, as in (23a); {é ɡbǎ xwé} > {é ɡbá xwé} ‘he built houses’. This is reinterpreted as leftward H spread in (32b). In (32c) rightward H spread parallels Fongbe domain-final L, as in (23c), or; {é sà} 3 sell {àsón wè} crab two > {é sâ} {àsón wê} ‘she sold two crabs’ (Lefebvre and Brousseau 2002: 22). This is reinterpreted as leftward H spread in (32d): (32)

a.

H H H

{womi

pεε} tutu >

b.

H H H

based on Fongbe domain-final H

{womi pεε} tutu ‘man played horn’

Chapter 3. Tone on quantifiers in Saramaccan

c.

H

H

H

{womi boo}ko tutu >

d.

H

{womi

H

H

based on Fongbe domain-final L

boo}ko tutu ‘man played horn’

Since verbs are heads, the resulting change could have generalized into the headleftward rule for leftward spread in Saramaccan. Since every word has at least one TBU specified for tone, and H spreads only to unspecified TBUs, the H spread would be limited to a single word to the left of the head, as seen in the sandhi block at the left edge of property items. For words with final TBUs with specified L, there would be no rightward H spread, and so no reinterpreted head-leftward H spread; lògòsò kulé > lògòsò kùlé ‘turtle ran’.

.

Rightward H spread on serial verb constructions

Only SVCs that meet the appropriate criteria will have rightward H spread. Where rightward H spread does not occur, normal head-leftward H spread is the default. And where rightward H spread does occur, it is in combination with aspects of leftward H spread, namely the leftward H spread of V2 and the constraints on sandhi junctures from the head-leftward H spread rule. Rightward H spread in Fongbe is constrained by disjunctive borders only. It would seem, then, that rightward rules apply to SVCs in Saramaccan where they would approximate H spread in SVCs in Fongbe. In Fongbe, adjacent serial verbs have a conjunctive border (Wiesemann 1991), or obligatory sandhi, which is the case for Saramaccan adjacent serial verbs as well. In Fongbe, non-adjacent verbs have H spread between all the words of an intervening constituent, with possible L at the domain left edge, since Fongbe has no head-leftward constraints and no TBUs with specified L. In Saramaccan SVCs with rightward H spread, the intervening constituent mimics the Fongbe H spread domain, with H on all TBUs and possible Ls at the domain edges. The verbs in these SVCs, then, mimic the Fongbe conjunctive border, the H spread between adjacent serial verbs in Fongbe with surface H on TBUs adjacent to the sandhi juncture. The separate rules for some SVCs might indicate that SVCs entered the language at a later time than the development of the default head-leftward rules. The headleftward rules probably developed at a stage of creolization when speakers of Fongbe were not dominant. If SVCs then entered the language at a later time when speakers of Fongbe were dominant, SVCs could be incorporated with their Fongbe H spread rules with little pressure to compromise, as there would be no already existing SVCs. The compromise would be with the head-leftward constraints that create TBUs with surface L, and of course with TBUs with specified L, within the intervening constituent. These TBUs with L would not match the Fongbe H spread, and so those SVCs would not mimic Fongbe SVCs. If the head-leftward rules came after the SVC rules, there would be no dominance of speakers of Fongbe to determine which SVCs match





Marvin Kramer

Fongbe SVCs. If the two rules came at the same time, Fongbe speakers would not be dominant enough to prevent the development of the head-leftward rules, and would have no motivation to preserve SVC rules, since in Fongbe SVCs have no special H spread rules. The easiest explanation for the subset of SVCs in Saramaccan that mimic the H spread rules for SVCs in Fongbe is that linguistically dominant speakers of Fongbe would acquire the existing head-leftward H spread rules while maintaining rightward H spread rules where possible as SVCs were incorporated into the language. The leftward H spread rules were likely already in place to serve as a default when conditions for rightward H spread on SVCs were not met, and to supply the leftward H spread on V2. If, on the other hand, leftward spread rules were to have replaced rightward spread rules, it would be more difficult to explain why certain SVCs were not affected. It would appear that the rightward spread rules for SVCs transferred along with SVCs during creolization, since there is no argument against the transfer of SVCs during creolization. A likely time for the development of the head-leftward rules would have been previous to 1699, when speakers of Gbe languages appear not to have been numerically dominant (Arends 1995: 243). From 1680 to 1699 speakers of Gbe languages represented roughly 40% of the people imported from Africa. But in the period from 1700 to 1719 Gbe speakers were numerically dominant at roughly 70% of imported people, a likely time for the introduction of SVCs. It might be necessary to assume, then, that the Surinamese creole predecessor to Saramaccan had tones, since Saramaccan is thought to have developed from an earlier Surinamese creole between 1690 and 1710 (Smith 2002: 142).

.

The blocking of rightward H spread rules on SVCs by quantifiers

The source of the exceptional tonal characteristics of quantifiers in Saramaccan is most likely Kikongo. These quantifier tonal characteristics would have transferred after SVCs, since it is easier to explain them as superimposed exceptions on already existing H spread rules on SVCs, rather than H spread rules on SVCs allowing exceptions for already existing quantifier rules that wouldn’t necessarily conflict with SVC rules. It is also easier to assume that SVC and quantifier rules transferred separately at times when speakers of the respective languages Fongbe and Kikongo were dominant, since the H spread rules are subtle and are not essential to communication. As mentioned, Kikongo is not a verb serializing language, so in SVCs with quantifiers speakers of Kikongo would prefer their own quantifier rules over existing SVC rules rather than attempt to use both sets of rules. At a time when speakers of Kikongo were dominant, there would likely be few speakers of Fongbe. There would be, then, little motivation to preserve rightward H spread rules on SVCs if they were being blocked by Kikongo influenced quantifier rules. If, on the other hand, SVCs were to have transferred after the quantifier rules, there would

Chapter 3. Tone on quantifiers in Saramaccan

be few speakers of Kikongo to preserve the quantifier rules at the expense of rightward H rules on SVCs. And if they entered the language at the same time, Kikongo speakers would not be dominant enough to cause blocking of rightward H spread on SVCs with quantifiers, since Fongbe of course also has quantifiers. Of the two major substrate languages, in the period from 1740 to 1803 speakers of Kikongo were the numerically dominant group of imported people, while the importation of Gbe speakers was negligible (Arends 1995: 243). This period of time is far later than the dates 1690–1710 generally considered to mark the creolization of Saramaccan (Smith 2002: 142). There is an indication, however, of Kikongo influence on Saramaccan at a late date, namely the loss of r in Saramaccan. While Fongbe has an r (Lefebvre and Brousseau 2002: 16), Kikongo has no r (Carter and Makoondekwa 1987). In the early texts and dictionaries of Saramaccan there are words with r that in the modern language have been replaced with vowel changes or l. Modern Saramaccan has no r. In the Riemer dictionary c. 1779 (Arends and Perl 1995), karo ‘corn’ is kálu in modern Saramaccan. There are several words in Riemer’s time that were in transition, such as bâli or bari ‘sweep’ from Portuguese varrer, which is now balí or baí. Another example is leti or reti ‘right’, which is léti in the modern language. There appears to have been pragmatically motivated variation with this item, as seen in the Wietz Bible of 1805 (Schuchardt 1914: 2–35). The form leti was used for ordinary items, while the form reti was used when referring to items relating to deities. In Acts 2:25, David says a de na mi leti sei ‘he is at my right hand’. Later, God says ɡo sinda na mi reti sei ‘sit at my right hand’ in Acts 2:34. When Peter cured a lame man, he reached down with his right hand, in Acts 3:7; a panja hem na hem reti mau ‘he took him by the right hand’. That right hand was no ordinary hand at this point, since Peter had just declared, in Acts 3:6, that the cure was na nem va Jesu Christo ‘in the name of Jesus Christ’. The r ~ l variation exhibits a recognition that r was associated with things external to Saramaccan maroon life, and l was associated with maroon life. This would be an important distinction for speakers of Kikongo who had escaped into Saramaccan territory and whose r-less native language would give their l-version of Saramaccan a distinct language identity.

. Conclusion If in fact the tonal characteristics of quantifiers in Saramaccan are transferred features from Kikongo, they would be the first reported influence on Saramaccan beyond lexical items from any substrate language other than Fongbe. The tonal characteristics for non-universal quantifiers involve maximizing the H in their phrases, and for universal quantifiers involve maximizing the L in their phrases. The effect of quantifiers on the H spread rules of SVCs is a development within Saramaccan, and not transfer, since, as mentioned, Kikongo is not a verb serializing language. But the effect of quantifiers on SVCs may offer a clue to the relative timing of the incorporation of SVCs and quantifiers, as may be the effect of some SVCs on the leftward default H spread rules.





Marvin Kramer

The first period of known Kikongo numerical dominance, 46.7%, over Gbe, 39.4%, was the decade 1680–1689 (Arends 1995: 243), but this dominance was slight compared to the Kikongo dominance over Gbe in the period 1740–1803 (ibid.). This period is far later than the generally accepted date for the creolization for Saramaccan, which is the period 1690–1710. Perhaps Saramaccan had gone through a period of stability, and then, with the Kikongo influence, gone through a minor recreolization. The Portuguese element in Saramaccan may have gone through creolization (Smith 1999: 278), which would have affected a version of early Sranan. It is not known if creolization had ended for early Sranan when the Portuguese element was incorporated to create Saramaccan, so that it would have been recreolization. The effect of quantifiers on H spread rules for SVCs and on the default H spread rules could likely be considered creolization, since shift with interference would entail a large population of Kikongo speakers separate from Saramaccan speakers, and would not have involved the extensive Kikongo lexical element in Saramaccan (Thomason and Kaufman 1988: 37), yet the effect would also not have been the result of borrowing, since the phonological interference was not accompanied by the borrowing of syntactic features (ibid.: 38). A case for creolization is the alteration of the original Kikongo rules so that only the effect on quantifier phrases is preserved, while the Saramaccan rules are also altered. This would be the result of negotiation, a cross-language compromise characteristic of creolization (ibid.: 154). If SVCs entered the language after the development of the default rules, then the development of the default rules occurred during creolization, since there is no argument that the incorporation of SVCs was not part of creolization. The treaty of 1762, where Saramaccans won their freedom and their territory, ended significant increases in their population from escaped slaves (Arends 1995: 253). But from 1740 to 1762 Kikongo speakers would be a major addition to Saramaccan society, and in these decades Kikongo speakers were very important in the struggle for independence (Price 1983: 135). An influential population of Kikongo speakers living with Saramaccan speakers could have brought about H spread rules that block some existing Saramaccan rules. These new rules would mimic the effect of Kikongo phrasal tone rules while changing them to local tone rules or the blocking SVC rightward H spread, the cross-language compromise. This then would be, in a small way, creolization. Perhaps recreolization is more likely in a creole language because language identity for its speakers involves awareness of its creole origin. Now the relative dominance of Kikongo speakers and the negligible importation of Gbe speakers in the latter 18th century would have affected Sranan as well as Saramaccan. In Sranan, ‘other’ is tra, and like Saramaccan óto is possibly from the Portuguese outro/a ‘other’. In any case, the Sranan etymon would appear to be the stressed or H second syllable of an earlier two-syllable word, which would not reflect the stress of the lexifier word, but could be the reflex of an earlier Kikongo quantifier rule.6 There is no .

A similar case might be made for taa ‘other’ in Ndjuka.

Chapter 3. Tone on quantifiers in Saramaccan

tra in the early text of Sranan, Herlein (1718), but it occurs as tara in Nepveu (1770), and variously as tra or tre in Van Dyk (1765) (Arends and Perl 1995). But in Herlein there is a sentence with entre translated as ‘other one’ (p. 74). Arends and Perl offer two possible interpretations; wan tra ‘one other’, and entre ‘other’ < Dutch ander(e) (ibid.: 24). Now in Van Dyk, tre only occurs before the vowel in tre estrede ‘day before yesterday’, contrasting with tra tem ‘other time’ (1965: 152), while in Herlein entre precedes ples ‘please’, contrasting with tara plessi in Nepveu (1770: 78). It would seem that entre is the word used in Herlein. The Herlein text precedes the period of dominance by Kikongo speakers, and appears to lack an example of the Portuguese etymon for ‘other’. If this word were tra, it would reduce the likelihood of any late Kikongo influence on quantifiers in the Surinamese creoles. The tonal characteristics of non-universal quantifiers, like óto, are subtle and are not necessary for communication, hardly a salient candidate for transfer. If the source of these characteristics were a tonal feature from Kikongo, it would seem to be a small and somewhat odd single feature to retain from a language that is rich in tonal morphology. But, as pointed out by Odden (1994), quantifiers in Kikongo are unique in the language. Their effect on the surface tones of nouns in quantifier phrases can be accounted for by semantics rather than syntax, since the syntax of quantifiers is variable, and even the presence of some is not necessary. For quantifiers in Kikongo, then, the phonological level must have direct access to the semantic level of representation (ibid.: 201). There is no argument that this applies to quantifiers in Saramaccan. But the special access to semantics in Kikongo gives quantifiers a special salience, since phrases rather than immediate constituents are the locus of effect. It is argued in this paper that the special salience of Kikongo quantifier phrases facilitated the transfer that now affects non-universal quantifier phrases in Saramaccan, and has come to affect phrases with universal and non-universal quantifiers in SVCs.

References Arends, J. 1993. Towards a gradualist model of creolization. In Byrne, F. and Holm, J. (eds). Atlantic meets Pacific. Amsterdam: John Benjamins. 371–80. Arends, J. 1995. Demographic factors in the formation of Sranan. In Arends, J. (ed). The early stages of creolization. Amsterdam: John Benjamins. 233–85. Arends, J. and Perl, M. 1995. Early Suriname creole texts. Frankfurt: Vervuert. Brousseau, A.M. 1991. Domaines et relations gouvernement dans le processus tonals du Fongbe. Recherches Linguistiques de Vincennes 20. 23–112. Carter, H and Makoondekwa, J. 1987. Kongo language course: maloongi makikongo. Madison: African Studies Program, University of Wisconsin, Madison. Devonish, H. 1989. Talking in tones: A study of tone in Afro-European creole languages. London: Karia Press. Good, J. 2003. Morphosyntactic tone raising in Saramaccan: The reanalysis of substrate phonology as tonal morphology. In Booij, G. and Marle, J. (eds). Yearbook of morphology 2002. Kluwer: Dordrecht. 105–34.





Marvin Kramer Good, J. 2004. Tone and accent in Saramaccan: Charting a deep split in the phonology of a language. Lingua 114. 575–619. Hayes, B. 1980. A metrical theory of stress rules. PhD thesis, MIT. Herlein, J.D. 1718. Beschrijvinge von de volksplantinge Zuriname. Leeuwarden: Injema. Kramer, M. 2005. High tone spread in Saramaccan serial verb constructions. Journal of Portuguese Linguistics 4. 31–53. Lefebvre, C. and Brousseau, A.M. 2002. A grammar of Fongbe. Mouton: Berlin. Migge, B. 1998. Substrate influence in the formation of the Surinamese plantation creole: A consideration of sociohistorical data and the linguistic data from Ndyuka and Gbe. PhD thesis, Ohio State University, Columbus. Mufwene, S.S. 1991. Pidgins, creoles, typology and markedness. In Byrne, F. and Huebner, T. (eds). Development and structures of creole languages: Essays in honor of Derek Bickerton. Amsterdam: John Benjamins. 123–43. Nepveu, J. 1770. Annotatien op de Surinaamsche Beschrijvinge van A0 1718. MS. Municipal Archives, Amsterdam, Marquette-archive, nr. 231, inv. Nr. 298 (complete version). State Archives, The Hague, collection Verspreide West-Indische Stukken, nr. 428 (incomplete version). Odden, D. 1994. Syntactic and semantic conditions in Kikongo phrasal phonology. In Cole, J. and Kisseberth, C. (eds). Perspectives in phonology. Stanford: Center for the Study of Language and Information. 167–202. Price, R. 1983. First time. The historical vision of an Afro-American people. Baltimore: Johns Hopkins University Press. Rountree, C. 1972. Saramaccan tone in relation to intonation and grammar. Lingua 29. 308–25. Schuchardt, H. 1914. Die Sprache der Saramakkaneger in Surinam. Verhandelingen der Koninklijke Akademie van Wetenschappen te Amsterdam. Afdeeling Letterkunde. Nieuwe reeks XIV.6. Amsterdam: Johannes Müller. 2–116. Smith, N. 1999. Pernambuco to Surinam 1654–1665? The Jewish slave controversy. In Huber, M. and Parkvall, M. (eds). Spreading the word [Westminster Creolistics Series 6]. London: University of Westminster Press. 251–98. Smith, N. 2001. Voodoo chile: Differential substrate effects in Saramaccan and Haitian. In Smith, N. and Veenstra, T. (eds). Creolization and contact. Amsterdam: John Benjamins. 43–80. Smith, N. 2002. The history of the Surinamese creoles II: Origin and differentiation. In Carlin, E. and Arends, J. (eds). Atlas of the languages of Suriname. Leiden: KITLV Press. 131–52. Smith, N. 2004. A case for substrate allophony 350 years on. Presented at SPCL Boston, Jan. 9. Thomason, S. and Kaufman, T. 1988. Language contact, creolization, and genetic linguistics. Berkeley: University of California Press. Van Dyk, P. 1765. Nieuwe en nooit bevoorens geziene onderwijzinge in het Bastert Engels, of Neeger Engels, zoo als hetzelve in de Hollandze colonien gebruikt word, etc. Amsterdam: De Erven de Weduwe Jacobus van Egmont. Voorhoeve, J. 1961. Le ton et la grammaire dans le Saramaccan. Word 17. 146–63. Wiesemann, U. 1991. Tone and intonational features in Fon. Linguistique Africaine 7. 65–90.

 

Morphophonological properties of pitch accents in Jamaican Creole reduplication Shelome Gooden University of Pittsburgh

Several researchers have discussed prosodic differences between segmentally identical reduplications in Creole languages. I show here that in Jamaican Creole, this is a stress difference. Stressed syllables in Jamaican Creole are realized phonetically with a fall in fundamental frequency (F0) onto that syllable. I focus on two types of adjectival reduplication, intensive and distributive, which produce segmentally identical outputs. I examine F0 traces of the words in three different prosodic contexts and show that while intensive reduplications are doubly accented, distributives are accented only once. These differences are a function of differences in their metrical structure that translates into a distinction between word-level and phrase level stress. The cumulative effect is an analysis which incorporates both the accentual and segmental properties of reduplicated words.

.

Introduction

Reduplication, a common morphological process in Creole languages, can create segmentally identical words. In these cases we can see pitch differences (often interpreted as stress or lexical tone), that correlate with differences in meaning. This is reported for Guyanese Creole (Devonish 2003), Sranan (Adamson and Smith 2003; Dioncie 1959 in Sebba 1981), Krio (Nylander 2003), Haitian (Sylvain 1936), Mauritian (Baker 2003) and Jamaican Creole (Gooden 2003a, b). However, since the prosodic typology of these Creole languages is still being debated, the precise nature of these prosodic differences or how they interact with reduplication processes has previously been unreported. At the same time, the potential to explore these prosodic distinctions has been strengthened by the increasing number of discussions on prosodic typology/description of creole languages (cf. Carter 1987; Devonish 1989, 2002; Devonish and Murray 1995; Finney 2004; Good 2004; Gooden 2005; Remijsen and van Heuven 2005; Rivera-Castillo 1998, among others). See also earlier descriptions (Lawton 1963; Rountree 1972; Voorhoeve 1961). This paper is somewhat of a fusion of two different aspects of my research on Jamaican Creole (JC) reduplication that separately examine some of these issues.



Shelome Gooden

Gooden (2003a) provided an Optimality Theory (OT) (McCarthy and Prince 1993, 1995a; Prince and Smolensky 1993) analysis of the prosodic requirements on the size and shape of the copied material (reduplicant). The analysis showed that JC reduplication is a systematic and phonologically predictable process that copies a trochaic foot. Gooden (2003b) described preliminaries of the fundamental frequency (pitch) differences between two segmentally identical types, distributive and intensive reduplication. The hypothesis for two prominent syllables in intensive reduplications and one prominent syllable in distributives is reexamined here. I propose that the prominence relationships must be interpreted as an accentual stress contrast in the intonational phonology of the words. The paper gives an account for the JC reduplication facts in OT. I show that both the conditions on segmental well-formedness and accent placement must be satisfied, otherwise the correct surface form of the words cannot be predicted. The paper is organized as follows. Section 2 is an overview of JC prosody including an overview of word-level stress with a brief discussion of stress in one phrase type. Section 3 gives details of adjectival reduplications and highlights the similarities between intensive and distributive reduplication. Section 4 is the phonetic analysis of intensive and distributive reduplications. This is restricted to two prosodic contexts, citation form and non-final position in a statement. The formal phonological analysis of reduplication is given in Section 5 and the conclusion in Section 6.

.

Background on Jamaican Creole prosody

Although Lawton (1963) and Sutcliffe (1986) discussed Jamaican Creole (JC) in ways that have been interpreted as JC having lexical tone, Gooden (2003c) shows that the contrasts in the lexicon and the phrasal phonology, involve intonational morphemes that depend on discourse function. These intonational morphemes are linked to stressed syllables in the same way that “pitch accents” (see below) are linked to the stressed syllables of words under focal prominence in many dialects of English and Dutch (cf. Beckman 1986). The analysis of stress contrast in reduplications follows the Autosegmental-Metrical (AM)1 account of F0 and its relationship to prosodic organization (cf. Beckman and Pierrehumbert 1986). The AM approach has been widely adopted in the literature on intonation and phrasing in a variety of languages, including languages which might be characterized in some typologies as pitch accent languages (e.g., Japanese), tone languages (e.g., Mandarin) and stress languages. Crucial to the understanding of this framework, is the way in which it distinguishes between autosegmental content features and the metrical structures that license them. In particular, the metrical part of the framework refers to the intersecting rhythms or hierarchically

.

The name is due to Ladd (1996).

Chapter 4. Pitch accents in Jamaican Creole reduplication

organized prosodic constituents (syllable, foot etc. Selkirk 1984) projected from the content features (distinctive features etc.) that are licensed at different positions. That is, the prosody of an utterance is viewed in terms of the layered phonological constituents and is therefore a raw organizational structure. Crucially prosody is not synonymous with the SPE notion of a suprasegment. Unlike distinctive features, prosody in the AM approach refers to organization within the metrical grid (see Beckman 1996). The autosegmental part of the framework refers to the specification of content features that are autonomously segmented. Another aspect of the AM framework has to do with the relationship between an autosegment and the metrical position that licenses it, i.e., association. Differences in the association of the autosegment to the metrical structure yields differences in the interpretation of the resulting structure. One such autosegment is tone. By tone I mean the local specification of F0 features independent of the segmental material with which they are aligned. In the AM approach, tones are treated similarly regardless of their function in individual languages. They differ from other types of autosegments in typically associating to a wide range of metrical positions. For example, a pitch accent is a tone (or closely linked group of tones) whose distribution is defined with respect to a prosodic prominence (Beckman 1996). Pitch accents are typically associated with the prominent (metrically strong) syllable or the mora constituent. Boundary tones are associated with higher level prosodic constituents at the edge of phrases like the intonational phrase. Tone specifications are inherently arranged in a scale from low (L) to high (H) corresponding to target F0 values speakers produce at different prosodic positions in an utterance. Viewing these as target F0 values (level tones H and L) ensures that linguistically unimportant variations in pitch range (such as by segmental perturbation) are not represented in the phonological system. Formally, the pitch accents associated with stressed syllables are the ‘starred’ (*) tones and can be monotonal (e.g., H*) or bitonal (e.g., H+L*). The asterisk (*) marks the specific part of the tone associated with the stressed syllable. The low F0 at the end of statements is marked with L% and the high F0 at the end of questions is marked with H%. From this perspective, we can view different levels of stress, i.e., primary, secondary etc. as differentials in the levels of prominence contrast for these prosodic constituents. Potential ambiguities such as that observed for intensive and distributive reduplication can therefore be resolved by reference to the prosodic organization of prominent syllables in the word and implies that the alignment of the associated F0 will necessarily differ.

.

Stress in Jamaican Creole

Stress in JC is distinctive contra Lawton (1963, 1968) and Akers (1981). This supports previous analyses of JC stress, (Alderete 1993; Gooden 2003c; Wells 1973). Phonetically, syllables bearing main stress have a fall in F0 onto that syllable and unstressed vowels may be deleted. The data, primarily from the Dictionary of Jamaican English (DJE)





Shelome Gooden

(Cassidy and LePage 1967), reflects stress patterns on monomorphemic content words from all lexical categories, compounds and reduplicated words. Additional data (especially reduplication data) are from my own fieldwork with native speakers. The DJE was the main source for primary stress assignment in the words, supplemented by native speaker intuitions on syllable prominence. In cases where they did not match, I relied on the native speakers’ judgment. JC has a weight-sensitive stress system (cf. Alderete 1993; Gooden 2003c) a common property of stress systems cross-linguistically (Hayes 1995). Coda consonants, long vowels and diphthongs contribute to syllable weight. The preferred foot type in JC stress assignment is a trochaic foot, which is constructed iteratively from left-to-right. The location of stress is phonologically predictable. Primary stress generally falls on the leftmost or only heavy syllable in a word and secondary stress is two syllables away from the main stress as shown in (1). In this respect (at least), the stress pattern in JC somewhat resembles that of other Caribbean Creoles, for example Saramaccan (Good 2004: 589–90) and Curaçao Papiamentu (Remijsen and van Heuven 2005). In these languages secondary stress is predictable given the position of primary stress.2 (1)

pa.'laav 'pai.zn 'aa.ln n.'lk.ful g.'laan.tin 'niiz.b.r 'dm.b.'ln 'ka.kan.'tar 'ba.da.'rie.∫an 'po.ko.'mie.nja

‘to lie about idly’ ‘poison’ ‘egret’ ‘neglectful’ ‘to fool around’ ‘type of fruit’ ‘type of tree’ ‘type of top’ ‘thing or person that is a source of annoyance’ ‘type of revivalist cult’

If there are no heavy syllables, then the penultimate syllable gets primary stress. There is no stress on a word final light syllable. (2)

'ma.ka su.'ma.d

‘thorn’ ‘somebody’

This generalization is consistent across bisyllabic, trisyllabic and quadrasyllabic words. A small percentage of the data with alternate stress patterns are not discussed here (see Gooden 2003c). All the compounds in the sample were nominal compounds up to five syllables long. Compounding in JC involves the concatenation of nouns and form a single prosodic word. One motivation for assigning this status is that they receive just one primary stress and in this way are similar to non-compound words which also have only one

.

Stress is marked with standard IPA notation.

Chapter 4. Pitch accents in Jamaican Creole reduplication

primary stress. Nespor and Ralli (1993) used a similar criterion in their description of Greek compounds. The stress pattern in compounds depends on the stress pattern of the input words. Primary stress on the first member of the compound is realized as secondary stress in the compound word. In addition, the rightmost member of all the compounds has main stress. For example, ('haas ‘horse’ + 'baat ‘bath’) 'haas'baat ‘weed used to scrub horses’; ('bl ‘belly’ + 'ad ‘god’) 'bl'ad ‘lutton’. These provide an important point of comparison with reduplicated words which show similar stress patterns.

. Stress in reduplication The stress pattern on distributive reduplications mimic the pattern in compound words in that they bear a single primary stress, i.e., word level stress. Similar reports have been made for other English-based creoles like Krio (Nylander 2003) and Guyanese (Devonish 2003). In these two varieties, as in JC, reduplication exhibits similar prosodic properties to compounding. Intensive reduplications have a different stress pattern from compounds and distributives and are able to bear two main stresses. I will argue that these instead form a phonological phrase and thus get phrasal stress. The metrical grids in (3) compare distributive reduplication with a quadrasyllabic monomorphemic word, and a quadrasyllabic compound. Distributive reduplications, like quadrasyllabic monomorphemic words and compounds, have only one accentable syllable as marked in bold. This is because they each have a single primary stress and only stressed syllables may be accented. That is, prominence at higher levels in the metrical grid is precluded by prominence at lower levels. (3) Distributive

Quadrasyllabic Word

Quadrasyllabic Compound

x x x x x x rinrin

PPh x PPh Prwd accented x Prwd accented Foot stressed x x Foot stressed Syllable x x x x Syllable ‘green all over’ j lo j lo ‘yellow all over’ x PPh x Prwd accented x x Foot stressed x x x x Syllable a l ε ta ‘alligator’ x PPh x Prwd accented x x Foot stressed x x x x Syllable bε l u man ‘pregnant woman’

When compared to the patterns for the intensive reduplications in (4), we see that both distributives and intensives have identical stress patterns at the level of the foot. However, since intensives have two metrically strong syllables at the level of the prosodic





Shelome Gooden

word, they are eligible to have two pitch accents. This stress pattern is identical to the phrase level stress pattern in the narrow focus phrase in (5). (4) Intensive

(5)

x PPh x x Prwd accented x x Foot stressed x x Syllable rinrin ‘very green’

x x x Narrow Focus x x Phrase x x x x j lo maɡo

x PPh x x Prwd accented x x Foot stressed x x x x Syllable j lo j lo ‘very yellow’

PPh Prwd accented Foot stressed Syllable ‘yellow mango’

This distinction in the potential to bear two versus one pitch accent is somewhat reminiscent of the distinction in English between compound words like paper clip or blackbird which are both accented on the first foot, and sequences of two words like paper tiger and black bird which have two accents, one on each foot. The former has been referred to in the literature as a compound and the latter as a phrasal compound (cf. Nespor and Vogel 1986; Vogel and Raimy 2002).

.

Jamaican Creole adjectival reduplication

Jamaican Creole has a reasonably rich morphological subsystem in which there are several productive processes of reduplication operative in all major word classes. The processes yield four different semantic types. In distributive reduplication, the meaning of the reduplicated word conveys the sense of ‘scattered’, ‘all over the place’, ‘here and there’ or ‘occasionally’ (Anagbogu 1995). In iterative reduplication, the meaning of the reduplicated word describes an action performed repeatedly while in intensive reduplication the degree of quantity (noun), the degree of action (verb) or the degree of quality (adjective) has more intensity or emphasis in the reduplicated word than in the unreduplicated word (cf. Moravcsik 1978). The fourth type is characteristic reduplication where the meaning expressed by the reduplicated form incorporates the feature X, where X is a semantic component of the base. For example, the word dk ‘to pierce’ reduplicates as in (6). (6)

a. b.

dkdk dkdk

c. d.

dkdk dkopdkop3

‘prickly object’ (Nominal) ‘characteristically prickly all over/ pierced all over’ (Characteristic/Distributive) ‘very prickly’ (Intensive) ‘to prick repeatedly’ (Iterative)

. This can occur in both predicative and attributive positions. In predicative position it indicates a process and in attributive position, it indicates a result of an action. There appears to be no prosodic differences between the two.

Chapter 4. Pitch accents in Jamaican Creole reduplication

Interestingly, two of these reduplication processes, distributives (6b) and intensives (6c), yield segmentally identical forms. These two types of reduplication are the focus of the discussion in this paper.

.

Contrasting reduplications

There are two possible output forms for distributive reduplication, either a quadrasyllabic form as seen in (7a) and (7c) or a bisyllabic form as seen in (7b). Notice though that the reduplicated words in (7c) do not precisely match the input word. They have an epenthetic vowel, [], which has been shown to be the default epenthetic vowel in JC (Gooden 2003a), inserted to fulfil a bisyllabic foot requirement. (7)

Distributive Reduplication a. n ‘grainy’ nn jlo ‘yellow’ jlojlo pisa ‘spliced’ pisapisa b. blak ‘black’ blakblak rin ‘green’ rinrin c. blak ‘black’ blakblak rin ‘green’ rinrin swit ‘sweet’ switswitI

‘grainy all over’ ‘yellowish all over’ ‘in pieces’ ‘black spots all over’ ‘green spots all over/greenish all over’ ‘black spots all over’ ‘green spots all over’ ‘sweet ingredients all over’

The examples in (8) show that intensive reduplications can also be bisyllabic as in (8a) or quadrasyllabic as in (8b). The difference here is that the input to reduplicated words in (8b) are already bisyllabic. (8)

.

Intensive Reduplication a. blak ‘black’ la ‘long’ lë ‘low’ rin ‘green’ b. maaa ‘skinny’ rεn ‘grainy’ om ‘gummy’ jlo ‘yellow’

blakblak lala lëlë rinrin maaamaaa rεnrεn omom jεlojεlo

‘very black’ ‘very long’ ‘very low’ ‘very green’ ‘very skinny’ ‘very grainy’ ‘very gummy’ ‘very yellow’

Previous accounts

Previous phonological analyses of JC reduplication include a segmental analysis and a syllable-based account. McCarthy (1982) proposed a segmental analysis of JC reduplication with the template in (9) defining the base for reduplication. (9)

Onset V (C(V))

where ‘onset’ is a possible syllable onset in JC

This template is based on data from DeCamp (1974) and so does not cover the range of data discussed here. As such, there are several aspects of JC reduplication that are





Shelome Gooden

not adequately covered under this approach. For example, the template incorrectly predicts that coda consonants are prohibited in reduplicants. Also, the template predicts that reduplicants have no long vowels or diphthongs. To the contrary, the examples seen above show input forms with long vowels e.g., maaa ‘skinny’ and diphthongs e.g., rεn ‘grainy’ and filled codas e.g., la ‘long’. Apart from the empirical issues, a segment based account misses the generalization that in all cases the JC reduplicant is a prosodic unit, a foot. Gooden (2003c) showed that the fact that some reduplicants are a bisyllabic foot is predicted by the same constraints that predict that other reduplicants are variably a bimoraic or a bisyllabic foot. A segment-based account would require different templates to account for these differences and thus provides a less economical account of the facts. Alderete (1993) proposed a syllable-based account of JC reduplication within a prosodic morphology framework. That analysis is also based on data from DeCamp (1974) and so is similarly restricted in terms of data coverage. Alderete proposed that the base for JC reduplication is a minimal prosodic word with at least one foot, a moraic trochee. The base is analysed as a foot comprised of two light syllables (LL) or a single heavy (H) syllable. The approach assumes that in HL cases, the final L syllable is unfooted. Alderete’s analysis is similar to the one presented here in that phonological conditions on the process are delimited in terms of authentic units of prosody following the assumptions of prosodic morphology (McCarthy and Prince 1995a). However, evidence that the moraic trochee analysis is not satisfactory can be taken from the fact that the portion of the word corresponding to the reduplicant is not necessarily a moraic trochee. For example, there are LL reduplicants as in jεloRjεloB ‘very yellow’; HL reduplicants as in rεnRrεnB ‘very grainy’; H reduplicants as in blakRblakBop ‘to make very black’; but there are also LH reduplicants as in blakopRblakopB ‘to blacken repeatedly’. Alderete’s analysis cannot explain why the LH reduplicant is permissible in words like blakopRblakopBop ‘to blacken repeatedly’ but not in words like blakRblakBop ‘to make very black’. This is directly related to the theoretical treatment of the base. In this approach, the base is not necessarily identical to the input form since it is the portion of the input from which segmental material is copied. As I discuss in detail elsewhere (Gooden 2003a, c), this conceptualization of the base is vital in accounting for differences between iterative reduplications whose base is a stem equivalent to the input and intensive reduplications (complex input) whose base is a root (i.e., the stem minus the affix of the input) as in blakRblakBop ‘to make very black’. From a theoretical perspective, both Alderete’s analysis and my own analysis specify conditions on the size and shape of the input in JC reduplication whereas this analysis specifies conditions on the reduplicant and the base. This is a reflection of the theoretical approach taken here where both the base and the reduplicant must satisfy phonological as well as morphological requirements (McCarthy and Prince 1995b). As I will show in this paper, these phonological conditions extend to structure above the word. In this way the analysis is significantly different, since neither of the previous approaches considered the role of

Chapter 4. Pitch accents in Jamaican Creole reduplication

JC prosody in disambiguating segmentally identical reduplicated words like the intensive (simple stem input) and distributive reduplications discussed here. The next section discusses how stress is related to the location of pitch accents in reduplicated words. While words must bear stress, not every word will have a pitch accent since this implies prominence at multiple levels in the metrical grid.

. Pitch accents in Jamaican Creole reduplicated words I use the fundamental frequency contours as a phonetic representation of tone features associated with each type of reduplication. Figure (1) is a graph plot of F0 over time. The graphs discussed here were made using P 4.026 (www.praat.org) and are annotated as illustrated below. He wants an undershirt and some yams

F0 (Hz)

300 H* H+L* 200

H+L*

L%

100 im 0.6

want wan 0.8

1

mirina

an som

1.2 1.4 1.6 time (seconds)

yam 1.8

2

Figure 1. Annotation conventions illustrated with the utterance ‘He wants an undershirt and some yams’ (female speaker)

In each graph, the F0 trace is marked off to show what portion of the utterance it corresponds with. The associated tones are overlayed on the F0 trace, at the point where the accented syllable is located for pitch accents, and at the edge of the intonational phrase for boundary tones. The bottom row has broad phonemic transcriptions in a modified version of the Cassidy and LePage (1967) writing system. The absolute F0 value may differ across utterances and is natural, so the interpretation of highs and lows is relative to the highs and lows elsewhere in the utterance and to the speaker’s pitch range. The F0 contours discussed below are representative samples of recorded utterances from fieldwork with 15 native speakers including myself. The speakers are from a rural farming community (Top Alston) in the parish of Clarendon and ranged in age from 29 to 87 at the time of the recordings. All the speakers use a conservative variety of JC





Shelome Gooden

and had a clear perception of the contrast between intensive and distributive reduplications. The target words were elicited in carrier sentences such that they were direct objects of main verbs or copula constructions. The contexts facilitated both a distributive reading and an intensive reading of the target word to avoid semantic bias, and were produced in the five different prosodic conditions as shown in (10). The utility of this methodology was first highlighted by Bruce (1977) who showed that by comparing F0 contours of contrasting words in different sentential positions, the individual F0 contributions of final contours, sentence accent and word accent could be isolated. (10)

Prosodic Condition (a) Isolation/Citation (b) Final (nuclear) (d) Question-final (nuclear)

Pragmatic Context (c) Non-final (prenuclear)

Statement

(e) Non-final

Yes-No Question (prenuclear)

I will focus on the accent pattern on the reduplicated words in citation form and statement final position, i.e., nuclear position, since the differences between intensive and distributive reduplications are most robust in these contexts. Furthermore, the types of pitch accent that are realized on the words in other prosodic contexts correspond to differences in the shape of the F0 contours associated with the contexts themselves. The H+L* pitch accent is not specific to nuclear position as it appears in prenuclear positions as well (see Gooden 2003c).

.

Pitch accent in distributive reduplication

The F0 patterns in monosyllabic and bisyllabic based words are identical. In Figure (2), the fall from a high F0 on the first syllable onto the second syllable of the bisyllabic word is analogous to the fall from the second syllable to the last two syllables in the quadrasyllabic word. In both cases the pitch accent is associated with main stress.

Figure 2. F0 contour of the reduplicated words rinrin ‘scattered green’ and jεlojεlo ‘scattered yellow’ produced in citation form (female speaker)

As noted above, syllables with main stress in unreduplicated words and compounds also show a fall in F0. The pitch accent in all these case is analysed as a H+L* pitch

Chapter 4. Pitch accents in Jamaican Creole reduplication

accent to capture the fact that there is a low tonal target on the stressed syllable of the word. The preceding high F0 is not associated with the stressed syllable. Figure (3) shows the words in final position in a statement. (a) He wants one that is scattered green

(b) He wants one that is scattered yellow 250

200

H*

H+L*

H+L*

150 100 0.2

im 0.4

want

0.8

1.0

1.2

1.4

200

H*

H+L*

H+L*

150 100 im

griingriin

wan we

0.6

L%

F0 (Hz)

F0 (Hz)

250

0.4

1.6

waant 0.6

wan we

0.8

1.0

1.2

yeloyelo 1.4

1.6

1.8

(c) The one Mary wants is scattered yellow 250 F0 (Hz)

200 H*

150

H+L* H+L* L%

100 50 di wan mieri want

yeloyelo

0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 Time (second)

Figure 3. F0 contour of the reduplicated words produced in statement intonation in final position in the sentence (female speaker (a) and (b) and male speaker (c))

Figure 4 shows a quadrasyllabic compound. Compounds have a similar stress pattern to distributive reduplicated words and thus have a similar F0 pattern in sentence final position, as shown in Figure (6). He wants a pregnant woman

F0 (Hz)

300 H+L*

200

H+L* L%

100 im

want

0.4

0.6

wan

beliuman

0.8 1 1.2 time (seconds)

1.4

1.6

Figure 4. F0 contour of bεlman ‘pregnant woman’ produced in sentence final position (female speaker)



Shelome Gooden

.

Pitch accent in intensive reduplication

In intensive reduplications (Figure 5), there is an initial rise to a high on the first stressed syllable, which is maintained into the second syllable, then the pitch falls to a low on the stressed second syllable. The two F0 peaks are thus associated with stressed syllables. I analyse the F0 peak on the initial syllable as a prenuclear rise to a H* pitch accent. I refer to the accent as prenuclear since it appears before the main pitch accent (H+L*) in the word. Notice that both the low and the preceding high of the H+L* are located within the second syllable whereas in distributives this fall appears much later in the word.

Figure 5. F0 contour of the reduplicated words  ‘very green’ and jεlojεlo ‘very yellow’ in citation (male speaker)

A similar F0 pattern is seen in both words in sentence final position as well (Figure 6). (a) The one Mary wants is very green 200 150

H* H*

H+L*

H+L*L%

100 50

(b) The one that Mary wants is very yellow 200 F0 (Hz)

F0 (Hz)



150

H*

H+L*

H* H+L* L%

100

di wan

mieri

want

griingriin

3.2 3.4 3.6 3.8 4 4.2 4.4 4.6 4.8 5 Time (seconds)

50

di wan we

mieri

want

jεlojεlo

3.2 3.4 3.6 3.8 4 4.2 4.4 4.6 4.8 Time (seconds)

Figure 6. F0 contour of the reduplicated words  ‘very green’ and jεlojεlo ‘very yellow’ in sentence final position (male speaker)

A comparison between intensives and narrow focus phrases is instructive as they both have phrasal stress. In these examples, narrow focus is on the adjective modifier in the phrase. This is by no means an exhaustive analysis of phrase types in JC but is meant to demonstrate that intensive reduplications pattern with a phrase with two

Chapter 4. Pitch accents in Jamaican Creole reduplication

accented syllables. In Figure (7) a similar prenuclear rise (H*) can be seen on the initial foot of the phrase, and a late fall on the second foot (H+L*).4 yellow (not red) mango

F0 (Hz)

300

H*

200

H+L*

100 yelo mango 0.2

0.4

0.6 0.8 time (seconds)

1

Figure 7. F0 contour for the phrase jεlomago ‘yellow mango’ in citation

To summarize, the prosodic difference between bisyllabic base intensive and distributive reduplicated words lies in the distribution of accented syllables in the words. Intensives are analysed as having two pitch accents, the first of which is a prenuclear rise to a high tonal target (H*) and the second is a low tonal target preceded by a high (H+L*). Unlike intensives, distributives have only H+L* as there is a single stressed syllable onto which the pitch falls. Since both intensive and distributive reduplicated words have a low F0 target on the head foot of the word, in this respect, they both pattern like compound words; having right-headed prominences. However, only distributive reduplications and compound words have one accented syllable. In neither case do we see the prenuclear accent that is observable in words with an intensive meaning and the narrow focus phrase. In the following section, I present the constraint-based analysis.

.

Constraint based analysis

To understand how the accentual constraints interact with the rest of the phonology and morphology we need an accurate formulation of the accentual constraints themselves and of the phonological representations to which they apply. Section 3.1 reviewed the prosodic structure of reduplicated words which served as the basis for interpreting the associated F0 described in Section 4. Against this background, I propose two types of constraints; constraints on segmental material and constraints on accent placement. . In the comparable phrase produced with broad focus, there is only a single H+L* pitch accent on the second foot in the phrase.





Shelome Gooden

.

Constraints on segmental material

A core assumption of OT is that phonological surface forms optimally satisfy a ranked series of constraints. The underlying form serves as the structure that is maximally preserved, but is also subject to conditioning by a set of universal constraints. Since faithfulness constraints typically interact with phonological markedness constraints, surface forms are not always identical to underlying forms. This is seen wherever a phonological markedness constraint is in conflict with a lower-ranking faithfulness constraint (Kager 1999, McCarthy 2002). In cases of full reduplication like those seen here, there must be complete segmental similarity between the base and the reduplicant. In OT this is accounted for by means of Identity constraints as in (11). (11)

Maximize Base-Reduplicant (MBR): Every element of the base has a correspondent in the reduplicant (McCarthy and Prince 1995b); No deletion.

This formalizes the observation that the reduplicant (the copied material) copies the entire base (the material available for copy). The inherent claim of this model is that the reduplicant copies the base in order to achieve complete identity with it, as is also assumed by earlier Copy and Association models of reduplication (cf. Bell 1983; Broselow and McCarthy 1983; Marantz 1982, among others). As noted above, complete identity in JC reduplication is not always attainable under these terms, since the process is regulated by prosodic requirements on the phonological shape and size of the reduplicant. In fact, when these requirements are not met, there is no reduplication (Gooden 2003a, c). I have assumed that the reduplicant is prefixed to the base. A given form incurs a violation of MBR if a segment in the base is not also in the reduplicant5 as in (12b) and (12c). (12)

 ‘truly dead’ / R + ε /

MAXBR


a. εRεB b.

deRεB

*!

c.

dRεB

*!*

The constraint R=F[] in (13) captures the observation that the JC reduplicant is a trochee. These can be comprised of two syllables as in ri.n.ri.n ‘scattered green/ green all over’ or can be a single heavy syllable as in ded.ded ‘truly dead’. (13)

R=F[]: the reduplicant is a trochaic foot6

The optimal form in (14) is selected because the reduplicant copies the whole base and is a trochee. Although candidate (b) has a heavy reduplicant and satisfies .

The reduplicant and base are marked in tableaux by subscripts.

.

Feet are demarcated by parentheses ( ).

Chapter 4. Pitch accents in Jamaican Creole reduplication

R=F[] it does not copy all of the base, and violates MBR which rules it out. There is no crucial ranking of the constraints. This constraint is never violated as all surface reduplicants are trochaic feet. In the analyses below, I will assume that all surface forms satisfy the foot requirement, so the constraint is excluded from the tableaux. (14)

 ‘grainy all over’ /Red[] + /


MBR Red=Ft[]

a. ()RB b. ()R B

*!

In cases like (14) where there is no epenthetic vowel, any restriction on the size and shape of distributives will also hold for the intensive reduplicant as shown in (15). (15)

 ‘very grainy’ /Red[] + /


MBR Red=Ft[troch]

a. ()RB b. ()R B

*!

Notice though that these constraints by themselves are unable to distinguish between the word with the distributive meaning and the one with an intensive meaning. I will discuss this issue further below. Different constraints are needed to account for words like rin → rinrin ‘scattered green/green all over’ where an epenthetic vowel, [] is present both in the reduplicant and the base even though it is not seen in the input form. The epenthetic vowel is only present in distributive reduplicated forms and augments a monosyllabic input to bisyllabic status (cf. Gooden 2003a). The groups of constraints in (16) simply formalize this observation. Morpheme-specific alignment constraints like Align Base are often made explicit in the literature (cf. Gussenhoven 2000) and are given the function of linearizing affixes relative to their base as is done here. The constraints penalize insertion in the reduplicant (DBR) and the output (DIO). (16)

a. b. c.

Align Base Right, [] (AB[][]): align the right edge the base of a distributive reduplicant with the vowel [].7 DBR: every segment in the reduplicant has a correspondent in the base (McCarthy and Prince 1995a). DIO: every segment in the output has a correspondent in the input (ibid.).

. A constraint such as Dep-Vowel could also capture this observation, explicitly prohibiting vowel insertion.





Shelome Gooden

In (17), candidate (a) is chosen as optimal although it violates DIO since it satisfies the highly ranked Max constraint. Candidate (b) is eliminated since it violates MBR. This establishes the crucial ranking of MBR over DEPIO. Candidate (c) violates DBR since it inserts a vowel in the reduplicant. Candidate (d) fails on DIO but is a poor candidate anyway with several violations of DBR and the AB constraint. (17)

 ‘green all over’ /R[] + /


MBR

a. ()RB b. ()RB

DIO

DBR

AB[][]

* *!

c. ()R B d. ()R B

*!

*!

*

*

***

A tableau (18) shows that distributive reduplicants must remain faithful to their bases, irrespective of the number of syllables in the base. Since the distributive reduplicant is not obligatorily a bisyllabic foot, this supports the low ranking of AB. Further, since there is no crucial ranking between AB and DIO, both of the possible output candidates (a) with the epenthetic vowel and (b) which is faithful to the input, are correctly predicted. Candidate (c) fails on MBR though it too is faithful to the input. Candidate (d) satisfies AB by vowel insertion but the reduplicant does copy all of the segments in the base, which rules it out. Candidate (e) also remains faithful to the input but inserts a segment in the reduplicant which creates a lack of identity with the base. (18)

 ‘black all over’/  ‘black all over’ /R[] + /

.




a. ()RB




b. ()RB

MBR

DIO DBR AB[][] * *

c. ()RB

*!

d. ()RB

*!

* *

Constraints on accent placement

As discussed above, the prosodic structure is important to understanding the intonational phonology of reduplicated words. This is what cues the meaning of the words for speakers. Distributives get default word level accent based on their stress pattern. As noted above, the analysis of intensives cannot rely on constraints on segmental well-formedness alone since these cannot predict the intensive reading. When distributive reduplications occur in nuclear position they have a H+L* pitch accent on the head foot. Intensive reduplications, have a H+L* pitch accent on the head foot but also have

Chapter 4. Pitch accents in Jamaican Creole reduplication

a H* pitch accent on the preceding foot. Without these pitch accents, monosyllabic base intensives would be indistinguishable from distributives, when they appear in the same sentential context. The constraint in (19) is a short hand way of formalizing conditions (19a and b) that together reflect the observation that the head foot in a reduplicated word has a H+L* pitch accent and is always the rightmost foot. (19)

A: A H+L* pitch accent is anchored to the prosodic head at the right edge of a prosodic word. (AHL(P)) a. H+L* is associated with the head foot of the word. b. The head foot is rightmost in reduplicated words.

This A constraint is partially reflective of a general constraint in the language and captures the observation that prominent syllables in lexical words have a HL melody in nuclear position. In this way, distributive reduplication behaves like single prosodic words bearing a single nuclear pitch accent. In tableau (20), as in subsequent ones, prosodic word constituency is indicated by square brackets [ ]. The optimal candidate in tableau (20) has a H+L* pitch accent properly associated with the right edge of the prosodic word. If the pitch accent is misaligned, the candidate will fatally violate AHL. (20)

 ‘green all over’ /R[] + /




AHL

H+L*  a. [()R (B)] H+L*  b. [()R (B)]

*!

When AHL is considered with the constraint ranking established so far, the crucial ranking of AHL above MaxBR is evident. This ranking rules out candidate (21b) with the misaligned pitch accent on the initial syllable. (21)

 ‘green all over’ /R[] + /




AHL

MBR

H+L*  a. [()R (B)] H+L*  b. [()R (B)] H+L*  c. [()R (B)]

DIO *

*!

*

*!

*





Shelome Gooden

Prominent syllables in words in nuclear position bear a pitch accent, so it is important to rule out words that would surface without an accent. The established constraint hierarchy cannot rule out the candidate in (22c) as a surface form even though it does not bear a pitch accent. (22)

 ‘green all over’ /R[] + /




H+L*  a. [()R (B)] H+L*  b. [()R (B)



AHL MBR

DIO *

*!

* *

c. [()R (B)]

The inclusion of the constraint in (23) requires that every prosodic word bears primary stress and would rule out forms like (22c). (23)

Obligatory Head (OHD)8: every lexical word has at least one syllable marked for the highest degree of metrical prominence (primary stress) (Hyman 2006).

The tableau in (24) shows the crucial ranking of OHD above DIO. With this, the ill-formed candidate (22c) is now ruled out since it incurs a fatal violation of OHD. (24)

 ‘green all over’ /R[] + /




AHL MBR

OHD

H+L*  a. [()R (B)] H+L*  b. [()R (B) c. [()R (B)]

DIO *

*!

* *!

*

To summarize, the discussion so far, the data shows that the distributive reduplicant is exactly one foot in size. It copies the base completely but must have a single H+L* pitch accent on the head-foot of the word. The hierarchy of constraints for distributives is given in (25). . Obligatory Head is necessarily parasitic on constraints which require syllables to be stressed.

Chapter 4. Pitch accents in Jamaican Creole reduplication

(25)

Constraint hierarchy for Distributive Reduplication AHL(P) MB, OHD DIO, DBR, AB[]

This constraint hierarchy does not reflect the occurrence of the second accent in intensive reduplications. Tableau (26) shows that without a constraint referring to the prenuclear accent we cannot correctly predict the surface prosody of intensive reduplications. (26)

 ‘very green’ /R[] + /

AHL MB OHD

H+L*  a. [()R (B)] H+L*  b. [()R (B)]

*!

*!

H+L*  

c. [()R (B)]

Candidate (b) incurs a violation of AHL because the H+L* pitch accent is on the leftmost foot instead of the rightmost one. Although candidate (c) was selected as the output form, it is not the correct surface form. The actual output has a H* pitch accent on the initial foot of intensive reduplications. The constraint given in (27) formalizes this. (27)

A (H* L): A H* pitch accent is anchored to the left edge of a phonological phrase, (AH(L, P)).

This constraint defifnes the domain of this accent as the phonological phrase. The appearance of the pitch accent in that location is predicated on there being a prominent syllable at the level of the prosodic word. The analysis in terms of the phonological phrase accounts for why the leftmost prosodic word in intensive reduplications does not get a nuclear accent. I suggest that this position is reserved for the first accent in the phonological phrase of this type; a prenuclear accent. In addition, since the presence of the additional grid mark at the level of the prosodic word licenses the presence of the prenuclear accent in intensive words, the accent cannot appear in distributives given that they have a different metrical structure.





Shelome Gooden

Tableau (28) shows the evaluation of a monosyllabic-based intensive word. Candidate (a) wins since it satisfies both of the constraints on accent placement. It has two accents, the first of which is a prenuclear H* accent and the second a H+L* accent. The other candidates incur fatal violations of AHL and AH. Candidate (b) has only one accent, a nuclear pitch accent, which is incorrectly anchored in the initial foot, thus it violates AHL. Candidate (c) likewise has only one accented foot, which is appropriately anchored to the head foot. However, in this case the candidate violates AH since it does not have a prenuclear accent. Candidate (d) shows that having only a prenuclear accent is also prohibited. Candidate (e) has both pitch accents but they are misplaced causing violations of both AHL and AH. The pitch accents in the candidate in (f) are identical to those in the predicted surface form in (a) but it is ruled it since it violates MBR. This shows that a well-formed surface form for intensive reduplication must have both a prenuclear H* pitch accent and a nuclear H+L* pitch accent appropriately associated with the initial and rightmost feet respectively. That is, the left and right edges of the phonological phrase. Bisyllabic based words are accounted for similarly as shown in (29). (28)

 ‘very green’ /R[] + /




AHL MB AH OHD (P) (L, PP)

H* H+L*   a. [()]R [()]B H+L*  b. [()]R [()]B

*!

H+L* 

*!

c. [()]R [()]B H* 

*!

d. [()]R [(B)] H+L* 

H* 

*!

*

e. [()]R [()]B H* H+L*   f. [()]R [()]B

*!

Tableau (29) shows the selection of a bisyllabic based word with the constraints on the segmental shape and prosodic well-formedness. As before, the optimal candidate

Chapter 4. Pitch accents in Jamaican Creole reduplication

is the one which satisfies both the segmental constraints and the prosodic constraints, candidate (a). Candidate (b) satisfies the constraint on the segmental shape of the output form but fails on the A constraints since the pitch accents are incorrectly associated with the word. Candidate (c) has the appropriate prosodic form, both a prenuclear and nuclear accent but the entire base is not copied. This reinforces the idea that intensive reduplications must equally satisfy conditions on segmental and prosodic well-formedness to produce grammatical surface forms. Candidate (d) has no pitch accents and fails. (29)

 ‘very gummy’ /R[] + /




AHL MB AH OHD (P) (L, PP)

H* H+L*   a. [()R (B)] H+L* H*   b. [()R (B)]

*!

H* H+L*   c. [()R (B)] d. R B

*

*! *!

In summary, the intensive reduplicant is a prosodic foot. In this subtype of intensives, the reduplicant copies the base completely while adhering to the prosodic requirements. Further, it must adhere to other prosodic requirements which permit its stressed syllables to be accented. It must have two such syllables in order to be a wellformed output. The hierarchy of constraints in summarized in (30). (30)

Constraint hierarchy for Intensive Reduplication AHL(P) MB, AH (L, P) OHD

.

Conclusion

This paper argues for a metrical stress contrast between intensive and distributive reduplications. In particular, differences between the reduplication processes were shown to be directly related to the type of prosodic restrictions imposed on them at the segmental and prosodic levels. This showed that an analysis of the surface realization





Shelome Gooden

of JC reduplicated words requires an account of both phonological requirements on segmental structure and surface intonational patterns on these words. This kind of interaction between segmental requirement and prosody is also observed in a variety of other languages (cf. Chan 2006; Downing 2003; Minkova 2002). Chan 2006 for instance highlights the importance of prosody in reduplication in Cantonese a language which, like many Creole languages, has little inflectional morphology. I will summarize the issues relevant to JC below. The first issue has to do with the shape and size of the reduplicant. In both the cases discussed, the JC reduplicant is completely faithful to the base, i.e., MBR is always satisfied (full reduplication). Though the input form was not always identical to the reduplicant, the size and form of the reduplicant was shown to be a trochaic foot. Second, the metrical structure of intensive and distributive reduplication determined the kind of pitch accents associated with them. While intensives were doubly accented in nuclear position, distributives were accented only once. Since doubly accenting the reduplicated word was observed in the intensive reduplication process and in the narrow focus phrase, this suggests that it may serve a pragmatic function of marking ‘emphasis’. However, unlike intensives, distributives and compounds have a single nuclear pitch accent associated with the syllable with primary stress. In addition, both intensive and distributive reduplications are right prominent, having their nuclear stressed syllables dominated by the rightmost foot in the word. In this regard, both are similar to compound words which are also right prominent. In sum, the constraints on the intensive and distributive reduplication processes aimed at forming well formed prosodic structures in terms of their segmental and rhythmic organization.

References Adamson, L. and Smith, N. 2003. Productive derivational predicate reduplication in Sranan. In Kouwenberg, S. (ed.). Twice as meaningful: Reduplication in pidgins, creoles and other contact languages. London: Battlebridge. 83–92. Akers, G. 1981. Phonological variation in the Jamaican Continuum. Ann Arbor: Karoma. Alderete, J. 1993. The prosodic morphology of Jamaican Creole iteratives. In Benedicto, E. (ed.). University of Massachusetts Occasional Papers 20: The UMOP in indigenous languages. Amherst: Graduate Linguistic Student Association. 29–50. Anagbogu, P.N. 1995. The Semantics of Reduplication in Igbo. Journal of West African Languages 25. 43–52. Baker, P. 2003. Reduplication in Mauritian Creole, with notes on reduplication in Reunion Creole. In Kouwenberg, S. (ed.). Twice as meaningful: Reduplication in pidgins, creoles and other contact languages. London: Battlebridge. 211–8. Beckman, M. 1986. Stress and non-stress accent. Dordrecht: Foris Publications. Beckman, M. 1996. The Parsing of Prosody. Language and Cognitive Processes 11. 17–67. Beckman, M. and Pierrehumbert, J. 1986. Intonational structure in Japanese and English. Phonology Yearbook 3. 255–310.

Chapter 4. Pitch accents in Jamaican Creole reduplication Bell, S. 1983. Internal C reduplication in Shuswap. Linguistic Inquiry 14. 332–8. Broselow, E. and McCarthy, J. 1983. A theory of internal reduplication. The Linguistic Review 3. 25–88. Bruce, G. 1977. Swedish word accents in sentence perspective. Lund: Gleerup. Carter, H. 1987. Suprasegmentals in Guyanese: Some African comparisons. In Gilbert, G. (ed.). Pidgin and creole languages. Honolulu: University of Hawaii Press. 213–63. Cassidy, F.G. and LePage, R. 1967. [1980]. Dictionary of Jamaican English. Cambridge: University of Cambridge Press. Chan, H.L. 2006. Prosodic morphology of Cantonese attenuative reduplication. In Otsu, Y. (ed.). Proceedings of the seventh Tokyo conference on psycholinguistics. Tokyo: Hitsujishobo. 87–102. DeCamp, D. 1974. Neutralization, iteratives and ideophones: The locus of language in Jamaica. In DeCamp, D. and Hancock, I. (eds). Pidgins and creoles: Current trends and prospects. Washington: Georgetown University Press. 46–50. Devonish, H. 1989. Talking in tones. London: Karia Press. Devonish, H. 2002. Talking rhythm stressing tone: The role of prominence in Anglo-West African Creole languages. Kingston, Jamaica: Arawak Publications. Devonish, H. 2003. Reduplication as lexical and syntactic aspect marking: The case of Guyanese Creole. In Kouwenberg, S. (ed.). Twice as meaningful: Reduplication in pidgins, creoles and other contact languages. London: Battlebridge. 47–60. Devonish, H. and Murray, E. 1995. On stress and tone in Papiamentu: An alternative analysis. UWILing – Working Papers from the Department of Language and Linguistics 1, University of the West Indies, Mona. 43–57. Downing, L. 2003. Compounding and tonal non-transfer in Bantu languages. Phonology 20. 1–42. Finney, M. 2004. Tone assignment on lexical items of English and African origin in Krio. In Escure, G. and Schwegler, A. (eds). Creoles, contact and language change: Linguistics and social implications. Amsterdam: John Benjamins. 211–36. Good, J. 2004. Tone and accent in Saramaccan: Charting a deep split in the phonology of a language. Lingua 114. 575–619. Gooden, S. 2003a. Reduplication in Jamaican Creole: Semantic functions and prosodic constraints. In Kouwenberg, S. (ed.). Twice as meaningful: Reduplication in pidgins, creoles and other contact languages. London: Battlebridge. 93–103. Gooden, S. 2003b. Prosodic contrast in Jamaican Creole reduplication. Linguistische Arbeiten 478. 193–208. Gooden, S. 2003c. The phonology and phonetics of Jamaican Creole reduplication. PhD thesis, Ohio State University: Columbus, OH. Gooden, S. 2005. Proposal for special session on Prosodic descriptions of Creole languages: Implications for Creole formation. Annual Meeting of the Society for Pidgin and Creole Linguistics, 6–7 January 2006, Albuquerque, NM. Gussenhoven, C. 2000. The lexical tone contrast of Roermond Dutch in Optimality Theory. In Horne, M. (ed.). Prosody: Theory and experiment. Dordrecht: Kluwer. 129–67. Hayes, B. 1995. Metrical stress theory: Principles and case studies. Chicago: The University of Chicago Press. Hyman, L.M. 2006. Word-prosodic typology. Phonology 22. 225–57. Kager, R. 1999. Optimality Theory Cambridge: Cambridge University Press. Ladd, D.R. 1996. Intonational phonology. Cambridge: Cambridge University Press. Lawton, D. 1963. Suprasegmental phenomena in Jamaican Creole. PhD thesis, Department of English, University of Michigan.





Shelome Gooden Lawton, D. 1968. The implications of tone for Jamaican Creole. Anthropological Linguistics 10. 22–6. Marantz, A. 1982. Re reduplication. Linguistic Inquiry 13. 435–82. McCarthy, J. 1982. Prosodic templates, morphemic templates and morphemic tiers. In van der Hulst, H. and Smith, N. (eds). The structure of phonological representation: Part I. Dordrecht: Foris. 191–223. McCarthy, J. 2002. A thematic guide to Optimality Theory. Cambridge: Cambridge University Press. McCarthy, J. and Prince, A. 1993. Generalized alignment. In Booij, G. and van Marle, J. (eds). Yearbook of Morphology. Dordrecht: Kluwer. 79–153. McCarthy, J. and Prince, A. 1995a. Prosodic morphology. In Goldsmith, J. (ed.). Handbook of phonological theory. Cambridge, MA: Blackwell Publishers. 318–66. McCarthy, J. and Prince, A. 1995b. Faithfulness and reduplicative identity. In Kager, R., van der Hulst, H. and Zonneveld, W. (eds). Proceedings of the Utrecht workshop on prosodic morphology. The Hague: Mouton. 249–384. Minkova, D. 2002. Ablaut reduplication in English: The criss-crossing of prosody and verbal art. English Language and Linguistics 6. 133–69. Moravcsik, E. 1978. Reduplicative constructions. In Greenberg, J. (ed.). Universals of human language. Vol. 3 Word Structure. California: Stanford University Press. 297–334. Nespor, M. and Ralli, A. 1993. Stress domains in Greek compounds: A case of morphologyphonology interaction. Amsterdam Studies in the Theory and History of Linguistic Science 117. 201–8. Nespor, M. and Vogel, I. 1986. Prosodic phonology. Dordrecht: Foris. Nylander, D. 2003. Reduplication and compounding in Krio. In Kouwenberg, S. (ed.). Twice as meaningful: Reduplication in pidgins, creoles and other contact languages. London: Battlebridge. 133–7. Prince, A. and Smolenksy, P. 1993. (ROA Version 8/2002). Optimality Theory: Constraints in Generative Grammar: Technical Report. Rutgers University Center for Cognitive Science. Remijsen, B. and van Heuven, V.J. 2005. Stress, tone, and discourse prominence in Curaçao Papiamentu. Phonology 22. 205–35. Rivera-Castillo, Y. 1998. Tone and stress in Papiamentu: The contribution of a constraint-based analysis to the problem of Creole genesis. Journal of Pidgin and Creole Languages 13. 1–38. Rountree, S. 1972. Saramaccan tone in relation to intonation and grammar. Lingua 29. 308–25. Sebba, M. 1981. Derivational regularities in a creole lexicon: The case of Sranan. Linguistics 19. 101–17. Selkirk, E. 1984. Phonology and syntax: The relationship between sound and structure. Cambridge, MA: MIT Press. Sutcliffe, D. 1986. The Black churches. In Sutcliffe, D. and Wong, A. (eds). The language of the Black experience: Cultural expression through word and sound in the Caribbean and Black Britain. Oxford: Blackwell. 15–31. Sylvain, S. 1936. Le creole haitien: Morphologie et syntaxe. Port-au-Prince: Imprimerie De Meester. Vogel, I. and Raimy, E. 2002. The acquisition of compound vs. phrasal stress: The role of prosodic constituents. Journal of Child Language 29. 225–50. Voorhoeve, J. 1961. Le ton et la grammaire dans le Saramaccan. Word 17. 146–63. Wells, J.C. 1973. Jamaican pronunciation in London. Publications of the Philological Society XXV. Oxford: Basil Blackwell.

 

Effort reduction and the grammar Liquid phonology in Haitian and St. Lucian Eric Russell Webb1 University of California, Davis

This chapter examines surface variability involving liquids in Haitian and St. Lucian creoles, attributing surface forms to the interaction of constraints promoting effort reduction in phonetically based Optimality Theoretic phonology. Following Articulatory Phonology, segmental primitives are construed as gestural tract variables; the contextual transitions between tract variables are formalized as more or less effort inducing, categorized as oppositions between putatively learned, more and less effortful gestural sequences reflected in a series of effort reduction constraints. In Haitian, surface forms are attributed to the ranking of constraints favoring liquid-vowel coarticulation, whereas in St. Lucian surface forms are predicted by the ranking of constraints promoting cluster gestural reduction.

.

Introduction

French Lexifier Creoles (FLC) provide interesting examples of divergent liquid phonological patterning in closely related languages. Haitian and St. Lucian /ʁ/ and /l/, for example, are subject to similar distributional restrictions; in the case of the former segment, this also involves allophonic alternations which obtain in distinct environments in each FLC. This paper reexamines Haitian and St. Lucian liquid phonology in a gestural model, in which surface forms are explained as the product of contextual

. The present work is an expanded and revised version of a paper initially presented at the 2004 Creole Conference in Curaçao. I wish to thank conference participants for useful feedback and, especially, Emmanuel Nikiema and Anne-Marie Brousseau for suggestions regarding terminology. I remain in debt to editors Viveka Velupillai and Magnus Huber for their helpful comments, and to two anonymous reviewers whose critiques have made for clearer presentation and analysis. Any remaining shortcomings are solely my own.



Eric Russell Webb

effort avoidance. The goal of this work is to test the hypothesis that allomorphy (in the case of /ʁ/) and distributional restriction (in the case of /l/ and /ʁ/) can be explicated with reference to phonological knowledge concerning the physical realization of gestural sequences. In this analysis, phonological primitives are conceived of as gestural components, adopting formalisms of Articulatory Phonology (AP; Browman and Goldstein 1989 et seq.), where segments are describable as tract variables and linear strings as gestural scores. In a broadly Optimality Theoretic (OT; Prince and Smolensky 1993) grammatical model, output gestural scores are evaluated by constraints referring to effort minimization (following Kirchner 1998), where effort is defined as a complex of gestural implications. The grammar captures relative effort reduction in a hierarchical ranking of markedness constraints, such that the difference between the two FLC is one of the relative dominance of constraints targeting a specific effort reduction outcome. This analysis departs from the hypothesis that phonological knowledge governs the actuation of mental representations and that a formal grammar of this knowledge should make reference to learned patterns of articulatory goodness. Of course, several other considerations must come into play, most notably the role of perception (see e.g., Peperkamp and Dupoux 2003; Steriade 2001; Wright 2001, 2004) and lexical knowledge (see e.g., Gess 2003, 2004; McMahon 2000; Myers 1999). While the present work is neither an apologia for nor an explicit promotion of phonetically based phonology, it is hoped that this will enrich debate in both phonological and Creolist linguistic circles, as well as underscore the need for increased interest in and collaboration among researchers of different points of view.

.

Data

Language data called upon throughout this analysis is taken from Tinelli (1970, 1981) for Haitian and from Carrington (1984) for St. Lucian. These and other sources are consistent as it concerns the regular surface forms of liquids, as well as their phonetic descriptions; while the data presented below and discussed throughout should not be understood as the only possible surface forms of /l/ and /ʁ/ in either language, they represent general tendencies within the respective speech communities. It also should be stressed that a phonetically based phonological analysis such as that proposed here is not meant to account for all surface phonetic detail, including the fine grained output variability inherent to all speech production and associated acoustic signals. Rather, such a grammar seeks to incorporate in its formal mechanism discrete, phonological knowledge which governs the actuation of underlying forms as surface structural descriptions, associable to eventual acoustic output (see discussion in Russell Webb fc a, b). Acknowledging the limitations of such data, the descriptions of the cited sources provide a sufficient, if not all-encompassing, point of departure for the analysis at hand.

Chapter 5. Effort reduction and the grammar

Unsurprisingly, given their relatedness, liquid distribution in Haitian and St. Lucian is similar in many ways. For example, liquids in both creoles are licensed in simple onsets both word-initially and medially, as in (1). (1)

Simple onset a. Haitian rache [ʁa∫e] diri [diʁi] la [la] kilès [kilεs] b. St. Lucian rète [ʁεtε] diri [diʁi] la [la] male [male]

‘tear off ’ ‘rice’ ‘’ ‘who’ ‘stay’ ‘rice’ ‘’ ‘I went’

In simple onsets, Haitian casual speech /ʁ/ is reported to surface most often as a labialized approximant [w] when preceding a labialized or rounded vowel, as in (2); by contrast, St. Lucian /ʁ/ surfaces as a fricative [ʁ] in similar environments. (2)

/ʁ/ before round vowels a. Haitian rose [wɔz] ‘rose’ wout [wut] ‘route, road’ b. St. Lucian rule [ʁule] ‘drive’ ròb [ʁɔb] ‘dress’

In addition to differences noted in simple onsets above, the two FLC provide distinct surface forms of /ʁ/ in complex onsets. Haitian surface forms are predictable by the following vowel; /ʁ/ surfaces as [w] preceding rounded vowels and as [ʁ] preceding a non-rounded vowel, following (1) and (2). In St. Lucian, /ʁ/ surfaces as [w] in any complex onset, regardless of the following vowel; Carrington also notes that /ʁ/ is frequently elided in these contexts (e.g., [tape] or [twape] for trape, ‘to catch’), although he does not specify the frequency of elision (1984: 55). As above, /l/ is unaffected in both FLC. (3)

Complex onsets a. Haitian twò [twɔ] twou [twu] prèmje [pʁεmje] entre [ãtʁe] vle [vle] legliz [legliz]

‘too much’ ‘hole’ ‘first’ ‘enter’ ‘want.’ ‘church’





Eric Russell Webb

b.

St. Lucian pren [pwã] drive [dwive] kribish [kwi.bi∫] gro [gwo] glo [glo] vle [vle]

‘take.V’ ‘drift about’ ‘crayfish’ ‘fat’ ‘water’ ‘to want’

Whereas onset /ʁ/ surface forms differ in Haitian and St. Lucian, tautosyllabic coda liquid distribution is similar in the two FLC. As seen in (4), both prohibit /ʁ/ in simple and complex codas; /l/ is by contrast licensed in these positions.2 (4)

Tautosyllabic coda restriction a. Haitian lamè [lamε] ‘sea’ flè [flε] ‘flower’ gade [gade] ‘look at’ pôt [pɔt] ‘door’ wòl [wɔl] ‘role’ ampil [ãpil] ‘many’ b. St. Lucian wè [wε] ‘to see’ haji [aji] ‘to hate’ gadijen [gadijɛ̃] ‘guard’ pwèl [pwɛl] ‘animal hair’ vil [vil] ‘town’

Several issues with regard to the input status of /ʁ/ must be addressed prior to any phonological analysis, as FLC data raise the question of whether /ʁ/ is present in the underlying form of tokens such as rose, pren, or lame, regardless of orthographic conventions. Two contrasting approaches can be taken as supportive of the present analysis, which includes /ʁ/ in the phonological input: the first looks to larger variations among speakers, as attested in the descriptive data and variability in the larger speech community; the second follows from theoretical principles that underlie the present analysis and OT grammars, in general. Indirect evidence for underlying /ʁ/ is gleaned from alternations seen in more emphatic and acrolectal speech in both creoles, e.g., where Haitian rose surfaces as [ʁɔz]

. Note that diachronically related tokens in the lexifier contain coda /ʁ/ (e.g. Fr. la mer /mεʁ/) and more acrolectal forms of the creoles provide surface forms including coda [ʁ]. The examples presented parallel data and discussion in Tinelli (1981), which conflate diachronic and synchronic phonological processes.

Chapter 5. Effort reduction and the grammar

and St. Lucian gro as [gʁɔ]. Carrington also reports several instances of idiosyncratic variability in St. Lucian, where older generations are shown to produce fewer surface [w] (/ʁ/) tokens than are younger speakers (1984: 27, see also Tinelli 1981: 13). More explicit evidence for the underlying status of coda /ʁ/, at least in the case of Haitian, comes in the form of derivations in which /ʁ/ and /l/ are maintained, as in (5). (5)

Extrametrical liquid erasure in Haitian sik [sik] ‘sugar’ vs. sikre [sikʁe] meg [mεg] ‘thin.’ vs. megri [mεgʁi] tab [tab] ‘table’ vs. tabli [tabli] misk [misk] ‘brawn’ vs. miskle [miskle]

‘sweet’ ‘thin.V’ ‘settled’ ‘brawny’

Regardless of data particularities and variability among speakers, OT’s principle of Richness of the Base (ROTB) establishes that a grammar must take into account and make verifiable predictions given any possible input, where input is defined universally rather than locally (see Archangeli 1997). In contrast to antecedent theoretical models, OT dispenses with language-specific rules in favor of universal, violable constraints which are arranged in a hierarchy, such that higher-ranking constraints have a greater effect in the selection of a given output form. Importantly, ROTB requires that no constraint hold over grammatical input and that all grammars contend with any input, including those which do not have underlying lexical status and/or which are distributionally restricted in the surface forms of the language in question. As such, the locus of language particularity is not the input, but the constraint ranking (Prince and Smolensky 1993: 207; see also Kager 1999: 19, 29–34). For the purposes of the present discussion, both arguments are satisfied by the assumption that /ʁ/ and /l/ are present in input forms, even though this may not be the case for all lexical representations. The phonological grammar must select output forms in the case of an input /ʁɔz/ (→ Haitian [wɔz], but → St. Lucian [ʁɔz]) or /mεʁ/ (→ [mε]), even in instances where there is no fully faithful surface correspondent to these. A summary of pertinent liquid allophony and distributional restriction in Haitian and St. Lucian is given in (6) using traditional notation. Shading in this table highlights instances of surface distinction between the two FLC; an asterisk (*) indicates an illicit or unattested output; elsewhere should be taken at face value. (6)

Summary of distribution and allophony /ʁ/ → [ʁ] Haitian

/ʁ/ → [w]

*[ʁ]

*[l], *[ʁ]

/l/ → [l]

/ __ V [− lab] / __ V [+ lab] / __ (C) ]w / C __ ]w elsewhere

St. Lucian / __ V

/ C __ V

/ __ (C) ]w / C __ ]w elsewhere

Following (6), any analysis of Haitian and St. Lucian liquid phonology must account for /ʁ/ allophony and distributional restriction. The same analysis must also explicate





Eric Russell Webb

the distinctions noted between different allophonic patterns, i.e., why /ʁ/ surfaces as [w] in complex onsets in St. Lucian, irrespective of the following vowel, while the same result in Haitian depends entirely on the vocalic context.

.

Analytical framework

In the majority of texts cited above, data in (1) through (5) are accounted for in traditional, rule-based analyses. For example, Tinelli proposes an R-rounding rule to account for Haitian data, as in (7) below: (7)

R-rounding rule (Tinelli 1981: 165, his DR 11, modified) C → C / ____ V [+ cons] [+ lab] [+ lab] [− obst] [− clos]

Similar rules are posited to account for deletion of /ʁ/ in Haitian post-vocalic and extrametrical positions, which are considered cases of laxing (Tinelli 1981: 173). Regardless of subsequent modification to these rules that might be necessitated by different phonological theories and/or models (e.g., substitution of an aperture condition for the feature [±fricative]), and ignoring issues of lexical representation as well as lexifier representational access in synchronic creole grammars, transformational rules such as the above are descriptively adequate. While analyses such as these do well to account for surface forms through the formalization of languagespecific rules, they do little to advance an understanding of these and related instances of liquid allophony and distributional restriction from a grammatically aspecifi c (i.e., universal) point of view. Similarly, Nikiema (2002) discusses the erasure of final, post-obstruent or extrametrical liquids, a process noted in nearly all FLC, although he shies from providing an explanation of why final liquids are erased in the first place. Such rules make no reference to causality, i.e., to the principles at work which shape phonological grammars, nor do they account for the assumed grammatical processes (rather than simple imperatives) which constitute speakers’ linguistic competence. A distinct approach to the question of liquid phonology is advocated in the present work, which assumes that explanation of any linguistic phenomenon should refer to extra-linguistic principles (following inter alia Haspelmath 2004; Penke and Rosenbach 2004). Primary to the present analysis is an understanding of the productive properties of segments and the relative effort implied by these. This is undertaken in two steps, one addressing the specific production characteristics of liquids, the other addressing effort, effort categorization and effort reduction. Both components of the

Chapter 5. Effort reduction and the grammar

present analysis seek to capture the mechanisms by which speakers’ competence incorporates learned patterns of effort reduction in a phonetically based phonological model. Brief mention should be made of what is meant by a phonetically based phonology, although readers unfamiliar with such approaches are encouraged to consult the works cited below and elsewhere. A growing body of theoretical work advocates increased attention to the ways by which surface sound patterns are influenced by phonetic principles, including the relative effort implied by the actuation and reception of speech, and to the incorporation of these principles in the grammar, a formal repository of linguistic competence (see Hayes et al. 2004 for a general background discussion; see also Boersma 1998; Kirchner 1998; and Steriade 2001). Fundamental to these approaches is not the understanding that phonetics leads directly to phonology, i.e., that the form can be immediately derived from function, but that functional oppositions are learned and categorized by speakers and that phonetic goodness is inductively grounded in phonology (Hayes 2004). As such, it is proposed that the forms of a sound grammar are indirectly catalyzed by the principles governing markedness (e.g., biomechanical ease, perceptual salience) and faithfulness (e.g., input-output cue preservation, gestural correspondence). Given the stated hypothesis that surface forms can be attributed to phonetically based phonological knowledge, the first task of the present work is to describe the data in (1) through (5) as gestural phonological units. This involves a recapitulation of the articulatory properties of these segments and their formalization within an AP model. Modifications to input gestural scores are subsequently explained with reference to the biomechanical principle of effort reduction or avoidance (see Boersma 1998: 149 and Kirchner 1998: 37–8), captured in phonetically based constraints, discrete units of phonological knowledge referring to the actuation of different gestural primitives. In Section 4, these are integrated into a grammar predicting surface forms; output is accounted for by the interaction of markedness constraints, promoting effort avoidance or minimization, and faithfulness constraints, promoting the faithful copying of input gestural information in the output.

.

Gestures in phonology

The production characteristics of the liquids under investigation are well known. These are summarized in (8), taking into consideration active and passive articulators, tongue contour or shape, oral cavity aperture and secondary articulator characteristics. Although [w] is not normally grouped with liquids, this segment is included in the following discussion based on its surface distribution in FLC. For more in depth discussion of the articulatory characteristics of dorsal continuants, in particular, see Delattre (1971); Tranel (1987: 139–43); Ladefoged and Maddieson (1996: 225–6) and Russell Webb (2002: 33, 155–62).





Eric Russell Webb

(8)

Articulatory properties of FLC liquids

segment description uvular (velar, ʁ uvulo-velar) fricative or approximant l alveolar or dental lateral approximant w

labio velar approximant

active tongue passive articulator contour/shape articulator velum/uvula tongue flat3 dorsum

secondary articulator lips (variable contraction)

tongue blade

tongue sides closed (verticalis, centrally, genioglossus) open laterally tongue dor- variable: sum (velum) close to wide

lips

lateral: tongue alveoli sides drawn down, center raised flat n/a

cavity aperture variable: close to wide

It should be noted that the above description assumes that [ʁ] articulation allows for variable lip rounding and that [w] involves the raising of the tongue body normal to the velum.4 The raw phonetic characteristics of speech segments are involve a good deal of contextual and speaker-dependent variability, such that the assumed underlying representation of different phonemes and allophones does not mirror their surface correspondents. In most phonological models, only those properties which provide distinctiveness are assumed to be present in the input, leaving questions of gradient surface forms to subsequent levels of processing (e.g., to EVAL; see Mielke 2004; Russell Webb fc b). In the present analysis, these properties are defined as gestural primitives following AP, i.e., as the productive characteristics which both allow for distinctiveness vis-à-vis other segments and reflect the prototypical gestural properties of output forms. AP conceives of phonological units as dynamic articulatory structures consisting of tract variables, corresponding to meaningful variations in cavity aperture and active articulator displacement. These variables reflect the productive properties of surface segments noted in (8). For example, narrow lip aperture reflects positive contraction of the lips, while a narrow tongue body constriction reflects the close aperture of the

.

The term “flat” is used to refer to the contour of a non-tensed tongue dorsum.

. Varying degrees of lip rounding of contraction are noted during the production of dorsal continuants in several languages, including those whose phonetic form is similar to Haitian and St. Lucian, e.g. French and Dutch (Russell Webb 2002) and German (Delattre 1971). Assuming the fricative formants and dispersion of spectral noise are the primary cues to the identification of [ʁ], lip rounding and protrusion have little effect on discrimination.

Chapter 5. Effort reduction and the grammar

tongue dorsum normal to a passive articulatory target (e.g., the velum); by contrast, wide lip aperture reflects the lack of labial contraction or rounding and wide tongue body constriction a relatively greater aperture between the tongue body and passive articulator in the oral cavity. A summary of relevant tract variables of these segments is given below, in (9). (9)

Tract variables (Browman and Goldstein 1992: 157) tract variable



w

variable

narrow

velum

velum

narrow

variable

l

LA (lip aperture) TTCL (tongue tip constriction location) TTCD (tongue tip constriction degree) TBCL (tongue body constriction location) TBCD (tongue body constriction degree)

alveolar closed-central open-lateral

In addition to segment specific tract variables, AP provides for the sequential organization of gestures along a time scale, referred to as a gestural score, wherein the gestural specification of one segment can influence that of another in a given domain. In the context of on-line speech, the tract variables are subject to coarticulatory gestural overlap, as in (10), which gives examples of Haitian /l/ and /ʁ/ in the context of both rounded and unrounded vowels. (10)

Sample gestural scores (Haitian la, rache, rose) wide

LA

var.

wide

wide

var.

narrow

TTCL

alveolar

alv

TTCD

lateral

grvd velum

TBCL TBCD l

palate

velum

wide

narrow

wide

narrow

wide

a







e

narrow wide 



z

Adjacent, differentially specified tract variables in (10) highlight transition sites, where coarticulatory gestural overlap is predicted. In cases where a tract specification is variable, as for /ʁ/, the phonology must predict the output specification of surface forms, i.e., if this will be realized as [wide] or [narrow]. Examples such as these draw attention to several important points which can be adduced from data in (1) through (5), namely that the languages in question provide different, albeit similar, restrictions



 Eric Russell Webb

on surface forms in gestural scores. Three such restrictions are of immediate concern to the present analysis, summarized in (11). (11)

Surface patterns: FLC liquid gestural scores a. Restriction of LA sequences: the sequence LA[narrow] – LA[wide] fails to surface in Haitian; the sequence TT or TB[close] or [closed] – LA[narrow] fails to surface in St. Lucian b. Restriction of TBCD: the sequence TBCD [narrow] – LA[narrow] fails to surface (Haitian only) c. Prohibition of edge sequences involving TT or TB: the sequence of TT [close]/[closed] and/or TB [close]/[closed] fails to surface in word final position (both FLC).

An intriguing observation is made from observations of Haitian and St. Lucian data. In Haitian, variation concerns only the coarticulatory effects stemming from adjacent LA tract variables, i.e., it is seemingly unconcerned with other aperture specifications relative to the oral cavity (TTCD or TBCD). Contrastively, St. Lucian rhotic variation depends upon tongue tip or body constriction degree; here, the sequence of a wide lip aperture and a narrow tongue constriction is prohibited, i.e., narrow TT/TBCD precedes narrow LA and wide TT/TBCD precedes wide LA. Whereas Haitian surface forms imply feature sharing from /ʁ/-to-vowel, St. Lucian forms are predictable by consonant-to-consonant coarticulatory factors. Given a phonological model built upon gestural primitives, phonological knowledge as it is cast upon these in the context of a holistic competence system, i.e., a grammar, must both describe how surface forms derive from underlying gestural specifications and explain why modifications occur – or fail to occur – in given generations. In Haitian, the grammar must account for the leftward spread of lip aperture and the restriction of surface gestural scores involving both lip and tongue body constriction, while in St. Lucian, the grammar must account for the restriction of certain sequences involving narrow lip aperture, regardless of any vocalic influence, where the combination of tongue tip or body closure and lip narrowing is prohibited. In both FLC, a combination of two close or closed TB variables at word edges is also prohibited and must be predicted by the grammar. It might be descriptively sufficient to simply posit a grammar which provides for ad hoc rules or constraints referring to specific tract variable contexts, e.g., *LA[close] − LA[wide] (‘no sequences of close and wide lip aperture’); indeed, this seems to be the case for prohibition of /ʁ/ in codas, as opposed to /l/, /s/ and other consonants, which are licensed in both FLC (see discussion in Section 4). In the case of onset variability and post-obstruent erasure, however, constraints such as these present explanatory disadvantages, as they deny a place for more generalized phonetic knowledge in phonological grammars. Prior to developing any constraints, however, attention must be given to the notion of gestural effort, as the principle of effort avoidance underlies the assumed discrete phonological competence captured in constraints and their ranking in a production grammar.

Chapter 5. Effort reduction and the grammar 

.

Effort in phonology

Effort and ease of articulation have long been viewed as contributors to linguistic form, notably to diachronic phonological patterns. As early as 1891, Passy distinguished between économie and emphase as the primary catalysts for sound change. Earlier twentieth century interest in effort is seen Grammont (1933) and Martinet (1964), whose advocacy of functionalism is largely based on articulatory considerations. More recent attempts to integrate productive considerations into grammar are provided in Boersma’s (1998) Functional Phonology, Kirchner’s (1998 et seq.) approach to lenition and in others advocating a review of the place of phonetics in phonology (see e.g., the introductory chapter to Hayes et al. 2004). Common among the different phonetically based, driven and inspired approaches is an assumption that phonological form is shaped in part by articulation and/or perception. A review of cited literature reveals that, while authors share assumptions about the influence of non-grammatical forces upon the development and/or emergence of grammar, there are divergent opinions as to the substance and content of these forces, including the notion of effort. There is general agreement that the relative difficulty implied by a given gesture and/or gestural configuration may underlay certain types of phonological knowledge; there is, however, no developed consensus about the scope or discrete form of effort in phonology, i.e., how effort is congruent with competence, nor is there a common definition of what is meant by the term effort. More troubling for some are difficulties associated with the quantification of effort. Such issues notwithstanding, a survey of effort in recent literature underscores common threads. For Lindblom, effort can be equated to a degree of force times a precise unit of time (1990, see also Lindblom et al. 1995). Boersma’s expanded definition includes variables for time or duration of muscular activation, special muscular displacement, velocity of muscle movement, mass of muscle involved in the displacement, counter forces exerted by other muscles and by gravitational forces, and a constant correlated to the energy required for muscular contraction (1998: 149–50). Kirchner’s definition is founded upon the notion of effort as a positive expenditure of energy; with articulator displacement, force and time, he includes a factor involving articulatory precision, i.e., the relative degree of complexity implied by a segment (1998: 35–51). Effort is necessarily gradient, such that constraints referring to effort make reference to idealized degrees or thresholds, e.g., the relative complexity, duration and precision implied by articulator movement away from a rest or neutral position. Such phonetic gradation is assumed to be categorized into levels of effort, based on speaker knowledge of category boundaries and systemic contrast acquired by speakers (see e.g., Kirchner 1998: 59–60, 87–93). This work follows largely from Kirchner’s understanding of effort, with the proviso that an effective understanding of articulatory effort considers both contextual and systemic implications. Effort can be intrinsically contrastive within a phonological system, i.e., the relative effort implied by one inventory member vis-à-vis another, or

 Eric Russell Webb

effort can be concatenative, i.e., the relative effort implied by one segment adjacent to another in the context of on-line speech. If articulatory effort is to be understood as a positive expenditure of energy for the purpose of realizing a given gesture, effort reduction is assumed to obtain whenever less positive neuro-muscular energy is spent on a given task. As such, effort reduction can be accomplished by avoiding precise gestures or favoring less-precise ones, by avoiding more distal articulator displacements or those which occur over a longer period of time in favor of displacements accomplished in a shorter period or implying reduced displacement, and by avoiding complex gestural sequences, such as those which obtain when the tongue is drawn forward, then back and the forward again. This section focuses on the relative effort implied by the sequential realization of gestural specifications in a specific domain (in this case, the word), including transitions and complex sequences of tract variables, and effort reduction implied by the non-realization of gestures and/or tract variables within these domains. Focus is on the gestures and gestural scores implied by data in (1) through (5); of specific concern are effort considerations applicable to lip aperture, velar constriction and tongue body displacement, as well as the effort implied by transitions between different specifications affecting these. Also at issue is an account of variability involving one segment (i.e., /ʁ/) that would seem to be impervious to others (/l/, /s/, etc.), i.e., why certain gestures and gestural specifications imply effort reduction whereas others are unaffected by it. Crucially, the relative effort implied by either a segment or sequence of segments is phonologically effortful only in the context of a system, where distinct output structural descriptions are categorically evaluated with regard to effort expenditure and/or minimization. These categorical effort scenarios are formalized in Section 3.3; markedness constraints promoting relative effort reduction are articulated in Section 3.4.

.. Lip aperture Following the definition of effort elaborated above, any avoidance of a particular neuro-muscular activity results in effort reduction. Absent of a sequential gestural context, this implies that any contraction of the lips – i.e., tensing or narrowing – is more effortful, as lip contraction (corresponding to LA[narrow]) requires positive energy expenditure. In a context including a linear sequence of segments, themselves a paradigmatic bundle of dynamic gestures, however, effort reduction might be accomplished by reducing the transition between adjacent segments, i.e., providing for more coarticulatory overlap. In such a context, effort reduction may obtain if one tract variable is maintained across a series of segments, rather than being manipulated two or more times in this context. In a sequence involving the contraction of the lips, overall effort is reduced if lip aperture is maintained; contrastively more effort is involved if the lips are alternately widened and narrowed or narrowed then widened. These considerations are modeled in (12), where displacement or

Chapter 5. Effort reduction and the grammar 

activation of the lips is presented as the vertical axis; solid red lines indicate more effort inducing gestural transition patterns, while dashed green lines indicate less effort inducing ones. (12)

Schematization of effort and LA

wide a p e r t u r e

more effortful (neutral/rest)

narrow less effortful time → S1

S2

In a gestural phonological model in which lip aperture is understood as a segmental primitive, modification to input gestural specification in a linear sequence of segments specified for lip aperture may result in effort reduction.

.. Multiple displacements The avoidance of multiple articulator displacements may also result in effort reduction, especially when displacement involves distinct targets to which the tongue must move in order to realize a gesture. In a sequence involving a stop + ʁ + vowel, for instance, the tongue must first be displaced making contact with its target (the alveoli, palate or velum), next positioned normal to the velum, and subsequently be re-positioned for vocalic production. Elimination or non-realization of a gesture or gestural transition for the intermediate segment results in a net reduction in overall effort, again as the avoidance of a positive articulatory displacement. In the case of alveolar obstruents, this involves a movement forward and up (to the alveoli and/or upper incisors), back (to the velum) and the a repositioning according to the concerned vowel; in the case of a labial obstruent, the tongue is neutral during initial occlusion, but must still be drawn back toward the velum and subsequently repositioned. When the initial obstruent is velar, the tongue is in position, but must still anticipate movement to the gestural configuration specified by the following vowel. A schematic representation of this is provided in (13).

 Eric Russell Webb

(13)

Effort and multiple constrictions wide less effortful TT/ TB CD

more effortful

closed C



V

Following (13), modification to a sequence of input TT or TB gestures in a sequence of gestures results in a net reduction in overall output effort. If constriction is maintained in this sequence, the least effortful output will be one which transitions from a more closed to a more open (or the reverse) gestural configuration absent of a tertiary or intermediate articulator displacement. A related, but distinct type of effort reduction involves that of gestural scores involving multiple overlapping constrictions. Such constrictions are phonetically correlated to complex articulatory profiles, i.e., the simultaneous activation of two or more articulators. Given the definition of effort in above, the non-realization of one or more overlapping gestures is also a reduction in overall articulatory effort. Of course, multiple effort-reducing solutions are possible to any given potentially higher effort inducing situation. Articulator displacements shared across the boundary of several segments may be more likely to be preserved (akin to lip aperture coarticulation), whereas those that involve transitions away from distal targets are likely to be reduced or not realized at all (akin to the discussion of displacement avoidance).

.. Gestural erasure A final instance of concern here involves the non-realization of multiple underlying gestures, i.e., the loss of an entire phonological segment in the output. This is a common occurrence at domain boundaries, in the FLC data under investigation, at the word edge. Given the definition of effort above, such instances of gestural erasure are easily understood; any reduction in displacement or force, such as obtains from not performing an action compared to performing it, is less effortful. It is interesting to develop even briefly a picture of the relative effort implied by the non-realization of gestures implied by segments at the word edge, such as the loss of extrametrical liquids noted in (5). In order for any extrametrical element to be produced, the active articulator must be released from a complete or nearly complete constriction (plosive or fricative, respectively) and positioned following the gestural specification for /l/ and /ʁ/,

Chapter 5. Effort reduction and the grammar 

segments involving distinct and relatively complex tract variables and which are of relatively greater aperture or distinct contour than are other obstruents. Mirroring statements regarding effort reduction via adjacent gestural harmony and the effacement of multiple constrictions, effort reduction may be accomplished by the reduction of transitions between gestures, when the ultimate transition into a vowel is not accomplished.

.

Categorization of effort

Hayes (2004) proposes that speakers inductively learn patterns of phonetic goodness; given an AP formalization of phonological segments and segment concatenation, this implies that speaker competence includes knowledge of the relative articulatory ease and difficulty of specific gestures and gestural scores. Given a range of output forms, more- and less-effort inducing forms are learned and subsequently categorized. As this concerns the relative effort of different segments and the sequential realization of segments in context, markedness may be correlated to the amount of overall effort required for a given production. Like classical interpretations of markedness, phonological effort is both discrete and locally contrastive; unlike most other markedness considerations, effort is categorical and gradient, such that effort can be incremental (Kirchner 1998: 201). Rather than knowing an infinite range of possibilities, speakers are assumed to learn oppositions between better and worse outcomes; for the purposes of the present analysis, multiple levels of effort reduction are not necessary and discussion is limited to instances of modification to input gestural specification, including the non-realization of a gesture or tract variable. Assuming that speakers inductively learn the relative effort implied by different prototypical gestural scores as they acquire language, and that the oppositions under discussion in the present work are universally available, three crucial categorizations can be asserted that are of relevance to FLC liquids. These are formalized in (14, lip aperture), (15, multiple constrictions) and (16, gesture avoidance or non-realization). (14)

Categorized lip aperture effort (> ‘is more effortful than’; S = any paradigmatic arrangement of gestures, i.e., a segment) ∀ S1 S2/S[LAα] ∨ S[LAβ] → S1[LAα] S2[LAβ] ∨ S1[LAβ] S2[LAα] > S1[LAα] S2[LAα] ∨ S1[LAβ] S2[LAβ] A sequence of LA gestures having distinct aperture specification is more effortful than a sequence of LA gestures sharing the same aperture specification

(15)

Effort categorization of multiple transitions ∀ S1S2S3/S1CD[close] ∨ [closed] ∧ S3CD[wide] → S2CD[close] > S2CD[wide] For all sequences S1S2S3, where S1 constriction degree is either close or closed and S3 constriction degree is wide, a close S2 constriction degree is more effortful than a wide S2 constriction degree

 Eric Russell Webb

(16)

Multiple gestures at word edge ∀S1S2]w/S1CD[close] ∨ CD[closed] ∧ S2CD[−wide] → S1S2 > S1−(S2) For all sequences S1S2 at the word edge, where S1 constriction degree is either close or closed and S2 constriction degree is not wide, the sequence S1S2 is more effortful than the sequence S1 alone.

Statements such as those in (14) through (16) are assumed to reflect what speakers have learned through trial-and-error during language acquisition. Among the many categorizations that are undoubtedly made based on the relative effort of different gestures and gestural scores, including those militating against all effort, these are most applicable to the present discussion. Such categorizations are reflected in principled constraints developed below and applicable to positional licensing of tract variables or gestural configurations.

.

Constraints

In OT, knowledge governing the processing of surface forms is embodied in violable constraints. Within a phonology built on gestural goodness, discrete knowledge promoting effort reduction is construed as markedness and interacts with faithfulness, promoting the maintenance of input featural specifications (here the gestural information contained in input representations).5 Kirchner introduces a cover markedness constraint promoting effort avoidance or reduction, given in (17), below. (17)

Articulatory effort minimization L: minimize articulatory effort (Kirchner 1998: 38, his 2-1)

This constraint is similar to other functionally defined constraints targeting articulatory economy or ease, such as *E of Boersma (1998), although it is theoretically distinct inasmuch as effort avoidance is a positive imperative, whereas effort prohibition militates against specific output forms or gestural outcomes. Assuming that constraints are universal (a basic OT premise), L must be very low-ranked in all of the world’s languages. Targeted effort avoidance, reflecting learned patterns regarding specific categorical oppositions and involving specific gestures and/ or gestural configurations, is captured in focused constraints, taking into account the context-sensitive oppositions of (14) through (16). Three targeted effort reduction constraints are of importance to the present work, given in (18) below. The original titles of Kirchner (1998) are reformulated as -E, which should be read as the positive imperative “reduce effort.” . Readers unfamiliar with OT, constraint theory and conventions used in OT analyses are referred to Kager (1999).

Chapter 5. Effort reduction and the grammar 

(18)

Targeted effort reduction -E(LA): for all sequences involving adjacent LA gestural specification in the input, the least effortful output is chosen -E(CD]w): for all sequences involving adjacent CD gestural specification at the right edge of words, the least effortful output is chosen -E(CD)C2: for all tautosyllabic consonant sequences, the least effortful CD specification of C2 is chosen

The logical complement to effort reduction is a series of constraints promoting the preservation of input, i.e., for the maintenance of correspondences between input gestural representations and output form. Following Kirchner, faithfulness is formalized here as P, provided in (19) with modifications specific to the gestural approach taken here. (19)

Preservation (Kirchner 1998: 74, his 3-7) P(erve) (F Input/Output): Given a set of features F with the specification α, for each input αF specification there is exactly one output corresponding αF specification, and for each output αF specification there is exactly one input αF specification (‘input featural specification should be preserved in the output; output featural specification should have an input correspondent’)

As with markedness constraints in (18), faithfulness may also be targeted to specific tract variables and gestures, this based on the assumption that constraints distinguish between different gestures and tract configurations. Two specific preservation constraints are provided in (20), concerning the preservation of input Lip Aperture and the preservation of Tongue Body Constriction Degree, respectively. (20)

Targeted preservation P(LA): input LA specification is preserved in the output P(TBCD): input TBCD specification is preserved in the output

Faithfulness as defined in (20) is not sufficient to account for all attested output, as Haitian and St. Lucian data clearly evidence instances where input gestures are modified, rather than maintained or lost, as the all-or-nothing P vs. –E dichotomy might imply. Of concern here is with the degree of preservation of input specification, rather than the presence or absence of the gesture itself. As noted in both Haitian and St. Lucian, the content of tract variables may at times be changed, but the tract variable is not removed as such. The grammar must therefore take into account gradient preservation, i.e., the preservation of the tract variable but not the gesture specific to that variable. This assumes the availability or emergence of constraints which refer not only to specific gestures, but to the variable content of these gestures, and that the grammar distinguishes between the two. (21)

Minimal preservation PM(LA): the tract variable LA is preserved in the output PM(TBCD): the tract variable TBCD is preserved in the output

 Eric Russell Webb

Rather than targeting the content of tract variables, i.e., input gestural specifications, as is the case for P(LA) and (TBCD), PM promotes the maintenance of the tract variable itself, absent a particular specification. As such, an output [ʁ] (an approximant dorsal continuant) in correspondence to /ʁ/ satisfies PM(TBCD), as it maintains a putative TBCD specification (here, [wide]), even though it violates P(TBCD), given the disparity between the output value and that of the input (i.e., TBCD [narrow]).

. Analysis The present analysis must account for allophony triggered by /ʁ/-vowel coarticulation in Haitian and by consonant cluster gestural simplification in St. Lucian. Given the phonological relatedness of liquids, an account must furthermore be made for the lack of /l/ variability in these cases. Finally, the phonological grammars proposed herein must account for word-final, post-obstruent liquid elision, affecting both /ʁ/ and /l/. The last incidence is one that is perhaps the easiest to explain and, at the same time, the least controversial. Numerous accounts for the erasure of word-final, post-obstruent liquids have been proposed, including French (see e.g., Rialland 1994; Russell Webb fc a) and Haitian (Nikiema 2002). Here, a unified account can be presented for both FLC in question, given the ranking –E(CD]w) >> P(TBCD), where the latter refers to the preservation of input gestural specification with regard to constriction. It should be noted that, as formulated in (18), –E(CD]w) is ignorant of simplex codas, as it works only against any two gestures specified for CD at the right edge of words and is not triggered by a single CD in these positions, as exemplified by output for vil (‘city’); this constraint also has no effect on bimorphemic, bisyllabic derivations, as in tabli (‘table-V’). Evaluation of these outputs is provided in (22), applicable to both Haitian and St. Lucian. Surface restriction on /ʁ/ patterning in codas is assumed to be accountable by a more general constraint (e.g., * ʁ/) and is not treated in further detail. (22)

Word-final multiple constriction erasure –E(CD]w) /tabl/ tabl
tab /tabl+i/
tabli tabi /vil/
vil vi

P(TBCD)

*! *

*!

*!

Chapter 5. Effort reduction and the grammar 

At the same time, the grammar must prohibit output variations triggered by consonant clustering, e.g., *[pwã], attested in St. Lucian, but not Haitian; both this and the relevant P constraints must outrank –E(CD)C2/CCV (hereafter –EC2 for reasons of economy) in the latter, as seen in (23). Note that P(TBCD) is assumed to be co-ranked with P(LA) and that, at this stage of analysis at least, there is no evidence that P(LA) and –EC2 cannot be co-ranked.6 (23)

Pre-vocalic LA harmony: Haitian PM(LA) –E(LA) P(LA) P(TBCD) –EC2

/ʁɔz/ *!

ʁɔz
wɔz ɔz

*!

*

*

*

*

/tʁɔ/ *!

tʁɔ
twɔ tɔ

*!

* *

*

*

*

*!

*

*

*

/ʁɛte/
ʁɛte wɛte ɛte

*!

/pʁɛmje/ *


pʁɛmje pwɛmje pɛmje

*!

*!

*

*

*

The question remains as to why /ʁ/ is affected in such a way, whereas /l/ is insensitive to this. The gestural scores relevant to both simple and complex prevocalic /l/ indicate that, in contrast to /ʁ/, there is no shared gestural specification across segment boundaries for the former, as is the case for LA of the latter (/l/ involves a distinct

. Here, it may be possible to subsume both P(LA) and P(TTCD) under a general constraint targeting tract variable specification of the input, e.g., P(tract).

 Eric Russell Webb

tongue tip contour, rather than a combined tongue body constriction and lip aperture). Such an observation is not, however, enough to provide for a phonological account of the differences between the two liquids’ phonotactics. Additional effort reduction constraints might include –E(LC), promoting avoidance lateral tongue constriction. Assuming a concurrent structure preservation constraint relative to the gesture involved in other non-obstruent consonant input, here provided as P(LC), the following ordering is obtained: (24)

Haitian constraint ranking –E(CD]w) >> PM(LA) >> –E(LA) >> P(LA), P(TBCD), P(LC) >> –EC2

This constraint ranking effectively eliminates unattested outputs involving obstruent + /l/ in complex onsets, such as vle ‘want’ /vle/ → [vle] *[ve], which would otherwise obtain were –EC2 ranked above P. St. Lucian /ʁ/ surface forms obtain by the manipulation of tract variables in onset clusters; the grammar must therefore eliminate sequences of close or closed constrictions in complex onsets, captured in the ranking –EC2 >> –E(LA). The relative dominance of –EC2 underscores an important distinction between St. Lucian and Haitian and promotes [pwã] as the optimal output for /pʁã/, in contrast to the latter.7 This constraint ranking alone is not sufficient to account for all output, however, as St. Lucian provides for TTCD[close] preceding rounded vowels, but only in simple onsets, establishing the relative ranking –EC2 >> P(TTCD) >> –E (LA). P(LA) must also be included in this ranking, in order to eliminate candidates such as *[pã], where both the tongue body constriction and lip aperture gestures are absent, as would obtain a constraint promoting input-output faithfulness with regard to LA low-ranked. In both instances of LA and TTCD, St. Lucian remains agnostic with regard to the degree of either velic closure or lip constriction in all cases; lacking more specific evidence, this analysis assumes that PM(LA) and PM(TTCD) are subsumed by the respective, co-ranked P, although the possibility for a higherranking minimal preservation must be left open for future analyses. As presented in (25), constraint ranking leads to several instances of over-prediction, e.g., the erasure or lenition of /l/ in complex onsets, for which there is no evidence. As such, a constraint promoting the input-output correspondence of a lateral constriction gesture must be ranked higher than that of –EC2, as is –E(CD]w), providing the following constraint ranking for St. Lucian.

. An interesting question, which extends beyond the scope of the present work, concerns the reasons underlying the adoption of different strategies by related grammars, especially by closely related languages (in this case, creoles with a common lexifier).

Chapter 5. Effort reduction and the grammar 

(25)

Cluster TBCD reduction: St. Lucian −EC2

P(LA)

P(TTCD)

−E(LA)

/ʁule/ *

ʁule wule

*!

ule

*!

*

/pʁã/ pʁã

*!

pwã

*

pã

*

*!

/gʁo/ gʁo

*!

gwo go (26)

* *

*

*!

St. Lucian constraint ranking –E(CD]w) >> P(LC) >> –EC2 >> P(TBCD), P(LA) >> –E(LA)

Comparison of the contrast rankings in (24) and (26) for the two FLC highlights an important distinction vis-à-vis the role of markedness in each; for Haitian, effort reduction takes the form of generalized consonant-vowel coarticulation, whereas in St. Lucian, it is accomplished by the removal or reduction of constriction in the context of temporally overlapping gestures. The phonological distinction between /ʁ/ and /l/ derives from distinct effort reduction constraint rankings, affecting specific gestures in pre-vocalic contexts, but which treat all multiple constrictions as the same in postobstruent, final contexts.

.

Discussion and conclusion

The present work reexamines liquid phonology in Haitian and St. Lucian, based on the assumption that categorized effort reduction is learned (and therefore subsumed under grammatical competence), serving as the catalyst for output variation and surface distinctions noted in the two FLC. Foundational to the approach taken here is the AP conceptualization of gestures as the minimal particles or units of speech, as well as the formalization of more and less effort-inducing gestural scores. Sequences of gestures which are judged to be more effortful and assumed to be learned and subsequently categorized as such by language users are subject to –E markedness

 Eric Russell Webb

constraints, which militate against more effort-inducing forms in the output. Competition between such constraints, as well as their interaction with faithfulness or gestural preservation, is captured in relative rankings. From these rankings, comparison between Haitian and St. Lucian reveals that the former has a relatively highly ranked constraint promoting regressive coarticulation, whereas a constraint promoting effort reduction via multiple constriction avoidance in prevocalic clusters is higher-ranked in the latter. Both grammars establish highly ranked effort avoidance relevant to post-obstruent gestural constriction, eliminating tautosyllabic vowel-consonant-liquid sequences. The approach taken here presents several advantages over previous explanations of liquid phonology. Firstly, it takes into account limitations on the human productive mechanism, as well as assumed cognitive categorizations of greater and lesser effort. This is advantageous as explanation in this instance rests on principles external to phonological grammars, which are assumed to be indirectly reflected therein. Secondly, the constraints in which effort reduction are captured allow for the indirect inclusion of phonetic principles in phonology and their relative and variable ranking allows for cross-linguistic comparison of the role and content of effort reduction in phonological grammars. As such, explanation is also provided for the distinctions noted between rhotic allophony in Haitian versus St. Lucian. Finally, the present approach has made a rigorous distinction between synchronic and diachronic processes and has avoided the pitfalls of antecedent analyses, many of which posit lexifier (i.e., French) representations as subjacent forms or as grammatical input. Naturally, the present work has not answered all questions relevant to liquid phonology in either FLC or in FLC, in general. Specifically left unanswered at this stage of analysis is the question of why tautosyllabic, post-vocalic (or coda) /ʁ/ is subject to historic erasure and why /l/ is similarly unaffected. The present analysis has attributed differences in the patterning of /l/ and /ʁ/ to their gestural specification, specifically the type of constriction implied by each; clearly there is more to be said with regard to non-obstruent consonants in this regard, also taking into account the glides or semi-vowels /j/ and /w/. Additional research is also warranted regarding the phonetic form of the /ʁ/ phoneme in different phonotactic contexts, examining the extent to which velic closure is manifest in contexts where it is judged to be more labialized, e.g., in Haitian /ʁɔz/ → [wɔz]. This line of inquiry demands closer phonetic study of speech samples in the respective languages, focusing on both articulatory and acoustic characteristics. Finally, and perhaps most problematic for approaches such as the present, which seek to attribute surface forms to phonetically-motivated phonological mechanisms, the question as to the input or lexical representation of different forms begs to be asked, i.e., what is the psychological reality of forms containing what has been asserted here as /ʁ/ before a rounded vowel (in Haitian) or in complex onsets (in St. Lucian)? It may very well be that, for many speakers, the lexical representation of words given here as /ʁɔz/ (Haitian) or /pʁã/ (St. Lucian) should be considered as having an underlying /w/, rather than /ʁ/, regardless of theory-specific principles concerning input constraint or lexical status.

Chapter 5. Effort reduction and the grammar 

References Archangeli, D. 1997. Optimality Theory: An introduction to linguistics in the 1990s. In Archangeli, D. and Langendoen, D.T. (eds). Optimality Theory: An overview. Oxford: Blackwell. 1–32. Boersma, P. 1998. Functional phonology. The Hague: LOT. Browman, C. and Goldstein, L. 1989. Articulatory gestures as phonological units”. Haskins Laboratories Status Report on Speech Research SR-99/100. 69–101. Carrington, L.D. 1984. St. Lucian Creole: A descriptive analysis of its phonology and morphosyntax. Hamburg: Helmut Buske. Delattre, P. 1971. Pharyngeal features in the consonants of Arabic, German, Spanish, French, and American English. Phonetica 23. 129–55. Gess, R. 2003. On re-ranking and explanatory adequacy in a constraint-based theory of phonological change. In Holt, D.E. (ed.). Optimality Theory and language change. Dordrecht: Kluwer. 67–90. Gess, R. 2004. Phonetics, phonology and phonological change in Optimality Theory: another look at the reduction of three-consonant sequences in Late Latin. PROBUS 16.1. 21–41. Haspelmath, M. 2004. Does linguistic explanation presuppose linguistic description? Studies in Language 28.3. 54–79. Hayes, B. 2004. Phonetically driven phonology: The role of Optimality Theory and inductive grounding. In McCarthy, J.J. (ed.). Optimality Theory in phonology. London: Blackwell. 290–309. Hayes, B., Kirchner, R. and Steriade, D. 2004. Phonetically based phonology. Cambridge: Cambridge University Press. Grammont, M. 1933. Traité de phonétique. Paris: Delagrave. Kager, R. 1999. Optimality Theory. Cambridge: Cambridge University Press. Kirchner, R. 1998. An Effort-Based Approach to Consonant Lenition. PhD thesis, UCLA. Ladefoged, P. and Maddieson, I. 1996. The sounds of the world’s languages. London: Blackwell. Lindblom, B. 1990. Explaining phonetic variation: A sketch of the HandH theory. In Hardcastle, W.J. and Marchal, A. (eds). Speech production and speech modeling. Dordrecht: Kluwer. 403–39. Lindblom, B., Guion, S., Hura, S., Moon, S-J. and Willerman, R. 1995. Is sound change adaptive? Rivista di Linguistica 7.1. 5–37. Martinet, A. 1964. Economie des changements phonétiques: traité de phonologie diachronique. Bern: Francke. McMahon, A. 2000. Lexical phonology and the history of English. Cambridge: Cambridge University Press. Mielke, J. 2004. The emergence of distinctive features. PhD thesis, the Ohio State University. Myers, J. 1999. Lexical phonology and the lexicon. MS, Rutgers Optimality Archives 330-0699. Nikiema, E. 2002. Phonétique et phonologie des créoles. Les (parlers) francais ont le schwa, les créoles on l’R. In Bavoux, C. and de Robillard, D. (eds). Linguistique et créolistique: Univers créole 2. Paris: Anthropos. 77–96. Passy, P. 1891. Etude sur les changements phonétiques et leurs caractères généraux. Paris: Firmin-Didot. Penke, M. and Rosenbach, A. 2004. What counts as evidence in linguistics? An introduction. Studies in Language 28.3. 480–526. Peperkamp, S. and Dupoux, E. 2003. Reinterpreting loanword adaptations: the role of perception. Proceedings of the 15th International Conference of Phonetic Sciences 367–70. Prince, A. and Smolensky, P. 1993. Optimality Theory: constraint interaction in generative grammar. MS, Rutgers Optimality Archive 537-0802.

 Eric Russell Webb Rialland, A. 1994. The phonetics and phonology of extrasyllabicity in French. In Keating, P. (ed.). Phonological structure and phonetic form. Cambridge: Cambridge University Press. 136–59. Russell Webb, E. 2002. The relational /R/: three case studies of rhotic integrity and variation. PhD thesis, University of Texas at Austin. Russell Webb, E. forthcoming a. Explanatory adequacy: A deductive approach to /r/ in French. Linguistica Atlantica. Russell Webb, E. forthcoming b. Formalizing Creole sound change: An Optimality Theoretic account. Journal of Pidgin and Creole Linguistics. Steriade, D. 2001. The phonology of perceptibility effects: the P-map and its consequences for constraint organization. MS, UCLA. Tinelli, H. 1970. Generative phonology of Haitian creole. PhD thesis, University of Michigan. Tinelli, H. 1981. Creole phonology. The Hague: Mouton. Tranel, B. 1987. The sounds of French. Cambridge: Cambridge University Press. Wright, R. 2001. Perceptual cues in contrast maintenance. In Hume, E. and Johnson, K. (eds). The role of speech perception phenomena in phonology. San Diego: Academic Press. 251–77. Wright, R. 2004. A review of perceptual cues and cue robustness. In Hayes, B., Kirchner, R. and Steriade, D. (eds). Phonetically based phonology. Cambridge: Cambridge University Press. 34–57.

 

 

Reflexivity in Capeverdean Predicate properties and sentence structure Maria Alexandra Fiéis and Fernanda Pratas Centro de Linguística da Universidade Nova de Lisboa With most Capeverdean verbs, reflexive contexts are not marked morphologically by a se-clitic of the type existing, for instance, in European Portuguese: Pedru laba (Capeverdean)/O Pedro lavou*(-) (European Portuguese), meaning ‘Peter washed (himself)’. We propose that in Capeverdean no functional projection is available to accommodate this reflexive marker. This proposal is supported by independent evidence favoring a negative value of the Split, IP parameter (Pratas 2006). In European Portuguese, on the contrary, an expanded functional domain allows for the se-type morpheme, which is inserted post-syntactically. The reflexive reading is obligatory for most Capeverdean transitive verbs occurring in mono-argumental entries. We propose that this unique argument is generated internally and moves to subject position, being allowed to receive two theta-roles. Only a small set of Capeverdean verbs require a SELF-type anaphor for a reflexive reading to be available. This is the case of mata ‘kill’: Djon mata *(si kabesa). ‘John killed himself ’.

. Goals and approach Reflexive contexts in Capeverdean1 (henceforth CV) can surface in two different forms: either [DP V] or [DP V  kabesa]. This SELF-type reflexivity marker, however, is not very productive. Furthermore, there is no se-type reflexivity morpheme available in the language. This paper has three main goals.2 Firstly, we show that in most contexts of reflexivity (the ones with the form [DP V]) the reflexive interpretation depends on the semantic

. Variety of Santiago Island: the data that support this work have been gathered particularly in Flamengos, and Picos de Santa Catarina. . This text is a revised version of the paper we presented at the 2004 meeting of the Associação Portuguesa de Linguística (Porto, September 2004) and at the 2005 SPCL meeting (Oakland, CA, January 2005). We thank the audiences in these meetings, three anonymous reviewers,

 Maria Alexandra Fiéis and Fernanda Pratas

properties of the predicates. Secondly, we defend that the language sentence architecture is coherent with the absence of any se-type morpheme: as proposed in Pratas 2002, 2004 and 2006 (cf. Baptista 1997, 2002), in CV there is only one functional projection, a Tense Phrase (TP) – we shall describe this below in greater detail – which cannot accommodate this type of reflexivity marker. And thirdly, in line with a Distributed Morphology approach (Embick and Noyer 2001; Halle and Marantz 1993), we point out syntactic constraints, and also principles from semantics, both lexical (argument structure and θ-grid of predicates) and compositional (sentence interpretation), in order to account for phenomena which, as we hope to show, occur after syntax. For the sake of argumentation, we have made the option of comparing these CV phenomena with the correspondent data in the lexifier of this Creole language, European Portuguese (henceforth EP). We shall see that there is a particularly relevant difference in the sentence architecture of the two languages: in contrast with CV, in EP it is widely assumed that there are distinct functional projections under an Inflectional Phrase (IP) (AgrP, AspP, NegP, among others3). The line of reasoning goes as follows: whereas in CV there is no functional slot to accommodate a se-type morpheme, in EP this slot is available and, thus, this morpheme plays a role as far as reflexivity is concerned.

. The data, the problem and a possible proposal In this section we introduce the relevant data in CV and their counterparts in EP (for the sake of clarity, English translations are provided, although English reflexive contexts are not at stake in the present paper). As we shall observe, the reflexive clitic of the se-type (cf. with the non-clitic anaphoric expression of the SELF-type, a si mesmo ‘himself ’) does not have a match in CV. Consider the following examples with human subjects.4 In (a) we have the transitive entries and in (b) the reflexive ones: (1)

a.

b.

Pedru laba5 mo. Pedru wash hand ‘Pedru washed his hands.’ Pedru laba./ O Pedru lavou-se. Pedru wash/ The Pedru wash..-3 ‘Pedru washed (himself).’

(CV)

(CV)/(EP)

the editors of the present volume, and in particular João Costa, for their insightful comments and suggestions. . On how Capeverdean functional domain accommodates these grammatical values, cf. Pratas 2006. .

These subjects can be interpreted either as agents, or as experiencers.

.

The absence of any TMA marker generally means a Past Perfect reading in CV.

Chapter 6. Reflexivity in Capeverdean 

(2)

a.

b.

(3)

a.

b.

Pedru fri-l. Pedru hurt-3 ‘Pedru hurt him/her.’ Pedru fri./ O Pedru feriu-se. Pedru hurt/ The Pedru hurt..-3 ‘Pedru got hurt.’ Djon perdi dinheru. Djon lose money ‘Djon lost his money.’ Djon perdi./ O Djon perdeu-se. Djon lose/ The Djon lose.-3 ‘Djon got lost.’

(CV)

(CV)/(EP)

(CV)

(CV)/(EP)

The Capeverdean example in (2b) would also be possible with the reflexive expression – Pedru fri si kabesa –, but this would mean that he hurt himself on purpose, which would have consequences on the subject θ-role: in this case it would be an agent, whereas in (2b) it is an experiencer. Now consider constructions with non-human subjects: (4)

Porta abri./ A porta abriu(-se). door open/ The door opened(-3) ‘The door opened.’

(5)

Es vidru li dja kebra.  glass   break Este vidro acaba de se partir. This glass finishes of 3 break ‘This glass just broke.’

(CV)/(EP)

(CV) (EP)

We might consider two possible accounts for this reflexive strategy in CV. Either CV does not have a se-type morpheme in the numeration, hence it cannot use it, or, following the proposal in Levin and Rappaport Hovav 1995 for English, lexical semantic properties are responsible for the different entries of some predicates. This kind of variation found in English also occurs in a crosslinguistic perspective, as shown in Fiéis 2003: (6)

a. b.

Jean a brûlé la maison. ‘John burnt the house.’ La maison a brûlé. ‘The house got burnt.’ (lit. The house burnt)

(7)

a.

Sortez la voiture du garage. ‘Take the car out of the garage.’ (lit. Exit the car of the garage) Jean est sorti. ‘John left.’

b. (8)

a.

b.

Personne ici ne serait capable de me tomber. ‘Nobody here would be able of putting me down.’ (lit. Nobody here would be able of falling me) Jean est tombé. ‘John fell.’

 Maria Alexandra Fiéis and Fernanda Pratas

Reinhart (1996, 2000, 2001) also assumes a lexicalist perspective for English. According to this account, some English verbal entries undergo lexical reduction. For instance in Max washed the reduction of the internal argument results in a reflexive reading; in Lucie rolled the reduction of the external argument results in an unaccusative entry. A similar proposal has been adopted in Pratas (2002, 2004) for the above CV data. Nevertheless, this analysis does not account for such cases as those in EP, where a se-clitic is realized and has been simply (and not satisfyingly) assumed as a trace of the accusative case.

.

Our proposal

.

Assumptions

.. Distributed Morphology approach (DM) The contrast between some reflexive entries in different languages suggests that the relevant “visible” phenomenon – presence-absence of a se-morpheme – is determined in the morphological (post-syntactic) module of the grammar. According to DM, in the syntax some terminal nodes are generated and moved by syntactic rules and constraints. The result of this constitutes the input of a morpho-phonological domain, in the Phonological Form (PF) branch of the grammar (the other branch is the Logical Form, LF). A morpheme is an atom of the morpho-syntactic representation. It is the terminal node and its content, and not the phonological expression of this terminal (this one is a Vocabulary Item). The content of an abstract morpheme, active in the syntax, is a bundle of abstract (syntactic-semantic) features extracted from an available set in Universal Grammar (UG). .. The split, IP parameter: positive in EP, negative in CV The Split, IP parameter (as proposed in Bobaljik 1995; Bobaljik and Thráinsson 1998; Thráinsson 1996), has a positive value in EP (IP is split in some other functional maximal projections, as TP, AgrP, AspP, etc.), and a negative value in CV. In CV, the agreement, mood or aspect features surface in none, one or some different Vocabulary Items, under the T (Tense) label. In CV (Pratas 2002, 2004, 2006) there is no empirical evidence to project any category other than T in the functional domain. In Pratas (2006) it has been shown in greater detail that for CV: (1) there is no evidence – provided neither by any overt agreement morphology nor by any sort of positional arguments – for positing an AgrSP node; (2) TMAs appear in sequences unbreakable by any intervening element (no XP may disrupt their adjacency). The same TMA morpheme may instantiate different meanings, depending on distinct factors, namely: (1) combination with other markers, expressions and discourse information, and (2) semantic properties of predicates. Hence, it is proposed that TMAs surface in successive head adjunction under the

Chapter 6. Reflexivity in Capeverdean 

label T. This label also accommodates the negative marker ka, which displays a head behavior, and also the subject clitic.

.. θ-roles (theta-roles or thematic roles) as features Hornstein 1999’s approach to θ-theory is crucial to our work. The author claims that θ-roles are features that license movement. Thus, this allows for the kind of representation we will be proposing ahead (see, for instance, (14)), in which a DP can receive more than one θ-role contradicting the θ-criterion. According to the θ-criterion, each argument cannot receive more than one θ-function, and each θ-function in an argument structure is assigned to only one argument. This disallows phrase movement from a θ-position to another. In other words, assuming Horstein’s rejection of this constraint, a DP can receive more than one θ-role. We will see that the data seem to favor this hypothesis. This proposal of double θ-role assignment to a single argument has been independently reached in Reinhart and Siloni (2005), where this operation is called bundling. Their analysis considers that there is a lexicon-syntax parameter: in some languages there is a lexical reflexivization, which bundles (unifies) two clusters of features (θ-clusters), resulting in a complex θ-role, and reduces the accusative Case feature of the verb (this is the case of English, as it could be the case of Capeverdean); in other languages, where the reflexivization is syntactic, there is no lexical reduction and so the verb still carries the Case feature, which then relies on the role of a clitic like se (this is the case of Romance languages, among others, and thus could be assumed as the case of Portuguese). This clitic is also considered by these authors as being non-argumental (as we do in the present paper), hence it cannot be assigned the “extra” θ-role; this must be handled also by an arity operation – the same bundling, which in this case occurs at a different level. .

Sentence structure

In EP there is a functional head (Refl, for some cases, Cause or Voice, for others) where the se-morpheme is inserted, in the morpho-phonological module. This allows for a unified analysis of the se-clitics – we propose that they are non-argumental, whereas previous analyses (Cinque 1988; Duarte 2001) considered some of them argumental (like in O Pedro lavou-se ‘Peter washed’) and some non-argumental (A porta abriu-se ‘The door opened’). In CV, no functional category other than T is projected. Hence there is no slot available for any se-morpheme. The interpretation of the sentence depends not only on the semantics of the predicates but also on the sentence composition (a reflexive reading, for instance, is the only one we obtain whenever there is no internal argument for a transitive verb).

.

Consequences of our proposal

The so-called reflexive reading, (i) Pedru laba (‘Pedru washed’), the inherent reading, (ii) Ana xatia (‘Ana got angry’), and the anticausative-unaccusative reading, (iii) Porta abri

 Maria Alexandra Fiéis and Fernanda Pratas

(‘The door opened’), are possible in CV in mono-argumental entries of some predicates, depending on its semantic properties. The relevant contrast is between the reflexive reading we obtain with laba (‘wash’), as in (i), and a verb like mata (‘kill’). In CV the latter needs a SELF-anaphor for the reflexive reading to be possible: Djon mata *(si kabesa) (‘John killed *(himself)’). In other words, the subject in (i) is assigned both the external θ-role, Agent, and the internal θ-role, Theme (these verbs assign these roles to different arguments in a regular transitive entry). In (ii), only the internal θ-role, Experiencer, is assigned. Although we cannot have a passive construction in EP (*A Ana foi zangada), we have a semantic information about a cause or an external agent not given in the sentence. This becomes clear by the counterpart in English, where a get-construction is needed. This is not, then, a case of inherent reflexivity, which would require an external argument, with a typical Agent θ-role. In (iii), only the internal θ-role, Theme, is assigned. Note that, by contrast with “real” unaccusatives, this type can have transitive entries. Now consider the θ-grids we propose for these verbal entries: (9)

Pedro laba./O Pedro lavou-se. ‘Peter washed.’

(CV)/(EP)

1 

2 

i

i

In Pedru laba, just like in O Pedro lavou-se, θ-roles 1 (external argument) and 2 (internal argument) merge in the subject. In reflexive entries the semantic representation shows that both arguments coincide: Pedru λ x [x laba x]. We could test the assignment of the Agent θ-role to the external argument of the verb laba ‘wash’ by introducing in the sentence, in adjunct position, an adverbial oriented to the Agent, like di xokota ‘deliberately’. However, we should say that this particular entry, laba, contains an inherent information of this intentionality. But this same test works for other transitive verbs, like xuxa ‘get dirty’: Pedru xuxa di xokota ‘Peter got dirty deliberately’. In (10), Ana xatia, only θ-role 2 is assigned. Although we cannot have a passive structure with the verb xatia (ex. Ana pintxadu ‘Ana was pushed’/*Ana xatiadu *‘Ana was angried’), we have a semantic information concerning a cause or external agent not expressed in the sentence: something or someone made her angry. (10)

Ana xatia./A Ana zangou-se. ‘Ana got angry.’ 1

 or  i

(CV)/(EP) 2 / i

Chapter 6. Reflexivity in Capeverdean 

In (11), Porta abri, only θ-role 2 – Theme, in this case – is assigned. These predicates, are the so-called verbs of causative alternation. (11)

Porta abri./A porta abriu-se. ‘The door opened.’

(CV)/(EP)

1

2 

 or 

i

i

.

Syntactic representations

The following syntactic diagrams stand for examples in EP – (12) and (13), and CV – (14) to (17). In (12), the DP O Pedro receives two θ-roles: the Theme θ-role is assigned in its base position (internal argument); the DP moves afterwards to Spec, VP, where it absorbs the Agent θ-role. This constituent, now with two θ-roles, moves further to Spec, IP, where nominative case is assigned. Given the availability, in EP, of a functional node, the se-type clitic is inserted postsyntactically in the morphophonological component of the grammar. This se-type morpheme is co-referent with the subject and must have its morphological shape for person and number (-me ; -te ; -se ; -nos ; -vos ; -se). Like in (12), in (13) this se-type clitic is a morpheme co-referent with the subject and it is inserted post-syntactically in the morphophonological module of the grammar. Nevertheless, in (13) there is only one θ-role assigned to the DP, the internal one (Theme, in this case). There is also no violation of the Case Filter: the internal argument is obligatorily moved to Spec, IP, where it gets nominative case. This is what happens in passives. AgrP

(12) DP

O Pedro Agr q1 ; q2 ; Nom

Agr’ IP I’ I

ReflP Refl’ Refl

VP

se-type DP clitic t lavar q1; (q2)

V’ t q2

 Maria Alexandra Fiéis and Fernanda Pratas

An analysis under a compositional semantics point of view shows that this is not an unaccusative entry: semantically we have intuitions about the relation (established by the predicate) between the only argument and some Cause or Agent (external θ-roles) not expressed in the sentence. Their existence is however suggested by the Portuguese clitic se. According to Kayne (1988) and Pesetsky (1995), the se clitic signals the presence of the external argument in the proper syntax. This is an argument in favor of an unaccusative analysis for these constructions. In our proposal, the se morpheme is inserted post-syntactically. We consider that the only valid argument for taking these constructions as unaccusative is the fact of them having a derived subject (for more arguments against the view of these constructions as unaccusatives, cf. Reinhart and Siloni 1999). AgrP

(13) DP

A Ana / Agr A porta q2 ; Nom

Agr’ IP I’ I

CauseP/VoiceP Caus’/Voice’ Caus/Voice

VP

se-type clitic

V’ V

DP

zangar/ abrir

t q2

Now consider the relevant representations for CV from (14) to (17). The reflexive reading in (14) depends on the contrast with the expected structure of a transitive verb like in (15). The absence of a DP – in the position of internal argument – represented as a trace of the deleted DP – provides the only possible interpretation: the one in which subject and object coincide. The DP subject has been moved to Spec, TP (the only functional projection in CV) carrying the complex θ-function Agent+Theme, where it gets nominative case. A se-type morpheme is not inserted, since there is no functional head available.6 . An anonymous reviewer has pointed out to us that the Split, IP parameter is not related to the availability of other functional projections such as ReflP or CauseP, where the clitic

Chapter 6. Reflexivity in Capeverdean 

A potential problem for this analysis is the accusative case feature of the verb: if the DP is assigned accusative case in its base (internal) position, it would be prevented from moving to a different case position (Spec, TP). However, if we consider that these entries do not select for a DP in their base external position, we may conclude that they are not able to assign accusative case. This follows one basic idea in Burzio’s Generalization: “All and only the verbs that can assign a θ-role to the subject can assign accusative Case to an object” (Burzio 1986: 178). The relevant relation at stake here is between the verb and the base external position; in these constructions, although the DP receives a subject θ-role, its base position is internal.7 (14)

TP DP

T’ VP

Pedru θ1; θ2; Nom.

DP

V’

laba t q1; (q2)

(15)

t q2

TP DP

T’

Pedru T q1; Nom.

VP DP t q1

V’ V laba

DP mo q2; Acc.

In (16), the subject(s) is(are) generated in the internal argument position, receiving the internal θ-role – Theme or Patient, depending on the semantics of the verb, as expected. However, the reading is not the reflexive one, since the subject is not an Agent (it is

se was supposed to be spelled out. We are assuming, indeed, that there are three domains in the sentence – CP, IP and VP (this last one being the lexical domain and the first two the functional domains) – and that these functional projections belong to the IP domain. . If this line of reasoning is on the right track, we might consider that this is also the case in EP – cf. (12) and (13) – and, hence, this would constitute additional evidence in favor of the non-argumental nature of the se-type morpheme. This would also support our post-syntactic insertion proposal.

 Maria Alexandra Fiéis and Fernanda Pratas

widely assumed that subjects of reflexive entries are Agents). ‘The door’ does not ‘open itself ’ (although we can say something like ‘The door opened by itself ’). Notice, however, (17), where the SELF-type anaphor si kabesa is obligatory in CV, brings more evidence to our proposal. Some language idiosyncrasy requires that the reflexive reading for the verb mata ‘kill’ be explicit in the sentence. We propose that this expression appears as an adjunct (and not the internal argument), assuring the reflexive reading, given that the absence of the internal argument of a transitive verb is not sufficient here. (16)

TP DP

T’

Ana/Porta T q2; Nom.

VP V’ V

DP

xatia/abri

t q2

TP

(17) DP

T’ VP

Pedru q1; Nom. DP t q1 V mata

VP VP

DP DP si kabesa t q2

In EP, for this verb matar, the se morpheme is enough for the reflexive reading, without the anaphoric expression; as in (12), se is inserted post-syntactically, in a functional head with the label Refl.

.

Final remarks

In this paper we have proposed that the reflexive interpretation of certain Capeverdean entries with the form [DP V] – which are, thus, not morphologically marked for

Chapter 6. Reflexivity in Capeverdean 

reflexivity – depends on the semantic properties of the predicates involved. Whereas the verb laba (‘wash’), fri (‘hurt’) or perdi (‘lose’), among many others, have an obligatory reflexive reading when no DP is spelled out in the internal position, a SELF-type anaphor is necessary for the well-formedness of reflexive sentences with some other verbs, such as mata (‘kill’): Pedru mata si kabesa (‘Pedru killed himself ’)/*Pedru mata. (*‘Pedru killed’). We have analyzed in greater detail the former type of predicates, which is much more productive in the language. We propose that the relevant property forcing this strategy in these reflexive entries – of the form [DP V] – lies at the syntactic level. The reasoning goes as follows: the value of the Split, IP parameter (Bobaljik 1995; Bobaljik and Thráinsson 1998; Thráinsson 1996) is negative in CV; its functional domain is limited to Tense Phrase (TP) (Pratas 2004, 2006); hence, there is no slot available in the functional domain for this clitic to be spelled out. We have shown that, with respect to this, there is a contrast between CV and EP; the latter language expanded functional domain allows for the late insertion, as defined within a Distributed Morphology approach (Embick and Noyer 2001; Halle and Marantz 1993), of a se-type reflexive morpheme (non-argumental). We have left open the questions about which specific syntactic role this EP morpheme plays in the structure (see Duarte 2001; Kayne 1988; Pesetsky 1995; Reinhart and Siloni 1999 for discussion on the syntactic role of this type of clitic in reflexive entries). Our proposal about CV relies on the assumption, following Hornstein (1999), that θ-roles are features that license movement. This allows for a DP to move from one (internal) θ-position to another (external), receiving more than one θ-role (the proposal of a complex θ-role for reflexive entries like English ‘wash’ has been independently developed in Reinhart and Siloni 2005). We have also argued that, since there is no DP generated in the external position (although a moved DP lands there, receiving an external θ-role, moving only then to Spec, TP, where it checks nominative case – as we can see in (14)), these entries do not assign accusative case. This implies following one basic idea in Burzio’s Generalization (Burzio 1986), and assume a dependency between an accusative case feature and the selection of a DP base generated in Spec, VP.

References Baptista, M. 1997. The morpho-syntax of nominal and verbal categories in Capeverdean Creole. PhD thesis, Harvard University, Cambridge, MA. Baptista, M. 2002. The syntax of Cape Verdean Creole, the Sotavento varieties. Amsterdam: John Benjamins. Bobaljik, J. 1995. Morphosyntax: The syntax of verbal inflection. PhD thesis, MIT. Bobaljik, J. and Thráinsson, H. 1998. Two heads aren’t always better than one. Syntax 1. 37–71. Burzio, L. 1986. Italian syntax: A Government-Binding approach. Dordrecht: Reidel.

 Maria Alexandra Fiéis and Fernanda Pratas Cinque, G. 1988. On si constructions and the theory of arb. Linguistic Inquiry 19. 521–81. Duarte, I. 2001. Tópicos de sintaxe comparada. Relatório para prestação de Provas de Agregação, Universidade de Lisboa. Embick, D. and Noyer, R. 2001. Movement operations after syntax. Linguistic Inquiry 32. 555–95. Fiéis, M.A. 2003. Ordem de palavras, transitividade e inacusatividade: Reflexão teórica e análise do português dos séculos XIII a XVI. PhD thesis, Universidade Nova de Lisboa. Halle, M. and Marantz, A. 1993. Distributed morphology and the pieces of inflection. In Hale, K. and Keyser, S.J. (eds). The view from building 20. Cambridge, MA: MIT Press. 111–76. Hornstein, N. 1999. Movement and control. Linguistic Inquiry 30. 69–96. Kayne, R. 1988. Romance se/si. GLOW Newsletter 20. Levin, B. and Rappaport Hovav, M. 1995. Unaccusativity. At the syntax-lexical semantics interface. Cambridge, MA: MIT Press. Pesetsky, D. 1995. Zero syntax: Experience and cascades. Cambridge, MA: MIT Press. Pratas, F. 2002. O sistema pronominal do caboverdiano. Variante de Santiago. MA thesis, Universidade Nova de Lisboa. Pratas, F. 2004. TP in Capeverdean: an almighty functional projection? Paper presented at the Lisbon Workshop on Alternative Views on the Functional Domain. Universidade Nova de Lisboa, July 8-9. Pratas, F. 2006. Tense features and argument structure in Capeverdean predicates. PhD thesis, Universidade Nova de Lisboa. Reinhart, T. 1996. Syntactic effects of lexical operations: Reflexives and unaccusatives. OTS working papers. Reinhart, T. 2000. The theta system: syntactic realization of verbal concepts. OTS Working Papers in Linguistics. Available for download at http://www.tau.ac.il/~reinhart/ling_dl/download.htm Reinhart, T. 2001. Experiencing Derivations. SALT lecture. New York University. Available for download at http://www.tau.ac.il/~reinhart/ling_dl/download.htm Reinhart, T. and Siloni, T. 1999. Against the unaccusative analysis of reflexives. Available for download at http://www.tau.ac.il/~reinhart/ling_dl/download.htm (later published in Alexiadou, A., Anagnostopoulou, E. and Everaert, M. (eds). 2004 Studies on unaccusativity: The syntax-lexicon interface. Oxford: Oxford University Press. 159–80). Reinhart, T. and Siloni, T. 2005. The lexicon-syntax parameter: Reflexivization and other arity operations. Linguistic Inquiry 36.3. 389–436. (version used here is the one available at http://www.tau.ac.il/~reinhart/ling_dl/download.htm). Thráinsson, H. 1996. On the (non-)universality of functional categories. In Abraham, W., Epstein, S.D., Thráinsson, H. and Zwart, C.J.-W. (eds). Minimal ideas: Syntactic studies in the minimalist framework. Amsterdam: John Benjamins. 253–81.

 

An additional pronoun and hierarchies in Lower Columbia Chinúk Wawa1 David D. Robertson University of Victoria

In this study I examine the expression of third person, especially among objects, in the extensive documentation of the Lower Columbia River variety (lcCW) of Chinúk Wawa (CW). I suggest that an additional form, null grammaticalized as a pronoun, is needed to approach descriptive adequacy. Comparing the syntactic behavior of this Ø with that of the sole previously described 3 form, yáka, I claim that these instantiate Differential Marking (Aissen 2003) along animacy and definiteness hierarchies – which are absent from previous lcCW descriptions although ubiquitous in the recorded data on this variety. I show that the key distinctive feature of third-person Ø is inanimacy, as opposed to number (in yáka as well as plural ɬáska), so it is most appropriately labeled 3. In the course of showing this, I argue for two refinements of Differential Marking theory. First, I point out that hierarchical distinctions occur not only among all NPs but also, replicated in microcosm, among pronouns and their referents. Second, I present evidence that not only objects and (by corollary) subjects are differentially marked, but also genitives. I suggest that the model for the lcCW indefinite vs. numbered third-person pronominal distinction was Lower Chinook and possibly Salish languages sharing this regional feature. I show that nulls in other regions’ CW differ in being nonobligatory, and may in fact be stylistic devices (implying the relative stability

. Acknowledgments: The Centre for Studies in Religion and Society at UVic (for a fellowship which made possible some of the work associated with this paper); Henry B. Zenk and Tony A. Johnson of the Chinúk Wawa program, Confederated Tribes of the Grand Ronde Community, Oregon (for sustained support of my research on the language); Tom Larsen; Linda Fink of Grand Ronde, Oregon; the audience at the 2006 Society for Pidgin and Creole Linguistics meeting, LSA, Albuquerque; Leslie Saxon, Ewa Czaykowska-Higgins, and Tom Hukari; Claire Turner and the other participants in Dr. Hukari’s LING 503/508 Syntax course, 2005; and the present volume’s diligent editors and reviewers. Any errors in this paper are my responsibility alone.

 David D. Robertson

of CW). The findings I present here significantly expand the description of endangered lcCW, promoting its accurate revitalization. Animacy . . . is so pervasive in the grammars of human languages that it tends to be taken for granted and become invisible. (Dahl and Fraurud 1996)

. Introduction Among the best-documented regional varieties of Chinúk Wawa (CW; synonyms Chinook, Chinook Jargon, CJ, Jargon) is that of the lower Columbia River and adjacent lands such as the Willamette Valley and Pacific coast, where this contact language can be presumed to have originated.2 I refer to the lower Columbia River-centred regional variety as ‘Lower Columbia CW’ (lcCW).3 Extensive documentation dates at least to the 1830s, when the Catholic missionaries Francis Norbert Blanchet and Modeste Demers conducted their work in CW and wrote significant amounts of material in it. Numerous authorities, including native speakers, have continued the literature on and in lcCW up to the present day, making this one of the better-documented and -described varieties of any language in north-western North America. Observable in the documentation of lcCW is the unanimity with which grammatical sketches and lexicons equate the third person singular with the pronoun yáka (and its variants ya and yáx. ka), of which (1) shows examples.4

. The earliest evidence of a speech variety recognizable as CW dates to December 10, 1805 on the lower Columbia, as recorded by Lewis and Clark (Thwaites 1959 (3): 276). The notion of CW’s pre-contact existence has attracted controversy, viz. Thomason (1983: 820–1), but confirming data are lacking. . The data used here are drawn from confirmed L1 speakers from the Grand Ronde, Oregon community (cf. Hymes and Hymes 1972; Zenk 1984) as well as from other lowerColumbia areas probably including L2 speakers. A laundry list of sources confirm that L1 and L2 lcCW were spoken alongside one another for decades (cf. Demers, Blanchet and St. Onge 1871 [1838]: 8; Hale 1968: 644, 1890: 19–20 [both ca. 1840]; Lewis and Murakami 1990 [referring to circa 1832]; Macdonald 1892 [referring to the 1830s, cited by Zenk 1984: 33]; Palmer 1906: 180 [mid-1840s]; and Swan 1989 [1857]: 199). Both are identical in nearly all features, including the pronoun distribution which is the subject of the present study. The primary difference was that (Grand Ronde) L1 lcCW was incipiently developing a few new inflectional morphemes: distributive reduplication and prefixes for e.g. progressive aspect (cf. Grant 1996). In the present study I remain agnostic as to the date(s) and location(s) of CW’s creolization. . Cf. Jacobs’ (1932: 36–8) description of the latter two as ‘tied’ (~bound) and ‘independent’ respectively. I will use yáka as the umbrella form in my discussion.

Chapter 7. Lower Columbia Chinúk Wawa 

(1)

5 álta ya ískam yáxka now 3 take 3 ‘Then he seized her. . .’

Victoria Howard (Jacobs 1936: 2)

Demers, Blanchet, and St. Onge (1871 [1838]), in the earliest discussion6 of the lower Columbia variety’s grammar, define the pronoun ïaka as ‘he, him, his’ (p. 26): as they note (p. 11) “Of the Nouns” and pronouns, these “are not declined; they have no gender nor number”.7 Horatio Hale (1968 [1846]), going into more detail, uses similar wording, “The personal pronouns are – . . . iáhka, he” (p. 641), also glossed as ‘he, she, it’ (p. 637), a description he retains in a revised presentation of his findings (Hale 1890). Melville Jacobs (1932) differs from the preceding mainly in including the short variant ya (p. 36) for ‘third person singular personal pronoun’. Anthony Grant (1996: 233) picks up the theme in his pronominal chart, where ‘3’ is yáxka, yáka [‘Emphatic’] and ya [‘Reduced’]. In summary, the sources form a tradition of CW grammatical description in which the way to express the third person singular is yáka and its variants. While it is true that all other regional varieties of CW normally (see discussion of pidgin CW, Section 7) have only yáka available to express all third person singulars, voluminous data challenge the above descriptions of lcCW. Example (2) reproduces a sentence from the earliest of my sources, together with my gloss and translation. (2)

wík mkhmәk úkuk pus msáyka mkhmәk áɬqi msáyka mímlus  eat   2 eat  2 die (‘Do not eat this [the apple]; if you [Adam and Eve] eat it, you will die.’) Demers, Blanchet, and St. Onge (1871 [1838]: 52)

. In citing examples of Chinúk Wawa usage herein, I preserve the original orthography to the greatest possible extent, thus their presentation within angled brackets. However, I have had to replace the ‘broken h’ character of Demers, Blanchet, and St. Onge (1871 [1838]) with χ. Jacobs’ (1932, 1936) ‘g with subscripted dot’ is here replaced with ġ. All example sentences are given glosses not only in English but also from the current Confederated Tribes of Grand Ronde dictionary (Zenk and Johnson 2003a). The orthography these use is essentially Americanist, with the following practical substitutions: ch = č, c´h = č’, sh = š, tɬ = λ, t´ɬ = λ’, ts = c, t´s = c’; dorsal followed by w = labialized; x. stands for uvular voiceless fricative. All morphemic glosses are supplied by me, as are all translations enclosed in parentheses. Abbreviations used: 1 first person, 2 second person, 3 third person, ani animate, cau causative, cnj conjunction, def definite, dem demonstrative, dist distributive, dscm discourse marker, fut future, incp inceptive, indef indefinite, irr irrealis, mod modal, neg negative, obj object, pl plural, prep preposition, sg singular. . This book’s treatment of lcCW grammar is admittedly quite brief; a dictionary and texts make up the bulk of the volume. . A reasonable inference is that the authors translated ïaka with the default ‘neutral masculine’ of European grammatical traditions.

 David D. Robertson

Context, to say nothing of the familiarity of this story, makes clear that a definite object is intended, yet no overt pronoun is found in object position, that is, immediately following the verb. (For clear exposition I underline the object in the English translations here.) This does not appear to be an isolated typographical error, either, because parallel examples such as (3–4) are found throughout the literature. (3)

pus íkta ya pálach yáx. ka ya ískam ya munk- wá.x  thing 3 give 3 3 take 3  pour h lamiyáy khapa yáx. ka k apa uk  3   old.woman ‘Whatever she gave him, he took [it], he threw it back at her, at the old woman.’ Victoria Howard (Jacobs 1936: 2)8

(4)

yaka lúlu sayá pi ya másh khapa ts qw uk lisák 3 carry far CNJ 3 throw  water  bag (‘He carried it [cat] far away and threw the bag into the water.’) Petite (n.d.) (‘Eeh ulman ya towen. . .’)

Not only do such sentences with null object pronouns abound, they appear to constitute a norm in this CW variety. A revision of previous accounts of CW grammar is called for, both in order for linguists to accurately understand this language and for current community revitalization efforts to be based on a model true to ancestors’ usages. Toward these aims, in this paper I will demonstrate the existence of animacy and definiteness as grammatical categories in the lcCW third person, in a corpus from some of the earliest records onward. I will show that null pronouns are obligatory for certain third person objects, constituting an additional pronoun in this variety. (I will provisionally assign this Ø a neutral gloss as ‘3’, in opposition to ‘3’ for yáka.)9

. By way of explaining the gloss I have provided, I note that Jacobs’ form wa´x. suggests the CW verb wáx. ‘to spill’ (Zenk and Johnson 2003a: 59), but the sense here allows a reading wә´x. , a variant of wә´x. t ‘again’. Thus munk-wә´x. ~ ‘give back’. The reading remains uncertain, however, as that form is unattested elsewhere; Zenk (p.c. August 18, 2005) sees this speaker (Victoria Howard) as unlikely to use the shortened form of wә´x. t. In any case, nothing in my paper turns on this point. . Tom Larsen (p.c. May 24, 2005) notes the possibility of analyzing many of lcCW’s 3sg subject constructions as involving a null, if what I here term the overt pronoun ya(ka) is seen as a sort of verbal agreement marker distinct from the pronominal system. Other considerations are then likely to come into play, for a sampling of which the reader is referred to Larsen (2002). In any case, such an analysis would not affect the claim of the present study, viz. that a null third-person pronoun having properties of inanimacy and indefiniteness must be recognized in lcCW.

Chapter 7. Lower Columbia Chinúk Wawa 

The remainder of this paper will be structured as follows: After a sketch of my methodology in Section 2, Section 3 will contrast Ø and yáka expressions of third-person object, leading to an examination of the range of senses of Ø in Section 4. Notes on implications of gaps, some imposed by unforeseeable sociolinguistic factors, in the data will be given in Section 5. Following this in Section 6 I will examine third-person exponences in various syntactic positions (ergo roles, in a configurational and analytic language such as lcCW), arguing that hierarchy effects are observable throughout. I will draw conclusions and suggest future research directions in Section 7.

.

Methodology

As virtually no speakers who acquired lcCW as a childhood home language are left, my methodology centred on already collected materials. The available sources of textual and grammatical data on this language variety were collected in a literature search and through personal contacts with the Native community and with scholars who work on CW.10 Unpublished data included Henry B. Zenk’s field notes and audio recordings of elders from the Grand Ronde Indian Reservation, which are of unparalleled value in establishing that the patterns observed throughout the corpus reflect actual usage. Zenk’s (1984) dissertation on the linguistic history of that speech community, as well as his and Tony A. Johnson’s (2003a) dictionary produced for tribal members’ use, were also highly useful. Another tribal source was elder Eula Petite’s lcCW class handouts, provided by her friend and student Linda Fink. Published sources for lcCW included the grammatical sketches, ranging from fleeting to fairly detailed, by Demers, Blanchet and St. Onge (1871 [1838]); Grant (1996); Hale (1968 [1846], 1890); Jacobs (1932); and Larsen (2002).11 Sources of example usages rich enough to rival Zenk’s field notes included Demers, Blanchet and St. Onge (1871 [1838]); and Jacobs (1936). From these sources I assembled a corpus, selecting relevant forms based on my knowledge of lcCW as well as considering any English (etc.) glosses provided. These glosses

. A point which is important for an institutionally-affiliated researcher in Canada to establish is that ethical considerations were a minor issue since I worked with documents only. All of these were either publicly available or provided with the explicit permission of the pertinent researcher. . Demers and Blanchet learned CW at Fort Vancouver (present-day Vancouver, Washington; 1871 [1838]: 8); St. Onge learned it among the Yakama people upstream on the Columbia (cf. Kamloops Wawa 1901b). Jacobs obtained most of his CW data from “the best Jargon speaker. . . Mrs. Victoria Howard”, a member of the Grand Ronde L1 community, so his study is “based mainly on her dictations and [is] valid only for the Jargon spoken in the northwest Oregon region” – i.e., lcCW (1932: 28, emphasis added).

 David D. Robertson

obtained and/or provided by past researchers were assumed to contain potentially irreplaceable information about meanings intended by lcCW speakers, nearly all of whom are deceased. Reference to glosses was intended to be judicious, with the data itself of course assigned greatest weight. It will be shown in Section 3 that English glosses have provided previous researchers with a misleading idea of the nature of lcCW third-person pronouns, a misconception which the present study aims to correct. The corpus was created in the form of a Microsoft Excel 2003 database in which I entered all examples of sentences including third-person pronominal objects. Entries were divided among separate sheets, one for each exponence of third person (yáka, Ø, code-switched ‘her’, etc.). Each example was tagged in separate fields with an English translation, citation of its exact source, and the name of the lcCW speaker who produced it, if known. Patterns were extracted from the database, in the form of correlations between a given pronominal exponence of third person and a given gloss (e.g., ‘he’, ‘it’, ‘some’, etc.). These correlations were compared and contrasted with one another by syntactic tests of commutability among exponences and among roles.

.

Expressing third person with and without yáka: Animacy hierarchies

Examples (2–4) imply that objects glossed in English as neuter-gender ‘it’ are the third persons which are represented by Ø in Chinúk Wawa. The data, as in (5, 6), again show that the English glosses are not the crucial metric and that a more nuanced description based on observed CW facts is needed.12 (5)

. álta ntsáyka munk- xl úkuk xá . úyma . wqaɬ impossible now 1  different  ‘We cannot make it different now.’ John B. Hudson (Jacobs 1936: 19)

(6)

ya wáwa yá.xka máyka munk- k’ílapay khapa ntsáyka uk lamiyáy 3 say 3 2  return  1  old.woman ‘He said to it, “Return the old woman to us” .’ Victoria Howard (Jacobs 1936: 3)

. Omitted from consideration are all cases where a gloss ‘it’ is clearly completely spurious or where it was felt impossible to determine the presence of Ø. Four patterns were apparent: 1. kánawi ‘all’ is given a gloss ‘it all/all .of it’ in six examples. 2. úma ‘to feed’ is glossed as ‘give it to’ (or ‘give them to’ – cf. §4) in nine. 3. wáwa ‘speak/say’ is glossed as ‘tell it, say it, talk about it, speak it, tell about it, talk it up’ in six. 4. kákwa ‘thus(ly)’ is glossed as ‘(do) it/that, (remember) it, (want) it, (tell) it, (talk) it (up)’ in nine.

Chapter 7. Lower Columbia Chinúk Wawa 

Such examples demonstrate the lack of predictive power of an English-inspired description equating gender ([±neuter]) with CW null 3. It should be reiterated that there is no grammatical reflection of natural gender in this language’s pronouns (already demonstrated in 1, repeated here as 7) any more than there is in the nouns, as mentioned above. (7)

álta ya ískam yá.xka now 3 take 3 ‘Then he seized her. . .’

Victoria Howard (Jacobs 1936: 2)

In addition, the preceding examples show that other broadly pronominal morphemes (the presumably non-neuter yáka and the presumably ungendered ) can express the ‘it’ of English. Both of these elements will be touched on again in a later section; for the moment it suffices to explain that the object úkuk in (4) is half the time (six of 11 occurrences found) given the more literal English gloss ‘that, this’. This  appears to be best understood as ‘it’ only in limited circumstances, those where its antecedent is taken as given but not specific information which must be further specified, if not by overt naming then by deixis. Crucially, in all cases object úkuk can equally well be understood as ‘this, that’, and (8a) versus (8b) demonstrates that once the antecedent is thus clarified, null is used in any further reference to it. (8a)

t´ɬúnas na m´әkhmәk úkuk, wík na m´әkhmәk Ø. uncertain 1 eat   1 eat 3 ‘I don’t know but what I ate it, but what I didn’t eat it.’ lit.: ‘I don’t know but what I ate that, but what I didn’t eat it.’ Wilson Bobb (Zenk field notes 8/14/83)

(8b)

na m´әkhmәk úkuk, wík na m´әkhmәk úkuk.) (*? t´ɬúnas uncertain 1 eat   1 eat 13

Referring momentarily back to (6), the mismatch between the CW yáka (elsewhere ‘he’, ‘she’) and its gloss ‘it’ derives from more mundane considerations. In that example the reference is to a supernatural being which turns out to resemble humans enough to be considered animate: It not only has at least one arm and leg but even speaks elsewhere in the same text. Talking “animal people” characters in myth texts (such as Fly, in Petite n.d. [Miga liplip lasäp. . .]) are treated as animate and referenced with yáka for the same reason. As seen above, yáka is also used for nonhuman, nonmythical beings,

. The asterisked form in (8b) is obviously not an acceptability judgments elicited by me from this deceased lcCW speaker. Instead it represents something more general: The absence of attested ‘either-or’ structures in which a demonstrative object is overtly repeated. Tom Larsen (p.c. May 24, 2005) suggests this second version of this sentence is not so much ungrammatical as “stylistically awkward”.

 David D. Robertson

e.g., cats, dogs, cows, and horses, to which such humanlike activities as speech are not attributed, but which can, again carefully invoking a comparison with English, be referred to as either ‘it’ or ‘he/she’. Apparently these beings are perceived as individuated and as possessing personalities, thus distinct from inanimates. Less individuated are the lower nonmythical animals such as the snake of (9), and the very young human baby(!) of (10).14 (9)

(10)

. . Ø qhá álta ya ískam uk úlq’ ya munk- míɬayt now 3 take  snake 3  be.located 3 where uk ya x.úmx. um  3 brain ‘He took a snake, he put it in where her brains (were).’ Jacobs (1936: 2) tutúsh álta ya munk- stúx Ø álta ya míɬayt ya ya úma‘ Ø now 3  untie 3 now 3 be.located 3 feed 3 breast ‘She untied it [=him, a young baby], she sat down, she gave it [=him] the breast.’ Victoria Howard (Jacobs 1936: 11)15

In a token of variation, one example has a young baby being referred to as 3 yá.xka: (11)

álta ntsayka ískam yá.xka now 1 take 3 ‘Now we will take him.’

Jacobs (1936: 11)

This seems a rare instance of optionality (cf. section 6) between yáka and Ø, and its occurrence just at the borderline between animate and inanimate categories suggests the gradiency of the animacy scale. We have seen that the glosses provided with our data can be misleading, and we must accordingly be sensitive to the actual use of the pronominals. Concise proof that yáka is to be associated with [+animate] and Ø with [−animate] third person is in the application, under principled conditions, of both to a cat which is first living, then dead, in (12).16 This example comes from a story told both extemporaneously and in writing by an elder who grew up in a lcCW-speaking household. It shows that the event of the cat’s death is what occasions a switch from the one pronoun to the other. .

Note that (10) exemplifies Ø in both direct and indirect object positions.

. Note that in the case of ditransitives, the indirect object normally precedes the direct object in lcCW; thanks to Tom Larsen for discussion of the placement of null in such constructions. Note also that Ø in reference to a young baby is sometimes glossed as ‘him’ elsewhere in the same story (Jacobs 1936: 11) but syntactically always patterns with the clearly inanimate examples given here. .

This example incorporates (4).

Chapter 7. Lower Columbia Chinúk Wawa 

(12)

ɬiil p’ús pi lúlu yáka álta (uk úlman/ya) ískam uk now  old.man/3 take  black cat  carry 3 mának’i sayá pi másh yá khapá. more far  throw 3 there (‘Then the old man took the black cat and carried it farther away, and left it there. . .’) [Then the old man kills the cat; the narration continues:] sayá pi ya(ka) uk p’ús ya másh khapa lisák yaka lúlu Ø  cat 3 throw  bag 3 carry 3 far  3 tsə́qw Ø/uk lisák másh khapa throw  water 3/ bag (‘. . .he threw the cat into the bag. He carried it far away and threw it/the bag into the water.’) Petite (n.d.) (‘Eeh ulman ya towen. . .’) and Zenk field recordings17

This is quite distinct from the free variation in (10, 11). Differential grammatical treatment of a living vs. a newly deceased referent appears also in (13), highlighting the difference between English and CW in this regard: CW treats its grammatical animacy as changeable, whereas English sees it as a permanent quality and therefore continues to use ‘him’ after the referent’s death. (13)

<wε´l, yaga-ma·´c kaba- i´li’i, ya- mak18-i x.bu uk- i´li’i.> ya mank- íx. puy uk íli‘i wél yaka másh Ø khapa ílii  3 throw 3  ground 3  cover  ground ‘Well, he put him [mythological character’s son] in the ground, he had the ground covered over.’ Coquille Thompson (Jacobs 1936: 26)

. In this example, alternates separated by slashes represent differences between Petite’s written and (extemporized) verbal renditions of the same story. The reader might wonder whether, given the second alternate ending, Petite had in mind ‘the bag’ as the referent of null here. Given known patterns of lcCW referent tracking, however, the default interpretation points to a referent ‘the cat’ since the latter is already in focus, and by corollary there are two 3obj in this example in the first alternate version. Note that in the text following the killing, ‘the cat’ is fronted from the default postverbal object position. In CW this is a way of focusing a clausal element. I claim no more than that this is simple object-fronting, but topicalization-movement (‘as for the cat. . .’) accompanied by another Ø object (uk p’ús[,] ya másh Ø) cannot be ruled out. My claims about the principled distribution of null vs. yaka are supported by either claim – by the latter more strongly than the former. Special thanks are due to Henry B. Zenk for sharing audio field recordings of several tellings by Petite. .

Sic; probably a typographical error.

 David D. Robertson

I have shown that the relevant property of CW null ‘it’ which distinguishes it from the other third-person pronoun yáka is not that of gender as the English glosses often imply, but roughly something to do with the difference between people and things. This difference I term animacy, a reasonable label but one which we must take with a grain of salt. (Cf. Carnie 2003: 10 “it is totally mysterious how we might measure or compute, for example, the relative ‘animacy’ of a set of NPs”. For example Yamamoto’s (1999) ‘cognitive’ scale of animacy is, though more thorough, of little more predictive force than prior discussions of animacy, since local cultural valuations play a role in each language’s assignment of animacy.) My intention is mainly to point out for the first time in the CW literature that, as in innumerable other languages of the world, such a hierarchical distinction is active in this language’s differential object marking (DOM) system. Aissen’s (2003: 8) animacy markedness hierarchy (14) predicts some such distinction among syntactic objects. (14)

Object/Inanimate

}

Object/Animate

}

Object/Human

According to the hierarchy in (14), inanimate entities are definitively less-marked as objects than are humans, which will always behave as animates in any system having such a distinction (op.cit.: 19). The borderline case is animate objects (living things other than humans): some languages will include at least some animals etc. in the animate category as does CW yáka, while others will oppose these as inanimates to animate humans. “Leakage across the animate-inanimate boundary is also found” (op. cit.: 20), such that various languages will draw the line between the two with greater or lesser restrictiveness. But the fact is that the distinguishing line will separate two sets of adjacent entities within (14) – languages are not found which treat both inanimate and human objects as a group over against animate objects. The basic soundness of the hierarchical animacy distinction rests on this observation. Aissen (2003) predicts that entities more marked as objects will appear with overt morphology tagging them as such, while less-marked objects will lack such morphology. CW, being isolating in nature, lacks affixation but the differentiation in its object marking is indeed morphological, i.e., presence (yáka) versus absence of overt pronoun (Ø). The other mechanism which CW could conceivably employ for distinguishing relative animacy, word-order (op.cit.: 35–6), is actually unavailable due to its already being used in all varieties of this analytic, configurational language to distinguish participants’ syntactic functions (Section 5 below).

.

‘Core’ and ‘extended’ senses of Ø: Definiteness hierarchies

So far we have examined data on CW null object which show its use as a clearly singular pronoun, as is the case with the majority of examples in my corpus (75 of a total of 108). The remainder of the null objects I found exhibit certain interesting properties which make a more complete description of 3 possible. Shown in (15) is one of the 10 instances where Ø is glossed with English ‘them’.

Chapter 7. Lower Columbia Chinúk Wawa 

(15)

. . uk míɬayt ya x. úmum ya məkhmək Ø qháx. where  be.located 3 brain 3 eat 3 ‘Where her brains lay (in her skull), (there) he ate (them).’ Victoria Howard (Jacobs 1936: 2)

In the instance of the very few sentences like (15), the plurality of ‘brains’ is an artifact of the English translation; note that ‘brain’ and ‘it’ are equally sensible glosses. The referent contains no identifiable, thus potentially countable, constituent parts, and is therefore a mass noun. By contrast, in a slightly more numerous type exemplified by (16–17), the objects do consist of identifiable parts. (16)

“áɬqi nay t’ík- t’ik Ø {just so far}, áɬqi nayka wáx Ø.”  1 drip  3 just so far  1 spill 3 ‘I will drip them [beans] just so far, (then) I will spill (them).’ Zenk field notes 2/1/83

(17)

nay shúz n nayka {now hang} Ø nayka munk- ɬáq 1  hang 1 shoe and 1 now hang 3 khapuk {tree limb} . tree limb ‘I take off my shoes and I now hang (them) on (a) tree limb.’ Zenk field notes 2/1/8319

The constituent ‘atoms’ of the direct-object referents in (16, 17) are countable in CW, to the best of my knowledge; it is just that in these examples it is superfluous to count them. As regards taking off one’s shoes, a default reading involving the complete pair as opposed to just one shoe seems reasonable, just as one normally deals with (measured) quantities of beans rather than either individual or carefully counted units. I suggest that (16–17) instantiate the ‘3’ analysis of the CW null pronoun, in that uncounted nouns are prototypically treated as non-plural if not overtly singular, and in that none of these object nouns are animate by the standards laid out in Section 2. Now, no explicit marking of number on nouns in CW is possible, let alone obligatory, given the analytic nature of this language. Only extrinsic marking is allowed, that is, discrete number words may optionally be added to bare nouns; however, neither agreement nor classifiers exist in this language to obligatorily clarify number distinctions. This is exactly parallel to Wilhelm’s (2006: 20–2) argument that Dëne Sųɬiné (Athabaskan) nouns have a semantic count-mass distinction (which is lexically specified) yet are completely unspecified for number. Wilhelm (p. 26) consequently analyzes such ‘barenoun languages’ as representing a new ‘third type of language in the typology of bare . Items thought to represent code-mixing with English are in curly braces. In kh apuk two CW words are here conventionally phonologically fused. This example sentence shows that code-switching into English does not prevent use of the null-object construction.

 David D. Robertson

nouns and countability/number systems’ in addition to the widely recognized ‘agreement’ and ‘classifier’ types. Chinúk Wawa thus constitutes an additional exemplar in support of her typological revision. Two examples, (18–19), show a possible extension from the already observed senses of 3 to indefinite animate reference: (18)

ɬúchmәn pus méri ɬáska mákuk Ø pus ɬáska tíki  3 want woman  marry 3 buy 3 (‘If they wanted a woman to marry, they bought one/her.’) Petite (n.d.) (Ahnkudty mithlite cuba ogook. . .)

(19)

<Eeh siwash tillicum cupshwalla hooloyma siwash tillicum. Hluska mamoonk elite[e].> íxt sawásh tílixam kapshwála .xlúyma sawásh tílixam ɬáska one Indian people steal other Indian people 3 mamunk- iláythix Ø  slave 3 (‘One Indian tribe would steal other Indian people. They enslaved some/them.’) Petite (n.d.) (Ahnkudty mithlite cuba ogook. . .)

Obviously the referent in each example is animate by the standards established above, so these are cases which impinge on the territory encompassed by 3 yáka. These conceivably remain on the 3 side of the boundary: (18, 19) are excerpted from descriptions of past aboriginal customs, perhaps intended as referencing general patterns rather than entities. Moreover example (18) is an irrealis construction, which could be expected to take a relatively indefinite object. As a result Ø in both examples may reflect objects which are not individuated. But it must be observed that the partitive reading in both examples is equally or less likely than an interpretation as (anomalously) definite. Also importantly, (18, 19) were written rather than spoken, by a native-speaking schoolteacher whose CW is known to have undergone hypercorrection in various ways when she wrote. For example, in (18) spus is ‘purely a dictionary form’ according to Zenk (p.c.), sometimes employed by Petite when writing, in order to match the de facto standard of available published sources on Chinúk Wawa. In unrehearsed speech this speaker used her local dialect’s norm pus. Consequently these last two examples are at best marginal to the corpus, and possibly are flukes. The best description of CW null pronoun will recognize its primary and secondary meanings as in (20), terms I assign based on relative frequency of use.20 . Compare Dahl and Fraurud’s (1996: 62–3) ‘fuzzy dichotomy’ between ‘core or prototypical’ hierarchical rank (‘essentially human beings perceived as agents’) versus ‘extended’ rank (‘the rest of the universe’). Thus they, like I, consider ‘collective nouns’ to be among the ‘types of NPs [which] tend to show up as exceptions or dubious borderline cases’; cf. (15–17) and possibly (18, 19). Similarly they note that ‘NPs referring to animals’ are likely to straddle, as in lcCW, the animate-inanimate borderline.

Chapter 7. Lower Columbia Chinúk Wawa 

(20)

‘Primary’ and ‘secondary’ meanings of CW Ø: ‘3Obj’ → 3Uncounted → 3Mass (→ 3Non-individuated animate?)

This scale is secondarily intended as hypothesizing ‘3’ as predecessor to the various other senses of Ø, which I suggest are still best understood as unified under the umbrella of inanimacy and indefiniteness. This recalls Aissen’s (2003: 8–10) linking of definiteness with animacy as the key factors determining DOM (comparable with Yamamoto’s (1999: 2–39) framework involving subscales of animacy, person, ‘individuation’/definiteness, semantic role and ‘politeness’/empathy).21 Aissen observes that crosslinguistically the ideal least marked object is not only a third-person inanimate but one which is as indefinite as possible. Her markedness hierarchy for definiteness, reproduced in (21), shows decreasingly marked objects as one reads to the right. (21)

Pronoun > Name > Definite > Indefinite Specific > NonSpecific

Aissen’s claims appear to be borne out by the CW data. The meanings of ‘3’ that I have presented have in common that they are low on this scale (non-definite, i.e., the two rightmost positions in (21)), low in animacy as in (14), or both. Reproducing another highly useful representation from Aissen (2003: 22), I show in Figure 1 her crossing of the animacy and definiteness scales. Differential Object Marking in any given language is predicted to follow this hierarchy in the following ways: 1. 2. 3. 4.

5.

“Higher (dominating) elements are more likely to be case-marked qua objects than lower (dominated) elements” (Aissen 2003: 21–4); items depicted side-by-side here are not crucially ranked with respect to one another; subcategories of object marking are predicted to occur within Figure 1: obligatory, optional, and totally prohibited marking; each of these subcategories will be composed of contiguous elements [only the ‘obligatory’ and ‘prohibited’ groups being necessarily present, though Aissen does not explicitly say so] (loc. cit.); DOM will diachronically [have] spread “along the paths defined in” Figure 1 (op. cit.: 32–4); and

. Næss (2004) offers the similar analysis that a subject’s degree of control and an object’s degree of affectedness are the relevant parameters. Many of these schemas can be traced back to the influential suggestion by Dowty (1991: 571–5) that all proposed thematic relations can be reduced to a scale between two broad polar concepts ‘Proto-Agent’ and ‘Proto-Patient’. These would be clusters respectively of the approximate sets of entailments [volition, sentience, causation, movement, existence independent of event] and [change, incremental theme, causally affected, stationary, non-independent existence]. Thus CW yáka would represent Proto-Agent versus Proto-Patient Ø.

 David D. Robertson

6.

when “α dominates β, then a. if an object of type β may be case-marked, then all objects of type α may be case-marked. b. if an object of type β must be case-marked, then all objects of type α must be case-marked”. (op.cit.: 22)

Most marked for objects

Human Pronoun Human PN

Human Definite Human Specific

Animate Pronoun

Animate PN

Animate Definite

Inanimate Pronoun

Inanimate PN

Human Non-Specific Animate Specific Inanimate Definite Animate Non-Specific Inanimate Specific Inanimate Non-Specific

Least market for objects

Figure 1. Relative markedness on the dimensions of animacy and definiteness (Aissen 2003: 22) [PN = proper name]

What all of this implies for CW is that the dividing line between overtly marked objects (i.e., yáka) and null-marked objects (3) should lie directly between adjacent elements (there being no clear cases of optional marking in my corpus, but cf. Fn. 8) for its entire length. I note that the DOM faultline in Chinúk Wawa, lying entirely within the third-person pronominal subsystem, does not literally instantiate the relations in Figure 1. The latter are conceived in terms of pronouns versus nouns.

Most marked for objects

Human Pronoun Human PN

Human Definite Human Specific

Animate Pronoun

Animate PN

Animate Definite

Inanimate Pronoun

Inanimate PN

Human Non-Specific Animate Specific Inanimate Definite Animate Non-Specific Inanimate Specific Inanimate Non-Specific

Least marked for objects

Figure 2. Differential object marking in lcCW referents, after Aissen (2003)

Chapter 7. Lower Columbia Chinúk Wawa 

Nonetheless it seems justifiable to use Aissen’s schema given the differences in behaviour, according to the type of NP they reference, between yáka and Ø. Thus I give in Figure 2a representation of DOM in CW. Inanimate proper names (those of places, to repeat Aissen’s example) are not pronominally referenced in the corpus I have worked with, but the dominating node ‘Inanimate Pronoun’ and dominated ‘Inanimate Definite’ are already known to be referenced with Ø. It can be hypothesized from (6. a, b) that any inanimate PNs will take null as well, when not referenced by spatial deictics (‘there’, ‘yonder’, etc.).

.

Corpus and sociolinguistic limitations: Variation and inanimate count plurals

In a prima facie contradiction of what I have said so far, a number of the sentences in my corpus which have transitive verbs and lack an overt object pronoun are nonetheless translated with English animates “him”, “her”, or even a local person, “me”. Examples (22–24) are representative of these cases, for each of which the context makes clear the implied object’s animacy: (22)

yáka t´ɬáp [Ø] khapa háws ya ɬátwa nánich [Ø] 3 go see [3] 3 find [3]  home ‘. . .he went to see him. He found him at home.’ Coquille Thompson (Jacobs 1936: 26)

(23)

“máyka áɬqi lúlu [Ø]”. 2  take [3] ‘You will take (her) [old lady].’

(24)

Zenk field notes 2/1/83

tunús mán munk- k’áw náyka pi másh tsә¯qw [Ø] little man  tied 1  throw water [3] (‘Boys tie me [grasshopper] up [to their fishing lines] and throw me into the water.’) Petite (n.d.) (Swakak. . .)22

From the foregoing discussion, it is clear that such examples could entail revising the analysis of Ø. What would have to be added is some range of animate elements in the Chinúk Wawa DOM scheme which are not third persons and which are optionally marked with an overt pronoun.

. Note the apparent null preposition in másh [Ø] tsә´qw, a common construction in fluent CW directional and locative phrases; cf. final section of this paper.

 David D. Robertson

Three observations are appropriate at this juncture. First, the first-person singular náyka in (24) should be viewed as a unique variation; it is the sole non-third person pronoun to putatively alternate with null, and no additional example of ‘I’ being expressed without náyka is to be found.23We then have only an extension of Ø marking within the third-person singular to deal with, i.e., a relatively minor exception to the ‘rule’. The second observation is three-pronged: (1) The examples of null personal animates total about 13% of the corpus of sentences with Ø (on the order of 20 out of 150). (2) The corpus has 82 examples of yáka for personal animate objects, about four for every null ‘he, she’. (3) Just two speakers out of circa 20 individually identified in documented lcCW (~10%) clearly allow this ‘alternation’ with any frequency (to say nothing of the copious material from unspecified speakers, where the alternation is absent). Third, it seems plausible that the null personal animates would not been a viable alternative had the null inanimate usage not existed. The various examples in the present study show that the latter is attested for a century (beginning with Demers, Blanchet and St. Onge 1871 [1838] before the first appearance of the null animate examples (in Jacobs 1936). The original meaning is thus presumably ‘inanimate’, from which ‘he, she’ are deviations. In sum, in the absence of additional information on the distribution of null ‘he, she’ in discourse, this usage can be footnoted as an infrequent free variant.24

. One reviewer’s question about this form was whether it might not be using an inanimate object pronoun due to the grasshopper narrator’s low degree of animacy; for reasons discussed here, no answer to this question has been possible. N.b. an apparent second example of Ø ~ 1 was at first included in the corpus [note: the sentence refers to an old woman who is facing away from the speaker]: qh áta ya wáwa “cháku” pi wík ya nánich [Ø]? how 3 say come   3 see [3] ‘How can she say “come in!” (to me), yet she doesn’t see (me)?’ Zenk and Johnson (2003a: 40) Here an additional null might be posited in the first clause, but support for that view is quite weak. Verbs of speech in lcCW frequently have no overt addressee, and in any case as Zenk points out (p.c.), a viable interpretation (and English translation) is as an intransitive, ‘. . .yet she doesn’t look?’ . It seems plausible that the null personal animates would not been a viable alternative had the null inanimate usage not existed. The latter are attested for a century (beginning with Demers, Blanchet and St. Onge 1871 [1838] before the first appearance of the null animate examples (in Jacobs 1936). The original meaning is presumably ‘inanimate’.

Chapter 7. Lower Columbia Chinúk Wawa 

It should be briefly noted in relation to the broader topic of CW referent tracking that certain sociolinguistic peculiarities are not predictable from the discussion up to this point. Because most native speakers of this highly endangered language have passed on, we are forced to appeal to the knowledge of and ‘feel for’ the Chinúk Wawa spoken by those late elders that a very small number of researchers and tribal members were able to develop. Henry Zenk (p.c. May 21, 2003) has pointed out to me that native speakers of lcCW with whom he worked “resisted abstract translation tasks” such as a linguist might use in trying to elicit object forms for the present study. Acceptability judgments are difficult to obtain under these conditions, it was suggested. An especially difficult unforeseen complication noted by Zenk is that native speakers have shown a preference in elicitation contexts for paraphrases that recast an object referent in (implicit) subject position, and even for repetition of overt NPs at the expense of pronoun use. The above considerations have impinged on my attempts to learn more about how inanimate count-plural objects are pronominally tracked, as these are scant in my data. ‘Count’ referents are straightforward when singular, as various examples above show. But reacting to three elicitation prompts of my construction, Tom Larsen (p.c. May 4, 2003) expressed substantial agreement with Zenk’s opinion that my prompts for plural object ‘them’ in (26a–c) would likely receive translations either like the ones using Ø shown here in bold italics, or equivalent to the italicized ones with recasts (underlined).25 (26a)

prompt: “I have three cars; do you like them?”: nánich úkuk ɬún t’síkt’sik, mayka tíki Ø? look  three car 2 like 3 ‘Look at these three cars, do you like them?’ OR ~ “Look at these three cars of mine; what do you think, nice, eh?”

(26b)

prompt: “I forgot where those four books of mine are; have you seen them?”: qhá úk nayka lákit búk, MÁYKA nánich Ø? where  1 four book 2 see 326 ‘Where are those four books of mine, have you seen them?’ OR ~ “Where are those four books of mine? Have you seen where?”

(26c)

“Mom bought me five new shirts; what do you think about them?”: úkuk qwínәm chxí lishát nayka mamá yaka pálach náyka,  five new shirt 1 mother 3 give 1

.

The latter were provided by Zenk (p.c. August 18, 2005).

.

i.e., with extra stress and elevated pitch on 2sg for emphasis.

 David D. Robertson

qháta mayka tíki Ø? how 2 like 3 ‘These five new shirts my mom gave me, how do you like them?’ OR ~ “These five new shirts, my mom gave me; do they seem nice (all five)?”

Though my ‘consultants’ also suggested the possibility for some speakers of expressing such count-noun objects with 3 ɬáska, my corpus nonetheless contains only animate referents for that pronoun. One hundred percent of lcCW speakers use ɬáska, as in (27), only for animate count plurals. (27)

áɬqi khánawi nayka munk- mímlus ɬáska  all 1  die 3 ‘I ought to go kill them all’ John Hudson (Jacobs 1936: 17)

I take the above facts together as suggestive that lcCW speakers kept ‘3’ null both as inanimate and as nonspecific as possible, again in the spirit of Aissen (1999, 2003) – but that nonspecificity is the more important factor in its meaning. Objects which are comparatively specific are at least optionally referred to with recasts, and those which are comparatively animate are obligatorily referred to with yáka and ɬáska. (Recall (18, 19) where animate mass/partitive plurals are expressed with null: their indefiniteness is more decisive than their animacy in deciding whether phonological exponence is used.) Due to the corpus and demographic limitations noted, I emphasize that my identification of an additional pronoun in lcCW does not complete the descriptive picture. There are additional factors which learners and researchers must take into account to accurately reflect the realities of this pidgin-creole language. Good examples are the above-mentioned preferences for overt recasts or for repetition of NPs. The discussion up to this point might seem to suggest that expressions with Ø would be expected in such circumstances. Yet a sociolinguistic rule (absent from published sources to date) may exist which is specific to the speech community, exploiting the productive morphosyntactic patterns of lcCW in unexpected ways. Native speakers’ preference for avoiding Ø in these constructions in no way implies the marginality of 3 to the language, given the large number of examples using the null pronoun that were collected from the 1830s to the 1980s. Instead it seems count-plural forms may simply not be nonspecific enough to come automatically under the umbrella of ‘3’, another indication (with (10, 11)) of variability at the fringes of the semantic range of lcCW Ø. Table 1 summarizes the observed interactions of animacy, definiteness (often emergent as [±count]), and number in the lcCW third person pronouns. This table shows clearly that it is when Ø bears the value [+definite] that it may exceptionally vary with the corresponding phonologically overt pronouns.

Chapter 7. Lower Columbia Chinúk Wawa 

Table 1. Animacy, definiteness and number in lcCW third-person pronouns −pl +ani +def

−pl +ani −def

−pl −ani +def

−pl −ani −def

+pl +ani +def

+pl +ani −def

+pl −ani +def

+pl −ani −def

yáka; !Ø

?Ø

Ø; !yáka

Ø

ɬáska

?Ø

Ø

animate count singular (1), (12); !(22,23)

animate noncount singular !?(18)

inanimate count singular (2), (4), (8–10), (12, 13), (25); !(11)

inanimate noncount singular (3), ((15–17))

animate count plural example (27)

animate noncount plural !?(19)

?Ø; !? ɬáska inanimate count plural ?(26); !?(no examples)

inanimate noncount plural ((15–17))

Key: ? = uncertain; ! = exceptional; (( )) = interpretable as [±PL]

Table 2 reduces these data to those features which are distinctive of Ø, yáka, and ɬáska. Redundant features are eliminated, demonstrating that the phonologically overt members of this set differ from one another primarily in the dimension of number, and from Ø in both definiteness and animacy. Table 2. Distinctive features of lcCW third-person pronouns

+PL

−PL

+ANI

+ANI

−ANI −DEF

ɬáska

yáka

Ø

The part of the semantic range of Ø which is most distinct from those of the overt pronouns is both [−animate] and [−definite], but by extension along the dimension [±definite], this pronoun can also easily refer to definites. Likewise, ɬáska and yáka may in exceptional cases be extended along their secondary semantic dimension, to inanimates. Based on its contrasting in the dimension of animacy with the other thirdperson forms, a more appropriate label than ‘3’ for the pronoun which I am proposing to add to the described inventory for lcCW is ‘3’. Yáka and ɬáska remain well described by the traditional labels ‘3’ and ‘3’. The finding of a split along the lines of gender (in the sense of a “classification of nouns into two or more classes with different grammatical properties” (Trask 1999: 100)), and confined to the third person singular, conforms exactly to the predictions of Harley and Ritter (2002: 486) for the structure of pronominal systems. Their feature geometry (Figure 3) makes explicit the “dependency of gender on number” (p. 514) – here respectively labeled CLASS and INDIVIDUATION.

 David D. Robertson

Referring Expression (= Pronoun)

PARTICIPANT

Speaker

Addressee

INDIVIDUATION

Group

Minimal

Augmented

CLASS

Animate

Feminine

Inanimate/ Neuter Masc ...

Figure 3. Universal pronominal feature geometry (Harley and Ritter 2002: 486)

In this model a given node may be used in a pronominal system if and only if its parent node is used as well. Since Chinúk Wawa indeed exploits a singular-plural distinction among its pronouns as (28) shows, a further elaboration into an animacy split is viable: (28) 1 2 3

singular náyka máyka animate yáka inanimate Ø

plural ntsáyka msáyka ɬáska

The confinement of this animacy split to the third person is explained ergonomically, as being ‘due to the fact that [e.g.,] a gendered 1st or 2nd person pronoun involves elaboration of both major organizing nodes [PARTICIPANT (local person) and INDIVIDUATION] but a gendered 3rd person pronoun involves only elaboration of the Individuation node’ (Harley and Ritter 2002: 509).

.

Subjects and possessors vs. objects: More evidence for hierarchies

I note that the lcCW null pronoun is restricted to object position. This distribution and my focus on objects largely to the exclusion of the other syntactically-specified argument positions – subject/agent and possessor – are explained in the following brief notes. Recall that Chinúk Wawa lacks inflectional morphology and is a strongly configurational language. Therefore differentiation of the subject and object functions of a given person-number combination is as in (29a, b) achieved by contrasting order of NP constituents in a phrase, as is the possessive function, to be discussed below. Examples (29c, d) show that subjects of copular constructions may be either initial or final in the phrase. (Cf. the positional table in Jacobs 1932: 37.)

Chapter 7. Lower Columbia Chinúk Wawa 

(29a)

(29b)

(29c)

(29d)

áyaq nayka másh máyka. qickly 1 beat 2 ‘I can beat you.’

John B. Hudson (Jacobs 1936: 14)

χáwqaɬ máyka másh náyka can’t 2 beat 1 ‘You can’t beat me.’

John B. Hudson (Jacobs 1936: 14)

<wi´k- ·ɬu.´c ya´xga> . wik- ɬúsh yaxka .  good 3 ‘She (is) not well.’

Victoria Howard (Jacobs 1936: 10)

t’ɬúnas ya sík perhaps 3 ill ‘Maybe [she] is sick.’

Victoria Howard (Jacobs 1936: 7)

Referring back to the concept of hierarchies such as those in (14) and (21), the optimal and least marked candidate for subjecthood is claimed to be a local (i.e., first or second) person, cf. Aissen (1999: 7). Correspondingly a nonlocal, especially an inanimate, subject is expected to be marked, while a nonlocal, inanimate syntactic object should be unmarked (cf. the above discussion and Aissen 2003: 34–5). Indeed in lcCW the latter has already been demonstrated; 3 is represented by Ø. Expression of inanimate third-person subject is tightly constrained, as (30a–d) show. In the absence of any overt NP such as that of (30a), the only recourse is to the DEM úkuk as in (30b).27 The latter is essentially a dummy construction (úkuk literally means ‘this/that/these/those’), is the only example of its kind in the corpus, and is probably a calque on English by a late 20th-century remember of lcCW. (30a)

(30b)

chaku- ɬáx. uk ya lagóm  out  3 pitch ‘its pitch came out’

Victoria Howard (Jacobs 1936: 7)

álta úkuk chaku- khíyәp now   dark ‘Now it’s getting dark.’ Ila Dowd (from Zenk field notes [Zenk p.c. November 26, 2005])

. The focus of (30a) is on the presence of overt subject NP lagóm (instead of a subject pronoun); ya here is in possessive use, which is not in discussion at the moment. (And this use of ya is in complete agreement with the observations made below on possessives.)

 David D. Robertson

(30c)

ɬátwa ískamt sә´qw Ø [2] go get water ‘Go fetch water!’ *‘It fetches/fetched water.’

(30d)

ya ískam uk úlq’ 3 take  snake ‘He took a snake’ *‘It took a snake.’

/ / / / /

[2] come now ‘Come now!’28 *‘It comes now!’29 Victoria Howard (Jacobs 1936: 2) / (idem: 12)

Victoria Howard (Jacobs 1936: 2)

The extension of Ø, as attempted in (30c), from object position to express an inanimate pronominal agent or subject, is unattested, albeit this incurs a violation of pattern symmetry. All other pronouns may freely occur in any syntactic position in CW, but the tendency against non-local, non-animate subjects is so complete that only other expressions of 3 than Ø (effectively only nouns as in (30a)) are found in this syntactic position. This and the fact that the only Ø subject allowed is a local person, 2/ – and only in the highly limited environment of imperatives, where additional marking (albeit e.g., tone of voice) might have been available – suggests that Aissen’s claims about both object- and subject markedness apply well to the CW data. I stress that yáka, as in (30d), is for nearly all speakers not a freely available option for expressing an inanimate third person subject or object in this pidgin-creole.30 The data on lcCW third-person possessors as in (31a–c) show similar patterning.

. The readings shown as acceptable for this utterance are so only if we do not posit 3. In other words the valid reading is phonologically identical to but syntactically distinct from the null-pronoun version. . The asterisked forms in (30c, d) are obviously not acceptability judgments elicited by me from this long-deceased lcCW speaker. Instead these represent something more general: The complete absence of inanimate agents/subjects marked with Ø or yaka anywhere in the corpus. . Zenk (p.c. August 18, 2005) notes rare examples of variation, in the speech of the Cowlitz Salish man Joe Peter, recorded by Jack Marr for John Peabody Harrington in the 1940s. As transcribed by Zenk and Johnson (2003b: 325), one example is: alta ukuk kh anawi- ikta mayka wawa mayka nanich alta wik yaka now  all thing 2 say 2 see now  3 kakwa dlet like true (Marr’s English model sentence:) ‘Now all the things you said, you see are not true.’ (My emphasis and glosses.)

Chapter 7. Lower Columbia Chinúk Wawa 

(31a)

(31b)

(31c)

ya tәnás 3 child (‘her children’) ~ ?(‘its children’)

Victoria Howard (Jacobs 1936: 1)

ukuk mán  man ‘that man’ ~ *‘its man’31

John B. Hudson (Jacobs 1936: 16)

< lasu·´p> lasúp [Ø] soup ‘soup’ ~ *‘its soup’

John B. Hudson (Jacobs 1936: 15)

Possessive phrases were impossible to find in isolation in my corpus; the above are extracted from larger constituents. (These could presumably occur by themselves as terse responses in conversation, but little of that type of CW text has been available to me.) What is observable in (31a–c) is that genitive case marking must be overt, i.e., have the exponent yáka, which however has a default [+animate] value, reinforcing the claim that hierarchies are active in the CW pronominal system. Now, a far more common use of third-person possessives in Chinúk Wawa is with an overt possessor. The latter can be as in (32a) animate or as in (32b) inanimate. (32a)

úk tәnәs-ɬúchmәn yaka tílixam /  - female 3 people ‘the relatives of the young woman’ [i.e., ‘the young woman’s relatives’] tílixam *úk tәnәs-ɬúchmәn Ø  - female [3] people *‘the young woman’s relatives’32[? ‘the young womenfolk’]33 John B. Hudson (Jacobs 1936: 17)

. As with (30c, d) the asterisked forms in (31b, c) are obviously not acceptability judgments elicited by me from this long-deceased lcCW speaker. Instead these represent something more general: The complete absence of inanimate possessors marked with anything other than yaka anywhere in the corpus. . As with (30c, d) and (31b, c) the asterisked forms in (32a, b) are obviously not acceptability judgments elicited by me from these long-deceased lcCW speakers. Instead these represent something more general: The complete absence of inanimate possessors marked with anything other than yaka anywhere in the corpus, and the universal trend for adjacent nouns to constitute compounds. An example of a compound in the sources consulted is <shit lametsin> ‘aperient’ (Demers, Blanchet and St. Onge 1871 [1838]: 19). . In (32a, b) the readings ‘the young womenfolk’, ‘the big fir tree bark’ seem potentially valid only when no Ø is assumed.

 David D. Robertson

(32b)

úkuk háyash lagóm- stík yaka skín /  big pitch tree 3 skin (‘the big fir tree’s bark’) [i.e., ‘the bark of the big fir tree’] úkuk háyash lagóm- stík Ø skín  big pitch tree [3] skin *(‘the big fir tree’s bark’) [? ‘the big fir tree bark’] Louis Kenoyer (De Angulo and Freeland n.d. [ca. 1928])

These latter examples are more revealing than (31a–c): In possessives, as with subjects/ agents, phonological exponence is more important than faithfulness to the Ø form of 3. Therefore a suppletive form yáka is enlisted.34 Recall from (31a) that without an overt NP possessor, a default interpretation of yáka as 3 (i.e., animate, never inanimate) is guaranteed. Without the exponent yáka, the result guaranteed by the syntactic properties of CW is an adjectival phrase (or compound) as seen in the ungrammatical readings of both Examples (32a, b). Additionally, it must be noted that inanimate possessor NPs are in practice rare; (32b) was the only example I could easily find. These facts suggest an extension of the animacy hierarchy (beyond the opposition subject/agent versus object, to include the role of possessor), in a well-motivated application of the ideas of Aissen et al. Harley and Ritter’s feature geometry (2002: 516) proposes to explain the kinds of ‘partial syncretism’ we have examined in the lcCW possessive and subject pronouns as an instance where, as in Katu and Palau, ‘3rd person forms do have distinct geometries for animate and inanimate in the singular, but there is no pronoun in the inventory to realize the distinct 3rd person animate geometry, which is hence realized by the elsewhere form’. Assuming the possibility of such mismatches between underlyingly specified forms and the set of surface realizations, lcCW can be analyzed as using an [animacy] distinction through its entire third-person pronoun system. In summary, there is evidence for the animacy distinction in pronouns outside of object position, but Ø itself cannot appear in subject/agent position (directly due to hierarchy effects) or in possessive-marking position (because formal distinctiveness from compound NPs must be maintained).

.

Conclusions; further research suggestions

I have shown that for the analysis of the Lower Columbia Chinúk Wawa pronominal system to approach descriptive adequacy, an additional third-person singular must . An Optimality-Theory style of analysis into ranked constraints readily suggests itself as a way of modeling alternation in subject/agent and possessive forms. That is, considerations of markedness (subjects/agents should be overt, pace Aissen 2003) here outweigh those of faithfulness to the canonical ‘inanimate’ form Ø.

Chapter 7. Lower Columbia Chinúk Wawa 

be posited. This pronoun has null form, denotes inanimates, and is characterized by person- and definiteness-hierarchy effects which distinguish it from singular yáka (the only form previously described for this area of the pronoun paradigm) and plural ɬáska . These hierarchies have themselves not been described in the previous literature on lcCW, but I have shown (some of) their unmistakable effects on argument selection in the language, especially in object vs. subject positions, in ways predicted by Aissen (1999, 2003). Two contributions to Differential Marking theory emerge from my paper. One is the finding that Aissen’s (2003) proposed network of hierarchically related animacydefiniteness combinations is instantiated not only among all NPs, but also among referents within the pronominal subsystem; additional work on inanimate proper names (Section 4) is predicted to solidify this claim. The other finding is that there is evidence for Differential Marking in more syntactic roles than Aissen’s proposal of object and by corollary subject, viz. lcCW genitives. Research on the historical development of present-day lcCW syntactic patterns such as yáka versus Ø will likely lead to a better understanding of pidgin CW genesis. The lcCW third-person subsystem differs markedly from those of the Indo-European languages (English and French) known to have been present in the lower Columbia River region in the earliest contact era. The latter have indefinite third-person pronouns (e.g., one, some(thing), somebody in English; on, en in French), but the syntactic distribution of these is governed by various pragmatic considerations, and these pronouns are still rigidly bound by number distinctions, while 3 is straightforwardly employed in all environments where numerically definite yáka and ɬáska are unavailable. Comparison can be drawn with the superstrate language family in the contact situation which produced CW, Chinookan, where an indefinite pronoun exists (q- ‘transitive’, [Ø] ‘intransitive’ and ‘possessive’) in opposition to the numbered third-person pronouns (Shoalwater Lower Chinook: Boas 1971 [1911]: 581; Kathlamet Upper Chinook: Hymes 1955: 73). The parallel with the CW system is more extensive than this, because in fact Chinookan third-person singular pronominals (and not other persons or numbers) make an effectively animate-inanimate (‘neuter’) distinction (the animates in turn being subdivided by quasi-‘natural gender’: male and female) as well as a number distinction (in this case singular-dual-plural). (Loc. cit.) Sherzer (1976: 72) ascribes overt animacy distinctions only to Lower Chinook, of all Northwest Coast languages, but a second potential source may be suggested. The Salish family of languages, whose member Łәw’ál’mәš (Lower Chehalis) was an adstrate spoken in the region of early contact with newcomers, in some regions is known to have third-person pronoun subsystems which use an indefinite vs. numbered distinction. Now, documentation of the Tsamosan Salish neighbours of Chinookan, especially on Łәw’ál’mәš and nearby Quinault, is not readily accessible. The published grammatical information on neighbouring Q’way’áy’iɬq’ (Upper Chehalis, viz. Kinkade 1991: 365–78) and λ’ púlmixq (Cowlitz, viz. Kinkade 2004: 217–81) is moreover mute

 David D. Robertson

on the possibility of an indefinite third-person pronoun.35 But indefinite third person is present in descriptions of some better-known Salish languages (e.g., Thompson and Thompson 1992: 63, 65).36 Kroeber (1999: 25) reconstructs *-Ø as the Proto-Salish third-person object pronominal, as well as overt third person subjects and possessives (ibid.: 14–5), all of which lack number distinctions. These may suggest models for the lcCW form. At this moment, any ascription of lcCW 3 to Salish influence is relatively tentative, but promising: analysis has recently been undertaken of both Łәw’ál’mәš and Quinault (E. Czaykowska-Higgins, p.c.), which may help establish whether null pronouns were used in those languages. It should be pointed out here that no analysis of K’alapuya grammar has been published (viz. Sherzer’s 1976: 56 note that he had access only to phonological data for that family), but that this is another known important participant in CW genesis, and therefore a third potential source of a pronominal definiteness distinction. In summary, until more extensive data on Lower Columbia-region aboriginal languages is made available, 3 can be posited as most broadly a weak areal, and most narrowly a specifically Chinookan, feature in the genesis of CW.37 In any case additional evidence mounts for the aboriginally-modeled nature of CW structures, syntactic, phonological and others, some of which have been proposed in Thomason’s (1983) influential theoretical study. It should be noted that CW null third-person objects are not strictly confined to the Lower Columbia variety. Although rare, parallel usages are to be found throughout pidgin CW of other regions, e.g., that of a native Snoqualmie Lushootseed speaker (33) and that of the newspaper Kamloops Wawa (34). (33)

(34)

<ma´·c saya´> másh Ø sayá throw Ø far ‘throw it away!’

Jacobs (1932: 39)

38 ɬaska ískam tunús {tin} ɬaska mámuk Ø kákwa sphún 3 take little tin 3 make Ø like spoon (‘They took a bit of tin and shaped it like a spoon.’) Kamloops Wawa (1901a)

. The closest parallel to an overt distinction of definiteness in Tsamosan Salish may be these languages’ ‘visible-invisible’ demonstrative opposition (Sherzer 1976: 75). . The latter are Interior Salish languages, belonging to the Plateau cultural-linguistic region where, unlike the Coast, grammatical animacy distinctions have been claimed to be absent (Sherzer 1976: 95). . The putative areal influence would be of limited range, given that there is no sign of an indefinite third person in nearby – unrelated – Sahaptin, a known but minor substrate language of CW. Viz. Jacobs (1931: 245–57). Sherzer (1976: 56–83) makes no mention of pronominal definiteness distinctions in his reasonably thorough survey of the region. .

Example 34 is my transliteration from the ‘Chinuk pipa’ shorthand of the original.

Chapter 7. Lower Columbia Chinúk Wawa 

In no case outside of lcCW, however, does null seem to be obligatory or common. Judging by trends observed in the corpora consulted, the usual pidgin formulations corresponding to (33, 34) would be phrased using overt pronoun yaka for the object. The same speaker can be found using both overt third-person pronouns and null for identical referents, so in the pidgin I do not treat Ø as a grammatically distinct pronoun. Instead I suggest that omission of the pronoun in object position is, for these pidgin varieties, a stylistic option. In terms that recall the hierarchical analysis above, Mendoza-Denton (1997: 237) points out that “item frequency, markedness, and social evaluation are all important factors in the crystallization of styles”. It is uncontroversial, regardless of one’s theoretical orientation, that third persons are extremely frequent and unmarked as objects. Sebba (1997: 106–7) claims that a hallmark of an expanded (versus a preceding ‘stable’ or incipient) pidgin is its possession of a range of established styles, with earlier stages’ toleration of significant variability (stemming from L1 influence, age, gender etc.) reinterpreted into clear stylistic distinctions. The existence of the pidgin null object may be one sign of some stage of CW expansion. In point of fact, more Native people than newcomers have exhibited null object in their CW speech, making it also a mark of ‘Indian’ style – perhaps this is the applicable ‘social evaluation’. Certainly the languages of north-western Europe which most newcomers spoke (English, French, etc.) disallow null objects, but we can hypothesize that the Pacific Northwest source languages of CW had some structure paralleling it.39 I suggest that the twin known facts of Native people’s use of CW amongst themselves, and sense of ownership of CW (cf. Robertson 2005) support the analysis of the pidgin null object as stylistic enhancement of Native identity. Whatever its nature, the lack of phonological exponence in the syntactic object position was perhaps exactly the source from which lcCW 3 grammaticalized. I suggest that the similar omission of prepositions in locative and directional phrases as in (35a) is likewise a sort of stylistic marker for the pidgin. Interestingly, the same omission not only exists in lcCW too, as in (35b), but is more common there. (35a)

Ø sayá ílii kánawi sán [á] nayka kiláy álta nayka [sísta] míɬayt all day oh 1 weep now 1 sister be.located Ø far place ‘Every day I cry, because my sweetheart [sic] is far away.’ Tralahaet (Barbeau 1951)40

(35b)

qhánchi líli mayka kúri Ø ɬúchmәn how.much long.time 2 run Ø female

. Thomason (1983: 844–5) characterizes object-final order as the norm among these languages. Perhaps further investigation of these source languages’ morphosyntax will reveal an analog to null third person object expression in postverbal position. .

These are the lyrics of a pidgin CW song, with the orthography slightly modified by me.

 David D. Robertson

‘How long has it been since you dated a woman?’ (Lit. ~ ‘How long have you been running after women?’) Jim Attwell (Zenk, p.c. July 13, 2002)

The usual wording of these phrases would use involve an overt preposition (khapa sayá íli?i; khapa ɬúchmәn). Additional investigation of null prepositions would help to establish the extent to which these reflect a grammatical category or, again, a stylistic device. For both lcCW and other endangered languages, I repeat my suggestion (Robertson 2004) that the closer to descriptive adequacy our linguistic treatments come, the greater the chances are that transmission to future generations will succeed. In the case of lcCW, which is being actively revitalized in Native communities of the Lower Columbia region, there is clearly a productive interface between descriptive and applied linguistic research, on the one hand, and application of the findings thereof on the other.

References Aissen, J. 1999. Markedness and subject choice in Optimality Theory. Natural Language and Linguistic Theory 17.4. 673–711. Aissen, J. 2003. Differential object marking: Iconicity vs. economy. Natural Language and Linguistic Theory 21.3. 435–83. Barbeau, M. 1951. Tsimshian songs. In Smith, M.W. (ed.). The Tsimshian: Their arts and music. New York: J.J. Augustin. Song #22. Boas, F. 1971 [1911]. The Chinook Indian language. Seattle: Shorey Book Store. Carnie, A. 2003. Some remarks on markedness hierarchies: A reply to Aissen (1999 and 2003). MS, University of Arizona. Retrieved in February 1, 2005 from: (http://dingo.sbs.arizona. edu/~carnie/Pages/Papers.html#Journal) Dahl, Ö. and Fraurud, K. 1996. Animacy in grammar and discourse. In Fretheim, T. and Gundel, J.K. (eds). Reference and referent accessibility. Amsterdam: John Benjamins. 47–64. De Angulo, J. and Freeland, L.S. [n.d.] [1928]. Short grammatical analysis of Tfalati Kalapuya with appended comparison between Tfalati and Chinook Jargon. Unpublished MS, Boas Collection, American Philosophical Society, Philadelphia. Demers, M., Blanchet, F.N. and St. Onge, L.N. 1871 [1838]. J.M.J. Chinook dictionary, catechism, prayers and hymns. Montréal: [n.n.]. Dowty, D. 1991. Thematic proto-roles and argument selection. Language 67.3. 547–619. Grant, A.P. 1996. The evolution of functional categories in Grand Ronde Chinook Jargon: Ethnolinguistic and grammatical considerations. In Baker, P. and Syea, A. (eds). Changing meanings, changing functions: Papers relating to grammaticalization in creole languages [Westminster Creolistics Series 2]. London: University of Westminster. 225–42. Hale, H. 1890. An international idiom: A manual of the Oregon trade language or Chinook Jargon. London: Whittaker. Hale, H. 1968 [1846]. Ethnography and philology. In United States Exploring Expedition/during the years 1838, 1839, 1840, 1841, 1842/under the command of Charles Wilkes. U.S.N. Ridgewood, NJ: Gregg. 635–50.

Chapter 7. Lower Columbia Chinúk Wawa  Harley, H. and Ritter, E. 2002. Person and number in pronouns: A feature-geometric analysis. Language 78.3. 482–526. Hymes, D. 1955. The language of the Kathlamet Chinook. PhD thesis, Indiana University. Hymes, D. and Hymes, V. 1972. Chinook Jargon as ‘mother’s tongue’. International Journal of American Linguistics 38.3. 207. Jacobs, M. 1931. A sketch of Northern Sahaptin grammar [University of Washington Publications in Anthropology 4.2]. Seattle: University of Washington. Jacobs, M. 1932. Notes on the structure of Chinook Jargon. Language 8.1. 27–50. Jacobs, M. 1936. Texts in Chinook Jargon [University of Washington Publications in Anthropology 7.1]. Seattle: University of Washington. Kamloops Wawa. 1901a. “Besieged in Pekin by the Boxers.” June. Kamloops Wawa. 1901b. “Rev. Father L.N. St. Onge, R.I.P.” December. Kinkade, M.D. 1991. Upper Chehalis dictionary [University of Montana Occasional Papers in Linguistics 7]. Missoula, MT: University of Montana. Kinkade, M.D. 2004. Cowlitz dictionary and grammatical sketch [University of Montana Occasional Papers in Linguistics 14]. Missoula, MT: University of Montana. Kroeber, P.D. 1999. The Salish language family: Reconstructing syntax. Lincoln: University of Nebraska. Larsen, T. 2002. Blue Munk. In Conathan, L. and McFarland, T. (eds). Proceedings of the 50th anniversary conference, June 8–9, 2002, University of California, Berkeley [Survey of California and Other Indian Languages, Report 12] Berkeley: Survey of California and Other Indian Languages. 108–18. Lewis, W.S. and Murakami, N. 1990. Ranald McDonald: The narrative of his early life on the Columbia under the Hudson Bay Company’s regime, of his experiences in the Pacific whale fishery, and of his great adventure to Japan. Portland: Oregon Historical Society. Macdonald, R. 1892. Letter to Eva Emery Dye, Oct. 31. Letters to Eva Emery Dye: MSS 1089. Portland: Oregon Historical Society. Mendoza-Denton, N. 1997. Style. In Duranti, A. (ed.). Key terms in language and culture. Malden, MA: Blackwell. 235–7. Næss, Å. 2004. What markedness marks: The markedness problem with direct objects. Lingua 114. 1186–212. Palmer, J. 1906. Palmer’s journal of travels over the Rocky Mountains. In Thwaites, R.G. (ed.). Early western travels, 1748-1846: Volume 30. Cleveland: Arthur H. Clark. Petite, Eula. [n.d.] [ca. 1985]. [Teaching materials in lcCW by this native speaker.] Robertson, D. 2004. On the meaning of Thompson resonant glottalization. In Brown, J.C. and Peterson, T (eds). Papers for ICSNL XXXIX: The thirty-ninth International Conference on Salish and Neighbouring Languages [UBC Working Papers in Linguistics 14]. Vancouver: University of British Columbia. 372–84. Robertson, D. 2005. Chinook Jargon as a BC Native language. [Lecture given in the University of Victoria Dean’s Lunchtime Lecture Series.] Sebba, M. 1997. Contact languages: Pidgins and creoles. New York: Saint Martin’s Press. Sherzer, J. 1976. An areal-typological study of American Indian languages North of Mexico [NorthHolland Linguistic Series 20]. Amsterdam: North-Holland. Summer Institute of Linguistics. 2005. Glossary of linguistic terms. Retrieved March 29, 2005: http://www.sil.org/linguistics/GlossaryOfLinguisticTerms/Index.htm Swan, J.G. 1989 [1857]. The Northwest coast, or, three year’s residence in Washington Territory. Fairfield, WA: Ye Galleon.

 David D. Robertson Thompson, S.G. 1983. Chinook Jargon in areal and historical context. Language 59.4. 820–70. Thompson, L.C. and Thompson, M.T. 1992. The Thompson language [University of Montana Occasional Papers in Linguistics 8]. Missoula, MT: University of Montana. Thwaites, R.G. (ed.). 1959. Original journals of the Lewis and Clark expedition, 1804-1806. New York: Antiquarian Press. Trask, R.L. 1999. Key concepts in language and linguistics. London: Routledge. Wilhelm, A. 2006. Bare nouns and number in Dëne Sųɬiné. MS, University of Victoria. Yamamoto, M. 1999. Animacy and reference: A cognitive approach to corpus linguistics [Studies in Language Companion Series 46]. Amsterdam: John Benjamins. Zenk, H. [var.]. Field notes on Grand Ronde Chinúk Wawa. Zenk, H. 1984. Chinook Jargon and native cultural persistence in the Grand Ronde Indian community, 1856–1907: A special case of creolization. PhD thesis, University of Oregon. Zenk, H. and Johnson, T.A. 2003a. Chinúk-Wawa: Kákwa ntsayka úlman tílixam ɬaska munkkә´mtәks ntsáyka/as our elders teach us to speak it. Grand Ronde, OR: Confederated Tribes of the Community of Grand Ronde. Zenk, H. and Johnson, T.A. 2003b. Chinúk Wawa translations by Joe Peter, 1941: Glimpses of a ‘Chinuk man’ in action. In Brown, J.C. and Kalmar, M. (eds). Papers for ICSNL XXXVIII: The thirty-eighth International Conference on Salish and Neighbouring Languages [UBC Working Papers in Linguistics 11]. Vancouver: University of British Columbia. 323–30.

 

Three irregular verbs in Gullah David B. Frank SIL International

Gullah exhibits a typical creole pattern for verbs, in that the verbs usually have invariant forms, and tense, mood and aspect are marked with a set of pre-verbal particles. Three prominent verbs, however, do have variant forms. This study examines Gullah verb phrases, starting with an overview of the patterns that are typical and could be considered regular. A stative-nonstative distinction for verbs is significant to the understanding of their use in context. After establishing the regular patterning of Gullah verbs, this descriptive analysis focuses on the irregular pattern exhibited by the verbs go, be and do, which have different forms for use in different contexts. Each of these verbs happens to be related to one of the four most common tense-mood-aspect markers.

.

Introduction

Gullah, also known locally as Geechee or to linguists as Sea Island Creole, is an English creole spoken along the coastal area of South Carolina and Georgia. The classic study Africanisms in the Gullah Dialect by Lorenzo Dow Turner (1949) brought Gullah into focus as a creole language native to the mainland United States.1 The majority of the Gullah-speaking population is descended from enslaved Africans brought from West Africa to grow rice in the marshy area along the south-eastern

. My own awareness of Gullah dates back to 1972, when I was attending college in Augusta, Georgia. After nineteen years of devoting my professional attention to St. Lucian Creole, most of that time living in St. Lucia, I am now living in North Carolina and focused on Gullah. My research, under the auspices of SIL International, has involved regular visits to the lowcountry of South Carolina, and specifically St. Helena Island, spending several days each month since the beginning of 2003 working with Gullah speakers. I am particularly indebted to Ardell Greene and Vernetta Canteen of Coosawhatchee, SC, and Emory Campbell of Hilton Head Island for sharing the details and love of their language with me, and to the Penn Center on St. Helena Island for hosting my research into Gullah. My own still-developing understanding of Gullah comes primarily from my contact with these people.

 David B. Frank

coastline of the American colony. The mosquito-infested lowcountry was not very hospitable, and during the eighteenth century the proportion of Africans and their descendants was high in relation to the population of whites (Wood 1974).2 Besides rice cultivation and cuisine, other connections between the Gullah population and West Africa include the weaving of sweetgrass baskets, songs and especially folktales and vocabulary, some of which has passed from Africa through Gullah and into English.3 There is an obvious relationship of Gullah to the Krio language of Sierra Leon, to Bahamian, to Jamaican Creole, to Belize Kriol, to Afro-Seminole spoken in Texas and Oklahoma, and to African American Vernacular English, though the exact nature and direction of these relationships is a matter of debate. A theory popularized by William Stewart is that Gullah and AAVE shared a common origin – basically that AAVE is a more decreolized form of Gullah. Today it is more commonly thought that Gullah had a significant influence on AAVE, but that the relationship is not one of direct descent.4 Speakers of Sea Island Creole would not know of their language by that name. In fact, for the most part they would not know of their language as being a creole. They might admit to being a speaker of Gullah, and if pressed to identify exactly what Gullah was, they might use the term ‘dialect’, meaning to them something not as well established as a language. At one time the names Gullah and Geechee – and the associated language, history and culture – were stigmatized. In the past several decades this situation has turned around, and the name Gullah today is used with pride. A visitor to the lowcountry will encounter Gullah tours, Gullah festivals, Gullah gift shops, restaurants and bookstores, the Gullah Sentinel newspaper, etc. The term ‘heritage’ is often used interchangeably with ‘Gullah’ in the context of tours and historic sites, and ‘lowcountry’ is often synonymous with ‘Gullah’ in the contexts of geography and cuisine. For several reasons, it is hard to say how many Gullah speakers there are today. This is partly because of the lack of any kind of language survey or census that takes

. In his book Black Majority, Peter Wood (1974: 36) estimates that 1708 was the year that the black population of South Carolina began to exceed the white population in number. The proportion of blacks in relation to whites continued to increase and remain high through the eighteenth century, though after about 1740 the number of those born in Africa dropped off in proportion to those born in the New World but of African descent (Wood 1974: 189). .

Jones-Jackson (1987: 137) gives the following examples: gumbo, goober, tote, yam.

. Mille and Montgomery (2002: xxxiii) write, “Much research has focused on the relationship between the two varieties as well as on their relationships with English. Researchers in the 1960s and 1970s often took the view that AAVE was once more or less identical to Gullah. Many linguists today believe that AAVE developed separately from Gullah but in a somewhat similar fashion because they share some ingredients (such as a tendency for not marking verb tense). Regardless of their actual lineages, a close historical relation is evident not only in their shared features but also in the obvious fact that their speakers share a similar heritage and common experience of de facto segregation”.

Chapter 8. Three irregular verbs in Gullah 

Gullah use into consideration. But a more difficult problem has to do with the creole continuum and bilingualism, in identifying who speaks Gullah and who does not. The answer is often not very clear. The Ethnologue (Gordon 2005) estimates that there is a Gullah-speaking population of about 250,000, among whom about 7–10,000 are monolingual. As a creole language, Gullah is somewhat distinct from English in its structure. The distinctiveness of Gullah is most obvious in its phonology, in the pronominal system, and in the structures of the noun phrase and the verb phrase. This paper focuses on the verb phrase. The basic pattern for the Gullah verb system is presented, and comparison and contrast are made with English and with other creole languages. Then we look at three particular Gullah verbs that do not consistently follow the basic verb pattern.

.

The basic pattern for Gullah verbs

Gullah has a pattern for time reference and the verb phrase that is distinct from English but agrees with the pattern common to Atlantic English and French creoles. The basic pattern is that there is no declension of verbs, and that tense, mood and aspect are denoted by the use of a set of particles coming before the verb. There are some exceptions to that general rule, and those exceptions are the focus of this study, but first we will look at the standard paradigm before turning attention to the few exceptions. We will take hab ‘to have’ as an example of a typical Gullah verb. The form of the verb does not change its phonemic shape for different uses such as past reference or imperfective aspect. Tense, mood and aspect are signaled by the use of a set of preverbal markers. The following is the standard set of preverbal TMA markers in Gullah: (1)

been gwine da done

anterior tense prospective mood imperfective aspect completive aspect

All four of these TMA markers would rarely, if ever, co-occur in the same verb phrase, but the relative order is as follows: (2)

done + been + da + gwine + VERB

The combination of been + da is common and takes the form beena.5 Combinations of been + gwine as preverbal TMA markers are not uncommon. As for the relative

. The spelling used here is orthographic and has only a loose relationship with the phonology. In other dialect literature this same Gullah word is spelled as bina or binnuh, but such differences are only in the written representation. The principle of Gullah orthography followed here is that English spellings are used, but adjustments are made when necessary to reflect the

 David B. Frank

order of gwine and da, the order of da + gwine is rare when the two are used as preverbal markers, but the order of gwine + da does not occur at all in the database used in this study, nor is it compatible with the intuitions of the Gullah speakers that were consulted.6 Now putting these common TMA markers together with the verb hab, we see the following typical combinations and associated meanings: (3)

A hab a heapa money. E hab a heapa money. E been hab a heapa money. E beena hab a heapa money. E gwine hab a heapa money. E been gwine hab a heapa money. E done hab a heapa money. E done been hab a heapa money.

‘I have a lot of money.’ ‘S/he has a lot of money.’ ‘S/he had a lot of money.’ ‘S/he used to have a lot of money.’ ‘S/he will have a lot of money.’ ‘S/he was going to/would have a lot of money.’ ‘S/he already has a lot of money.’ ‘S/he already had a lot of money.’

Note that hab is a stative verb in Gullah, so while beena can precede the verb, the combination of just da plus stative verb would be unusual. Other stative verbs include wahn ‘to want’, lob ‘to love’, tink ‘to think’, and know ‘to know’. With a nonstative verb such as gii ‘to give’, the combinations of preverbal markers and their meanings (as seen from an English perspective) are slightly different, as follows: (4)

A gii ya de money. A da gii ya de money. A been gii ya de money. A beena gii ya de money. A gwine gii ya de money. A been gwine gii ya de money. A done gii ya de money. A done been gii ya de money.

‘I gave you the money.’ ‘I give/am giving you the money.’ ‘I had given you the money.’ ‘I was giving/used to give you the money.’ ‘I will give you the money.’ ‘I was going to/would give you the money.’ ‘I already gave/have already given you the money.’ ‘I had already given you the money.’

distinctive Gullah pronunciation. Thus been is spelled in Gullah the same way as the English word been, because the pronunciation is more or less the same (at least in South-eastern U.S. regional English): [bn]. Gullah beena is pronounced as [bn]. The Gullah word meaning ‘to have’ is spelled as hab to indicate that a voiced bilabial stop is used in Gullah where there is a voiced bilabial fricative in the corresponding English word. This orthography for Gullah is not the most efficient from a technical point of view, leaves some room for guess-work, and might give the wrong impression that Gullah is a dialect of English, but is preferred by Gullah speakers who are literate in English and would like to extend their literacy skills to Gullah. . One possible source of confusion and ambiguity is that gwine can be used as a main verb and not just as a preverbal TMA marker. The combination of da + gwine is common when gwine is used as a main verb meaning ‘to go’. The analysis of relative order presented here is based only on the uses of been, gwine, da, and done as preverbal TMA markers.

Chapter 8. Three irregular verbs in Gullah 

Most Gullah verbs follow the nonstative pattern, whereby the unmarked form of the verb – i.e., the bare verb without the anterior tense marker preceding it – has past reference. Present reference is expressed by the use of the imperfective particle da preceding the verb.7 The following are other examples of nonstative verbs in Gullah: (5)

aks bring ceebe come cyaa git hep laan leh lib mek nyam pick

‘to ask’ ‘to bring’ ‘to deceive’ ‘to come’ ‘to carry’ ‘to get’ ‘to help’ ‘to learn/teach’ ‘to allow’ ‘to live’ ‘to make’ ‘to eat’ ‘to pick’

pit ron sabe say see sleep speck staat tek tote waak wok yeh

‘to put’ ‘to run’ ‘to save’ ‘to say’ ‘to see’ ‘to sleep’ ‘to expect’ ‘to start’ ‘to take’ ‘to carry’ ‘to walk’ ‘to work’ ‘to hear’

Tense-mood-aspect patterns have been a topic of considerable discussion and debate among creolists. While different observations and generalizations have been made, it has been noted that creole languages tend to share a common pattern for TMA and time reference in the structure of the verb phrase, even when there is no overlap in the specific forms of the words due to the different creoles coming from different lexifiers.8 For example, there is very little lexical overlap between Gullah and St. Lucian Kwéyòl, because the former is an English creole and the latter a French creole.9 Nevertheless, the two languages have a strikingly similar pattern for the verb phrase and TMA reference. Kwéyòl té, kay and ka correspond to Gullah been, gwine and da. The following sets of data from the two languages help demonstrate the common creole pattern:10

. Turner (1949) considered da + verb to be indeterminate as to time reference, with the time reference supplied by the context, but the present analysis agrees with Cunningham (1992) that da + verb is the normal expression of present (or habitual) time reference. See also the discussion in Hopkins 1994: 67. .

See especially Bickerton 1981, and a good summary twenty years later by Winford (2001).

. Three exceptions I have been able to find are Gullah gumbo, yam and jook, corresponding to Kwéyòl gonbo, yanm and djouké. (The phonological forms of these pairs of words are not as different as the orthographic representation would seem to suggest.) . For an analysis of the St. Lucian Kwéyòl verb phrase that goes into more detail than this analysis of Gullah, see Frank 2004.

 David B. Frank

Gullah

Kwéyòl

English

We come. We da come. We been come. We beena come. We gwine come. We been gwine come.

Nou vini. Nou ka vini. Nou té vini. Nou té ka vini. Nou kay vini. Nou té kay vini.

‘We came.’ ‘We are coming.’ ‘We had come.’ ‘We were coming/used to come.’ ‘We will come.’ ‘We were going to/would come.’

Though there is some difference between stative and nonstative verbs in terms of cooccurrence with the TMA particles and the resultant meanings, the point is that for the vast majority of Gullah verbs, the verb is not declined for person, number, or time reference.11

.

The copula

Among the languages of the world, copular verbs – if any verb at all is used in formulas of equation and attribution – tend to be very irregular (Payne 1997: 117). Gullah is no exception in this respect, and while the Gullah copula patterns are different from English, there is nothing about the Gullah copula that is unique among languages in general (cf. Payne 1997: 114–23). The copula is an example of a Gullah verb that diverges from the basic pattern and has a variety of forms. For one thing, there is the copula da and the locative copula dey. It is debatable whether or not these two forms are variations of the same word. But there is also a null copula in some contexts. Again, it could be debated whether zero morphemes have a place in linguistic description. And finally, there is an infinitive form be in some contexts. Adding to the complexity of the conditioned variation of forms of the copula in Gullah is the confusion that arises from apparently free variation of forms as well. For example, one Gullah speaker12 gave the following three variants in answer to the question of how the sentence “How old is she?” would be expressed in Gullah: (6)

How ole e is? How ole she be? How ole e da?

. There are a few cases where the verb takes on different phonological shape according to the following context, e.g., shim ‘to see him/her/it’ (< see + um). .

Leo Gaston, age 59, from Coosawhatchee, SC.

Chapter 8. Three irregular verbs in Gullah 

This kind of variation, which has been noted by others, would seem to be the result of exposure to English, decreolization and the creole continuum. However, Mille’s study of Gullah over time shows the same variation a century ago as in more recent times, and she concludes that “there is no movement of the TMA system toward the acrolect as predicted by the decreolization hypothesis” (1990: 217). She adds, “Important among the findings of this study is the fact that this variation does not establish a direction – its patterning is very similar in both the historical and modern data. It is, in fact, stable variation” (ibid.). A further complication is that da is commonly used as a TMA particle, and it is somewhat unclear in certain contexts which way it is being used. Despite all the uncertainty, an analysis is presented below that attempts to sort out the different forms and contexts of the Gullah copula, based primarily on contemporary, as opposed to historical, sources of data.

.

da

The Gullah copula da is used as the verb in an equative clause, involving two noun phrases joined by the copula, and the second one is equated with the first. Hopkins (1994: 67) has found the same pattern: “as a copula da takes only noun complements.” The following is a formula for the structure of an equative clause, followed by examples: (7)

Eq. Cl. = NP + Copula + NP Ya da David. ‘You are David.’ E da me fada. (or: E me fada.) ‘He is my father.’ Dis man da de leada. (or: Dis man, e de leada.) ‘This man is the leader.’ Dis time yah da a ebil time. ‘This time here is an evil time.’ Da God wok. ‘[It] is God’s work.’

The last example above is from Turner (1949: 213), who called da (which he wrote phonetically as d) “a verb of incomplete predication” (ibid.), grouping together its use as an aspect marker and as a copula and equating it with English ‘to be’. What is called here an equative clause can be distinguished from a descriptive clause, in that the former has a noun phrase as its complement while the latter has a modifier as its complement. In Gullah, this makes a difference, as the equative clause has da as the copula but a descriptive clause (to be described below) has no explicit copula. However, there can be transformations of both of these related types of clauses

 David B. Frank

such that the copula comes at the end of the sentence, and in this case, the copula da does appear, as in the examples below: (8)

Be who ya da. ‘Be who you are.’ Tell we who ya da. ‘Tell us who you are.’ Man, who ya tink ya da? ‘Man, who do you think you are?’ Dohn mek no neba mind who ya da. ‘It doesn’t matter who you are.’ A ain know wa dat da. ‘I don’t know what that is.’ Wa good e da ‘What good is it?’

So the Gullah word da has two distinguishable uses – as a verbal auxiliary denoting imperfective aspect for the main verb, and as a copula.13 This fact might lead some (such as Turner, as noted above) to equate da with English ‘to be’. English uses a form of ‘to be’ as a copula, as in the sentence ‘This is good/mine/a funny story/in poor taste’. And English uses a form of ‘to be’ in its present progressive form, used for present reference, as in ‘The water is running down the stairs.’ It is tempting for someone going back and forth between Gullah and English to equate da with is, but the patterns for Gullah need to be appreciated independently of the patterns of English.

.

dey

There is a locative copula dey ([de]) in Gullah as well, which is close to but not identical in form with the copula da ([d]). The meaning of the locative copula is something like ‘is (located)’ and what comes after dey is an adverbial that describes a location, as in the following examples: (9)

E dey deh. ‘He/she/it is there.’ We sista dem dey yah wid we. ‘Our sisters are here with us.’ De cat dey deh een de house. ‘The cat is there in the house.’ Dem dey right out deh front ob de do. ‘They are right out there in front of the door.’

. Payne notes (1997: 117) that “Copular verbs tend to function as auxiliaries in other constructions. . . . In fact, when a language develops auxiliaries, the first verbs to become auxiliaries are the copular verbs. Second are the verbs of motion”.

Chapter 8. Three irregular verbs in Gullah 

This last example is adapted from Cunningham (1992: 80), who calls dey14 a ‘locating verb’. Note that dey is only used in present reference. Future and past reference will be explained below.

.

Ø

In some contexts, Gullah has a null copula. Whereas an equative clause involves the copula da followed by a noun phrase, a descriptive clause simply has the subject noun phrase followed by an adjective or a prepositional phrase, with no explicit copula verb. (10)

Descr. Cl. = NP + Adj/PP E nice. ‘He/she/it is nice.’ Dem skeeta bad. ‘Those mosquitos are bad.’ E mo olda den me. ‘S/he is older than I.’ Dat gyal purty down. ‘That girl is quite attractive.’ De house ron down. ‘The house is run down.’ Ya ain good fa nottin. ‘You are not good for anything.’15 E in a fambly way. ‘She is in a family way (i.e., pregnant).’

Note that when a descriptive clause is transformed such that the verb would go at the end of the sentence, as in a question, the copula does show up as da. Thus the declarative sentence E ole ‘He/she/it is old’ is transformed into the question How ole e da? ‘How old is he/she/it?’ Both da and dey have a null variant following the anterior marker been. That is, except where da is being used as a preverbal imperfective aspect marker, neither *been da nor *been dey is a combination that occurs in Gullah, but rather instead simply been. The following example sentences are analyzed as instances of the Gullah anterior

. Only the written representation is changed from what Cunningham had. She spelled the ‘locating verb’ as deh. Both her spellings and mine are orthographic rather than phonemic, and are somewhat arbitrary. I changed the spelling just to regularize it with the spelling used in the rest of this paper. Her original example sentence is Them deh right out there front the door. The gloss provided above is hers. . The negator ain can apply to any verb, except that in a descriptive clause it goes before the “null” copula. The sentence E ain tell nobody ‘S/he didn’t tell anybody’ is an example of how ain is used before an explicit verb.

 David B. Frank

marker been (not to be confused with the English ‘been’, which is a form of the verb ‘to be’) plus a null copula: (11)

E been too opsot. ‘S/he was quite agitated.’ We ron ta de place weh Joe been. ‘We ran to the place where Joe was.’ Dey ain recognize who e been. ‘They did not recognize who he was.’ Dat man been a faama fo a been. ‘That man was a farmer before I was.’ A school been yah. ‘A school was here.’ She been a good ooman. ‘She was a good woman.’

The last two sentences above are adapted from Cunningham (1992: 89–90), who posits a transformational rule in Gullah whereby the locating verb (dey) and the equating verb (da) are obligatorily deleted following Aux (been).

.

be

All three forms of the copula described thus far share the conditioned variant form be following gwine, as the following set of examples illustrates: (12)

Da a haad time. ‘[It] is a hard time.’ E gwine be a haad time. ‘It will be a hard time.’ E real purty. ‘It is real pretty.’ E gwine be real purty. ‘It will be real pretty.’ E dey deh. ‘He/she/it is there.’ E gwine be deh. ‘He/she/it will be there.’

Another use of the be form of the copula is following a modal, as in: (13)

Oona mus be too happy. ‘You. must be very happy,’ Mus dohn be scaid. ‘Don’t be scared.’ Dat cyahn be good. ‘That can’t be good.’

Chapter 8. Three irregular verbs in Gullah 

Chullun kin be heapa trouble. ‘Children can be a lot of trouble.’

Also, the be form is used after fa to form a sort of Gullah infinitive: (14)

Dat boy ain fit fa be ya fren. ‘That boy is not fit to be your friend.’ Dey gwine drag ya fa be jedge een de coat. ‘They will drag you to be judged in the court.’ Dey ain deh weh dey spose fa be. ‘They are not where they are supposed to be.’ Me fren come fa be jes like a broda ta me. ‘My friend came to be just like a brother to me.’

Other Gullah verbs, apart from the ones described in this paper, do not have a conditioned variant form to be used in an infinitive construction.

.

To go

The Gullah verb meaning ‘to go’ has three main forms: go, gwine and gone. The distribution of each of these forms will be described below.

.

gwine

Besides serving as a prospective mood marker, gwine has a distinct function as a form of the Gullah verb meaning ‘to go’. It is used after da and beena, as in example (15): (15)

We da gwine to Beaufort. ‘We are going to Beaufort.’ All de people da gwine hine um. ‘All the people are going after him/her.’ We ain know weh e da gwine. ‘We don’t know where he is going.’ Wiles dey beena gwine, dey see dey sista. ‘While they were going, they saw their sister.’ E beena gwine dey fa see dem. ‘S/he was going there to see them.’ Peter beena gwine all oba de countryside. ‘Peter was going all over the countryside.’

However, note that sometimes go is used in this environment, as in the following: (16)

E gwine pass oba wiles e da go bout e bidness. ‘He will die while he is going about his business.’ De boat reach lan weh dey beena go. ‘The boat reached land where they were going.’

 David B. Frank

Note also that the combination been gwine is not the Gullah verb for ‘to go’ preceded by the anterior marker, but is rather a combination of the anterior tense marker plus the prospective mood marker, as in the following examples: (17)

.

Dey ain know wa been gwine happen. ‘They did not know what would happen.’ De boat been gwine sail fa Daufuskie. ‘The boat was going to/would sail for Daufuskie.’

gone

The gone variant form of the Gullah verb meaning ‘to go’ is used for past reference, as a past tense form of the verb. (18)

E gone fa wok fa she. ‘S/he went to work for her.’ De man gone an marry de gyal. ‘The man went and married the girl.’ Dey gone on weh dey beena gwine. ‘They went on where they were going.’ De chullun done gone eenta de house. ‘The children already went into the house.’ Wa oona gone fa see? ‘What did you go to see?’

Cunningham (1992: 49) says that there is an obligatory transformation in Gullah whereby in the underlying been + ‘to go’ combination, the main verb is deleted, as in Example (19): (19)

A been ta Charleston Tuesday. ‘I went to Charleston Tuesday.’

However, since both Cunningham’s data and my own show gone as a common form of ‘to go’ in Gullah, a simpler analysis might be to consider the underlying verb in sentences like this to be the locative copula dey rather than a form of ‘to go’. In that case the deletion of the surface structure verb has already been explained.16

.

go

Like be, go is a form of a Gullah verb that is used in infinitive constructions (after fa), in commands, following gwine and following modals, as in these examples: (20)

Oona mus go ta school. ‘You. must go to school.’

. In this case, been is the anterior tense marker rather than a form of the Gullah copula, and there is no explicit verb.

Chapter 8. Three irregular verbs in Gullah 

E been scaid fa go deh fa lib. ‘S/he was afraid to go there to live.’ Ya cyahn go noweh. ‘You can’t go anywhere.’ Leh we go an bury we fambly. ‘Let us go and bury our relative.’ Dey sen we fa go ta de sto. ‘They sent us to go to the store.’ Take ya money an go home! ‘Take your money and go home!’ Dey ain haffa go way fom yah. ‘They don’t have to go away from here.’ Go ta de riba an ketch we sompn fa nyam. ‘Go to the river and catch us something to eat.’ A gwine go do um. ‘I will go do it.’ Ya kin go now. ‘You can go now.’

.

To do

The Gullah verb meaning ‘to do’ has two forms: do and done. The former has the widest distribution, occuring in a variety of contexts. The latter is a past tense form.

.

do

The Gullah verb do can be preceded by the tense-mood-aspect markers been, gwine, da and beena, or in an infinitive construction following fa, or following a verb such as leh or mek. It can also stand alone in certain dependent clauses to have an indefinite time reference. The combination da do would mean ‘habitually do’ or ‘to be doing’. (21)

Dey da do de same ting. ‘They do/are doing the same thing.’ Ef ya do dat, ya gwine be sorry. ‘If you do that, you will be sorry.’ E study haad bout wa e gwine do. ‘S/he studied hard about what he/she will do.’ Dey ain know wa dey spose fa do. ‘They don’t know what they are supposed to do.’ Oona mus do wa A tell oona fa do. ‘You. must do what I tell you to do.’ Leh we do um de way we wahn. ‘Let us do it the way we want.’

 David B. Frank

De Debil da shry fa mek we do ebil ting. ‘The Devil tries to make us do evil things.’ We gwine come atta we done do we wok. ‘We will come after we have (already) done our work. A done yeh bout ebryting ya beena do. ‘I have heard about everything you have been doing.’

.

done

Like the Gullah verb for ‘to go’, the ‘to do’ verb has a specific past tense form. These two verbs are unique in Gullah in that regard. (22)

E done um fa oona. ‘S/he did it for you..’ We sorry fa wa we done. ‘We are sorry for what we did.’ De ooman done all e been able fa do. ‘The woman did all she was able to do. A done done the wok ya gii me fa do. ‘I already did the work you gave me to do.’

The verb form done only occurs in past time reference and is not preceded by the anterior tense marker been, the prospective mood marker gwine, or the imperfective aspect marker da. There are many examples of done done in my data, and it seems to have some free variation with done do to express a fairly broad range of meaning, including ‘already did,’ ‘had (already) done,’ and ‘has/have (already) done.’ The reader should keep in mind that not every instance of done in Gullah is the past tense form of do. The word done also has a distinct function as a completive aspect marker, as in the last example above.

.

Conclusion

The verb system in Gullah is for the most part quite regular, but three particular, common verbs have an irregular pattern. The Gullah verbs meaning ‘to be’, ‘to go’ and ‘to do’ each have a set of variant forms for different contexts. Each of these common but irregular verbs also happens to have a form that doubles in function as a tense-moodaspect marker. Da is a form of the copula, alongside dey and be, but also an imperfective aspect marker. Gwine, go and gone are variant forms of ‘to go’ in Gullah, but the former doubles as the prospective mood marker. Done is the past tense form of do but also the completive aspect marker. The use of these basic verbs also as TMA markers is not unusual among the languages of the world. This descriptive analysis of Gullah verbs and verb phrases serves as an attempt to document a creole language that has been elusive because it has been unwritten,

Chapter 8. Three irregular verbs in Gullah 

because it has been stigmatized, because the language has undoubtedly changed over time, and because its use is spread out over a wide area and there are thus natural variations. No claim is made here that this snapshot captures the essence of Gullah for all places and all time. As this analysis demonstrates, Gullah has forms and patterns that are distinct from English and at the same time quite similar to other creole languages, but also similar to other non-creole languages as well. As we work toward a more accurate understanding of Gullah as a creole language, for a more complete picture to emerge the present study needs to be compared with historical language data, data from other venues in the wider Gullah area, and with other creole varieties.

References Bickerton, D. 1981. Roots of language. Ann Arbor: Karoma. Cunningham, I.A.E. 1992. A syntactic analysis of Sea Island Creole [Publication of the American Dialect Society 75]. Tuscaloosa: University of Alabama Press. Frank, D.B. 2004. TMA and the St. Lucian Creole verb phrase. In Escure, G. and Schwegler, A. (eds). Creoles, contact and language change: Linguistics and social implications. Amsterdam: John Benjamins. 237–57. Gordon, R.G. Jr., (ed.). 2005. Ethnologue: Languages of the world, 15th edition. Dallas: SIL. Hopkins, T. 1994. Variation in the use of the auxiliary verb da in contemporary Gullah. In Montgomery, M. (ed.). The crucible of Carolina: Essays in the development of Gullah language and culture. Athens: University of Georgia Press. 60–86. Jones-Jackson, P. 1987. When roots die: Endangered traditions on the Sea Islands. Athens: University of Georgia Press. Mille, K.W. 1990. A historical analysis of tense-mood-aspect in Gullah Creole: A case of stable variation. PhD thesis, University of South Carolina. Mille, K.W., and Montgomery, M.B. 2002. Introduction. In Turner, L.D. Africanisms in the Gullah dialect. Columbia: University of South Carolina Press. xi–lvii. Payne, T.E. 1997. Describing morphosyntax: A guide for field linguists. Cambridge: Cambridge University Press. Turner, L.D. 1949. Africanisms in the Gullah Dialect. Chicago: University of Chicago Press. Wood, P.H. 1974. Black majority: Negroes in colonial South Carolina from 1670 through the Stono Rebellion. New York: Norton. Winford, D. 2001. On the typology of creole TMA systems [Society for Caribbean Linguistics Occasional Paper 29]. St. Augustine: SCL.

c 

Afro-Bolivian Spanish The survival of a true creole prototype John M. Lipski The Pennsylvania State University

This study analyzes data on a unique Afro-Hispanic dialect, spoken in remote areas of Bolivia by descendents of Africans who arrived in the 16th century. AfroBolivian speech, now confined to the oldest members of a few isolated communities, has preserved configurations that demonstrate the feasibility of early creolization of Spanish in independent locations. The present investigation concentrates on the Afro-Bolivian VP and DP: quasi-invariant verbs; incipient particle-based verbal system; null definite articles; lack of gender concord; stripped and invariant plurals. There exists a series of unidirectional implicational relationships that span the range from the basilect to the acrolect, and which provide a template for gradual decreolization based on relative markedness and ease of language processing.

.

Introduction

In creole studies the notion of “simplification”, while all-pervasive, rarely acquires an empirically grounded definition. In particular the manner in which Ibero-Romance verbal, nominal, and adjectival morphosyntax becomes transformed into creole paradigms such as those found in Palenquero, Papiamentu, Cape Verdean and São Tomé creole Portuguese has not been satisfactorily explained. Although all of these creole languages have eliminated Spanish and Portuguese morphological agreement in favor of invariant forms, occasional fossil remains of fully agreeing combinations suggest a gradual step-wise restructuring. One of the most significant obstacles in reconstructing the formation of Romance-derived creole languages and their implications for general theories of creolization is time depth: there is little reliable information on the earliest stages of creolization and restructuring, and there are no contemporary configurations containing enough remnants of the first stages of the “big bang” of creole formation to provide corroborative data. This is nowhere more evident than in the case of creoles arising from the contact between Spanish and a variety of African languages, of which the remaining specimens are few and of debatable origins.

 John M. Lipski

There is considerable historical, literary, and folkloric evidence that the speech of Africans who acquired Spanish as a second language (known as bozales) existed over a large enough territory and in some regions for sufficient time as to have produced at least significantly restructured varieties of “Afro-Spanish”, if not fully developed creole languages (Lipski 2005). Many of the linguistic features of these imitations are typical of all learners of Spanish: unstable subject-verb and noun-adjective agreement, use of disjunctive object pronouns instead of clitics, confusion of the copulas ser and estar, misuse of common prepositions, and avoidance of grammatically complex sentences containing subordinate clauses. Other traits are not common in learners’ Spanish but are found in Afro-Iberian creoles and probably represent the influence of African areal features: prenasalized consonants, paragogic vowels used to produce open CV syllables, in situ questions, double negation, and use of adverbial particles instead of verbal inflection for tense, mood and aspect. Finally, many of the literary imitations are simply grotesque racist parodies, devoid of any resemblance to the true results of Afro-Hispanic language contacts. Central to the debate over the reconstruction of bozal language is the extent to which this pidginized Spanish speech exhibited consistent traits across time and space, and the possibility that Afro-Hispanic pidgins may have creolized across large areas of Spanish America. The abundant bibliography of studies based on corpora of literary, musical, and folkloric texts has broadened the discussion to include a wide range of hypotheses and scenarios, but ultimately the texts in question are imitations or recollections produced by non-bozal authors, and therefore of debatable validity. Little contemporary evidence has survived the transition from bozal language to ethnically unmarked regional varieties of Spanish. There exist several isolated Afro-Hispanic speech communities throughout Latin America where traces of apparently post-bozal Spanish coexist with regional vernacular varieties. In most cases deviations from standard Spanish are limited to occasional lapses of agreement not found among monolingual Spanish speakers lacking the former bozal connection. In a few cases words or grammatical elements once found in bozal speech have survived, but more robust evidence is still needed in order expand the empirical bases for discussion beyond literary imitations.

.

The traditional Afro-Bolivian dialect of the Yungas

Previously unknown to linguists and even to most neighboring citizens, a unique Afro-Hispanic speech community survives in Bolivia’s most remote valleys. To a greater extent than any Afro-Hispanic variety studied to date, this community exhibits enough apparent continuations of early colonial bozal Spanish as to attempt a reasonable reconstruction of this set of contact vernaculars. In its most basilectal form – now confined to a tiny group of the oldest residents – this Afro-Bolivian dialect offers a blueprint for the formation of Afro-Hispanic contact language throughout the Americas. Highland Bolivia, known in colonial times as Alto Perú [upper Peru], then

Chapter 9. Afro-Bolivian Spanish 

the Audiencia de Charcas, was the site of the earliest massive importation of African slaves in Spanish America.1 In Potosí, Bolivia, at the time the world’s richest silver mine, African slaves primarily worked in the royal mint (casa de la moneda) and as domestic servants; the actual mining was done by indigenous workers. A small collection of songs and indirect descriptions of Africans’ dances and language from the 17th century survives as testimony of a much larger cultural patrimony (Lipski 1994, 1995a). Ultimately, the population of African descent blended into the overwhelmingly mestizo (mixture of European and Amerindian) nation, and today only a tiny fraction of the national population is obviously of African origin. Most contemporary Afro-Bolivians live in scattered communities in the provinces of Nor Yungas and Sud Yungas, in the department of La Paz, surrounding the capital city. The Yungas are tropical valleys no more than 3000 feet above sea level, surrounded by some of the most forbidding mountain terrain in all of South America, with peaks reaching more than 15,000 feet. The torturous terrain, nearly vertical geography, lack of adequate roads and other infrastructure, and frequent mud and rock slides, has cut off the Yungas communities from the rest of Bolivian society. Most communities are less than 150 miles from La Paz, but to reach even the closest settlements one must travel upwards of six hours in crowded and decrepit vehicles along a one-lane muddy mountain road with steep drop-offs and no guard rails. The region is principally inhabited by an Aymara-speaking indigenous population, while the Afro-Bolivians live in scattered houses on the mountainsides, in the most precarious conditions and without electricity, running water or sanitary installations. As a result of the social and geographic isolation, residents of the Yungas communities have retained cultural and linguistic traits that have faded from more populated urban areas. Black Yungueños in Sud Yungas have frequently intermarried with Aymaras, speak Aymara and wear traditional Aymara clothing. In the Nor Yungas communities, where Afro-Bolivian speech still survives and where the present research was conducted, black Bolivians remain linguistically and culturally separate from Aymaras; they learn enough Aymara to function efficiently in the Aymara-dominant local markets, but maintain a completely separate life style through networks of extended families. In the remainder of this study the term Afro-Yungueño will be used; Afro-Bolivians in general have no word for themselves or their unique dialect, other than the ubiquitous and racist negro and moreno. The origin of the black population in the Bolivian Yungas is not known with certainty. They almost certainly descend from the slave population originally working in the highland mining towns. The first documentation of black slaves in the Yungas comes at the end of the 18th century and by the middle of the 19th century the use of black slave labor on the plantations or haciendas of the Yungas was well established. Nowadays, Afro-Bolivians in the Yungas occupy (usually as squatters) tiny parcels of land that once belonged to large haciendas. Prior to the land reform that came with the Bolivian revolution of 1952, they worked as peons on these same haciendas. The regional . Bowser (1974); Crespo (1977); Cuche (1981); Harth-Terré (1971, 1973); Millones Santagadea (1973); Pizarroso Cuenca (1977); Portugal Ortiz (1977).

 John M. Lipski

African origins of modern Afro-Bolivians are uncertain, since by the turn of the 17th century, when the importation of African slaves to the mining regions reached its peak, Portuguese slave dealers that supplied Spanish colonies were drawing their slaves from across West Africa, from Senegambia down to the Congo/Angola region. Only the African surnames Angola and Maconde (apparently of Kongo origin) persist in the Yungas. These surnames were given by ship captains and slave dealers, but typically approximate the coastal African areas from which the respective slaves were drawn. There are no religious or cultural practices that allow for a more precise reconstruction of the various regional African influences that permeated the colonial slave trade. The most important Nor Yungas communities containing high concentrations of Afro-Bolivians, and where the traditional dialect still can be heard, are Tocaña, Mururata, Chijchipa, Coscoma, Dorado Chico, and Khala Khala. In Sud Yungas the principal black community is Chicaloma, with black Bolivians scattered in many neighboring settlements. Only a few of the traditional dialect traits are heard in Sud Yungas and in the remaining Afro-Bolivian settlements. All contemporary Afro-Yungueños speak rustic highland Bolivian Spanish (known locally as castellano), whose sociolinguistic features vary according to level of education and contact with other areas of Bolivia. Most older Afro-Yungueños in the Nor Yungas settlements mentioned above have at least passive competence in the Afro-Bolivian dialect; an undetermined but evidently shrinking number have total active fluency. Most younger community residents disavow any knowledge of this dialect, evidently associating the traditional speech with the illiteracy and degradation of the pre-1952 plantation environment. However, observation of these closely-knit extended families suffices to demonstrate the considerable passive competence in the traditional dialect possessed by all community residents. It is impossible to estimate the number of fluent speakers of the Afro-Yungueño dialect, but full active competence is probably limited to at most a few hundred individuals, possible even fewer. The maps in Figure 1 show the relative location of the Afro-Yungueño communities. Afro-Bolivian communities Traditional area Immigration area

Figure 1. Map of principal Afro-Bolivian communities

Chapter 9. Afro-Bolivian Spanish 

In order to obtain extensive samples of Afro-Bolivian Spanish, field work was conducted in June, 2004, August, 2005, October 2005, and August, 2006.2 Accompanied by Lic. Juan Angola Maconde, a native of Dorado Chico and the author of the only major study on contemporary Afro-Bolivian culture (Angola Maconde 2000), interviews and recordings were made in Coroico, Arapata, Coripata, Coscoma, Khala Khala, Dorado Chico, Dorado Grande, Tocaña, Mururata, and Chijchipa in the province of Nor Yungas, as well as Chicaloma in the province of Sud Yungas. I would like to express my profound gratitude to this remarkable intellectual, writer, activist, and friend. Sincere thanks are also due to Antonia Pinedo Zalles of Mururata, a remarkable activist and voluntary literacy teacher in that community. During these initial field studies, a total of fifty five Afro-Yungueños were interviewed, men and women, with ages ranging from 35 to 92.

.

Overview of traditional Afro-Yungueño speech

The Afro-Bolivian dialect differs from other highland Bolivian dialects in terms of segmental and suprasegmental phonology. It also differs from all other monolingual varieties of Spanish worldwide in the structure of DPs and VPs, both of which exhibit morphosyntactic reduction suggestive of Afro-Hispanic restructured language. The phonetic differences are immediately apparent as Afro-Bolivians shift seamlessly between colloquial highland Bolivian Spanish and their own unique dialect. Highland Bolivian Spanish belongs to the Andean dialect cluster, and is characterized by strong sibilant pronunciation of syllable- and word-final /s/ (e.g., hasta [ásta] ‘until’, by the groove fricative realization of the multiple trill phoneme /rr/ (carro [káo] ‘car’), and assibilation of word-final /r/ (comer [koméɹ] ‘to eat’), by retention of the palatal lateral phoneme /λ/ (calle [káλe] ‘street’), and by devoicing and shortening of atonic vowels, especially those in contact with /s/ (presidente [presiδénte] ‘president’). In the Afro-Bolivian dialect final /s/ is either aspirated or lost, a feature found in eastern (lowland) Bolivian dialects as well as throughout the Caribbean and southern Spain (entonces [entónse(h)] ‘then’). Word-final /r/ usually drops (mujer [muhé] ‘woman’), and is completely absent in all verbal infinitives (hablar [aβlá] ‘to speak’). The palatal lateral phoneme /λ/ has merged with /y/ (a phenomenon known as yeísmo), as in much of the rest of the Spanish-speaking world (llover [yoβé] ‘to rain’). Unlike in the highland Bolivian dialect, atonic vowels are not reduced, and stressed vowels are frequently elongated far in excess of the duration found in other Spanish dialects. These elongated stressed vowels are accompanied by a striking circumflex intonation (sharply rising and falling contours on the stressed syllable, with pitch change of up to an octave) entirely unlike patterns found in other Latin American dialects. Afro-Bolivian . The fieldwork in 2005 was partially supported by a grant from Penn State’s Africana Research Center, for which the author expresses deep gratitude.

 John M. Lipski

speech makes its greatest contribution to the reconstruction of early bozal Spanish contact dialects in the restructured DPs and VPs, since it is here that one can observe patterns that deviate from all other contemporary and historically documented varieties of Spanish.

.

The Afro-Yungueño DP

.. Lack of gender agreement In the quintessential Afro-Yungueño DP, at least five phenomena distinguish these configurations from patrimonial Spanish dialects: (1) lack of noun-adjective gender agreement; (2) invariant plurals, that is, no plural suffixes on nouns, adjectives, or determiners; (3) use of a single invariant definite article; (4) elimination of definite articles in generic constructions; (5) “stripped plurals”, i.e., the retention of plural /s/ only on the first element of plural DP. Given more than 400 years of constant contact with standard Spanish, the Afro-Bolivian dialect does not always exhibit restructured DPs in their pure form (with all five points of divergence from modern Spanish), but enough examples of each remain in contemporary speech to allow for a reasonable extrapolation. The gender and number marking in the Afro-Bolivian DP represent the most consistent deviation from other varieties of Spanish and constitute a link to the time when bozal pidginized Spanish was in use. In all dialects of Spanish, all elements of a DP agree in gender and number; masculine gender is typically marked by the suffix -o and feminine gender with -a, although many nouns and adjectives end in -e or a consonant, in which case gender marking is opaque. In the Afro-Yungueño basilect, lack of gender concordance is always manifested by the masculine gender, the most frequent and presumably unmarked form in Spanish. In the basilect there is no gender concord at all; the Spanish masculine form combines with masculine determiners such as el (singular) and lu(s) (plural and sometimes singular). In the decreolizing contemporary speech lapses in gender agreement co-occur with correct gender concord, with and without plural concord, but when lapses occur it is always the masculine form that prevails. Actually recorded examples include the following (modern Spanish equivalents are in square brackets, with the relevant morphological endings underlined): (1)

loh persona mayó . person older ‘the older people’

[las personas mayores]

camisa blanco shirt white.. ‘white shirts’

[camisas blancas]

hartos viuda many.. widow ‘many widows’

[hartas viudas]

Chapter 9. Afro-Bolivian Spanish 

unos quince mula some.. fifteen mule ‘fifteen mules’

[unas quince mulas]

nuestro cultura antigo our.. culture old.. ‘our old culture’

[nuestra cultura antigua]

ese mujé .. woman ‘that wife’

[esa mujer]

otro cosa other.. thing ‘another thing’

[otra cosa]

algunos enfermedá some.. illness ‘some illnesses’

[algunas enfermedades]

These examples demonstrate that plural marking is more robust than gender marking, but that both may disappear in Afro-Bolivian speech.

.. Stripped and invariant plurals All varieties of Spanish maintain morphological signaling of the distinction singularplural, although erosion of final and intervocalic consonants may lead to non-canonical combinations.3 The widespread loss of word-final consonants, e.g., in rustic dialects of southern Spain, often leads to phonological restructuring in vernacular speech, resulting in the loss of canonical plural endings: árbo < árbol [árboles] ‘trees’; re < res [reses] ‘cows’; etc. (Carrasco Cantos 1981: 99; Salvador Plans 1987: 40). In Afro-Iberian creole languages nouns and adjectives remain invariable, in a form derived from the singular (and in the case of adjectives the masculine gender); when pluralization must be indicated it is by means of established plural markers (e.g., the postposed 3rd person plural pronoun nan in Papiamentu, the preposed plural particle ma in Palenquero). In AfroBolivian speech not only does the frequent elision of word-final /s/ obliterate overt plural marking in nouns and adjectives for which this morpheme is the only indication of plurality, but plural marking also disappears in nouns and adjectives ending in consonants, for which the usual Spanish plural allomorph -es is required. The end

. For examples in some rustic dialects of Andalusia, the common loss of word-final /l/ is accompanied by loss of intervocalic /l/; this may lead to a new type of plural formation: arbo/ árboe < arbol, perá < peral/perae, animá < animal/animae (Moya Corral 1979: 81–4).

 John M. Lipski

result is a complete suspension of plural marking in nouns and adjectives. Examples of invariant plurals include: (2)

lu peón lu mujé persona mayó algunos enfermedá lu profesor tres real

< los peones < las mujeres < personas mayores < algunas enfermedades < los profesores [< reales]

‘the peasants’ ‘the women’ ‘older people’ ‘some illnesses’ ‘the teachers’ ‘three reales [a small coin]’

In what appears to be partially decreolized evolutionary step past the complete lack of overt plural marking, the traditional Afro-Yungueño DP at times presents an alternative form of plural marking that separates it from all contemporary and most historically attested Afro-Hispanic dialects and brings it closer to vernacular Brazilian (including Helvécia; Lipski 2006b) and Angolan Portuguese (Lipski 1995b): marking of plural /s/ on only the first element (usually a determiner) of plural DPs. This form of plural-marking is not found in any European Portuguese dialect, and is one of the features that form the basis for theories that basilectal Brazilian Portuguese is a semicreole formed through contact with African languages during the colonial period. Typical Afro-Yungueño examples include: (3)

esos fiesta .. party ‘those parties’

[esas fiestas]

luh guagua jóven .. child young. ‘the young children’

[los guaguas jóvenes]

algunos cosa some.. thing ‘some things’

[algunas cosas]

personah mayó person. older. ‘older people’

[personas mayores]

unas muñeca some.. statue ‘some statues’

[muñecas]

Occasional examples of similar constructions have been found in the Afro-Hispanic dialects of the Chota Valley in highland Ecuador (Lipski 1986a) and in the Chocó regional of Colombia (Caicedo 1977, 1992), as well as among the L2 Spanish speakers of Equatorial Guinea (Lipski 1985), but never in systematic fashion. In the Afro-Yungueño dialect “stripped plurals” are not as frequent as in vernacular Brazilian Portuguese but they occur often enough to be considered a legitimate feature of this dialect. Marking plural /s/ on the first element of DPs has been a component of Afro-Iberian

Chapter 9. Afro-Bolivian Spanish 

language at least since the late 16th century, and several clear examples are found in early 17th century Portuguese texts. The first unequivocal examples of stripped plural DPs in Spanish come in the villancicos or Christmas carols written by the Spanish nun Sor Juana Inés de la Cruz in Mexico in the late 17th century. These songs are written in bozal Spanish and purport to represent the speech of African slaves newly arrived in the Americas. The poems contain several instances of stripped plurals, such as: (4)

las Leina .. queen ‘the queens’

[las reinas]

estos Parre Mercenaria .. priest mercenary.. ‘these mercenary priests’

[estos padres mercenarios]

los demoño .. demon ‘the demons’

[los demonios]

lus nenglu .. black.. ‘the blacks’

[los negros]

turo las Negla all.. the.. black.. ‘all the black women’

[todas las negras]

Among varieties of contemporary Portuguese, in addition to occurring in vernacular Brazilian and Angolan Portuguese, stripped plural DPs have been documented for the vestigial Portuguese of Macau, the last decreolized vestiges of the formerly vigorous Portuguese-derived creole spoken in Macau and Hong Kong. In earlier times, pluralization of nouns was effected by simple reduplication: chino-chino ‘Chinese.’, coisa-coisa ‘things’, etc. This was eventually replaced by an DP in which plural /s/ was marked only on the first determiner (Batalha 1974: 10): (5)

as casa .. house ‘the houses’

[as casas]

três pataca [três patacas] three pataca ‘three patacas’ (monetary units) dois mão two hand ‘two hands’

[dois mãos]

Similar stripped plurals were attested for the pidginized Spanish spoken by Chinese contract laborers in 19th century Cuba (Lipski 1998c, 1999b), many of whom learned

 John M. Lipski

Spanish from African-born bozales, Papiamentu-speaking laborers from Curaçao, and other creole-speaking cane-cutters. Some Chinese laborers also spoke Macau Portuguese creole, since this was the port from which the Portuguese recruited laborers. (6)

un tlaje le tleinta a.. suit of thirty ‘a suit for 30 soles’

sol [soles] sol (Santa Cruz 1982: 294) [Peru]

lo tlen [los trenes] palan y no pasa na .. train stop.3 and  happen.3 nothing ‘the trains come by and nothing happens’ (Feijóo 1981: 150-1) [Cuba]

Agreement in the Afro-Yungueño DP is much more rudimentary than, for example, in the Helvécia Portuguese DP as described by Baxter and Lucchesi (1993) and Baxter et al. (1997), and totally restructured DPs are the norm in Afro-Yungueño, as opposed to their vestigial and ephemeral character in Helvécia Portuguese. For example Lucchesi (1998: 90) summarizes Helvécia data that show that even Helvécia speakers over the age of 60 (in 1990) produced nearly 90% correct noun-adjective gender agreement, with younger speakers approaching 100% agreement. When speaking entirely in the traditional dialect, elderly Afro-Bolivians produce upwards of 90% of their DPs with no gender or number marking.

.

The Afro-Yungueño VP

.. Invariant 3 verb forms Together with the restructured DP, the nature of the Afro-Yungueño VP is the most significant departure from patrimonial Spanish patterns, in that most person and number agreement is absent. No other variety of Spanish past or present departs from the full subject-verb agreement system inherited from Latin. Afro-Yungueño Spanish exhibits more fully restructured VPs than its closest “competitor” among Afro-Iberian speech communities, contemporary Helvécia or vernacular Brazilian Portuguese, where some person and number agreement remains. In nearly all cases the Afro-Yungueño verb lacking person and number agreement is derived from the Spanish 3rd person singular (3). The most “creole-like” combination in Helvécia Portuguese, the combination of first-person singular reference with 3 verbs (e.g., eu trabalha ‘I work’), not found in even the most vernacular forms of other Brazilian dialects, only occurs 35% of the time among speakers born before 1930, and virtually disappears from the speech of younger community members (Lucchesi 1998: 89). Naro (1998) has presented evidence of very occasional lapses of subject-verb agreement in contemporary and earlier European Portuguese, but never involving the first-person singular. Comparative evidence from historical reproductions of bozal Spanish, and from contemporary L2 varieties and vestigial or semi-speaker varieties of Spanish (Lipski 1986c, 1996b), as well as from the first stages of Spanish child language, confirm the unmarked status of the 3 verb form (Bybee 1985; Bybee Hooper 1980). The Spanish of Equatorial Guinea and

Chapter 9. Afro-Bolivian Spanish 

the Portuguese of Angola also make frequent use of the 3 as invariant verb, although the respective metropolitan languages are the official standard. Early child language in both Spanish and Portuguese also favors the 3 as the unmarked form (Simões 1976: 47; Simões and Stoel-Gammon 1979). Unlike vestigial, L2, and literary examples, in the basilectal Afro-Yungueño dialect use of the 3 as default invariant verb is categorical, not occasional. Some spontaneously produced examples are: (7)

nojotro tiene [tenemos] jrutita 1 have.3 fruit. ‘we have fruit’ yo no entiende [entiendo] eso 1  understand.3 that ‘I don’t understand that’ yo creció [crecí] junto con Angelino 1 grew.3 together with Angelino ‘I grew up with Angelino’ nojotro creció [crecimos] loh do 1. grew.3 .. two ‘the two of us grew up’ ello salía [salían] mi avisá 3 leave.3 1 warn. ‘they came to warn me’ ¿de qué nojotro pobre va [vamos] viví? of what 1 poor. go.3 live. ‘What are we poor folks going to live on?’ nojotro trabajaba [trabajábamos] hacienda 1. work.3 hacienda ‘we worked on the haciendas’ lo patrón siempre tenía [tenían] partera .. owner always had.3 midwife ‘the landowners always had midwives’ leña no cargaba como nojotro cargaba [cargábamos] firewood  carried.3 like 1 carried.3 ‘[nobody] carried firewood like we carried firewood’ yo sí lo carga [cargo] 1 yes it.. load.3 ‘I really do load [the coca]’ lo que nojotro ta [estamos] hablando that.  1 .3 speaking ‘What we’re talking about’

 John M. Lipski

para ele e extraño como nojotro habla [hablamos] for 3 be. strange how 1 speak.3 ‘for him it’s strange the way we talk’ nojotro no sabía [sabíamos] nada 1  knew.3 nothing ‘we didn’t know anything’ qué día yo va í [voy a ir] what day 1 go.3 go. ‘what day I’m going to go’

.. Possible combinations of ta +  One of the most controversial issues in the study of Afro-Iberian language is the possible monogenesis of Afro-Romance creoles. A key feature of such a putative creole would be a verb system based on preverbal TMA particles combined with invariant verb stems derived from the infinitive. The various proposals are summarized and analyzed in Lipski (1986b, 1987, 1991, 1992, 1993, 1996a, 1998a, 1999a, 2002a). Ziegler (1976) and later Castellanos (1990) offer grammatical sketches of what such a bozal-derived creole grammar would look like, based on Afro-Cuban examples taken from literary and folkloric texts. A key feature of this putative Spanish creole is a VP based on the imperfective or progressive particle ta (presumably derived from Spanish/Portuguese estar), a common denominator found in all Spanish- and Portuguese-derived creoles worldwide. The existence of the particle ta in all creoles derived from Spanish and Portuguese has lent considerable support to monogenetic theories based on the notion of a pidginized Portuguese “reconnaissance language” (e.g., as in Naro 1978: 342).4 However, despite the presence of ta or its congeners among Ibero-Romance creoles, major differences in the use of ta among these creoles suggest multiple routes of evolution. In the pre-19th century Spanish bozal corpus, there are no examples of ta or está(r) used with an invariable verb stem in a fashion suggestive of its use as a preverbal particle, contrary to the proposal of Naro (1978). However, the 19th century Cuban bozal corpus provides a different panorama, with some apparent instances of ta as an aspectual particle. This has led to claims that Cuban bozal Spanish became a true creole, sharing with Palenquero and Papiamento (and with Cape Verdean) an earlier Afro-Lusitanian heritage (e.g., Megenney 1984): (8)

como que like 

yo ta cuchá la gente yo ta  .3 hear. .. people 1 .3

. Indeed, a creole-like verb structure using the particle ta is not present in any AfroPortuguese pidgin texts, from the 15th century to the 20th, although found in some fashion in all Afro-Portuguese creoles. Found in many texts is the portmanteau verb sar (apparently a fusion of ser and estar) or santar, possibly a fusion of sentar and estar (Lipski 1999c, 2002b).

Chapter 9. Afro-Bolivian Spanish 

mirá gente mucho look. people much ‘Since I’ve heard the people I look at people a lot’ (Manuel Cabrery Paz, “Exclamaciones de un negro”; Guirao 1938) yo ta robá un gaína jabá  .3 steal. one.. hen spotted.. ‘I stole a spotted hen’

(Cabrera 1976)

horita ta bení pa cá now .3 come. for here ‘Now (she) is coming here’ (Ignacio Villa, “Drumi, Mobila”; Guirao 1938) ta juí, ta pujá mí .3 flee. .3 push. 1 ‘(he) flees, (he) pushes me’

(Suárez y Romero 1947)

primero ta llorá na má first .3 cry. nothing more ‘First (she) cries no more’

(Santa Cruz 1908)

In the Afro-Caribbean literary examples, ta typically combines with an invariant stem derived from the Spanish infinitive lacking final /r/, an established phonetic reduction which began as early as the 16th century (Lipski 1995a). A probable source for some instances of ta in Afro-Cuban texts is Papiamentu, which was present in 19th century Cuba and Puerto Rico as thousands of sugar cane cutters were taken from other Caribbean islands during the sugar plantation boom (Lipski 1993, 1996a, 1998a, b, 1999a, 2002a). Pockets of Papiamentu speakers are documented for Cuba and Puerto Rico, and other Papiamentu elements penetrated Afro-Cuban Spanish.5 Adding to the plausibility that the Afro-Cuban instances of ta are due to language contact is the virtually complete absence of such constructions in the many other Afro-Hispanic corpora from Latin America. This proposal does not completely invalidate the monogenetic hypothesis, since the presence of the particle ta in Afro-Cuban texts could in principle stem from an earlier Portuguese-derived creole, twice-removed from Caribbean Spanish via the intermediate stage of Papiamentu. The search must be extended to instances of ta as putative preverbal particle in circumstances in which relexification of a Portuguese-derived creole is unlikely. Outside of Cuba and Puerto Rico, ta used as putative verbal particle in AfroHispanic speech makes only a few ghost appearances, none of which holds up robustly under closer scrutiny. Tompkins (1981: 311) cites an older Afro-Peruvian informant in Cañete, who recalled a line from an old song: Lima ta hablar y Cañete ta pondé ‘Lima

. Another source of ta + Vinf constructions in Afro-Caribbean Spanish is phonetic reduction of the gerund, suggested by such half-reduced constructions as pavo real ta bucán paló ‘the peacock is looking for a tree’ (Cabrera 1983).

 John M. Lipski

speaks and Cañete responds’. My own fieldwork in the same region (conducted in 2003) failed to uncover any recollection of this song, or other attestations of ta used as particle in Afro-Peruvian speech or song (the informant who had provided the quotation to Tompkins was deceased). This suggests that at least some creoloid verb forms may have occasionally surfaced in Afro-Peruvian speech, although apparently never coalescing into a consistent pattern. In her analysis of vestigial Spanish in Trinidad, Moodie (n.y.) uncovered one instance of what she believes to be a creoloid construction with ta: la esposa cuasi ta olvidá el español ‘his wife is almost forgetting Spanish’. According to Moodie this combination is very infrequent in the speech of even the oldest community members. Moodie’s putative ta example may be the result of phonetic erosion not only of forms of the verb estar to ta (frequent in all colloquial varieties of Caribbean Spanish), but also of the gerund, which accompanies estar in progressive constructions: está hablando > [es]tá hablá[ndo] ‘is talking.3’. Vestigial Trinidad Spanish has many other examples of the erosion of final syllables (Lipski 1990; Moodie 1986). Another possible instance of the construction ta +  found outside of Cuba appears in an enigmatic bozal poem by the Panamanian writer Víctor Franceschi (1956: 30): (9)

[. . .] Si te pica por allá, cuando tu tá tlabajá yo te puee asegurá que tu vaj a recordá lo que mama tá avertí . . .

if it bites you there when you are working I can assure you that you will remember what your mother warned you

The construction tú tá tlabajá ‘you work’ and mama tá avertí ‘mother warns you’ are the only known literary examples of the construction ta +  outside of Cuba and Puerto Rico. In Afro-Yungueño Spanish TMA particles are not a usual part of the VP. However, the basilectal Afro-Yungueño dialect occasionally allows, in rapid speech, for constructions based on the auxiliary verb estar (usually reduced to invariant ta in casual speech) plus what has the superficial appearance of be the Spanish infinitive (lacking the final /r/ as in all Afro-Yungueño speech), instead of the usual gerund ending in -ando or -iendo. Recorded examples from the Afro-Yungueño corpus include: (10)

yo ta vení [estoy viniendo] di tal parte 1 .3 come. from such place ‘I’m coming from such and such a place’ nojotro ta hablá [estamos hablando] bien 1 .3 speak. well ‘we talk well’ ¿qué oté ta tomá [está tomando]? what 2 .3 drink. ‘what are you drinking?’

Chapter 9. Afro-Bolivian Spanish 

yo ta tomá [estoy tomando] mi plato 1 .3 take. my.. plate ‘I am eating my food’ ¿ande pue oté ta [estás yendo] í? where then 2 .3 go. ‘where are you going?’ ¿qué oté ta hacé [estás haciendo]? what 2 .3 do. ‘what are you doing?’ eje perro ta ladrá [está ladrando] .. dog .3 bark. ‘that dog is barking’

In the Afro-Bolivian corpus the appearance of ta +   is infrequent,6 but when explicitly queried on this combination, speakers of the Afro-Yungueño dialect do not consider it as a true component of their speech; when presented with specific examples extracted from the corpus all acknowledge the construction as possible. They view the use of ta +  as a performance phenomenon, and indeed a comparison with the remainder of the Afro-Yungueño corpus points to phonetic erosion in rapid and unguarded speech as the locus of this construction, which is not an integral part of the dialect’s verb system. This is amply demonstrated in the following fragment, from a man remembering how acquaintances greeted one another during chance encounters on the road: (11)

¿di ande pue compa ta viniendo? from where then compadre .3 come. ‘where are you coming from, compadre?’ yo ta vení [estoy viniendo] di mi casa 1 .3 come. from my house ‘I’m coming from my house’ ¿ande pue compa ta indo? where then compadre .3 go. ‘Where are you going, compadre?’

In this example ta combines with a rapidly pronounced gerund, which then erodes to a form similar to the infinitive, only to re-emerge as a full gerund a moment later. The free alternation between progressive constructions with a fully realized gerund and eroded combinations that resemble the Spanish infinitive offers a model for . Some speakers interviewed never used this combination. Among those that do produce combinations involving ta plus truncated infinitive/gerund, the overall frequency is less than 10% of all instances in which a simple present or present progressive verb could be used.

 John M. Lipski

how ta +   structures could arise spontaneously in emergent AfroIberian creole languages. The transitory performance nature of such configurations in Afro-Yungueño speech and the fact that the speakers themselves do not regard ta +   as the “correct” form precludes including the latter in a definitive genealogy of creole languages based on the use of the particle ta. The Afro-Bolivian data are in principle incompatible with the monogenetic hypothesis (although there is no convincing documentation that any form of Portuguese or Afro-Portuguese pidgin or creole language ever reached colonial Bolivia), but they do provide a scenario for multiple sources of the particle ta.

.

Decreolization and implicational relationships in Afro-Yungueño Spanish

The basilectal Afro-Yungueño dialect shows evidence of an early restructuring of Spanish in favor of simpler and morphologically less marked configurations. This supposition is supported by data on variation from the contemporary dialect. Given the gradual displacement of the traditional Afro-Bolivian dialect by modern Bolivian Spanish over the past three generations, there is considerable morphosyntactic variation across generations and as a factor of exposure to standard Spanish. There are also regional isoglosses radiating outward from a geographical locus where the most basilectal variety is spoken. Both the regional and generational variation are systematic enough as to reveal unidirectional implicational relationships. Ranging from the most creole-like features to the least, the Afro-Bolivian implicational scale is as follows: Table 1. Implicational relationships in Afro-Yungueño speech TRAIT 3 verb forms for 1 3 verb forms for 1 3 verb forms for 3 Invariant plural article lu(s) No gender concord in adjectives Stripped plurals Invariant plurals

EXAMPLE → → → → → → →

yo va trabajá ‘I’m going to work’ nojotro va trabajá ‘we’re going to work’ eyu(s) va trabajá ‘they are going to work’ lu(s) mujé ‘the women’ esos fiesta ‘those parties’ algunos cosa ‘some things’ lu(s) patrón ‘the landowners’

In other words speakers who exhibit a given trait will also use all the traits found lower on the chart (e.g., those who say yo va trabajá ‘I am going to work’ will also say nojotro va trabajá ‘we are going to work’, lu(s) mujé ‘the women’, lu(s) patrón ‘the landowners’,), while features higher on the chart will be absent (there are, for example, speakers who say lu(s) patrón but not lu(s) mujé, eyu(s) va trabajá ‘they are going to work’, nojotro va trabajá, etc.). These implicational relationships provide robust

Chapter 9. Afro-Bolivian Spanish 

evidence for the maximally unmarked status of the 3 form in Spanish7 (as opposed to the infinitive; Lipski 2002c) and to the masculine singular. An additional complication is found in the grammars of some mesolectal Afro-Yungueño speakers, who effect most cases of subject-verb agreement when there is a single main verb, but suspend agreement on auxiliary verbs, particular ir in the periphrastic future combination ir (a) + , and in estar +  progressive combinations. Examples include: (12)

¿de qué nojotro pobre va [vamos] viví? of what 1 poor. go.3 live. ‘What are we poor folks going to live on?’ lo que nojotro ta [estamos] hablando este rato that  1 .3 speak.  time ‘What we’re talking about right now’ qué día yo va í [voy a ir] what day 1 go.3 go. ‘what day I’m going to go’ eyo va leé [van a ler] 3 go.3 read. ‘they are going to read’ nojotro va leé[vamos a ler] 1 go.3 read. ‘we are going to read’

It is also more common to hear the 3 used instead of the 1st person singular in the preterit than in the present tense. Baxter, in Mello et al. (1998: 126-7) documents a similar tendency for Helvécia Portuguese semicreole. In Afro-Yungueño Spanish (and judging by published examples also in Helvécia and Angolan Portuguese), gender agreement is suspended more frequently than number agreement, and in an implicational fashion. In other words, whereas an AfroBolivian speaker may produce combinations such as: (13)

esos hierba .. herb ‘those herbs’ algunos enfermedá some.. illness ‘some illnesses’

[esas hierbas]

[algunas enfermedades]

. In Afro-Bolivian Spanish the position of the stress accent strongly correlates with the distinction infinitive.3 finite verb form in Spanish; forms derived from the infinitive end in stressed vowels, while 3 verb forms have antepenultimate stress. In some creole languages (e.g., Philippine Creole Spanish, Papiamentu) there are many ambiguous verbs that could derive either from the infinitive or the 3 but this does not appear to be the case in Afro-Bolivian Spanish.

 John M. Lipski

luh persona mayó [las personas mayores] .. person older. ‘the adults’

in which number is marked (albeit only on the first element) but in which there is no gender marking, there are no observed configurations such as *esa hierba for plural esas hierbas. From the perspective of restructuring under imperfect language acquisition and subsequent decreolization, this suggests that number features are acquired before gender features, and that in decreolization number features will spread to extended projections before gender features. This is not surprising in view of the fact that grammatical number marks a semantically prominent distinction (one versus many), whereas in Ibero-Romance grammatical gender concordance is almost always semantically empty. Even with semantically feminine nouns such as mujer ‘women’, yegua ‘mare’, gallina ‘hen’, the feminine gender markers attached to determiners and adjectives in the DP serve no semantic function. Only in cases of epicene gender – very rare in Spanish and even less frequent in the daily discourse of the Bolivian Yungas and almost always identifiable through other discourse markers – does grammatical gender marking on determiners and adjectives serve to distinguish the gender of the noun: el/la agente ‘the agent[/]’, el/la estudiante ‘the student[/]’. Experimental studies demonstrate that processing of grammatical gender in the absence of conceptual/biological information about the sex of the object is more costly; i.e., the processing of gender in la casa ‘the house’ requires more processing time than for la mujer ‘the woman’ (e.g., Vigliocco and Franck 1999).8 The predominance of number over gender has been well documented cross-linguistically in first- and second-language acquisition as well as language impairments (e.g., De Vincenzi 1999; Di Domenico and De Vincenzi 1999; Eberhard 1997), and the decreolizing Afro-Bolivian data bear out this hierarchy. Moreover there is much cross-linguistic experimental evidence that production of grammatical gender in bare nouns requires additional processing time (e.g., Cubelli et al. 2005 for Italian; also Schriefers and Jescheniak 1999). In other words grammatical gender information is selected by the speaker whether or not “needed” for the utterance about to be produced. This occurs independently of the phonological form of the word, i.e., whether or not one of the canonical -o/-a endings is present. Cubelli et al. (2005: 52) analyze this result as “reflecting a competitive lexical selection due to an abstract grammatical gender feature rather than to the morphological or phonological similarity of [the relevant nouns]”. In addition, they state that [. . .] to produce a given noun, the corresponding lexical-semantic and lexicalsyntactic representations, specifying meaning and grammatical properties respectively, have to be selected before accessing its phonological form [. . .] the . There are dissenting viewpoints, however. For example Domínguez et al. (1999) report experimental findings that suggest that in Spanish “gender information is accessed more straightforwardly than number in an inflected word” (ibid.: 495).

Chapter 9. Afro-Bolivian Spanish 

selection of semantic and grammatical features is conducted independently and [. . .] the selection of the lexical form of a given noun is achieved only when competition at both semantic and syntactic levels has been resolved. (ibid.: 53)

They postulate that the semantic information is selected before the syntactic information. By extension, eliminating the grammatical gender category consequently eliminates the need for gender selection, and produces a more efficient (“faster”) processing strategy. De Vincenzi (1999) provides Italian experimental data that indicate that “number information is used in an earlier stage of antecedents identification (where syntactic information is used), while gender information is used at a later stage (where lexical and semantic information are used)” (ibid.: 551). She also suggests that while number heads an autonomous syntactic projection, gender never does so, not even in the case of variable gender. Finally, Caramazza et al. (2001) provide cross-linguistic information that production and processing of determiners is complicated by the necessity to retrieve gender and number information; “determiner selection in a given language occurs at the same point for all determiners, even though some of them could be selected earlier. In the case of Spanish, even though masculine determiners could be selected early, they are nevertheless selected at the same late point as feminine determiners” (ibid.: 223). Many of the L2 Italian data on the acquisition of grammatical gender presented in Chini (1995) reinforce these observations. Afro-Yungueño Spanish reduces this complexity by effectively eliminating the gender and number marking on the determiners. The processing of grammatical number, on the other hand, is not symmetric, in that there is experimental evidence suggesting that whereas plural is a semantically and syntactically marked category, words lacking a plural affix are not semantically singular but rather unmarked for number (e.g., Berent et al. 2005). This hypothesis is supported by the Afro-Yungueño preference for invariant plurals of the sort (luh) mujé [las mujeres] ‘the women’, luh varón [los varones] ‘the male children’ even in the most acrolectal forms of the dialect. Rather than representing a mismatch between semantic and syntactic features – a configuration that should augment rather than reduce production and processing difficulty – the bare plural is an unmarked form, whose plural reference can easily be extracted from the preceding plural determiner or from the surrounding discourse.

.

Conclusions: Afro-Yungueño Spanish as a creole prototype

Do the Afro-Bolivian data represent the final stages in the decreolization of an earlier Palenquero-like Spanish creole or the remnants of a stable restructured but not creolized variety of Spanish that co-existed with highland Bolivian Spanish since its inception? A glance at the radically simplified VP and DP of the basilectal Afro-Yungueño dialect suggests that a full-fledged creole once existed here. On the other hand the

 John M. Lipski

known historical and demographic data do not offer the conditions typically associated with creolization. Afro-Bolivians never lived in maroon communities nor were they totally cut off from native speakers of Spanish.9 Whether or not a true Afro-Hispanic creole once existed in Bolivia, the Afro-Bolivian data represent the consolidation of bozal Spanish into a stable restructured variety fueled by considerations of processing simplicity. The contemporary speech community exhibits clear signs of decreolization, gradually bringing the Afro-Yungueño grammar into alignment with Spanish in a series of steps that reflect the relative markedness of Spanish agreement features. This represents a trade-off between increased processing requirements and a closer approximation to the national linguistic standard; the latter has been the stronger factor since the arrival of widespread public education in the Afro-Yungueño communities. The survival of this restructured variety in a robust state through the end of the 20th century, together with the more recent decreolization in the direction of standard Spanish nominal and verbal morphology, suggest that the traditional Afro-Yungueño dialect can stand as a prototype for the spontaneous creolization of Spanish in isolation. By extension, the directional implications found in the decreolizing dialects provide independent confirmation of hypotheses about the relative markedness of gender and number inflection in Romance languages. The data obtained from recently “discovered” Afro-Hispanic isolates such as Helvécia Portuguese and Afro-Yungueño Spanish should energize the search for similar linguistic tide pools, before they succumb to the inevitable threat of mass-produced language globalization.

References Angola Maconde, J. 2000. Raíces de un pueblo: cultura afroboliviana. La Paz: Producciones CIMA, Embajada de España, Cooperación. Batalha, G. 1974. Língua de Macau. Macau: Imprensa Nacional. Baxter, A. and Lucchesi, D. 1993. Procesos de descrioulização no sistema verbal de um dialeto rural brasileiro. Papia 2. 59–71. Baxter, A., Lucchesi, D., and Guimarães, M. 1997. Gender agreement as a “decreolizing” feature of an Afro-Brazilian dialect. Journal of Pidgin and Creole Languages 12. 1–57.

. However, it appears that early in their history, Afro-Bolivians were surrounded by a monolingual Aymara-speaking majority and had little contact with native speakers of Spanish, so that conditions favoring creolization could have existed. Although Afro-Yungueño Spanish has co-existed with Aymara-influenced varieties for several centuries and has absorbed many Aymara elements, the qualitative differences between Aymara-Spanish interlanguage and AfroYungueño Spanish are significant enough to render it unlikely that Afro-Yungueño speech is simply a replica of the L2 Spanish spoken by Afro-Bolivians’ Aymara neighbors. Lipski (2006a, b, fc) summarizes these differences.

Chapter 9. Afro-Bolivian Spanish  Berent, I., Pinker, S., Tzelgov, J, Bibi, U. and Goldfarb, L. 2005. Computation of semantic number from morphological information. Journal of Memory and Language 53. 342–58. Bowser, F. 1974. The African slave in colonial Peru 1524-1650. Stanford: Stanford University Press. Bybee Hooper, J. 1980. Child morphology and morphophonemic change. In Fisiak, J. (ed.). Historical morphology. The Hague: Mouton. 157–88. Bybee, J. 1985. Morphology: A study of the relation between meaning and form. Amsterdam: John Benjamins. Cabrera, L. 1976. Francisco y Francisca: chascarrillos de negros viejos. Miami: Editorial C.R. Cabrera, L. 1983. El monte. 3rd ed. Miami: Editorial C.R. Caicedo, M. 1977. Chocó mágico folklórico. Quibdó: Gráficas Universitarias del Chocó. Caicedo, M. 1992. Poesía popular chocoana. Bogotá: Departamento de Publicaciones de Colcultura. Caramazza, A., Miozzo, M., Costa, A., Schiller, N. and Alario, F.-X. 2001. A crosslinguistic study of determiner production. In Dupoux, E. (ed.). Language, brain, and cognitive development: essays in honor of Jacques Mehler. Cambridge: MIT Press. 209–26. Carrasco Cantos, P. 1981. Contribución al estudio del habla rural de Baeza (Jaén). Jaén: Instituto de Estudios Giennenses. Castellanos, I. 1990. Grammatical structure, historical development, and religious usage of Afro-Cuban bozal speech. Folklore Forum 23.1-2. 57–84. Chini, M. 1995. Genere grammaticale e acquisizione: aspetti della morfologia nominale in italiano L2. Milano: Francoangeli. Crespo, A. 1977. Esclavos negros en Bolivia. La Paz: Academia Nacional de Ciencias de Bolivia. Cubelli, R., Lotto, L., Paolieri, D., Girelli, M. and Job, R. 2005. Grammatical gender is selected in bare noun production: evidence from the picture-word interference paradigm. Journal of Memory and Language 53. 42–59. Cuche, D. 1981. Perou nègre. Paris: L’Harmattan. De Vincenzi, M. 1999. Differences between the morphology of gender and number: evidence from establishing coreferences. Journal of Psycholinguistic Research 28. 537–53. Di Domenico, E. and De Vicenzi, M. 1999. A distinction among phi-features: gender and number in the retrieval of pronoun antecedents. Rivista di Linguistica 11. 41–71. Domínguez, A., Cuetos, F., and Segui, J. 1999. The processing of grammatical gender and number in Spanish. Journal of Psycholinguistic Research 28. 485–98. Eberhard, K.M. 1997. The marked effect of number in subject-verb agreement. Journal of Memory and Language 36. 147–64. Feijóo, S. (ed.). 1981. Cuentos populares cubanos de humor. Havana: Editorial Letras Cubanas. Franceschi, V. 1956. Carbones. Panama: Departamento de Bellas Artes y Publicaciones del Ministerio de Educación. Guirao, R. 1938. Orbita de la poesía afrocubana 1928–1937. Havana: Ucar García. Harth-Terré, E. 1971. Presencia del negro en el virreinato del Perú. Lima: Editorial Universitaria. Harth-Terré, E. 1973. Negros e indios: un estamento social ignorado del Perú colonial. Lima: Librería/Editorial Juan Mejía Baca. Lipski, J.M. 1985. The Spanish of Equatorial Guinea. Tübingen: Max Niemeyer. Lipski, J.M. 1986a. Lingüística afroecuatoriana: el valle del Chota. Anuario de Lingüística Hispanica (Valladolid) 2. 153–76. Lipski, J.M. 1986b. Sobre la construcción ta + infinitivo en el español ‘bozal’. Lingüística Española Actual 8. 73–92.

 John M. Lipski Lipski, J.M. 1986c. Creole Spanish and vestigial Spanish: evolutionary parallels. Linguistics 23. 963–84. Lipski, J.M. 1987. The construction ta + infinitive in Caribbean bozal Spanish. Romance Philology 40. 431–50. Lipski, J.M. 1990. Trinidad Spanish: implications for Afro-Hispanic language. Nieuwe WestIndische Gids 62. 7–26. Lipski, J.M. 1991. Origen y evolución de la partícula ta en los criollos afrohispánicos. Papia 1.2. 16–41. Lipski, J.M. 1992. Origin and development of ta in Afro-Hispanic creoles. In Byrne, F. and Holm, J. (eds). Atlantic meets Pacific: A global view of pidginization and creolization. Amsterdam: John Benjamins. 217–31. Lipski, J.M. 1993. On the non-creole basis for Afro-Caribbean Spanish. Research Paper No. 24. Latin American Institute, University of New Mexico. Lipski, J.M. 1994. El español afroperuano: eslabón entre Africa y América. Anuario de Lingüística Hispánica 10. 179–216. Lipski, J.M. 1995a. Literary ‘Africanized’ Spanish as a research tool: Dating consonant reduction.” Romance Philology 49. 130–67. Lipski, J.M. 1995b. Portuguese language in Angola: Luso-creoles’ missing link? Presented at the annual meeting of the American Association of Teachers of Spanish and Portuguese (AATSP), San Diego, California, August 1995. Available at http://www.personal.psu.edu/ jml34/papers.htm. Lipski, J.M. 1996a. Contactos de criollos en el Caribe hispánico: contribuciones al español bozal. América Negra 11. 31–60. Lipski, J.M. 1996b. Los dialectos vestigiales del español en los Estados Unidos: estado de la cuestión. Signo y Seña 6. 459–89. Lipski, J.M. 1998a. Latin American Spanish: Creolization and the African connection. Publications of The Afro-Latin American Research Association (PALARA) 2. 54–78. Lipski, J.M. 1998b. El español bozal. In Perl, M. and Schwegler, A. (eds). América negra: Panorámica actual de los estudios lingüísticos sobre variedades criollas y afrohispanas. Frankfurt: Vervuert. 293–327. Lipski, J.M. 1998c. El español de los braceros chinos y la problemática del lenguaje bozal. Montalbán 31. 101–39. Lipski, J.M. 1999a. Creole-to-creole contacts in the Spanish Caribbean: the genesis of Afro Hispanic language. Publications of the Afro-Latin American Research Association (PALARA) 3. 5–46. Lipski, J.M. 1999b. Chinese-Cuban pidgin Spanish: Implications for the Afro-creole debate. In Rickford, J. and Romaine, S. (eds). Creole genesis, attitudes and discourse. Amsterdam: John Benjamins. 215–33. Lipski, J.M. 1999c. Evolución de los verbos copulativos en el español bozal. In Zimmermann, K. (ed.). Lenguas criollos de base lexical española y portuguesa. Frankfurt: Vervuert. 145–76. Lipski, J.M. 2002a. Contacto de criollos y la génesis del español (afro)caribeño. In Díaz, N., Ludwig, R., and Pfänder, S. (eds). La Romania americana: Procesos lingüísticos en situaciones de contacto. Frankfurt: Vervuert. 53–95. Lipski, J.M. 2002b. Génesis y evolución de la cópula en los criollos afro-ibéricos. In Moñino, Y. and Schwegler, A. (eds). Palenque, Cartagena y Afro-Caribe: Historia y lengua. Tübingen: Niemeyer. 65–101.

Chapter 9. Afro-Bolivian Spanish  Lipski, J.M. 2002c. ‘Partial’ Spanish: Strategies of pidginization and simplification (from Lingua Franca to ‘Gringo Lingo’). In Wiltshire, C. and Camps, J. (eds). Romance phonology and variation. Amsterdam: John Benjamins. 117–43. Lipski, J.M. 2005. A history of Afro-Hispanic language: Five centuries and five continents. Cambridge: Cambridge University Press. Lipski, J.M. 2006a. El dialecto afroyungueño de Bolivia: En busca de las raíces el habla afrohispánica. Revista Internacional de Lingüística Iberoamericana (RILI) 3.2. 137–66. Lipski, J.M. 2006b. Afro-Bolivian Spanish and Helvécia Portuguese: Semi-creole parallels. Papia 16. 96–116. Lipski, J.M. 2007. Afro-Yungueño speech: The long-lost ‘black Spanish’? Spanish in Context. 4.1–43. Lucchesi, D. 1998. A constitução histórica do português brasileiro como um processo bipolarizador: tendências atuais de mudança nas normas culta e popular. In Grosse, S. and Zimmermann, K. (eds). “Substandard” e mudança no português do Brasil. Frankfurt TFM. 73–99. Megenney, W. 1984. Traces of Portuguese in three Caribbean creoles: Evidence in support of the monogenetic theory. Hispanic Linguistics 1. 177–89. Mello, H.R. de, Baxter, A., Holm, J. and Megenney, W. 1998. O português vernacular do Brasil. In Perl, M. and Schwegler, A. (eds). América negra: Panorámica actual de los estudios lingüísticos sobre variedades criollas y afrohispanas. Frankfurt: Vervuert. 71–137. Millones Santagadea, L. 1973. Minorías étnicas en el Perú. Lima: Pontíficia Universidad Católica del Perú. Moodie, S. 1986. El español de Trinidad: Variabilidad y desgaste articulatorio. Anuario de Lingüística Hispánica (Valladolid) 2. 177–96. Moodie, S. n.y. Basilectal survivals in post creole Caribbean Spanish. MS, University of the West Indies, St. Augustine, Trinidad. Moya Corral, J.A. 1979. La pronunciación del español en Jaén. Granada: Universidad de Granada. Naro, A. 1978. A study on the origins of pidginization. Language 45. 314–47. Naro, A. 1998. O uso da concordância verbal no português substandard do Brasil: Atualidade e origens. In Grosse, S. and Zimmermann, K. (eds). “Substandard” e mudança no português do Brasil. Framkfurt: TFM. 139–51. Pizarroso Cuenca, A. 1977. La cultura negra en Bolivia. La Paz: Ediciones ISLA. Portugal Ortiz, M. 1977. La esclavitud negra en las épocas colonial y nacional de Bolivia. La Paz: Instituto Boliviano de Cultura. Salvador Plans, A. 1981. ¿Tres pueblos de habla extremeña en Andalucía? Estudio lingüístico. Anuario de Estudios Filológicos 4. 221–31. Santa Cruz, M. de. 1908. Historias campesinas. Havana: Imprenta y Librería de M. Ricoy. Santa Cruz, N. 1982. La décima en el Perú. Lima: Instituto de Estudios Peruanos. Schriefers, H. and Jescheniak, J. 1999. Representation and processing of grammatical gender in language production: A review. Journal of Psycholinguistic Research 28. 575–600. Simões, M.C.P. 1976. Descrição do sistema verbal de uma criança brasileira com dois anos. In I Encontro Nacional de Lingüística, conferéncias. Rio de Janeiro: Pontifícia Universidade Católica de Rio de Janeiro, Departamento de Letras. 45–55. Simões, M.C.P. and Stoel-Gammon, C. 1979. The acquisition of inflections in Portuguese: A study of the development of person markers on verbs. Journal of Child Language 6. 53–67.

 John M. Lipski Suárez y Romero, A. 1947. Francisco. 2nd ed. Havana: Ministerio de Educación. Tompkins, W. 1981. The musical traditions of the blacks of coastal Peru. PhD thesis, University of California, Los Angeles. Vigliocco, G. and Franck, J. 1999. When sex and syntax go hand in hand: Gender agreement in language production. Journal of Memory and Language 40. 455–78. Ziegler, D.-V. 1976. A preliminary study of Afro-Cuban creole. MA thesis, University of Rochester.

 

Copula patterns in Hawai‘i Creole Creole origin and decreolization Aya Inoue University of Hawai‘i at Mānoa/East-West Center

The present paper demonstrates a multivariate analysis of copula absence in Hawai‘i Creole (HC). Results show that the patterns of use of copula absence in HC parallel those of African American Vernacular English (AAVE) and Caribbean creoles in terms of the constraints on the following grammatical category. Evidence from social constraints shows that similar distributional patterns were observed across age groups. Copula absence has been the center of the debate in terms of the possible links between AAVE and Caribbean creoles. It is discussed that the pattern of copula variability in the following grammatical categories reflects protocreole features and, furthermore, the structure of the substrate African languages (Holm 1984; Holm et al. 1999). Close examination of the creole copula stay in HC suggests that the pattern of copula variability in HC in terms of the following grammatical categories should not be attributed principally to the creole copula or to the copula-like structures in the substrate language Portuguese.

.

Introduction*

Copula absence has been extensively investigated in the literature of Englishbased creoles (hereafter, EBCs) and African American Vernacular English (AAVE)

* An earlier version of this paper was presented at the Society for Pidgin and Creole Linguistics Annual Meeting, Albquerque, NM in January 2006. I would like to express my gratitude to Jeff Siegel and Andrew Wong for their advising on this paper. Jeff Siegel’s project entitled “External Influences and Internal Variation in Current Hawai‘i Creole” was funded by NSF grant BCS 03-45959. The support of the National Science Foundation is gratefully acknowledged. I would also like to thank East-West Center for supporting my graduate study. Several others, including the editors of this volume, and two anonymous readers helped improve this paper with their insightful and encouraging comments, but all remaining errors are of course my own.

 Aya Inoue

(Rickford 1991, 1998). It has been the center of the debate over the putative creole origin of AAVE, and often discussed as an African influence on the process of creolization of Caribbean creoles (Winford 1993: 155). As Walker and Meyerhoff (2004) pointed out, however, a handful of EBCs (such as Barbadian Creole, Jamaican Creole, Trinidadian Creole, Guyanese Creole) have received most of the attention, and there is a need for more variationist analyses of copula absence in other EBCs. In this paper, I demonstrate a multivariate analysis of copula absence in Hawai‘i Creole (HC),1 a nonAtlantic English-based creole spoken in the islands of Hawai‘i. There have not been any variationist multivariate analyses using the VARBRUL program2 on copula absence in HC so far. The results imply a pan-creole tendency as well as the issue of whether there is a current decreolization process in HC or not. Amongst the research on copula absence in EBCs and AAVE, the hierarchical ordering of the following grammatical environment has been reported as most significant. Rickford and Blake (1990) argue that if a prior creole origin were to leave its vestiges in a decreolizing or decreolized variety, we would expect to find the following patterns of copula absence in which the environment before _gonna favors copula absence the most and the environment before _NP does so the least: _gonna > _V+ing > _Adj > _Loc > _NP

In the rest of the present paper, the term ‘creole pattern’ is used to refer to the above mentioned hierarchical pattern regarding the following grammatical categories. It is argued by Holm (1984) that this pattern reflects protocreole features and, furthermore, the structure of the substrate African languages. The question arises whether this pattern is also observed in other non-Atlantic EBCs which do not have African languages as substrates. HC is geographically remote, and has different substrate languages from Atlantic EBCs. Its substrate languages are Hawaiian, Portuguese, Cantonese, Japanese and Visayan as well as other Asia-Pacific languages. However, although HC substrates do not involve any African languages, the possible influence from Atlantic creole features to HC grammatical

. The creole in Hawai‘i has been referred to as “Hawai‘i Creole English (HCE)” in many of the previous examples of literature. However, following the recent studies by Sakoda and Siegel (Sakoda and Siegel 2004a, b), I will refer to it as Hawai‘i Creole (HC) in the present study based on two reasons: First, calling Hawai‘i Creole English is misleading in that it suggests Hawai‘i Creole to be a dialect of English. Second, Hawai‘i Creole is a more consistent reference considering how other creoles around the world (Jamaican Creole, Guyanese Creole, Bahamian Creole just to name a few) are referred to in linguistic literatures. . VARBRUL is a probabilistic-based multivariate regression analysis that indicates the relative influence of various linguistic or social factors on a linguistic process. It produces factor weights between 0 and 1 where a weighting greater than .5 indicates a favoring effect while a factor weight below .5 indicates an inhibiting effect (Young and Bayley 1996).

Chapter 10. Copula patterns in Hawai‘i Creole 

structure was pointed out by Holm (1985). Holm (1985) argued the possibility that features in HC have their origin in the diffusion of Atlantic creole features through jargonized versions of Atlantic creole spoken by sailors and traders. This raises another question whether the distributional pattern of copula absence in current HC has any implications in terms of the Atlantic creole diffusion to HC grammar. Previously, copula absence in HC was quantitatively investigated with distributional analyses some thirty years ago (Day 1973; Perlman 1973). Day (1973) observed an internal implicational pattern within the use of null copula in four environments. Copula absence occurs most frequently before the V+ing construction, followed by the environment before adjectives, locatives, and noun phrases. _V+ing > _Adj > _Loc > _NP Perlman (1973) also found a similar hierarchy of the variability of copula in five environments. Perlman’s (1973) study quantified both stay copula and zero copula. _gon > _V+ing > _Adj > _Loc > _NP If HC shares the creole pattern with other EBCs, the argument that claims that copula absence in EBCs reflects influence from African languages might require alternative explanations.

.

Data and methodology

The present study analyzed sociolinguistic interviews with rural O‘ahu residents collected through the project External Influences and Internal Variation in Current Hawai‘i Creole. This project was conducted by Jeff Siegel in 2001–2006. HC speech from people who were born and raised in the rural areas of the island of O‘ahu was analyzed in the present study. Compared with the outer islands, the island of O‘ahu is reportedly where the most “decreolized” and therefore the most acrolectal varieties are found (Romaine 1994). Data from the twenty speakers stratified by age and gender was quantified and analyzed using GoldVarb 2001 (Robinson, Tagliamonte and Lawrence 2001). In the present paper the term copula is used to cover any form (or zero marked with the symbol Ø) corresponding to a form of Standard English be. The following three forms were quantified as variants of copula absence: Full form copula (am/is/are), contracted form (-’m/-’s/-’re), and zero copula (Ø). I looked at both copula environment in a narrow sense that occurs before NP (as in “My sista Ø one bus driver”), Adjective (“My sista Ø skinny”), and Locative (“Kent them Ø inside da house”), and auxiliary be that occurs before V+ing (as in “He Ø helping me”), and gon(gonna, goin)+V (“She Ø goin miss da prom”). As for the computation, the present study employs “straight deletion” methodology in which tokens of zero are counted out of tokens of zero, contraction, and full forms combined.

 Aya Inoue

The two age groups (Post-60’s vs. Pre-60’s) represent the speakers who grew up before and after the American statehood in 1959. Statehood brought social impacts relevant to the linguistic environment in Hawai‘i, such as rapid development and a huge influx of new people. Gender of the speakers (Male vs. Female) was also considered as possible source of copula variability. Tokens were coded for the five linguistic factors as shown in the Table 1: following grammatical category, grammatical person and type of subject, and preceding as well as following phonological context.

Table 1. Linguistic factors for variation of copula absence Grammatical categories

Linguistic environments

1. Following grammatical category

_gonna (gon, goin) _V+ing _Adjective _Locative _NP Personal pronoun Other pronoun (e.g., there) NP 1st  (am) 3rd  (is) Other (are) _Consonant _Vowel Consonant_ Vowel_

2. Subject

3. Grammatical person and number of subject

4. Following phonological segment 5. Preceding phonological segment

The contexts described below were excluded from the analysis. First, nonfinite (negation) and past tense contexts as in “He Ø not coming” and “He Ø not there yesterday” were excluded. Past tense copula has a different pattern from present tense copula in HC. For example, grammatical person and number are often neutralized as in “She was there” and “We was there”. Secondly, neutralization contexts were excluded. This is because tokens which are followed by a sibilant, as in “He’s sick”, which, in rapid speech, are phonetically difficult to distinguish from deletions, as in “He Ø sick”. Thirdly, reported Standard English speech was excluded. Clause final context as in “You know what that is?” was also excluded because it is not clear whether HC copula absence is possible in such contexts. HC has a creole copula stay. This stay copula was also excluded. It is not semantically/functionally equivalent to other English-like copula because it has a different co-occurrence constraint. For example, although the stay copula in HC may be used before adjective and locative phrases, it is possible only when they denote “a nonpermanent or nonintrinsic quality, or a change

Chapter 10. Copula patterns in Hawai‘i Creole 

in conditions”3 (Sakoda and Siegel 2003: 77). In the present study, the stay copula was counted and analyzed separately. There were fourteen examples of the stay copula in the data. In terms of the following grammatical environment, two tokens occurred before V+ing, seven tokens occurred before the adjective, and five tokens occurred before the locative environment. When examining the distribution of the stay copula in terms of the social background of the speakers, stay tokens were observed more with the older Pre 60’s speaker group.

.

Copula variability in Hawai‘i Creole

The result of the distributional analysis of copula absence concerning age and gender groups suggests that the two social factors do not seem to play the crucial roles in the occurrence of copula absence in HC. Tables 2 and Table 3 show the distributional analyses of copula absence with the age and gender groups. Table 2 compares the two age groups. The distribution of the three variants across age groups does not vary to a large extent. Table 3, below, shows the distribution across gender groups. The pattern of copula absence is also similar across the two gender groups. Table 2. Distributional analysis of copula absence with age groups Full

Contracted

Deleted

Total

Pre-60’s Post-60’s

26 (13%) 43 (19%)

69 (36%) 61 (27%)

92 (49%) 114 (52%)

187 (46%) 218 (53%)

Total

69 (17%)

130 (32%)

206 (50%)

405

Table 3. Distributional analysis of copula absence with gender groups Full

Contracted

Deleted

Total

Male Female

36 (17%) 33 (16%)

62 (30%) 68 (33%)

105 (51%) 101 (50%)

203 (50%) 202 (49%)

Total

69 (17%)

130 (32%)

206 (50%)

405

Table 4 below demonstrates a distributional analysis with a linguistic factor – the effect of the following grammatical category. It shows more applications of copula absence for the environments of auxiliary be, namely before gonna, gon, goin, and V+ing, when compared with the environments of the copula in narrow sense, namely, before NP, . For example, the sentences “He stay free aswhy” and “She stay sick” are acceptable because stay is used before adjectives that denote nonpermanent quality. The sentence “Da wahine stay short”, on the other hand, is not acceptable because the adjective denotes a permanent quality (Sakoda and Siegel 2003: 77–8).

 Aya Inoue

Adjective, and Locative. The relative ordering of the frequency matches the patterns expected by the creole origin hypothesis argued in Rickford and Blake (1990). Table 4. Distributional analysis with following grammatical category Grammatical categories Following grammatical category

Linguistic environments

Full

Contracted

_gonna (gon, goin) _V+ing _Adjective _Locative _NP Miscellaneous

1 (1%) 4 (5%) 19 (18%) 1 (4%) 30 (34%) 14 (26%)

9 (13%) 12 (16%) 32 (31%) 14 (56%) 37 (43%) 26 (50%)

Total

69 (17%)

130 (32%)

Deleted

Total

58 (85%) 68 (16%) 55 (77%) 71 (17%) 52 (50%) 103 (25%) 10 (40%) 25 (6%) 19 (22%) 86 (21%) 12 (23%) 52 (12%) 206 (50%) 405

In order to see whether older speakers and younger speakers show different patterns, it is necessary to confirm if this pattern is observed in both age groups. Table 5 below demonstrates distributional analysis of the effect of the following grammatical environment for the two age groups. The pre 60’s and the post 60’s groups show similar frequency. Table 5. Copula absence for Pre 60’s and Post 60’s Pre 60’s Following grammatical environment _gonna (gon, goin) _V+ing _Adjectiv _Locative _NP Miscellaneous

Post 60’s

%

N (Total 92/187)

%

N (Total 114/218)

79% 68% 51% 46% 15% 29%

27/34 20/29 27/52 7/15 6/40 5/17

91% 83% 49% 30% 28% 28%

31/34 35/42 25/51 3/10 13/28 7/35

The similar patterns in the two age groups in terms of the effect of the following grammatical environment are demonstrated graphically in the Figure 1 below. This creole pattern was also reported in the two previous study of HC in 1970’s. The distributional analyses for the following grammatical category in the two previous studies are compared with the present study. All three studies have the same hierarchical order. 1.

Day (1973:111)with 23 male/female speakers from the island of O‘ahu, Hawai‘i, and Maui. _V+ing(94%)>_Adj(72%>_Loc(62%)>_NP(63%)

Chapter 10. Copula patterns in Hawai‘i Creole  100

Frequency

80 60

Pre 60's Post 60's

40 20 0 _ NP

_ V-ing _ gonna _ Adj _Loc Following grammatical environment

Figure 1. Copula absence in two age groups in Hawai‘i Creole

2.

3.

Perlman (1973: 120) based on his research carried out in the island of O‘ahu with 29 male speakers (working-class informants, ages 36–75). _gon (89%) > _V+ing (37%) > _Adj (33%) > _Loc (25%) > _NP (20%) *Stay copula occurred only in _Loc environment (37% of the tokens in locative environment) The present study with 20 male/female speakers from the rural areas of the island of O‘ahu. _gon (86%) > _V+ing (74%) > _Adj (52%) > _Loc (38%) > _NP (18%)

Frequency

Figure 2 below compares the patterns in the three studies graphically. In terms of the frequency of copula absence the distributional pattern in the present study is located 100 90 80 70 60 50 40 30 20 10 0

Day (1973) Perlman (1973) The present study

_ NP

_Loc _ Adj _ V-ing _ gonna Following grammatical environment

Figure 2. Copula absence in three studies on Hawai‘i Creole; Day (1973), Perlman (1973) and the present study

 Aya Inoue

between the patterns of the two previous studies. All three studies have the same hierarchical order. The one difference is that there is not any data for before the _gonna environment for Day’s (1973) study. Are there any signs suggesting that decreolization processes are currently in progress in HC? Neither (1) the distributional pattern in the Pre 60’s and Post 60’s, nor (2) the comparison of Day (1973), Perlman (1973) and the present study suggest evidence to support a decreolization process in terms of copula absence within the past thirty years. The results of multivariate analysis show that three out of five linguistic factors turned out to play a significant role in the application of copula absence in HC: following grammatical categories discussed above, types of subjects, and grammatical person of subject. As already mentioned, the two social factors, age and gender were not significant. Table 6 shows Variable Rule analysis of the factors contributing to the application of copula absence in HC regarding the following grammatical categories. Table 6. Constraints on Copula Absence (Deleted variant) in terms of the following grammatical categories as analyzed by Variable Rule (Varbrul) Program Factors Significance = 0.000 _gonna (gon, goin) _V+ing _Adjective _Locative _NP Miscellaneous

Factor weights Input = 0.514

%

N Total N = 405

.81 .65 .49 .40 .26 .34

85% 77% 50% 40% 22% 23%

58/68 55/71 52/103 10/25 19/86 12/52

Table 7 below shows variable rule analysis of the type of subject and the grammatical person of subject. Table 7. Constraints on Copula Absence (Deleted variant) in terms of type of subject and grammatical person of subject as analyzed by Variable Rule (Varbrul) Program Factors Significance = 0.000 Type of subject Personal pronoun Other pronoun NP Grammatical person/number of subject 1st  (am) 3rd  (is) Other (are)

Factor weights Input = 0.514

%

N Total N = 405

.69 .15 .51

73% 10% 45%

165/226 12/115 29/64

.36 .43 .81

66% 33% 92%

52/78 85/252 69/75

Chapter 10. Copula patterns in Hawai‘i Creole 

So far, we looked at the effect of the following grammatical environment. As for the effect of the preceding grammatical environment, namely, the type of subject, fewer studies of creoles report data on this environment than for following grammatical environment (Rickford 1998: 183). In the present study, personal pronoun subject favors copula absence more than NP does. This [Personal pronoun > NP] pattern for favored environment for copula absence is also observed in Bahamian English (Reaser 2004) and AAVE (Baugh 1980). However, in some other EBCs such as Barbadian Creole (Rickford and Blake 1990), Jamaican Creole (Rickford 1996) and Liberian Settler English (Singler 1991), the reversed order [NP > personal pronoun] is observed. More study is required to explain the pattern of the preceding grammatical environment.

. Discussion – Creole copula in Atlantic vs. non-Atlantic creole While English is considered as one of the languages “in which zero copula encoding is never allowed” (Stassen 2005: 486),4 copula absence is typologically common, hence does not only occur in EBCs. In terms of the geographical distribution, Stassen (2005) recognized a number of areas where zero copula encoding is highly conspicuous or prominent: Pacific region, South America, as well as in the northern part of Africa. As for the major substrate languages of HC, many of them (Portuguese, Cantonese, Japanese, and the languages in the Philippines) belong to the areas in which a zero copula constitutes a minor or marginal option; Indo-European, China, Japan, and the Philippines (ibid.: 487). Creole languages (especially in their most conservative variety or basilect) often have their own copula forms (referred to as the creole copula in the present paper). For instance, Guyanese Creole has a/de, Sranan has na/de, and Tok Pisin has stap just to mention a few. In HC, the form of the creole copula is stay as mentioned in Section 2. The creole pattern of copula absence has been repeatedly observed across many EBCs. There have been discussions on the source of the creole pattern: _gonna/goin+V as most favorable, _Adj fairly high, and _NP and _Loc least favorable environment for copula absence. Holm (1984) and Holm et al. (1999) discussed three possible sources for this creole pattern in Atlantic creoles: (1) protocreole system5 (2) creole copula and (3) copula-like structures in substrate languages. According to Holm’s (1984: 298) explanation, the hierarchy observed in the creole pattern presumably reflects the structure of the protocreole grammar. Gonna is a calque for a protocreole preverbal marker

.

Note that Stassen (2005) limits its scope to stative predicate nominal sentences only.

. My understanding of the term ‘protocreole’ discussed in Holm (1984) and Holm et al. (1999) is that it refers to “linguistic factors common to the origin of [Jamaican Creole and Gullah], namely the language situation in West Africa and the English-speaking plantations of the New World” (Holm 1984: 292).

 Aya Inoue

indicating irrealis, V+ing is a calque for a protocreole verb and its preverbal progressive/ iterative marker, Adj is a subclass of verb in the protocreole. Loc and NP require copulas because of protocreole locative and equative copula, respectively. Especially, protocreole equative copula was obligatory before NP and hence copulas preceding NP are least likely to delete. As Walker (2000: 37) puts it, Holm’s (1984) explanation “implicitly predicts higher rate of absence before gonna, V+ing and Adj and lower rates before Loc and NP.” Holm et al. (1999) reports comparative creole copula patterns based on a database that includes not only Atlantic creoles but also five non-Atlantic creoles. It focuses on the relationship between the form of the copula and the three following syntactic environments represented in copula hierarchy: before NP, before locatives, and before adjectives or adjectival verbs. They summarized their observation as follows: Among the Atlantic creoles, with few exceptions, an expressed copula is required before NPs; a copula of a different form occurs before locative expressions, but this can often be deleted; no copula usually occurs before adjectives (ibid.: 98, 114).6

Furthermore, Holm et al. (1999) states that the pattern of the creole copula reflects the common copula pattern found in the Niger-Congo languages forming the substrate for EBCs. In his study of Bahamian copula absence, Reaser (2004: 5) pointed out that the determination of the kinds of copula forms realized in a language’s grammar is a fundamental methodological priority, and thus, an important first step in the detailed analysis of the copula system is a qualitative description of the overall copular system. It is necessary to have a detailed description of the creole copula in HC (stay copula) as well as the copula pattern in the substrate languages of HC in order to examine the validity of Holm’s argument. Distribution of HC creole copula stay has not been fully described due to the variability of HC. Some of the following grammatical categories have been individually discussed in the description of stay in the literatures on HC grammar (Cheng 1969; Day 1973; Perlman 1973; Sakoda and Siegel 2004b). Based on the study with two informants from the Island of O‘ahu, Cheng (1969) summarized the optionality of stay copula as absent before NP, expressed before Loc, optionally expressed before Adj. More specifically, in _NP environment, copula is generally absent as in “They students” “I one student” (ibid.: 105–6). Stay is optional before _Adj as in “She real pretty” and “I stay hungry” (ibid.: 105). This optionality was later explained as “present if it

. Their observations include that of an equative copula, which was excluded from citation due to the lack of the relevance to the present discussion, which I note here. “. . . a highlighter of the same form as the equative copula often occurs before fronted constituents” (Holm et al. 1999: 114).

Chapter 10. Copula patterns in Hawai‘i Creole 

denotes nonpermanent/nonintrinsic quality” in Sakoda and Siegel (ibid.) In _Loc environment, the copula verbs in standard English are usually replaced by stay as in “My parents stay home” (ibid.: 106). Based on the data from 10 speakers from different areas on O‘ahu, Day’s (1973: 21) preliminary analysis described ste (stay) as a copula used before locatives. Perlman’s (1973: 120) study with 29 working-class informants observed 35 tokens of stay copula (out of 1104 tokens) all of which occurred in locative environment. Previous literature on the HC creole copula stay suggests that it occurs most extensively before Loc environment although it is also observed before NP and before Adj environments to some extent. Sakoda and Siegel (ibid.) stated that stay was “not used when the location is ‘here’ or ‘there’ or in ‘where question’”. Cheng (1969) and Sakoda and Siegel (ibid.) described the occurrence of stay in _NP environment as “in general, no copulative verb is used” (Cheng 1969: 105), and “usually absent” (Siegel and Sakoda 2004b). According to Sakoda and Siegel (2004b), stay before V+ing is “frequently left out nowadays”. As mentioned in Section 1, the substrate languages of HC involve variety of languages such as Hawaiian, Portuguese, Cantonese, Japanese, and some Philippine languages. Regarding these substrate languages, Siegel (2000) pointed out the dominance of two ethnic groups, the Chinese and Portuguese at the time of the emergence of HC. According to Siegel (2003: 197), the functions of the HC copula stay are similar to the copula-like structure estar of the substrate language Portuguese rather than to the nonpunctual aspect markers in other creoles. Siegel (2003: 197) argued that the functions of HC stay (both copula and progressive aspect marker) are overlapping with those of Portuguese estar. Table 8. Functions of HC stay compared to Portuguese estar and nonpunctual markers in other creoles [adapted from Siegel 2003: 197, Table 2.] Functions Progressive marker Habitual marker Inchoative with statives Copula with adjectives Copula with locatives Perfective marker

HC stay

Portuguese estar

Other creoles nonpunctual aspect marker

+ − − + + +

+ − − + + +

+ + + − − −

The overlap between functions of HC stay and Portuguese estar suggests that the substrate Portuguese is a good explanation of the pattern in the creole copula stay. Our next question is whether the pattern of current HC copula absence reflects that of both the creole copula stay and the copula-like structures of the substrate language Portuguese. Holm et al. (1999: 99) discussed that the difference of the pattern of creole copula between Atlantic creoles and non-Atlantic creoles is that creole copulas are present in _NP environment in all the Atlantic creoles (13 out of 13 creoles)

 Aya Inoue

he examined, but that they are absent in the majority of the non-Atlantic creoles (4 out of 5 creoles) he examined. In terms of the _NP category, HC shows the pattern of the majority of non-Atlantic creoles. Holm et al. (1999) pointed out that the patterns of copula absence in EBCs are similar to the pattern of the creole copula at least in term of the relative ordering of _Loc, _Adj, and _NP environments. It might suggest that the low rate of copula absence in the _NP environment is explained by the pattern of creole copula that is present in the _NP environment. However, the pattern of copula absence in HC shows the low rate of absence in _NP environment. The low rate of absence before _NP does not match the fact that the pattern of its creole copula stay is described as “the copulative verb in SAE [Standard American English] is usually replaced by stay” in _Loc environment, but “in general, no copulative verb is used” in _NP environment (Cheng 1969: 105–7). In HC, relative absence of copula in current HC speech does not reflect the relative absence of creole copula stay at least in terms of _NP environment. In other words, although the creole copula in HC is “generally absent” with NP – which should lead to a high rate of copula absence if the pattern in the creole copula is to be attributed as the source of the creole pattern – the rate of absence before NP is low as in the cases of creoles that have an expressed creole copula before NP. The present study suggests that the pattern in creole copula is not the principal source of the creole pattern. This raises another question whether the distributional pattern of copula absence in current HC has any implications for the Holm’s (1985) theory that HC grammatical structures are influenced by Atlantic creoles through diffusion. The use of copula is one of the HC grammatical features that Bickerton (1981) discussed the parallels with Atlantic creoles. As we have seen, although the syntactic distribution and function of the HC creole copula stay overlap with those of the substrate language Portuguese, they are not reflected in the distribution of copula absence in current HC speech. This leaves room for other sources including the diffusion of Atlantic creoles to be responsible for the observed creole pattern in current HC speech. Only one of the substrate languages was examined in the present study. More research on the distributional description of the substrate languages of HC might shed light on this issue.

. Summary and conclusion This study provided a multivariate analysis of copula absence in Hawai‘i Creole (HC), an EBC spoken in the islands of Hawai‘i. Variation in copula absence in HC in the present study does not provide evidence to support the hypothesis of decreolization within the past 30 years in terms of the HC speakers from rural areas of the O‘ahu island. Constraints on the following grammatical category on copula absence in HC are like other EBCs. Creole patterns are often argued to reflect creole copula pattern and the substrate patterns observed in African languages (Holm 1984). This study provided a non-Atlantic case study that replicated the creole patterns in hierarchical

Chapter 10. Copula patterns in Hawai‘i Creole 

ordering involving the following grammatical category. It suggests that in the case of HC, the pattern should not be attributed principally to the creole copula. This is because the pattern of the creole copula and substrate language do not match the observed creole pattern in current HC speech especially in terms of the rate of the copula absence in _NP environment; different from what we might have expected from the creole copula and substrate which leads to a high rate of absence, the observed pattern has a low rate of absence in _NP.

References Baugh, J. 1980. A reexamination of the Black English copula. In Labov, W. (ed.). Locating language in time and space. New York: Academic Press. 83–106. Bickerton, D. 1981. Roots of language. Ann Arbor: Karoma. Cheng, T. 1969. Preliminary description of Hawaiian Pidgin sentence-types with an introduction to its phonology. Report No. 1, In Peterson, R. (Principal investigator), Chuck, H. (Coordinator) and Coladarci, A. (Research methodologist). Appendix A to Final report, teaching Standard English as a second dialect to primary school children in Hilo, Hawaii, Project No. 5-0692, Volume I. Hilo, HI: State of Hawaii Department of Education. Day, R.R. 1973. Patterns of variation in copula and tense in the Hawaiian post-creole continuum. Working Papers in Linguistics 5.2. Department of Linguistics, University of Hawai‘i. [Published version of Day’s 1972 University of Hawaii dissertation.] Holm, J., Adamson, L., Bhattacharjya, D., Chapuis, D., DeGraff, M., de Kleine, C., Faraclas, N., Green, K., Lorenzino, G., Mello, H., Michel, A., Owens, J., Khamis, C., Patrick, P., Santoro, S., Simon, R., Suzuki, M., and Yillah, S. 1999. Copula patterns in Atlantic and non-Atlantic Creoles”. In Rickford, J.R. and Romaine, S. (eds). Creole genesis, attitudes and discourse: Studies celebrating Charlene Sato. Amsterdam: John Benjamins. 97–119. Holm, J. 1985. Substrate diffusion. In Muysken, P. and Smith, N. (eds). Substrata versus universals in creole genesis. Amsterdam: John Benjamins. 259–78. Holm, J. 1984. Variability of the copula in Black English and its creole kin. American Speech 59.4. 291–309. Perlman, A.M. 1973. Grammatical structure and style shift in Hawaiian Pidgin and Creole. PhD thesis, University of Chicago. Reaser, J. 2004. A quantitative sociolinguistic analysis of Bahamian copula absence: Morphosyntactic evidence from Abaco Island, The Bahamas. Journal of Pidgin and Creole Languages 19.1. 1–40. Rickford, J.R. 1998. The creole origins of African-American vernacular English: Evidence from copula absence. In Mufwene, S.S., Rickford, J.R., Bailey, G. and Baugh, J. (eds). AfricanAmerican English: Structure, history and use. London: Routledge. 154–200. Rickford, J.R. 1996. Copula variability in Jamaican Creole and African American Vernacular English: A reanalysis of DeCamp’s text. In Guy, G., Feagin, C., Schiffrin, D. and Baugh, J. (eds). Toward a social science of language. Vol. 1: Variation and change in language and society. Amsterdam: John Benjamins. 357–72. Rickford, J.R. 1991. Rappin on the copula coffin: Theoretical and methodological issues in the analysis of copula variation in African-American Vernacular English. Language Variation and Change 3. 103–32.

 Aya Inoue Rickford, J.R., and Blake, R. 1990. Copula contraction and absence in Barbadian English, Samaná English and Vernacular Black English. In Hall, K., Koenig, J.P., Meacham, M., Reinman, S., and Sutton, L.A. (eds). Proceedings of the sixteenth annual meeting of the Berkeley Linguistics Society, February 16–19. 1990. Berkeley, California: Berkeley Linguistic Society. 257–68. Robinson, J.S., Tagliamonte, S. and Lawrence, H. 2001. GoldVarb 2001: A multivariate analysis application for Windows. University of York. URL: http://www.york.ac.uk/depts/lang/ webstuff/goldvarb/ Romaine, S. 1994. Hawai‘i Creole English as a literary language. Language in Society 23. 527–54. Sakoda, K. and Siegel, J. 2004a. Hawai‘i Creole: Phonology. In Kortmann, B. and Schneider, E.W. (eds) with Burridge, K., Mesthrie, R. and Upton, C. A handbook of varieties of English, Volume I: Phonology. Berlin: Mouton de Gruyter. 729–49. Sakoda, K. and Siegel, J. 2004b. Hawai‘i Creole: Morphology and syntax. In Kortmann, B. and Schneider, E.W. (eds) with Burridge, K., Mesthrie, R. and Upton, C. A handbook of varieties of English, Volume II: Morphology and syntax. Berlin: Mouton de Gruyter, 742–69. Sakoda, K. and Siegel, J. 2003. Pidgin grammar: An introduction to the creole language of Hawai‘i. Honolulu: Bess Press. Siegel, J. 2003. Substrate influence in creoles and the role of transfer in second language acquisition. Studies in Second Language Acquisition 25.2. 185–209. Siegel, J. 2000. Substrate influence in Hawai‘i Creole English. Language in Society 29. 197–236. Singler, J. 1991. Copula variation in Liberian Settler English and American Black English. In Edwards, W. and Winford, D. (eds). Verb phrase patterns in Black English and creoles. Detroit: Wayne State University Press. 129–64. Stassen, L. 2005. Zero copula for predicate nominals. In Haspelmath, M., Dryer, M.S., Gil, D. and Comrie, B. (eds). The world atlas of language structures. New York: Oxford University Press. 486–7. Walker, J.A. 2000. Rephrasing the copula: Contraction and zero in early African American English. In Poplack, S. (ed.). The English history of African American English. Oxford: Blackwell. 35–72. Walker, J. and Meyerhoff, M. 2004. Zero copula in the Caribbean: Evidence from Bequia. Paper presented at New Ways of Analyzing Variation 33. University of Michigan, Ann Arbor, MI. Winford, D. 1993. Predictions in Caribbean English creoles. Amsterdam: John Benjamins. Winford, D. 1990. Copula variability, accountability, and the concept of “polylectal” grammars. Journal of Pidgin and Creole Languages 5.2. 223–52. Young, R., and Bayley, R. 1996. VARBRUL analysis for second language acquisition research. In Bayley, R. and Preston, D.R. (eds). Second language acquisition and linguistic variation. Amsterdam: John Benjamins. 235–306.

 

 

On the properties of Papiamentu pa Synchronic and diachronic perspectives* Claire Lefebvre Université du Québec à Montréal and MPI-EVA

Isabelle Therrien Université du Québec à Montréal

This chapter discusses the properties and the historical derivation of the lexical item pa in Papiamentu. First, it documents the fact that pa may fulfil several functions (preposition, mood marker, complementiser, and possibly case marker), and it provides an account of this multifunctionality. It is proposed that pa may head several syntactic projections. Second, the paper traces the origin of the properties of pa. It is shown that, if the form pa is derived from Portuguese para, its semantic and syntactic properties are derived from those of corresponding lexical entries in the West African substratum languages of P. Third, the paper provides an account of this division of properties between the source languages within the framework of the relexification account of creole genesis. It is shown that, in this case, two substrate lexical entries, one fulfilling the functions of preposition, complementiser and case marker, and one that of mood marker and complementiser, have been relexified on the basis of a single superstrate form para, yielding the creole lexical entry pa, cumulating the functions of the two substrate entries. The match between the creole entry and the substrate corresponding lexical entries is not perfect. The mismatches are considered in detail, showing that the make up of pa has also involved reorganisation of the distribution of labour between related lexical entries, through reanalysis and grammaticalisation.

.

Introduction

This paper will discuss the properties and the historical derivation of the multifunctional lexical item pa in Papiamentu, henceforth P. P is a creole spoken mainly in * This paper has been written as part of the project entitled ‘Constraints on Relexification’ financed by a SSHRC grant to Claire Lefebvre. We would like to thank Lucille Berry for providing us with original Papiamentu data. Thanks to the anonymous reviewers, and to the editors of this volume, Magnus Huber and Viveka Velupillai, for their comments and questions on an earlier version of this paper.

 Claire Lefebvre and Isabelle Therrien

Aruba, Bonaire and Curaçao. Its lexifiers or superstratum are romance languages, Portuguese and Spanish, (60%), and Dutch, (30%) (Maurer 1986).1 Parkvall (2000) identifies its substratum languages as those that were spoken on the Slave Coast in the 17th/18th century, among which are the Gbe languages. This is congruent with Postma (1990) who shows that the majority of the slaves (56%) who were sent to Curaçao prior to 1700 were from the Slave Coast. The form pa fulfils several functions: preposition, mood marker, complementiser, and possibly case marker.2 The first objective of the paper is to provide a description of the properties of pa for each of its functions (Section 1). This will be done mainly on the basis of published sources (see below). Following the analysis in Lefebvre (1998a) and in Lefebvre and Massam (1988), we will assume that multifunctional lexical items may head more than one syntactic projection. It will be shown that pa may head PP, MoodP, CP and possibly K(=case)P. The second objective of the paper is to trace the origin of the properties of pa. To our knowledge, this topic has not been addressed in the literature on P. We will take up the problem on the basis of a three-way comparison of the properties of pa with those of corresponding lexical items in the contributing languages of P (Section 2). On the one hand, it will be shown that, if the form of the P lexical item pa is assumed to be derived from Portuguese para (see e.g., Birmingham 1970: 104), very few of its semantic and syntactic properties are derived from those of this lexical item. On the other hand, it will be shown that several of its semantic and syntactic properties are derived from those of corresponding lexical entries in the substratum languages of P. For the purpose of our comparison, we will use Portuguese as the superstrate language, and Fongbe, a West African language of the Gbe cluster spoken in Benin, as the substrate language. This methodology is justified in Section 2. The results of our comparison will lead us to considering a scenario of the origin of the properties of P pa that involves the process of relexification, as defined in Lefebvre (1998b

. The authors agree that, while the forms of the lexical entries of modern Papiamentu are predominantly derived from Spanish, they were first predominantly derived from Portuguese (see e.g., Goodman 1987; Kouwenberg and Murray 1994: 5; Kouwenberg and Muysken 1995: 205; Navarro Tomás 1953; VanWijk 1958; etc.). Historical evidence supports this claim. The Portuguese were first in establishing colonies in the area. The Spaniards came later. Several authors propose that relexification is the process by which the hypothesised Portuguese variety was turned into a Spanish one (see e.g., McWhorter 1995; Voorhoeve 1964). According to Lenz (1928), Dutch, as well as English constitute later additions to the lexicon of Pp. . Pa is treated as a single etymon because, as will be seen throughout this paper, pa has a core meaning across its various functions. With other authors (e.g., Bouchard 1995; Cowper 1989, 1995; Gomeshi and Massam 1994; Johns 1992; Nida 1948; Ruhl 1989; etc.), we assume that monosemy is to be preferred over polysemy wherever possible (see Lefebvre 1999 for an extensive discussion of this issue).

Chapter 11. On the properties of Papiamentu pa 

and the references therein) (Section 3). It will be shown that, in this case, however, two substratum lexical entries (for Fongbe: nú, preposition, complementiser and case marker, and ní, mood marker and complementiser) appear to have been relexified on the basis of a single superstratum form para, yielding the creole lexical entry pa, cumulating the functions of the two substrate entries. It will also be shown that the match between the creole entry and the substratum corresponding lexical entries is not perfect. The mismatches between the lexical entries will be considered in detail in Section 4, showing that the make up of the lexical entry pa has also involved reorganisation of the distribution of labour between related lexical entries, through reanalysis and grammaticalisation. Section 5 concludes the paper. The bulk of the P data is drawn from written sources: Andersen (1990); Kouwenberg (1990); Kouwenberg and Michel (fc); Kouwenberg and Murray (1994); Kouwenberg and Muysken (1995); Maurer (1988, 1993); Muysken (1977). Data have also been collected by Isabelle Therrien from Lucille Berry-Haseth, a native speaker of P currently living in Curaçao. Finally, data collected by Silvia Kouwenberg are cited from Kouwenberg and Lefebvre (in press). The Portuguese data are from written sources. The Fongbe data are likewise mainly drawn from written sources with a few additions from Claire Lefebvre’s field notes. Before getting into the heart of the topic, a few remarks are in order. First, the examples are given in the orthography used by each authors.3 Second, the glosses used by the various authors for a given lexical item are not always uniform. In the examples cited in this paper, we have made the glosses uniform. Some authors do not provide glosses. In these cases, we have added glosses.

.

The various functions of pa

The lexical item pa fulfils the functions of preposition, mood marker, complementiser and possibly of case marker. This section documents the properties of pa as it occurs in its various functions, and it takes up the issue of the multifunctionality of this lexical item.

.

Pa occurring as a preposition

As a preposition, pa may select either NPs or clauses. The properties of the complements of pa will each be discussed in turn. Given the fact that, in the following sections, the P data will be compared with corresponding ones in the source languages, some distinctions are being made here that otherwise need not be.

.

For discussions on Pp orthographic conventions, see Kouwenberg and Murray (1994).

 Claire Lefebvre and Isabelle Therrien

..

Pa may select NPs

First, the NP complement of pa may be interpreted as a benefactive. This is exemplified in (1). (1)

Lo mi larga e buki pa bo den e kas (P)  1 leave  book  2 in  house ‘I will leave the book for you in the house.’ (Kouwenberg and Muysken 1995: 210)

Second, the NP complement of pa may be a particular type of benefactive, best rendered as ‘one’s own sake, one’s own good’ as in (2). (2)

Ora ku e trabou ta kla, bo por bai pa bo mes bon. (P) Hour   work be finished 2 can go  2 self good ‘When the work is finished, you can go for your own good.’ (Isabelle Therrien’s field notes)

Third, with the verb meaning ‘to pay’, the phrase introduced by pa is interpreted as ‘in exchange of ’, illustrated in (3). (3)

Mi a paga veinte guilder pa e bistí aki (P) 1  pay twenty guilder   dress here ‘I paid twenty guilders for this dress.’ (Isabelle Therrien’s field notes)

Fourth, the phrase introduced by pa may be interpreted as ‘given NP’, exemplified in (4). (4)

Mayoría di e outo nan (…) ta mucho karo pa Majority of  car 3 (…)  very expensive  nan tamaño (P) 3 size ‘Most of the cars (…) are too expansive for/given their size.’ (Kouwenberg and Michel fc: 3)

Fifth, pa may be interpreted as ‘on behalf of ’, as is shown in (5). (5)

Lo mi papia kuné pa bo  1 talk with.3  2 ‘I will talk with him for you (i.e., on your behalf).’

(P)

(Isabelle Therrien’s field notes)

Sixth, the preposition pa may also occur with the demonstrative esei ‘that’, as in (6). In this case, it is interpreted as ‘on that account’, ‘for that (reason)’. (6)

Ta p’ esei mi’ n po’ laga di hib’ e pan p’ e (P) be  that 1  can leave of bring 3 bread  3 ‘That is why I have to bring her the bread.’ (Maurer 1988: 384, from de Jongh 1970)

Chapter 11. On the properties of Papiamentu pa 

Pa can also be interpreted as ‘because’, as in (7).4 (7)

Pa su bon comportacion su patron ta consider’ e mucho (P)  his good behaviour his boss  regard 3 much ‘Because of his good behaviour, his boss has much regard for him.’ (Maurer 1988: 364, from Kroon 1979)

Seventh, the complement of pa can be a Goal. In (8), the Goal is a place. (8)

M’ a haña mi ta bai konvoi di binti barku fo’i merka I  find 1  go convoy of twenty ship from U.S. pa fransha (P)  France ‘I found myself going on a convoy of twenty ships from the U.S. to France.’ (Andersen 1990: 76)

In (9), the Goal is a person. (9)

Pedro a manda un karta pa Maria Pedro  send  letter  Maria ‘Pedro sent a letter to Maria.’

(P) (Isabelle Therrien’s field notes)

Eight, the complement of pa may express duration, as is shown in (10).5 (10)

Juan a traha pa tres ora Juan  work  three hour ‘Juan worked for three hours.’

(P) (Isabelle Therrien’s field notes)

In all these examples, pa is a preposition. It can be distinguished from verbs on the basis of a simple syntactic test. Verbs may be stranded, as in (11). (11)

Kiko Juan a kumpra? what Juan  buy ‘What did Juan buy?’

(P) (Isabelle Therrien’s field notes)

Prepositions cannot be. The contrast in grammaticality between (12)a and (12)b shows that the preposition ku ‘with’ cannot be stranded. (12)

a.

Ku ken bo a papia? with who 2  talk ‘With whom did you talk?’ b. *Keni bo a papia ku ti

(P) (Isabelle Therrien’s field notes) (P)

Likewise, stranding of pa in (9) would yield an ungrammatical sentence. . Note that there is also a separate P word meaning ‘because’: pasobra (Birmingham 1970: 110). . Isabelle Therrien’s informant points out that she prefers the use of largu or of durante to express duration: Juan a traha tres ora largu or . . .durante tres ora.

 Claire Lefebvre and Isabelle Therrien

.. Pa may select purposive clauses The purposive clauses selected by pa can be either tensed or infinitival.6 When the purposive clause is tensed, its subject is overt, and it is referentially free. In (13), the overt subject of the purposive clause is coreferential with the subject of the main clause. (13)

Mi a pone un kama den kurá pa mi bolbe drumi pafó (P) 1  put  bed in yard  1 return sleep outside ‘I put a bed in the yard for me to sleep outside again.’ (Kouwenberg and Murray 1994: 54)

In (14), the overt subject of the purposive clause is disjoint from that of the main clause. (14)

Mi ta bai paga lus pa nos drumi 1  go switch-off light  1 sleep ‘I am going to switch the light off so we can sleep.’

(P) (Maurer 1993: 248)

When the purposive clause is infinitival, its subject is covert and it is necessarily interpreted as coreferential with the subject of the main clause, as in (15). (15)

Mi a bai ‘Job Centre’ pa buska trabou (P) 1  go Job Centre  find work ‘I went to the Job Centre to find work.’ (Kouwenberg and Murray 1994: 46)

Pa occurring with an infinitival purposive clause has often been analysed as a complementiser (e.g., Dijkhoff 1993; Kouwenberg and Murray 1994; Maurer 1993). In our view, this analysis does not hold for the following reasons. First, if pa is a purposive preposition in (13) and (14), on what basis can it be analysed as a complementiser in (15), since in both contexts, pa has the meaning ‘in order that/to’. Second, as will be argued below, in one of its functions, pa is a [+T] complementiser. With standard views (e.g., Chomsky 1981), we assume that complementisers are either [+T] or [−T], but not both. If pa were a [−T] complementiser in (15), the complementiser pa would have to bear both the + and the – value of the [Tense] feature. Third, the clausal complement of a preposition can be questioned. In this case, the preposition and the Wh-pronoun are fronted, as is shown in (16). (16)

Pa kiko tin ora bo mester tapa bo kurpa? (P)  what have our 2 must cover your body ‘What for, sometimes, do you have to cover yourself?’ (Isabelle Therrien’s field notes)

. Our analysis runs counter to that in Mufwene and Dijkoff (1989) who claim that there is no finite/non-finite distinction, therefore no infinitive, in P.

Chapter 11. On the properties of Papiamentu pa 

In contrast, the clausal complement of a complementiser cannot be questioned in this way, as is shown by the contrast in grammaticality between the sentences in (17)a and b. (17)

a.

Mi ta deseá pa e bai 1  wish  3 go ‘I want him to go.’ b. *Pa kiko bo ta deseá  what 2  wish

(P) (Maurer 1988: 49) (P) (Isabelle Therrien’s field notes)

On the basis of the above tests, it appears that, whether the purposive clause complement of pa is tensed or not, the lexical item pa that introduces it has the status of a preposition. In addition to the results of these tests, there is a theoretical argument based on the Government and Binding (henceforth GB) framework (see Chomsky 1981) that argues for the analysis of pa as a preposition in the context of purposive clauses. Recall from (15) that, when the purposive clause is infinitival, its subject is covert, and that it is necessarily interpreted as coreferential with the subject of the main clause. Within the GB framework, this interpretation follows from the fact that the subject of the purposive clause is a covert pronominal form, represented as PRO, that is obligatorily interpreted as coreferential with a nominal antecedent, in this case the subject of the main clause, and that cannot be in a position where it is governed (that is, PRO is ungoverned). If pa were a complementiser in (15), it would govern PRO, the subject of the embedded clause. Since PRO needs to be ungoverned, pa cannot be a complementiser in this context. As a preposition, pa governs its complement, that is the whole purposive clause, but not the subject position of this clause. This theoretical argument constitutes an additional argument for not analysing pa as a complementiser in the context of a purposive clause. Its analysis as a purposive preposition meaning ‘in order that/to’ provides a unified account of the fact that, as a preposition, pa can select either tensed or infinitival clauses.

.

Pa occurring as a mood marker

Kouwenberg and Lefebvre (in press) argue that, at least for some speakers, P pa may also have the function of an irrealis mood marker conveying obligation.7 As such, pa occurs between the subject and the verb, as part of the TMA system. In (18) and

. Therrien’s informant has a prospective interpretation of pa occurring in this context. It thus appears that for both Kouwenberg’s and Therrien’s informants, pa occurring between the subject and the verb is irrealis. Speakers appear to differ as to whether pa is modal or aspectual. A similar ambiguity is discussed in Lefebvre (1996) on the basis of Haitian Creole data. Pending further research on the aspectual interpretation of P pa, in this paper, we concentrate on its use as a mood marker.

 Claire Lefebvre and Isabelle Therrien

(19), pa follows tabata ‘past imperfective’ and ta predominantly ‘non-past’, respectively. As per the analysis in Kouwenberg and Lefebvre (in press), tabata and ta are in complementary distribution as head of TP, and pa heads MoodP. (18)

Mi tabata pa bai buska dos hamster serka nan (P) 1   go search two guinea.pig near 3 ‘I was supposed to go fetch two guinea pigs from them.’ (Silvia Kouwenberg’s field notes)

(19)

Huan ta pa yuda mi mañan. Juan   help 1 tomorrow ‘Juan is supposed to help me tomorrow.’

(P) (Silvia Kouwenberg’s field notes)

As is pointed out by Kouwenberg and Lefebvre (in press), the properties of pa in the above examples are reminiscent of the behavior of the mood marker fi/fu in Caribbean English based creoles (see e.g., Lefebvre and Loranger 2006; McWhorter 1997), and of pou in Caribbean French based creoles (e.g., Koopman and Lefebvre 1981, 1982; Sterlin 1989). Kouwenberg and Lefebvre (in press) analyse mood pa as the head of MoodP, a projection occurring between the subject and the verb. So, in addition to fulfilling the function of a preposition, P pa may fulfil the function of a mood marker, occurring between the subject and the verb, as part of the TMA system.8

.

Pa occurring as a complementiser

As a complementiser, pa is selected by emotive verbs such as ke ‘to want’ and deseá ‘to desire’, as is exemplified in (20) and (21). (20)

(21)

Hose ke pa Ramon a bai (kaba) Hose want  Ramon  go (already) ‘Hose wishes that Ramon were gone.’ Mi ta deseá pa e bai. 1  wish  3 go ‘I want him to go.’

(P) (Maurer 1988: 178) (P) (Maurer 1988: 49)

. As a mood marker, pa may also occur with restricted classes of verbs such as pidi ‘to ask’, as is exemplified in (i). (i)

E tata a pidi Hose i Thomas pa ø bai buska Wansito (P)  father  ask Hose and Thomas   go fetch Wansito ‘Father asked Hose and Thomas to go and fetch Wansito.’ (Maurer 1993: 247)

This use of Pp pa is reminiscent of Haitian mande pou ‘to ask’ (see Sterlin 1989). A more elaborate discussion of this use of Pp pa would require more data than what is currently available.

Chapter 11. On the properties of Papiamentu pa 

The complementiser pa is also selected by effective verbs such as perkurá ‘to see to’ in (22), or pidi ‘to ask’, as in (23). (22)

(23)

El a perkurá pa nan siña franses. 3  take.care  3 learn French ‘He saw to it that they learned French.’

(P) (Maurer 1988: 182) ø9

bai E tata a pidi Hose i Thomas pa nan  father  ask Hose and Thomas  3  go buska Wansito (P) fetch Wansito ‘Father asked Hose and Thomas to go and fetch Wansito.’ (Maurer 1993: 247)

Other such verbs that select the complementiser pa include mira ‘to see to’ and sòru ‘to see to’. The complementiser pa may also be selected by verbs that are basically classified as propositional such as bisa and di both meaning ‘to say’. When occurring with pa, however, these verbs are interpreted as effective verbs meaning ‘to order’ (see Maurer 1988: 51). This is exemplified in (24) and (25). (24)

(25)

El a bisa nan pa nan bai. (P) 3  tell 3  3 go ‘He told them to go.’ (Maurer 1988: 51) Mi di kuné pa e bai buska e tapa unbé (P) 1 tell with-3  3 go look.for  cover one.time ‘I told him to go look for the cover right now.’ (Maurer 1988: 175, from Kleinmoedig-Eustatia 1981)

Clauses introduced by the complementiser pa are tensed. With Kouwenberg and Lefebvre (in press), we take it that a major test for tensedness in P is obligatoriness of an overt subject. Clauses introduced by the complementiser pa require an overt subject. In (24) and (25), the clauses introduced by the complementiser pa contain an overt subject. Suppression of the overt subject from these clauses would yield ungrammatical sentences. While clauses introduced by the complementiser pa require an overt subject, infinitival clauses introduced by di require a covert subject. In this case, the covert subject of the embedded clause is interpreted as being coreferential with the subject of the main clause, as shown in (26). Following the convention in (Chomsky 1981), we refer to this covert subject as PRO.

. Given that pa introduces clauses that are in the subjunctive mood, on the one hand, and given the fact that there is no subjunctive mood marker showing up between the subject and the verb in clauses introduced by pa, several authors (e.g., Andersen 1990; Maurer 1993) have proposed that P has a null subjunctive mood marker occurring between the subject and the verb of a clause introduced by pa. This topic is not discussed in this paper.

 Claire Lefebvre and Isabelle Therrien

(26)

Ela primintí di duna mas informashon 3- promise  give more information ‘He promised to give more information.’

(P) (2004-07-13: 1)10

The two sentences in (27) and (28) lack an overt subject. (27)

(28)

Ajera mi tabata ke pa tabata jobe Yesterday 1  want   rain ‘Yesterday I wanted it to rain.’ Mi ke pa piki koko 1 want  pick coconut ‘I want that there be coconuts picked.’

(P) (Kouwenberg 1990: 46) (P) (Kouwenberg 1990: 44)

In spite of the fact that they lack an overt subject, these clauses are not infinitival. Their covert subject is, however, of a different type than the covert subject of the sentence in (26). Indeed, in (27) and (28), the subject of the embedded clauses is an expletive (e.g., ‘it’, ‘there’ in English). In P, expletive subjects are always covert, that is, they are pronominals lacking a phonological representation. In contrast to PRO occurring as the subject of infinitival clauses, the covert subjects of the embedded clauses in (27) and (28) are not coreferential with the subject of the main clause. According to the analysis in Therrien (2004), the covert subjects of (27) and (28) have the properties of small pros, i.e., covert pronouns with pronominal features. Following this analysis, the sentences in (27) and (28) do not constitute counter examples to the fact that P tensed clauses require an overt subject. P pa is excluded from infinitival clauses introduced by di, as in (26). This fact further argues for the claim that P clauses containing the complementiser pa are [+T]. We thus conclude that, as a complementiser, pa is identified by the feature [+T].11 In addition to being [+T], pa is subjunctive as opposed to indicative.12 As we saw above, verbs that select pa are effective and emotive verbs. Cross-linguistically, verbs belonging in these semantic classes select a subjunctive complementiser (see Farkas 1992 for further discussion). This is congruent with the fact that the event described by the clause introduced by pa is not presupposed to have occurred nor to occur in the near future. Furthermore, in its function as a complementiser, pa is mutually exclusive with the complementiser ku. As a complementiser, ku is selected by propositional verbs or verbs that are interpreted as such. Typically, verbs belonging in this semantic class select an indicative complementiser such that the event described by the clause

.

This example is drawn from the P daily newspaper Èxtra.

. Mufwene and Dijkoff (1989: 312) analyse clauses like those in (27) and (28) as being infinitival, hence their conclusion that, as a complementiser, pa can be either [+/−T]. . We are using subjunctive and indicative following the terminology proposed by Winford (1993: 290) for Caribbean English based creoles.

Chapter 11. On the properties of Papiamentu pa 

introduced by an indicative complementiser is presupposed to have occurred, or to definitely occur in a near future. As is illustrated in (29), P ku has the properties of an indicative complementiser. (29)

Nan a bias papa ku mi n’ ta bai winkle di sapaté (P) 3  tell Papa  1   go shop of shoemaker ‘They told Papa that I was not going to go (to work in) the shoemaker’s shop.’ (Andersen 1990: 87)

Extensive discussion of ku in its use as an indicative complementiser may be found in Kouwenberg and Lefebvre (in press), and in Therrien (2004). We thus conclude that, as a complementiser, pa is tensed, as opposed to di which is infinitival, and that it is subjunctive, as opposed to ku which is indicative. Several arguments support the analysis of pa as a complementiser. First, it is a property of complementisers to occur at the periphery of the clause. As a complementiser, pa always occurs at the left periphery of the clause, before the subject. Second, it is a property of complementisers to be selected by specific classes of verbs. As a complementiser, pa is selected by effective and emotive verbs, or by verbs that behave as such. This contrasts with its function as a mood marker where it is not selected. Third, it is a property of complementisers to be specified for a value of tense. As we saw above, as a complementiser, pa is identified for the feature [+T]. This contrasts with its function as a preposition selecting either + or – tensed purposive clauses. Fourth, it is a property of complementisers to encode indicative or subjunctive mood. As a complementiser, pa is subjunctive, as opposed to ku which is indicative. Finally, it is a property of complementisers to delete in specific contexts. Like other complementisers, pa may be deleted, as is shown in (30). (30)

Ma mi no ke (pa) bo grita òf kanta But 1  want  2 scream or sing ‘But I don’t want for you to scream or sing.’

(P) (Maurer 1988: 173)

As is pointed out in Maurer (1988: 173), however, the conditions under which pa may be deleted still remain to be determined. What is the position occupied by pa in its function as a complementiser? In discussing this question, we adopt the split CP analysis of Rizzi (1997), according to which CP in fact corresponds to four projections – For(ce)P, Top(ic)P, Foc(us)P, and Fin(iteness)P. In this analysis, ForP is the projection where the relation of dependence between the embedded and the matrix clause is expressed. Hence, ForP is the position for [+T] complementisers. Since pa has the function of a [+T] complementiser, as such, it would be expected to be the head of ForP.

.

Pa is a multifunctional lexical item

Throughout this section, it was shown that pa fulfils several functions: preposition, mood marker and complementiser. Assuming an underspecification analysis of

 Claire Lefebvre and Isabelle Therrien

multifunctional lexical items, pa would be able to head various projections. Hence, as a preposition, pa would head PP, as a complementiser, it would head ForP, as a mood marker, it would head MoodP. These syntactic positions are illustrated in (31). (31)

a.

Figure 1. pa occurring as a preposition PP P' pa NP / clause [+/–T]

b.

Figure 2. pa occurring as a complementiser V' ForP

V

For' pa

c.

……

Figure 3. pa occurring as a mood marker between the subject and the verb T' T

MoodP Mood' pa

……

The various functions of pa and the projections that it heads are not entirely foreign to each other. First, there is a semantic link between its various functions. Indeed, pa is [−realis] in all of its functions: as a mood marker heading MoodP, as a subjunctive complementiser heading ForP, or as a purposive preposition meaning ‘in order that/to’ selecting clauses. The feature [−realis] cannot readily be claimed to be part of the definition of the preposition pa selecting NPs. However, in many languages, the same preposition that selects purposive clauses also selects benefactive NPs. This suggests that there must be a semantic link between the two uses of this preposition, with NPs and with clauses. This link can be seen in the relationship

Chapter 11. On the properties of Papiamentu pa 

that there is between a purposive clause denoting an event yet to occur and a Goal argument yet to be reached. Second, the syntactic categories that pa can be the head of are not entirely foreign to each other either. For example, the shared properties of P and C are extensively discussed in Emonds (1985),13 and Haegeman (1998: 167) refers to English for as a prepositional complementiser. The relationship between Mood and C is also well documented. On the one hand, the lexical items occupying either one of the two positions must agree in mode, such that both C and Mood must be either indicative [+realis] or subjunctive [−realis] within a given clause. On the other hand, the same lexical item may occupy either one of the two positions. Third, several cases of reanalysis involving these categories have been reported on in the literature, showing that they are not unrelated. For example, Hinds (1973) and Washabaugh (1975) document cases of development of complementisers from prepositions. Koopman and Lefebvre (1981) document cases of development of complementisers from mood markers. So, the various functions fulfilled by pa are, in some way or other, related to each other. A more elaborate discussion of this issue is far beyond the scope of this paper. However, work along the lines of the above cited authors, and others, suggests that the postulated links are pointing in the right direction.

.

Pa occurring as a case marker?

In the examples below, the complements of pa bear various thematic roles. In (32), the complement of pa is a possessor. (32)

Mi ker a pidi Kompa ku Kompa por para padrino di 1 want  ask Kompa  Kompa can stop godfather of e yu pa mi (P)  child  1 ‘I would like to ask you (Kompa) if you (Kompa) would be the godfather of my child.’ (Maurer 1988: 364, from Kleinmoedig 1981)

In (33), the complement of pa is an agent. (33)

E kara a keda studiá pa un religioso i un  face  stay study   man.of.religion and  professor (P) professor ‘The face has been studied by a man of religion and by a professor.’ (Maurer 1988: 330)

The status of pa in the above examples is far from being clear. Pa appears to manifest the properties of a generalised preposition or of a case marker. First, while the

.

Emonds goes as far as to claiming that P = C, a conclusion that we do not share.

 Claire Lefebvre and Isabelle Therrien

preposition pa discussed in Section 1.1 has semantic content with the general meaning ‘for’, pa in the above examples has less semantic content. Rather, it appears to be relational. While prepositions do have semantic content, case markers are semantically lighter. Second, in all the contexts where pa occurs as a preposition, pa has a core meaning. In the above examples, however, pa appears to be deprived of meaning since it is relational. Furthermore, its complements bear thematic roles such as possessor or agent. These thematic roles are not related to those of the NP complements of the preposition pa. It is a property of case markers to be able to associate with s bearing various thematic roles (Lefebvre and Brousseau 2002: 47). The properties of pa as it occurs in examples (32)–(33) are properties that are generally associated with the syntactic category case rather than preposition. Although more syntactic tests could be used in order to substantiate the case function of pa, for the sake of the present discussion, we will assume that, in the above examples, pa has the function of a case marker, and that, as such, it heads the projection KP (=CaseP) (see Travis and Lamontagne 1992), as in (34). (34)

Figure 4. pa occurring as a case marker KP K' pa

NP

On the basis of the data discussed in this section, we conclude that, in addition to its other functions, discussed in Sections 2.1, 2.2, and 2.3, pa possibly has the function of a case marker.

.

Summary

The various functions of P pa are summarised in Table 1. Table 1. The functions of P pa ●

Preposition + NP ‘for’ + benefactive ‘for’ + one’s own sake ‘in exchange of NP’ ‘given property x’ ‘on behalf of NP’ pa + ‘this’: ‘for this reason’ ‘because’ + clause ‘for’ + Goal place ‘for’ + Goal person ‘for’ + NP expressing duration Continued

Chapter 11. On the properties of Papiamentu pa 

Table 1. Continued ●

●

●

●

Preposition + [+/−T] purposive clause + tensed: embedded subject is disjoint from main clause subject + tensed: embedded subject is coreferential with main clause subject − tensed: embedded subject is coreferential with main clause subject Mood Marker between the subject and the verb Complementiser tensed/irrealis/subjunctive Case marker of possessor agent

P pa can function as a preposition. In that case, it heads a PP. It may select NPs or purposive clauses. Purposive clauses can be either [+T] or [−T]. Moreover, P pa can function as a mood marker. As such, it heads MoodP. P pa can also function as a tensed irrealis/subjunctive complementiser. With this function, pa heads ForP. Finally, pa appears to have one other function that remains to be further documented: that of a case marker. Having identified the functions of P pa, we now turn to the discussion of the source of its properties.

.

The source of the properties of pa: a three-way comparison

In order to identify the source of the properties of P pa, we will adopt the general methodology designed during the Haitian Creole project, as reported in Lefebvre (1998b: 65–70). We will thus undertake a three-way comparison between the creole lexical entry, pa, and the closest lexical entries in the source languages of the creole. Due to limitations of various kinds, we will limit our comparison to one language in each source. Our choice of languages takes into account the approximate date where P was created: around 1650 (e.g., Kouwenberg and Muysken 1995; Maurer 1986). Since the P form pa is hypothesised to be derived from para by syllable truncation (e.g., Birmingham 1970), it could be derived from either Portuguese or Spanish para. Various sources point out the widespread exposure of the African population in Curaçao to the Portuguese language around 1650 (e.g., Fouse 2002; Lenz 1928; and authors cited in Note 1). Furthermore, according to Lenz (1928), Spanish would have had an important input to P only in the 19th century, when business men from Colombia and Venezuela established themselves in Curaçao. On the basis of this information, our comparison with the superstrate will be based on Portuguese, rather than on Spanish.14 . Other colonial languages such as Dutch of English are not considered here for two reasons. First, according to the sources cited in the introduction they were not involved in the make up of early Papiamentu. Second, they did not contribute the form of the lexical entry under

 Claire Lefebvre and Isabelle Therrien

Since the variety of Portuguese the creators of P have been exposed to has not been identified in the literature on P, our references include work on both European and Brazilian Portuguese. As for the substratum language, our comparison will be based on Fongbe, one of the Gbe languages spoken by members of the African population in Curaçao at the time the creole was formed. Although Gbe languages, among which Fongbe, were not the only ones spoken when P was formed (Parkvall 2000), they were important ones during the 16th/17th century slave-trade in the Caribbean area in terms of number of speakers (see e.g., Singler 1996, and the references cited therein). Another reason for choosing Fongbe for comparative purposes, is that this language is the best documented among the P potential substrate languages (see Lefebvre and Brousseau 2002, and the references cited therein). In no way does this methodological choice implies that Fongbe was the only substrate language of P. There were certainly other Gbe languages involved, other Kwa languages involved, and possibly some Bantu languages involved in the formation of P. Another methodological issue has to do with the varieties of the languages selected for the comparative analysis. Ideally, varieties of the languages as spoken in the 16th/17th century should be the ones under analysis here. While there might be data available for 16th/17th century Brazilian Portuguese, there are none for P nor for Fongbe. Given this situation, the three-way comparison presented in this section is based on modern data. In no way does this methodological choice entail that there has been no change in the languages involved since 1650. The test we are presenting here is not the ideal one. However, we are convinced that it is a valid one. The choice of Portuguese for the discussion of the properties of the superstrate data is both linguistically and historically motivated since pa appears to have been derived from para by syllable truncation. The choice of Fongbe for the discussion of the properties of the substrate data is also historically motivated. Potential criticisms of the various components of a methodology along these lines are addressed in Lefebvre (1998b: 65–70). Space limitations prevent us from reproducing them here. The markedness issue is also addressed in Lefebvre (1998b: Chapter 3). The three-way comparison that we are presenting below thus involves Portuguese para, P pa, and Fongbe nú/ní. The comparison shows that, while the form of the P lexical item pa is most probably derived from Portuguese para, its other properties are not derived from this Portuguese lexical item. Rather, they seem to be derived from corresponding lexical entries in the P substratum languages.

.

P pa and Portuguese para

P pa is most probably phonologically derived from Portuguese (henceforth listed as P in the examples) para ‘for’ through syllable truncation. Syllable truncation is consideration in this paper. Indeed, the P form pa can only be traced to the Portuguese/Spanish form para, as specified in the text.

Chapter 11. On the properties of Papiamentu pa 

widespread in P (e.g., Birmingham 1970). Examples of this process are provided in (35). (35)

P P cansado > kansa ‘tired’ pecado > pika ‘sin’ pesado > pisa ‘heavy’

(Birmingham 1970: 31)

In the paragraphs that follow, we thus compare the properties of P pa with those of Portuguese para. Like P pa, Portuguese para may serve as a preposition. As such, it may select NPs. The NP selected by para may be a benefactive as in (36), which corresponds to P (1). (36)

Este livro é para ti this book be for 2 ‘This book is for you.’

(P) (Cantel 1999: 130)

The phrase introduced by para may be interpreted as ‘given NP’, exemplified in (37), in which the use of para corresponds to that of P pa in (4). (37)

Esta criança é grande para sua idade This child be tall for his age ‘This child is tall for his age.’

(P) (Isabelle Therrien’s field notes)

It may also select a Goal (place), as in (38), which corresponds to P pa in (8). (38)

O sertanejo abandonou o Recife e voltou para  sertanejo abandon  Recife and come.back for a terra natal (P)  soil native ‘The sertanejo left Recife and came back to his village.’ (Boncompain 1989: 57)

On the basis of Boncompain (1989), Cantel (1999) and Dunn (1928), the above data exhaust the range of NP complements that para may select as a preposition. In contrast to P pa, para does not select a NP meaning ‘one’s own sake’, ‘one’s own good’, as Portuguese (39), which parallels P (2), is not grammatical. (39) * Pode ir para você mismo Can go for 2 self [Lit. ‘You can go for yourselves.’]

(P) (Isabelle Therrien’s field notes)

Unlike P pa in (3), Portuguese para cannot be interpreted as ‘in exchange of ’, as (40) is not grammatical. (40) * Eu paguei vinte euros para este vestido 1 pay twenty euro for this dress [Lit. ‘I paid twenty euros for this dress.’]

(P) (Isabelle Therrien’s field notes)

 Claire Lefebvre and Isabelle Therrien

Unlike P pa in (5), Portuguese para cannot be interpreted as ‘on behalf of ’, as is shown by the ungrammaticality of (41). (41) *Eu posso falar-lhe para você 1 can talk-3 for 2 [Lit. ‘I can talk to him for you.’]

(P) (Isabelle Therrien’s field notes)

Unlike P pa in (6), Portuguese para cannot be used with the meaning ‘for this reason/ because’. This is shown by the ungrammaticality of (42).15 (42) *É para isto eu voltei be for this 1 come.back [Lit. ‘This is why I came back.’]

(P) (Isabelle Therrien’s field notes)

Unlike P pa in (9), Portuguese para cannot introduce a Goal of the type of that in (43). (43) *Escrevi para pais de Maria write for parent of Maria [Lit. ‘I wrote for Maria’s parents.’]

(P) (Isabelle Therrien’s field notes)

Unlike P pa in (10), Portuguese para cannot express duration, as is shown by the ungrammaticality of (44). (44) *João trabalhou para três horas Juan worked for three hour [Lit. ‘Juan worked for three hours.’]

(P) (Isabelle Therrien’s field notes)

The discrepancy between the two languages shows that Portuguese para has a smaller semantic range than P pa. In fact, the semantic range covered by P pa is shared by several lexical items in Portuguese. Para is one of them. But Portuguese also has por. For example, the notion ‘in exchange of ’ is rendered by por, as is shown in (45). (45)

Pagei um dinheirão por esta casa Pay  large.amount for this house ‘I paid a fortune for this house.’

(P) (Boncompain 1989: 59)

The notion of ‘on behalf of ’ is also rendered by por, as shown in (46). (46)

Não posso assinar por êle  can sign for 3 ‘I cannot sign for him.’

(P) (Dunn 1928: 594)

Likewise, the notion ‘because’ is rendered by por, as in (47).

. A reviewer points out that the sentences in (42) and (44) are grammatical in modern European Portuguese.

Chapter 11. On the properties of Papiamentu pa 

(47)

Veio por ser amigo dele come for be friend of-3 ‘He came because he is a friend of his.’

(P) (Cantel 1999: 130)

While the expression ‘for ever’ makes use of para as in para sempre, duration may also be expressed by means of the lexical item por, as is shown in (48). (48)

Viajou por una semana travel for  week ‘He travelled for a week.’

(P) (Dunn 1928: 598)

Finally, when the Goal argument is a person, it is introduced by the preposition (or case marker) a, as is illustrated in (49). (49)

Escrevi aos pais de Maria Write to.. parent of Mary ‘I wrote to Mary’s parents.’

(P) (in Cantel 1999: 58)

When the Goal argument is a place, the Goal may also be introduced by a, as in (50). (50)

Viajou de Paris a Londres travel from Paris to London ‘He travelled from Paris to London.’

(P) (Isabelle Therrien’s field notes)

As we saw in (38), however, a Goal argument that is a place may also be introduced by para (see Dunn 1928) for a discussion of the subtle difference in meaning between the two forms). So, the semantic load of one lexical item in P is divided between several lexical items in Portuguese: para, por and a. In addition to selecting NPs, Portuguese para, like P pa, selects purposive clauses. As is the case in P (see (13)–(15)), these purposive clauses may be tensed or infinitival, as in (51) and (52), respectively. (51)

(52)

Esconde-te para que eles não te vejam Hide-2 for  3  2 see ‘Hide yourself so that they don’t see you.’ O menino subiu à arvore para apanhar laranjas  child climb to tree for gather orange ‘The child climbed the tree in order to gather oranges.’

(P) (Boncompain 1989: 43) (P) (ibid.: 57)

Recall from Section 2.1.2 that the subject of the tensed purposive clause selected by P pa is referentially free, such that it can be coreferential with the subject of the main clause, as in (13), or it can be disjoint from it, as in (14). In Portuguese, the subject of the tensed purposive clause selected by para cannot be coreferential with that of the main clause, as is shown by the ungrammaticality of the sentence in (53).

 Claire Lefebvre and Isabelle Therrien

(53) *O meninoi subiu à arvore para elei apanhara laranjas (P)  child climb to tree for 3 gather orange [Lit: ‘The child climbed the tree in order for he gather oranges.’] (Isabelle Therrien’s field notes)

Typically, infinitival clauses exhibit no overt subject, and the main verb of the clause bears neither agreement, nor tense morphology. The sentence in (52) above illustrates such a typical infinitival clause. In addition to the standard infinitival structure in (52), Portuguese has another infinitival structure that has quite unusual properties. In this structure, not only does the infinitival complement of para exhibit an overt lexical subject, but this lexical subject bears nominative case. In (54), the subject of the infinitival complement clause is the nominative pronominal form eu that contrasts with the accusative form me. Furthermore, although the verb is unmarked for tense, it agrees in person and number with its subject. According to Dunn (1928: 495), these endings are: 1sg: ø, 2sg: -es, 3sg: ø, 1pl: -mos, 2pl: -des, 3pl: -em. This unusual infinitive is exemplified in (54). (54)

Emprestou-me dinheiro para eu comprar un casaco. (P) lend-1 money in.order.to 1 buy  coat ‘He lent me money in order for me to buy a coat.’ (Dunn 1928: 496)

Some authors (e.g., Boretzky 1983) have suggested that the unusual infinitive construction of Portuguese is reminiscent of the for constructions in creoles (see e.g., P (13)). Holm (1988) has argued against such a view: The Portuguese personal infinitive construction is only partially parallel to the creole ‘for’ constructions: while the Portuguese infinitive can take a [nominative] subject and even an inflectional ending indicating agreement in person and number . . . it is still an infinitive and thus – unlike the corresponding creole verbs – untensed (Holm 1988: 170).

With Holm, we conclude that the unusual infinitival structure of Portuguese is not attested in P. We have shown that P pa has the function of an irrealis mood marker (see (18)). Unlike P pa, Portuguese para does not have this function. We have also shown that in P, pa also has the function of a complementiser. As such, it introduces tensed irrealis/subjunctive clauses (see e.g., (20)). Subjunctive sentential complements that are introduced by pa in P are generally introduced by the tensed irrealis complementiser que, in Portuguese, as is illustrated in (55). (55)

Quero que você me faça um favor want  2 1 do  favor ‘I want you to do me a favor.’

(P) (Boncompain 1989: 43)

The same sentential complements may also be rendered in a structure involving the unusual infinitival construction described above. This is illustrated in (56).

Chapter 11. On the properties of Papiamentu pa 

(56)

Eles pediram para eu voltar 3 ask  1 return ‘They asked me to return.’

(P) (Holm 1988: 169)

As irrealis complementisers, both P pa and Portuguese que select a [+T] subjunctive clausal complement. Portuguese also offers the option of replacing que by the [−T] complementiser para, in which case the complement clause appears as the unusual infinitival construction described above. Again, the latter option is not available in P. So even though the surface position of Portuguese para and that of P pa coincide, their properties and those of the clauses they introduce are quite distinct. The similarities and dissimilarities of P pa and Portuguese para are summarised in Table 2.

Table 2. Comparative properties of P pa and of Portuguese para

●

●

●

●

●

Preposition + NP ‘for’ + benefactive NP ‘for’ + ‘one’s own sake’ ‘in exchange of NP’ ‘given property x’ ‘on behalf of NP’ pa/para + ‘this’: ‘for this reason’ ‘because’ + clause ‘for’ + Goal place ‘for’ + Goal person ‘for’ + NP expressing duration Preposition [+/−T] purposive clause + tensed: embedded subject; disjoint from main clause subject + tensed: embedded subject; coreferential with main clause subject − tensed: usual infinitival structure − tensed: unusual infinitival structure Mood marker between the subject and the verb Complementiser tensed, irrealis, subjunctive unusual infinitival structure pa/para + NP possessor agent

pt para

Pp pa

+ − − + − − − + − −

+ + + + + + + + + +

+ − + +

+ + + −

−

+

− +

+ −

− −

+ +

On the basis of this comparison, our conclusion is that, although Portuguese para has most probably provided the label for the P lexical entry pa, the two lexical entries only share a small portion of their semantic and syntactic properties.

 Claire Lefebvre and Isabelle Therrien

.

P pa and Fongbe nú and ní

The Fongbe lexical items closest to P pa are nú and ní. While nú serves as a preposition, as a complementiser, and as a case marker, ní serves both as a mood marker and as a complementiser. As will be seen below, in some contexts, nú and ní are interchangeable. Thus, even though nú and ní are separate lexical items, in some environments, their meanings and functions intersect. In this section, it is shown that, to a large extent, the properties of P pa cumulate those of Fongbe nú and ní. Like P pa, Fongbe nú selects NPs. The complement of nú may be benefactive, as is shown in (57), which corresponds to P (1). (57)

K ɔ̀kú x ɔ̀ às ɔ̀n nú Àsíbá. Koku buy crab ú Asiba ‘Koku bought crab for Asiba.’

(F) (Lefebvre and Brousseau 2002: 302)

The NP complement of nú may be an expression meaning ‘one’s own sake’. This is exemplified in (58). The Fongbe data in (58) parallel P data in (2). (58)

Mí yì nú mí.ɖéè 2 go ú you..self ‘You may go for yourselves.’

(F) (Claire Lefebvre’s field notes)

In the context of verbs of paying, Fongbe nú may also be interpreted as ‘in exchange of ’, as in (59), which parallels P (3). (59)

Àsíbá ná àkwɛ́́ nú àsɔ́n ɔ̀ Asiba gave money ú crab  ‘Asiba gave money in exchange for the crab.’

(F) (Claire Lefebvre’s field notes)

In Fongbe the phrase introduced by nú may also be interpreted as ‘given NP’, as in (60), which parallels P (4). (60)

Mótò ɔ̀ vè-àhì nú kloklo é dó é car  cost ú size 3 have  ‘The car is expensive for its size.’

(F) (Claire Lefebvre’s field notes)

Fongbe nú may also be interpreted as ‘on behalf of ’, as in (61), which parallels P (5). (61)

À ná dó gbè nú tɔ̀ tówé nú mí (F) 2  throw voice ú father your ú me ‘You will say hello to your father for me.’ (Segurola and Rassinoux 2000)

Fongbe nú may also occur with NPs denoting duration, as is shown in (62). This parallels P pa in (10)a. (62)

É kán-wèzùn nú àzán àtɔ̀n 3 run ú day three ‘He ran for three days.’

(F) (Claire Lefebvre’s field notes)

Chapter 11. On the properties of Papiamentu pa 

Unlike P pa which can be used with a demonstrative pronoun yielding the interpretation ‘for this reason’ (see (6)), Fongbe nú cannot be used in this way. When the notion ‘reason’ is intended, the postposition wútú (also realised as útú/ú) is used, as in (63). (63)

a.

b.

énɛ́́ ú ɔ́ that cause  ‘for that (reason)’ étɛ́ wútú what cause ‘why’

(F)

(F) (Claire Lefebvre’s field notes)

(This discrepancy between the two languages will be taken up in Section 4). However, in some contexts, Fongbe nú may be interpreted as ‘because’, as is shown in (64). The meaning of nú in (64) corresponds to that of pa in (7). (64)

À wá nú àzɔ́ ɔ́ 2 come ú work  ‘You came for/because of your work.’

(F) (Claire Lefebvre’s field notes)

Fongbe nú may also introduce a Goal. The Goal may be a place as in (65) which parallels P (8), or a person, as in (66) which parallels P (9). (65)

(66)

Sɔ́ wèmá élɔ́ yì nú nɔ̀ví tówè bring book this go ú brother  ‘Bring this book to your brother.’ Ùn gò sín Pàrí nú Kùtónú 1 leave from Paris ú Cotonou ‘I left Paris for Cotonu.’

(F) (Segurola and Rassinoux 2000) (F) (Claire Lefebvre’s field notes)

The data presented above show that the range of NPs that P pa can take corresponds to the range of NPs that Fongbe nú can take, with one exception (to which we come back in Section 4). Like P pa, the Fongbe preposition nú may select purposive clauses. These can be tensed, as is illustrated in (67), which corresponds to P (14). (67)

Àjwá wà àzɔ́ ɔ́ nú Bàyí ní mɔ̀ àkwɛ́́ Ajwa do work  ú Bayi í find money ‘Ajwa did the work in order that Bayi would have money.’

(F) (da Cruz 1994: 25)

In (67), the subject of the embedded clause is referentially disjoint from that of the main clause. Recall from (13) that in P, the subject of the embedded clause can also be coreferential with that of the main clause. In Fongbe, this option is not available. Recall from (15) that P pa also selects infinitival purposive clauses. In contrast to P pa, the Fongbe preposition nú does not select infinitival purposive clauses. In section 4, we come back to these two discrepancies between P and its substratum language.

 Claire Lefebvre and Isabelle Therrien

As for the function of mood marker, the Fongbe morpheme is ní. With this function, ní occurs between the subject and the verb, as part of the TMA system (see Lefebvre 1996; Lefebvre and Brousseau 2002). It may express an obligation, as in (68), a wish, as in (69), or an order, as in (70). (68)

(69)

(70)

Bàyí ní ɖà wɔ́ Bayi í prepare dough ‘Bayi should prepare dough.’

(F) (Lefebvre and Brousseau 2002: 117)

wè Máwú ní cɔ́ God í protect 2 ‘May God protect you.’ Mí ní ɖù nú 1 í eat thing ‘Let’s eat.’ [Lit. ‘We must eat thing.’]

(F) (Anonymous 1983: V, 4) (F) (Lefebvre and Brousseau 2002: 93)

This use of Fongbe ní parallels that of P pa occurring as a mood marker between the subject and the verb (see (19)). Whether P pa may have the same range of meanings as Fongbe nú in this function is a topic for future research. What matters at this point is the fact that, like P pa, Fongbe ní may function as a mood marker that is part of the TMA system. Fongbe ní may also occur in an imperative construction lacking an overt subject, as is illustrated in (71). (71)

Ní ɖù! í eat ‘He must eat!’

(F) (Avolonto 1992: 53)

In this context, nú may also be used, as is illustrated in (72). (72)

Nú à wá! ú 2 come ‘Come!’

(F) (Anonymous 1983: V, 4)

In (72) nú precedes the subject. By hypothesis, ní in (71) is in the same position as nú, thus in a pre-subject position. Available data on P pa do not attest that a pre-subject position is available for pa in P. Fongbe ní and nú also occur in conditional clauses as is exemplified in (73) and (74), respectively. In this context, they have an epistemic reading. (73)

Ní jì jà ɔ́, Kɔ̀kú ná wà àzɔ́ a˘ (F) í rain fall  Koku . do work  ‘If it rains, Koku will not work.’ (Lefebvre and Brousseau 2002: 175)

(74)

ɔ́, mí ná wà àzɔ́ Nú à wá sɔ̀ ú 2 come tomorrow  1 . do work ‘If you come tomorrow, we will work.’

(F) (Kinyalolo 1993)

Chapter 11. On the properties of Papiamentu pa 

Ní and nú are interchangeable when they occur in the context of (73) and (74) (Lefebvre and Brousseau 2002: 175). Furthermore, according to Kinyalolo (1993: 162), ní and nú are optional in this context. To our knowledge, P exhibits no such use of pa. This discrepancy will be addressed in Section 4. As is the case of P pa, Fongbe ní and nú may also serve as complementisers. Since they select tensed clauses, they are marked for the feature [+T]. They are selected by emotive verbs, as is exemplified in (75) for nú, and in (76) for ní. The event described by a clause introduced by ní/nú is not presupposed to have occurred, nor to occur in the near future; ní and nú are thus tensed, irrealis, subjunctive complementisers. (75)

Ùn jló nú à ní wá (F) 1 want ú 2 í come ‘I want you to come.’ [Lit. ‘I want that you come.’] (Lefebvre and Brousseau 2002: 116)

(76)

Ùn jló ní à ní wá (F) 1 want í 2 í come ‘I want you to come.’ [Lit. ‘I want that you come.’] (Lefebvre and Brousseau 2002: 117)

The above examples show that, as complementisers, ní and nú are interchangeable (see Lefebvre and Brousseau 2002: 117). Examples (75) and (76) correspond to P (21). As a complementiser, Fongbe nú is also selected by effective verbs such as ‘to see to’, as is shown in (77). (77)

Bló nú mɛ̀ bì ní m ɔ̀ nùɖúɖú See.to ú person all í have food ‘See (to it) that all the people have food to eat.’

(F) (Rassinoux 2000)

The Fongbe sentence in (77) parallels the P one in (22). Several arguments support the analysis according to which ní/nú are complementisers in (75)–(76). First, like complementisers, they occur at the periphery of the clause, in this case the left periphery. Second, like complementisers, they are selected by specific classes of verbs, in this case, emotive and effective verbs. Third, like complementisers, they are specified for tense; since ní/nú select tensed clauses, they are [+T]. Fourth, like complementisers, they are part of a paradigm, and, as such, they are in complementary distribution with the other complementisers in the language. They are thus in complementary distribution with ɖɔ̀, which serves as a complementiser selected by propositional (utterance, cognition and perception) verbs, as is shown in (78)–(80).16 . As is thoroughly discussed in Lefebvre and Loranger (in press) the form ɖ ɔ̀ is a multifunctional item serving as a verb meaning ‘say’, as a complementiser meaning ‘that’, as a quote introducer and as a marker conveying similarity and manner.

 Claire Lefebvre and Isabelle Therrien

(78)

(79)

(80)

É ɖ ɔ̀ nú mì ɖ ɔ̀ à ná wá 3 say ú 1 ɖɔ̀ 2 . come ‘(S)he told me that you would come.’ mɔ́tò ɔ́ K ɔ̀kú lìn ɖ ɔ̀ Àsíbá gbà Koku think ɖ ɔ̀ Asiba destroy car  ‘Koku thinks that Asiba destroyed the car.’

ɔ̀ yàví É m ɔ̀ ɖ ɔ̀ v í 3 see ɖ ɔ̀ child  cry ‘(S)he saw that the child cried.’

(F) (Anonymous 1983: X, 1) (F) (Lefebvre 1992) (F) (Tossa 1994: 181)

While ní/nú are subjunctive, ɖɔ̀ is indicative (see Lefebvre and Loranger in press). So, like P pa, in one of their functions, ní/nú serve as complementisers. As a complementiser, Fongbe nú introduces the clausal complements of both káká ‘until’ and có ‘before’, respectively, as is shown in (81). (81)

a.

b.

Káká nú nù nɛ́́ lɛ́́ ná jɛ́́ ɔ́, é kó before ú thing    happen  3 already yì (F) hɔ̀n run.away go ‘Before these things happened, he had already gone.’ (Rassinoux 2000) àmlɔ́n (F) Có nù à ná wá ɔ́, dɔ́ before ú 2 . come  sleep sleep ‘Before you come, sleep.’ (Lefebvre and Brousseau 2002: 172)

According to Isabelle Therrien’s informant, P pa does not introduce the clausal complement of corresponding adverbs in P. This is shown by the ungrammaticality of the sentence in (82). (82) *Promé pa bo bias algu before  2 say something [Lit. ‘Before you say anything.’]

(P) (Isabelle Therrien’s field notes)

As for P pa and Fongbe nú occurring with other NPs (most probably in their function as case markers), they compare in the following way. In contrast to P pa, Fongbe nú neither occurs with a possessor nor an agent, a topic which will be taken up in section 5. However, Fongbe nú occurs as a case marker with a small class of verbs such as ɖ ì-xɛ̀sì ‘to fear’, as in (83). (83)

Àsíbá ɖ ì-xɛ̀sì nú mɛ̀ Asiba fear ú someone ‘Asiba fears someone’

(F) (Rassinoux 1987)

To our knowledge, P pa is not used as a case marker in similar contexts. Informants consulted use the form di (< Portuguese de) instead of pa in this context, as is shown in (84).

Chapter 11. On the properties of Papiamentu pa 

(84)

Mi tin miedu di bo 1 have fear of 2 ‘I fear you.’

(P) (Isabelle Therrien’s field notes)

The similarities and dissimilarities of P pa and Fongbe ní/nú are summarised in Table 3. The striking fact about the distribution in Table 3 is that P pa has all the syntactic functions of Fongbe nú/ní. Like Fongbe nú/ní, P pa may fulfil the function of a preposition selecting NPs and tensed purposive clauses whose subjects are free with respect to the subject of the main clause, the function of mood marker occurring between the subject and the verb, the function of a tensed irrealis/subjunctive complementiser, and the function of a case marker. Differences in details between P pa and Fongbe nú/ní involve mainly the properties of the complement clause of the preposition, the fact that, as mood marker, pa does not occur in conditional clauses while nú/ní do, and finally, the fact that, as case markers, pa and nú do not have the same distribution. Even though there are these differences between P pa and Fongbe nú/ní, we consider that they are out ranked in importance by the qualitative and quantitative similarities between them. In Section 5, we further discuss the discrepancies between P pa and Fongbe nú/ní. Table 3. Comparative properties of P pa and of Fongbe nú and ní

Pp pa ●

●

●

Preposition + NP ‘for’ + benefactive NP ‘for’ + ‘one’s own sake’ ‘in exchange of NP’ ‘given property x’ ‘on behalf of NP’ ‘for’ + NP expressing duration pa/nú + ‘this’: ‘for this reason’ ‘because’ + clause ‘for’ + Goal place ‘for’ + Goal person Preposition [+/−T] purposive clause + tensed: embedded S* disjoint from main clause S + tensed: embedded S coreferential with main clause S − tensed Mood marker between the subject and the verb above the subject occurs in conditional clauses

fongbe nú fongbe ní

+ + + + + + + + + +

+ + + + + + − + + +

+

+

+

−

+

−

+ − −

+ +

+ + + Continued

 Claire Lefebvre and Isabelle Therrien

Table 3. Continued

●

●

●

Complementiser tensed, irrealis, subjunctive Introduces complements of the prepositions ‘before’ pa/nú +NP possessor agent ‘to fear’ + pa/nú + NP

Pp pa

fongbe nú

+

+

−

+

+ + −

− − +

fongbe ní +

*S = subject.

.

Summary

Table 4 compares the properties of P pa with those of Portuguese para, and of Fongbe nú/ní. The striking fact about the distribution in Table 4 is that, while P pa shares with Portuguese para the sole syntactic functions of preposition and complementiser, it has all of the syntactic functions of Fongbe nú/ní: preposition, mood marker, complementiser and case marker. Moreover, in its function as a preposition, P pa, like Fongbe nú, selects NPs that have a much wider variety of thematic roles than those selected by Portuguese para. In its function as a complementiser, P pa, like Fongbe nú, selects tensed irrealis clausal complements while Portuguese para selects the unusual infinitive structure. The overall distribution of pa and related lexical items in the contributing languages of P thus reveals that, if the label of P pa most probably comes from Portuguese para through syllable truncation, its other properties generally come from those of the closest Fongbe lexical items, nú/ní. Table 4. Comparative properties of Portuguese para, P pa, and Fongbe nú/ní

Pt para Pp pa fongbe nú fongbe ní ●

Preposition + NP ‘for’ + benefactive NP ‘for’ + ‘one’s own sake’ ‘in exchange of NP’ ‘given property x’ ‘on behalf of NP’ ‘for’ + NP expressing duration pa/para/nú + ‘this’: ‘for this reason’ ‘because’ + clause ‘for’ + Goal place ‘for’ + Goal person

+ − − + − − − − − +

+ + + + + + + + + +

+ + + + + + − + + + Continued

Chapter 11. On the properties of Papiamentu pa 

Table 4. Continued

Pt para Pp pa fongbe nú fongbe ní ●

●

●

●

●

Preposition [+/−T] purposive clause + tensed: embedded S* disjoint from main clause S + tensed: embedded S coreferential with main clause S − tensed − tensed unusual infinitival structure Mood marker between the subject and the verb above the subject occurs in conditional clauses Complementiser tensed/irrealis/subjunctive unusual infinitive structure Introduces complements of prepositions meaning ‘before’ pa/para/nú + NP possessor agent ‘to fear’ + pa/nú + NP

+

+

+

−

+

−

+ +

+ −

− −

− − −

+ − −

+ +

+ + +

− +

+ −

+ −

+ −

−

−

+

− − −

+ + −

− − +

*S = subject.

In light of this three-way comparison, we now turn to the account of such a division of properties of the creole lexical entry between its contributing languages. Mismatches between the properties of the creole lexical entry and those of the corresponding substratum ones will be taken up in Section 5.

.

A relexification account of the properties of Pp pa

This section provides a historical analysis of the properties of P pa, set within the framework of the relexification account of creole genesis, as is outlined in Lefebvre (1998b and the references therein). We begin with providing a definition of the process of relexification and with stating the constraints that are acting upon it.

.

Relexification as relabelling

Relexification is a cognitive process that consists in assigning a lexical entry of a language L1 a new label drawn from a language L2. The process of relexification thus reduces to relabelling.17 This process can be represented as in (85). Given a lexical .

See e.g., Lefebvre 1998b, Lefebvre and Lumsden 1994.

 Claire Lefebvre and Isabelle Therrien

entry as in (85)a, assign this lexical entry a new phonological representation drawn from another language, as in (85)b, and eventually remove the original phonological representation. The resulting lexical entry is as in (85)c. The semantic and syntactic properties of the new lexical entry are drawn from one language, while its label is drawn from another language. (85)

a.

Figure 5. /phonologyi/ [semantici] [syntactici]

b.

Figure 6. /phonologyi/ /phonologyj ¢/ [semantici] [syntactici]

c.

Figure 7. /phonologyj ¢/ [semantici] [syntactici]

The process of relabelling thus creates new lexical entries that have the semantic and syntactic properties of the original ones, and phonological representations derived from phonetic strings drawn from another language.18 As was shown in Section 3, P pa manifests this division of properties: while its phonological representation is derived from Portuguese para, many of its semantic and syntactic properties are derived from corresponding lexical items in Fongbe, one of the P important substratum languages. According to Muysken (1981: 62), relexification is semantically driven: “For relexification to occur, the semantic representations of source and target language entries must partially overlap; otherwise, the two entries would never be associated with one another. Other features of the two entries may, but need not, be associated with each other.”

As a consequence, lexical entries which have semantic content can be relabelled. Those which don’t, such as operators or case markers, cannot be relabelled. In Lefebvre and Lumsden’s (1994) scenario, the latter lexical entries are assigned a phonologically null representation at relabelling. They may acquire a label later through grammaticalisation. In light of these preliminary remarks we now turn to the relexification account of the creation of the P lexical entry pa. . For various representations of the process and extensive discussions of how it applies in various contexts, see e.g., Lefebvre 1998b; Lefebvre and Lumsden 1994; Mous 1995; Muysken 1981.

Chapter 11. On the properties of Papiamentu pa 

.

P pa results from the relabelling of substratum closest lexical entries

We propose that the P lexical entry pa was created through the relabelling of lexical items of the type of Fongbe nú and ní on the basis of Portuguese para. We provide a hypothetical historical derivation on the basis of these two Fongbe lexical entries. On the one hand, Fongbe nú and Portuguese para share some semantics. As prepositions, they both select NPs that can be interpreted as benefactive. Fongbe nú and Portuguese para could have been associated on the basis of this shared semantic property. So, let us assume that Fongbe nú was relabelled on the basis of Portuguese para, yielding the P lexical entry pa. As a consequence of relabelling, the properties and functions of Fongbe nú, that are not shared by Portuguese para, that is, that of selecting NPs meaning ‘one’s own sake’ or NPs with various thematic roles, that of functioning as a tensed subjunctive complementiser, and that of functioning as a case marker, found themselves as part of the definition of the P lexical entry pa. On the other hand, Fongbe ní and Portuguese para also share some semantics. Recall from examples (68)– (71) that ní is an irrealis TMA marker. Portuguese para may occur with a prospective, thus irrealis, meaning in contexts such as Juan está para salir ‘Juan is about to go out’. On the basis of this shared semantics, and shared position between the subject and the verb, Fongbe ní and Portuguese para could have been associated. The properties of Fongbe ní, that are not shared by Portuguese para, that is, that of being a mood marker, that of being a tensed subjunctive complementiser, found themselves as part of the definition of the P lexical entry through relabelling. The fact that Fongbe nú and ní appear to have merged in some of their functions (see Section 3) probably contributed to the conflation of both their properties in P pa. One of the reviewers is asking us to motivate the assumption that cumulation of substratum features has priority over cumulation of superstratum features “since otherwise we could also assume that P pa is a blend of Portuguese para + por with some overlap of que functions”. The answer to this comment is quite straightforward in view of the nature of the process of relabelling described above, and in view of the claim that it plays a major role in creole genesis. Indeed, if two substratum lexical entries (here nú and ní) are relabelled on the basis of the same superstratum form, (in this case para), the resulting creole lexical entry has the combined properties of both substratum lexical entries, provided that they are semantically related. The very process by which the blended creole lexical entries are created provides the motivation for cumulation of substratum features. In this analysis, the fact that P pa may constitute a “blend of Portuguese para + por with some overlap with que functions” is not pertinent to the question at stake, that is: how did the properties of P pa come to be the way they are? The relabelling of lexical entries of the type of Fongbe nú and ní on the basis of a single superstratum form para provides a straightforward account of the fact that most properties of P pa correspond to those of corresponding lexical entries in the substratum languages, rather than to those of Portuguese para.

 Claire Lefebvre and Isabelle Therrien

.

Mismatches between the substratum and the creole lexical entries

As can be seen from Table 4, the correspondences between the substratum and the creole lexical entries are not perfect. Among the differences observed between the substratum and the creole lexical entries, some are trivial, some are more important. We discuss these differences one by one.

.

The ‘for this (reason)’ discrepancy

While Fongbe nú does not occur in the expression meaning ‘for this reason’, P pa does (see (6)). Instead of using the preposition nú to convey the meaning ‘for this (reason)’, Fongbe uses the postposition wútú ‘cause’ as in énε´ wútú ɔ́ ‘for this reason’. In P there is no postposition corresponding to Fongbe wútú. What the above data show is that, in addition to assuming the uses of the Fongbe preposition nú, P pa also assumes some (or all?) of the uses of the Fongbe postposition wútú. A precise account of how much of the semantics of Fongbe wútú P pa has undertaken is a matter for future research.

.

The non-finite property of the purposive clause selected by P pa

Recall from Section 3.2 that, as prepositions, both Fongbe nú and P pa may select purposive clauses that are tensed. While P pa may also select purposive clauses that are not tensed, as in (15), Fongbe nú does not have this option. In Fongbe, tenseless purposive clauses are rendered in one of two ways. The first one involves a [−T] (nominalised) clause selected by the postposition gbé ‘for’. According to Segurola and Rassinoux (2000), this postposition encodes purpose, intention and Goal. It is illustrated in (86). (86)

Ùn jà [[nú ɖù] gbé] 1sg arrive nú eat for ‘I arrived in order to eat.’

(F) (Anonymous 1983: IX, 4)

The second one involves the conjunction bó, illustrated in (87). As is extensively discussed in Lefebvre (2004), Fongbe has two clausal conjunctions: b which conjoins two clauses that have different subjects, and bó which conjoins two clauses that have coreferential subjects. In (87), the conjunct clause containing bó has thus no overt subject. The conjunction bó and the future marker ná that follows it are interpreted together as meaning ‘in order to’. (87)

Ùn ná yì bó ná wà àzɔ́ (F) 1 . go  . do work ‘I will go in order to work.’ (Lefebvre and Brousseau 2002: 174)

Can the discrepancy observed between P pa and Fongbe nú be attributed to differences between the substratum languages with respect to corresponding lexical items? This possibility cannot be rejected a priori for, as was shown in Lefebvre (1998b), there

Chapter 11. On the properties of Papiamentu pa 

are cases of variation between the substratum languages of a given creole that are reflected in the creole. In this case, however, it appears that the availability of [−T] purposive clauses in the creole cannot be attributed to the availability of such purposive clauses in other substratum languages of P, because, as was pointed out by Koopman (1986: 244), in West African languages in general, infinitival complements do not occur with prepositions (nor with Ns or As either). Can the discrepancy observed between P pa and its corresponding substratum languages be attributed to the properties of Portuguese para? It is possible, since Portuguese para may select infinitival clauses of the standard type, as was shown in (52). However, there is also the possibility that the creators of P did not perceived the infinitival character of the complement clause of Portuguese para. In this case, they would have had to use their own lexicons and grammars as the source of the infinitival complement of P pa. The two purposive constructions discussed above may well have served this purpose in the following way. The Fongbe postposition gbé ‘for’ does not seem to have been reproduced as such in the creole. Rather, it appears that the preposition pa has taken over the functions of the postposition gbé, also meaning ‘for’, of selecting [−T] complements. In this view, the Fongbe postpositional structure in (86) would be the first source for the P [−T] clauses occurring as the complement of pa. A second source might very well be the structure in (87) involving the bó ná construction. As is extensively discussed in Lefebvre (2004), Fongbe bó has not been relexified, and it could not have been, because there was no appropriate form in the superstratum language to relabel it. It appears, however, that the P preposition pa has taken over the function of bó ná in structures such as those in (87). This would have involved the reanalysis of the structure in (88)a as a structure of the type in (88)b. (88)

a.

NPi

b.

NPi

. . .bói proi ná wà àzɔ́  . do work ‘. . . in order to do the work’ . . . pa PROi VP

(F)

(P)

Thus, as a result of the relabelling of Fongbe nú on the basis of Portuguese para, P pa would only select [+T] clauses. Its selectional properties will have been extended so as to included [−T] clauses, thus creating a P counterpart to the Fongbe structures in (86) and (87).

.

A puzzling syntactic difference between P pa and Fongbe ní in their function as mood markers

In Section 2.2, we saw that both P pa and Fongbe ní may occur as mood markers between the subject and the verb. There is, however, a puzzling difference between the distribution of pa and ní occurring in this function. In Fongbe clauses introduced by the tensed subjunctive complementisers nú or ní, the mood position between the subject and the verb must be filled by the mood marker ní. This was illustrated in (75)

 Claire Lefebvre and Isabelle Therrien

and (76). In contrast, in P clauses introduced by the tensed subjunctive complementiser pa, the mood position between the subject and the verb must remain empty, as in (89) (see also the examples in Section 2.3). (89)

Mi ta deseá pa e bai. 1  wish  3 go ‘I want him to go.’

(P) (Maurer 1988: 49)

Something has happened here that prevents pa from occurring between the subject and the verb in the embedded clause introduced by the complementiser pa (see Note 9). The same phenomenon is also observed in Haitian Creole (see Koopman and Lefebvre 1981, 1982; Sterlin 1989), in Saramaccan (see Lefebvre and Loranger 2006), and possibly in other Atlantic creoles. We do not have an explanation for this fact as yet. Another difference between Fongbe ní and P pa involves their surface positions. While both may surface as head of MoodP between the subject and the verb (see (19) and (68)), only ní may occupy the syntactic position preceding the subject (see (71)). Although we do not have an explanation for this fact as yet, it is possible that there be a link between the unavailability of the pre-subject position for pa and its nonoccurrence as conditional clause introducer.

.

Mood markers as conditional clause introducers

Recall from Section 2.2 that Fongbe nú and ní may occur as epistemic mood markers in conditional clauses, as in (73) and (74). P pa does not occur in this context. Rather, P conditional clauses are introduced by si or ku as is shown in (90) and (91), respectively. (90)

(91)

Si awaseru ta kai, nos no por bai kas if rain  fall 1  can go house ‘If it is raining, we cannot go home.’

(P) (Maurer 1988: 232)

Ku bo’ n ta tende, mi ta kita bo garganta afó (P) if 2   obey 1  remove 2 throat out ‘If you don’t obey, I will cut your throat.’ (Maurer 1988: 235)

Whether this discrepancy between P pa and Fongbe ní/nú is linked to the non-availability of the pre-subject position for pa will have to await future research.

.

Pp pa and Fongbe nú as case markers

In Section 3.2, we saw that both P pa and Fongbe nú may have the function of case marker. However, as case markers, P pa and Fongbe nú do not have the same distribution. While P pa may appear with a possessor and an agent (see Section 2.5), Fongbe nú does not occur in this context (see Section 3.2). In Fongbe, possession is expressed

Chapter 11. On the properties of Papiamentu pa 

by means of two case markers: objective sín ‘of ’, and genitive tɔ̀n ‘‘s’ as per the analysis of Brousseau and Lumsden (1992). These two cases markers are illustrated in (92). (92)

a.

b.

f ɔ̀tóò K ɔ̀kú tɔ̀n lɛ́ picture Koku   ‘Koku’s picture’ K ɔ̀kú sín f ɔ̀tóò lɛ́ Koku  picture  ‘pictures of Koku’

(F) (Brousseau and Lumsden 1992) (F) (Brousseau and Lumsden 1992)

Likewise, in contrast to P pa, Fongbe nú does not select an agent. The same two case markers as above fulfil this function, as is illustrated in (93).19 (93)

a.

b.

yòvó ɔ́ sín wìwε´n lɛ́ stranger   message  ‘messages by the stranger’

ɔ́ tɔ̀n lɛ́ wìwε´n yòvó message stranger    ‘messages by the stranger’

(F) (Brousseau and Lumsden 1992) (F) (Brousseau and Lumsden 1992)

As we saw in Section 4.1, case markers cannot be relabelled because they lack a semantic representation. They may, however, be assigned a phonological representation in the creole by grammaticalisation (e.g., the grammaticalisation of a preposition as a case marker). By hypothesis, sín and tɔ̀n will have been assigned a null form at relabelling. Could it be that P pa has become a case marker through grammaticalisation, in some of the contexts corresponding to those where sín and tɔ̀n occur? We leave to future research the detailed implementation of this proposal. In Section 3.2, we saw that Fongbe nú has the function of a case marker in the context of the complement of verbs such as ‘to fear’ (see (83)). In the same context, P has the form di (from Portuguese de) instead of pa (see (84)), showing that, in this case, the superstrate lexical entry has been adopted.

.

Summary

Several of the mismatches between the Fongbe and the P data concern the ‘division of labour’ between the lexical entries involved. The ‘for this (reason)’ discrepancy (Section 5.1) is an example in point, as it involves a difference in the division of labour between the preposition meaning ‘for’ and a postposition meaning ‘cause’ in the P . As per the analysis in Brousseau and Lumsden (1992), the two case markers sín and t ɔ̀n may also be interpreted as ‘from’ or ‘to’ (e.g. ‘message from/to/made by the stranger’). . The principles defining the choice between -nù and -tɔ́ are not known. Whether the preposition nú and the affix -nù are semantically related, is a question for further research.

 Claire Lefebvre and Isabelle Therrien

substratum language. The discrepancy in finiteness property of the purposive clause introduced by pa and corresponding lexical items in the P substratum language (Section 5.2) also involves a difference in the division of labour between the substratum language lexical items involved. In this case, in addition to the preposition meaning ‘for’, a postposition meaning ‘for’ was also shown to be involved. Likewise, the distinctive distributional properties of P pa and Fongbe nú as case markers were shown to involve a difference in the division of labour between the various lexical entries involved. The reanalysis of substratum structures has been hypothesised to play a role in the make up of the P purposive infinitival clauses (Section 5.2). The grammaticalisation of pa as a case marker was also hypothesised to have played a role in the make up of the properties of pa as a case marker (Section 5.5). The incorporation of a superstratum form, in this case di, into the creole lexicon was also shown to be involved. The syntactic difference observed between P pa and Fongbe ní in their function as mood markers in Section 5.3 and 5.4 is of a different kind, and a sound explanation of this difference is awaiting further research.

.

Conclusion

The first objective of this paper was to provide a detailed description of the properties of pa for each of its functions: preposition, mood marker, complementiser, and case marker. It was shown that the multifunctional character of pa may be captured in a simple way if pa is allowed to head several syntactic projections: PP, ForP, MoodP, and KP. The second objective of the paper was to trace the origin of the properties of pa. This was achieved by means of a three-way comparison of the properties of pa with those of the closest lexical items in the source languages of P. This comparison has led to the following conclusion: while the form of the P lexical item pa is most probably derived from Portuguese para through syllable truncation, most of its semantic and syntactic properties are derived from those of corresponding lexical items in the P substratum languages. A relabelling account of the data was proposed. In this analysis, the properties of pa would result from the relabelling of two substratum lexical entries, e.g., Fongbe nú and ní, on the basis of a single superstratum form. This explains why P pa cumulates the functions of the two closest substratum lexical entries. Unlike other cases of relexification, where a (near to) perfect match is found between the properties of a creole lexical entry and those of the corresponding substratum ones (see e.g., Lefebvre 1998b; Lefebvre and Loranger in press), the case of P pa does not offer as a perfect match, as was shown in Section 4. Why does this situation obtain? In our view, this follows from several factors. The first one is the possibility for more than one substrate lexical item to be relabelled on the basis of the same superstratum form (see Lefebvre 1998b for a discussion of several such cases). The fact that P pa cumulates the functions of both Fongbe nú and ní was hypothesised to follow from their having been relabelled on the basis of the same superstratum form pa(ra).

Chapter 11. On the properties of Papiamentu pa 

The fact that P pa also appears to have undertaken the labour performed by the postpositions wútú and gbé ‘for’ may also be attributed to the relabelling of these lexical items on the basis of pa(ra). A second factor is the availability of other cognitive processes involved in linguistic change, among which reanalysis and grammaticalisation. These two processes were hypothesised to have played a role in the make up of the properties of P pa. In the scenario of creole genesis argued for in Lefebvre (1998b), these other processes occur after relabelling, that is on the output of the relexified lexicons. By definition, these processes modify the properties of the original lexical entries. A third factor is the availability of levelling in the development of a creole. This process reduces the variation that there is between dialects of the same language (Trudgill 1986). Lumsden and Lefebvre (1994) hypothesise that in a creole’s development, levelling applies in the following way. Relexification of several lexicons on the basis of a single (or main) language provides the members of an early community with a common vocabulary. However, since relabelling is an individual process, the variation between the substratum lexicons will be reproduced in the creole through relabelling, such that a creole’s lexical entry may not have exactly all the same features for all speakers of the early creole. When speakers quit targeting the superstratum language and start targeting the relexified lexicons, they may reduce the variation created by relabelling through levelling (for a discussion of the interaction between relabelling and levelling, see Lefebvre 1998b Chapter 2, 84–7, 137–9, Lefebvre 2004: 27, Chapter 9; Lefebvre and Therrien in press, etc.). Again the process may modify the properties of the lexical entries produced by relexification. A fourth factor is the possibility of adopting a superstratum lexical entry. For example, we mentioned that the Portuguese lexical entry de was incorporated in P as di. Finally, as is the case in other languages, there is the possibility of internal linguistic change applying to any lexical entry, such that three hundred years later, the early creole lexical entries may have undergone some changes. One of the reviewers claims that the data presented in this paper seem to support a restructuring view of creolisation rather than a relexification account of it. In our view, adopted from Lefebvre (1998b), the data do support a relexification account of creole genesis. As stated above, however, other processes may apply on the relexified lexicons with the consequence that three hundred years later, the properties of some creole lexical entries may depart from those of the relexified/relabelled ones. The availability of all the processes referred to above shows the complexity of the historical derivation of a creole’s lexical entry.

References Andersen, R.W. 1990. Papiamentu tense-aspect, with special attention to discourse. In Singler, J.V. (ed.). Pidgin and creole tense-mood-aspect systems. Amsterdam: John Benjamins. 59–96. Anonymous 1983. Éléments de recherche sur la langue fon. Cotonou: n.p.

 Claire Lefebvre and Isabelle Therrien Avolonto, A. 1992. De l’étude sémantico-syntaxique des marqueurs préverbaux à la structure de la phrase en f ɔ ̀ngbè. MA thesis, Université du Québec à Montréal, Montreal. Birmingham, J.C. 1970. The Papiamentu language of Curaçao. PhD thesis, Virginia, Université de Michigan. Boncompain Grelier, M. 1989. Portugais, l’essentiel de la grammaire. Casteilla: Paris. Boretzky, N. 1983. Kreolsprachen, Substrate und Sprachwandel. Wiesbaden: Harrassowitz. Bouchard, D. 1995. The semantics of syntax. Chicago: Chicago Press. Brousseau, A.-M. and Lumsden, J.S. 1992. Nominal structures in Fongbe. In Lefebvre, C. (ed.). Topics on the syntax and semantics of Fongbe. Special issue of the Journal of West African Languages 22.1. 5–26. Cantel, R. 1999. Précis de grammaire portugaise. Paris: Vuibert. Chomsky, N. 1981. Lectures on government and binding. New York: Mouton de Gruyter. Cowper, E.A. 1989. Thematic underspecification: The case of have. Paper read at the Canadian Linguistics Association Annual Meeting, Université Laval, Quebec City. Cowper, E.A. 1995. English participle construction. Revue canadienne de linguistique 40. 1–38. Da Cruz, M. 1994. Les constructions sérielles du fongbè: Approche sémantique et syntaxique. In Lefebvre, C. (ed.). Travaux de recherche sur le créole haïtien 20–21. Montréal: Université du Québec à Montréal. 1–154. Dijkhoff, M. 1993. Papiamentu word formation. PhD thesis, University of Amsterdam. Dunn, J. 1928. A grammar of the Portuguese language. Washington, DC: National Capital Press. Emonds, J.E. 1985. A unified theory of syntactic categories. Dordrecht: Foris Publications. Farkas, D.F. 1992. On the semantics of subjunctive complements. In Hirschbühler, P. and Koerner, K. (eds). Romance languages and modern linguistic theory: Papers from the 2nd linguistic symposium on romance languages (LSRLXX), Ottawa. Amsterdam: John Benjamins. 127–51. Fouse, G.C. 2002. The story of Papiamentu: A study in slavery and language. Lanham/Oxford: University Press of America. Ghomeshi, J. and Massam, D. 1994. Lexical/syntactic relations without projection. Linguistic Analysis 24. 175–217. Goodman, M. 1987. The Portuguese element in the American Creoles. In Gilbert, G.G. (ed.). Pidgin and creole languages. Honolulu: University of Hawaii Press. 361–405. Haegeman, L. 1998. Introduction to Government and Binding Theory. Oxford: Blackwell Publishers. Hinds, J. 1973. Case markers and complementizers: Korean and Japanese. Stanford University Working Papers on Language Universals 13. 93–6. Holm, J. 1988. Pidgins and creoles. Vol. 1: Theory and structure. Cambridge: Cambridge University Press. Johns, A. 1992. Deriving ergativity. Linguistic Inquiry 17. 587–622. Jongh, E.A. de 1970. E dia mas históriko [: 30 di mei 1969] 3rd ed. Curaçao: E.A. de Jongh Kinyalolo, K.K.W. 1993. On some syntactic properties of ɖɔ̀ in Fon. Lingua 91. 201–33. Kleinmoedig-Eustatia, A. 1981. Mi koto di kuenta 1. Curaçao: Sekshon di Kultura. Koopman, H. 1986. The genesis of Haitian: Implications of a comparison of some features of the syntax of Haitian, French and West African Languages. In Muysken, P. and Smith, N. (eds). Substrata versus universals in creole genesis. Papers from the Amsterdam creole workshop, April 1985. Amsterdam: John Benjamins. 231–58. Koopman, H. and Lefebvre, C. 1981. Haitian Creole pu. In Muysken, P. (ed.). Generative studies in creole syntax. Dordrecht: Foris Publications. 201–21.

Chapter 11. On the properties of Papiamentu pa  Koopman, H. and Lefebvre, C. 1982. PU: Marqueur de mode, préposition et complémenteur. In Lefebvre, C., Magloire-Holly, H. and Piou, N. (eds). Syntaxe de l’haïtien. Ann Arbor, MI: Karoma. 64–91. Kouwenberg, S. 1990. Complementizer PA, the finiteness of its complements, and some remarks on empty categories in Papiamento. Journal of Pidgin and Creole Languages 5.1. 39–51. Kouwenberg, S. and Lefebvre, C. (in press). A new analysis of the Papiamentu clause structure. Probus. Kouwenberg S. and Michel, A. forthcoming. Papiamentu. Kouwenberg, S. and Murray, E. 1994. Papiamentu. Munich: Lincom Europa. Kouwenberg S. and Muysken, P. 1995. Papiamento. In Arends, J., Muysken, P. and Smith, N. (eds). Pidgins and creoles: An introduction. Amsterdam: John Benjamins. 205–18. Kroon, W. 1979. Giambo bieuw a bolbe naa wea: Novela intima Curazoleña (3rd ed.). Curaçao: Servicio di Enseñansa i Kultura/Gobiernu Insular. Lefebvre, C. 1992. AGR in languages without person and number agreement: The case of the clausal determiner in Haitian and Fon. In Lefebvre, C., Lumsden, J.S. and Travis, L. (eds). Functional categories, special issue of The Canadian Journal of Linguistics 37.2. Toronto: Canadian Linguistic Association. 137–56. Lefebvre, C. 1996. The tense, mood and aspect system of Haitian Creole and the problem of transmission of grammar in creole genesis. Journal of Pidgin and Creole Languages 11.2. 231–313. Lefebvre, C. 1998a. Multifunctionality and variation among grammars: The case of the determiner in Haitian and in Fongbe. Journal of Pidgin and Creole Languages 13.1. 93–150. Lefebvre, C. 1998b. Creole genesis and the acquisition of grammar: The case of Haitian Creole. Cambridge: Cambridge University Press. Lefebvre, C. 1999. Substratum semantics in the verbal lexicon of Haitian creole. Studies in Language 23.1. 61–103. Lefebvre, C. 2004. Coordinating constructions in Fongbe with reference to Haitian Creole. In Haspelmath, M. (ed.). Coordinating constructions. Amsterdam: John Benjamins. 123–64. Lefebvre, C. and Brousseau, A.-M. 2002. A grammar of Fongbe. Berlin: Mouton de Gruyter. Lefebvre, C. and Loranger, V. 2006. On the properties of Saramaccan fu: Synchronic and diachronic perspectives. Journal of Pidgin and Creole Languages 21.2. 275–335. Lefebvre, C. and Loranger, V. (in press). A diachronic and synchronic account of the multifunctionality of Saramaccan táa. Linguistics. Lefebvre, C. and Lumsden, J.S. 1994. Le rôle central de la relexification dans la genèse des langues créoles. Plurilinguisme 8. 47–93. Lefebvre, C. and Massam, D. 1988. Haitian Creole syntax: A case for DET as head. Journal of Pidgin and Creole Languages 3.2. 213–43. Lefebvre, C. and Therrien, I. (in press). On Papiamentu ku. In Siegel, J., Eades, D. and Lynch, J. (eds). Language description, history and development: Linguistic indulgence in memory of Terry Crowley. Amsterdam: John Benjamins. Lenz, R. 1928. El papiamento, la lingua criolla de Curazao (la gramática más sencilla). Santaigo de Chile: Imprenta Balcells. Lumsden, J.S. and Lefebvre, C. 1994. The genesis of Haitian Creole. In Lefebvre, C. and Lumsden, J.S. (eds). The central role of relexification in creole genesis: The case of Haitian Creole., 15 pages. Research report prepared for SSHRCC on the project La genèse du créole haïtien: un cas particulier d’investigation sur la forme de la grammaire universelle, Université du Québec à Montréal, Montreal.

 Claire Lefebvre and Isabelle Therrien Maurer, P. 1986. Le papiamento de Curacao: Un cas de creolisation atypique? Etudes créoles 9.1. 97–113. Maurer, P. 1988. Les modifications temporelles et modales du verbe dans le papiamento de Curaçao (Antilles Néerlandaises). Buske: Hamburg. Maurer, P. 1993. Subjunctive mood in Papiamentu. In Byrne, F. and Holm, J. (eds). Atlantic meets Pacific: A global view of pidginization and creolization. Amsterdam: John Benjamins. 243–50. McWhorter, J.H. 1995. The scarcity of Spanish-based creoles explained. Language in Society 24. 213–44. McWhorter, J.H. 1997. Towards a new model of creole genesis. New York: Peter Lang. Mous, M. 1995. Language intertwining. Paper read at the Amsterdam Creole Workshop: Creole genesis and language contact, Amsterdam, 1995. Mufwene, S.S. and Dijkoff, M. 1989. On the so-called ‘infinitive’ in Atlantic creoles. Lingua 77. 297–330. Muysken, P. 1977. Movement rules in Papiamentu. Amsterdam Creole Studies 1. 80–102. Muysken, P. 1981. Half-way between Quechua and Spanish. The case for relexification. In Highfield, A.R. and Valdman, A. (eds). Historicity and variation in creole studies. Ann Arbor: Karoma. 52–79. Navarro Tomás, T. 1953. Observaciones sobre el Papiamento. Nueva revista de filologia hispanicá 7. 183–89. Nida, E. 1948. The identification of morphemes. Language 24. 414–41. Parkvall, M. 2000. Out of Africa: African influences in Atlantic Creoles. London: Battlebridge. Postma, J. 1990. The Dutch in the Atlantic slave trade, 1600–1815. Cambridge: Cambridge University Press. Rassinoux, J. 1987. Dictionnaire français-fon. Cotonou: Imprimerie Dumas. Rassinoux, J. 2000. Dictionnaire français-fon. Cotonou: Société des Missions Africaines. Rizzi, L. 1997. The fine structure of the left periphery. In Haegeman, L. (ed.). Elements of grammar: A handbook in generative syntax. Dordrecht: Kluwer Academic publishers. 281–337. Ruhl, C. 1989. On monosemy: A study in linguistic semantics. Albany, NY: State University of New York Press. Segurola, B. and Rassinoux, J. 2000. Dictionnaire fon-français. Cotonou: Société des missions africaines. Singler, J.V. 1996. Theories of creole genesis, sociohistorical considerations, and the evaluation of evidence: The case of Haitian creole and the relexification hypothesis. Journal of Pidgin and Creole languages 11. 185–231. Sterlin, M.-D. 1989. Les caractéristiques de pou: Un modal en position de complémenteur. In Lefebvre, C. and Lumsden, J.S. (eds). Le créole haïtien. Special issue of the Revue Québécoise de Linguistique 18.2. Montréal: Université du Québec à Montréal. 131–47. Therrien, I. 2004. La multifunctionnalité de ku et pa en papiamento. MA thesis, Université du Québec à Montréal, Montreal. Tossa, C.Z. 1994. Adjonctions et séries verbales dans les langues gbé. PhD thesis, Ottawa University, Ottawa. Travis, L. and Lamontagne, G. 1992. The case filter and licensing of empty K. In Lefebvre, C., Lumsden, J.S. and Travis, L. (eds). Functional categories. Special issue of The Canadian Journal of Linguistics 37.2. Toronto: Canadian Linguistic Association. 157–74. Trudgill, P. 1986. Dialects in contact [Language in Society 10]. Oxford: Basil Blackwell.

Chapter 11. On the properties of Papiamentu pa  Van Wijk, H.L.A. 1958. Origines y evolucion del papiamentu. Neophilologus. 42. 169–82. Voorhoeve, J. 1964. Creole languages and communication. In Symposium on multilingualism. London: Committee for Technical Cooperation in Africa. 233–42. Washabaugh, W. 1975. On the development of complementizers in creolization. Working Papers on Language Universals 17. 109–38. Winford, D. 1993. Predication in Caribbean English Creoles [Creole Language Library 10]. Amsterdam: John Benjamins.

 

No exception to the rule The tense-modality-aspect system of Papiamentu reconsidered* Nicholas Faraclas, Yolanda Rivera-Castillo and Don E. Walicek University of Puerto Rico, Río Piedras

In this chapter, it is demonstrated that although the Tense, Modality, and Aspect (TMA) system of Papiamentu has been cited by a number of researchers (Andersen 1993; Bickerton 1980, 1981) both as being exceptional in relation to other Creoles of the Caribbean and as being deviant from universal strategies for marking TMA attributed by some to Creole languages worldwide, Papiamentu TMA operates essentially on the basis of the same system found in most Atlantic Creoles as well as in most of their West African substrate languages. All of the features which Andersen (1993: 89–91) and others cite as ‘aberrant’ in Papiamentu, including: (1) the near obligatory use of the markers a or ta before verbs; (2) the absence of a ‘Ø marker’ for perfective aspect; (3) the existence of two irrealis markers (lo and Ø) and the ‘deviant position’ of lo; (4) the dual (tense and aspect) function of the marker tabata; and (5) the lack of a specific morpheme that functions exclusively as an anterior marker; are all shown to be the result of features and patterns of grammatical change found throughout the Afro-Atlantic.

.

Introduction

The Tense, Modality, and Aspect (TMA) system of Papiamentu has been cited by a number of researchers, including Andersen (1993) and Bickerton (1980, 1981), both as being exceptional in relation to other Creoles of the Caribbean and as being deviant from universal strategies for marking TMA attributed by some to Creole * We extend our sincere appreciation to those in Aruba who made this research possible, especially those individuals who agreed to have their conversations recorded. We thank the University of Alabama, Tuscaloosa for sponsoring Rivera-Castillo’s data collection there through its Summer Research Grant (2000) program. The University of Puerto Rico at Río Piedras provided additional support for this research through its Fondo Institucional para la Investigación (FIPI), award 8-80-502. We would also like to thank Rose Vázquez for her assistance in the transcription and translation of interviews.

 Nicholas Faraclas, Yolanda Rivera-Castillo and Don E. Walicek

languages worldwide. Andersen’s (1993) extensive database and insightful analysis confirm Maurer’s (1993) assertion that the feature [–realis] is normally realized as Ø (i.e., by the absence of either a or ta before a verb). He then goes on to show very convincingly that the markers a and ta are perfective (non-durative) and imperfective (durative) aspectual markers respectively, rather than tense markers as previously supposed, and that the presence or absence of these markers is to some extent sensitive to whether the verb that follows them is stative or not. By focusing on basilectal speech samples and steering clear of what he calls a ‘rigid European latinate tradition’ of interpretation, Andersen thus succeeds in demonstrating that TMA marking in Papiamentu (hereafter PA) does not differ as much from that which might be predicted by the Language Bioprogram Hypothesis (LBH) than Bickerton himself had conceded. Nevertheless, he concludes that significant aspects of the grammar of TMA in PA remain highly idiosyncratic. According to Andersen (1993: 89–91), a number of features distinguish the PA TMA system from those of other Creoles as well as from that proposed in Bickerton’s LBH. These include the following five characteristics: 1. 2. 3. 4. 5.

the near obligatory use of the markers a or ta before verbs; the absence of a ‘Ø marker’ for perfective aspect;1 the existence of two irrealis markers (lo and Ø) and the ‘deviant position’ of lo; the dual (tense and aspect) function of the marker tabata; and the lack of a specific morpheme that functions exclusively as an anterior marker.

With mainly data from other Caribbean Creoles and the hypothetical constructs of the LBH to use as points of reference, Andersen appears to be at somewhat of a loss to account for how a set of TMA phenomena that in some contexts may appear to be unique to PA could have developed. In this study, we attempt to demystify some of the questions surrounding TMA in PA. We suggest that this can be done by building upon the solid foundation of Andersen’s research and by supplying additional points of reference, including some which Maurer hinted at in his work, namely the languages of the coast of West Africa. Thus we situate TMA features in PA alongside a number of the characteristics of TMA systems found in West African languages. The characteristics described in this chapter as typical of West African TMA come from the systematic description of TMA in Nigerian Pidgin (hereafter NP) and its substrate languages found in Faraclas (1987, 1988, 1990, 1996). In this work, the term ‘TMA system’ will be used in its broadest sense to encompass both TMA auxiliaries as well as other means used to signal TMA-like meanings.

.

Methodological questions: What constitutes a ‘substrate sample’?

Both typologically and genetically, the languages of the south of Nigeria are particularly interesting for the study of African substrate influences on PA, NP, and the other . Andersen (1993) terms the presence of an overt marker as the lack of a ø marker for the perfective aspect.

Chapter 12. No exception to the rule 

Atlantic Creoles. These varieties emerged from the maritime and agro-industrial political economies made possible by the massive enclosure and alienation of indigenous American, African, and European land and labor during the colonial era and the ensuing contact that occurred among indigenous American, African, and European cultures. We contend that claims which purport to link the Afro Caribbean populations of given islands to a single area of origin along the west coast of Africa at any given time in the history of the slave trade are at best extremely tenuous. Such assertions are complicated by factors such as the incomplete and inaccurate nature of the documentation of the legal trade of slaves, the pervasiveness of illegal slave trading, the massive movements of undocumented populations from one part of the West African coast to another, and complex patterns of migration and cultural contact within the Caribbean. Therefore, alongside the need for studies of particular West African languages or geographically or genetically restricted groups of West African languages, there is also a need for studies of African substrate languages that consider West African languages on the whole from an areal/typological perspective. From this viewpoint, Southern Nigeria is the area that best represents a balanced sample of the languages of the west coast of Africa from the east of the Ivory Coast, through Ghana, Togo, Benin, Nigeria, Cameroon, Equatorial Guinea, Gabon, the Congo, Angola, Namibia, and South Africa. Among contemporary Africanists, there is general consensus that all of the present day Bantu languages are the result of a series of relatively recent waves of migration over the past few thousand years from the south of what is now Nigeria and Cameroon and therefore Bantu constitutes a sub-branch of the Benue-Congo branch of the Niger Congo language family (Hinnebusch 1989: 454). Creolists seem to have been slow to recognize the extremely close genetic and typological relationships both among the Bantu languages themselves and especially between the Bantu languages and southern Nigerian Benue-Congo languages such as Efik/Ibibio of the Cross River sub-branch, Igbo of the Igboid sub-branch, Edo/Bini of the Edoid sub-branch and Yoruba of the Defoid/Yoruboid sub-branch (of Benue-Congo, Williamson 1989b: 261). Another common misunderstanding among creolists concerns changes in the classification of African languages that took place during the last quarter of the twentieth century. Since the 1980s, the Nigerian Defoid/Yoruboid, Igboid, and Edoid subbranches have been removed from the Kwa branch of Niger-Congo and reassigned to the Benue-Congo branch (Stewart 1989: 217; Williamson 1989b: 269–72). Despite the fact that this reclassification is over two decades old, creolists unfortunately still frequently speak of Yoruba, Igbo, and Edo as ‘Kwa languages’. Even more creolists are unaware that there is little or no evidence of a distinction between the Kwa branch (including ‘Akan’, Ewe, Twi, Fante, Asante, Guang, Ga, Gbe/Fongbe and others) and the Benue-Congo branch of Niger-Congo in the first place. This realization has lead many Africanists to abandon such a division altogether in favor of a combined Benue-Kwa ‘macro-branch’, whose homeland would be somewhere in southern Nigeria, just southwest of the Benue-Niger confluence (Williamson 1989a: 17–8; 1989b: 269–72).Thus, creolists have grossly underestimated the genetic and typological connections among

 Nicholas Faraclas, Yolanda Rivera-Castillo and Don E. Walicek

‘key’ West African substrate languages such as ‘Akan’, Twi, Gbe/Fongbe, Yoruba, Edo, Igbo, Efik/Ibibio, Kikongo, and Mbundu. All of these languages not only belong to the same language family (i.e., Niger-Congo), but also to the same ‘macro-branch’ of that family (i.e., Benue-Kwa). In addition, most belong to an even more closely related grouping, the ‘micro-branch’ Benue-Congo). The sample of the languages of Southern Nigeria used in the present study includes both some of the major languages spoken by those Africans who were taken as slaves to the Americas (e.g., Yoruba, Igbo, Edo, Efik/Ibibio, Ijo) as well as languages which have close genetic and typological relationships with the ‘Kwa’ languages spoken from Benin to the Ivory Coast to the west, and the Bantu and Bantoid languages spoken from Cameroon to South Africa to the east. Since the features of the TMA systems of these languages have already been systematically compared to those of NP, and because NP is one of the Atlantic Creoles which has remained in close contact with these languages, we have found it practical to use it as a point of reference from which to begin this inquiry. Given the fact that the TMA system of NP is so similar to those found in the substrate sample languages, we generally use the TMA system of NP to exemplify the TMA systems found in the substrate sample. How does Papiamentu fit in? By considering the TMA system of PA on the one hand and the TMA systems which typify NP and a broad sample of languages from southern Nigeria (i.e., the Benue-Congo/Benue-Kwa languages spoken along the coast) on the other hand, the ‘aberrant’ features found in PA can be seen as logical and predictable consequences of prominent patterns and processes found in the linguistic traditions and practices of the African populations who were responsible for the creation and the subsequent elaboration of TMA marking in PA and other Atlantic Creoles. NP and the languages of southern Nigeria share with both PA and the TMA model proposed in the LBH the relative prominence of aspectual marking over tense marking. But pervasive marking for realis modality distinguishes NP and the languages of southern Nigeria from predictions made in the LBH. Below we will show how PA can be seen as actually patterning with NP and the languages of southern Nigeria in as far as it comprehensively marks realis modality. In the languages of West Africa, NP, and a number of other Atlantic Creoles, realis markers such as clause final o and reflexes of the verb ‘to come’ are extremely widespread (Singler 1985). Realis modality is also linked, albeit indirectly, by default to markers for both completive (perfective) and incompletive (imperfective) aspect in NP and the languages of southern Nigeria. If we postulate both an aspectual and a realis modality marking function for a and ta in PA and a scenario whereby compulsory realis marking spreads progressively from nonstative to stative verbs, we can easily explain three phenomena: 1. 2.

that a and ta are obligatory on all realis non-stative verbs (thus eclipsing any Ø marking for perfective aspect) in non-focused environments; that where a and ta are not present (Ø) an irrealis reading has become the norm;

Chapter 12. No exception to the rule 

3.

.

that this innovative form of realis marking has exerted structural pressure on other markers such as the irrealis markers bai (not mentioned in Andersen 1993) and lo and a hypothesized anterior marker taba.2

Data samples

Regarding spoken data utilized in this work, the features and constructions identified as typical of NP are those found in approximately 30 one hour recordings of spontaneous speech collected to form a representative sample of a NP speaking community.3 Similar concerns were taken into consideration in the collection of data on PA. Approximately fifteen speakers of PA from the island of Aruba supplied samples of spontaneous speech.4 Additional data come from secondary sources such as Maurer (1993). Archival materials are used for diachronic comparisons. Following the substrate/areal model proposed by Gilman (1986); Faraclas (1988, 1990, 1996) and McWhorter (2000), the substrate sample was collected on the basis of detailed surveys of those languages of southern Nigeria for which comprehensive grammars were available at the time of our research.5 The abbreviations shown in Table 1 will be used to identify the indicated studies of the West African languages which were used in the substrate data sample. To ensure maximum comparability and minimum bias, all of the data in the West African substrate sample have been analyzed using Comrie and Smith’s Lingua Descriptive Questionnaire (1977). The results shown in Table 2 suggest that in one TMA subsystem after another, the grammar and diachronic development of NP cannot be adequately described or analyzed without extensive reference to its substrates. The analysis included in this work reconfirms this conclusion for PA and proposes that the same may hold for many other Atlantic Creoles.

. All of these markers play a less typical and more marginalized role in PA in comparison to their counterparts in the other Creoles of the Caribbean and beyond. It is tempting to see tabata as a compound but we do not have synchronic evidence of this. Bickerton’s LBH includes an anterior marker, a function that taba would have had according to Andersen. . An attempt was made to collect a representative sample of the NP-speaking community of Port Harcourt, Rivers State, Nigeria. NP speakers were selected on the basis of ethnolinguistic background, sex, age, amount of formal education, whether NP was learned as a first or second language, and the extent to which NP was used by each in day to day interactions. . One native speaker of PA was also consulted about some aspects of the analysis and the interpretation that are presented in this paper. . Some elements of the analysis of NP and its substrates found below were taken from Faraclas (1987 and 1990) where the results of a systematic comparison of NP and its substrate languages can be found.

 Nicholas Faraclas, Yolanda Rivera-Castillo and Don E. Walicek

Table 1. Studies of West African substrate languages West African Substrates: Yoruboid (YD): Yo=Yoruba (Bamgbose 1966) Edoid (ED): En=Engenni (Thomas 1978) Ijoid (JD): Central (CT): Ko=Kolokuma (Williamson 1965) Eastern (EA): Ka=Kalabari (Jenewari 1977) Igboid (ID): Iz=Izi (Meier et al. 1975); Ig=Igbo (Emmenanjo 1978, 1985); Ek=Ekpeye (Clark 1969) Cross River (CR): Lower Cross (LC): Ob=Obolo (Faraclas 1984); IE=Ibibio/Efik (Kaufman 1969); Or=Oro (Kuperus 1978) Upper Cross (UC): Mb=Mbembe (Barnwell 1969) Bendi (BN): Be=Bekwara (Stanford 1967)

Table 2. Tense, aspect, and modality in NP, West African substrate languages, and Spanish (Sp, one of the Iberian lexifier languages of Papiamentu; Faraclas 1990) YD ED FEATURE Tense: Tense not prominent Future most prominent Ø present marking Factative tense Link future-irrealis Aspect: Aspect prominent [+/– completive] ‘finish’ [+completive] Locative copula [–completive] Factative aspect Modality: Modal verbs common ‘leave’ or ‘make’ subjunctive Ø subject imperative [+realis] usually unmarked ‘come’ [+realis] phrase final o [+realis]

NP

Yo

En

JD CT Ko

JD EA Ka

ID

ID

ID

Iz

Ig

Ek

CR LC Ob

CR CR LC LC IE Or

CR UC Mb

CR BN Be Sp

+ + + + +

+ + + + +

+ + + ? +

+ + + ? ?

+ + + + +

+ + + + +

+ + + + +

+ + + + ?

+ + + + +

+ + + + +

+ + + ? +

+ + + + +

+ + + ? +

– – – – (+)

+ + + +

+ + + +

+ + + +

+ + + ?

+ ? + +

? + ? +

+ ? ? +

+ + + +

+ + + +

+ + + ?

+ + + +

+ ? ? +

+ + ? ?

– – – (+)

+

+

?

?

+

+

+

+

+

+

+

?

?

–

+ +

+ +

+ –

+ +

+ +

+ +

+ +

+ ?

+ –

+ +

+ +

? ?

? +

+ –

+ +

+ +

+ ?

+ +

+ +

+ +

+ ?

+ +

+ +

+ +

+ +

+ +

+ +

+ +

+ +

+ +

+ +

+ ?

+ ?

+ +

+ +

+ +

+ +

+ +

+ +

+ +

+ ?

– –

Chapter 12. No exception to the rule 

.

Comparison between the PA TMA system and the TMA system of NP and West African substrate languages

As shown in Table 3, there are some remarkable parallels between the TMA system of PA and those of the languages of southern Nigeria, exemplified here by NP: Table 3. Similarities between the TMA systems in NP, West African substrate languages, and PA6 FEATURE ASPECT Aspect plays a predominant role in TMA system Main aspectual distinction is [±perfective/completive] Imperfective marker from copula Perfective marker from the verb ‘to finish’ MODALITY Modality plays an important role in TMA system Irrealis marked by modal verbs Realis marker for non-statives TENSE/SEQUENCE Tense plays a relatively minor role in TMA system Future most consistently marked tense Future marked by irrealis marker/modal Future from verb ‘to go’ Anterior marker from past form of copula in lexifier

NP and its substrates

PA

+ + dè fínish

+ + ta kaba

+ + +

+ + +

+ + gò gò bìn

+ + lo/ta bai (ta) bai taba(ta)

When Table 3 is compared with Table 2, PA seems to be operating with essentially the same TMA system as that found in NP and the West African substrate languages included in our sample, with some exceptions. Here we note five key areas of comparison: 1. Aspect prominence and factative aspect: While aspect plays a much more important role than does tense/sequence in NP (Faraclas 1996), the West African substrates (Faraclas 1990), and PA (Andersen 1993), a key difference between the West African and PA TMA systems is the availability of factative aspect/tense (i.e., the assignment of [+completive] aspect and [+past] tense to otherwise unmarked [–stative] verbs and the assignment of [–completive] aspect and [–past] tense to otherwise unmarked [+stative] verbs, see Examples 1 and 2 below) in West Africa and the lack of

. Realis markers for non-statives merit further attention (see last entry for modality). Regarding a realis marker from the verb ‘to come’, we know that a form similar to kom exists for NP and its substrates and speculate that bin may play a similar role for PA. Further research and archival investigation are necessary.

 Nicholas Faraclas, Yolanda Rivera-Castillo and Don E. Walicek

factative aspect (i.e., obligatory aspectual marking on [–stative] verbs in non-focused environments, i.e., outside of negative, imperative, subjunctive/irrealis, subordinate clauses, question-word questions, emphatic and focus constructions) in present day PA. 2. Imperfective (incompletive) aspect: The main aspectual distinction in NP, the West African substrate languages, and PA is [±perfective/completive] with both the West African languages and PA using a reflex of the copula (NP dè and PA ta) to mark the imperfective. While no quantitative studies yet exist to prove this point, it appears that in PA ta is used with much greater frequency than de is in NP and the West African substrates, as in PA ta is used obligatorily with [–stative] verbs in non-focused environments to signal imperfectivity. 3. Perfective (completive) aspect: NP, the West African substrate languages, and PA all use a perfective marker which is a reflex of the verb ‘to finish’ often alongside another (more frequently occurring) perfective/completive marker (dón in NP and a in PA). However, in NP, this more common perfective marker dón is used only slightly more often than is fínish, while in PA the more common perfective/completive marker a is obligatorily used with all [–stative] verbs in non-focused environments to signal the perfective, therefore a occurs very much more frequently than does kaba. 4. Modality prominence and irrealis modality: Modality distinctions are generally more prominent than tense/sequence distinctions in NP, the West African substrate languages, and in PA. The central modality distinction in West African languages generally and in PA is [±realis], with [–realis] modality signaled primarily by the use of modal verbs, but also by an auxiliary which is a reflex of ‘go’ (i.e., NP gò and historically, at least, PA bai) specifically used to mark the future.7 5. Tense, sequence, anterior, future: Both NP and PA have an anterior marker which is a reflex of a past form of the copula in the lexifier language (NP bin < English been and PA taba < Portuguese/Spanish estaba); however, in present day PA the anterior taba may only occur with the [–perfective] marker ta, yielding the fused forms tabata (also pronounced tawata, ta’ata). Similarly, the irrealis/future marker bai (a reflex of ‘go’) occurs synchronically principally with the [–perfective] marker ta, yielding ta bai (also pronounced tei), although cases of an auxiliary-like use of bai preceded by the [+perfective] marker a have been attested in contemporary PA, a phenomenon that requires more detailed study in the future.8

. Realis modality. In NP and its substrates, a form similar to kom, a realis marker which is a reflex of the verb ‘come’ is frequently used, especially with [–stative] verbs in less focused environments and in narrative texts, where the speaker wishes to insist on the fact that what s/ he is asserting is objective reality. A similar realis usage may occur in PA for bin (also a reflex of ‘come’), occurring with a much lower frequency than its counterparts in NP and its substrates. . In PA, the adverb-like form lo which appears to be undergoing a process of grammaticalization as an irrealis auxiliary is used far more frequently than is (ta) bai to mark the future.

Chapter 12. No exception to the rule 

Factative aspect/tense in West African languages and NP is illustrated by the following examples: (1)

(2)

.

[–stative] verbs ‘write’ and ‘cut’ with no overt aspect/tense markers are interpreted as [+completive] and [+past/anterior]: Obolo (Faraclas 1984): Ó-gé íkpá. you-write book ‘You wrote the book.’

NP (Faraclas 1996): Yù rayt dì buk. you write the book ‘You wrote the book.’

Ó-kép úkwá. you-cut yam ‘You cut (past) yams.’

Yù kot nyam. you cut yam ‘You cut (past) yams.’

[+stative] verbs ‘know’ and ‘have’ with no overt aspect/tense markers are interpreted as [–completive] and [–past/anterior]: Obolo (Faraclas 1984): Ó-ryòng íkpá. you-know book You know the book.’

NP (Faraclas 1996): Yù sàbi dì buk. you know the book ‘You know the book.’

Ó-káang úkwá. you-have yam ‘You have yams.’

Yù get nyam. you have yam ‘You have yams.’

Perfective vs. completive aspect

Because the distinction between perfective and completive aspect is not always clear-cut and because this distinction is not central to the general lines of argumentation put forward in this article, for convenience in this discussion we conflate our description of completives and perfectives under the cover feature [±completive], which conforms to the terminology favored by most Africanists, or under the composite feature [±completive/ perfective]. In other contexts, the perfective is considered to focus on an event as a whole (its limits) while the completive focuses on the end point of a structured event (its termination) (Salas González 1996). In the first century BC, Marcus Terentius Varro distinguished between completed and non-completed events in his description of the perfective. While more recent research has not invalidated his claims, it has shown that the perfective may also have an inchoative meaning (indicating the inception of an event) or represent an event with no indication of its beginning or end (Binnick 1991). Therefore, completives can be considered to constitute a subset of the perfective, equivalent to telic perfectives, i.e., those which refer to events terminated within a bounded interval of time. While the perfective is compatible with a wide range of tenses in some languages, in most Indo-European languages it typically entails a past interpretation (Comrie 1976: 66–71). In terms of their diachronic typology, completives often originate in dynamic verbs meaning ‘to finish’ such as PA kaba which then can become anterior and later perfective markers (Bybee 1994: 59). PA a probably originated from different sources, including

 Nicholas Faraclas, Yolanda Rivera-Castillo and Don E. Walicek

the auxiliary for the perfect in Romance (habere, ‘to have V-ed’), which also has a modal meaning that intersects with the meaning of tener (‘to have to’ – obligation). Since completives do not usually combine with stative predicates, so it seems that PA a functions to some extent as a completive, like awe in Sango (see Bybee 1994: 71 on Samarin 1967). In PA it is kaba that combines with both dynamic and stative verbs, and thus it could be said that kaba functions more properly than a as a general perfective. In Sranan, however, the form kaba has not become a general perfective marker, but instead acts as a “marker of terminative aspect with following verbs” as do reflexes of verbs meaning ‘to finish’ in most West African languages (Winford 2000: 393–4). As has become apparent in this discussion, the fluidity and ‘fuzziness’ of the boundaries between aspectual categories such as ‘perfective’ and ‘completive’ is abundantly evident in the Atlantic Creoles, their substrate languages, and their lexifiers. We therefore make only a limited attempt to differentiate between them in this work, especially since the hypotheses which we are advancing here do not crucially depend on such a distinction.

.

How the Papiamentu TMA system could have developed from a TMA System similar to that found in Nigerian Pidgin and West African substrates

We propose that both the NP and PA TMA systems began their separate but overlapping historical trajectories with TMA systems that were similar in many respects to those of their West African substrate languages. On the basis of this assumption we will offer an explanation as to how the PA TMA system came to differ: (1) from the TMA systems found in West African substrate languages (exemplified by NP); (2) from the TMA systems found in other Caribbean Creoles (again exemplified by NP); and (3) from hypothetical prototypical Creole TMA systems such as that proposed as part of Bickerton’s (1981) LBH.

.

Parallel trajectories for kóm in NP and a/ta in PA

A typical example of the use of a reflex of the verb meaning ‘to come’ as a realis auxiliary in West African languages and in the Atlantic Creoles is the following excerpt from a procedural text (on how to convince the mother of one’s girlfriend to allow her daughter to marry). It was recorded in the spontaneous NP speech of a worker at an oil refinery near Port Harcourt, Nigeria (Faraclas 1987: 54): (3)

Yù kom sìdón. . . mébi dì màma ì gò don kom kol-am tel-am you  sit maybe the mother she    call-her tell-her se, “Hu bi dis pesin nàw?” Yù gò kom rìmáynd dèm se nà dì that who be this person now you   remind themthat HL the pesin we bìn giv-am dat mòni dat de. Nà ím dèm gò kom kom. person who  give-her that money that day  it they   come (NP) ‘You sit down . . . maybe her mother will have called to her and asked her “Who is this person?” You will remind them that it is the person who gave her that money that day. So (then) they will come.’

Chapter 12. No exception to the rule 

The striking parallels between the patterns of propagation of kóm-type realis markers in NP and West African substrate languages on the one hand, and the propagation of obligatory a/ta aspectual marking in PA on the other, suggest that these markers have similar histories in their respective languages. Variational studies (Aikhionbare and Chumbow 1982; Hyman and Magaji 1970; Faraclas 1987; Marchese 1983) attest to the ongoing extension of realis markers such as kóm in NP and its substrates from [–stative] verbs and non-focused environments to [+stative] verbs and more focused environments. The term ‘focus’ is used here in the Africanist sense and applies to negatives, imperatives, subjunctive/irrealis constructions, subordinate clauses, question-word questions, as well as emphatic and focus constructions. While no such variational studies have been done thus far on TMA in PA, it is worth noting that obligatory a and ta aspectual marking in PA occurs only with [–stative] verbs in non-focused environments, with a an ta being used optionally with [+stative] verbs and in a much more restricted way in focused environments. PA texts from 1776 (Maurer 1998) and 1803 (Martinus 1994) show a different distribution for a and ta than at present. In these texts, a significant number of verbs (both [–stative] and [+stative]) occur without any overt aspectual markers, even in non-focused environments, including non-subordinate and non-subjunctive clauses. Because we are proposing that TMA in PA originally resembled that of the West African substrates more than it currently does, this is not surprising. The web matrix outlined in Diagram 1 (Foley 1986; Heine and Reh 1984) allows us to schematize the TMA systems of NP and the West African substrate languages. It shows how factative aspect and tense would be automatically assigned to these unmarked verbs based on their value for the feature [±stative] and/or from other contextual cues.9 ASPECT

COMPLETIVE dón ‘finish’ unmarked [–stative] verbs

INCOMPLETIVE dè ‘be’ unmarked [+stative] verbs

PAST/ANTERIOR bìn PRESENT unmarked [–stative] verbs

REALIS kóm ‘come’

FUTURE

IRREALIS gò ‘go’

MODALITY Diagram 1. The basic Tense, Modality, and Aspect system of NP and West African substrate languages (forms taken from NP) . Diagram 1 shows lines that link poles. The lines represent default paths, showing how subsystems interrelate. The aspectual subsystem has two poles, completive and incompletive, both

 Nicholas Faraclas, Yolanda Rivera-Castillo and Don E. Walicek

In these early documents, while ta is normally repeated before each verb in a particular series (even in cases of serialized verbs), single instances of a seem to have scope over more than one verb in a series (even when these are not part of the same serialized verb construction), as shown in the following passage from Martinus (1994: 237): (4)

[. . .]comandant Engles a baay na kaas die comandeur commander Engles / go to house the commander kibra canon die canpania y bira toer caion nan roepert, break cannon the field and become all cannons  military.term claba cañon, toema toer scopette y polber die fortie Cargaar, stick cannon take all guns and gunpower the fort Cargaar, pa deen die tera y pidie asistensie, [. . .] (PA) by interior the land and ask.for assistance ‘[. . .] Commander Engles went home, the commander broke the formation of the cannons in the field, making all the cannons (immobilized?), taking all the guns and gunpowder to Fort Cargaar, by an inland route and asking for assistance. . .’

Given the fact that we are assuming, using West African languages as our model, that ta has been used primarily as an aspectual marker from the very earliest forms of PA up until the present, it is not surprising that its scope would be limited to a single verb. The marker a is used with greatest frequency in the purely narrative portions of these early texts. Maurer (1987: 39) identifies the following sources for a in PA: (1) the auxiliary for the perfect in Romance (habere); (2) Dutch hebben; (3) the adverb já; and (4) and the morpheme a from the Angolan Bantu language Mbundu. Both habere and reflexes of verbs meaning ‘to be’ are also typical sources for anterior markers in Romance (Bybee et al. 1994: 64). But in a somewhat similar way to that found in the earlier documents cited above, in contemporary PA a marks the first verb in a series, and not each verb individually as anteriors typically do, as shown in Examples (5) and (6): (5)

El a hap, frega su wowo-nan, rèk su kurpa. (PA) He  yawn rub his eye- stretch his body ‘He/she yawned, rubbed his/her eyes, and streched.’ (Munteanu 1996: 363)

of which have markers which are used whenever the aspectual properties of a verb phrase cannot be determined by factative aspect (active verbs → completive, stative verbs → incompletive) or by other contextual cues. Similarly, the modality subsystem has two poles, realis and irrealis, only one of which (irrealis) has a marker which is frequently used. There are markers for the realis pole as well, but these are only used where the speaker wishes to stress the truth value of an assertion or wishes not to be perceived as an interested party or controlling factor in the matter under discussion. The tense subsystem is of less importance than the aspect and modality systems. The most commonly marked tense, the future, borrows its marker from the irrealis pole of the modality system. Factative tense (active verbs → past, stative verbs → present) or other contextual cues determine present and past tenses in most cases. There is a past marker which is rarely used, and there is no marker for the present tense.

Chapter 12. No exception to the rule 

(6)

.

El a baha y mira [. . .] he  lower and look ‘He/she bent and took a look at [. . .]

(PA) (Maurer 1998: 163)

The status of a as a realis marker in PA

The greater scope of a and its use in narratives is best explained if we assume that it was not always used primarily as an aspectual marker, but instead also as a modality marker like kom in NP. This would explain not only the greater scope exhibited by a in comparison with ta, but also its similar pattern of propagation through the PA system to the realis modality marker kom in NP and its substrates (i.e., from [–stative] to [+stative] verbs, from non-focused to focused environments, and from narratives to other registers).10 Additional evidence for the realis marking function of a is provided not only by the definition attributed to it by Andersen (1993: 83) as a marker of “a singular realized event,” but also by the following examples from Maurer (1993: 248): (7)

Mi a baha lus di mi lanterna pa mi Ø sende di bo. (PA) I  lower light of my lantern so I light of you ‘I have lowered the light of my lantern in order to light yours (which I have yet to do).’

(8)

Mi a baha lus di mi lanterna pa mi a sende di bo. (PA) I  lower light of my lantern so I / light of you ‘I have lowered the light of my lantern in order to light yours (which I have done).’

We propose then, that a was used as a realis marker in earlier stages of PA, spreading rapidly just as kom is presently doing in NP (Faraclas 1987, 1988, 1996) until its use became obligatory with [–stative] verbs in non-focused constructions. This would have established an opposition between any modality [–realis] marking vs. a [+realis] marking in these environments. In older texts, the use of a as a realis marker is suggested by the Ø marking in subordinate (tin in Example 9) and negative clauses (no Saby in Example 10): (9)

Samuel Costa Andrade’s declaration (16 January 1776): boste a-tende kiko tin? You /-listen what have Have you heard what is going on?

(PA) (Maurer 1998: 205)

. If a were primarily a [+perfective] marker which has always been in opposition to ta, it would presumably have a wide distribution before [+stative] verbs (which are by default [–perfective] and thus require overt marking to signal [+perfective]), but this is not the case. Izione Silva (p.c. in Andersen 1993: 93) suggests that [+perfective] was originally marked by Ø in PA.

 Nicholas Faraclas, Yolanda Rivera-Castillo and Don E. Walicek

(10)

Letter from Abraham da Costa Junior to Sara Pardo (5 October 1775): My dusie bo pay a-manda bo ruman Aronchy ku My sweet you father /-order you brother Aronchy with Tony & Maria Koge na Kamina dy piter may es nigrita Tony and Maria take  road  Pietermaai the black antunyica & nan a-ybee tras dy forty & nan Antunyica and they /-bring   fort and they a-manda sutee guatapana mas my no Saby paKy razon. (PA) ⁄-order beat guatapana but I no know for.what reason ‘My love, your father ordered your brother Aronchy and Tony and Maria to pick up black Antunyica at Pietermaai road and they brought (her) behind the fort and they ordered that she be beaten with guatapana, but I have no idea why.’ (Maurer 1998: 203)

In some cases, markers are optional in subordinate conditional clauses as well: (11)

Sy bo sabi manda gabla Ku my dios paga-bo. If you know send speak that my God pay-you ‘If you know why, inform me, that God will compensate you.’

(PA)

The use of TMA markers is not obligatory after the first verb in coordinated sentences unless there is a change in tense/aspect. In the following example, a verb in the past (borbe) carries no marker when it follows a verb (biny) bearing the marker a: (12)

.

Bida manda gabla Ku my Kico bechy a-biny Dear send speak with me what Bechy /-come busca na punta & borbe bay a-Sina presto. (PA) find in Punda and return go that-way fast ‘My dear, inform me what Bechy was looking for in Punda and why she left so fast.’

a and ta become portmanteau TMA polarity markers in PA, leading to the elimination of factative aspect

Because realis markers are by default linked to [+past] tense and because [–stative] verbs would be interpreted as being [+past] and [+perfective] by factative tense/aspect, pressure was exerted on ta to become obligatory on [–stative] verbs to mark [–past] and [–perfective] in opposition to a. This oppositional dynamic between a and ta would have had the effect of reinforcing the aspectual default interpretation of a, gradually transforming it into a dual function [+realis] and [+perfective] marker. Because of its status as a [+realis] marker, however, a would remain unavailable as a [+perfective] marker in incompatible [–realis] environments, such as negatives, imperatives, subjunctives, question-word questions, or in verb complexes containing modal auxiliaries, where the [+perfective] marker kaba must be used instead of a to signal a perfective meaning. As indicated by Maurer (1998: 165), a does not co-occur

Chapter 12. No exception to the rule 

with statives and modals such as por ‘can,’ ta ‘to be,’ tin ‘to have,’ or ke ‘to want.’ However, this restriction does not apply to kaba, as confirmed in recordings made in Aruba in 2000: (13)

o bol kaba ke pálo. . . or ball / want stick ‘or a ball wanted a stick. . .’.

(PA)

At the same time, the obligatory use of both a and ta before [–stative] verbs in nonfocused constructions would have had the effect of incorporating ta into the [+realis] pole of an earlier opposition between modal [–realis] marking vs. a [+realis] marking in these environments, thus transforming ta into both a [–perfective] and [+realis] marker. Evidence for the [+realis] use of ta to emphasize the objective truth value of an assertion can be found as far back as in PA texts from 1776 (Wood 1972). In present day PA, the main opposition in the modality system is between [–realis] modals vs. [+realis] a/ta in main clauses. The obligatory use of a and ta as markers of aspect and modality before [–stative] verbs in non-focused environments would have eventually led to the elimination of any possibility of a factative (‘unmarked’) aspect reading for most verbs. The elimination of factitive aspectual interpretation, however, would not eliminate the default TMA links illustrated in diagram 1 above. That is, the default links between [+perfective] aspect and [+past] tense and [+anterior] sequence would have given a secondary factative-like [+past/+anterior] value to a, while the default links between [–perfective] aspect and [–past] tense and [–anterior] sequence would have given a secondary factative-like [–past/–anterior] value to ta. This would explain why a was considered by linguists to be a past tense auxiliary and why ta was considered to be a present tense auxiliary before Andersen’s reanalysis of them as aspectual markers in 1993.11 In this way, PA a has become a portmanteau TMA polarity marker which can signal [+realis], [+perfective], [+past], and/or [+anterior], while PA ta has become a portmanteau TMA polarity marker which can signal [+realis], [–perfective], [–past], and/or [–anterior] according to lect.12

.

Anterior marking function of PA taba is attenuated, leading to conflation with ta

Both Andersen (1993: 62, 70) and Bickerton (1981: 87) posit an earlier [+anterior] marker taba for PA, whose counterpart would have been [+anterior] bin in NP. . It also accounts for their more tense-like properties in acrolectal speech and writing, where the tense prominence of the superstrate reinforces their secondary tense marking function and de-emphasizes their primarily aspect marking function in the basilects. . One of the characteristics of lectal variation is that the acrolect tends to be tense-dominant and the basilect aspect-dominant.

 Nicholas Faraclas, Yolanda Rivera-Castillo and Don E. Walicek

With the extension of a/ta aspect-modality marking to all non-focused [–stative] verbs, taba would have come to be used much more commonly with ta than with a, because as a [–perfective] marker, ta would automatically confer a factative-like default [–past/anterior] interpretation on the verb. Therefore taba would need to be used together with ta whenever a [–perfective] [+past/anterior] interpretation was needed. Because a is a [+perfective] marker, a would automatically confer a factativelike default [+past/anterior] interpretation on the verb, thereby limiting the need for taba with any [–stative] verb marked by a. For this reason, taba only survives in present day PA as part of the conflated form tabata, which combines [+past/anterior] taba with [–perfective] ta. Here PA shows similar patterning to NP, where default paths on the TMA web are normally unmarked while non-default paths (such as [+past/anterior] [–perfective]) often require special TMA markers or combinations of TMA markers. The combination of taba with ta is evident, for example, in the Aruban recordings from the year 2000: (14)

Si paso mi tawa-tin un amigo di kapverdiaan yes since I /-have a friend  Capeverdean ‘Yes, since I had a Cape Verdean friend.’

(PA)

(15)

un ku ami tawata kere. . . k’e tawat’un dialect. . . a that I / believe that-it ⁄-a dialect ‘a. . . what I believed [. . .] that it was a dialect. . .’

(PA)

Unlike a, tabata combines with stative and non-stative verbs, to the point of undergoing conflation with the stative verb tin, as shown in Example 14; this conflation is, however, optional: (16)

.

Mi tata tawata tin un auto. I father / have a car ‘My father used to have a car.’

(PA)

Irrealis/future marking function of PA bai is attenuated, leading to conflation with ta and replacement by lo and Ø

If we consider PA bai, the counterpart of [–realis] gò in NP, to be one of the original future markers in PA, it would have undergone a somewhat similar process to that just described for taba. As a [–perfective] (and by default [–past] marker), ta would be more semantically compatible with bai than the [+perfective] (and by default [+past]) a (leaving aside some ambiguous cases of a + bai in present-day PA speech, which merit further examination beyond the scope of the present work). Eventually, ta was conflated with bai to form ta bai or tei. This would have left a gap in the inventory for marking other [–realis] clauses, which is currently being filled by lo or the Ø ‘subjunctive.’ Evidence for the grammaticalization of lo as a future marker is the fact that it still straddles the boundary between adverbial and auxiliary (see Example 19 below). Unlike other auxiliaries, lo often occurs

Chapter 12. No exception to the rule 

in the adverbial slot before the negator no and subject pronominals (Kouwenberg and Murray 1994), with a very similar distribution to the future marker bai in Tok Pisin, a paradigmatic case for the synchronic transformation of an adverbial into an auxiliary (Sankoff and Laberge 1973). The use of bai as a future marker is attested in the following exchange taken from the Aruba recordings in which speakers A (17) and B (18) talk about an upcoming event: (17)

Wel, dia binticuater di e luna aki tin dera gai, no? Well day twenty.four  the month here have holiday Q ‘Well, the 24th of this month we have the “dera gai,” don’t we?’

(PA)

(18)

Hopi kandela, Aruba ta bai ta bou huma. much fire Aruba   be under smoke ‘(With) so much fire, Aruba will be covered with smoke.’

(PA)

When used as the main verb meaning ‘go’ bai may combine with the future marker lo (Example 19) and with the perfective marker a (Example 20): (19)

Lo ta bai. . . den siman ta te un’or.   go during week be until one-hour ‘[It] will happen. . . during the week until one in the morning.’

(PA)

(20)

Si, bo sa, [. . .] mi a bai cerca Andreína, yes you know I  go yard Andreína ‘Yes, you know [. . .], I visited Andreína’

(PA)

.

Tense, modality and earlier treatments of TMA in PA

As stated above, early descriptions of TMA in PA categorized a as a past or perfect marker and ta as a present marker (see, for example, Littmann 1945: 53). Given our analysis of the evolution of these markers, a tense-based interpretation of the functions of a and ta is neither surprising nor entirely without justification, since both have come to encode a broad range of TMA polarities, including [+past] for a and [–past] for ta. While Maurer (1993: 243) essentially maintains a tense reading for a and ta, he makes a convincing and perceptive case for the prominence of modality in PA and even suggests that the modality systems of West African languages should be considered a probable source for this PA feature. He posits a ‘subjunctive zero’ marker for verbs in focused constructions (where a and ta cannot occur), citing the existence of subjunctive constructions in Bantu languages as a possible substrate source (ibid.: 249). A significant milestone in the treatment of TMA in Papiamentu was reached in Andersen’s reanalysis (1993) of a and ta as markers of perfective and imperfective aspect, rather than as markers of past and present tense. Bickerton’s LBH (1981), which predicts the existence and prominence of [±perfective] aspectual marking in Creole languages worldwide, proved to be a useful guide and tool for this strand of Andersen’s

 Nicholas Faraclas, Yolanda Rivera-Castillo and Don E. Walicek

work. Unfortunately, Andersen does not follow up on Maurer’s suggestion to take a closer look at the substrate languages in order to shed light on the role of modality marking in PA. As a result, he is inclined in the end to concede that his analysis does not remove the label of ‘atypical’ or ‘exceptional’ from PA, both in terms of its relation to other Caribbean Creoles as well as in relation to the LBH.

.

Conclusion: PA has conserved the basic features of both the substratal and ‘universal/prototypical’ creole TMA systems

In the conclusion to his analysis of aspect in Papiamentu, Andersen (1993: 89) cites the following five features as ‘deviant’ or ‘aberrant’ characteristics of its TMA system. In the light of the scenario outlined in Section 3 above, each of these apparently exceptional features can be seen to have developed from earlier PA TMA systems that differ minimally from those found in other Caribbean Creoles and their African substrate languages: 1. The near obligatory use of the markers a or ta before verbs: We have suggested above that the spread of a realis marker such as a from [–stative] verbs in non-focused constructions in narrative texts to other environments is a widespread phenomenon in West African languages. The ‘factative’ default interpretation between [+realis] marking and [+perfective] and [+past] that eventually led to a becoming a portmanteau morpheme marking all of these polarities also typifies West African languages. Once a begins to function in this way and its use is extended to more and more environments, it is predictable that the frequency of usage of ta would increase along with its semantic scope (from [–perfective] to [–past] and finally [+realis]), resulting in the near obligatory use of the markers a and ta before non-subjunctive verbs. 2. The absence of a ‘Ø marker’ for the perfective aspect: In NP and most of its substrate languages, an unmarked [–stative] verb is normally interpreted to be [+perfective], and [+past/anterior] while an unmarked [+stative] verb is normally interpreted to be [–perfective], and [–past/anterior] in a ‘factative’ TMA reading. Because all [–stative] verbs are obligatorily marked by the portmanteau TMA markers a and ta, such ‘zero marking’ is rarely attested in main clauses in PA. However, we argue that the same system within which factative TMA value assignment operates in West African languages and most other Creole languages still defines the basic parameters of the PA TMA system. In the PA TMA system, a and ta have become obligatory, taking over the default reading of non-stative verbs as past and perfective. These readings for a and ta are consistent with the factative/default matrix in Diagram 1 above, which typifies West African languages and probably most Creole languages as well. Thus the lack of “zero marking of perfective aspect” (Andersen 1993: 89) is a relatively superficial feature that masks fundamental congruity between TMA in PA, West African languages, and other Caribbean Creoles.

Chapter 12. No exception to the rule 

3. The existence of two irrealis markers (lo and Ø): As explained above, once a and ta marking become obligatory markers of [+realis] modality with [–stative] verbs, it is predictable that: (1) cases where a/ta marking are absent (‘zero marked’ cases) will be interpreted as [–realis]; and (2) the bai irrealis marker would conflate with ta thus attenuating its [–realis] function, paving the way for the incorporation of a different marker lo to signal [–realis]. The predictable result is “the existence of two [or even three!] irrealis markers” ([ta bai], lo and Ø) (Andersen 1993: 89). Rather than signifying a major difference between PA on the one hand and West African languages and Atlantic Creoles on the other, the proliferation of [–realis] markers in PA suggests that a central role is played by [+realis] modality marking in all of these languages. 4. The dual (tense and aspect) function of the marker tabata: With the propagation of a/ta marking on [–stative] verbs, the earlier use of a [+past/anterior] marker taba becomes unnecessary in [+perfective] environments because of the default [+past/anterior] reading of a. Thus taba comes to be used only in the [–perfective] environments marked by ta, leading to an eventual conflation of the two markers and the dual (tense and aspect) function of the marker tabata. 5. The lack of a specific morpheme that functions only as an anterior marker:13 With the conflation of taba and ta, the only way to mark [+past/anterior] is by a default factative-like reading of a. This “lack of a clear anterior marker” (Andersen 1993: 89) not only underscores the fact that the ‘unmarked’ default/factative TMA parameters found in these West African languages and most Caribbean Creoles are still very much operational in PA, but also that, just as in most West African languages and most Caribbean Creoles, the marking of tense/sequence is not as crucial as the marking of aspect and modality. In the analysis that we have presented above, we have used archival and spoken data from PA and an overview of typological features in West African languages to show that the PA TMA system is not ‘deviant’ or ‘exceptional.’ At the same time, we have argued that TMA in PA has developed in predictable ways from TMA systems that typified earlier forms of PA. We contend that these earlier systems resemble the TMA systems found in the Benue-Kwa languages spoken by the majority of those Africans who were taken as slaves to the Americas. Because these populations played a major role in the creation of the colonial era plantation Creoles, it is no accident that many of the features that typify TMA systems in West African languages are also the features that have been proposed as ‘universal’ or ‘prototypical’ Creole features (Bickerton 1981). . Andersen (1993: 89) also mentions the “deviant position of lo” as a feature of the TMA system. A discussion of this observation requires that the forms bai, ta, and lo all be examined. It also demands that the possibility of homophony for lo be taken into consideration. These are issues we hope to address in future work but for the time being they fall outside the scope of this paper.

 Nicholas Faraclas, Yolanda Rivera-Castillo and Don E. Walicek

This demonstrates how scenarios that recognize convergence among a matrix of factors, such as substrates, universals, superstrates. and monogenetic influences, can account for the data in an optimal way, providing viable explanations not only for phenomena that conform to ‘regular’ patterns and processes observed in Creole languages generally, but also for phenomena that appear on the surface to be ‘aberrant’ or ‘deviant.’

References Aikhionbare, M.O. and Chumbow, B.S. 1982. Focus in Nigerian Pidgin English, Niger Delta. Cahiers Ivoiriens de Recherche Linguistique 12. 21–45. Andersen, R.W. 1993. Papiamentu tense-aspect, with special attention to discourse. In Winford, D. (ed.). Predication in Caribbean English Creoles [Creole Language Library 10]. Amsterdam: John Benjamins. 59–95. Bamgbose, A. 1966. A grammar of Yoruba [West African Language Monographs 5]. Cambridge: Cambridge University Press. Barnwell, K.G.L. 1969. A grammatical description of Mbembe (Adun dialect) – A Cross River language. PhD thesis, University College London. Bickerton, D. 1980. Creolization, linguistic universals, natural semantax, and the brain. In Day, R.R. (ed.). Issues in English Creoles. Papers in the 1975 Hawai’i conference. Heidelberg: Groos. 1–18. Bickerton, D. 1981. Roots of language. Ann Arbor: Karoma. Binnick, R.I. 1991. Time and the verb: A guide to tense and aspect. Oxford: Oxford University Press. Bybee, J. 1994. The grammaticalization of zero: Asymmetries in tense and aspect systems. In Pagliuca, W. (ed.). Perspectives on grammaticalization. Amsterdam: John Benjamins. 235–54. Bybee, J., Perkins, R. and Pagliuca, W. 1994. The evolution of grammar: Tense, modality, and aspect in the languages of the world. Chicago: University of Chicago Press. Clark, D.J. 1969. A grammatical study of Ekpeye. PhD thesis, School of Oriental and African Studies, University of London. Comrie, B. 1976. Aspect. Cambridge: Cambridge University Press. Comrie, B. and Smith, N. 1977. Lingua descriptive studies: Questionnaire. Lingua 42. 1–71. Emmenanjo, E.N. 1978. Elements of modern Igbo grammar. Ibadan: Ibadan University Press. Emmenanjo, E.N. 1985. Auxiliaries in Igbo syntax: A comparative study. Bloomington: Indiana University Linguistics Club. Faraclas, N. 1984. A grammar of Obolo. Bloomington: Indiana University Linguistics Club. Faraclas, N. 1987. Creolization and the tense-aspect-modality system of Nigerian Pidgin. Journal of African Languages and Linguistics 9. 45–59. Faraclas, N. 1988. Nigerian Pidgin and the languages of Southern Nigeria. Journal of Pidgin and Creole languages 3.2. 177–97. Faraclas, N. 1990. From Old Guinea to Papua New Guinea: A comparative study of Nigerian Pidgin and Tok Pisin. In Verhaar, J. (ed.). Melanesian Pidgin and Tok Pisin. Amsterdam: Benjamins. 91–169. Faraclas, N. 1996. Nigerian Pidgin. London: Routledge.

Chapter 12. No exception to the rule  Foley, W. 1986. The Papuan languages of New Guinea. Cambridge: Cambridge University Press. Gilman, C. 1986. African areal characteristics: Sprachbund, not substrate? Journal of Pidgin and Creole languages 1.1. 33–50. Heine, B. and Reh, M. 1984. Grammaticalization and reanalysis in African languages. Hamburg: Helmut Buske. Hinnebusch, T. 1989. Bantu. In Bendor-Samuel, J. (ed.). The Niger-Congo languages. Lanham, MD: University Press of America. 450–76. Hyman, L. and Magaji, D. 1970. Essentials of Gwari grammar [Occasional Publication 27]. Ibadan: Institute of African Studies, University of Ibadan. Jenewari, C.J. 1977. Kalabari syntax. PhD thesis, University of Ibadan. Kaufman, E.M. 1969. Ibibio grammar. PhD thesis, University of California, Berkeley. Kouwenberg, S. and Murray, E. 1994. Papiamentu [Languages of the World/Materials 83]. München: Lincom Europa. Kuperus, J. 1978. Esqisse du systeme verbal de l’oro du Nigeria. MA thesis, Universite Paris V. Littmann, J. 1945. The windblown language: Papiamentu. Hispania 28.1. 50–9. Muntenau, D. 1996. El papiamento, lengua criolla hispánica. Madrid: Editorial Gredos. Marchese, L. 1983. On assertive focus and the inherent focus nature of negatives and imperatives: Evidence from Kru. Journal of African Languages and Linguistics 5.2. 115–30. Martinus, F. 1994. The origin of the adjective participle in Papiamentu. In Zimmerman, K. (ed.). Lenguas criollas de base lexical espanhola y portuguesa. Frankfurt/Madrid: Vervuert. 231–49. Maurer, P. 1993. Subjunctive mood in Papiamentu. In Byrne, F. and Holm, J. (eds). Atlantic meets Pacific. Amsterdam: John Benjamins. 243–50. Maurer, P. 1998. El PA de Curaçao. In Perl, M. and Schwegler, A. (eds). América negra: Panorámica actual de los estudios lingüísticos sobre variedades hispanas. portuguesas y criollas. Frankfurt/Madrid: Vervuert/Iberoamericana. 139–217. McWhorter, J. 2000. The missing Spanish creoles: Recovering the birth of plantation contact languages. Berkeley: University of California Press. Meier, P. Meier, I. and Bendor-Samuel, J. 1975. A Grammar of Izi. Norman: Summer Institute of Linguistics and the University of Oklahoma Press. Salas González, E. 1996. A semantics for the Spanish perfective and imperfective forms. PhD thesis, University of California-Davis (UMI 9627753). Samarin, W.J. 1967. A grammar of Sango. The Hague: Mouton. Sankoff, G. and Laberge, S. 1973. On the acquisition of native speakers by a language. Kivung 6.1. 32–47. Singler, J. 1985. The story of o. Paper presented to the 16th conference on African linguistics, Yale University, April 1985. Stanford, R. 1967. The Bekwarra language of Nigeria: A grammatical description. PhD thesis, School of Oriental and African Studies, University of London. Stewart, J.M. 1989. Kwa. In Bendor-Samuel, J. (ed.). The Niger-Congo Languages. Lanham, MD: University Press of America. 217–46. Thomas, E. 1978. A grammatical description of the Engenni language. Arlington: Summer Institute of Linguistics and the University of Texas Press. Williamson, K. 1965. A grammar of the Kolokuma dialect of Ijo [West African Language Monographs 2]. Cambridge: Cambridge University Press. Williamson, K. 1989a. Benue-Congo Overview. In Bendor-Samuel, J. (ed.). The Niger-Congo Languages. Lanham, MD: University Press of America. 247–74.

 Nicholas Faraclas, Yolanda Rivera-Castillo and Don E. Walicek Williamson, K. 1989b. Niger-Congo Overview. In Bendor-Samuel, J. (ed.). The Niger-Congo Languages. Lanham, MD: University Press of America. 3–46. Winford, D. 2000. Tense and aspect in Sranan and the creole prototype. In McWhorter, J. (ed.). Language change and language contact in pidgins and creoles. Amsterdam: John Benjamins. 383–442. Wood, R.E. 1972. New light on the origins of Papiamentu. Neophilologus 61.1. 18–30.

 

A look at so in Mauritian Creole From possessive pronoun to emphatic determiner Diana Guillemin The University of Queensland

This paper argues that at particular stage in the development of the determiner system in Mauritian Creole (MC), the 3 possessive pronoun so, inherited from the French son, was used as a singular definite determiner as well as a possessive pronoun. It was used when there was a need to single out a unique element in the discourse, or to introduce a new referent which was to become the focus of attention. So was mostly used with genitive constructions, where the phonologically null determiner, represented as δ, was equally grammatical, as shown in (1) and (2): (1)

(2)

So frère ça mamzelle là  brother  girl  The brother of this young girl “Le frère de la jeune fille”1

(Baissac 1888: 154–5)

ène loulou dire δ mari ça femme là a wolf tell husband  woman  A wolf says to the husband of this woman “Un des loups dit au mari” (de cette femme)

(Baissac 1888: 156–7)

This use of the possessive pronoun form is not grammatical in French. However, I argue that, in the early creole, the genitive construction in fact licensed its use as a determiner, when a new referent was being introduced in the discourse. This use of so with genitive constructions is no longer . Most of my MC examples will be glossed as follows: MC on the top line, 2nd line is morpheme-by-morpheme gloss in English, 3rd line is a Standard English translation equivalent, 4th line is a standard French equivalent. In the case of long quotes, I will provide only a ‘block’ translation. English glosses will be in single quotation marks and French glosses will be in double quotation marks. For all the quotations from Baissac (1888), I provide his own French translation of the creole text, while the English as well as the French translations in all other examples are my own. For the sake of reader friendliness, the sources for the data appear at the end (i.e., after the French translations), except for the New Testament passages, since they are all from the same source, Anderson 1885, unless otherwise specified. For examples that are in languages other than MC, the language will be identified.

 Diana Guillemin

grammatical in modern MC, but this particle continues to be used as an emphatic determiner with both singular and plural NPs.

.

Introduction: A new determiner system

Early in the genesis of Mauritian Creole, the singular French definite articles (le/la) and partitive determiner (du), incorporated into a significant number of the nouns that they modified, e.g.,: ● ● ● ●

Le roi (the king) La fenêtre (the window) (De) la farine (flour) Du monde (people)

→ → → →

lerwa (king) lafnet (window) lafarin (flour) dimun (person, people)

In the case of count nouns starting with a vowel, it was the plural phonological form of the noun, with an initial [z], as a result of liaison with the plural determiners des/les, which transferred into the creole. The plural determiners themselves did not incorporate into the nouns, as shown: ● ●

des/les enfants (children/the children) des/les oiseaux (birds/the birds)

→ →

zanfan (child) zwazo (bird)

It seems that the French determiners, which serve to mark the semantic contrasts of (in)definiteness and singular vs. plural were not recognized as separate morphemes, but were taken to be an integral part of the nouns that they modified (Baker 1984; Chaudenson 1981; Grant 1995; Strandquist 2005).2 Not all nouns ended up with an incorporated article, but the French determiner system collapsed, and the immediate consequence was that, in the early creole, all nouns were bare, yielding ambiguous interpretations between [±definite] singular and [±definite] plural interpretations. Gradually, over a period of some 100 years, from the late 18th century to the end of the 19th century, new functional items emerged to mark these semantic contrasts, namely: ●

● ●

The singular indefinite determiner enn, which appears early in the 19th century, to unambiguously mark nouns as singular and [−definite]. The post-nominal definiteness and specificity la, which also appears around 1820. The pre-nominal demonstrative sa, which is mostly used in conjunction with post-nominal la.

. Determiners also serve to mark gender in French, but this is not relevant to my analysis as gender is not a grammatical feature of nouns in MC.

Chapter 13. A look at so in Mauritian Creole  ●

●

The plural marker bann, which grammaticalizes towards the end of the 19th century. It is unspecified for the feature [±definiteness]. I propose that a null [+definite] determiner was present early in the creole.

Despite the emergence of new determiners, modern MC continues to admit bare nouns in various syntactic configurations, yielding either a [−definite] plural or [+definite] singular interpretation. In existential constructions, which admit only indefinite NPs (Milsark 1979), a bare noun is [−definite] plural, as is the case with the NP manb in (3): (3)

Dan sak su-komite ena manb LPT in each sub-committee have member LPT ‘In each sub-committee there are members of LPT’ “Dans chaque sous-comité il y a des membres de la LPT” (Ledikasyon pu Travayer 2006)

When a bare noun occurs in an argument position, it is always [+definite] singular, as is the case with Zirispridanss, the subject of the verb ‘be’ (se), and kad, the object of the preposition dan, in (4) and (5) respectively: (4)

Zirispridanss setenn konsept ziridik ki itilize kan, jurisprudence is.a concept legal  use when, ‘Jurisprudence is a legal concept which is used when, . . .’ “La jurisprudence est un concept légal qu’on utilise quand, . . .” (Carpooran 2002: 69)

(5)

dan kad Liniversite Moris, . . . in context university Mauritius ‘in the context of the University of Mauritius,’ “dans le cadre de l’université de Maurice,”

(Carpooran 2002: v)

Assuming that ‘DP can be an argument, NP cannot’ (Longobardi 1994: 628), though D may be a phonologically null element, the bare nouns in (4) and (5) are analyzed as DPs as opposed to NPs. Interestingly, null D forces a singular interpretation of its complement NP.3 The plural marker bann is unspecified for definiteness. In existential constructions, bann + NP is [−definite] plural, and in argument positions bann + NP is [+definite] plural, as shown in (6) and (7) respectively: (6)

Lontan ti ena bann baryer kiltirel in.the.past  have  barrier cultural ‘In the past there were cultural barriers’ “Autrefois il y avait des barrières culturelles”

(Hookoomsing 2002: 30)

. In a separate paper (fc.), I propose that the null definite determiner δ selects Num(ber)P, as opposed to NP. The head of Num is singular by default, and bann is the phonological expression of the feature [+plural] associated with the Number node. Bann and cardinal numerals are in complementary distribution. Thus, δ + N is [+definite] singular and δ + bann + N is [+definite] plural.

 Diana Guillemin

(7) E kan bann premye syantis ti dir ki later vire, . . . and when  first scientist  say  earth revolve ‘And when the first scientists said that the earth revolves,’ “Et quand les premiers scientistes ont dit que la terre tourne,” (Ah-Vee 2002: 72)

The determinative use so in the 19th century creole manifested itself at a particular stage of the development of the new determiner system. It served to mark an NP as referential, and like the null determiner, it forced a singular interpretation of its complement NP. I argue that its occurrence was not arbitrary, but was motivated by both pragmatic and syntactic considerations, when the interpretation of a bare noun may have been ambiguous. Its use as singular definite determiner disappeared following the grammaticalization of the plural marker bann towards the end of the 19th century. This paper is organized as follows: In Section 2 I look at the use of so in the early creole; Section 3 is a brief overview of the determinative use of possessive pronouns in other languages of the world. In Section 4 I compare the pragmatic use of so vs. the null determiner in early MC, and in Section 5 I propose that the genitive constructions with which this morpheme was used in fact licensed its use as a definite determiner. Section 6 looks at the use of so in modern MC, where it functions as an emphatic determiner. Section 7 is my conclusion.

.

So in the early creole

The first determinative use of the possessive pronoun appears in the first official imprint in the creole language of the Proclamation for the emancipation of the slaves in 1835, where it still has the French form, son: (8)

zautre tous va perdi son nom Esclaves 2 all  lose  name Slaves ‘You will no longer be called Slaves’ (lit. You will all lose the name Slaves) “Vous perdrez tous le nom d’Esclaves” (Nicolay 1835, in Chaudenson 1981: 119)

Its occurrence becomes more frequent during the second half of the century: (9)

(10)

Qui na.pas capav’ mêm’ senti so l’oder rose, . . . who  can even smell  scent rose ‘Who cannot even smell the scent of the rose, . . .’ “Qui ne peut même pas sentir le parfum de la rose,”

(Lolliot 1855)

Qui li capav’ vini so la.taill’ ein’ taureau  3. can become  size a bull ‘That he can reach the size of a bull’ “Qu’il peut atteindre la taille d’un taureau”

(Lolliot 1855)

Chapter 13. A look at so in Mauritian Creole 

In the creole version of the Bible, Anderson consistently translates the French definite article as so: (11)

(12)

li ti appel so nom Jésu. 3.  call DET name Jesus ‘he named him Jesus’ “auquel il donna le nom de Jésus”

(1993: 2) (1955: Ch. I, v. 25, p. 1456)

nettoy an promié, so andan bol, é so andan pla, . . . clean at first  inside cup and  inside dish ‘First clean the inside of the cup, and the inside of the plate,’ (1993: 26) “Purifie d’abord l’intérieur de la coupe et de l’écuelle,” (1955: Ch. 23, v. 26, p. 1492)

In his translations of folk tales, as recounted by Ppâ Lindor’s and Mmâ Télésille’s, two elderly native creole speakers, Baissac interprets their use of so as a definite article:4 (13)

ça qui pour arrive dans so finition mo zistoire    happen in  end . story ‘What will happen at the end of my story’ “Ce qui va arriver à la fin de mon histoire” (Baissac 1888: 328–9)

(14)

Paulin ti son frère Pauline, Pauline ti son seir Paulin Paulin   brother Pauline, Pauline   sister Paulin ‘Pauline was the brother of Pauline, Pauline was the sister of Paulin’ “Pauline était frère de Pauline, Pauline était sœur de Paulin” (Baissac 1888: 290–1)

It would not be possible in English or French to use the 3 singular possessive pronoun in any of the above contexts, as shown: (15)

a. b. c. d.

*his name Jesus *son nom de Jésus *its scent of the rose *son parfum de la rose *its end of my story *sa fin de mon histoire *his brother of Pauline *son frère de Pauline

Furthermore, given that bare NPs would have been equally grammatical, in the above examples, what precisely does so bring to the interpretation of the noun phrase?

. Although Baissac was a French speaker, I will assume that the creole examples in his folk tales are genuine renditions of the spoken creole of his day, on the grounds that he did not actually invent the stories, but merely transcribed them and translated them into French. He writes in the Préface to his 1888 collection: ‘Ce n’est pas, en effet, oeuvre de conteur que nous avions à faire, mais de simple rapporteur, voire de sténographe, …’ (Baissac 1888: viii). (‘In fact, our role was not one of storyteller, but of simple reporter, nay stenographer, …’)

 Diana Guillemin

Prior to addressing this question, I will refer to studies on the determinative use of possessive pronouns in other languages of the world.

.

The use of the possessive pronoun as a determiner

The determinative use of a possessive pronoun has been documented in other languages of the world by Fraudud, who notes that its ‘essential semantic/pragmatic property has to do not with possessivity but with referentiality and, more specifically, with focus of attention’ (2001: 260, emphasis in original).5 In the following examples, the possessive, or 3 suffixes, are used ‘determinatively’, as ‘definite articles’ or instead of definite articles: (16)

(17)

Ver-as lim j-is silema n’in forest-..3 snow-. melted already ‘In the forest the snow melted already’

(Komi, Uralic) (Fraurud 2001: 248)

Düş-tü-m. Sonra-si–ni bil-me-m (Turkish) fall--1 after-.3- know--1 ‘I fell. I don’t know what happened after (that)’ (lit: ‘its after’, ‘the after of it’) (Fraurud 2001: 248, 261)

Himmelmann (2001) claims that ‘there is evidence that in several languages possessive pronouns can be used well beyond the typical contexts of use for possessive pronouns. This is particularly clear in instances where there is no antecedent for the pronoun,’ as in the following example from Indonesian: (18)

Karena sungai-nya keruh because river-3 muddy ‘Because the river was muddy’

(Indonesian) (Himmelmann 2001: 839)

He notes that ‘The river referred to here is the river known to everybody in the speech community’ but it is clear that there is no possessor for the river and that the possessive pronoun is used as a definite determiner. Himmelmann comments that in none of the above languages (Uralic, Turkish and Indonesian) does ‘it appear that the use of a possessive pronoun is obligatory in all . . . definite or specific contexts.’ (2001: 839) I am not aware of the determinative use of the 3 possessive pronoun in other creole languages, but the reanalysis of the 3 pronoun as a marker plurality and definiteness has been documented by DeGraff (fc), Deprez (2005) and Lefebvre (1998, 2001) for

. Languages include, among others, the Semitic language Amharic, Komi, Udmurt, Turkish and Yucatec Maya. For a list of references see Fraurud (2001).

Chapter 13. A look at so in Mauritian Creole 

Haitian Creole (HC), where the morpheme yo serves both as a plural marker and 3 pronoun, as shown: (19) a.

krab yo crab  ‘the crabs’

b.

yo pati 3. leave ‘They left’

(HC) (Lefebvre 2001: 8)

DeGraff and Deprez also indicate that yo is ambiguous between a definite plural marker, and a 3 possessive pronoun: (20)

Bouki deja pase rad yo

Bouki has already ironed the(ir) clothes (21)

Jan achte liv yo John bought the books/their books

(HC)

(DeGraff 1994: 115) (HC) (Deprez 2005: 861)

In her work on the role of grammaticalization in French Lexifier Creoles, Deprez (fc) notes that post nominal number in Guyanese Creole is marked by the morpheme ye, which like yo in HC, also functions as 3 pronoun, and that ‘both markers are associated with definiteness in their pronominal use and in their determiner uses’. This view is seemingly shared by DeGraff, who glosses yo as a [+definite] plural determiner: (22)

Li kouri dèyè makout yo 3 run after thug - ‘He/she ran after the thugs’

(HC) (DeGraff fc: 21)

What is interesting to note is that the determinative use of pronouns in all these languages, like the use of so in MC, share the feature [+definiteness]. However, while it is the plural pronoun which is reanalyzed in Haitian and Guyanese Creole, it is the singular possessive pronoun which is reanalyzed in MC.

.

So versus the null determiner in MC

I suggested that the reanalysis of the possessive in early MC was not arbitrary, but must have been driven by both pragmatic and syntactic considerations. In this section, I give examples to support my claim. The following example is from a MC translation of the Gospel according to St Mark: (23)

E l’her là mem, comman zot sorti dan sinagog, zot vine avec Jac et Jan, dan la caz Simon é André. So bel mér Simon, ti apré dormi, li malad avec la fiévr; e l’her là mem, zot parl Jésu pour li. (1888: 5) ‘As soon as they left the synagogue, they entered the house of Simon and Andrew, with James and John. Now Simon’s mother-in-law was in bed with a fever, and they told him about her at once.’ (1993: 34)

 Diana Guillemin

“Et aussitôt, en sortant de la synagogue, il alla dans la maison de Simon et d’André avec Jacques et Jean. Or la belle-mère de Simon était au lit avec la fièvre, et aussitôt on lui parle d’elle.” (1955: Ch. 1, v. 29–30, p. 1505)

The bel mér is the newly introduced individual, who becomes the focus of the narrative, as opposed to the other previously introduced characters. She is the one who is sick with fever, and who is about to be blessed by Jesus. A null [+definite] determiner would have been equally grammatical, but it would fail to identify the bel-mer as an important character in the discourse. The entity marked by so is obviously more salient than if zero marked. Note that in the French text, the clause is introduced by or, a discourse marker, which is defined as: “Marquant un moment particulier d’une durée ou d’un raisonnement” (‘marking a particular moment in a duration or an argument’) (Robert 1972). It serves to introduce a contrasting viewpoint in an argument, for which the translator has chosen the particle so. The notion of contrastive focus is also implied in Example (24): (24)

Mais mo éne vié doumounde, so fille léroi zène zène but 1. an old person  daughter king young young ‘But I am an old person, (whereas) the king’s daughter is quite young.’ “Mais je suis vieux et la princesse est jeune.” (Baissac 1888: 52–3)

In the 1880’s the plural marker bann had not yet fully grammaticalized, and a bare NP could be interpreted either as [+definite] singular as with lacase in (25), as [+definite] plural as loulous in (26) or as [−definite] plural, as robes and cimises in (27): (25)

Zène fille couri lacase léroi Young girl run house king ‘The young girl runs to the king’s palace’ “La jeune fille court au palais du roi”

(26)

Loulous6

(27)

(Baissac 1888: 140–1)

trouve ça la.forêt là; qui zaute va faire? wolves find  forest   3  do ‘The wolves find this forest; what will they do?’ “Les loups rencontrent la forêt, que vont-ils faire?” (Baissac 1888: 174–5)

Zotte mette éne bande coutirières dan louvraze: coude robes, 3 put a group dressmaker in work: sew dress, coude cimises7, . . . sew shirt, . . . ‘They put a number of dressmakers to work: (they) sew dresses, (they) sew shirts, . . .’

.

The noun is in the plural, but the final s would be silent, as in French.

. Although enn band is used frequently in the folk tales to mark an indefinite number, i.e., for [−definite] plural, there is only one example of band as a marker of plurality:

Chapter 13. A look at so in Mauritian Creole 

“On met un régiment de couturières à l’ouvrage: elles cousent des robes, des chemises. . .” (Baissac 1888: 364–5) The presence of so in (24) eliminates this ambiguity with regard to the interpretation of the NP fille léroi, by picking out a single daughter of the king. It also serves to emphasize the contrast in age between the princess and the narrator, who, in scheming to save his life, must make believe that he is much too old to marry her.

As early as 1880 Baissac identified this emphatic use of so, and claimed that: “So est en réalité le seul adjectif possessif. Il a en créole un emploi emphatique assez original.” (1880: 14) (‘So is in fact the only possessive adjective. It has in creole quite an original emphatic use.’) He gives the following example, with the French equivalent, where he has translated so as the definite article le/la: (28)

So mâle cardinal rouze, so femelle zaune zaune  male cardinal red,  female yellow yellow ‘The male of the cardinal is red, the female is yellowish’ “Le mâle du cardinal est rouge, la femelle jaunâtre”

(Baissac 1880: 14)

Once again, so places emphasis on its complement NP while also encoding a notion of contrast, here between the male and female of the species cardinal. However, the fact remains that pragmatics alone cannot justify the use of the possessor as determiner, such use must also be grammatical. What is interesting about the determinative use of so in the early creole is that it always occurs with genitive constructions, and I propose that such constructions were in fact able to license the determinative use of this particle.

.

The use of a definite article There are broadly two functions associated with definite articles. One of these is an anaphoric function, to refer back to something mentioned in the preceding discourse. The other is a nonanaphoric function, to refer to something not mentioned in the preceding discourse but whose existence is something that the speaker assumes is known to the hearer. This assumed knowledge may be based on the general knowledge (as in the sun) or it may be based on inferences that the hearer can make in context (for example, inferring from mention of a house that the house has a door, thus making it possible to use a definite article in referring to the door of the house (Haspelmath et al. 2005: 154).

(i)

tout ça bande malices là all   trick  ‘all these tricks’ “tous ces tours-là” (Baissac 1888: 106–7)

 Diana Guillemin

A necessary and sufficient condition of the use of the definite article, thus, is for the hearer to be able to identify some set of elements in discourse, and to locate a referent in it. When a definite NP is first introduced, it relies in the context for its interpretation, which can be provided either by a discourse antecedent, or in the form of an associative relationship with a discourse antecedent. When there is a discourse antecedent, the definite article has ‘direct anaphoric’ use, e.g., a book . . . the book. When there is no discourse antecedent, there must be a ‘trigger’ (Hawkins 1978: 123) to license the use of the definite article, in which case, it has ‘associative anaphoric’ use, as in a book . . . the author, where book is the ‘trigger’ which licenses the use of the definite article with the NP author. In the case of ‘direct anaphora’ the antecedent of the definite NP must be relatively close in the discourse. In the case of ‘associative anaphora’ , where there is no discourse antecedent, the use of the definite article is licensed by shared knowledge of the relationship between the newly mentioned NP and its ‘trigger’. In MC, as in English and French, a new referent is introduced in the discourse by the indefinite article éne, equivalent of a/an and French un/une, from which it is derived: (29)

Ti éna éne fois éne zéne homme qui ti marié sembe éne zéne fille. Côment li té pour travaille dan éne carreau cannes morceau loin so lacase, li donne so femme quate lacloces: éne lacloce couivre, éne lacloce larzent, éne lacloce lor, éne lacloce diamant. Acthère là li dir li: coute bien: quand to va vlé mo tourne lacase pour dir moi quiqueçose, sonne lacloce couivre, quand to va pressé, sonne lacloce larzent; quand to va bisoin méme mo rentré, sonne lacloce lor; mais ça lacloce diamant là, to tendé, zamais zamais sonné, néque lhère grand malhor av toi. ‘Once upon a time there was a young man who was married to a young woman. As he worked in a sugar cane field a little far from his house, he gave his wife four bells: a copper bell, a silver bell, a golden bell, a diamond bell. Then he said to her: listen carefully, when you will want me to come back home to tell me something, ring the copper bell, when you are in a hurry, ring the silver bell, when you really need me to come back home, ring the golden bell; but this diamond bell, you hear me, never never ring, except if you are in terrible danger.’ “Il y avait une fois, un jeune homme qui avait épousé une jeune fille. Comme il devait aller travailler dans un champs de cannes assez éloigné de sa case, il donna à sa femme quatre cloches: une cloche de cuivre, une cloche d’argent, une cloche d’or et une cloche de diamant. Puis il dit: “Ecoute-moi bien. Quand tu voudras me voir revenir à la maison pour me dire quelque chose, tu sonneras la cloche de cuivre; quand tu seras pressée, sonne la cloche d’argent; si tu as vraiment besoin de moi sonne la cloche d’or; mais pour cette cloche en diamant, ne la sonne jamais que si quelque danger terrible te menace.” (Baissac 1888: 180–1)

In the above text, every new concept is introduced by éne, exactly as in the French and English translations. The indefinite singular article picks out a specific individual from a potentially larger set. With every subsequent mention of the noun that refers to a previously mentioned object, English uses the definite article ‘the’, French uses the equivalent le/la, while MC has bare nouns. The noun phrases lacloce couivre, lacloce larzent, lacloce

Chapter 13. A look at so in Mauritian Creole 

lor, are nevertheless referential expressions. They are definite in that they have a ‘strong’ discourse antecedent, and they refer to the only member of a relevant set, introduced by the indefinite singular article éne. The null determiner in MC thus behaves like the English and French definite article when used in direct anaphora. The environment in which the determinative use of so occurs in 19th century MC is with genitive constructions, which Hawkins (1978: 139) defines as ‘associative clauses’ because they incorporate both the trigger and the associate. Genitives function just like relative clauses in that they enable the hearer to first identify, and then locate the referent in the appropriate shared set of objects. While, for example, the NP ‘the end’ is meaningless on its own as a new referent in a discourse, the NP ‘the end of my story’ is grammatical because the modifying clause ‘of my story’ turns the NP ‘the end’ into a referential expression, thus licensing the use of the definite article. However, these very genitive constructions that licensed the determinative use of so in early MC have been analyzed as misconstrued genitives, which marked a transition between two types of genitive constructions that co-exist in MC (Syea 1994), namely: (30)

Type A: sapo mo papa hat 1. father ‘My father’s hat’ “Le chapeau de mon père”

(31)

Type B: mo papa so sapo 1. father 3. hat ‘My father’s hat’ “Le chapeau de mon père”

While Type A is derived from the French (minus the definite article le and the preposition de), Type B patterns with West African possessive constructions, namely Twi and Ewe, as well as those of Hindi and Bhojpuri, all substrate languages.8 Syea analyzed so in (13), repeated here as (32), as a possessive pronoun, co-indexed with the NP zistoire: (32)

[soi finition]k mo zistoirei ek 1. storyi ek [3.i end]k *its end of my story *sa fin de mon histoire

He claims that such constructions eventually disappeared because they were at odds with universal principles of grammar. However, if so is analyzed as a definite determiner,

.

The following examples are from Syea (1994):

Twi

Ewe

(ii) Ata ne na Ata 3. mother Ata’s mother

(iii) Ale fe afo Sheep 3. foot Sheep’s foot

Bhojpuri

(iv) Tor burhia ke

Hindi

hal

Your wife 3. condition Your wife’s health

(v) Apki patni ki

sthiti

Your wife 3. condition Your wife’s health

 Diana Guillemin

and not as a co-referential pronoun in the way assumed in (32), then binding violations do not occur. Furthermore, if so is to be analyzed as a possessive pronoun co-indexed with its complement NP, one would also expect to find examples with the well established plural form of the possessor, zaute/zotte (their), such as: (33) *Zaute lezos tout zanimaux *their bones all animals

But there are no such uses of the plural form of the possessive pronoun in any of the 19th century texts. We find only Type A genitives in the early texts, as follows:9 (34)

lézos tout zanimaux qui tigue là fine manzé bones all animals  tiger   eat ‘the bones of the animals that the tiger had eaten’ “les ossements des animaux qu’avait dévorés le tigre”

(Baissac 1888: 380–1)

What looks like a Type B genitive, is in fact a tensed clauses, minus the copula: (35)

Mais Zeanne son pitit! But Jeanne 3. child ‘But Jeanne is her daughter’ “Mais Jeanne est sa fille” ≠but Jeanne’s daughter ≠mais la fille de Jeanne

(Baissac 1888: 94–5)

While so had been reanalyzed as a determiner in early MC, it retained its original meaning as a possessive pronoun, and continued to be used just like the other possessives mo, to, nu, zot (my, your, our, your/their), as in so mari (her husband) so léquér (his/her heart), so lébras (his/her arm).

.

So in modern MC

These possessive pronouns also continue to be used in modern MC, exactly as in the early creole, and both Type A and Type B genitives still co-exist in the creole. However, genitive constructions with so as a determiner are no longer grammatical: (36)

*So sapo mo papa * hat 1. father

. I have come across only one example of Type B genitives in the early texts (19th century):

(vi)

grand Misié

son

causé

big master 3. talk ‘The master’s words’ “Les paroles du maître”

(Descroizilles 1867)

Chapter 13. A look at so in Mauritian Creole 

This change correlates with the emergence of other determiner elements, more specifically, that of the plural marker bann, which first occurs in Anderson’s translation of the Bible, where a large number of [+definite] plural NPs are translated as band + NP, as shown: (37)

(38)

éne dan band profet one in  prophet ‘one of the prophets’ “quelqu’un des prophètes”.10

(1993: 18) (1955: Ch. 16, v. 14, p. 1480)

E kan band cef prétr avec zancien ti accuz li and when  chief priest with elder  accuse 3. ‘And when he was accused by the chief priests and elders,’ (1993: 32) “Puis, tandis qu’il était accusé par les grands prêtres et les anciens,” (1955: Ch. 27, v. 12, p. 1500)

In modern MC, a [−definite] singular NP is marked by enn, just like in the early creole. A bare noun in an argument position is always [+definite] singular. Its definiteness feature is either anaphoric, when it has a discourse antecedent, or associative, when it is qualified by a relative clause or modified by a genitive, as in: (39)

Sapo mo papa hat 1. father ‘My father’s hat’ “Le chapeau de mon père”

The prenominal plural marker bann eliminates all ambiguity with regard to a singular or plural interpretation. An NP modified by bann is [−definite] plural in existential constructions, and [+definite] plural when in an argument position, or when the noun is modified by a genitive, as in: (40)

Bann sapo mo papa  hat 1. father ‘My father’s hats’ “Les chapeaux de mon père”

Following the grammaticalization of bann, the use of so to force a singular reading of bare nouns became redundant, but the use of this particle to mark emphasis in discourse persists in modern MC, where it is showing up in phrases without an overt modifier: (41)

.

Sa travay la  job  bizin dan sa need in 

enn travay lesuf. So premye kiksoz ki u a job stamina.  first thing  2 travay la lesuf job  stamina

Where des is a contraction of de les, i.e., prophètes is [+definite] plural.

 Diana Guillemin

‘This job is one that requires stamina. The first thing that you need in this job is stamina.’ “Ce travail est un travail qui réclame du souffle. La première chose dont vous avez besoin dans ce travail, c’est du souffle.” (1980: 8)

A bare noun, though grammatical, would fail to emphasize that the first and foremost requirement for this very difficult job is plenty of stamina. The use of so is similarly optional in the following example: (42)

Enn nui damour dan zot nik damour, apré so landémain matin11 a night of.love in 3. nest of.love, after  next.day morning, ‘A night of love in their love nest, and the next morning,’ (lit. ‘its next morning’) “Une nuit d’amour dans leur nid d’amour, ensuite, le lendemain matin,” (Maingard 2002: 23)

However, the ‘next morning’ in the narrative happens to be a decisive moment in the story, which the author has chosen to emphasize this fact by using so, and thus draw the attention of the hearer to that tragic moment in time. Although so continues to be used in MC as a 3 possessive pronoun in modern MC, it seems that it has totally lost its original ‘associative’ anaphoric meaning when used as a determiner. The following example, where so modifies plural NPs, provides evidence that this morpheme is now totally bleached of its singular feature when used as an emphatic determiner. The use of bann instead of so would have been equally grammatical: (43)

Me, mem avan solda reysi grinp sa montayn si difisil la, sef tribi donn lord so disip mont ziska enn ros pli lao ki tu ros laba, e li obliz zot, enn apre lot, mem so zanfan, mem so vye dimunn, mem so mama tibaba dan lebra, li obliz zot tu zet zot lekor depi lor sa ros la ver enn lamor serten anba. ‘But, even before the soldiers had completed the arduous ascent, the chieftain ordered his followers up to the highest cliffs on the rock, and forced them, one by one, even the children and old people, and mothers with their babies, to throw themselves off the rock to their certain deaths below.’ (Author’s translation) “Mais, même avant que les soldats aient réussi à grimper cette montagne si escarpée, le chef de la tribu donna l’ordre à ses disciples de monter jusqu’au rocher le plus haut là bas, et il les obligea, l’un après l’autre, même les enfants, même les vieilles personnes, même les mères avec leurs bébés dans les bras, il les obligea tous de se jeter du rocher vers une mort certaine en bas.” (Moss 2000: 5)

The narrators have chosen so to emphasize the tragic fact everybody, even the children, the old people, and the mothers with young babies, had to throw themselves over

.

Where lendemain matin is a fused expression and not a genitive construction.

Chapter 13. A look at so in Mauritian Creole 

the cliff to their certain death. The emphatic determiner serves to draw our attention to the plight of these helpless individuals, caught in a tragic situation.

. Conclusion What does this analysis tell us about the process of grammaticalization and creolization? With regard to determiner systems, Fraurud suggests that it may well be that the ‘lack of articles in a language is something that it will strive to make amends for, either by developing articles or by using some other means for signaling definiteness’ (2001: 262). She does not necessarily agree with this statement, on the grounds that many languages do not have articles, but still have the means to express definiteness either through word order, or specific morphemes, other than articles. This could hardly have been the case with the early creole, which had retained the strict SVO word order of its lexifier, but lost the determiners, as well as all inflectional morphology. The determinative use of the possessive pronoun in 19th century MC seems to have been motivated by both grammatical and pragmatic considerations, when new determiners were still emerging, and the language lacked the means to clearly express the semantic contrasts of [±definiteness] and [±plurality]. Prior to the grammaticalization of the plural marker bann, so was reanalyzed as a [+definite] determiner, and like the null ?, it forced a singular interpretation of its complement NP. It picked out a unique referent in the discourse, that needed emphasis for clear interpretation. It was commonly used with genitive constructions, where the interpretation of the head noun may have been ambiguous with regard to a singular or plural interpretation, and its use with these genitive constructions became redundant following the grammaticalization of bann. In modern MC, the use of so with both singular and plural NPs seems to be driven purely by pragmatic considerations, to mark emphasis in discourse, when such emphasis cannot be succinctly expressed by other morphosyntactic means. I assume that the features (in)definiteness, singularity and plurality to be universal semantic notions that must find expression in natural language, and which manifest morphosyntactic reflexes in the grammar of a language. The speakers of the early creole clearly lacked to means to express these contrasts, yet the process of grammaticalization of new determiner elements suggests that they had access to the semantics associated with the new morphemes. The process of grammaticalization, or creolization, may thus well be the manifestation of ‘internal’ language change driven by universal principles of grammar. The aim of this paper has been to understand what motivated the determinative use of a possessive pronoun in a new creole, and to situate such use within the context of the grammaticalization of other determiner elements in MC. It has also pointed to a similar phenomenon in other languages of the world, including other creoles, where a 3 pronoun serves to mark both plurality and definiteness. I am not aware of other typological studies on the determinative use of possessive pronouns, other than works

 Diana Guillemin

cited in this paper. Future research may help establish the semantic and syntactic factors that motivated the reanalysis of these pronouns as definite determiners, and why it is that plural pronouns are reanalyzed in Haitian and Guyanese Creole, as opposed to a singular pronoun in MC.12

References 1888. L’Évangil selon S. Marc (dan langaz créol Maurice). In The Gospel according to St. Mark (in Mauritian Creole). London: British and Foreign Bible Society. 1955. La Sainte Bible. Traduite en Français sous la direction de l’École Biblique de Jérusalem. Paris: Editions du Cerf. 1980. Bord la mer. Port Louis, Mauritius: Port-Louis Harbour and Docks Workers Union (PLHDWU). 1993. The Holy Bible, containing the Old and New Testaments: Catholic Bible Press. Ah-Vee, A. 2002. Lartik dan ‘Kwin syantifik’ ki finn paret dan Lagazet Lalit de Klas. In Ledikasyon pu Travayer (ed.). Langaz kreol zordi. Port-Louis, Mauritius: Ledikasyon pu Travayer. 72–3. Anderson, S.H. 1885. L’Evangile sélon S. Matthié (dan langaz créol Maurice). The Gospel according to St Matthew (in Mauritian creole). London: British and Foreign Bible Society. Baissac, C. 1880. Étude sur le patois créole Mauricien. Nancy: Imprimerie Berger Levrault. Baissac, C. 1888. Le folk-lore de l’Ile Maurice. vol. XXVII: Littératures populaires de toutes les nations. Paris: G. P. Maisonneuve and Larose. Baker, P. 1984. The significance of agglutinated French articles in the creole languages of the Indian Ocean and elsewhere. York Papers in Linguistics II. 18–29. Carpooran, A. 2002. Konsep drwa langaz ek drwa lingwistik: propozisyon pou enn aplikasyon dan Moris. In Ledikasyon pu Travayer (ed.). Langaz kreol zordi. Port-Louis, Mauritius: Ledikasyon pu Travayer. Carpooran, A. 2005. Diksioner morisien: Version prototip Let A ziska E. Quatre Bornes, Mauritius: Bartholdi. . The source of the morpheme yo in Haitian Creole has been identified as the French pronoun eux, (Goodman 1964; Sylvain 1936; Lefebvre 2001). However, Lefebvre also stipulates that, in its use of yo as a plural and definiteness marker, HC patterns with one of its substrate languages, Ewe, where the morpheme wo also indicates both notions (2001: 8). While the source of se in Martinique and Guadeloupean Creoles has been identified as the French plural demonstrative ces, Deprez (fc.) suggests that ‘alternatively, se could perhaps just as well derive from the French possessive pronoun ses.’ She further comments that although ‘the se marker does not by itself encode definiteness, it must obligatorily be associated with a definiteness marker’. It must occur with the post-nominal marker la as shown:

(vii)

a. b.

se tab nla (Martinique) ‘the tables’ sé boug la (Guadeloupe) ‘the men’

Chapter 13. A look at so in Mauritian Creole  Chaudenson, R. 1981. Textes créoles anciens: la Réunion et Ile Maurice: comparaison et essai d’analyse. Hamburg: H. Buske. DeGraff, M. 1994. The morphology-syntax interface in creolization (and diachrony). Studies in the Linguistic Sciences 24. 115–31. DeGraff, M. forthcoming. Kreyol Ayisyen, or Haitian Creole (‘Creole French’). Available at: http://web.mit.edu/linguistics/www/degraff/hc-sketch.pdf [Accessed March 2003] Deprez, V. 2005. Morphological number, semantic number and bare nouns. Lingua 115. 857–83. Deprez, V. forthcoming. Probing the structuring role of grammaticalization: nominal constituents in French Lexifier Creoles. Available at: http://www.rci.rutgers.edu/çdeprez/FLC.pdf [Accessed January 2007] Descroizilles, H.C. 1867. Navire fine engaze, or The Mauritius in danger. In Chaudenson, R. (ed.). Textes créoles anciens: (La Réunion et Ile Maurice): comparaison et essai d’analyse. Hamburg: Helmut. 125–31. Fraurud, K. 2001. Possessives with extensive use: A source of definite articles? In Baron, I., Herslund, M. and Sørensen, F. (eds). Dimensions of possession. Amsterdam: John Benjamins. 243–67. Goodman, M. 1964. A comparative study of French creole dialects. The Hague: Mouton. Grant, A.P. 1995. Article agglutination a in Creole French: a wider perspective. In Baker, P. (ed.). From contact to creole and beyond. London: University of Westminster Press. 149–76. Haspelmath, M., Dryer, M.S., Gil, D., and Comrie, B. (eds). 2005. The world atlas of language structures. Oxford: Oxford University Press. Hawkins, J. 1978. Definiteness and indefiniteness: A study in reference and grammaticality prediction. London: Croom Helm. Himmelmann, N.P. 2001. Articles. In Haspelmath, M., König, E., Wulf Österreicher, W.R. (eds). Language typology and language universals: an international handbook. Berlin: Mouton de Gruyter. 831–41. Hookoomsing, V. 2002. Kreol: Lang interfas an leritaz patrimwann et modernite. In Ledikasyon pu Travayer (ed.). Langaz kreol zordi. Port-Louis, Mauritius: Ledikasyon pu Travayer. 25–30. Ledikasyon pu Travayer. 2006. Diksyoner elektronik inter-aktiv. [Accessed January 2007] Available at: http://www.kiltir.com/diksyoner/info/introdiksyon.php Lefebvre, C. 1998. Creole genesis and the acquisition of grammar: The case of Haitian Creole [Cambridge Studies in Linguistics 88]. Cambridge: Cambridge University Press. Lefebvre, C. 2001. The interplay of relexification and levelling in Creole genesis and development. Linguistics 39. 371–408. Lolliot, P. 1855. Poésies créoles. Ms. Port Louis, Mauritius. Longobardi, G. 1994. Reference and proper names: A theory of N-movement in syntax and logical form [fall]. Linguistic Inquiry 25. 609–65. Maingard, J. 2002. Lagrin tambarin: 13 ti zistoires en kréol. Port Louis, Mauritius: Editions Maurice Milsark, G. 1979. Existential sentences in English. New York: Garland. Moss, R. 2000. Le Morne: Lemorn. Port Louis, Mauritius: Ledikasyon pu Travayer. Nicolay, W. 1835. Proclamation pour noirs esclaves dans Maurice. Traduction en langue créole. In Robert Chaudenson (ed.). Textes créoles anciens (La Réunion et Ile Maurice): Comparaison et essai d’analyse. 118–20.

 Diana Guillemin Robert, P. 1972. Petit Robert: Dictionnaire alphabétique et analogique de la langue française. Paris: Dictionnaire le Robert. Strandquist, R.E. 2005. Article incorporation in Mauritian Creole. MS. Department of Linguistics, University of Victoria: MA. Syea, A. 1994. The development of genitives in Mauritian Creole. In Adone, D. and Plag, I. (eds). Creolization and language change. Tubingen: Max Niemeyer Verlag. 85–97. Sylvain, S. 1936. Le créole haitien: Morphologie et syntaxe. Wetteren, Belgium: Imprimerie de Meester.

 

Chinese Spanish in 19th-century Cuba Documenting sociohistorical context* Don E. Walicek University of Puerto Rico, Río Piedras

Between 1847 and 1874, the international trade of Asian indentured laborers often referred to as the ‘Coolie Trade’ took the first large-scale arrivals of Chinese to colonial Cuba. During this period, there existed a system of racial hierarchy that divided the island’s population into three main groups: whites, free people of colour, and slaves. The 125,000 Chinese arriving in this period added another dimension to the cultural and linguistic heterogeneity of the island. This work pulls from archival information that chronicles this trans-Atlantic migration, including metalinguistic commentaries found in interviews done with these laborers. These records provide a micro-level view of the plantation system relevant for theoretical debates and related work within the field of Creole Studies. This work pays special attention to the relations between Chinese and African-born slaves and those Chinese who became bilingual in Spanish and served as intermediaries for their fellow countrymen.

. Introduction Between 1847 and 1874, the international trade of Asian indentured laborers known as the ‘Coolie Trade’ took the first large-scale arrivals of Chinese to colonial Cuba. During this period, there existed a system of racial hierarchy that divided the Caribbean island’s population into three main groups: whites, free people of color, and slaves. The roughly 125,000 Chinese who arrived in this period added another dimension to the linguistic and cultural heterogeneity of the island. The main problem addressed in this paper is that in describing the resulting situation of language contact, linguists have not taken full advantage of the primary-source archival information that chronicles this trans-Atlantic migration. These records, many * I extend my appreciation to Nicholas Faraclas, Armin Schwegler, and an anonymous reviewer for comments on an earlier version of this paper. Any remaining errors are my own and will hopefully be discussed in future work on this topic. A draft of this essay was presented at the meeting of the Society for Pidgin and Creole Languages in Oakland, CA, USA 7–8 January 2005.

 Don E. Walicek

of which have become widely available in the second half of the twentieth century, provide a surprisingly detailed micro-level view of the plantation system that is highly relevant for creolistics. As noted by Bakker (2002: 75), many claims about plantation life made by linguists who study language contact and linguistic creolization in the context of colonialism in the Caribbean are “purely speculative and largely unscientific”, ignoring the work of historians and the rich literature that they have produced on relevant topics.1 Arends (2001: 291) may provide a more nuanced view of this problem when he states that important progress has been made at the investigation of sociohistorical factors in Creole genesis at the macro level, but that little attention has been devoted to historical factors at the micro level. The primary historical sources examined here contain surprisingly detailed information about the nineteenth-century Spanish plantation system, information that has a direct bearing on understanding linguistic contact as it occurred between the Chinese and other groups in colonial Cuba. Relating the social to the linguistic, this essay offers a four-part account of language contact for this historical event. The first and second sections below examine some of the circumstances leading up to the Trade in China and Cuba, respectively. The third part provides an overview of a previous linguistic analysis of Chinese Cuban Spanish.2 The final and most extensive section of this paper utilizes primary source data, particularly interviews with the Chinese laborers themselves, to describe the experiences of these indentured laborers as they relate to sociolinguistic phenomena. The testimonies cited below have been included because they include migrants’ comments on language use and language contact.3

.

British Imperialism and the international dimension of the trade

Great Britain’s presence in Asia’s treaty ports made the onset of the “Coolie Trade” to Latin America possible. In 1842, approximately five years before the first Chinese indentured laborer reached Cuba, China and Great Britain signed the Treaty of Nanking, the international agreement that marked China’s surrender and the official end of the Second Opium War (1840–1842).4 The treaty held Europeans and other . Bakker (2002) mentions a few notable exceptions, creolists who study the situations described in historical sources, specifically the work of Philip Baker, John Singler, and Jacques Arends. . I refer to the varieties of Spanish used by Chinese laborers in nineteenth-century Cuba as Chinese Cuban Spanish and reject its classification as a pidgin. In contrast, Lipski (2000a, b) uses the term “Chinese-Cuban Pidgin Spanish.” . The conventions of romanized spelling and capitalization of proper names that appear in The Report have not been altered. . Although several international agreements eventually brought bloodshed between Great Britain and China to a close, they came only when China was forced to surrender. In addition to

Chapter 14. Chinese Spanish in 19th-century Cuba 

Westerners living and doing business in China exempt from Chinese laws, and enabled them to live and trade in China as they wished.5 The system of extraterritoriality that was established allowed for the majority of Chinese migrants to be exported from China’s five major treaty ports: Amoy, Macao, Swataw, Hong Kong, and Wampoa. During this era, the treaty ports were among China’s fastest growing cities. The voyages taking laborers to Cuba were expeditions organized as independent business ventures through joint agreements among entrepreneurs and government officials in southern China, the Philippines, Spain, and Cuba (Corbitt 1971, Turner 1974). Table 1. Distribution of the 19th-century trade of Chinese laborers, 1847–1874. Country Brazil British Guiana British Honduras Chile Costa Rica Ecuador French Guiana Guadeloupe Jamaica Martinique Panama Peru Suriname Trinidad Subtotal Cuba Total

Laborers surviving passage 983 13,830 474 2,000 650 300 3,331 208 515 426 1,000 110,000 2,839 2,643 139,199 124,815 264,014

Sources: Total for Cuba comes from Table 2; statistics for other countries comes from Huesmann (1991: 219a).

Men who came to be known as chy chau tau (“swineherds” in Cantonese) worked as recruiters of potential immigrants in Kwangtung, the province from which most migrants originated, and the areas surrounding these five southern ports. The chy chau tau, many who spoke English, are examples of “treaty-port Chinese”, a small minority who followed “Western ways” in pursuit of modernization and financial gain (Jiménez Pastrana 1963: 26; Yung 1909: 54, 176). The testimonies of men who were taken to

the Treaty of Nanking, other treaties guaranteed and negotiated foreign authorities’ terms of negotiation with China; among them are the American Treaty of Wang-hea and the French Treaty of Wampoa. .

The laws that non-Chinese were subject to were usually those of their home countries.

 Don E. Walicek

Cuba from Kwangtung and Fukien relate that crimps, lower-scale agents, worked for the chy chau tau and assisted them in their efforts to obtain migrants. These crimp runners directly engaged in entrapping laborers. Some were men who had travelled from their provinces of birth to find employment in the coastal region of southern China, as were many of those they kidnapped. A long history of circular migration to places like the Philippines and unfamiliarity with the notion that they could be sold as property contributed to the vulnerability of potential migrants (Pan 1990: 32–8). Cheng A-mou, a victim of the Trade recalls in an 1876 testimony recorded in Cuba, we observed on the signboards of foreign buildings the words ‘agencies for the engagement of laborers,’ and believed that they truthfully described the nature of the establishments, little expecting that having once entered the latter, exit would be denied us, [. . .] it became evident that we were not to be engaged as laborers but to be sold as slaves. (Helly 1993: 37)

In some instances the runners deluded even agents, doctors, translators, and linguists who were employed on voyages to the West, selling them as laborers once they reached the Caribbean.6 Table 2. Chinese migration to Cuba, 1847–1874. Year 1847 1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866 1867 1868 1869 1870

Departing from China 612 5,150 1,750 3,130 6,152 10,116 10,414 8,459 7,204 7,252 356 1,045 2,664 6,794 13,368 15,616 8,100 6,720 1,312

Deaths during voyage 41 843 39 145 1,084 1,575 3,019 1,345 1,011 279 12 93 511 394 977 1,353 732 1,060 85

Surviving arrivals 571 4,307 1,711 2,985 5,068 8,541 7,395 7,114 6,193 6,973 344 952 2,153 6,00 12,391 14,263 7,368 5,660 1,227 (Continued)

. Examples include Ch’en Ming-Yuän and Yao Wen, two physicians sold as laborers upon arrival in Havana, and Chang-T’ai-lung, a linguist who met the same fate; see Helly (1993: 47).

Chapter 14. Chinese Spanish in 19th-century Cuba 

Table 2. Continued. Departing from China

Year 1871 1872 1873 1874 Totals

1,577 8,915 5,856 2,863 141,391

Deaths during voyage 89 755 763 373 16,576

Surviving arrivals 1,488 140 5,093 2,490 124,815

Source: Pérez de la Riva (1967: 6).

As shown by Table 1, Cuba is merely one of fifteen locations to which the Chinese were taken. However, the number of indentured laborers to reach the island was greater than that of any other destination, as shown in Table 2. About half of all the Chinese to reach Latin America and the Caribbean alive were taken to Cuba.7 The overwhelming majority of these servants were male. Only a handful of Chinese females, a dozen or less, migrated to the island.8

.

Changes in Cuba’s social order

Forced migration of Asian laborers to Cuba took place at a time when the island’s sociocultural order was changing rapidly. The gradual abolition of African slavery and the push for independence from Spain by Cuban-born whites, criollos, were the main factors driving this change. In this period, criollos were clearly distinguished from peninsulares, persons born in Spain. When the 1841 Cuban census revealed for the first time that blacks outnumbered whites, white criollos, especially those who were merchants, became increasingly uneasy about threats to their dominant position in society. Planters and others of European descent feared a race war, as it had been repeatedly predicted that blacks would rebel against them like had happened in Haiti. Some white intellectuals desiring independence claimed that Spain deliberately inflated the number of blacks on the island to frighten and control whites.9 While groups defined as white came from various backgrounds and maintained diverse interests, their legal status as such ascribed them economic and social privileges that non-whites .

Sources for this statistic are Pérez de la Riva (1967: 6) and Huesmann (1991: 219a).

. Unfortunately, no women were interviewed by the commissioners in 1874. In the nineteenth century most females served as domestic laborers in Havana. In the twentieth century a larger number of females arrived. . According to an 1836 correspondence by Van Ness, the U.S. envoy to Spain, Caltrava, the Spanish minister at the time, believed that the fear of blacks “[. . .] worth an army of 100,000 men” and would prevent white criollos from making any revolutionary attempts”. Quoted by Paquette (1988: 80).

 Don E. Walicek

generally did not have. In the words of Pérez (1988: 93), whites were “[. . .] united in their defense of sugar and slavery, upon which rested prosperity, property, and privilege”. As shown in the graph below, the number of Africans arriving in Cuba began to decline in the years preceding the first large-scale arrival of Chinese in 1847. This augmented planters’ fears of a shortage of exploitable labor.10 Blacks on the island included two main groups: slaves who had recently arrived from Africa, referred to in Spanish as bozales or negros de nación, and those born in the Caribbean. As the trade of African laborers continued, persons of African descent born in Cuba were increasingly working off the plantations, and the population of free people of color of mixed racial descent (gente de color) grew. The number of African slaves imported each year did increase somewhat in the 1860s, reaching 24,895 and 23,964 in 1860 and 1861, respectively (Murray 1980: 244). Though the importation of black slaves to Cuba was officially outlawed in 1865, both legal African slavery (limited to certain groups defined in European law) and the clandestine trade of Africans continued after that date (see D’Auvergne 1933: 245).11 It was only in 1873 that the last known legal shipment of African slaves reached Cuba. And despite legislation directed at manumission and the abolition of the Atlantic Slave Trade, no laws prevented the importation of indentured laborers from Asia. These are just some of the pull factors that brought Chinese to the Pearl of the Antilles, as Cuba was sometimes called. Graph 1.

African Slaves Imported to Cuba (1840-1847)

Number of Persons

14.470

10.000

9.776 8.000

3.000 1.300

1.500 1.000

1840

1841

1842

1843

1844

1845

1846

1847

Years

From the point of view of planters and others in need of exploitable labor, four conditions framed the possibilities of Chinese indenture and made them ideal laborers in . Statistics for Graph 1 are estimates made by the British Foreign Office and the British Commissioners, both in Havana; see Murray (1980: 244). . Paquette (1988: 35–6) states that in the 1840s slave shipments from the Bight of Benin and Mozambique increased, while those from the Bight of Biafra and the Gold Coast declined. D’Auvergne (1933) notes that more than 3,140 African slaves arrived in 1842 and that shipments continued after this date.

Chapter 14. Chinese Spanish in 19th-century Cuba 

the eyes of many: first, they were available in large numbers; second, unlike Africans, they remained legally unprotected by a Western power; and third, important in allaying the planters’ afrophobia, Asians were not considered black. In fact, apart from the stipulations of indenture, the Chinese laborers were to be voluntary migrants and legally free. Therefore, for white planters, merchants, and colonial officials eager to find a solution to what they saw as perennial labor problems, Chinese labor in Cuba symbolized the possibility of future economic profits and a relatively smooth transition from slave to “free” labor. And finally, significant for Spanish authorities in Madrid because it could at least temporarily reduce the island’s labor shortage, the introduction of Asian labor represented the possibility of easing tensions between Crown and colony. In the Royal Decree of 1847 the Spanish Crown encouraged Spaniards in Cuba to import large numbers of Chinese. Despite evidence of a long history of racial conflict between the Chinese and Spaniards in the Philippines, the decree asserts that the experience of the Filipino colony proved Asians to be “docile, industrious, frugal, temperate, hardened to rural labor” and accustomed to a “rigorous climate” (British and Foreign State Papers, 1217). One Spanish official claimed that this history had demonstrated that the Chinese “. . . were more intelligent, although not so robust and athletic as the negro” and required treatment “combined with justice and kindness, but by no means relaxed in character”; according to the Spanish Crown, proper treatment could make the Chinese “most valuable servants, and in a great measure supersede the slave labour of the island [Cuba]” (quoted by Corbitt 1971: 3). Once the Trade was well underway, increased controversy surrounded it, in China, Europe, and in the Americas (Yung 1909: 176). International criticism led to the closure of the ports of Amoy, Swataw, Hong Kong, and Wampoa in 1859. In 1862 the U.S. and British governments prohibited their ships from engaging in the transport of involuntary migrants. The Trade was left open only to Macao, the oldest European settlement in East Asia and southern China’s hub of European activity. In the years after 1859, as earlier in the Trade, the majority of recruits were from the Four Districts area in Kwangtung (Lai 1993: 40; Martin 1939: 18).

.

“Chinese-Cuban Pidgin Spanish” and Afrogenesis

In Lipski (2000b) two provocative hypotheses are put forward: (1) that Macao Creole Portuguese played a significant role in the communication strategies of Chinese laborers, and (2) that surviving examples of habla del chino (“speech [Spanish] of the Chinese”) reflect “creoloid constructions” in Chinese Cuban-Pidgin Spanish and bozal Spanish. In describing the language of Chinese laborers in nineteenth-century Cuba, Lipski (2000b) pays attention to the possible linguistic input that they could have received both in Asia and the Caribbean. In addition to recognizing Macao as the most significant point of departure, he isolates a period that might have been especially important in determining what linguistic knowledge the migrants arrived with, the time just

 Don E. Walicek

preceding their departure from Asia. The suggestion that some spoke a Portuguese Creole is made: Since most of the Chinese were recruited through the Portuguese colony of Macao, where a Portuguese pidgin and creole was spoken among the native Chinese population, there exists the possibility that some of the Chinese workers added their knowledge of a Portuguese creole to the already rich mix of creole and creoloid elements present in nineteenth-century Cuba. (Lipski 2000b: 226)

Lipski supports the argument that some Chinese migrants spoke a Portuguese Creole by stating that many of them spent about one year in Macao awaiting the departure of ships going to Spanish America, a claim discussed in more detail below. He states that such an extended period of confinement might have given them ample time to learn Macao Creole Portuguese (hereafter referred to as MCP). Lipski relates the acquisition of a Portuguese pidgin on the part of the Chinese to the “Afro-Creole debate”. Central to this argument is the proposal that bozal Spanish, the variety of Spanish said to have been spoken in the Caribbean by newly arrived African slaves, originated in West Africa when Africans there learned a Portuguese-influenced Pidgin that was used throughout the Atlantic Slave Trade. But how do the Chinese and the linguistic varieties they may have spoken fit into this complex puzzle? Lipski (2000b: 222) suggests: The implications for the study of bozal Spanish are immediate and far-reaching, for if it can be demonstrated that Chinese workers in Cuba brought with them, at least some fragments of Macao creole Portuguese, and added it to the linguistic mix in which bozal Spanish was formed in the Caribbean, this provides yet another route of entry of certain creoloid constructions in attestations of AfroCuban Spanish.12

Lipski holds that MCP shares with Chinese Cuban Spanish some features common among Afro-European Creoles found in the Caribbean. In his analysis of passages from a “Chinese Spanish corpus” he finds that the variety of Spanish spoken by the Chinese exhibits at least three grammatical features also found in MCP: a.

the use of disjunctive pronouns instead of clitics Comandante Lupelto, pa mi no mila . . . “Major Ruperto, I didn’t see anything” (Feijóo 1981: 145; quoted in Lipski 2000b: 223)

b.

the elimination of definite articles Mila, Ginilá, coje tlella pa ti, dásela otlo gente, que yo no quiele dejá máuse que tu no lá pa mi… “Look, General, take these stars for yourself, give them to somebody else; I don’t want to give up the mauser that you won’t let me have” (Consuegra and Guzmán 1930: 163-4; quoted in Lipski 2000b: 224)

. Two related ideas are also put forward by Lipski. First, he argues that MCP was imported (rather than formed or transformed in Malacca) to Macao from Malacca. Second, Lipski proposes that black Africans present in Malacca could have been previously exposed to some of the linguistic features of a Malacca-derived variety of MCP.

Chapter 14. Chinese Spanish in 19th-century Cuba 

c.

the omission of verbs Ciudadano cubano tó . . . Yo digo, junto tó nosotlo . . . “Cuban citizens, all [of us] . . . I say, all of us [are] together” (Jiménez-Pastrana 1983: 92; quoted in Lipski 2000b: 224).

The proposal that MCP and Chinese Cuban Spanish share these features is used to support two hypotheses: (1) that a significant number of Chinese migrants were familiar with MCP, and (2) that MCP influenced Chinese Cuban Spanish, and possibly AfroCuban Spanish. In arguing for the existence of cross-fertilization and convergence among the Spanish of the Chinese, MCP, and Afro-Cuban Spanish, Lipski focuses on the proposal that “the linguistic model for the Chinese was frequently the speech of bozales who had already learned some Spanish, as well as the Spanish spoken as a second language by workers from (Afro-American creole speaking) Caribbean territories” (2000b: 217). Lipski takes his “Chinese-Cuban Pidgin Spanish” data from a diverse group of sources. These works, usually written by native speakers of Spanish in the century that followed, typically feature a stereotypical male character, often referred to simply as el chino (“the Chinese”), who speaks not an Asian language but a non-standard variety of Castilian Spanish.13 The main characteristics cited for this literary variety of Spanish are a.

conversion of /r/ to /l/ in all positions e.g., miedo > mielo, ‘fear’; emperador > empelaló, ‘emperor’

(Lipski 2000b: 219)

b.

frequent use of null subjects e.g., poqué lice esa cosa ahola? “Why do [you] say those things now?” No quelé tlabajá . . . No sabel, capitán . . . Yo no sabel. “I don’t want to work . . . [I] don’t know captain . . . I don’t know” (Bueno 1959: 54–73; quoted by Lipski 2000b: 220–1)

c.

use of subject pronouns for contrastive focus or emphasis e.g., Tú tlabaja mucho. Tú tumba mucha caña y ganá mucho dinelo. “You work hard, you cut a lot of sugar cane, and you earn a lot of money” (Feijóo 1981: 153–4; quoted by Lipski 2000b: 221)

Lipski maintains that these texts feature speech forms representative of a variety of Spanish used by nineteenth-century Chinese laborers. He (2000b: 219) writes: “[s]o familiar was the habla del chino ‘Chinese talk’ to the average Cuban, that a literary stereotype quickly developed, almost always portraying the Chinese in a somewhat comical but never totally unfavorable light”. Though some of the texts in the corpus are folkloric or historical in scope, most perpetuate racial stereotypes that can be traced back to Spanish colonialism in the . Works featuring Chinese Spanish seem to have been created in contexts quite distinct from those including habla bozal from the eighteenth century, such as the surviving folk songs discussed in Arrom (1941) and Guirao [1937] 1970). Roberts (fc: 145) points to Kikongo features in habla bozal and suggests that these songs “[. . .] are consistent in their appearance with later (nineteenth century) stereotypical representations of habla bozal ”. For related work on the role of African-influenced speech in contemporary ritual practices in Cuba see Schwegler (2005).

 Don E. Walicek

Philippines in the preceding century (see Mörner 1967: 5). And though they may be characterized by a consistent set of features, it is not clear that these stereotypes are based on reliable accounts of spoken language. It is perhaps at least partially for this reason that widely distributed linguistic stereotypes are still familiar to the “average” Cuban. For these and other reasons detailed below, I argue that these representations of habla del chino fail to accurately and adequately represent Chinese interactions and verbal utterances.

.

The Report

Testimonies by Chinese indentured laborers living in Cuba provide information that is key to understanding the unique linguistic situation in which they found themselves, forcing the reexamination of a few of the assertions made in Lipski (2000b). Although primary source information of this type is indeed scarce, a wealth of it is available in The 1874 Cuba Commission Report (hereafter referred to as The Report). Published in Chinese, English, and French in 1876, 1877, and 1970 (and republished as Helly 1993), this work is one of the few sources to offer a firsthand description of the experiences of Cuba’s nineteenth-century Chinese population. The Report contains petitions and depositions collected in Cuba during 1874. That year – the twenty-seventh year of the shipment of involuntary laborers from Asia to Latin America – an international commission travelled to the island. Led by the Chinese dignitary Ch’en Lan Pin, the multilingual commission worked in cooperation with another that visited Peru. Under orders of the Qing government in Beijing, its mission was to investigate abuses against Chinese by collecting depositions from survivors of the Trade. In cooperation with Spanish authorities, representatives of the commission visited plantations, barracoons, jails, and sugar warehouses, personally accepting the written petitions and oral testimonies of the laborers over a period of approximately seven weeks. They collected depositions from a total of 2,841 Chinese. Of the men they consulted, only 90 claimed to have left China freely and only 2 attested to humane treatment by their Spanish employers. In their written introduction to the testimonies (which were conducted in Chinese), the commissioners noted that ordinarily “administrators and overseers stood by with whip in hand,” but fortunately “attempted no more overt intimidation” (Helly 1993: 121). Some of the interviews the commission conducted were only minutes long, others more extensive.14 Most of the interviews were done orally. In “Despatch to Yamên,” the preface to their summary of interviews, the commissioners write, “[. . .] these men were seen by us all, and these words were heard by us all” (see Helly 1993: 34). However, other evidence indicates that some migrants did not respond to the inquiry verbally. Fearful of . Despite Lipski’s assertion of the contrary, it is clear that the overwhelming majority of the Chinese were involuntary migrants. Those that did leave voluntarily did so only because of the lies told them about the nature of their indenture and their future status in Cuba.

Chapter 14. Chinese Spanish in 19th-century Cuba 

reprimands and beatings at the hands of overseers, these laborers believed that written communication would be quicker, safer, and perhaps even more informative than face-to-face oral interviews. P’an To-li, for example, delivered his testimony in writing. Representing himself and two others, his testimony states, “Hearing of the inquiry which you are instituting we intended to in person present our statement, but our employer said that he would only allow (other) eight men to visit you, and as his severity is great we did not dare to disregard the order and therefore transmit this written petition” (Helly 1993: 121). In other cases, migrants found out that the commission was coming only shortly before its arrival, composed a joint list of grievances that they all signed, and then sent a representative to deliver it.

.

The linguistic backgrounds of migrants

While an unknown percentage of migrants may have been proficient in MCP, data from The Report suggest that the majority were not. There are several reasons for this. First of all, while The Report confirms Lipski 2000b’s assertion that a large proportion of the Chinese in Cuba were “speakers of Cantonese,” it also indicates that the pool of migrants included a significant number of natives from Fukien, Hu-kwang, Kiangsu, Manila, Shanghai, and Cheh-kian (Helly 1993: 113). These are regions identified as zones where Cantonese was not the dominant language (Ramsey 1987: 21–26). Martin (1939: 3–6) also underscores the substantial diversity within the group of languages spoken by Cuba’s Chinese populations and rejects the idea that they can be described as one spoken in language. Second, thousands of migrants did not live or work in or in proximity to the area around Macao where MCP was spoken. Third, migration internal to China had increased the diversity of languages spoken in southern China so that varieties earlier found elsewhere were also spoken in Kwangtung. Coupled with the fact that not all of the ships involved in the Trade left from Macao, when contextualized historically these three factors suggest that a substantial number of Chinese migrants would not have known MCP. Lipski’s focus is, however, on those who did leave from the port of Macao. He shifts attention away from the importance of regional origin as an indicator of linguistic ability, and notes that few Macao natives were actually taken to Cuba. Furthermore, the importance of linguistic interaction between would-be migrants and MCP speakers is underscored. As noted above, he speculates that many of those kidnapped and eventually taken to Cuba might have learned MCP during the year that they purportedly spent awaiting departure. Nevertheless, interviews with migrants suggest a different scenario. The Report alludes to a situation in which would-be migrants were often shipped out quickly and indicates that communication was highly restricted in the area where they waited.15 . It is likely that the length of the period ships spent waiting in Asia was affected by a number of factors, including the season. The seasons affected ocean currents and therefore travel by ship. Labor demands of the sugar harvest also varied according to the time of the year.

 Don E. Walicek

A petition signed by Liu A-shou and four others states that they left shortly after they were lured to Macao and abducted. It notes, “we were decoyed to Macao barracoons, and though not inspected by any Portuguese officials, were, after the evening meal, – our queues having been tied together, and guarded by foreign soldiers armed with firearms, – forced to embark, whilst no one heeded the cries for aid we uttered on the way” (Helly 1993: 37). Testimonies indicate that quick departures and waiting periods usually lasting less than a month were the norm and that this facilitated the crimps’ efforts to deceive and conceal unwilling migrants. Moreover, in the case of Liu and countless others, the conditions in barracoons were far from conducive for learning a new language. Consider the description offered by Yeh Fu-chün, who begins his recollection with the image of a foreigner closing gates behind him: [. . .] As all exit was prevented we perceived how we had been betrayed but their was no remedy; in the same chambers were more than one hundred others, most of whom passed their days and nights in tears, whilst some were dripping in blood, – the result of chastisements inflicted on account of a suspected intention of escape, or of a declaration of unwillingness, when interrogated by a Portuguese inspector. (Helly 1993: 39)

Yeh’s testimony suggests that there was little, if any, substantial dialogue between the person closing the gate (possibly a speaker of MCP, Portuguese, and/or a variety of Chinese) and those who were detained. Of course there was undoubtedly verbal exchange among those trapped in these chambers and other holding areas, many of whom were probably multilingual in several Chinese languages; however, it seems that unlikely that MCP or Portuguese was generally used for sustained interactions and conversations. Moreover, it is not altogether clear that the speakers involved would have used MCP, a variety that Holm (1989: 296–8) indicates was used mainly by the minority Eurasian population in Macao for trade with Chinese merchants in Canton, rather than a variety of Chinese commonly used for interethnic communication among the diverse populations found in the province of Kwangtung and surrounding areas.

.

Linguistic difference and the organization of the trade

Testimonies in The Report also shed light on the migrants’ language abilities at the time of their departure, confirming the existence of a substantial degree of linguistic diversity among them. T’ang Yü and six others comment on a situation also reported by others, stating “the contracts were read to us but we did not know what they contained” (Helly 1993: 42). The initial contracts that these men refer to were often printed in two languages, one side in Portuguese (or sometimes Spanish), and the other side in Chinese characters.16 Given that speakers of disparate varieties of . Article 31 of the Regulations of September 14, 1872 stipulates “. . . the contract shall be made out in Spanish”. Some migrants were apparently required to sign an initial contract in China and one in Cuba. See Helly (1993: 145).

Chapter 14. Chinese Spanish in 19th-century Cuba 

Chinese shared a written language, in some circumstances these documents were relatively accessible to migrants from diverse linguistic backgrounds; however, not all of those indentured were literate and many were not allowed to see their contracts, as is evident in a petition signed by Fêng A-kai and six others: “The contract was read to me and though it was read to me but I did not comprehend what the nature of the document was; and besides I myself cannot read” (Helly 1993: 42). Even in instances when they were read, a variety of linguistic factors and abuse at the hands of Europeans conspired to prevent migrants from understanding them (whether they were read in “Chinese” or another language). As Fêng’s recollection clarifies, oral communication was sometimes impeded by the diversity of Chinese languages among migrants and the fact that some were mutually unintelligible (Martin 1939: 3–4, 18). Whether the contracts were usually read in Portuguese, Spanish, or a variety of Chinese is unclear, but the repeated reference to foreigners and interpreters, many of whom are indeed likely to have spoken Portuguese and/or MCP, confirms the presence of languages that many would-be migrants did not understand. Ch‘ên Lung, for example, recalls the role of an interpreter in his kidnapping, stating, “when inspected by the Portuguese official, I declared that I was unwilling to go abroad. The interpreter then told me that if I persisted in my refusal I would be confined under a privy for three years” (Helly 1993: 40). Other evidence indicates that a smaller portion of the migrants may have spoken MCP and Portuguese. Li Chao-ch’un, possibly bilingual and in physical possession of a contract of indenture, noted that much was passed over in the process of translation and that only a few clauses of the contract were read aloud. Privy to information communicated in the language of their masters and others in positions of power, bilinguals like Li shared knowledge of translators skipping over passages with their peers (see Helly 1993: 41). Additional information describing recruitment and linguistic barriers comes from one of the agents involved in organizing the trade. According to Francisco Abella, an entrepreneur who financed the transportation of thousands of laborers, “the Portuguese authorities would address a group of coolies through an interpreter, who often knew only one of the eleven dialects spoken in China. . .” (quoted in Corbitt 1971: 41). Abella verifies the importance of Chinese languages but does not make reference to MCP. His declaration also points to the existence of a substantial degree of linguistic diversity among the Chinese. In addition, it suggests that officials could not communicate with speakers of all “dialects” and that this facilitated the deception of some migrants. The Report refers to occasions when officials and Chinese crimps took advantage of migrants’ ignorance of European languages and Western culture. In one case, Hsü A-fa, a would-be migrant unfamiliar with Latin America, asked where Havana was. Told that “Havana” was the name of a ship he signed a contract of indenture, he believed that he was merely engaging for service aboard a vessel (Helly 1993: 42). Others were misled by claims such as the one that “the eight foreign years specified in the contracts were equivalent to only four Chinese [years], and that at the termination of the latter

 Don E. Walicek

we would be free” (Helly 1993: 37, 42).17 In hundreds of other cases, Chinese were given contracts with their names written in the Roman alphabet rather than Chinese script, only to discover later, when they became literate in Spanish (or obtained access to their contract or someone who could read it for them), that that these were fictitious names rather than their own. Tactics such as these point to ways in which those in charge profited from and exploited migrants’ lack of familiarity with European languages and traditions.

.

The role of interpreters on voyages to the West

From the onset of the Trade, linguistic difference also affected the organization of voyages to the West. The apparently frequent presence of interpreters on voyages to the Caribbean verifies the general absence of a common language between a significant number of the Chinese and Europeans. Men such as Tsêng A-Tai testify that the interpreters aboard the ships abused them: “the interpreter was a Portuguese, and constantly kicked us” (Helly 1993: 44). Another by the name of Hsieh Kuan-Chieh implicates an interpreter in the suicide of a passenger, stating “one man jumped overboard because he was constantly beaten by the master and interpreter [. . .]” (Helly 1993: 44). A number of migrants remark on their inability to understand the language of the Europeans, despite the presence of interpreters. They vividly recall difficulties that language barriers created in obeying commands while at sea and relate that they were severely beaten as a result of not understanding the orders they were given. Some were even killed due to breakdowns in communication. Hundreds committed suicide. According to P‘ang A-tung, Europeans had little tolerance for instances in which language barriers impeded communication. He explains: “if through not understanding the language the work was not well performed, we were beaten, and one of our number threw himself overboard” (Helly 1993: 45). It seems that some who could understand neither a European language nor the interpreters on board turned to others who could for translation and assistance, as also happened later on plantations; however, due to Europeans’ apparently constant fear of mutiny, ship crews tried to limit contact of this type. In cases such as that of Tsêng A-yang, migrants were altogether forbidden from speaking while on board, at least in the presence of those responsible for their transport (Helly 1993: 46). The 1874 testimonies indicate that those in control of the ships felt threatened by migrants communicating among themselves in their own (non-European) languages, languages that most of them could not understand. This fear may have been heightened when Chinese were brought up on deck for exercise, food, washing, and air. . For examples of the manipulation of language to create confusion about differences between the Chinese calendar and that of the West, see petitions of Chêng A-mou and Hsieh-A-fa (Helly 1993: 42).

Chapter 14. Chinese Spanish in 19th-century Cuba 

Huang A-ch‘êng recalls the following: “In consequence of my talking with some others, the master suspected me of some mutinous intentions and directed the sailors to tie up, strip, flog, and chain, me and thirty-one others” (Helly 1993: 46). The Report tells of instances in which passengers were punished for speaking Chinese to demand better treatment. Ch’ê-n Lung remembers that he and others voiced a collective demand for more humane treatment during the voyage; he says, “When being prohibited from going on deck and being deprived of water we created an outcry, we were regarded as mutinous by the master, and over ten were killed” (Helly 1993: 46). These recollections describe a pattern on the ships that also emerged in Cuba: among themselves the Chinese tended to speak the languages of their home regions, but when speaking across racial lines or in the presence of speakers of European languages, they were expected to use the language of their captors. The most notable exceptions to this latter generalization are those instances when the Chinese stood up for their rights and asserted their humanity, actions that were usually deemed unacceptable even when they were communicated and justified in Spanish.

. Relations with persons of Spanish descent On the plantations of Cuba, Chinese workers, even those who had spent a relatively short period on the island, were expected to be able to communicate with those overseeing their work. There was punishment and suffering for not speaking Spanish. An individual by the name of Sung Hai notes in his testimony, “at present there are here forty Chinese who though recently arrived and ignorant of the language work in chains” (Helly 1993: 60). Some in positions of power had little patience for new arrivals’ ignorance of Spanish. As Huang Hsing remembers, “I saw three newly arrived Chinese killed by blows from the overseer, who declared that they could neither speak nor work” (Helly 1993: 105). Another migrant, Chou Tê-‘hua testifies, “they broke my finger because I was unable to express myself in Spanish” (Helly 1993: 61–2). Têng Kuan’s testimony makes it clear that not speaking Spanish could have ghastly consequences: During the two years which I passed on the plantation I was constantly beaten with great severity. I made a complaint to the officials, but through my imperfect acquaintance with Spanish, the result was that I was sent to the dépôt, where, with chained feet, I had to work without wages on the roads and breaking stones. (Helly 1993: 54)

Li Yü saw the punishment and abuse that stemmed at least partially from not understanding Spanish as altogether unjust. He explains: [. . .] when I was at work on account of not understanding the language I was beaten with a cowhide whip, and found the chastisement unendurable, and as I saw that another man poisoned himself through being unable to support the cruelties, I preferred a charge to the officials. They gave no heed to my accusation. I afterwards was chastised even more severely [. . .] (Helly 1993: 61)

 Don E. Walicek

Hundreds of the 1874 testimonies indicate that mistreatment of Chinese was widespread. As testimonies such as that of Ch’ên A-lin indicate, inability to communicate in Spanish or the absence of a translator made it even more difficult for contract laborers to defend themselves: “I was accused of committing a theft, was arrested, and imprisoned, owing to my ignorance of the language I was sentenced to one year’s imprisonment” (Helly 1993: 87). Liang A-ch‘êng relates a similar incident: “My master had promised that on Sundays labour might cease at noon, but when I wished to act in accordance with this assurance the overseer forbade my so doing and struck and kicked me, and as I am ignorant of the language I could not complain” (Helly 1993: 64). Spaniards generally considered Chinese language(s) and Chinese phenotypical features to be signs of Asians’ biological inferiority, while they believed Catholicism and “white” skin color attested to their own superiority (Mörner 1967: 5). Within this ideology of racial hierarchy, the ongoing use of Chinese language(s) among migrants, as well as the existence of learner varieties of Spanish influenced by Asian languages, would most often have only strengthened the Spaniards’ beliefs in the genetic and linguistic inferiority of Chinese laborers. Still there were expectations that the Chinese would assimilate culturally and learn Spanish; however, for the Chinese, speaking the language of the island’s dominant group was a way of pushing for the allowances provided in the contract of indenture (e.g., food, clothing, wages, time off for rest, passage to return home). At the same time, fluency in Spanish among Chinese plantation workers could be a threat to Spaniards’ desires to quell Chinese resistance and create a manageable labor force of which they could readily take advantage. Despite the insistence of colonial officials and masters that they learn Spanish, a number of testimonies indicate that their “superiors” were, perhaps ironically, reluctant to listen to pleas of laborers who communicated in Spanish. As Li Ying-sung recalls, “on the termination of the agreement I was refused the necessary papers, and my master was even wrathful at my applying for them. I was delivered to the officials, who would not permit me to speak, and was confined in prison for six years, labouring but receiving no wages. . .” (Helly 1993: 74). Chu Ts‘un-fang contextualizes the practice of silencing Chinese as a tactic of supervisors and recalls how the use of Spanish to challenge authority led to his brother’s death: my brother [. . .] placed on a wagon a smaller than ordinary quantity of sugarcane, and in reply to an overseer explained that the oxen were thin, and not able to move a heavier load. The overseer became very angry, said he would not permit Chinese to argue with him, and seizing a stick, inflicted him with heavy blows. This occurred at 4 p.m. and at 6 p.m. my brother was dead [. . .]. (Helly 1993: 104)

In dozens of accounts laborers are punished and chastised for using Spanish to assert the rights granted them under Spanish law, as well for defending what might be seen as more basic rights that the Chinese believed many slaves of African descent had at the time. Ch‘iu Têi offers one such example: “[. . .] on the streets, too, boys constantly throw stones at us, without any attempt at prevention on the part of the

Chapter 14. Chinese Spanish in 19th-century Cuba 

Government, and the underlings of the officials arrest us on fictitious charges, and drag us before the authorities, who close our mouths and impose fines” (Helly 1993: 112). A petition in the name of Hsien Tso-pang and thirteen others describes treatment of this sort as a common occurrence: in the streets we are constantly exposed to insult, to being struck by stones, and unless we submit in silence we are soon assaulted by a mob and are finally dragged off to jail. If a Chinese desirous to effect a purchase tries to bargain, he is at once abused, but a Spaniard or Cuban entering the shop of a Chinese wounds or even kills the latter if he attempts to ask for payment. (Helly 1992: 112)

Chinese were silenced by authorities, their masters, and persons they encountered in public when they used Spanish to defend the rights that they believed differentiated their status from that of slaves and animals. Chinese resistance to mistreatment played a role in Feijóo Sotomayor’s decision to oppose the purchase of Chinese contract laborers throughout the 1850s and 1860s. Early on a vocal supporter of Chinese immigration, he later argued that colonial Cuba was confused by the heterogeneity of its people. Sotomayor concluded that it would be “highly imprudent to introduce on a grand scale another race with a different language, religion, and culture” (quoted in Villanueva 1877: 343). Spanish intellectual José Antonio Saco held similar views. Initially unconcerned with cultural and linguistic differences, he encouraged Chinese immigration in the 1840s; but in the 1850s he insisted that it cease. In the latter period, Saco describes the Chinese as males (“evil people”) in the moral and political order of the island, a group whose presence “all good Cubans should lament;” he claims that the Chinese are indifferent to religion and espouse “anti-Christian” beliefs (Saco 1960: 570).18 It seems that some Chinese chose to emulate the relatively prestigious variety of Spanish used by persons of European descent, despite racism and abuse from whites. That some Chinese spoke a variety of Spanish similar to that of whites is corroborated by Esteban Pichardo, the nineteenth-century researcher providing the most detailed analysis of the island’s language in this period. In the introduction to his 1862 dictionary of Cuban Spanish, he writes, “The Chinese or Asians that are already numerous, mainly in Havana, have not formed a dialect . . . they pronounce with clarity the Castilian sounds that they learn quickly, although with a Creole accent like the Yucatecans” (1862: vii).19 He includes two “strange words” from popular Cuban Spanish (i.e., “el vulgo”) that the Chinese used, cha-chao, and tabacúa. Remarking on the conversion between /r/ and /l/ in words like amal (amar) and sordado (soldado), one of the features also discussed by Lipski (2000b), Pichardo indicates that it was a “defect” not .

This passage was translated from Spanish by the author.

. This passage was translated from Spanish by the author. The “Yucatecans” included Mayo prisoners of war and enslaved Mayo and Yaqui indigenous peoples from northern Mexico; in Cuba they worked primarily in the provinces of Caney and Ciguabos; see Knight (1990: 186).

 Don E. Walicek

common to all of the island nor to all of the Chinese (1862: viii). Also significant in terms of the claims made by Lipski, Pichardo makes no connection between Chinese Cuban Spanish and the speech of persons of African descent. In fact, Pichardo implies that the speech of Africans did not consistently serve as a target language for those Chinese who learned Spanish. His observations suggest instead that the speech of Chinese speakers of Spanish may have been relatively unmarked. For many whites and possibly for Chinese, the use of habla bozal was relatively stigmatized in most social contexts.

.

Relations with persons of African descent

Relations between the Chinese and Europeans are not the only set of cross-cultural dynamics to take into consideration in identifying the factors that shaped contact between speakers of Chinese and Spanish. As Lipski (2000a, b) argues, the relations between Chinese and blacks (i.e., Africans and Afro-Cubans) and linguistic contact between these groups should also be examined. But do the 1874 testimonies reveal that bozales may have influenced the acquisition of Spanish among the Chinese or indicate a substantial degree of contact between these two groups? Though there are dozens of references to Africans in The Report, there is little, if any, evidence of positive relations, camaraderie, or frequent interaction between the races. Generally, animosity prevailed between Chinese and blacks. As Helly (1993: 22) points out, “[. . .] the Chinese refused to be assimilated to the category of nonwhite/ slaves and they developed no concord with the black foremen who directed the field labor”.20 The same appears to be true for most relationships between Chinese and black slaves, persons with whom they may have had little contact. According to Lipski (2000b), the Chinese worked in separate work gangs and were housed apart from blacks, living in separate quarters. In addition, there was always a much larger population of African than Chinese descent on the island, suggesting that the opportunity for interaction would be limited to only some Africans. A few testimonies documenting interaction indicate that owners and overseers encouraged slaves to beat or assist in abusing the Chinese.21 The few examples of inter-racial alliances that do appear are either limited or temporary. Though notable, most resulted in increased hostility between Chinese and blacks. Wu Yeh-ch‘êng relays one such case: “four negroes in league with certain recently arrived Chinese killed the new administrator. By an outlay of money on the part . For remarks on the relationships between Chinese and foremen see, for example, the testimonies of Lai Ch‘uan-shou, Lü A-chên, Chang A-hsiu, Chu Ts‘un-fang-kuei, and Ho A-shan in Helly (1993: 54, 66, 88, 103–4, 118, respectively). . For evidence describing the negative relationship between Chinese and blacks see, for example, the testimonies of Lü A-Ch‘ên, Lü A-wu, ‘Ho Chang-yu, Li A-‘hui, and Wu A-ch‘ing in Helly (1993: 66, 87, 96, 98, 112, respectively).

Chapter 14. Chinese Spanish in 19th-century Cuba 

of our employer, the participation of the negroes was not mentioned, and the crime was imputed to us – ten in all – whose contracts were on the eve of expiration, and we were consequently imprisoned” (Helly 1993: 88). Inter-racial reliances between Chinese and white criollos provide the context for many of the literary attestations of Chinese Cuban Spanish (e.g., Jiménez Pastrana 1963: 76–9, 1983: 92; Varela 1980: 10). Many of these appear in anecdotes that detail Chinese participation in the Ten Years’ War (1868–1878) and other Cuban struggles for independence from Spain. The Chinese are depicted as patriotic, faithful, and obedient volunteers with a limited command of Spanish. While they are quoted as speaking in Spanish, many of these sources point to communication barriers. Jiménez (1963: 76–7), for example, includes examples of Chinese speaking Spanish but also cites evidence which indicates that the Chinese had “little knowledge of the language” and that most criollos had a difficult time understanding them. Helg (1995: 11) argues that whites and non-whites who joined the ranks of those who took up arms against Spain in the Ten Years’ War did not necessarily share the same ideas of freedom, but did perhaps look forward to forming new social relationships based on “envisioned notions of justice and equality”. While this is not the idea highlighted in excerpts from the corpus in which Chinese insurrectionists use Spanish, it is supported by evidence that participation was limited to some Chinese and by remarks suggesting that motivation varied among those who took up arms. Some Chinese, including thirteen interviewed by the commissioners, chose to separate themselves from their owners and the battle. Others joined because of geographical proximity and the criollos’ promise that they would be rewarded with freedom from indenture (Helly 1993: 93–4). LePage and Tabouret-Keller (1985) shed light on the sociolinguistic significance of data documenting race relations and diversity among the Chinese of the sort found in The Report and other historical research. As LePage (2000: 29) explains, their theory of identity as a series of acts points to the significance of the individual and posits that speakers: [. . .] create (the use of the word does not imply consciousness or “rationality”) their linguistic systems so as to resemble those of the groups from they wish from time to time to be identified with, or as to distinguish themselves from those they wish to distance themselves from.

Work by historians such as Scott suggests that due to reference group membership and differential power relationships the Chinese would have rejected bozales as a group with which they wished to identify themselves. She writes (1985: 109–10): The importation of Chinese contract laborers [. . .] required the invention of a third category between slave and free. This often led to tensions in the work force by creating an invidious distinction between slaves and free workers. The Chinese insisted on being treated differently from slaves, and the law – at least in theory – upheld this distinction. But the pattern of labor on the plantation could place the

 Don E. Walicek

Chinese in situations they felt obscured this distinction, and violence, flight, and conflict with slaves were frequent results.

Nevertheless, there are parallels between the experiences of these two groups: both bozales and the Chinese were subject to forced acculturation or “seasoning” in the hands of masters and overseers and, as established above, this process of adaptation had a linguistic component. “Unseasoned” Chinese and Africans were new arrivals with limited Spanish input. As the amount of time individuals spent on the island increased, slaves and indentured laborers received more and more input particular to the island, including more exposure to Cuban Spanish. And as testimonies by a minority of Chinese bilinguals demonstrate, fluency in Spanish over time had much to do with personal experience and socialization, just as it did for Africans who learned Spanish in Cuba. The 1874 testimonies show clearly that there was substantial motivation to learn Spanish for some, as well as a group of Chinese across the island who, like the majority of Africans, achieved fluency in the language. As Lipski (2000a) holds, in Cuba interaction with native or near-native varieties of the European lexifier was never removed to the extent found in Afro-American societies where Creole languages emerged (e.g., British, Danish, Dutch, and French colonies in the Caribbean). Nevertheless, it seems that this may be more true for persons of African ancestry (whether creoles or bozales) than it was for the Chinese categorically. A substantial portion of the latter maintained their ancestral language(s) and either never learned Spanish or acquired only a limited command of the language. A key difference is that the larger size and demographics characterizing the population of persons of African descent (e.g., the presence of women, multiple generations, a large mixed-race population) shaped and reflected patterns of sociolinguistic interaction that differentiate them from the Chinese. Nineteenth-century commentaries describing the language of speakers of African ancestry blur linguistic boundaries and demonstrate that the usage of Spanish among Africans appears to have been much more extensive, frequent, and widespread (and common) than it was for Chinese as a group. They suggest that generalizations about the island as whole and linguistic boundaries retrospectively drawn by linguists among blacks may overemphasize place of birth and bozal status as concomitant characteristics that effectively index variation in language use. According to Pichardo (1875: 12), Cuban-born blacks spoke like white criollos and those of Matanzas and Havana shared linguistic traits with Africans. He states that the speech of bozal slaves could be heard frequently and in all parts of the island, that it was common and identical among members of this group regardless of their specific nation of origin. According to Pichardo, the habla bozal spoken by African-born slaves was conserved eternally unless they reached the island as young children. Reviewing references to habla bozal in Pichardo’s work and other nineteenth-century sources, Roberts (fc: 207) holds that habla bozal was in practice characterized “[. . .] by a level of stability that belies the idea that idiosyncratic attempts at second language use suggest” and describes it

Chapter 14. Chinese Spanish in 19th-century Cuba 

as “[. . .] a fixed variety which Africans acquired”. This assertion, like attestations and discussions of bozales’ speech in the nineteenth and early twentieth centuries, (e.g., De La Torre 1854; Guirao [1938] 1970; Pichardo 1862, 1875), raises questions about the classification of habla bozal as an “Afro-Hispanic pidgin” and the statement that it “was by and large a transitory phenomenon, emerging in different guises in each Afro-Hispanic speech community and converging with native varieties of Spanish in a single generation” (Lipski 2000a: 438). A closer look at how shifts in identity over time may have differed for Africans and Chinese casts additional doubt on the assertion that the learner’s variety of Spanish that the Chinese spoke converged with a pidgin that was spoken by newly arrived blacks. According to Paquette (1988: 38), once bozales became conversant in “AfroSpanish patois” or Spanish they were called ladinos.22 A similar term is not known to have existed for acculturated Chinese laborers; however, it seems that for the Chinese a de facto category of Asian ladinos, however small the group might have been, certainly did exist. One explanation for the absence of the wider identification of a group of Chinese Spanish-speakers as ladinos may be that during this era Cuban society at large considered Asians such a foreign element that the idea of a ladino Asian was for many not only undesirable, but also a contradiction in terms. In contrast, Spaniards and creoles in Cuba made a clear social distinction between those persons of African ancestry who were recognized as having assimilated to Cuban culture and those born in Africa (and purportedly spoke differently), despite de facto cultural and linguistic features that complicate these distinctions. In the case of Africans, these distinctions may have had more to do with aforementioned racial fears, the stigma of slavery, and emergent notions of cubanidad, than it did with language use itself. In the early years of Chinese indentured servitude on the island, it is likely that few Chinese spoke Spanish fluently and that the laborers frequently used mutually intelligible Asian varieties, (e.g., native languages, second languages) in communication among themselves. During the first decade or so of migration, the “first generation” of Chinese probably learned Spanish through interactions centered around the plantations (e.g., talking and interacting with a diverse group of speakers including overseers, owners, and Afro-Cubans). Some individuals may have even been pressured by overseers to translate for those Chinese with whom they were grouped. However, sinocentrism among migrants, their physical separation from blacks, and the association of bozales with slavery (and white skin color with freedom), make it unlikely that the Chinese had the type of linguistic interactions with newly arrived Africans that would lead them to “target” or be directly influenced by habla bozal. Certain shared predispositions

. Examples of habla bozal from Cuba have frequently been taken from the speech of Africans in who had become or were in the process of coming ladino (See Roberts fc: 128–9, especially his discussion of carbela).

 Don E. Walicek

such as a common first language and consistent exposure to similar varieties of Spanish may have contributed to regularities in the Spanish spoken by some Chinese; however, factors like the existence of different languages among migrants, diverse language attitudes, dissimilar communication patterns, and varying degrees of exposure to Spanish would have contributed to variation. These tendencies and related sociolinguistic data suggest, as Parkvall (2000: 21) contends, that the variety of Castilian the Chinese used in Cuba was not a pidgin, but a learner’s variety of Spanish.

.

Chinese-Spanish bilinguals

Despite my use of the term “learner variety”, there was substantial variation in Spanish proficiency among migrants. Some undoubtedly learned Spanish and spoke it well. While some of these bilinguals were plantation workers, others were persons who managed to either purchase their freedom or complete the term of their contract of indenture. I suggest that these individuals, like whites on the island, are more likely to have been linguistic models for those Chinese who acquired Spanish than are Africanborn bozales.23 Data in The Report suggest that some bilinguals, perhaps in terms of language ability the Asian equivalent of the African ladino, held important social positions in places where groups of Chinese lived together, such as on the plantations and in the depósitos centrales, areas where authorities held laborers not under a contract to a single master (Martin 1939: 10; Scott 1985: 100).24 Frequently, bilinguals were persons who demanded that the rampant abuse against indentured laborers cease. Li Tai, for example, remembers,

. An extremely small number of formerly indentured Chinese did establish families in which Spanish was the dominant language, despite social prejudices and laws against miscegenation (Martinez-Alier 1991: 76–79). Most of these were with women of African ancestry described as “mulattoes and negresses” (Helly 1993: 115). Two men are stated to have married white women. One spoke of being disliked and injured because of his marriage. See references to Chang Ch‘ang-kuei and Chêng A-lai, Helly (1993: 115). In the twentieth century Chinese migrated to Cuba under very different circumstances and unions of various types (e.g. between Chinese, interracial) were much more common. . A parallel can be drawn between the Chinese interpreter and the black overseer, the basya, that Arends (2001) describes on the plantations of Suriname. In both instances individuals acted in intermediary positions and their contacts with whites on the plantation were probably stronger and more frequent than those of most other indentured laborers. As Arends proposes in his interpretation of the social network model of language variation and change (L. Milroy 1980; J. Milroy 1992), this suggests that the basya is likely to have been a linguistic innovator. A similar situation may have existed with bilingual Chinese who acted as interpreters and group leaders. The anonymous reviewer of this chapter points to an additional parallel: Portuguese who may have served a similar role between laborers and plantation owners in Hawai‘i.

Chapter 14. Chinese Spanish in 19th-century Cuba 

one of us, able to speak Spanish, told our employer that rice was the customary diet of the Chinese, and that, whilst at Macao it had been distinctly understood that Sundays were to be days of rest, the existing practice was different. Our employer gave way to anger, regarding these words as too daring; and afterwards our companion fell sick and was sent to the hospital where in a few days he died. (Helly 1993: 61)

Testimonies show these bilinguals as resources for Chinese speakers both within and outside plantations (e.g., Helly 1993: 67, 74). A testimony by Ch’ên I, confirms that individuals proficient in Spanish acted as mediators and leaders for those who were not. Ch’ên offers an anecdote that shows a bilingual speaker intervening to secure better treatment for his fellow countrymen: when our contracts expired nine months’ wages were due to us, and we applied to the new master for payment. He directed us to continue our labour and to await the return of our former employer. We – 16 in all – then inquired what wages would be issued to us, and were told that we should receive only the former rate of $4. We replied that after our original engagement had been completed at a higher rate ought to be conceded, and that if it were not granted we should prefer to be delivered over to the authorities. The administrator then declared that our first contracts had not yet expired, and that it mattered not whether willingly or reluctantly we must work on the old conditions; and when we persisted in our refusal the overseer brought cords and chains, and stood round us, as if intending to place us in fetters. Upon this one of our number conversant in Spanish urged the administrator that our contracts had already expired, and that we refused to work on the former terms [. . .] (Helly 1993: 79)25

In other instances, laborers who could not file petitions to report the violation of their contracts or other forms of abuse (because of their lack of familiarity with Spanish or lack of access to the proper authorities), relied on bilinguals among them for assistance and support. As told by an ill-treated worker named Lo-A-êrh: On the expiration of the contract, I was sent to a dépôt where I passed four months. I was then removed by my master, who forced me to labor in chains during three months. He daily beats me and yesterday flogged me with great severity. Five months have now elapsed, during which no wages have been issued to me. On account of this and the other ill usage, I requested a fellow workman to, on my behalf, lay a complaint before the officials. He has done so on three occasions and in each case it has been rejected, and I have thus to go on labouring; [. . .] (Helly 1993: 55) . In this case the intervention of Ch’ên’s bilingual companion led to an investigation of the administrator. Eventually Ch’ên and five others were taken to Colon’s depósito as they requested. The remaining ten in his group were returned to the plantation.

 Don E. Walicek

Such bilingual intermediaries could be targets of officials’ abuse, as Chinese who held semi-official roles as interpreters were sometimes accused of attempts to deceive authorities. Hsieh A-‘hou recalls such a case: four men out of a gang of thirty newly arrived, died in the hospital four hours after they had entered it, on account of alleged sickness. Upon this twenty men laid before the authorities a charge of murder, and at the request of the administrator an official of low rank visited the estate. I, acting as interpreter, translated the evidence of two witnesses who declared that sickness had been the cause of death. On the following day other officials continued the enquiry. All were present, and on a question being put as to the prior existence, the general answer was that the men had not been ill. The officer of low rank then declared that, on the previous occasion, I had interpreted incorrectly, and I was placed in prison. (Helly 1993: 67)

Regarding language structure, Chan and Tai (1989: 50–58) explain that the grammar of Chinese languages exposes the inadequacy of traditional Western grammatical concepts (e.g., the word, subject, object) as universal classes always appropriate for the analysis of natural languages. They point out that formal word classes and “parts of speech” are largely inappropriate for the study of Chinese, noting characteristics that differentiate languages such as those spoken by many migrants from Spanish. These include the existence of nominal predicatives, the absence of inflectional and derivational morphology (as traditionally defined), the absence of subject-verb agreement, and the general lack of subject/object case distinction. Their description of Chinese suggests that two indigenous grammatical concepts, shizi (“full words”/words with lexical meaning) and xuzi (“empty words”/words with grammatical meaning) can assist linguists in recognizing the significance of these differences (Kratochvil 1968: 106–15). This same pair of concepts may be helpful in describing the emergence of bilingualism and the use of Spanish among the aforementioned Chinese laborers.

.

Looking back on the Chinese diaspora

Testimonies from The Report serve as reminders that language contact among humans is always shaped by lived experience and suggest that analyses of language contact can be strengthened when they take sociolinguistic phenomena into consideration as a constellation of co-existing and complex factors that overlap and interact with another. Accordingly, this essay, a response to creolists’ calls for what Arends (2002: 56) terms “historical correctness”, has shifted the focus away from grammar to look back on language contact in sociohistorical context. The discussion offered above establishes nine premises that assist in staking out this complexity as it relates to Asian indentured servitude in Cuba: a.

Push as well as pull factors (including Britain’s presence in China and the gradual abolition of the African Slave Trade in the Spanish Caribbean) led to the forced

Chapter 14. Chinese Spanish in 19th-century Cuba 

b.

c.

d.

e.

f.

g.

h. i.

migration of Chinese to Cuba in the nineteenth century and to the acquisition of Spanish by some Chinese. A diversity of native languages were spoken by laborers from diverse provinces and social classes; many can be traced to Kwangtung, the province of southern China from which most migrated. The period in which laborers were held awaiting transfer to Cuba was relatively short, providing little exposure to MCP, a nineteenth-century Creole probably spoken by only a very small minority among indentured laborers. Persons in positions of power (e.g., Portuguese traders, European interpreters, owners of contracts, Spanish authorities, overseers) exploited and relied on linguistic difference in subjecting the Chinese to semi-enslavement. Linguistic contact between Chinese and bozales was limited, due to infrequent linguistic contact, acrimonious race relations, owners’ pitting of Chinese and Africans against one another, and the stigmatization of slavery and Africans in various sectors of Cuban society. Varieties of Spanish associated with whites probably maintained relative prestige among Chinese and may have been the variety to which they were most frequently exposed, making it influential for those who learned Spanish. Two opposing tendencies characterize Spaniards’ attitudes towards laborers: first, the Chinese were expected to use Spanish in linguistic exchanges with owners, overseers, authorities, and other Spanish speakers; second, Chinese were frequently silenced and or punished when they used Spanish to defend themselves against mistreatment. Bilingual Chinese acted as intermediaries for their fellow countrymen, sometimes serving as linguistic models for others learning Spanish. The overwhelming majority of Chinese maintained their native languages for ingroup communication; this speech community included many individuals who appear not to have acquired proficiency in Spanish.

As the 1874 Chinese testimonies suggest, the documentation of language contact sometimes requires the reconsideration of premises. Regarding written examples of “el habla del Chino”, a close reading of these materials reveals that many belittle the fact that the Chinese were conscious of their victimization in a social system dominated by white racism. Often these examples of speech are deemed humorous and noteworthy precisely because they correlate with popular beliefs that identify the linguistic behavior of Chinese as distinct, foreign, predictable, and error-ridden (see examples in Espina Pérez 1972: 59; Varela 1980: 18–24). These representations of speech minimize the linguistic variation that appears to have characterized the migrants’ use of Spanish as a non-native language and fail to allude to the impact that racism and social disenfranchisement may have had on Chinese laborers’ acquisition of Spanish. In contrast, metalinguistic data from The Report draw attention to variation and show that individual agency and historical context mediate between social facts and

 Don E. Walicek

linguistic ones. The future study of this corpus should take into consideration features of Chinese languages and the striking grammatical parallels between “Chinese-Cuban Pidgin Spanish” and the literary representations of Spanish of non-Chinese (e.g., blacks and whites in Cuba) in what is known about the pragmatic, social, and historical contexts in which they were produced. Further research on the early presence of Chinese in the Caribbean and Latin America can be done on a number of additional topics, including: the investigative missions to Cuba and Peru that led to the publication of The Report and related Chinese and Spanish language archives; the experience of the Chinese in nineteenthcentury Cuba in comparison to that of the Chinese in nineteenth-century Peru; the effects of the absence of Chinese women on language attitudes and language contact; the evolution and perpetuation of racial and linguistic stereotypes in the nineteenth and twentieth centuries; the features of grammar and discourse in literary examples of “el habla chino” and their comparison to written examples of non-standard Spanish found in costumbrista texts and works featuring habla bozal; language use in Chinese speech communities across systems of colonization (e.g., comparing communities in Cuba, Puerto Rico, and Costa Rica to those in non-Spanish settings like Jamaica, Trinidad, and Suriname). The collection of oral histories from descendants of Chinese Cubans in New York City, Miami, Puerto Rico, and other settings could contextualize these and other inquiries and perhaps, as a reviewer of this essay suggested, take the sociolinguistic reconstruction proposed in this article one step further. Sociolinguistic research on the aforementioned topics can contribute to debates about diffusion, language contact, creolization, and the categorization of Creoles and non-Creoles across diverse sociohistorical and geographical situations. And perhaps at least equally significant, this type of investigation can be used to assess the adequacy of theoretical models and ideological assumptions that shape some of the questions linguists ask. As DeGraff (2004: 838) suggests, “..creolists just cannot take it for granted that we ‘had it right the first time’ (or even the last time); we must confront old dogmas with new, and old, data in order to come up with empirically and theoretically better accounts”.

References Arends, J. 2001. Social stratification and network relations in the formation of Sranan. In Smith, N. and Veenstra, T. (eds). Creolization and contact. Amsterdam: Benjamins. 291–307. Arends, J. 2002. The historical study of Creoles and the future of Creole studies. In Gilbert, G. (ed.). Pidgin and Creole linguistics in the twenty-first century. New York: Peter Lang. 49–68. D’Auvergne, E. 1933. Human livestock: An account of the share of the English-speaking peoples in the development, maintenance, and suppression of slavery and the slave trade. London: Grayson and Grayson. Bakker, P. 2002. Some future challenges for Pidgin and Creole studies. In Gilbert, G. (ed.). Pidgin and Creole linguistics in the twenty-first century. New York: Peter Lang. 68–92.

Chapter 14. Chinese Spanish in 19th-century Cuba  British and foreign state papers, 1814-1977. Volumes 35–37. London. Bueno, S. (ed.). 1959. Los mejores cuentos cubanos, Tomo I. La Habana: Segundo Festival del Libro Cubano. Chan, K.M. and Tai, J.H.Y. 1989. A critical review of Norman’s Chinese. Journal of the Chinese Language Teachers Association XXIV.1. 43–61. Consuegra y Guzmán, I. 1930. Mambiserías: Episodios de la Guerra de Independencia 1895– 1898. La Habana: Imprente del Ejército. Corbitt, D.C. 1971. A study of the Chinese in Cuba, 1847-1947. Wilmore, North Carolina: Asbury College. DeGraff, M. 2004. Against Creole exceptionalism (Redux). Language 80.4. 834–9. Espina Pérez, D. 1972. Diccionario de cubanismos. Barcelona: M. Pareja. Feijóo Sotomayor, F. (ed.). 1981. Cuentos populares cubanos de humor. La Habana: Editorial Letras Cubanas. Guirao, R. [1938] 1970. Orbita de la poesia afrocubana 1928-1937: (Antología). Selección, notas biograficas y vocabulario. Nendeln: Kraus. Helg, A. [1876] 1995. Our rightful share, the Afro-Cuban struggle for equality, 1886-1912. Chapel Hill: University of North Carolina Press. Helly, D. (ed.). 1993. The Cuba Commission report: A hidden history of the Chinese in Cuba. Baltimore: Johns Hopkins University Press. Holm, J. 1989. Pidgins and Creoles: Volume two reference survey. New York: Cambridge University Press. Huesmann, J. 1991. The Chinese in Costa Rica. Historian 53.4. 711–2. Jiménez Pastrana, J. 1963. Los chinos en las luchas por la liberación cubana, 1847–1930. La Habana: Instituto de Historia. Jiménez Pastrana, J. 1983. Los chinos en la historia de Cuba: 1847–1930. La Habana: Editorial de Ciencias Sociales. Knight, F.W. 1990. The Caribbean: the genesis of a fragmented nationalism. New York: Oxford University Press. Kratochvil, P. 1968. The Chinese language today: Features of an emerging standard. London: Hutchinson and Company. Lai, L.W. 1993. Indentured labor, Caribbean sugar: Chinese and Indian migrants to the British West Indies, 1838–1918. Baltimore: Johns Hopkins Press. LePage, R.B. 2000. The evolution of a sociolinguistic theory of language. In Coulmas, F. (ed.). The handbook of sociolinguistics. Malden, Massachusetts: Blackwell. 15–32. LePage, R.B. and Tabouret-Keller, A. 1985. Acts of identity: Creole-based approaches to ethnicity and language. Cambridge: Cambridge University Press. Lipski, J. 2000a. Bozal Spanish: Restructuring or creolization. In Neumann-Holzschuh, I. and Schneider, E.W. (eds). Degrees of restructuring in Creole languages. Amsterdam: Benjamins. 437–68. Lipski, J. 2000b. Chinese-Cuban Pidgin Spanish: Implications for the Afro-Creole debate. In Rickford, J.R. and Romaine, S. (eds). Creole genesis: Attitudes and discourse. Studies celebrating Charlene J. Sato. Philadelphia: Benjamins. 215–33. Martin, J.L. 1939. De dónde vinieron los chinos de Cuba: Los jaca, los joló, los puntí, y los amoyanos, en la vida cubana. La Habana: Editorial Atalaya. Martinez-Alier, V. 1991. Marriage, class, and colour in nineteenth-century Cuba: A study of racial attitudes and sexual values in a slave society. Ann Arbor: University of Michigan Press. Milroy, J. 1992. Linguistic variation and change. Oxford: Blackwell.

 Don E. Walicek Milroy, L. 1980. Language and social networks. Oxford: Blackwell. Mörner, M. 1967. Race mixture in the history of Latin America. Boston: Little, Brown, and Company. Murray, D. 1980. Odious commerce: Britain, Spain, and the abolition of the Cuban slave trade. Cambridge: Cambridge University Press. Pan, L. 1990. Sons of the Yellow Emperor: A history of the Chinese diaspora. Boston: Little, Brown, and Company. Paquette, R. 1988. Sugar is made with blood: The conspiracy of La Escalera and the conflict between empires over slavery in Cuba. Middletown, Connecticut: Wesleyan University Press. Parkvall, M. 2000. Out of Africa: African influences in Atlantic Creoles. London: Battlebridge Publications. Pérez, L.A. 1988. Cuba: Between reform and revolution. New York: Oxford University Press. Pérez de la Riva, J. 1967. Demografía de los culíes chinos en Cuba, 1853–1874. Havana: La Biblioteca Nacional. Pichardo, D.E. 1862. Diccionario provincial casi-razonda de vozes cubanas. Third edition. Havana. Pichardo, D.E. [1875] 1976. Diccionario provincial casi-razonda de vozes cubanas. Third edition. La Habana: Editorial de Ciencias Sociales. Ramsey, S.R. 1987. The languages of China. Princeton: Princeton University Press. Roberts, P. forthcoming. The shaping of colonial identity in the Caribbean. Kingston: University of the West Indies Press. Saco, J.A. 1960. Colección de papeles cientificios, historicos, politicos, y de otros ramos sobre la isla de Cuba. Havana: Ministry of Education and Culture. Schwegler, A. 2005. Bozal Spanish: Captivating new evidence from a contemporary source (Afro-Cuban “Palo Monte”). In Fuller, J. and Thornburg, L.L. (eds). Studies in contact linguistics: Essays in honor of Glenn G. Gilbert. New York: Peter Lang. 61–90. Scott, R. 1985. Slave emancipation in Cuba: The transition to free labor, 1860-1899. Princeton: Princeton University Press. De la Torre, J.M. 1854. Compendio de geografia fisica, politica, estadistica, y comparada de la isla de Cuba. Habana: Imprenta de Soler. Turner, M. 1974. Chinese contract labor in Cuba, 1847-1874. Caribbean Studies 14. July. 6–81. Varela, B. 1980. Lo chino en el habla cubana. Miami: Ediciones Universal. Villanueva, M. 1877. La emigración de colonos Chinos. Revista Contemporánea 7. 339–76. Yung, W. 1909. My life in America and China. New York: Henry Holt and Company.

 

Comparative perspectives on the origins, development and structure of Amazonian (Karipúna) French Creole Jo-Anne S. Ferreira UWI, St. Augustine/SIL International

Mervyn C. Alleyne UWI, Mona/UPR, Río Piedras

Together known as Kheuól, Karipúna French Creole (KFC) and Galibi-Marwono French Creole (GMFC) are two varieties of Amazonian French Creole (AFC) spoken in the Uaçá area of northern Amapá in Brazil. They are socio-historically and linguistically connected with and considered to be varieties of Guianese French Creole (GFC). This paper focuses on the external history of the Brazilian varieties, and compares a selection of linguistic forms across AFC with those of GFC and Antillean varieties, including nasalised vowels, the personal pronouns and the verbal markers. St. Lucian was chosen as representative of the Antillean French creoles of the South-Eastern Caribbean, including Martinique and Trinidad, whose populations have had a history of contact with those of northern Brazil since the sixteenth century. Data have been collected from both field research and archival research into secondary sources.

.

Introduction

This study focuses on a group of languages/dialects which are spoken in Brazil, French Guiana and the Lesser Antilles, and to a lesser extent on others spoken in other parts of the Americas (as well as in the Indian Ocean). This linguistic group is variously referred to as Creole French, French Creole, French-lexicon Creole, French-lexifier Creole, French Creole languages/dialects, Haitian/Martiniquan/St. Lucian (etc.) Creole, and more recently by the adjective of the name of the country, particularly in the case of the Haiti (cf. Haitian, le haïtien; but cf. also le mauricien, le seychellois, le guyanais, etc.). The designation of these languages/dialects has not been standardised in the literature, nor in other media of academic discourse. Other aspects of these languages have also not come under general agreement among linguists; such as, for example, the very definition of “creole language”. In this study, French Creole is the usage adopted by the authors. The varieties under study here are referred to by the names of

 Jo-Anne S. Ferreira and Mervyn C. Alleyne

the groups that speak these varieties, for example, Karipúna French Creole (KFC) and Galibi-Marwono French Creole (GMFC). There is a general implication in the literature that these languages/dialects form a single “family” of languages. The cover term “French Creole” is very often and routinely used to include all of them, certainly in a generic sense, if not in a genealogical sense. Throughout the history of Creole Linguistics, it has been assumed that creole languages (of whatever lexical base) constitute a specialised group of languages requiring special theories and principles outside the general frameworks established by the science of linguistics to deal with the languages of the world. A counter movement dating as far back as the 1960s (cf. Alleyne 1966, 1971), has sought, explicitly or implicitly, to question the principle of what is now being referred to as “creole exceptionalism” (see DeGraff 2003) and to show that creole languages can be accommodated within general principles established for the languages of the world (although in some cases it is the rather marginalised principles of general linguistics that are invoked), as well as through multi-disciplinary approaches including History, Social Psychology, Psycholinguistics (see, for example, the recently established sub-branch of Linguistics, “Contact Linguistics”, which highlights the role of contact in language change; see Thomason and Kaufman 1988; Winford 2003). In the context of this creole exceptionalism, Creole Linguistics has, at least implicitly, assumed that creole languages emerged abruptly within one generation, crystallised immediately and remained unchanged over the more than three centuries of their emergence, except in those cases where importations from the official languages are made, whether in the form of loans (where the two languages do not share a common lexicon, e.g., Dutch importations into Sranan) or in the context of what has come to be known as “decreolisation” (e.g., Jamaican in contact with an official language, English, of the same lexical base). In fact, very little attention has been paid to internally motivated changes of the type undergone by all human language (cf. Historical Phonology, the well-established sub-branch of Linguistics). Comparative Creole Linguistics has been content to identify and analyse structurally the common forms that these languages quite definitely exhibit (cf., for example, Goodman 1964; Holm 1989; Valdman 1978). In some cases Creole Linguistics attempts to specify the origins of these common forms either in French (Chaudenson 1992, 1995) or in African languages (Alleyne 1980) or in language universals (Bickerton 1981; Muysken and Smith 1986), giving rise to three categories of creolists: superstratists, substratists and universalists. These studies establish the typological affinity of these languages but they have not cleared up in any definitive way the genealogical affinity, as, generally speaking, Creole Linguistics, adhering consistently to the principle of creole exceptionalism, has not tackled this question (cf., however, Alleyne 1976, which examines the complexity of this question). Nor have these studies treated creole languages as the outcome of a steady historical process over the centuries with changes induced by contact as well as internally motivated changes (but cf. Arends 1995; Migge 2002; Mintz 1971; Neumann-Holzschuh and Schneider 2000).

Chapter 15. Amazonian (Karipúna) French Creole 

It is, however, the case that creole languages have been, and continue to be, dynamic systems with internally generated change at all levels of language structure. As we find in other language “families,” creole languages exhibit chronological layers, that is, historically based variation which reveals changes that have not (yet) gone to completion leaving older forms side by side with the newer forms and allowing comparative and internal reconstruction of the historical forms and processes. Dialects such as those spoken in Uaçá, Amapá, Brazil, become very important as they may be considered with a fair degree of authenticity to have preserved older forms which are crucial in the mapping of historical processes and the reconstruction of earlier forms.

.

Karipúna and Galibi-Marwono1: The socio-historical context

This is the first study emerging from a research project entitled “A Comparative Study of French Creoles in the Southern Caribbean and Northern South America”. The larger project aims to compare varieties of Lesser Antillean French Creole with varieties of French Creole spoken in the north-eastern Amazon region of Brazil and north-eastern Venezuela.2 This present study is based on fieldwork carried out in February 1998 and July 2000 (Ferreira 1998; Wiesemann 2000), and then in May 2003,3 as well as on secondary sources. This article examines the French Creole language of Brazilian Amazonia in the overall context of Creole studies and in its local sociolinguistic context, mainly in Brazil and also in French Guiana. There already exist some introductory studies of Amazonian4 French Creole (AFC), that is, Karipúna and Galibi-Marwono of Amapá (see de Andrade 1988; Ferreira 1998; Ladhams 1995; Montserrat and Silva 1984;

. Or Galiby-Marwono, according to FUNAI (Fundação Nacional do Índio or the National Amerindian Foundation of Brazil). The Instituto Socioambiental lists Galibi do Uaçá, Uaçauara, and Mum Uaçá as alternate names. Among the Palikúr, they were also known as Maraoni (Diana Green, pc, February 1998). (They are not to be confused with the Galibi or Galibí do Oiapoque of Brazil and French Guiana.) .

This article does not deal with the French Creole of Güíria and El Callao, Venezuela.

. The UWI-based team (Ferreira, Alleyne and Nathalie Charlery) was accompanied by Stan Anonby (SIL Brazil), Peter Samuel (St. Lucia, SIL Americas) and Adriana Carla Alves e Silva (Brazil, formerly of Missão Novas Tribos do Brasil or New Tribes Mission – NTM). The research team comprised two native speakers of French Creole varieties (Charlery of Martinique and Samuel of St. Lucia), and one L2 speaker of French Creole (Alves e Silva of Brazil). This phase of the research was funded by the UWI Research and Publications Fund Committee. . Or Amerindian French Creole. Oiyapoque, Amapá (AP) is situated in the north-eastern part of the Amazon.

 Jo-Anne S. Ferreira and Mervyn C. Alleyne

Picanço 20035; Picanço Montejo 1996; Tobler 1983, 1987; Wittmann 1987). The aim of the larger project, of which this paper is a part, is to identify, study and classify these little studied dialects of French Creole spoken in relatively isolated enclaves of Amerindian peoples of Brazilian Amazonia, and to go deeper into the phonological, morpho-syntactic and lexico-semantic structure of the language. The overall purpose is to further expose linguists, historians, anthropologists, and other Caribbean Studies scholars to a group of Amerindian peoples whose first language is French Creole, to account historically and sociolinguistically for the language shift from an Amerindian language to French Creole within Portuguese-speaking Brazil and to analyse and document the development and current state of the language. The Amazonian French Creole dialects of Uaçá in northern Amapá are important by virtue of the fact that they are spoken by relatively isolated peoples, many of whom are now cut off from close contact with the lexifier language French and who can be assumed to be linguistically conservative rather than innovative. This present study will locate the French Creole dialects of these Amazonian peoples historically and comparatively within the broad family of French Creole. The hypothesis is that these dialects are the best evidence of nineteenth century French Creole and are vital for the reconstruction of the history of French Creole in general.

.

Creole languages in Brazil

In Brazil, there are four creolophone groups that speak two creole languages, one French-lexified, the other Portuguese-lexified. The variety of Portuguese-lexicon Creole is spoken in the state of São Paulo,6 while creole languages lexically based on French are spoken in the state of Amapá in North Eastern Brazil, the Brazilian state bordering French Guiana. One variety of French Creole is spoken natively in the northern Uaçá area of Amapá by two Amerindian groups, the Karipúna7 and the Galibi-Marwono, and the other is found in the southern area of Macapá, the capital of Amapá. The

. These varieties of French Creole are mentioned in do Couto 1997 and Tarallo and Alkim 1987. The groups are mentioned among other Brazilian indigenous groups, but relatively little attention is paid to them in these works as their language is not a traditional indigenous Brazilian Amerindian one (Arnaud 1966; Novaes 1994; Rodrigues 1986). . According to Gordon (2005: 226), the Portuguese-based creole language, Cafundo Creole, is a secret language spoken by a group of people located about 150 miles from the city of São Paulo, and “a similar language was recently discovered in the state of Minas Gerais.” Very little is known about these two languages. . The Karipúna do Amapá, also known as Caripuna, are not to be confused with the Karipuná de Rondônia. The latter group speaks Karipuná (also known as Karipuná do Guaporé, Caripuna, Jau-Navo, Juanauo, and Kagwahiva). This almost extinct language belongs to the Tupi family. (See also Montserrat 1994: 101; Teixeira 1995: 306.) This Rondônia group consists of only

Chapter 15. Amazonian (Karipúna) French Creole 

latter is referred to in the current international literature in English as Amapá French Creole (APFC), and in the Brazilian literature as Lanc-Patúa, and is said to be spoken by descendants of French Creole-speaking immigrants from the (English-official) Lesser Antilles who number approximately 25,000 (Gordon 2005: 224).8 While the two varieties are essentially similar, the northern Amapá variety shows some evidence of both recent French (via GFC) and Portuguese influence. This is because of the proximity of some of the speakers to French Guiana, as well as a relatively high degree of bilingualism in French Creole and Portuguese, especially among Galibi-Marwono young people (Suely Santos, pc, 1998).9 The southern variety, Lanc-Patúa, is said to show some English influence because of relatively large numbers of migrants from St. Lucia and Dominica, and also Portuguese influence. French Creole varieties in Brazil

Northern Amapá (Uaçá)

Amazonian French Creole (AFC) (or Kheuól/Crioulo/Patuá)

Southern Amapá (Macapá)

Amapá French Creole (APFC) (or Lanc-Patúa)

• Karipúna (KFC) • Galibi-Marwono (GMFC) Figure 1. French Creole varieties in Brazil

12 to 30 people, although there may be more (Gordon 2005: 228; Ricardo 1995: 39). Their language comprises two main dialects: Jacaria and Pama (or Pamana) and some of the speakers may be bilingual in Tenharim, a neighbouring Tupi language (Gordon 2005: 228). See also Instituto Socioambiental 1999–2005. . During the 2003 fieldtrip, an attempt was made by Stan Anonby and Peter Samuel to carry out fieldwork in Macapá, the capital of the state of Amapá, but this effort yielded no fruit. No Amapá French Creole (APFC) speakers were found, despite the fact that Julieta de Andrade completed her book on a group of Lanc-Patúa speakers in 1988, only nineteen years ago, and despite the number of 25,000 speakers recorded by the Ethnologue (print and web versions). Ethnologue. com notes that that variety of the language shows influences from English and French. . FUNAI notes that “um certo número de índios fala o francês devido à proximidade com a Guiana francesa, onde muitos residem ou residiram, trabalhando em Saint-Georges ou Caiena” (2001: 9). [Translation: a certain number of Amerindians speak French owing to the proximity to French Guiana, where many reside or have resided, working in St. Georges or Cayenne.]

 Jo-Anne S. Ferreira and Mervyn C. Alleyne

According to Gordon (2005: 228), the ancestors of the Karipúna of Amapá10 once spoke Karipúna (also known as Karipúna do Uaçá11), an unclassified language, now extinct, that may or may not have been a Tupi-Guarani language.12 There has been considerable psycholinguistic activity leading to language learning/acquisition and language death among Amerindian groups along the Brazilian-French Guiana border. It is evidently very important to study the linguistic, social, political and ideological conditions that underlay the death of indigenous dialects/languages and the adoption of French Creole dialects, and this area of study will be included in the larger project to be undertaken by the authors.

.

Location of the Amazonian French Creoles

The geographical area of study in question is the north-eastern corner of Brazil, specifically the northern point of the state of Amapá, an area bordered by the Oiyapoque13 River, which flows northeast and forms the border with French Guiana; the Uaçá River, which flows northwest; and Highway BR156, which connects the city of Oiapoque with Macapá, the state capital. (See maps below.) There are three Amerindian reserves in this area: Uaçá, Juminã, and Galibi divided among four Amerindian ethnic groups: Karipúna, Galibi-Marworno, Galibi do Oiapoque . The original Karipúna were possibly from the island of Marajó, according to Gordon 2005, citing Arnaud 1969. The Ethnologue.com notes that the ancestors of the modern Karipúna spoke Karipúna, an unclassified language. Picanço Montejo notes that “Algumas vezes os pajés afirmam cantar numa “língua Karipuna” que teria sido utilizada pelos seus antepassados, o que associa este conhecimento lingüístico-musical a uma reconstrução do passado do povo, aumentando seu valor sagrado” (2000). [Translation: At times the pajés (‘medicine men’) have confirmed that they sing in the “Karipúna language” which would have been used by their ancestors, and this linguistic-musical knowledge is linked to a reconstruction of the people’s past, thereby increasing the sacred value of the language.] See also S.J. Tobler 1979/1980. . The name for Uaçá in Palikúr (an Arawak language) is Karipura, and this Palikúr word appears to be the origin of the name of the Karipúna who come from the Uaçá area (Diana Green, pc, February 1998). The Palikúr, among whom Green and Green lived and worked for several years, have their own stories about the origins of the relative newcomers, the Karipúna and the Galibi-Marwono. Ladhams (1995) touches briefly on the Carib origin of the name Karipúna, and notes that Karipúna was the name of a Brazilian lingua franca, and also used for any nonTupi group. . FUNAI notes that many lexical items attributed to old Galibi are similar to Trió (Tyrió), a Carib language. . Oyapock is the French spelling and Oiapoque is the Portuguese spelling. Oiyapoque (or Oyapoque), the spelling chosen here, is sometimes used in the Anglophone literature, cf. Ethnologue. com, and is the choice of the authors, as it represents a combination of the two orthographies.

Chapter 15. Amazonian (Karipúna) French Creole 

(or Kaliña, cf. Tassinari 1998), and Palikúr. Of the three Amerindian reserves, the Uaçá is by far the largest and most important. While most of the Amerindians in this area speak French Creole to varying degrees, it is only the Karipúna and the GalibiMarworno who speak it as their mother tongue. Location of French creole speakers in Brazil

(© 2001, SIL International. Used by permission.)

Map 1. Location of Karipúna and Amapá French Creole speakers in Brazil

This paper will deal mainly with these two groups, and their language varieties will be referred to as KFC and GMFC, respectively, and together grouped as Amazonian French Creole (AFC), Amapá French Creole (APFC) being reserved for the southern variety of the city of Macapá. Kheuól, Crioulo and Patuá14 are written forms for the names of the language in the Brazilian literature, and these will be used, where appropriate, as cover terms for both varieties of northern Amazonian French Creole. (In the literature, Karipúna has generally been used as a cover term for both varieties. Here we are using AFC interchangeably with KFC and GMFC.)

. FUNAI also uses Patoá or Kreoul.

 Jo-Anne S. Ferreira and Mervyn C. Alleyne

CAYENNE % SEE INSET Atlantic Ocean

Atlantic Ocean

French Guiana

Riv

Jum ina

St. Georges ! er iv eR ! u q po Oiapoque ia O

er

French Guiana

er

iv

iR

ip

r Cu Hwy 156

r

ve

ua

Brazil

Ri

a uc

r

U Mouths of the Amazon

!

Ywauka

!

Kumaruma

Brazil Macapa #

Uaca River

(Maps © 2003, SIL International. Used by permission.)

Map 2. The State of Amapá Brazil

.

Map 3. Northern Amapá

Social history of the Karipúna

While French Creole is spoken as a mother tongue by the Karipúna and the Galibi- Marwono, it is also the trade language of the area, and is understood and spoken to varying degrees by neighbouring Amerindian groups such as the Palikúr and the Wayampí (descendants of the original inhabitants of the area). The Palikúr have been more exposed to both varieties of AFC and the Wayampí to Guianese French Creole (GFC, called ‘khéuol fin’ by speakers of AFC). There is limited bilingualism among the Karipúna, some speakers having varying degrees of competence in both Portuguese and KFC and others being completely monolingual in KFC. There is, however, a higher degree of bilingualism to be found among the Galibi-Marwono, according to FUNAI (2001). Ladhams (1995: 118; cf. Röntgen 1998) provides possibly the best synopsis of the social history of the Karipúna (very little research has been done on the origins of the Galibi-Marwono). Seeking to respond to previously unanswered questions raised by Holm (1989: 381), Ladhams’s investigation provides clear answers to the uncertainty surrounding the sociohistorical origins of KFC. An understanding of the social history of the Karipúna people (and by extension, the Galibi-Marwono) is necessary for understanding the origins and development of their language. According to Ladhams (1995: 118), “the historical scenario for the Karipúnas might be summarized as follows”:

Chapter 15. Amazonian (Karipúna) French Creole 

having moved first in 1830 from Pará State in Brazil some 500 km northwards to eastern Guiana, the Amerindians were joined by a diverse mixture of ethnic groups on the Curipi River, in what is now the Brazilian State of Amapá, having probably moved there in the 1840s. By the beginning of the present century, this amorphous community had expanded to the banks of the Oyapock River, and they were in continuous contact with French and/or Guyanais speakers.

Based on Ladhams’s analysis, the various stages in the history of the Karipúna are enumerated as follows (see Ladhams 1995 for a map of both French Guiana and Brazil): 1.

2. 3. 4. 5.

6. 7.

Because of the Cabanagem Revolt15 in the state of Pará during the mid-1830s, several Tupi-speaking Amerindians moved away from the Breves Strait, Pará at the mouth of Amazon. This group moved north-west to the Ouanary River16 in French Guiana. They then moved further south to the Oiyapoque area (Brazil), which was once territory hotly disputed by French Guiana and Brazil. They then moved back east, to the north of the Curipi River (Brazil). Because of a smallpox epidemic, they then moved to the south of the Curipi River. They were later joined by non-Amerindian Brazilians, of a variety of origins, including French Guianese Creole speakers. These were “displaced Brazilians … called Karipúnas” from the Curipi and Oiyapoque rivers, some of whom spoke a dialect of Wayampí, a Tupi language. They may have come into contact with French and French Creole-speaking missionaries in the Uaçá area. A gold rush in Approuague, French Guiana (another river further west) in 1854 caused many gold miners and would-be gold miners of various origins (including French and French Creole-speaking Martiniquans17 and others) to move to Oiapoque and Curipi in search of gold.

Ladhams (1995: 117) also notes that in the late nineteenth century, “all Amerindians, except the Palikúr, were fluent in Creole” (citing Coudreau 1893: 378).18 Today that has changed and many Palikúr also speak Kheuól (Diana Green, pc, February 1998).19 . The Cabanagem Revolt of 1835–37 was a separatist movement, leading to the creation of the “Republic of Pará”, and involved oppressed Amerindians and Africans (formerly enslaved), as well as some Pará-based members of the Brazilian plantocracy who resented governance from the south of Brazil. The word cabanagem refers to a type of Amazonian dwelling used by ribeirinhos (river-dwellers, comprising Amerindians, Africans, and mestiços). . The commune of Ouanary is just north of the commune of St. Georges (both communes on are the Oiyapoque river, bordering Brazil). .

Interestingly, the Oiyapoque area was also once known as Martinica, according to informants.

. It would be interesting to ascertain the extent to which the history of the Galibi-Marwono is similar to that of the Karipúna and at what point the two groups emerged as different ethnic groupings. . See also the Palikúr-Portuguese-Kheuól word list by Green and Green 1996, and FUNAI 2001.

 Jo-Anne S. Ferreira and Mervyn C. Alleyne

Kheuól or Patuá was the lingua franca of this area that was becoming increasingly multilingual due to contact among native Amerindians speaking diverse languages from both Tupi and Carib families, Guianese French Creole speakers, Brazilian Portuguese speakers, and others. Since the Karipúna themselves were becoming more and more mixed and in need of an in-group language, GFC, as Ladhams (1995: 118) puts it, “was a prime candidate” for the language choice of the Karipúna, since it was no doubt in use among (some of) the Amerindian and (the majority of) the nonAmerindian members of the group for a number of years. Another factor in the choice of French Creole “would have been the continuous contact with speakers of French and GFC on both sides of the Oyapock River, even after the settling of the border dispute in 1900” between Brazil and France (Ladhams 1995: 118). In fact, the language of the Oiyapoque area in the 1900s was either French or French Creole (Ladhams 1995: 117–8, referring to another source, Reis 1949: 171). Today, the Karipúna people are a highly mixed group, comprising not only descendants of Amerindians, but also of Africans, Asians and Europeans, and mixtures thereof. AFC is the only French Creole natively spoken in lusophone territory in the Americas. As the official language of Brazil, Portuguese remains the prestige language, and continues to dominate the educational system of the Oiyapoque area of Uaçá, northern Amapá. However, children in the creolophone Amerindian reserves receive bilingual education during the first three years of primary schooling, which helps to reinforce the community role of Kheuól or Patuá while at the same time introducing the children to Portuguese (Spires 1987; Tassinari 1997).20 The result for some Karipúna and for most Galibi-Marwono is bilingualism in French Creole and Portuguese.

.

Demographic survey

According to FUNAI (2001: 5), the current figures for the mostly creolophone Karipúna are approximately 1,726, while the Galibi-Marwono of Uaçá number approximately 1,787 (see Tables 1 and 2). FUNAI also counts 69 bilingual mixed Palikúr-Karipúna21 in the village of Flecha, which would bring the total up to 1,795 Karipúna (in contrast to Ethnologue.com’s figures of 672 for the whole group). The Galibi-Marwono are concentrated mostly in one large village, Kumarumã, comprising 300 families, nearly all of whom are bilingual in Kheuól and Portuguese. . Several didactic materials in AFC are used in the CIMI bilingual education programme among the Karipúna (Picanço Montejo 1983, 1985a, b, c, d, 1988; Tobler n.d.). There exist published traditional and translated stories also in use in the schools (Forte et al 1983; Ruffaldi and Spires 1996a, b; Spires 1997a, b; Tobler and Tobler 1983a, b). . FUNAI notes that approximately 969 of the 999 Palikúr are bilingual in Palikúr (an Arawakan language) and Kheuól, and there are 30 Galibi do Oiapoque who speak Galibi (a Carib language), Kheuól and French, bringing the total number of northern Amazonian French Creole speakers up to 4,493 in the year 2000.

Chapter 15. Amazonian (Karipúna) French Creole 

The Karipúna, on the other hand, are scattered among about sixteen villages, the largest of which is Manga, with 465 residents belonging to about 76 families. The population of two other villages, in the area of Juminã, comprises a mixture of Karipúna and Galibi-Marwono people. Interestingly, the more accessible of the two villages, Juminã-Uahá (nearer to the Oiyapoque River), remains more conservative and more monolingual in Kheuól, while the more remote Juminã-Kunanã (further away from the Oiyapoque River, up the igarapé or creek, and therefore farther away from the city of Oiapoque) is largely bilingual in Kheuól and Portuguese. The differences are not only linguistic, but religious, since Juminã-Uahá is distinctly more Catholic, while Juminã-Kunanã is more Protestant.

Table 1. Population of Karipúna Villages

Table 2. Population of Galibi-Marworno Villages

Table 3. Population of Palikúr Villages

Manga Zacarias Japiim Paxiubal Santa Isabel Taminã Espírito Santo Jõdef Açaizal Encruzo Ariramba Kunanã Piquia Curipi Karia Estrela Total:

Kumarumã Uahá Samauma Tukay Total:

Kumenê 572 Flecha 69 Tawary 44 Kamuywá 49 Kwikwit 29 Amomin 31 Puaytyeket 68 Urubu 42 Mangue II 46 Ywauka 49 Sub-total: Palikúr in Brazil 999 Palikúr in French Guiana* 500 Total: 1,499

465 32 33 44 262 54 351 64 91 25 62 76 19 32 29 87 1,726

1,578 81 58 70 1,787

Table 4. Population of Galibi do Oiapoque Village Galibi Total:

*(Queixalós 2000: 299)

30 30

Total number of French Creole speakers: 4,493 Sources: FUNAI 2001 and Queixalós 2000.

The preceding tables list the populations of the Uaçá reservations of northern Amapá, showing the numbers according to villages. A look at population statistics from 1943 to 2001 shows that the Karipúna population in 2001 is six times larger than it was in 1943. No comparable information is available for the Galibi-Marwono, since most academic interest has focused on the Karipúna to date (see also Azevedo 2001).

 Jo-Anne S. Ferreira and Mervyn C. Alleyne

Table 5. Population statistics of the Karipúna of Amapá from 1943–2001

.

YEAR

TOTAL

SOURCE

1943 1969 1983 1988 1989 1993 1998 2001

262 439 400–600 880 500 1,353 1,250 1,726

Rondon 1953: 282 Arnaud 1969: 8 S.J. Tobler 1983: 10 Picanço Montejo 1996: 8 Holm 1989: 381 Ricardo 1995: 39 D. Santa Rosa (pc 1998) FUNAI 2001

French Guianese, Karipúna and Galibi-Marwono French Creole

Social historical evidence suggests that AFC is an offshoot of one of the dialects of nineteenth century Guianese French Creole (Corne 2002). There are conflicting reports about how different it is from GFC (for analyses of GFC, see Corne 1971; Reighard 197422; Saint-Jacques-Fauquenoy 1974, 1972). According to a survey done by Graham in 1985 (referred to in Holm 1989: 381), speakers of GFC have a 77% understanding of tape recordings of KFC, and a 78% comprehension rate for St. Lucian Creole.23 Within AFC itself, phonetic differences between KFC and GMFC lie mainly in vowel choice. For example, of the mid front unrounded vowels, GMFC prefers the open-mid vowel [ɛ] in open syllables, while KFC prefers the close-mid vowel [e], as in late [late] (KFC) vs. laté [latɛ] (GMFC) ‘earth’, and bõswe [bõswe] (KFC) vs. bõswé [bõswɛ] (GMFC) ‘good afternoon/evening’ (Ferreira 1998). Vidal (2000) and Picanço Montejo (2000) confirm this. The latter states: Entre os Karipuna e os Galibi Marworno que falam a mesma língua, a única diferença entre eles é de pronúncia. Os Karipuna pronunciam com um som mais fechado e os Galibi com um som mais aberto, porém a grafia é a mesma. Exemplo: Piebua—árvore (grafia comum entre os dois povos). [piêbua]24 (pronúncia Karipuna—som fechado) [piébua] (pronúncia Galibi—som aberto)25

.

See also Jennings 1995.

. The differences among the South American and the Lesser Antillean varieties appear to be in the areas of phonology and lexicon, with minor morpho-syntactic differences. (See Appendix II.) .

Note that the value of ê in Portuguese is [e], and é is [ɛ]. See Appendix I.

. [Translation: Among the Karipúna and the Galibi-Marwono, speakers of the same language, the only difference between the two groups is that of pronunciation. The Karipúna use a more close sound and the Galibi more open, but the orthography is the same. For example piebua ‘tree’ (same orthography for the two groups): [piêbua] [piebua] (Karipúna pronunciation—close sound), and [piébua] [piɛbua] (Galibi pronunciation—open sound).]

Chapter 15. Amazonian (Karipúna) French Creole 

Other differences include the following: kaná [kaˈna]26 (KFC) vs. kanã [kaˈnã] (GMFC) ‘duck’ gho [ɡwo] (KFC) vs. gro [ɡro] (GMFC) ‘large’, and pãi [pãɪ̃] (KFC) vs. pãn [pãn] (GMFC).

.

Sources of Kheuól lexicon

The vast majority of AFC lexicon is, of course, derived from French. Of particular historical interest is the case of lexical items that contain agglutinated forms of the French articles. There are basically two types of agglutination: consonantal and syllabic. The following data are samples of agglutination and were retrieved from A.W. Tobler (1987: 85–86) and Corrêa and Corrêa (1998). They give an idea of the extent of the historical productivity of the process of article agglutination. According to Grant 1995 (on which the following lists are based27), some classes may remain open, such as Class I; others are closed, such as Class II. (Note that the transcriptions used here are for Modern AFC, as per Montserrat and Silva (1984), and for Modern Standard French. The orthography used here is the official AFC—Karipúna/Galibi—orthography.) C I

Karipúna nominal

French definite singular non-partitive article (la, l’) + Nominal Note that this class may be an open class and new words borrowed from French may keep the French article. labu /labu/ (la) boue /labu/ ‘mud’ laduan /laduan/ (la) douane /ladwan/ ‘customs’ lãfé /lãfɛ/ (l’) enfer /lãfɛʀ/ ‘hell’ lafiév /lafiɛv/ (la) fièvre /lafjɛvʀ/ ‘fever’ lafime /lafime/ (la) fumée /lafyme/ ‘smoke’ laxas /laʃas/ (la) chasse /laʃas/ ‘hunting’ lide /lide/ (l’)idée /lide/ ‘idea’ lizin /lizin/ (l’)usine /lyzin/ ‘factory’ ló /lɔ/ (l’)or /lɔʀ/ ‘gold’ lótél /lɔtɛl/ (l’)autel /lɔtɛl/ ‘altar’ C II

Karipúna nominal zabapẽ /zabapɛ̃/ zafé /zafɛ/ .

French definite (les)/, indefinite (des) non-partitive plural article + Nominal (les) arbres à pain /lezaʀbʀapɛ̃/ ‘breadfruit’ (les) affaires /lezafɛʀ/ ‘business’

The orthographical acute accent on a represents primary stress.

. There are many more data samples, but we have chosen the lexemes most common across varieties.

 Jo-Anne S. Ferreira and Mervyn C. Alleyne

zahiko /zahiko/ zãj /zãʒ/ zam /zam/ zéklé /zɛklɛ/ zepẽ /zepɛ̃/

(les) haricots /lezaʀiko/ (les) anges /lezɑ̃ʒ/ (les) armes /lezaʀm/ (les) éclaires /lezeklɛʀ/ (les) épingles /lezepɛ̃ɡl/

‘bean’ ‘angel’ ‘gun’ ‘lightning’ ‘pin’

zépól /zɛpɔl/ zétuél /zɛtwɛl/ zo /zo/

(les) épaules /lezepol/ (les) étoiles /lezetwal/ (les) os /lezo/

‘shoulder’ ‘star’ ‘bone’

C III

Karipúna nominal nak /nak/ nam /nam/ nót /nɔt/

French indefinite singular non-partitive article (un, une) + Nominal (un) arc /œnaʀk/ ‘arch, ark’ (une) âme /ynam/ ‘soul’ (un) autre /œnotʀ/ ‘another’ C IV

Karipúna nominal djife /dife/ djilét /dilɛt/ djilo ~ dlo /dilo/ ~ /dlo/ djipeẽ /dipɛ̃/ djisã /disã/ djisél /disɛl/ djize /dize/ djivẽ /divɛ̃/ or djivã duhi /duhi/

French (singular or plural) partitive article (du, de l’, des) + Nominal (du) feu /dyfø/ ‘fire’ (du) lait /dylɛ/ ‘milk’ (de l’) eau /dәlo/ ‘water’ (du) pain /dypɛ̃/ ‘bread’ (du) sang /dysã/ ‘blood’ (du) sel /dysɛl/ ‘salt’ (des) oeufs /dezœ/ ‘egg’ (du) vin /dyvɛ̃/ ‘wine’ (du) riz /dyʀi/ ‘rice’

These agglutinated forms are interesting in several ways. On the one hand, a considerable number of them are common to all French creoles. But there are also a number of forms that are attested in only one variety of French Creole or in just a few. The common forms represent a similarity that could hardly be explained as chance. They therefore support a hypothesis that French creole language varieties go back historically to some single form which was later diffused to other areas and there developed further forms. In a future project, a comparative study of French Creole varieties will be undertaken in order to determine: 1. 2. 3.

if there is any linguistic motivation for agglutination if these languages can be organised into sub-groupings based on shared agglutinated forms if there are any inferences that can be made concerning the genesis and history of these languages.

Apart from the obvious French influence, a large number of words for flora and fauna are clearly Amerindian in origin, from both the Tupi and Carib families. As Corne

Chapter 15. Amazonian (Karipúna) French Creole 

(1985: 233) notes, “of Tupi there appears to remain but little: some lexical items (flora, fauna. . .), the content of some tales and riddles, and perhaps other traces”. Examples include the following: ahe – parrot atxipa – type of fish bakurau – type of fish iaia – type of fish iauanaú – janaú (a type of animal) ipapu – type of fish kanaxi/kamaxi – type of fish kuabio – cassava by-product kulubhi – colibri kusiri – type of animal masuhu – type of fish tauahu – tracajá (a species of turtle) ture – ceremonial dance (probably of Tupi origin) urariri – type of animal

Some Kheuól words are used side by side with Portuguese words, for example, stilo and kanét (‘pen’), lafinét and janél (‘window’). There is also some Portuguese influence in the lexicon and in the phonological shape of Portuguese loan words.28 Some examples of Portuguese influence are seen in Portuguese words that have been integrated into AFC, which appears to prefer closed syllables. These loan words have been adapted to the phonology of AFC by syllable restructuring, specifically apocope, with the deletion of the final vowel as in List 1, and the deletion of a CV in List 3; for example: List 1 – apocopation of final V amig (> amigo) ‘friend’ Joán > Joana proper name kam (> cama) ‘mattress’ (lit. ‘bed’) kanét (> caneta) ‘pen’ kap (> capa) ‘cover’ lag (> lago) ‘lake’ pak (> paca) ‘lappe’ (agouti paca) puls (> pulso) ‘pulse’ zagai (> zagaia) ‘type of bird’

The word kaz ‘house’ is found elsewhere in other varieties of French Creole. Informants link this word to Portuguese casa ‘house’.

. FUNAI notes that the Karipúna variety is more aportuguesado (Portuguese-influenced) than the Galibi-Marwono variety. This is an interesting observation, since informants in the Oyapoque area noted that most of the Galibi-Marwono village of Kumarumã was bilingual in Kheuól and Portuguese, more than for the Karipúna.

 Jo-Anne S. Ferreira and Mervyn C. Alleyne

List 2 – apocopation of V.CV and VC depós (> depósito) ‘deposit’ lap (> lápis) ‘pencil’ List 3 – apocopation of final CV or Glide +V benefis (> benefício) ‘favour’ hadj (> rádio) ‘radio’ heloj (> relógio) ‘clock’ hemed (> remédio) ‘medicine’ hos (> rosca) ‘coil’ List 4 – apocopation of final C deve (> dever) ‘to owe (money)’ fika (> ficar) ‘to stay’ (the origin is probably an older Portuguese stratum in GFC rather than due to modern Portuguese influence, cf. Goodman 1987)

Others directly from Portuguese with no syllabic restructuring include: List 5 kaho (> carro) ‘car’ laghimas (> lágrimas) ‘agony’ (lit. ‘tears’) bhiga (> briga) ‘fight’ (the origin is probably an older Portuguese stratum in GFC rather than due to modern Portuguese influence, cf. Goodman 1987)

.

Comparative perspective

In this final section of this paper, we deal with several issues within the scope of Historical Comparative Linguistics and of Creole Linguistics. 1. 2.

First, to situate KFC comparatively within the family of French creoles (however a priori or putative may be the concept of “family” here). Then, to consider the historical evolutionary development within French creoles and to examine how KFC fits into the evolutionary schema.

These issues will lead us to look at (1) diffusion as a factor in accounting for the comparative picture, and (2) genesis in the Creole Linguistic tradition vs. evolution in the Historical/Comparative tradition as the competing explanatory models to account for KFC. The earlier work on Karipúna did not pay much attention to the socio-historical circumstances in which this language originated, apart from an apparent assumption that it arose independently of le guyanais (Ladhams 1995: 115). The history of French colonial expansion, however, supports the view that the emergence of French Creole dialects in the different regions of the Americas was more of a “stepping stone” process

Chapter 15. Amazonian (Karipúna) French Creole 

than an independent “mushrooming”. Alleyne (1996: 26-8) summarises the historical scenario of French colonial expansion as follows: Les premiers établissements dans le Nouveau Monde furent d’ordre militaire ou paramilitaire, aussi bien légaux qu’illégaux. Des pirates et des filibustiers menaient, dès 1504, des opérations sur la côte brésilienne, puis ils établirent rapidement des colonies stratégiques rudimentaires aux Antilles. A partir de 1625, l’occupation se stabilise. En 1627, Belain d’Esnambuc débarqua à Saint-Christophe avec un contingent d’émigrants normands. Ce fut le point de départ d’un mode de peuplement qui se survivra jusqu’à la moitié du XVIIIème siècle, l’engagement. En 1630, un groupe d’aventuriers français et anglais occupèrent l’île de la Tortue. De là, les Français ne tardèrent pas à amorcer l’occupation de la partie occidentale de l’Ile d’Hispaniola, connue sous le nom de Saint-Domingue. Sous l’égide de la compagnie des Isles de l’Amérique fondée par Richelieu, des colons arrivèrent à la Guadeloupe en 1635, et plus tard dans la même année à la Martinique. . . . La première occupation de la Guyane se situe vers 1626, mais la première tentative de colonisation se situe vers 1642, avec l’arrivée de trois cents engagés. Après plusieurs tentatives infructueuses, une nouvelle expédition, en 1664, aboutit à la création d’une colonie plus stable. [La colonie française de la Nouvelle Orléans fut fondée en 1718].29

A diffusionist hypothesis is also supported by the high degree of mutual intelligibility among the dialects, consistent with the relatively short time depth of separation. In a diffusionist scenario, an evolving language (located initially in St. Kitts or Hispaniola), variable as a result of the different regional dialectal origins of the French colonists and engagés and as a result of the contact with the different languages of Africans (and of other peoples), began to display a further variation between older forms and newly emerging forms. In each new location, this evolving language underwent further changes, independent and divergent, but with some contact and exchanges between the different parts of the French colonial empire. We thus have a configuration

. [Translation: The first New World settlements were military or paramilitary, both legal and illegal. As early as 1504, pirates and filibusters were carrying out operations on the Brazilian coast, followed quickly by rudimentary strategic colonies in the Antilles. From 1625, occupation stabilised. In 1627, Belain d’Esnambuc came ashore in St. Kitts with a contingent of Normand emigrants. This was the beginning of a model of settlement which would survive up to mid 18th century, indentureship. In 1630, a group of French and English adventurers occupied Tortuga. From there, the Frenchmen soon began the occupation of the western part of Hispaniola, known as Saint-Domingue. Under the aegis of the Compagnie des Iles d’Amérique founded by Richelieu, colonists arrived in Guadeloupe in 1635, and later in the same year in Martinique. . . The French presence in Guyane dates from about 1626, but the first attempt at colonisation took place about 1642 with the arrival of 300 indentureds. After several unsuccessful attempts, a new expedition, in 1664, led to the creation of a more stable colony, [The French colony of New Orleans was founded in 1718.]]

 Jo-Anne S. Ferreira and Mervyn C. Alleyne

of dialects showing a common inherited core but with a number of idiosyncratic developments in particular territories or in particular sub-zones (Haiti, Lesser Antilles, French Guiana, Louisiana, Indian Ocean). Another feature of this historical scenario is that at the present time each dialect may show changes that are not complete, that is, are not totally “regular” (i.e., they do not affect all the pertinent forms), so that there are many cases of new idiosyncratic developments existing side by side, in one and the same dialect, with earlier forms. Both the comparative configuration and the picture of individual dialects reveal a pattern of chronological layers. This scenario is akin to that of the earlier development of Romance languages. Latin (not at all a homogeneous language but one with geographical, social and stylistic variation) is taken, to some extent progressively and chronologically, to the different regions of the Roman Empire. In each location, this relatively common core undergoes changes which show different levels of divergence from the inherited model. Both comparatively across the different emerging dialects and within any one dialect, there is variation between older forms and newer forms, representing different chronological layers. The different regions and their dialects can thus be characterised, relatively speaking, as “innovative” (French), “conservative” (Italian, Rheto-Roman), “isolated” (Romanian) (cf. Diez who states “si l’on embrasse l’ensemble de la langue française, on s’aperçoit bien vite que l’élément latin y est moins fort que dans l’espagnol et l’italien”30 (1874: 107)). It is very important to give KFC a rating on a conservative to innovative scale. This rating will be based on comparative and internal linguistic evidence as well as on nonlinguistic socio-cultural historical data. The geographical location of KFC on the periphery of the French Creole geographical zone and its relative isolation, cutting it off from significant contact with, and influence from, other central innovating areas, are the main non-linguistic factors in rating KFC as relatively conservative. The significance of isolation is that these peoples have been cut off from meaningful societal contact with other FC speakers and the number of innovations due to borrowing will be consequently relatively small. GFC (or le guyanais), which, as we said above, the social history suggests as the immediate ancestor of Karipúna, is itself, by virtue of its location on the periphery, to be considered as a relatively conservative dialect. To a large extent, KFC retains the conservative forms of GFC. It is reasonable to suppose that these forms belong to an earlier historical level of French Creole, rather than that they are innovations. Some examples are provided in the discussions below. Because of the very complex (and complicated) regional dialect configuration of the French language in the sixteenth/seventeenth century, with le francien consolidating its status as the national official language, it is not always easy to determine the status of

. [Translation: If we take an overall look at the French language, we immediately realise that the Latin element is less significant there than in Spanish and Italian.]

Chapter 15. Amazonian (Karipúna) French Creole 

some forms in terms of the chronological layer to which they belong and their source. French Creole dialects often extend vowel nasalisation beyond the phenomenon of nasalisation in standard French. But so do many French regional dialects. French Creole dialects inherit all the nasal vowels of French (except [œ̃], the nasalised openmid front rounded vowel). In addition, vowels in an open syllable and followed by a nasal consonant undergo regressive nasalisation, in some cases sporadic, in French Creole dialects (they normally do not in standard French): Table 6. Examples of nasalisation in four varieties of French Creole Haitian

Martiniquan

St. Lucian

Guadeloupean

Gloss

zãmi lãni bã nu

zãmi lãni bã nu

zãmi lãni bã nu

zãmi lãni bã nu

‘friend’ ‘anise’ ‘give us’

The most outstanding nasal feature is that three of these dialects—Haitian, Martiniquan, St. Lucian—show a strong tendency towards progressive nasalisation (which is unknown in standard French), as in the following examples (Table 7): Table 7. Haitian

Martiniquan

St. Lucian

Guadeloupean Karipúna

Gloss

fẽmẽ rẽmẽ bẽɲẽ ɡumẽ zawẽɲẽ pãɲẽ ʒãmẽ

fẽmẽ ẽmẽ bẽɲẽ ɡumẽ

fẽmẽ ẽmẽ bẽɲẽ ɡumẽ zaɡriɲẽ

feme ẽme bẽɲe ɡume

‘close’ ‘love’ ‘bathe’ ‘fight’ ‘spider’ ‘basket’ ‘never’

feme eme beɲe ɡume areɲe

Only two such forms have so far been attested in GMFC: kanã ‘duck’ (>French canard), and kanũ ‘canoe’. Both types of nasalisation, progressive and regressive, are attested in north-western dialects (Norman and Picard) and central dialects (Angevin) of French (Faine 1939: 59). One interpretation of the French Creole phenomenon is that Haitian, Martiniquan and St. Lucian retained this early French dialectal progressive nasalisation, while the other dialects simply never had it or else underwent denasalisation change. This would, however, be counter-intuitive. Nasalisation of vowels, progressive or regressive, is far more common as a process in the languages of the world than denasalisation. In creole languages, although both processes are present, nasalisation is the rule and denasalisation the exception. In addition there is other evidence that suggests that progressive nasalisation in the three French Creole dialects is a later exceptional development. The personal pronouns of first and second singular are based on a French root with initial [m-], [t-] respectively. (The earlier forms show a case system which

 Jo-Anne S. Ferreira and Mervyn C. Alleyne

has been well preserved only in the Indian Ocean dialects.) For example, an eighteenth century Haitian text—the poem/song “Lisette Quité la Plaine” by Duvivier de la Mahautière (1757) (DucœurJoly 1802)—records nominative [mo] and accusative [mwe], and this latter form exists in Haiti today only in the fixed expression [ã mwe], literally ‘to me’ (French “au secours”). Haitian and the dialects of the Lesser Antilles are the only ones in which progressive nasalisation has produced the form [mwẽ]. It is reasonable to suppose that Haitian and the Lesser Antillean dialects innovated with [mwẽ]. The other dialects have no earlier attestation of [mwẽ]. In fact, the Guianese and Karipúna forms [mo] and [to] are most likely the unstressed forms of [mwe] and [twe] (cf. zozo > French “des oiseaux” [dezwazo] > [zwazo] > [zozo]). There are other forms in Haitian which we may rate as relatively innovating, supporting our evaluation of Haitian on the basis of several non-linguistic features: geographical centrality, cultural dynamism (e.g., the number of innovative forms of music, art, religion, cuisine), social political revolution. Haitian possesses a number of highly innovative forms not to be found anywhere else and not attested in earlier historical stages of Haitian. For example: ●

● ● ●

●

nasalised high vowels: /ĩ/ and /ũ/; for example, pĩgã ‘take care’ (cf. Guadeloupe [pẽɡã]); [ũɡã] ‘priest’) 2nd pers. plur. pron /nu/ definite article /–la/ attached to expanded NPs and relatives extensive syntactic reanalysis (capable, après, finir, aller, gagner, rester, sortir, mettre, connaître, venir become grammatical morphemes of Tense, Mood, Aspect) fuller Predicate category, with a wider range of sub-categories: Proper Noun, Noun, Adjective (in addition to Verb), otherwise stated as a higher degree of multifunctionality.

Needless to say, there is little or no evidence in AFC of the type of variation observable in cases where a French Creole language exists side by side with French; sometimes referred to as a creole (or post-creole) continuum. In the case of KFC, variation is observable in the context of juxtaposition with Brazilian Portuguese and manifests itself in the phonology and in the lexicon. There are, however, some problems of analysis that require further investigation. For example, dental stops are always palatalised before the close front unrounded vowel [i] (as they are in most varieties of Brazilian Portuguese). This palatalisation is also observable in Haitian, albeit to a lesser degree (e.g., djab (