sign in sign up
<<
Home , Aymara language, Carib language, Carijona language, Carlos Chinin, Huni Kuin, Languages of Bolivia

This file is freely available for download at http://www.etnolinguistica.org/illa This book is freely available for download at http://www.etnolinguistica.org/illa

References:

Crevels, Mily, Simon van de Kerke, Sérgio Meira & Hein van der Voort (eds.). 2002. Current Studies on South American , [Indigenous Languages of America (ILLA), vol. 3], [CNWS publications, vol. 114], Leiden: Research School of Asian, African, and Amerindian Studies (CNWS), vi + 344 pp. (ISBN 90-5789-076-3)

CURRENT STUDIES ON SOUTH AMERICAN LANGUAGES INDIGENOUS LANGUAGES OF (ILLA)

This series, entitled Indigenous Languages of Latin America, is a result of the collaboration between the CNWS research group of Amerindian Studies and the Spinoza research program Lexicon and , and it will function as an outlet for publications related to the research program.

LENGUAS INDÍGENAS DE AMÉRICA LATINA (ILLA)

La serie Lenguas Indígenas de América Latina el resultado de la colabora- ción entre el equipo de investigación CNWS de estudios americanos el programa de investigación Spinoza denominado Léxico y Sintaxis. Dicha serie tiene como objetivo publicar los trabajos que se lleven a cabo dentro de ambos programas de investigación.

Board of advisors / Consejo asesor:

Willem Adelaar (Universiteit Leiden) Eithne Carlin (Universiteit Leiden) Pieter Muysken (Katholieke Universiteit Nijmegen) Leo Wetzels (Vrije Universiteit, Amsterdam)

Series editors / Editores de la serie:

Mily Crevels (Katholieke Universiteit Nijmegen) Simon van de Kerke (Universiteit Leiden) Hein van der Voort (Katholieke Universiteit Nijmegen)

The word illa means ‘amulet’ in Aymara and Quechua. INDIGENOUS LANGUAGES OF LATIN AMERICA (ILLA) 3

CURRENT STUDIES ON SOUTH AMERICAN LANGUAGES

Contribuciones seleccionadas del 50 Congreso Internacional de Americanistas en Varsovia y del Taller Spinoza de Lenguas Amerindias en Leiden, 2000

Selected papers from the 50th International Congress of Americanists in Warsaw and the Spinoza Workshop on Amerindian Languages in Leiden, 2000

Edited by

Mily Crevels Simon van de Kerke Sérgio Meira Hein van der Voort

Research School of Asian, African, and Amerindian Studies (CNWS) Universiteit Leiden The Netherlands 2002

This file is freely available for download at http://www.etnolinguistica.org/illa CNWS PUBLICATIONS Vol. 114

Indigenous Languages of Latin America (ILLA), Vol. 3

CNWS PUBLICATIONS is a series produced by the Research School of Asian, African and Amerindian Studies (CNWS), Leiden University, The Netherlands.

Editorial board: F. Asselbergs; . Forrer; F. Hüsken; . Jongeling; H. Maier; P. Silva; B. Walraven

All correspondence should be addressed to: Dr. .. Vogelsang, CNWS Publications, c/ Research School CNWS, Leiden University, PO Box 9515, 2300 RA Leiden, The Netherlands.

Tel. +31 (0)71 5272171 Fax. +31 (0)71 5272939

Current studies on South American languages

Current studies on South American languages; edited by Mily Crevels, Simon van de Kerke, Sérgio Meira and Hein van der Voort - Leiden 2002: Research School of Asian, African and Amerindian Studies (CNWS), Universiteit Leiden. CNWS Publications, Vol. 114, ISSN 1567-813X)

ISBN: 90-5789-076-3

Printing: Ridderprint, Ridderkerk

Front cover design: Nelleke Oosten. Illustration shows members of the Chono family from San Juan de Kelekelera, browsing through the primer of the Leko that was written and distributed among them by Simon van de Kerke in July 2000. Theodoro Chono, one of the youngest semi-speakers of Leko, is sitting in the middle.

© Copyright 2002 Research School CNWS, Leiden University, The Netherlands

Copyright reserved. Subject to the exceptions provided for by law, no part of this publication may be reproduced and/or published in print, by photocopying, on microfilm or in any other way without the written consent of the copyright-holder(); the same applies to whole or partial adaptations. The publisher retains the sole right to collect from third parties fees in respect of copying and/or take legal or other action for this purpose. CONTENTS

The editors 1 Introduction

Sociolinguistics and history of the field

Mily Crevels (L) 9 Why speakers shift and languages die: An account of in Amazonian Wolf Dietrich (W) 31 Guaraní criollo y guaraní étnico en , y Brasil Sieglinde Falkinger (L) 43 Diferencias entre el lenguaje de hombres y mujeres en (Besiro) Utta von Gleich (L) 57 Multilingual literacies in Bolivia Rotger Michael Snethlage (W) 75 Leben, Expeditionen, Sammlungen und unveröffentlichte wissenschaftliche Tagebücher von Dr. Emil Heinrich Snethlage Life, expeditions, collections and unpublished field notes of 89 Dr. Emil Heinrich Snethlage

Phonology & phonetics

Astrid Alexander-Bakkerus (W) 103 Nominal morphophonological processes observed in Pedro de la Mata’s Arte de la Lengua Cholona (1748) Esther Herrera Z. (W) 111 Las estructuras fonéticas de la lengua emberá Esteban Emilio Mosonyi (W) 127 La fonología suprasegmental y otras particularidades del baniva de Maroa, idioma tonal arawak del Río Negro, Pedro Viegas Barros (W) 137 Fonología del Proto-Mataguayo: Las fricativas dorsales

Morphosyntax

Eliane Camargo (W) 149 Cashinaua personal pronouns in grammatical relations Alain Fabre (W) 169 Algunos rasgos tipológicos del Kamsá (Valle de Sibundoy, Alto Putumayo, sudoeste de ) vistos desde una perspectiva areal Elena Filimonova (L) 199 Person hierarchy and its implication: The case of Aymara Colette Grinevald (L) 215 Nominal classification in Movima Simon van de Kerke (L) 241 Complex verb formation in Leko Sérgio Meira (W) 255 A first comparison of pronominal and demonstrative systems in the Cariban Odile Renault-Lescure (W) 277 Le parfait en kali’na Jeanette Sakel (L) 287 Gender agreement in Mosetén Hein van der Voort (L) 307 The quotative construction in Kwaza and its (de-) grammaticalisation Mary Ruth Wise (W) 329 Applicative in Peruvian

The symbols (L) and (W) indicate the articles that originated in the Leiden and Warsaw symposia, respectively. The languages discussed in the present volume are indicated on the map at the end of the introductory chapter, on pages 6 and 7.

INTRODUCTION

This book contains a selection of articles based on papers presented at two different symposia on the indigenous languages of . The first of these two symposia took place July 10–14, 2000, at Warsaw University, during the 50th International Congress of Americanists in Warsaw, Poland. This symposium was entitled Lenguas Amazónicas y de las áreas adyacentes: aspectos descriptivos y comparativos / Languages in the Amazon and its neighbouring areas: descriptive and comparative aspects and was organised by Marília Facó Soares, José Alvarez and Hein van der Voort. The second symposium took place two and a half months later, on September 28–30, 2000, at Leiden University, Leiden, The Netherlands. This symposium was entitled Workshop on Bolivian and Rondonian indigenous languages and was organised by Mily Crevels and Simon van de Kerke. It represented the 3rd Workshop on Amerindian Languages organised at Leiden University. Because of the high degree of similarity and overlap between the topics of these symposia, the editors of the planned proceedings decided to collaborate on a single volume of articles on the indigenous languages of South America. The indigenous languages of South America form a highly diverse group. This part of the world boasts of an amazing wealth of genetic linguistic phyla and isolated languages and stocks. A large part of this wealth has not yet been documented and studied to any degree of satisfaction. Therefore, many proposals for genetic linguistic relationships are still completely unproven and, since a thorough attempt at linguistic classification is also a prerequisite for the label ‘isolated’, most of the so-called isolated languages are better considered as unclassified so far. Furthermore, many of the special grammatical and phonological characteristics of these languages will continue to go unnoticed by general and typological until serious fieldwork is conducted and subsequent publications become available. Unfortunately, the majority of the approximately 375 indigenous languages spoken in South America are in danger of disappearance under the pressure of national and colonial languages. This situation is, among other things, directly related to the fact that the indigenous peoples have been suffering physical, social, cultural and economic discrimination and genocide at the hands of Western invaders since the sixteenth century, and that the ecological destruction of their traditional habitat continues relentlessly until this very day. Therefore, the study of indigenous languages should not only bring scientific data and insights within the confines of its own discipline, but it should also provide a scientific basis for language preservation efforts and be co-instrumental in the emancipation of the indigenous peoples. Furthermore, the sharing of field experiences in wider circles may result in a higher level of general awareness of the value of cultural and linguistic diversity. With this collection of articles, the editors intended to provide an outlet for publication of the results of ongoing research on South American Indian languages. Hopefully it will inspire 2 SOUTH AMERICAN LANGUAGES discussion and exchange between colleagues and stimulate others to participate in carrying out the tasks that lie ahead. The majority of the contributions to the present volume relate to research that is the direct result of linguistic fieldwork by the authors. Most of the languages discussed are spoken in the Amazon basin, some are spoken in the region of the eastern foothills of the , and a few are from the Andes region itself. One linguistic family is from the . A fair number of the languages under discussion are genetic isolates from Bolivia. The major part of the articles are about morphosyntactic aspects of a specific language, a few deal with and phonetics, and a few others have a more typological or historical-comparative orientation. This volume also includes several sociolinguistic articles. According to these categories, the contributions to this volume are grouped in the following order: Sociolinguistics and History of the field (Crevels, Dietrich, Falkinger, von Gleich, and Snethlage); (Morpho-)phonology/Phonetics (Alexander-Bakkerus, Herrera, Mosonyi, and Viegas Barros); Morphosyntax (Camargo, Fabre, Filimonova, Grinevald, van de Kerke, Meira, Renault-Lescure, Sakel, van der Voort, and Wise). MILY CREVELS’ contribution deals with possible causes for language death in Amazonian Bolivia. The author sketches the historical background against which the colonization of the region took place, and discusses the events that have led to the extremely alarming current state of some of the languages in the region. In his paper, WOLF DIETRICH reports on the development of the Atlas Lingüístico Guaraní-Románico (ALGR), a linguistic atlas of the areas in which Guaraní Criollo is in contact with Spanish and Portuguese. After a brief historical introduction, in which the several varieties of Guaraní are mentioned, and a brief description of the current sociolinguistic situation in Paraguay, Argentina, and , the author presents several differences between the Guaraní varieties, based on data obtained in the atlas project. SIEGLINDE FALKINGER gives an insightful account of the differences between male and female language use in Chiquitano. Extensively quoting historical manuscript sources that are little known and even less accessible, the author sketches the panorama of the specific differences in the speech of males and females. UTTA VON GLEICH’s article reports on an ongoing research project on the everyday use of literacy among bilinguals in Bolivia. The aim of the project is to find out whether literacy in more than one language may become important in the near future, in everyday life situations beyond the education system and beyond specialised professional groups. The described research approach is innovative as far as the indigenous are concerned. The last contribution in the sociolinguistic section is by ROTGER MICHAEL SNETHLAGE, son of the German ethnographer Emil Heinrich Snethlage, who travelled in eastern Brazil from 1923 to 1926 and in the Guaporé region from 1933 to 1934. The author gives an overview of the scientific career of his father, whose life unfortunately came to an end at an early age. Snethlage’s death was a setback for the INTRODUCTION 3 scientific study of a region that was still untouched by Western culture at that time. The incomplete bibliography at the end of the article serves as proof of Snethlage’s dedication to his field of interest, which was wider than ethnography alone. Because of the symbolic and historical importance of this article, the original German version is also included in this volume. ASTRID ALEXANDER-BAKKERUS discusses morphophonological processes in Cholón, a probably extinct Cholonan language of the eastern slopes of the Peruvian Andes. Her work is mainly based on an eighteenth-century manuscript by Pedro de la Mata. This important manuscript contains a description of Cholón which is thorough enough to provide the specialist with useful evidence even on a component of language as volatile as morphophonology. In her contribution, ESTHER HERRERA Z. examines the phonetic correlates of some of the most important phonological features in the Chamí dialect of the Emberá language: those related to the quality of the middle , and to the stop and series. She then compares the various (and often contradictory) descriptions in the literature, and, based on her acoustic data, proposes a new analysis to settle the question. ESTEBAN EMILIO MOSONYI’s article contains a brief description of the suprasegmental features of the language from the Negro (Xié) river, which the author analyses as a case of ‘tonal accent’. He proposes three contrasting accents, which correspond to different tonal melodies. He also describes the current situation of the few remaining speakers of this language, taking a strong position against the inevitability of ‘language death’. PEDRO VIEGAS BARROS presents new data on the reconstruction of Proto- Mataguayo dorsal . This part of the system is very problematic from a diachronic point of view. The author compares two previous reconstructions and shows that they do not account regularly for the dorsal . He proposes the reconstruction of two fricatives — * (velar) and *X (uvular) — to account for the correspondences that, in the previous analyses, had remained irregular. Although a few problems remain (dutifully reported by the author), the author’s proposal clearly represents progress in the understanding of Proto-Mataguayan phonology. In her article, ELIANE CAMARGO discusses the personal pronouns and grammatical relations in Cashinahua (Kashinawa, a Panoan language). Case marking follows a person-based split ergative pattern. After describing the system, the author proposes a correction to previous analyses by Bob Dixon and Doris Payne. ALAIN FABRE discusses some typological features of the Kamsá language from an areal perspective. Basing himself on published data and studies, the author looks at noun incorporation and noun classification, which he places in a wider context by considering a set of eleven neighbouring languages, from the Andean, Amazonian, and Pacific regions of Colombia. The author shows that Kamsá shares features with languages of the Andean as well as the Amazonian area, as a consequence of the role of its speakers as intermediaries between these two worlds. 4 SOUTH AMERICAN LANGUAGES

ELENA FILIMONOVA launches an innovative approach to person hierarchy in Aymara by claiming that irregularities of person marking in this language are influenced by a special configuration of the person hierarchy, according to which the person spoken to overranks the Speaker: A>S>N. The author suggests that this layout of person hierarchy is determined by a special conception of ego in the Aymara culture. COLETTE GRINEVALD discusses the productive nominal classification system of Movima, an unclassified language of the Amazonian lowlands of Bolivia. Her data reveal that Movima has a mixed system that functions mostly on the basis of morphophonological characteristics and partly on the basis of semantic organization. It appears to be fairly typical of the not very grammaticalized kind of noun class systems found in languages of the Amazon region and, as such, it challenges the morphosyntactic typologies of nominal classification systems conceived so far on the basis of data from other parts of the world. In his article, SIMON VAN DE KERKE presents a semantic characterization of the different suffixes that are used in Leko, an unclassified language of the Andean foothills region of Bolivia, to form complex verbal expressions. Since many of these suffixes also occur in the language as independent verb roots, it remains an unresolved question whether complex verbs in Leko should be better analyzed as the result of derivation or of compounding. SÉRGIO MEIRA’s article compares the pronominal and demonstrative systems of , attempting to provide a preliminary reconstruction of the Proto- Cariban system. With a representative sample of 27 languages, and using previous preliminary reconstructions of Proto-Cariban phonology as a starting-point, MEIRA proposes tentative protoforms for the various pronouns, as well as some ideas concerning their origin and evolution. ODILE RENAULT-LESCURE describes the semantic values of a specific past tense in Kali’na (also known as Cariña, Galibi, or Carib), marked by the suffix -i. After summarising the person- and TAM-marking system for finite verbs, she presents a number of exemples illustrating the (temporal and aspectual) uses of -i as a perfect marker, and also some examples of a modal use (‘let me…’, ‘I want to…’). She concludes that the meaning of this verb form centers around the idea of ‘experience’. The contribution by JEANETTE SAKEL deals with gender agreement in Mosetén, which is also an unclassified language of the Andean foothills region of Bolivia. The richly developed gender agreement system in Mosetén is rather unique among the indigenous languages of South America and demonstrates once again the necessity of thorough linguistic investigations in the field on every single language. In his article, HEIN VAN DER VOORT reports on certain modal derivational suffixes in Kwaza, an unclassified Amazonian language of Rondônia, Brazil. He attributes the emergence of these suffixes to a specific grammatical construction that expresses quoted speech and he hypothesises that processes of degrammaticalisation were involved in their creation. INTRODUCTION 5

MARY RUTH WISE’s article presents an overview of characteristics and functions of applicative affixes in a number of Amazonian languages spoken in . She deals among others with the question of whether the involved similarities across these languages, which belong to six different families, point to genetic relationships or to the coexistence of these languages in a linguistic area. The editors thank Eithne Carlin for comments. The edition of this book has been carried out with the generous support of the Spinoza research program Lexicon and Syntax, coordinated by Pieter Muysken in The Netherlands.

Amsterdam, December 2001 The editors

6 SOUTH AMERICAN LANGUAGES

Language Affiliation Researcher

1. Chayahuita (Cahuapana) (Wise) 2. Kali’na (Carib) (Renault-Lescure) 3. various (Carib) (Meira) 4. Emberá (Chocó) (Herrera) 5. Cholón (Cholon) (Alexander-Bakkerus) 6. Aymara (Jaqi) (Filimonova, von Gleich) 7. Baniva (Maipure-Arawak) (Mosonyi) 8. Yanesha' (Maipure-Arawak) (Wise) 9. various (Mataguayo) (Viegas Barros) 10. Mosetén (Mosetén) (Sakel) 11. Cashinahua (Pano) (Camargo) 12. Yaminahua (Pano) (Wise) 13. Yagua (Peba-Yagua) (Wise) 14. Nomatsiguenga (Pre-Andine Maipure) (Wise) 15. Quechua (Quechua) (von Gleich) 16. Guaraní (Tupí-Guaraní) (Dietrich) 17. Guaraní criollo (Tupí-Guaraní) (Dietrich) 18. Chiquitano (unclassified) (Falkinger) 19. Kamsá (unclassified) (Fabre) 20. Kwaza (unclassified) (van der Voort) 21. Leko (unclassified) (van de Kerke) 22. Movima (unclassified) (Grinevald) 23. Amazonian Bolivian (various) (Crevels) 24. Upper Madeira (various) (Snethlage) 25. Arabela (Zaparo) (Wise)

INTRODUCTION 7

Map of South America indicating the locations of the languages from which data are presented in the current volume

WHY SPEAKERS SHIFT AND LANGUAGES DIE: AN ACCOUNT OF LANGUAGE DEATH IN AMAZONIAN BOLIVIA1

Mily Crevels University of Nijmegen

1. Introduction In recent years the issue of language endangerment and language death in South America has received quite some attention through linguistic publications (Adelaar 1991, 1998, forthcoming; Crevels & Adelaar 2000-2002, 2001; Grinevald 1998; Grinevald Craig 1997; Moore forthcoming), and fortunately also through non- linguistic channels, such as, for instance, James Geary’s article in Time (July 7, 1997), or the VPRO Dutch television documentary Verloren taal2 (May 2001). In this paper I will try to give some possible factors for the extremely alarming situation of the native languages of the lowland Department of Beni3 in north-eastern Amazonian Bolivia. Section 2 sketches the historical background against which the colonization of the region took place, Section 3 gives an overview of the current status of the region’s indigenous languages, Section 4 discusses some possible causes of the deplorable state of some of these languages, and in Section 5, finally, a tentative conclusion is drawn.

2. Historical background In the 16th Century, when the Spaniards first arrived in the region, the area was populated by some 400 groups or tribes with an estimated total of 350,000 individuals who spoke about 39 different languages, most of which belonged to the Arawakan family (Baptista Morales 1995: 71). Nowadays only three are spoken in the Department of Beni: seriously endangered Baure, the Mojo language4, and Machineri5.

1 This article is dedicated to Don Desiderio Espíndola, who has done his utmost to help me in my search for the last traces of the Canichana language, and to my dear consultants Don Ignacio Aulo († 2000), Don Lauro Chanato, Don Ascensio Cacharana, Don Manuel Guasase, Doña Juanita Volome, and Doña Concha Gualu († 2001), the last speakers of Itonama. Furthermore, I would like to thank , Laércio Bacelar, Peter Bakker, Simon van de Kerke, and Hein van der Voort for reading a prefinal version of this paper and commenting on it. 2 ‘Lost language’ 3 The Spaniards called this region Gran Mojos, but it was also known as Gran Paitití, El Dorado, Isla de la Canela, Tierra de Enín, or Candire. Nowadays it is also known as the Llanos de Moxos (the Moxos Plains). 4 The Mojo language consists of the two subgroups Trinitario and Ignaciano. Although the differences between the two subgroups are minimal, speakers of Trinitario and Ignaciano claim that they do not understand each other (Fr. Enrique Jordá, S.J., p.c., San Ignacio de Moxos, June 2000). In this paper both subgroups are treated as varieties of the same language. 5 Machineri is spoken by a small group that, due to the negative social pressure with which it had to deal in Brazil, moved in 1985 from the State of (Brazil) into the Department of Pando (Bolivia). 10 CREVELS

The first Spaniards who came to the area discovered that it was not the savage and inhospitable land of their imagination. They encountered many villages and farms, and even the remains of great hydraulic works, which provided a clear token of the technical and organizational skills of the indigenous people of the region. Thousands and thousands of artificial hills with a height up to 60 feet marked the landscape, along with hundreds of artificial rectangular ponds up to three feet deep, all part of a system of cultivation and irrigation. The built-up high ground was used for farming purposes and canals were dug to connect ponds and rivers that caught water in this flood-prone region (Fr. Enrique Jordá p.c. 2000).6 The colonization process of the Gran Mojos area was not an easy task for the Spaniards. Protected by its clouds of insects, its extreme climate and its inhabitants’ fierce reputation, it was one of the last regions in South America reached by Europeans. From 1536-1537 onwards the first expeditions took place from what is now Cuzco in Peru to the legendary region of Gran Mojos, also known at that time as Gran Paitití. Many adventurers had long dreamed of conquering this land, because it was rumored to house the fabulous golden city of El Dorado. Finally, in 1617, a group of explorers established that El Dorado did not, in fact, exist in the Gran Mojos area.

2.1. The Jesuit Missions7 In 1569 the Jesuits first arrived in , and in 1587 Governor Lorenzo Suárez de Figueroa called them to Santa Cruz de la Sierra8 mainly to evangelize the Chiriguano, a belligerent Tupian group from eastern Bolivia. All the intents to establish reductions9 among the Chiriguano failed, mainly because of the continuous war between the Chiriguano and the Spaniards. On the other hand, they were able to establish flourishing reductions among the inhabitants of the Mojos area, the Mojeños, and the Chiquitano. Thus, in 1595, Jerónimo de Andión was the first Jesuit to reach the area with one of the expeditions organized by Governor Lorenzo Suárez de Figueroa. Jesuits also joined the expeditions of 1602, 1617, and 1624, but it would take almost 74 years until a Jesuit community was founded in Trinidad, today the capital of the Department of Beni. In 1668 the Mojeños asked the authorities of Santa Cruz de la

6 William Denevan’s 1961 discovery of the massive prehispanic earthworks over broad areas in the eastern lowlands of Bolivia, led to a drastic change in perspective regarding cultural development in the Amazon basin. A traditional archaeological approach of the Amazon points out the environmental limitations to cultural development, predominance of simple societies (bands and tribes), and subsistence systems based on hunting, gathering, and fishing with some limited agriculture in the form of slash and burn (cf. Steward 1963). However, the type of prehispanic raised field agriculture as documented in Denevan (1966), demonstrates that intensive agriculture was indeed possible and that large, dense populations were supported in these areas. 7 See Block (1994), Baptista Morales (1995) and Menacho (1995) for a detailed account of the establishment of the Jesuit missions in the Mojos area and the subsequent expulsion of the Jesuits in 1767. 8 The foundation of Santa Cruz de la Sierra by Ñuflo de Chávez in 1561 was initially inspired by the idea of having a staging post for the conquistadors who came in search for the riches of the empire of Candire. 9 Congregated settlements in which large numbers of Indians of different ethnic groups were brought together. LANGUAGE DEATH IN BOLIVIA 11

Sierra for protection against the belligerent Cañacuras and subsequently an expedition of 80 soldiers under the command of Captain Juan de la Hoz y Otalora was sent into the Mojos area. The Jesuits José Bermudo and Juan de Soto joined this expedition and once the area had been reached, an encampment called Santísima Trinidad was established. From here on Pedro Marbán, Cipriano Barace, and José del Castillo founded El Pueblo de Nuestra Señora de Loreto in 1682, the first of Mojos and the cradle of the in Beni. In the following years the Jesuits founded 26 other settlements along the upper courses of the rivers Mamoré, Iténez (Guaporé), and Beni, all three tributaries of the Madera River (cf. Map 1). Eventually, many of these settlements were relocated or simply abandoned due to floods, epidemics, or attacks by the mamelucos10. In the second half of the 16th and in the 17th century continuous raids had taken place from the Santa Cruz area into the territories of the missions, and as a result hundreds of Mojos’ native inhabitants had been captured and taken away as slaves. The presence of the Jesuits diminished but in no way made an end to the slavery expeditions. In 1720, the Jesuits were able to prohibit by royal decree the entrance of Spaniards into the reductions. Nevertheless, the advance and attacks of inhabitants of Brazil, who were protected by the Portuguese authorities, proved to be an even major threat. In 1767, when the expulsion of the Jesuits from all the domains of King Charles III of Spain began, only 15 settlements were left. Table 1 gives an overview of these settlements.

Mission Main ethnic group Linguistic family Foundation Loreto Mojo Arawakan 1682 Trinidad Mojo Arawakan 1686 San Ignacio Mojo Arawakan 1689 San F. Javier Mojo Arawakan 1691 San F. Borja Chimane Unclassified 1693 San Pedro Canichana Unclassified 1697 Concepción Baure Arawakan 1708 San Lorenzo Movima Unclassified 1709 Exaltación Cayubaba Unclassified 1709 San Joaquín Baure Arawakan 1709 San Reyes Maropa (Reyesano) Tacana 1710 San Martín Baure Arawakan 1717 Santa Ana Movima Unclassified 1719 Magdalena Itonama Unclassified 1720 San Simón Baure Arawakan 1744 Table 1. Jesuit missions in Mojos in 1767

As Block (1994: 179) points out, the Jesuit mission system caused important as well as painful changes in the lives of the Indians. Possibly, one of the most painful

10 Brazilian mestizos. 12 CREVELS changes was the concentration of formerly dispersed villages into central settlements. Thus, the native people became not only fully exposed to epidemic disease but also to cohabitation with their former aboriginal rivals. In the long run, however, mission populations emerged with a significant resistance to disease.

Map 1. (Re-)locations of the Jesuit Missions in Mojos 1682-1767 (copy of Lehm Ardaya 1999: 29)

At the end of the 17th century, the linguistic policy of the Jesuits was initially aimed at using only the Spanish and the Mojo language for educational and religious purposes. Thus, an effort was made to replace all the different languages that were LANGUAGE DEATH IN BOLIVIA 13 spoken in the region at that time by a single lengua general11, the Mojo language. Concentrating members of different ethnic groups in the same settlement supported this goal.12 Eventually, however, as the Jesuit settlements expanded more to the east, at least six other lenguas generales were recognized: Canichana, Movima, Cayubaba, Itonama, Baure, and Sapive. Once the missions had a solid base and were functioning smoothly, the Jesuits –realizing that their linguistic policy had only succeeded partly– started focusing on education in the native languages. As a result a number of and religious tracts were produced in the Mojo language, Baure, Movima, Cayubaba, Itonama, Canichana and a few other languages. The principle works that are still accessible today are Pedro Marban’s Arte de la lengua Moxa con su vocabulario y cathecismo (1701) and Antonio Magio’s Arte de la lengua de los indios Baures de la provincia de los Moxos (1749). During the Jesuit era the economic focus in the missions of Mojos was still on subsistence production. Apart from traditional food crops such as manioc, yams and maize, new crops were introduced, like rice, cotton, cacao and sugarcane. The biggest change, however, was brought about in 1682 by the introduction of cattle raising, which to this day still is one of the major sources of income in the Department of Beni. After their expulsion in 1767, the missions were turned over to civil administrators and curas13, most of whom were more or less recruited in the streets of Santa Cruz and other districts of the Audiencia of Charcas14. Generally these curas had close ties with secular Spanish society and were completely ignorant of the ways in which the Jesuits used to manage the missions. This led to a turning point in the history of mission culture in Mojos. Block (1994: 125) describes the situation in the missions at the time of the expulsion as follows:

“When the Jesuits departed the savanna in 1767, they left behind a resilient mission culture based on firm foundations. The population was resistant to European diseases and cohesively organized into small urban communities. A native political elite, enjoying popular support, had gained valuable experience during the Jesuit years. By 1767 members of the communities had become fluent in Spanish and showed remarkable skill in European arts and industries. The missions also retained a considerable material base. Although deprived of their access to Jesuit coffers, the Indians inherited the sumptuously appointed churches, well-equipped workrooms, and, most important, the riverfront lands and cattle herds amassed during the Jesuit century. But as the Jesuits left their posts, they were replaced by men of a different stripe.”

11 ‘’ 12 On the basis of the data provided by Eguiluz (1884), Métraux (1942: 54-55) points out that, even though Mojo was becoming the official and compulsory language in all the missions except San Francisco Borja, Mojo Indians actually were to be found only in Loreto, San F. Javier, and perhaps in Trinidad. 13 Spanish curates. 14 Administrative unit for the eastern territories founded in 1559 by King Felipe II of Spain. 14 CREVELS

Whereas the different missions functioned quite autonomously during the Jesuit era and, as pointed out above, economic activities were mainly aimed at subsistence production, this situation changed radically after the Jesuits’ departure. According to Block (1994: 177), after 1767 commercially viable activities became the order of the day. Thus, cacao cultivation pushed manioc gardens far from central settlements, cotton textiles replaced feather weaving, and no more additions were made to the sumptuary inventories of the mission churches. This economic reorientation converted the missions into revenue producers of the Spanish Crown, which inevitably led to the decline of the mission riches. During his travels in Bolivia the French savant Alcides d’Orbigny visited the Canichana Mission of San Pedro in 1832. He described the bloom and decline of the mission, which used to be the most prosperous of all missions, as follows (1958: 774- 75):

“[…] los jesuitas establecieron la Misión de San Pedro, cuya posición central la convirtió muy pronto en capital de la provincia. Concentraron en ella todas sus riquezas, todas sus grandezas y por sus monumentos, por el número de sus estatuas de santos, por las joyas que adornaban a sus Vírgenes y a sus Niños de Jesús, por las planchas de plata que decoraban sus altares y, más que nada, por las hermosas tallas de madera de su iglesia, San Pedro no tardó en rivalizar no sólo con las catedrales de Europa, sino también con las más ricas iglesias del Perú. […]. San Pedro fué dilapidada bajo el régimen de los curas, primero, y luego bajo el de los gobernadores. Lo mismo ocurrió bajo los administradores. […] la redujeron a la mayor miseria, y no hay duda que hoy es la más pobre de todas.”15

The new secular and economic orientation made the Indians easy victims to all sorts of abusive practices, either by the curas or the administrators. D’Orbigny’s travels in Bolivia also brought him in 1832 to the Mission of El Carmen16, where he was able to observe the deplorable situation of the native Baure (d’Orbigny 1958: 751):

“[…]. Basta para prueba la paciencia con que soportaron durante largos años la infame conducta de sus administradores y de su cura, quienes, habiéndose

15 ‘[…] the Jesuits established the Mission of San Pedro, which due to its central position soon became the capital of the province. They concentrated in the mission all their riches, all their great works and because of its monuments, because of its number of statues of saints, because of the jewels that adorned its Virgins and its Christ-childs, because of the silver plates that decorated its altars and, more than anything, because of the beautiful wooden sculptures of its church, it was not long before San Pedro not only competed with the cathedrals of , but also with the richest churches of Peru. […]. San Pedro was first wasted under the rule of the curas, and then under that of the governors. The same happened under the administrators. […] they reduced it to the greatest misery, and there is no doubt that today it is the poorest of all.’ [Translation MC]. 16 The Mission of El Carmen was founded in the Baure territory with Baure and Guarayo Indians after the expulsion of the Jesuits. LANGUAGE DEATH IN BOLIVIA 15

repartido la Misión como un harén común, se hacían traer sucesivamente a todas las indiecitas en cuanto habían llegado a la edad de ocho o diez años, bajo la pena de cincuenta azotes. No reproduciré ni el número de víctimas de esos monstruos ni otros horrendos detalles que supe de labios mismos de los intérpretes; hacen estremecer a la Humanidad. […].”17

Sharp conflicts arose between the Indians, the curas and the civil administrators, and in 1810 the populations of the missions of Trinidad and Loreto revolted against the Spanish tyranny. When the Bolivian Republic was proclaimed in 1825, mission culture still existed in Mojos, but in the next fifty years political and economic changes would end it. In 1842 President José Ballivián formed the new Department of Beni by combining Mojos with the area of Caupolicán in the Andean rain forest.

2.2. The rubber boom Halfway the 19th century the new Department of Beni slowly started to open up to commerce. This implied the settlement of some mestizos and criollos18 in the former reductions, the progressive creation of farming and cattle raising establishments in their immediate surroundings and a growing demand for indigenous workforce. Since the Indians traditionally worked on their own chacos19 and since indigenous workforce therefore was scarce, new ways of recruiting workforce were thought of. One of these ways was called el concierto20. It implied that the proprietors of the big farming and cattle raising establishments provided the Indians with an advance in goods and/or money for which the Indians committed themselves to work exclusively for the proprietor who had given the advance (cf. Vaca Diez 1989: 27). Around 1870, the exploitation of rubber was still limited to some areas on the upper Beni, the Madera and the lower Mamoré. However, from 1880 onwards –when the confluence of the Beni and the Mamoré rivers was discovered by the American physician and explorer Edwin Heath and the fluvial transportation route Reyes- Madera-Amazon came into use– rubber activities were expanded more to the north and started to play an important economic role. The rubber boom resulted in a massive invasion of the Beni territory by criollos and mestizos. Whereas in 1832 d’Orbigny (1845: 299) recorded 57 white residents in the Mojos settlements, Heath (1882), who explored the lower Beni River, observed 200 non-indigenous rubber workers in the region. Only a few months later this number had increased to 1,000- 2,000. With the expansion of the rubber activities the enganche21 recruiting system of

17 ‘[…]. Proof enough is the patience with which they bore during long years the infamous behaviour of their administrators and of their priest, who, having divided among themselves the mission as a joint harem, had brought to them successively all the Indian girls as soon as these reached the age of eight or ten years, on penalty of fifty lashes. I will neither reproduce the number of victims of those monsters nor other horrendous details I came to know from the very lips of the interpreters; they make Humanity shudder.’ [Translation MC]. 18 Persons of Spanish extraction 19 Piece of arable land where consumption goods such as manioc, yams and maize are cultivated. 20 ‘the agreement’ 21 ‘bait’, ‘hook’, ‘recruitment’ 16 CREVELS indigenous workers became common practice in the Department of Beni. Commercial companies such as Casa Suárez Hnos provided advances in money, provisions and goods to the so-called enganchadores22 who, in turn, went to the indigenous villages to deliver the advances and subsequently move the Indians to the rubber plantations (cf. Lehm Ardaya 1999: 51). Many went by their own free will, but others were forced. Moreover, uncontacted groups of Indians were taken prisoner and forced to work in the extraction of rubber sap. Initially young and adult males from all ethnic groups in the Beni were recruited. This was effectuated to such an extent that some villages that once had a numerous population were only left with old men, women and children. Thus, in 1874, the village of El Carmen had 750 female and 15 male inhabitants (cf. Moreno 1973: 74). With the removal of large numbers of adult males from the indigenous villages, not only their proper base, but also their demographic and social structure were deeply affected. The rubber plantations with their life-threatening malarial fevers, the deadly rapids that made transportation of the shipments of rubber moulds extremely complicated, the very harsh regime that only worked in favor of the masters, all this in combination with the workers’ debts that never got paid off and even passed on from father to son, turned out to be devastating to the indigenous workforce that had landed in this maelstrom of exploitation, violence, terror, adventure, fortune, mystery and alcohol (cf. Lijerón Casanovas 1998: 84). In the years of the rubber boom (1870- 1910) the population of the Department of Beni was decimated. As a member of the Arvid Hernmarckschen Expedition to Bolivia (1908-1909) the Swedish zoologist and anthropologist Erland Nordenskiöld visited the Beni. He described the dissolution of all ethnic and cultural ties of the indigenous groups under the rubber barracks regime as follows (1922: 123):

“Hier kann man nicht mehr von Indianern verschiedener Stämme sprechen, sondern nur noch von Kautschukarbeitern. , Baure, Itonama, Mestize, alle sind unter ein Dach zusammengebracht. Sie wohnen in Baracken. Jeden Familienband ist gelöst. Sie haben keinen selbständigen Ackerbau, keinen selbständigen Haushalt. Nur wenige sind verheiratet. Die meiste haben ungebundene Frauen, die von einem Arm in den anderen wandern. Dadurch werden die Geschlechtskrankheiten verbreitet und die Frauen mit Kinder sehr selten. Die sterblichkeit bei zarten Kindern ist sehr groß. Es wird immer schwerer neue Arbeiter anzuwerben. Die einzige Möglichkeit ist solche zu kaufen, zu kaufen die Schuldsklaven von anderen; und ein arbeitstüchtiger Mensch ist hier mindestens tausend Bolivianos wert. Ein Knabe wird auf vierhundert Bolivianos geschätzt. Ein Mädchen wird meist nach dem Aussehen bezahlt. […] Jede Spur indianischer Kunstfertigkeit und Kultur ist hier verschwunden. Das Leben bietet diesen Menschen nichts anderes als Arbeit für den Herrn an

22 ‘recruiters’ (= persons who had to hook rubber workers) LANGUAGE DEATH IN BOLIVIA 17

Wochentagen, Saufen an Feiertagen und Vielsaufen an den großen christlichen Festen. In den Dörfern feiert man diese mit Messen und Branntwein. Hier gibt es keine Messen, hier gibt es nur einen Gott, und dieser Gott ist der Brannt- wein.”23

Not only the Indians from the former Jesuit settlements fell victim to the rubber activities. The search for new zones with resinous trees during the rubber boom also had its genocidal and ethnocidal effects on the Indians who had not been catechized in the Jesuit missions, such as the Araona, Chácobo, Pacahuara, and Ese Ejja, who at their turn also had to undergo the effects of epidemic disease, alcohol, debt peonage, forced labor, etc.

Year Canichana Movima Cayubaba Baure Itonama 1687 4,000-6,000 1693 4,000-5,000 10,000-12,000 1704 6,000 1712 40,000 1749 2,930 3,000 1767 4,400 4,000 1780 1,860 1797 2,544 1831 1,939 1,238 2,073 5,178 4,815 1909 100 300 1962 2,000 1967 30-50 3,000-4,000 300 1974 500 1994 583 6,528 794 631 5,090 Table 2. Demographic data of some ethnic groups in the Department of Beni24

Table 2 gives an overview of some demographic data that have been roughly established or estimated by Jesuit priests, explorers, or local censuses between the

23 ‘Here one can no longer speak of Indians of different tribes, but only of rubber workers. Chiquitano, Baure, Itonama, mestizos, all are brought together under the same roof. They live in barracks. All family ties have been cut. They do not have independent fields, they do not run an independent home. Only a few are married. Most have unattached women, who wander from one arm into another. This causes venereal diseases to spread and women with children to be very scarce. Mortality among newborn infants is very high. It is steadily becoming more and more difficult to recruit new workers. The only possibility is to buy them, to buy the debt slaves of others; and here a hard-working person costs at least one thousand bolivianos. An adolescent boy is estimated at four hundred bolivianos. A girl is usually paid for according to her looks. […] Every trace of Indian craftsmanship and culture has disappeared here. Life offers these people nothing but work for the master on weekdays, drinking on rest days and getting extremely drunk on the big Christian holidays. In the villages these are celebrated with masses and brandy. Here there are no masses, here there is only one God, and this God is the brandy.’ [Translation MC]. 24 Data adapted from Métraux (1942), Plaza Martínez & Carvajal Carvajal (1985), and Crevels & Adelaar (2000-2002). 18 CREVELS late 17th century and the late 20th century. Although the data may not be very reliable, being sometimes based on estimations, Table 2 –with the exception of Movima– does show a sharp decline after 1831 as far as the numbers for the different ethnic groups are concerned. The relatively high demographic numbers for 1994 will be discussed in Section 3. All in all the numbers underline the devastating effects of –among other things– the rubber boom on the inhabitants of the Department of Beni. Around 1912, when the rubber boom was coming to an end, many people who worked on the rubber plantations in the north came back to the center and the south of the department. This implied a new impulse for the farming and cattle raising establishments and a reinforcement of the coercion practices to recruit Indian workforce. Once again the Indian population was subjected to debt peonage. This time la matrícula25 –a term that refers symbolically to the registration of Indians for the payment of territorial contributions– was applied via the state authorities to provide the proprietors with labor force (cf. Lehm Ardaya 1999: 70).

2.3. The Chaco War and its consequences In 1932, the Chaco War broke out between Bolivia and Paraguay. The origin of the conflict went back to the outcome of the War of the Pacific (1879-84), in which defeated Bolivia and annexed its entire coastal region. Bolivia attempted to break out of its landlocked situation by seeking access to the Paraguay River to ship oil to the sea. On that route lay part of the Gran Chaco, the Chaco Boreal, a rough and deserted region of about 259,000 square kilometres north of the Pilcomayo river and west of the Paraguay river. The Chaco Boreal was thought to have large oil preserves. Pushed by English oil companies and the United States, the two countries started a war to decide for once and for all on the political status of the region. More than 100,000 lives were lost, and the war ended in 1935 only when both sides were exhausted. After three years of mediated negotiation following the end of hostilities, Bolivia and Paraguay signed a treaty in 1938 at the Chaco Peace Conference in , which included Argentina, Brazil, Chile, Peru, , and the United States. Three quarters of the disputed Chaco Boreal went to Paraguay; at the same time Bolivia was granted a corridor to the Paraguay River, the privilege of using Puerto Casado, and the right to construct a Bolivian port. Argentina was given the main credit for the settlement, and Argentinean investors profited greatly from Paraguay’s territorial gain. Years later, oil companies explored the Chaco region but they were unable to find significant oil fields. After the Chaco War, Bolivia’s economy was so seriously disrupted that the impoverished masses demanded reforms. The defeat marked a turning point: the enormous loss of life and territory discredited the traditional ruling classes and service in the army had started to awaken some political awareness among the indigenous groups. The agricultural sector lacked capital, and by 1950 food imports had increased to 19% of total imports. Land was unequally distributed –92% of the

25 ‘the registration’ LANGUAGE DEATH IN BOLIVIA 19 arable land was in hands of owners of estates of 1,000 hectares or more. The expansion of mining from the end of the 19th century had attracted foreign investors and led to the establishment of three large foreign mining corporations that became dominant in Bolivia’s economic and political life. At that point these mining corporations controlled the country, together with the big landowners and the military. On April 8, 1952 a popular revolt took place in and elsewhere. The Movimiento Nacionalista Revolucionario (MNR), with the support of armed workers, civilians, and campesinos26 overthrew the ruling military junta and seized control of the government. Subsequently the MNR introduced universal adult suffrage, nationalized all the tin mines of the three big mining corporations, implemented an agrarian reform, by which the large estates were divided among former tenants and peasants, and promoted rural education, by implementing the Reforma Educativa27. By the time that the MNR was deposed by the army in 1964, it had introduced far- reaching economic and social changes. It also committed many serious violations of human rights. Over the next 28 years (1964-2002) Bolivia has 20 presidents, of whom 13 were generals; only two completed a full term in office. The Revolución Nacional28 of 1952 disregarded the indigenous lowland –and highland– groups by lumping them together under the label of campesinos. According to the 1953 Reforma Agraria29, the land that was occupied by Indians was to be considered fallow, and, therefore, its ownership could be granted to all sorts of entrepreneurs. The idea was to turn the rural Indian into a campesino who could make a contribution to economic and consumptive needs. The new agricultural law only permitted him to work on a single plot of land. This disintegrated the unitary and integral conception that the indigenous inhabitant had of his territory, and in the long run it would turn out to have disastrous consequences for the indigenous cultures and day-to-day life in the communities. From the perspective of the indigenous groups in the Department of Beni, not only the agrarian reform but also the educational reform had devastating effects on the rural communities, in which – up till that point– important features of the authentic indigenous cultures and traditions had been conserved to a certain degree. Even though the rural schools that were established during this period were the first ones to which most of the indigenous communities had access, the extremely hispanicizing education had genocidal effects on all the indigenous groups implied, insofar that it provoked the progressive loss of indigenous languages and other forms of cultural heritage. The rural teacher and his guava whip, which was used to punish the children that spoke in their native language, soon became decisive factors of a progressive acculturation process (cf. Lijerón Casanovas 1998: 119).

26 ‘peasants’ 27 ‘Educational Reform’ 28 ‘National Revolution’ 29 ‘Agrarian Reform’ 20 CREVELS

3. Current status of the languages To this day some 20 indigenous languages are still spoken in Amazonian Bolivia. They represent five language families and a total of seven language isolates or unclassified languages. Of these languages 15 are spoken in the Department of Beni: Reyesano, Cavineña, Ese Ejja, Chácobo, Pacahuara, Moré, Baure, Trinitario/Ignacia- no (Mojo), Sirionó, Yuracaré, Chimane, Canichana, Movima, Cayubaba, and Itonama.

Map 2. Location of the indigenous languages of Amazonian Bolivia.

Table 3 gives an overview of the indigenous languages of Amazonian Bolivia, their affiliation, speakers’ and demographic numbers, and state of endangerment. The general location of the languages is indicated on Map 2 as numbered in Table 3. LANGUAGE DEATH IN BOLIVIA 21

Language Language Number of Number of Degree of family speakers ethnic group endangerment 1 Reyesano Tacanan <10 4,118 Moribund 2 Tacana Tacanan 1,821 5,058 Endangered 3 Araona Tacanan 81 90 Seriously endangered 4 Cavineña Tacanan 1,180 1,736 Potentially endangered 5 Ese Ejja Tacanan 502 584 Endangered 6 Yaminahua Panoan 137 171 Seriously endangered 7 Chácobo Panoan 550 767 Endangered 8 Pacahuara Panoan 18 19 Moribund 9 Moré Chapacuran 76 200 Seriously endangered 10 Machineri Arawakan 140 155 Seriously endangered 11 Baure Arawakan ± 40 631 Seriously endangered 12 Trinitario Arawakan <10,00030 20,80531 Endangered 13 Ignaciano Arawakan Endangered 14 Jorá Tupian 0-532 Possibly extinct 15 Sirionó Tupian 399 419 Endangered 16 Yuki Tupian 125 156 Seriously endangered 17 Yuracaré Unclassified 2,675 3,333 Endangered 18 Leko Unclassified 20 80 Moribund 19 Mosetén33 Unclassified 585 1,200 Endangered 20 Chimane Unclassified 5,316 5,907 Potentially endangered 21 Canichana Unclassified 334 583 Possibly extinct 22 Movima Unclassified 1,452 6,528 Seriously endangered 23 Cayubaba Unclassified <10 794 Moribund 24 Itonama Unclassified <10 5,090 Moribund Table 3. Indigenous languages of Amazonian Bolivia35

When establishing the estimated number of speakers of each language, the biggest problem usually is the continuous mix-up of the number of the ethnic group with the actual number of speakers. Yet another problem involves establishing the proper number for the ethnic groups. A striking example is provided by Itonama, for which the 1994 Censo Indígena Rural de las Tierras Bajas (CIRTB)36 gives a total of 5,090 for the ethnic group. This high number is due to the fact that anyone who is born in Magdalena

30 Number applies to both subgroups of the Mojo language. 31 Number applies to the whole Mojo group. 32 Although Grimes (1996) reports that Jorá has become extinct since 1963, data collected by the German anthropologist Jürgen Riester in 1974 mention the existence of 5 speakers. In the literature Jorá is often stated to be a subgroup of Sirionó. 33 Although Mosetén and Chimane are often mentioned as separate languages, they form, together with Mosetén de Santa Ana, the small linguistic family Mosetenan (cf. Sakel this volume). 34 These speakers only remember a few words and one or two phrases. 35 Data adapted from Crevels & Adelaar (2000-2002). 36 ‘Rural Indigenous Census of the Lowlands’ 22 CREVELS

–or anywhere in the Province of Iténez for that matter– is considered to be, or considers himself to be Itonama. Since the majority of the population consists of whites, mestizos and criollos it is practically impossible to decide on the exact number of ethnic Itonama. The same problem occurs with the other ethnic groups mentioned in Table 2. The degree of endangerment depends on many factors, such as the percentage of speakers, the mean age of the speakers, whether the language is still spoken by children, etc. Thus, Yuracaré, with 2,675 speakers, is classified in Table 3 as ‘endangered’, while Cavineña, with 1,180 speakers is classified as ‘potentially endangered’. The difference lies in the fact that in the Yuracaré communities children no longer speak the language, while they still do in the Cavineña communities. The viability of Ese Ejja is quite good, but the language is classified as ‘endangered’ because the ethnic group is relatively small. On the other hand, Movima, with 1,452 speakers, has been classified as ‘seriously endangered’ due to the fact that all speakers are over 50 years old. Likewise, Araona, a language that is spoken by practically the whole community, is classified as ‘seriously endangered’, because the ethnic group is very small.37 A closer look at the languages that are spoken in the Department of Beni shows the possibility of making a subdivision in three sociolinguistic scenarios: a) predominant Spanish monolingualism, b) co-occurrence of bilingualism and Spanish mono-lingualism, and c) predominant native monolingualism.

a) PREDOMINANT SPANISH MONOLINGUALISM Itonama, Cayubaba, Baure, Reyesano b) CO-OCCURRENCE OF BILINGUALISM AND SPANISH MONOLINGUALISM More Spanish monolingualism than bilingualism: Movima, Mojo (Ignaciano & Trinitario), Moré More bilingualism than Spanish monolingualism: Yuracaré, Cavineña, Chácobo, Ese Ejja, Sirionó, Pacahuara c) PREDOMINANT NATIVE MONOLINGUALISM Chimane

Table 4. Current sociolinguistic situation in the Department of Beni38

Recent research39 has led to the belief that the Canichana language is no longer spoken and that within the next few years the same might apply to Itonama, Cayubaba, Reyesano, and even Baure. This would mean that three language isolates,

37 According to the demographic data, there exists a small tendency towards population growth: 39 (1963), 43 (1969), 50 (1975), 90 (1994), and finally 94 (1996). 38 Data adapted from the 1994 CIRTB. 39 In the period 1999-2001 I visited the Canichana of San Pedro on several occasions. The community would very much like to revive its language, but unfortunately all my attempts to track down speakers of Canichana have proved in vain. It now looks as though the language might be extinct. LANGUAGE DEATH IN BOLIVIA 23 one of the in total four , and one of the three Arawakan languages that are still spoken in Bolivia would disappear. The scenario in which there is a combination of bilingualism and Spanish mono- lingualism shows that Mojo, Movima, and Moré are on the way to be replaced complete- ly by Spanish. Even though Mojo still has a considerable number of speakers, an alarm- ingly rapid loss of language is reported among the youth. Yuracaré, Cavineña, Chácobo, Ese Ejja, Sirionó, and Pacahuara show a higher percentage of bilingualism than Spanish monolingualism, but even in the case of these languages there is a growing percentage of children who no longer acquire the language. Finally, Chimane, one of the dialects of the Mosetenan family, is by far the most viable language spoken ìn the Department of Beni. Generally the Chimane, especially the women and elders, speak very little Spanish. Children do not acquire Spanish until they are 15, unless they attend a school in one of the so-called colla40 communities. For many years economic development on the one hand, and preservation of the indigenous ethnic, cultural and social identity on the other hand, were considered to be two conflicting issues: working on the first would inevitably mean sacrificing the second. Moreover, it was often argued that the very culture of indigenous peoples stands in the way of their development. Consequently, programs to help indigenous peoples aimed at substituting their language, culture, and productive and agricultural practices. However, in the early 1990s the relationship between the indigenous groups and the Bolivian government changed. Whereas up to this point the relationship was based on exclusion, and even extermination, the State started to take the demands of the native groups into account. This change of attitude towards the Indians was not only triggered by the impulse of the newly arisen indigenous movement, but also by the growing awareness that the needs of the native population of Bolivia should be attended as well. Thus, in 1993, a state institution for the management of indigenous affairs was created, the Subsecretaría de Asuntos Étnicos (SAE)41. In 1997, this institution was converted into the Vice Ministerio de Asuntos Indígenas y Pueblos Originarios (VAIPO)42. Currently the Programa Amazónico de Educación43 –a program that is aiming at the implementation of intercultural bilingual education in the Bolivian lowlands– is being financed with the help of the Danish government. The 1990 and 1996 Indian marches held in support of territorial claims have led to the recognition of the fact that the struggle for territory lies at the base of all of the indigenous demands and should, therefore, be treated with priority. A consequence of this recognition has been the ratification in 1996 of the so-called Ley INRA44 (Instituto Nacional de Reforma Agraria45). This new agrarian reform law permits community land ownership and legalizes the creation of TCO’s (Tierra

40 Word used for referring to highland Indians. 41 ‘Bureau of the Assistant Secretary of Native Affairs’ 42 ‘Vice-Ministry of Indigenous Affairs and Native Peoples’ 43 ‘Amazonian Program of Education’ 44 ‘INRA Law’ 45 ‘National Institute for Agrarian Reform’ 24 CREVELS

Comunitaria de Origin46). Furthermore it allots land that is not fulfilling its economic and social function. As Lema Garrett (1998: 27) points out, by 1997, Amazonian Bolivia had not only captured the attention of the State but also that of the civilian society. Within the context of the decentralization of the administration on the one hand, and the growing interest of NGO’s on the other hand, the attention for regional and local indigenous problems has grown considerably. Nevertheless, it remains to be seen what will be the real impact of these development in the long run.

4. Discussion It is not possible to attribute the sometimes more than alarming state of some of the indigenous languages spoken in the Department of Beni –or in the whole of Amazonian Bolivia for that matter– to a single factor, or even to a few factors. In the previous sections I have tried to sketch some possible causes for language loss and language death in this region. It all started in the 16th century when the first Spanish expeditions with explorers and conquistadors entered Mojos in search of El Dorado’s . They were immediately followed by the slavery expeditions from the Santa Cruz area that came in search of human workforce. Although these expeditions usually did not stay on for a prolonged period in the Mojos area, they lasted long enough to contaminate the native populations with epidemic diseases and unbalance them by taking away young and strong men. By the end of the 17th century, the Jesuits, in search of souls, formed the third group to enter Mojos. The Jesuit missionaries were strongly opposed against slavery, an institution long condemned by Rome. Even though the missions in Mojos bloomed and brought important economic advantages to the area, it should be stressed that since the early days of the conquest of America Christian mission activity has been a salient component of the European aim to transform and assimilate the native inhabitants of America. The missions were run in a remarkably democratic way. Land, tools and workshops were the property of the community and work was obligatory for all able- bodied members. Nevertheless, the Jesuits’ prime concern was to “save the Indians' souls”, and, therefore, the indigenous customs and beliefs were discouraged. Christian values and moral were imposed on the people in such an effective way that today very little is known about the pre-Columbian indigenous cultures of this region. Furthermore, Block (1994: 58-59) points out that there is evidence that the demands of mission culture strained Mojos’ natural resources. Settling thousands of people into single communities led to changed cultivation patterns, which gradually diminished the fertility of the soil. Historical records show that the activities developed during the Jesuit era reduced the available food resources for gathering. The often more than inhumane ways, in which the local clergy and administrators who ruled the missions after the expulsion of the Jesuits in 1767 treated the indigenous inhabitants of the missions, has been sufficiently illustrated in,

46 ‘Indigenous Community Territory’ LANGUAGE DEATH IN BOLIVIA 25 among others, d’Orbigny’s descriptions of his travels in Bolivia. Mission culture came to an end when the rubber boom of the late 19th century brought white, criollo and mestizo settlers in sufficient numbers to overwhelm the indigenous culture –in so far as it still existed. Although the natives remained, they became marginalized to the Westernized society that had developed in the core regions, in which fluency in Spanish, affluence, and light skin were highly valued. As discussed in Section 2.2, the rubber boom formed one of the most disruptive factors in the progressive destruction of indigenous cultures and languages. After visiting the rubber barracks on the Iténez River, Nordenskiöld (1922: 124) could not but conclude the following:

“Eines aber steht fest, eine kunstfertige, freie indianische Bevölkerung ist ausgerottet oder in ein elendes branntweindurstiges Proletariat verwandelt worden. Um den Bedarf der Industrie in Europa und Nordamerika zu befriedigen, werden diese Gegenden vernichtet. Mit welchem Recht? Der Zivilisation wegen müssen die Schwachen untergehen. Nicht so sehr die Grausamkeit einzelner Personen gegen die Indianer regt mich auf, sondern die Versklavung einer ganzen Rasse.”47

When the Chaco War broke out, most had very little if any interest in the remote Chaco region. Nevertheless, the authorities drafted all young men into the Bolivian army to fight against the Paraguayans in this terrible war, in which Bolivia lost 50,000 lives. The Department of Beni sent many young natives who were noted for their courage and endurance under the most difficult circumstances. After the war those that had survived returned home bitterly disappointed because of all the sacrifices that had been made more or less in vain for a cause that was physically and conceptionally so remote from their own world. The educational reform that was carried through after the 1952 revolution may have been the decisive factor in the endangerment –and sometimes even extinction– of the languages of the Beni. The majority of the teachers of the newly established rural schools came from outside the region, did not know the local indigenous languages and were trained to severely punish the use of these languages within school settings. It did not take most of the pupils very long to shift to Spanish, thus causing their own native mother tongue to lose even more prestige. This process started almost half a century ago and by now the few speakers that are left in the case of some of the languages have to overcome a deep-rooted feeling of shame in order to be able to speak their native tongue.48

47 ‘But there is no doubt about one thing, a skillful, free Indian population has been extirpated, or changed into a miserable alcoholic proletariat. In order to satisfy the need of the industry in Europe and North America these areas are being destroyed. With which right? For the sake of civilization, the weak have to go down. It is not so much the cruelty of individuals against Indians that makes me angry, but the enslavement of an entire race.’ [Translation MC]. 48 Enrique Jordá (p.c. 2000) reports that in San Ignacio de Moxos the elderly women, who used to greet him and speak to him in Ignaciano, lately have stopped doing so. When asked why, the answer was that 26 CREVELS

In 1954, the Bolivian government invited the Summer Institute of Linguistics (SIL) to Bolivia with the objective to establish the integration of the indigenous peoples in national society. SIL stayed on until 1980 and worked on 18 indigenous, mostly Amazonian, languages. The institute played a pioneering role as far as bilingual education in the Amazonian lowlands is concerned, but once again missionaries imposed their Western norms and moral on the indigenous inhabitants of the Department of Beni by using bilingual education as a tool to launch their cultural and religious philosophy (cf. Castro Mantilla 1997: 80). Other missionary groups that are closely related to the SIL and that are still active in Bolivia today are the New Tribes Mission and the Swiss Mission. In 1951, the anthropologist François-Xavier Beghin stayed for two months with the last group of 14 surviving Jorá (Tupian) at the Bolsón de Oro Lagoon, to the east of Magdalena and the San Joaquín River. During his stay he found traces of a massacre in an abandoned Jorá village49. According to the surviving Jorá, the major part of the village’s inhabitants had been exterminated a year and a half before during one of the “punitive” expeditions undertaken by whites, criollos and mestizos in order to “reduce” and “domesticate” one of the last tribes that had the audacity to defend its hunting grounds and crops (cf. Beghin 1976: 130). In 1955 only five Jorá were left, and today the group, and therefore the language, is most probably extinct. On November 9, 2001, 10 landless campesinos50 were massacred and more than 20 injured, about 40 km from Yacuiba, the capital of the Province of Gran Chaco, Department of Tarija, in the south of Bolivia near the Argentinean border. Today in Bolivia 4,5% of the landowners still possess 70% of the total disposable land. Confronted with the unjust land distribution and having lost hope that the Ley INRA will allot them land, campesino families decided in May 2000 to organize Núcleos de Campesinos Sin Tierra51 and started to occupy land of big estates that did not fulfil any economic or social function52 in the Province of Gran Chaco. On June 23, 2000, 180 families occupied the first big estate, the Pananti estate of 3,000 hectares near Yacuiba. Today more than 18 Núcleos de Campesinos Sin Tierra exist. The 180 families of the Núcleo de Campesinos Sin Tierra de Pananti were attacked by landowners and paramilitary groups with the help of police and army. These attacks have been systematically repeated since the families first occupied the Pananti estate in June 2000, and the persecution of the landless campesinos has not stopped after the massacre at the Pananti estate. The Bolivian authorities have not taken any measures to curb the violence against the campesinos. The 1996 INRA Law was meant to put a stop to the illegitimate assignment and unequal distribution of land, but up till now the land distribution problems still have not been solved. The Instituto Nacional de

they had gotten used to Spanish, that their grandchildren always spoke to them in Spanish and did not understand Ignaciano. 49 Whenever one of the Jorá died, the other members of the group abandoned the village. 50 See Section 2.3, 3rd paragraph. 51 ‘Centers of Landless Peasants’ 52 See Section 3, penultimate paragraph. LANGUAGE DEATH IN BOLIVIA 27

Reforma Agraria has not succeeded in reverting to the State the abandoned, idle or illegal estates, which preferably should be allotted to indigenous peoples and communities without land. This is partly due to the ambiguity of the INRA Law, but more than anything to the fact that the State keeps giving in to the pressure of the influential class of large landowners that is opposed to the planned changes in the structure of agrarian landownership.53 Exactly 50 years have gone by since Beghin’s visit to the Jorá in 1951 and the massacre at the Pananti estate –which in view of the tension built up around the land allotment question could just as easily have taken place in the Department of Beni–. Unfortunately very little seems to have changed in this time span. As long as the indigenous inhabitants of the Department of Beni get exterminated over conflicts about land or other economic matters, as long as they keep on living in a form of debt servitude in which they get paid less than what the landowners charge them for food, clothing and other goods, as long as they are completely dependent on the landowners for cash money54, and as long as they are treated as some kind of third- rate population group, there will be very little chance that the dissolution of the ethnic groups and the disappearance of their languages will come to an end.

5. Conclusion The pending loss of languages, such as Itonama, Cayubaba, and Reyesano55, will have serious consequences for the transfer of knowledge to a younger generation concerning the surroundings and cultural traditions of the ethnic groups in question. After all, language reflects the unique world-view and culture complex of its speech community. Wurm (1999: 33-34) points out that with the disappearance of each language, an irreplaceable unit in our knowledge and understanding of human thought and world- view gets lost for ever. This results in a reduction of the human knowledge base as it may be expressed through language. Every language is a symbol of the ethnic identity of its speakers, and its documentation tends to keep it alive. However, the mere documentation of (seriously) endangered or moribund languages is not enough. As stressed by Grinevald Craig (1997: 270), linguists should combine salvage linguistics56 and archiving efforts with efforts of revitalizing or maintaining endangered languages. In the case of the indigenous languages of the Beni, this is not an easy task, since generally numbers of speakers are very low and, moreover, speakers are not convinced of the need to maintain the indigenous language. However, there is hope that the growing ethnic self-consciousness and pride of many groups in the Department of Beni in the past decade may lead to a resurgent interest of the speakers in their own native languages.

53 Full report at www.movimientos.org/cloc/show_text.php3?key=836/ 54 Especially for health issues cash is needed to pay a doctor or to buy medicine. 55 See Section 3. 56 Label that is sometimes used for the linguistic documentation of dying languages. 28 CREVELS

References Adelaar, Willem F.H. 1991 ‘The endangered languages problem: South America’, in: .H. Robins and .H. Uhlenbeck (eds.) Endangered Languages, Oxford/New York: Berg, pp. 45-91. 1998 ‘The endangered situation of native languages in South America’, in: Kazuto Matsumura (ed.) Studies in endangered languages (Papers from the Inter- national Symposium on Endangered Languages, Tokyo, 18-20 November 1995), Tokyo: Hituzi Syobo, pp. 1-15. fc. ‘Threatened languages in South America’, in: Matthias Brenzinger (ed.). Baptista Morales, Javier 1995 ‘Los Misioneros Jesuitas de Mojos’, Yachay 21: 71-90, : Universidad Católica Boliviana. Beghin, François-Xavier 1976 ‘Exacciones a las poblaciones indias de Amazonia’, in: Jaulin (ed.) El etnocidio a través de las Américas, México D.F.: Siglo Veintiuno Editores S.A., pp. 127-67, [Spanish translation of Jaulin (ed.) [1972] Le livre blanc de l’ethnocide en Amérique, Paris: Librairie Arthème Fayard]. Block, David 1994 Mission culture on the upper Amazon: native tradition, Jesuit enterprise, & secular policy in Moxos, 1660-1880, Lincoln/London: University of Nebraska Press. Brenzinger, Matthias (ed.) fc. Language Diversity Endangered, Berlin/New York: Mouton de Gruyter. Castro Mantilla, María Dolores 1997 La viva voz de las tribus. El trabajo del ILV en Bolivia 1954-1980, La Paz: CID. Crevels, Mily and Willem F.H. Adelaar 2000- UNESCO Red Book of Endangered Languages: South America, on-line ver- 2002 sion: www.tooyoo.l.u-tokyo.ac.jp/Redbook/index.html. 2001 ‘South America’, in: Stephen A. Wurm (ed.) Atlas of the World’s Languages in Danger of Disappearing (Second edition, revised, enlarged and updated), Paris: UNESCO Publishing, pp. 80-81. Denevan, William M. 1966 The aboriginal cultural geography of the Llanos de Mojos of Bolivia, Berkeley: University of California Press. Eguiluz, Diego de 1884 Historia de la Misión de Mojos [1696] Lima: Imprenta del Universo de C. Prince. Geary, James 1997 ‘Speaking in tongues’, Time 150/1: 38-44. LANGUAGE DEATH IN BOLIVIA 29

Grimes, Barbara E. (ed.) 1996 . Languages of the World, 13th edition, Dallas (TX): Summer Institute of Linguistics. Grinevald Craig, Colette 1997 ‘Language contact and language degeneration’, in: Florian Coulmas (ed.) The Handbook of Sociolinguistics, Oxford: Blackwell Publishers Ltd., pp. 257-70. Grinevald, Colette 1998 ‘Language endangerment in South America: a programmatic approach’, in: Lenore A. Grenoble and Lindsay I. Whaley (eds.) Endangered languages: current issues and future prospects, Cambridge: Cambridge University Press, pp. 124-59. Heath, Edwin R. 1882 ‘Exploration of the Beni river’, Journal of the American Geographical Society of New York 14: 116-65. Lehm Ardaya, Zulema 1999 Milenarismo y movimientos sociales en la Amazonía boliviana. La búsqueda de la Loma Santa y la marcha indígena por el territorio y la dignidad, Santa Cruz de la Sierra: APCOB-CIDDEBENI-OXFAM América. Lema Garrett, Ana María (ed.) 1998 Pueblos indígenas de la Amazonía boliviana, La Paz: AIP/FIDA/CAF/ TCA/PNUD. Lijerón Casanovas, Arnaldo 1998 Mojos-Beni: Introducción a la historia amazónica, Trinidad, Bolivia: Editorial RB. Magio, Antonio 1880 Arte de la lengua de los indios baures de la Provincia de los Moxos conforme al manuscrito original del P. Antonio Magio de la Compañía de Jesús, [1749], edited by Lucien Adam and C. Leclerc, Paris: Maisonneuve & Cia. Marbán, Pedro 1894 Arte de la lengua moxa con su vocabulario y cathecismo [1701], published by Julio Platzmann, Leipzig: B.G. Teubner. Menacho, Antonio 1995 ‘Crónica de una expulsión’, Yachay 21: 93-118, Cochabamba: Universidad Católica Boliviana. Métraux, Alfred 1942 The native tribes of Eastern Bolivia and Western Matto Grosso, Smithsonian Institution. Bureau of American Ethnology Bulletin 134, Washington. Moore, Denny fc. ‘Endangered Languages of Lowland Tropical South America’, in: Matthias Brenzinger (ed.). Moreno, Gabriel René 1973 Catálogo del archivo de Mojos y Chiquitos, La Paz: Editorial Juventud. 30 CREVELS

Nordenskiöld, Erland 1922 Indianer und Weisse in Nordostbolivien, Stuttgart: Strecker und Schröder, [German translation of Nordenskiöld [1911] Indianer och hvita i nordöstra Bolivia. Stockholm: Albert Bonniers Förlag]. Orbigny, Alcide d’ 1958 Viajes por América del Sur, edición con estudio y notas, de los textos de d’Orbigny [Viajes y viajeros. Biblioteca indiana. Colección de textos anotados 3], edited by Charles Wiener and Carlos M. de la Condamine, Madrid: Aguilar. Plaza Martínez, Pedro and Juan Carvajal Carvajal 1985 Etnias y Lenguas de Bolivia, La Paz: Instituto Boliviano de Cultura. Steward, Julian H. (ed.) 1963 Handbook of South American Indians, vol. 3, The tropical forest tribes, Copper Square Publishers. Vaca Diez, Antonio 1989 De Santa Cruz a Reyes: Crónica de un viaje, Serie Biblioteca Médica Boliviano, vol. 1, Santa Cruz, Bolivia. Wurm, Stephen A. 1999 ‘Languages in danger: The life and death of the world’s languages’. Multiethnica 24-25: 28-35, Uppsala: University of Uppsala. GUARANI CRIOLLO Y GUARANI ETNICO EN PARAGUAY, ARGENTINA Y BRASIL

Wolf Dietrich Universität Münster

1.1. Objetivos En esta comunicación quisiera dar información sobre el proyecto de un atlas lingüís- tico que abarca la zona del contacto de dos lenguas de origen europeo, el castellano y el portugués, y del guaraní, hablado por más del 90% de la población del Paraguay y de regiones adyacentes de la Argentina y del Brasil. En otra ocasión (Dietrich 1994) ya presenté algunos primeros resultados de las encuestas que hicimos en el curso de este proyecto. Este guaraní, que aquí llamamos “guaraní criollo” porque no coincide simplemente con el guaraní paraguayo en el sentido estricto, es hablado por cerca de 5 millones de hablantes, todos no indígenas, sino nacionales de los respectivos países. Uno de los objetivos del futuro atlas lingüístico es saber hasta dónde se extiende el guaraní criollo en los tres países mencionados y cuál es el grado de competencia de los hablantes en guaraní, cuántos monolingües hay y, en el caso de los bilingües, con qué destreza manejan cada una de las dos lenguas. Para esto, se están haciendo en- cuestas en un mínimo de ochenta lugares de la zona, de los cuales más de cincuenta ya han sido explorados. Presentaremos a continuación algunos detalles del método y del cuestionario empleado.

1.2. La situación histórica Aparte de las lenguas europeas y del guaraní criollo, fruto de las condiciones muy particulares de la misión jesuítica de la época colonial, hay tres lenguas guaraníes étnicas existentes hasta hoy en la zona mencionada: son el mbyá, el chiripá (así lla- mado en Paraguay) o nhandeva (nombre dado en Brasil) y el kaiwá o kaiová (nombre usual en Paraguay) o paĩ (denominación usual en Paraguay). Si el guaraní criollo se caracteriza por un mayor o menor grado de influencia del castellano, las lenguas étni- cas mencionadas representan algo como el punto de referencia de un guaraní puro, original. No sabemos exactamente cuál fue el origen dialectal del guaraní paraguayo de hoy ni tampoco sabemos en qué medida este guaraní se relaciona con el de la mi- sión jesuítica, el llamado guaraní antiguo o clásico. Sin embargo, queda claro que, en el momento de la expulsión de los jesuitas en 1768, los indios guaraníes que habían vivido en las reducciones, en parte volvieron a la selva y se juntaron con sus compa- triotas que no habían entrado en dichas reducciones. El chiripá y el kaiwá son lenguas étnicas que, en su fonética, su morfología, sintaxis y léxico, se distinguen muy poco del guaraní criollo literario, auténtico, con la diferencia de que desconocen el gran número de hispanismos léxicos y sintácticos. Lamentablemente hasta hoy no existe 32 DIETRICH ninguna gramática de las tres lenguas y sólo del mbyá tenemos dos diccio-narios, bastante incompletos (Dooley 1982, Cadogan 1992). El mbyá, aunque difiera del chiripá y del kaiwá tanto en su sistema fonológico como en su morfología y sintaxis, se puede, con lo todo, considerar como un pariente muy cercano del conjunto guaraní para que sirva igualmente de lengua de comparación en nuestro proyecto.

1.3. La situación actual en el Paraguay y la Argentina Desde 1992 se está elaborando el Atlas Lingüístico Guaraní-Románico (ALGR) en un proyecto común que el colega Harald Thun, de la Universidad de Kiel (Alemania), y yo estamos realizando junto con la Facultad de Lenguas Vivas de la Universidad Evangélica del Paraguay, institución que se dedica también a la formación de profe- sores de guaraní en la enseñanza bilingüe del Paraguay. El objetivo del proyecto es registrar la variación lingüística en las regiones que corresponden a la antigua zona guaranítica del Paraguay y del Río de la Plata, es decir a la parte oriental del Para- guay moderno, situada al este del Río Paraguay, y las regiones limítrofes de la Ar- gentina y del Brasil. Incluimos, sin embargo, todas las regiones a las que el gua-raní criollo se ha extendido en el curso de los siglos XIX y XX, es decir, el Chaco para- guayo, las Provincias de Corrientes, Misiones, Formosa, Chaco y el Norte de Santa Fe, así como el Sur del Estado brasileño de do Sul y el Oeste del Estado de Paraná. Todavía no sabemos nada concreto sobre la situación del no- roeste de Río Grande do Sul y del oeste de . La situación lingüística del Paraguay actual está caracterizada por una presencia fuerte del guaraní, en cuanto lengua materna de la mayoría de la población, sobre todo rural, y de la comunicación solidaria, que se da normalmente entre personas que se conocen. Esta presencia es apoyada hoy en día por la oficialización del guaraní, que fue realizada en 1991 y que desencadenó una serie de medidas en la enseñanza del guaraní en las escuelas del país. Frente a esto, la situación de las regiones limítro- fes de la Argentina es bastante diferente ya que falta no sólo el apoyo político, sino también el prestigio del que goza el guaraní en el Paraguay. Se trata de las Provincias de Corrientes, con un guaraní muy auténtico y vivo en el noroeste de la Provincia, de Formosa, Chaco, norte de Santa Fe y Misiones. En muchas partes, el guaraní está en una situación de fuerte decadencia, con fenómenos de pérdida lingüís-tica de la que vamos a dar ejemplos concretos a continuación.

1.4. La situación actual en el Brasil En el Brasil, en el oeste de los Estados de Paraná y Mato Grosso del Sur, la situación es muy diferente según las regiones y se basa principalmente en la imigración re- ciente de paraguayos. La población de la región occidental del Estado de Paraná se compone mayormente de imigrantes de polacos, ucranianos, italianos, alemanes y "gauchos", es decir brasileños del extremo sur del país, del Estado de Río Grande del Sur, que han llegado allá a partir de finales del siglo pasado. De este modo el guaraní es inexistente allí, con la excepción de los lugares de frontera, Foz do Iguaçu y Guaí- GUARANI CRIOLLO Y GUARANI ETNICO 33 ra, donde se manifiestan contactos tradicionales con el Paraguay desde los años ochenta del siglo pasado. En esa zona no queda, sin embargo, ningún vestigio de la presencia de los españoles ni de la misión jesuítica de principios del siglo XVII. En , la zona del guaraní minoritario se extiende desde la frontera con el Paraguay hasta la capital, Campo Grande, con un centro alrededor de Dourados, pero la extensión exacta queda todavía por definir.

2.1. Metodología de las encuestas Las encuestas, en cada lugar de exploración, se hacen en guaraní y a partir del guara- ní, con un hablante nativo paraguayo. Se hacen con un mínimo de cuatro personas o grupos diferentes, según el grado de instrucción (clase alta y clase baja, Ca/Cb) y según la edad (oponemos la generación de los ancianos, a partir de 50 años (GI), a la de los jóvenes (entre 18 y 36 años, GII). Así, en cada lugar, tendremos una infor- mación pluridimensional, representada por una cruz, símbolo que permite dar cuatro informaciones. Arriba se colocan las respuestas de la clase sociocultural alta, abajo las de la clase baja (parámetro diastrático); a la izquierda colocamos las de la GII, a la derecha las de la GI: CaGII/CaGI CbGII/CbGI El cuestionario consta de unas 400 preguntas que se refieren a datos sociolingüísticos y a la actitud del hablante ante la lengua guaraní, a la fonética y fonología del guaraní y al léxico. Preguntamos, en esta sección, por las partes del cuerpo, actividades cor- porales, como ‘roncar’, ‘eructar’ o ‘regoldar’, defectos corporales, como ‘bizco’, ‘jo- robado’, o ‘legaña’, ‘orzuelo’ y denominaciones humorísticas de la cabeza, de la ore- jas, de la nariz, de personas sordas etc., tanto en guaraní como en castellano. Otra parte léxica se refiere al parentesco, donde, por ejemplo, se observa, por un lado, la conservación de la estructura tradicional de la familia y de los térmi-nos reservados a la designación de las hermanas y hermanos mayores y menores, tanto desde el pun- to de vista de la mujer como del varón:

(1) che ryk’y ‘mi hermano mayor’ (dice el varón), che ryvy ‘mi hermano me- nor’ (dice el varón), che reindy ‘mi hermana’ (dice el varón), che kyvy ‘mi hermano’ (dice la mujer), che ryke ‘mi hermana mayor’ (dice la mujer), che kypy’y ‘mi hermana menor’ (dice la mujer).

Esta terminología, que hace la subcategorización cada vez en el propio sexo, pero no en el ajeno, está conocida y usada sólo en las lenguas étnicas; en el guaraní criollo del Paraguay, los términos son conocidos por una parte de los informantes ancianos, so- bre todo en el campo, pero generalmente no se usan, sino que se usa el sistema del castellano adaptado simplemente a la morfología del guaraní: che ermáno ‘mi her- mano’, che ermána ‘mi hermana’, ne ermáno/ermána ‘tu hermano/ -a’, iñermáno/ iñermána ‘su hermano/,-a’. Al mismo tiempo, se observa la desagregación parcial o total del sistema y la sustitución por el sistema castellano, tanto en lo que se refiere al 34 DIETRICH conocimiento pasivo como, y tanto más, en lo que concierne el uso. Se ha abando- nado la subcategorización en el propio sexo, que antes era usual en el guaraní. En la Argentina, donde la desagregación está casi completa en este campo semántico, algu- nos informantes ancianos de las regiones donde el guaraní está todavía muy vivo co- nocen algunos términos, pero no todos. Si se mantiene un par de términos son siem- pre el de che ryke’y ‘hermano mayor’ (del varón), y che reindy ‘hermana’ (del va- rón). Entre los jóvenes argentinos de la región, sobre todo de la clase de poca instruc- ción escolar, se registra un desconocimiento casi total de los términos específi- cos.Siguen en el cuestionario otras secciones, que se refieren a la orientación en el espacio, a la denominación de los colores y partes que miden el grado del dominio de la sintaxis castellana o portuguesa, respectivamente. Se pide una traducción cas- tellana o portuguesa de frases del guaraní que se refieren a la valencia sintáctica, al uso de los pronombres personales de objeto, a la expresión de conceptos locales, etc. Al final, hay una sección de conversación dirigida sobre temas etnográficos y una de lectura. Se pide la lectura de un texto altamente literario, en este caso bíblico, tanto en guaraní como en castellano o portugués. Salvo en el caso de analfabetos, la lectura en guaraní no causa problemas fundamentales en el Paraguay, donde la gente, en gran parte, tiene una escolarización por lo menos parcial en guaraní, lo que no es el caso en absoluto en la Argentina y el Brasil, donde los hablantes del guaraní, que, en gene- ral, nunca tuvieron la ocasión de ver un texto escrito en guaraní, se niegan al princi- pio a tratar de leer. Sin embargo, la lectura del texto de la parábola del hijo pródigo (NT, Mt 15, 11-32), con sus arcaísmos y sus construcciones sintácticas desacostum- bradas, sí presenta, en general, grandes difi-cultades para los informantes, tanto en guaraní como en castellano o, respectiva-mente, en portugués. La lectura, parámetro diafásico en nuestra exploración pluri-dimensional, es altamente reveladora con res- pecto a la familiaridad del informante con ciertos lexemas, ciertas formas gramatica- les, ciertas construcciones sintácticas. Los errores cometidos dejan entrever las cos- tumbres lingüísticas del informante respectivo.

2.2. Arcaísmos en Corrientes frente a neologismos en Paraguay En el campo de la fonogía y fonética no hay grandes diferencias entre el guaraní étni- co y el criollo en aquellas zonas en las que el guaraní está vivo y no en vía de desapa- rición. La africada /č/ tradicional, que se ha vuelto fricativa, [], en Paraguay, se man- tiene en las lenguas étnicas salvo en el caso, bastante frecuente, de los indios acultu- rados. Muchos de ellos son trilingües, es decir que hablan su propio idioma, el guara- ní paraguayo y, aunque imperfectamente, el castellano. El guaraní criollo, para ellos, no tiene ningún prestigio frente a su propia lengua, pero lo saben hablar para adap- tarse a los nacionales y demonstrar que no son inferiores a ellos. La [č] se man-tiene, en forma de la realización arcaizante [č] también en el guaraní muy particular de la provincia argentina de Corrientes, que no es simplemente una extensión del guaraní paraguayo, sino un guaraní independiente que se ha formado desde finales del siglo XVIII. Parece que en aquella zona, al sur del Río Paraná, hubo guaraníes ya antes de GUARANI CRIOLLO Y GUARANI ETNICO 35 la creación de las reducciones jesuíticas. Después de la abolición de las reducciones, una parte de los habitantes de ellas huyó al sur, se juntó con los guara-níes libres y los colonos españoles, gauchos en su mayoría, y, en la época de las guerras de inde- pendencia, se escondieron en los esteros del Iberá (de y vera ‘aguas brillantes’). So- brevivieron y, en el nuevo Estado de la Argentina, se desarrollaron de manera com- pletamente independiente de sus antiguos connacionales del Paraguay. En ambos estados, hay una diglosia en el sentido de que, tradicionalmente, el guaraní es la lengua de la intimidad y solidaridad, el castellano la lengua de la vida pública y de la oficialidad. En el Paraguay, desde 1992, año de la oficialización del guaraní, esta lengua se está enseñando en muchas escuelas y la enseñanza se está des- arrollando en guaraní en siempre más materias. Esto crea varios problemas prácticos y teóricos: hay que formar docentes para la enseñanza y decidir cuál será el nivel del guaraní enseñado. ¿Se enseñará el jopara, esa mezcla de guaraní con muchos térmi- nos y muchas construcciones tomadas del castellano, que es el guaraní hablado en las calles y en las casas, o un guaraní más puro, menos popular, pero más auténtico, más literario? Pero, ¿cómo crear, en el segundo caso, una terminología propia para con- ceptos modernos como ‘teléfono’, ‘pagamiento’, ‘democracia’, ‘universidad’, etc., necesidad inevitable si se quiere que el guaraní no sea sólo la lengua de la tradición y del folclore. Sobre estas problemáticas hay dicusiones muy vivas en la actualidad, es decir sobre el problema de saber si son aceptables neologismos descriptivos como los que siguen (cf. Krivoshein de Canese/Acosta Alcaraz 1997):

(2) pumbyry ‘sonido lejano’ para designar ‘teléfono’ jehepyme’ẽ ‘REFL/NOM-precio-dar’, ‘lo que se da’ para designar el ‘paga-miento’ porokua pavê reko ‘modo de gobernar todos a la gente’ para designar el concepto ‘democracia’ mbo’eha guasu ‘enseñar-NOM loc-AUGM’, ‘casa grande de la ense- ñanza’ para designar ‘universidad’.1

2.3. Evolución fonológica en Corrientes Para dar más información sobre la desagregación del guaraní criollo en la provincia argentina de Corrientes, diremos que, en la fonética, se observa la frecuente debilita- ción de las vocales nasales, que puede ir hasta la desnasalización total. Este fenóme- no se observa sólo en el habla de personas que ya no hablan con fluidez y no tienen muchas ocasiones de hablar guaraní. Así, parece que es un fenómeno de la lenta pér- dida de la lengua. En regiones rurales del noroeste de la Provincia, donde la lengua está viva, y en el habla de informantes que hablan fluentemente, no se observó ni siquiera la mínima huella de este fenómeno. La desnasalización parece producirse sobre todo en la sílaba fonológicamente decisiva, en la final, por ejemplo, en:

1 REFL ‘reflexivo, NOM ‘nominalizador’, NOM loc ‘nominalizador locativo’, AUGM ‘augmentativo’, ELAT ‘grado elativo’. 36 DIETRICH

(3) porá [po’ra] en lugar de porã [põ’rã] ‘lindo, bueno’

En la mayoría de los casos, la nasalidad de la palabra fonética o del sintagma, que en guaraní se extiende a partir de la sílaba tóncia, se mantiene en la sílaba precedente, de modo que no hay problema de orden fonológico, por ejemplo:

(4) chẽ tí ‘mi nariz’ en lugar de chẽ tĩ (5) iñãký [iã’kĩ] ‘está mojado’ en lugar de iñã’kỹ [iã’k] (6) õké [õ’ke] ‘puerta’

Pero se escucha también oké [o’ke] en lugar de õkẽ [õ’kẽ] ‘puerta’. En el caso de oké, el signo sería homófono con la forma verbal

(7) oké ‘durmió’ pero tendremos que analizar con métodos más sofisticados si en la conciencia de los locutores hay oposición u homofonía sólo para nuestros oídos inadaptados al dialecto o si hay realmente homofonía. Además, la nasalidad se mantiene en muchas palabras en el habla de ciertas personas, sobre todo de la clase baja, pero irregularmente. El abandono, por lo menos parcial de la oposición ‘nasal’ – ‘oral’ parece ser evidente si se tiene en cuenta que muchas de las parejas mínimas tradicionales no se dan en el guaraní correntino ya que uno de los elementos léxicos no se usa, como, por ejemplo:

(8) apu’a ‘redondo’ frente a apu’ã ‘levantamiento, levantarse’

En el guaraní correntino existe sólo [apua] o [ãpua], sin oposición semántica, con el sentido de ‘levantarse’ y ‘(vientre) levantado’, es decir ‘(mujer) embarazada’, mientras que el concepto ‘redondo’ se expresa por el hispanismo i-redóndo‘es redon- do’.

2.4. La nasalidad en el guaraní correntino La nasalidad no se extiende progresivamente en el sintagma, de modo que no funcio- na ni la prenasalización de las oclusivas sordas: /p/ > [mb], // > [nd], /k/ > [g], co- mo en el pasaje morfofonológico de una base como

(9) oký ‘llovió’ a la forma causativa omongý ‘hizo llover’ pero, en el guaraní correntino, omboký, ni la nasalización de /p/ > [m], /t/ > [], /k/ > [] con las consecuencias morfonológicas previsibles: GUARANI CRIOLLO Y GUARANI ETNICO 37

(10) (guaraní correntino) nde tí ‘tu nariz’ en lugar de ne tĩ ñũpe ‘en el campo’ en lugar de ñũme guaranípe ‘en guaraní’ en lugar de guaraníme (véase también Adelaar 1994: 128-133)

Este fenómeno, que existe también en lenguas étnicas, como, por ejemplo, en mbyá, debe considerarse como un arcaísmo lingüístico en el guaraní correntino, y no como un elemento de decadencia. El guaraní correntino, en este como en otros casos, no ha sufrido la normalización lingüística ejercida por parte de las instituciones paraguayas.

2.5. La vocal central cerrada en el guaraní correntino La vocal central //, en el habla imperfecta de muchos hablantes, se realiza de otra manera que en el guaraní paraguayo: es menos labializada y, sobre todo, más palatal. A veces parece confundirse con la /i/ aunque sea quizás incorrecto decir que la oposi- ción //-/i/ se haya suprimido. Hasta en casos como [aiki’ti] en lugar de [aik’tĩ] ‘yo [lo] corté’ la vocal respectiva parece ser algo distinta de la /i/ final. Los jóvenes mu- chas veces la realizan como una velar labializada y más abierta delante de una sílaba acentuada del tipo:

(11) C + /u/: [pun’tu], [p’tú] en lugar de [p’tũ] ‘oscuridad, noche’

El fenómeno aparece también en el habla de jóvenes de las regiones brasileñas ya visitadas y está todavía por analizar más detalladamente. Eventualmente podría in- terpretarse como un fenómeno más general de la inseguridad fónica de ciertos ha- blantes.

2.6. Reducciones en el léxico del guaraní correntino Además, se observan reducciones gramaticales y léxicas en el guaraní correntino de todos los lugares explorados. Esto se refiere a la perdida de la numeración de más de ‘dos’, a veces ya a partir de ‘uno’, es decir, de peteĩ ‘uno’, mokõi ‘dos’, mbohapy ‘tres’, irundy ‘cuatro’ y a la pérdida de los términos de parentesco tradicionales (no sólo de la designación de hermanos mayores y menores del varón y de hermanas mayores y menores de la mujer, términos sustituidos por che ermáno ‘mi hermano’ y che ermána ‘mi hermana’, sino también la frecuente sustitución de:

(12) che ména ‘mi esposo’ por che espóso, y de che rembireko ‘mi esposa’ por che kuña ‘mi mujer’ (13) t-amõi, che r-amõi ‘antepasado’ y ‘abuelo’

La palabra t-amõi, che r-amõi (13), que en lenguas étnicas como el mbyá significa ‘antepasado’ y ‘abuelo’, ha perdido el segundo significado en el guaraní paraguayo y 38 DIETRICH se usa sólo para ‘antepasados’. El ‘abuelo’ se llama che taitá guasú o, más común- mente, por che abuelo. En el guaraní correntino, sin embargo, che r-amõi se ha per- dido como tal y se sustituye por el hispanismo respectivo.

(14) che jarýi ‘mi abuela’

Che jarýi ‘mi abuela’ es un lexema vivo en mbyá como en el guaraní paraguayo, pero muy poco conocido en el guaraní correntino. Igualmente desconocidas en el guaraní correntino son las designaciones del ‘nieto’ y de la ‘nieta con respecto al abuelo’ (15a), y del ‘nieto’ y de la ‘nieta con respecto a la abuela’ (15b), palabras completa- mente vivas en mbyá.

(15) a che r-emimenõ b che r-emiarirõ

En el guaraní paraguayo, se observan una reducción del sistema guaraní de los térmi- nos del parentesco y un cambio semántico: La distinción guaraní entre la referencia al abuelo o a la abuela ha sido abandonada y sustituida por un término único, che r- emiarirõ ‘mi nieto, -a’. De este modo, el sistema está entre el guaraní puro, que no distingue el sexo de la persona designada, pero sí el de la persona de referencia, y el castellano, que preve distinciones contrarias. Un elemento de semi-romanización. El viejo término che remimeno, así en la ortografía oficial paraguaya, es un arcaísmo inusitado en Paraguay y una palabra completamente perdida en el guaraní correntino. Thun y Aquino (1998: 12-14) han descrito los resultados obtenidos, en el Para- guay, con respecto a (16):

(16) che uke’i ‘mi cuñada (dice la mujer)’, y che r-ovaja ‘mi cuñado/cuñada (dice el varón)’

La distinción que tradicionalmente se hace con respecto al varón, y que refleja la po- sición social de la esposa en la familia del marido, está completamente viva en las sociedades tradicionales de las etnias guaraníes, pero ya no en las de los nacionales paraguayos y argentinos. Sin embargo, mientras que uke’i se usa muy poco en Para- guay, pero se conoce bastante como arcaísmo, la palabra está completamente desco- nocida entre los guaranihablantes de la Provincia de Corrientes. Allí, che r-ovaja es el término único para ‘cuñada, cuñado’ en todos los respectos, como también en el habla moderna de muchos paraguayos. En el guaraní correntino se observa además: i) la pérdida de la orientación local. Ejemplos de este fenómeno son: GUARANI CRIOLLO Y GUARANI ETNICO 39

(17) che akatúa ‘mi derecha’ sustituido por che derecha, che asúa ‘mi izquier- da’ por che iskjérda, palabras vivas y usuales tanto en mbyá como en el guaraní paraguayo; ii) la pérdidad de la terminología del cuerpo humano en casos como:

(18) che kwã ‘mi dedo’ sustituido por che dédo; iii) la pérdida de términos para colores básicos como:

(19) hový ‘azul y verde oscuro’ sustituido por el hispanismo; iv) la pérdidad de conceptos fundamentales como:

(20) kuimba’é ‘varón’, sustituido por ómbre, etc.

3. Morfología, categorías gramaticales y sintaxis En la morfología y sintaxis no hay grandes diferencias entre el guaraní étnico y el criollo. Es verdad que el mbyá no conoce el doble futuro del kaiwá, del chiripá y del guaraní criollo. De los sufijos ’-ta ‘futuro con respecto al presente’ y ’-ne ‘futuro absoluto’ sólo usa ’-ta. En este caso es una de las lenguas étnicas que no ha aceptado la innovación ’-ta en la historia relativamente reciente de las lenguas tupí-guaraníes. Pero es verdad también que el futuro absoluto, en el guaraní paraguayo y en el cor- rentino, hoy en día más bien tiene el valor de un modo de la probabilidad epistémica que el de un tiempo:

(21) o-hó-ne 2 3-ir-FUT.ABS ‘probablemente irán’

Todas las lenguas mencionadas mantienen la diátesis reflexiva en je-, ñe, la recíproca en jo-, ño-, la factitiva en mbo-, mo-, la factitivo-comitativa en ro- y la causativa en -uka y toda la demás morfología que las lenguas guaraníes tienen en común. El gua- raní criollo hace un uso excesivo de la voz reflexiva para expresar acciones que ocur- ren sin agente, por ejemplo:

(22) o-je-he’i upéicha 3-REFL-decir como esto ‘así se dice’, según el modelo castellano.

2 FUT.ABS ‘futuro absolutivo’. 40 DIETRICH

Esta construcción, que sigue el modelo del castellano, es sólo un abuso normativo, no una desviación sistemática. Con respecto a la gramática y sintaxis, en las grandes líneas, hasta el guaraní criollo forma parte del conjunto de las lenguas indígenas.

4. Conclusión En resumen, se puede constatar un orden gradual que toma su punto de partida en las lenguas étnicas, más o menos “puras”, aunque ellas también sufren el influjo de las lenguas nacionales que las rodean, el castellano y el portugués, y que pasa por el gua- raní criollo del Paraguay, cuya forma culta, el llamado guarani-eté ‘guaraní-ELAT’, está muy cerca de las lenguas étnicas, para llegar al jopará ‘mezcla’, ‘lengua mixta’, caracterizado por muchas influencias léxicas y sintácticas del castellano (cf. Lustig 1996). Éstas pueden ser préstamos tanto directos como indirectos. Frente a este or- den, el guaraní correntino presenta características arcaizantes, por una parte, y, por otra parte, rasgos de una lengua ya reducida en varios aspectos de la fonética y del léxico.

Bibliografía Adelaar, Willem F.H. 1994 ‘The nasal/oral distinction in Paraguayan Guaraní suffixes’, en: Mary Ruth Wise (coord.), Lingüística tupí-guaraní/caribe. Etudios presentados en el 47º Congreso Internacional de Americanistas, 7-11 de julio de 1991, Nueva Or- leans, Revista Latinoamericana de Estudios Etnolingüísticos (Lima, Perú), VIII: 125-133. Cadogan, León 1992 Diccionario Mbyá-Guaraní – Castellano, Edición preparada por Friedl Grün- berg bajo la dirección de Bartomeu Melià, Asunción: CEADUC – CEPAG. Dietrich, Wolf 1994 ‘Mbyá, guaraní criollo y castellano: El contacto de las tres lenguas estudiado en un grupo mbyá de Misiones’, Signo & Seña. Revista del Instituto de Lingüística (Buenos Aires) 3: 55-71. 2001 ‘Zum historischen Sprachkontakt in Paraguay: Spanische Einflüsse im Guaraní, Guaraní-Einflüsse im regionalen Spanisch’, en: Gerda Hassler (dir.), Sprachkontakt und Sprachwandel, Münster: Nodus (im Druck). Dooley, Robert A. 1982 Vocabulário do Guaraní. Vocabulário Básico do Guaraní Contemprâneo (Dialeto Mbüá do Brasil), Brasília: SIL. Krivoshein de Canese, Natalia y Acosta Alcaraz, Feliciano 1997 Ñe’ẽryru avañe’ẽ - karaiñe’ẽ, karaiñe’ẽ - avañe’ẽ, Diccionario guaraní- español, español-guaraní, Asunción: Universidad Nacional (Colección Ñemitỹ). GUARANI CRIOLLO Y GUARANI ETNICO 41

Lustig, Wolf 1996 ‘Mba’éichapa oiko la guarani? Guaraní y jopará en el Paraguay’, Papia 4,2: 19-43. Thun, Harald y Aquino, Almidio 1998 ‘El Atlas Lingüístico Guaraní-Románico (ALGR), un trabajo necesario para actualizar informaciones lingüísticas sobre el español y el guaraní del Para- guay’, Ñemitỹ, Asunción, 36, págs. 8-14. Trad. portuguesa: ‘O Atlas Lingüístico Guaraní-Românico (ALGR). Um trabalho necessário para atu- alizar informações lingüísticas sobre o guarani e o espanhol do Paraguai’, Ca- dernos de Tradução (Porto Alegre) 5, 1999, págs. 53-66.

DIFERENCIAS ENTRE EL LENGUAJE DE HOMBRES Y MUJERES EN CHIQUITANO (BESIRO)

Sieglinde Falkinger Österreichisches Lateinamerika Institut

1. Introducción En el siglo pasado el fenómeno de diferencias en el hablar debido al sexo se juzgó algo muy exótico y fue localizado primero en pueblos exóticos. En la literatura lingüística, la diferencia en el lenguaje hablado por mujeres fue vista como desviada de la norma. Jespersen (1998), presentando ejemplos en chiquitano, habla de la “debilidad femenina” y Sapir (1990, 1991), quien encuentra ese fenómeno en la lengua Yana, lo pone entre “abnormal types of speech”. En general se pensaba que el vocabulario de mujeres era reducido. El lenguaje femenino tenía que ser simple y emocional. El hablar de los hombres por el contrario se consideraba como constructivo, práctico, abstracto y más complejo (vea Key 1975: 15). Basándose en esto se tomó por supuesto la variante mas-culina como norma.

2. Descripción del fenómeno por los jesuitas El fenómeno de un lenguaje diferente entre hombres y mujeres en el chiquitano fue mencionado primero en los relatos y gramáticas de los jesuitas en el siglo XVIII. En ese entonces los jesuitas fueron a tierras lejanas para convertir los indígenas a la religión católica. Para ese fin aprendieron lenguas indígenas usándolas en la catequesis para que los paganos entendieran bien la religión cristiana (vea Furlong 1946). Escribieron muchas gramáticas y vocabularios para poder traducir a esas lenguas los textos mas importantes del catecismo, los sermones etc. Así también para el “chiquitano”, o la “lengua chiquita” como la llamaron entonces. Los jesuitas fundaron en Chiquitos 10 reducciones1 en las cuales se hablaba como lengua general el chiquitano, la lengua de un grupo de los indígenas de la zona, probablemente de San Javier o San Rafael, porque esas fueron las primeras fundaciones en las tierras bajas de Bolivia. En muchas cartas se quejan los jesuitas de las dificultades extraordinarias y las complicaciones que causaban en el chiquitano las diferencias al hablar entre el hombre y la mujer:

“... la dificultad que mas ponderan en la Chiquita comunmente diciendo, que las mugeres hablan lengua distinta dela de los hombres. (...) Para entenderlo supongase que todos los nombres, pronombres, verbos, preposiciones, y de adverbios, declinables ó conjugables, tienen (para quando se haya de hablar de tercera persona) una inflexion,

1 De la palabra latina reductio = reducción: “Ad ecclesiam et vitam cvilem essent reducti” 44 FALKINGER que podemos llamar varonil, y otra mugeril, en singular, y lo mismo en plural, como se verá en las tablas de Declinacones y Conjugaciones. Ahora pues, las mugeres jamas pueden usar dela inflexion varonil; sino es solo quando refieren el dicho, ò clausula que habló un hombre. Al contario los hombres usan de ambas inflexiones asi en singular, como en plural, pero con esta diferencia: que de la varonil usan solo quando hablan de Dios, ò de las divinas Personas, ò de Angeles, Demonios, Hombres, Dioses falsos, y en suma de todo lo que los Pintores pintan en figura de varon: mas quando hablan de muger, ò mugeres, o qualquier otra cosa, que no sean las dichas, usan dela inflexion mugeril, como queda dicho (Ms-Camaño: 4).”2

Los jesuitas encontraron en Chiquitos, una lengua que tenia variación debido al sexo, la cual constituía una forma ideal para su tarea misional porque los hombres distinguían en forma gramatical cuando hablaban de “Dios, Angeles, demonios u otros hombres”. ¡Las mujeres no debian pronunciar esas palabras! La meta de los jesuitas era la conversión al catolicismo. Sus métodos eran general- mente pragmáticos y no tenían escrúpulos para introducir nuevos conceptos, pero en el caso del chiquitano tenían que someterse a las exigencias de esa lengua porque como “hombres que hablaban de Dios” tuvieron que usar la lengua adecuada, y eso no les fue fácil:

“No hai duda que tal diferencia de lenguage causa a los principios no pequeño embarazo; y que un misionero aún despues de practicar por algun tiempo el confesionario de los hombres, suele todavia rehusar el sentarse a oir confesiones de mugeres: porque no teniendo aun (por falta de trato) acostumbrado el oido al lenguage de ellas, temen y con razon, que no las entenderan (Ms-Camaño: 6).”

En los confesionarios se hacian entre otras, las siguientes preguntas:

Hombre hablando con hombre sobre mujer: “Unchunca nPais naxiñatimo? Has deseado alga muger? Axiñaca apaezo imo? La deseaste mucho? Taizoo3 atacu aemo? Consintio? Aenza4 nemozoe5? Era tu pariente? Aapi emotozo naquipoci6?” Era pariente de tu muger?

2 Las descripciones históricas que presento, vienen en su mayoría de la “gramatica de la lengua Chiquita” escrita por el Padre Joaquin Camaño jesuita de La Rioja Argentina, quien después de la expulsión llegó a Italia y escribió bosquejos gramaticales para el Abate Lorenzo Hervás. El Manuscrito hizo una gran peregrinación antes de llegar a las manos de Guillermo de Humboldt. Hoy se encuentra entre su herencia en Kracovia. Uso aquí la ortografía original y también la terminología, las abreviaciones, subrayados y símbolos tal como se encuentran en los diversos manuscritos. 3 Ichaiçoca. 1a. 3.o / r. po. Consentir permitir. Con ichacu, atacu & 1a. (= Präp. “por”) (MsM-V: 166) 4 Aenza . Parta. Interroga. Item “antes bien, quanto mas” (MsP-21) 5 Consanguinidad, emozoequis: ñemozoequi. Consanguineo, emozoez, zemozoe, emozoe, emozotostii, Pl. oñemozoe, zomemozoe, amemozoe, +, omemozoez (MsM-V: 165) LANGUAJE DE HOMBRES Y MUJERES EN CHIQUITANO 45

Hombre hablando con mujer sobre hombre: “Axiñaqu imo naqui ñoñeis? Has deseado â algun he.? (hombre) Axiñaca apaezo ymotij? Le deseaste mucho? Taizotij itacu aemo? Consintio? Aenza naqui nemozoe? Es tu pariente? Aenza emotozo naqui ycan‘7? Es pariente de tu marido? (Ms Paris 19: 209f).”

En este ejemplo se ve que el pronombre masculino naqui y el sufijo -tii caracterizan el lenguaje del hombre en la tercera persona singular (vea tabla 2 y 4 y 5 en el appendix).

De hombre a hombre, hablando de mujer se usa la versión femenina : objeto [+ fem] De hombre a mujer, hablando de hombre se usa la versión masculina : objeto [+ masc] El objeto define el género si el hablante es masculino.

Un detalle que llama la atención es el hecho de que las preguntas en los confesionarios se hicieron en ambas variantes, pero oraciones elementales como el “Padre nuestro” y el “Ave Maria” no. Estas solamente se encuentran en la versión masculina en los manu- scritos. Los jesuitas eran monjes que vivían el celibato. Evitaron en lo posible el trato con mujeres. Se dice por ejemplo del padre Esmid que él nunca entró solo en una casa particular de indígenas, ni daba la mano a una mujer. Tampoco enseñó a las mujeres ninguna labor. Este mismo Padre Esmid instruyó a los sacristanes para corregir a los sacerdotes cuando dijeran alguna “barbaridad”. Eso ilustra que han sido hombres los informantes y colaboradores de los padres jesuitas en cuanto a la lengua. Con eso se entiende la perspectiva que tenían. La versión masculina la establecieron como norma. Eso se ve en la construcción de las tablas de las declinaciones y conjuga- ciones. Lógico que la forma masculina ocupa el primer lugar (vea las tablas 1, 2 y 3 en el appendix). La forma femenina se explica al comienzo y no se menciona más.

“Quarto. La tercera persona del plural de varones en qualquiera declinacion es la misma, que la tercera persona del singular, mudando el tii, en ma, ò en o segun algunas parsialidades de indios. Y assi de la tercera persona singular ipoostii mudando tii en ma sale la tercera plural de varones ipoosma, y asi por todos los demas nombres. Item la tercera singular de mugeres, o de cosa que no sea Dios, Angel, varon, o Demonio, es la tercera singular de varon, quitado el tii; V:g: ipoostii, quitando el tii, queda ipoos su casa de ella, de muger, perro, gallina &, y esto en todos los nombres; Por tanto en adelante se omitiran la tercera singular de muger, y la tercera plural de varones, por ser cosa tan clara (MsM-G: 7f).”

6 quipocis esposa, muger casada (MsP-21) Esposa, izipoci, aquipoci, & 1a./ (MsM-V:219) 7 marido naqui icana (MsM-V:430) Esposo, n‘icana ñy dice la muger (MsM-V: 290) 46 FALKINGER

Lo mismo vale para el verbo (vea el esquema en la tabla 1en el appendix).

femenino vs mugeril masculino vs varonil 3..sg. ipoos ipoos-tii 3.Ps.pl. ipoos ipoos-ma

En este ejemplo se ve que la versión femenina es vista como diferente de la masculina y además incompleta. A los hombres los consideraron más inteligentes y con capacidad de distinguir algo como ‘género’ - porque el fenómeno de la diferencia en el hablar fue considerado un tipo de género. Si definimos género según Hockett “Gender are classes of nouns reflected in the behaviour of associated words” (Hockett 1958: 231, cit. según Corbett, 1991: 1), entonces los jesuitas tenían cierta razón. Pero en el chiquitano el ‘género’ no se limita al nombre, sino que aparece con la misma morfología en verbos y otras clases de palabras. Las terminaciones del posesivo en la declinación son casi idénticas a las que indican la persona en la conjugación.

“No tiene la Lengua Chiquita, hablando propriamente, aquella accion de generos, masculino, femenino, neutro & que tiene la Lengua Latina en sus nombres; pero tiene una cosa mui semejante; como di á entender hablando (dela) diferencia que hai entre el parlar varonil, y mugeril. (...) Supuesto esto, se puede mui bien decir que los hombres en lengua Chiquita, quando no hablan (en) persona de alguna muger, ò como muger, hacen distincion de genero; consideando como masculinos todos los nombres proprios y apelativos de Dios, Angeles, Demonios & y como femenino todos los de mugeres, bestias, y cosas inanimadas; y que usando dela inflexion varonil ò masculina delas voces declinables quando se refieren en la oracion a alguno de aquellos nombres, y dela mugeril ò femenina, quando se refieren á alguno de estos otros, vienen hacer una especie de conordancia en genero semejante àla que se hace (en) la lengua Latina (Ms-Camaño: 45).”

Sorprendente es que la diferencia solamente se da en la tercera persona en singular y plural. Vea las listas de las declinaciones y conjugaciones (vea las tablas 1, 2 y 3 en el appendix).

“Al contrario en Chiquito no hai concordancia, quando (se) habla en primera ò segunda persona, ò en primero y segundo posesivo; Porque nombres y verbos & no tienen en estas mas que una inflexion. La concordancia es solo en las terceras personas, ò terceras posesivos, donde hai dos inflexiones una: varonil, y otra mugeril. (...) el Articulo concuerda con el nombre, y el pronombre con el objeto; y hablo de pronombre indeclinable; porque aunque este y ... articulo no tienen inflexiones, con todo eso como hai pronombres y articulos varoniles ò masculinos, que no puede usar la mujer, y el varon usa so(la)mte quando habla de Dios, Angeles, Demonios, hombres & y otros LANGUAJE DE HOMBRES Y MUJERES EN CHIQUITANO 47 pronombres (ar)ticulos equivalentes femeninos, que usa siempre la muger, y usa tambien el varon quando habla de mugeres, bestias &. Este usar el hombre un articulo (pro)nombre quando habla de cosas que considera como femeninas, y otro qdo habla de entes que considera como masculinos, viene à ser una especie de concordancia en genero (Ms-Camaño: 45).”

Más notable es esa diferencia en el plural.

“Singular mandole el mi padre á él, que fuese otra vez en pos de aquel Padre hombre: baquipuco8ty naqui iyai moty, m‘airoty9 t‘ato isiuty manuqui Pais mujer: baquipuco n’ixup‘ imo, m‘airo t‘ato isiu manu Pais Plural hombre: baquipucoma unama oñeica moma, m‘airoma t‘ato isiu manuma Paica mujer: upaquipuco n‘oñeica ñome, m‘ameno t‘auto yosiu man‘iño Paica. les mandaron los hombres à ellos, que fuesen otra vez en pos de aquellos Padres (Ms-Camaño: 4).”

Mirando el paradigma de la conjugación y declinación según Camaño (tablas 1 y 3 en el appendix) se puede ver que la partícula que indica persona o posesión se antepone: es un prefijo. Las mujeres usan la 3. Ps.pl. siguiendo al paradigma de los prefijos, mientras los hombres usan ambas formas y rompen con esta regla haciendo uso del sufijo –tii en la 3a persona singular y -ma en la 3a persona plural, hablando de Dios, Angeles, hombres, etc. Camaño nos da un ejemplo bien detallado:

“Despues de sabido todo esto, enque hai increible multitud de minudisimas reglas, resta todavia gran peso de dificultad en la practica: porque, como arriva dixe, hai que atender a concordar adjetivos, sustantivos, verbos, advervios, preposiciones, articulos, pronombres, relativos; finalmte. todo lo declinable y conjugable, y mucha parte de lo indeclinable, cada voz, ó cada palabra con el ente à quien se refiere. De modo que muchas veces en una oracion, o periodo, se van alternando las voces, una en inflexion mugeril, la siguiente en varonil de singular, la otra en mugeril de plural & el adjetivo en inflexion mugeril y su sustantivo en varonil, ò al contrario. El verbo en varonil y su preposicion ò caso en mugeril, ò al reves. La preposicion en mugeril y su caso en varonil && Cosa tan extraordinaria, y de tanta algarabia bien digna es de algun exemplo: y asi pongo aqui el primero que me ocurre con su romance literal debajo: y paraq mejor se note la variedad de concordancias, pongo sobre cada voz una ò dos

8 Mandar, iquipuca, 2a, 3 co, r. to (MsM-V.425) Mandato, mandamiento, Baquipucus. Yaquipucu, & 1a./ Niyaquipuco obi, con iñemo, aemo, &. 3a./ Pl. Niyaquipucu oboi iñemo, & (MsM-V: 425) 9 Sg: yicaty, aiquity, zirotiity, Pl. oicati o oiquity, zomecaty, amenoty, ziromaty, omenoty – ellas (MsM-G: 117) 48 FALKINGER letras, que notan la inflexion en que dha voz está. La m. significa inflexion mugeril. La v. signifca varonil. Si alguna de ellas se añade s. quiere decir que es de singular. Si se añade p. quiere decir que es de plural.

m.p. v.s. m.p. m v.p. v.p. Ome’enzoro10 oi-ty Pedro maniño ni ñacunomoco11 unama fueron perdidos por Pedro aquellos escritos de los

v.s. m.p. m.s. m. v.p. m. yai-taiki-ty omixima12 au n’asarus-ma13 Paica, ni ñaquichoni- sus atentatos, bellos segun el juicio del los Padres, los quales leyó

v.s m.p. v.s. m.s. m. v.s. m.s. ma-ta-ty yocuu14 aetama-ty, mta ito isucari15 n’ipaqui-tos-ty atoñe el á solas, y también delante de su madre propria omeenzoro es inflexion mugeril de plur. del verbo ñe‘nzoca; porque la persona paciente, que son los Escritos, es femenina. Si fuera mascula. diria e‘nzoroma. Por la misma causa se pone el pronombre mugeril de plural maniño; y no manu de singular; ni manuma, ni maniqui varoniles, aquel de plural, y este de singular. Por lo mismo se pone el articulo mugeril ni, que es de ambos numeros; y la inflexion mugeril de plural omixima del adjetivo oxima; y finalmte. la inflexion tambien mugeril del plural yocuu, dela preposicion icuu; que es caso del verbo Absoluto ñaquichonimaca, leer, y por tanto debe concordar con los escritos leidos. Oity es inflexion varonil singr. dela preposicion zobi; porque su caso es Pedro masculino = y por lo mismo se pone yaitaiquity inflexion varonil del tercer posesivo de singular del sustantivo plural yaitaiqui, mis antenatos: porque siendo los antenatos de Pedro, Pedro es el genitivo de posesion, con quien hade concordar ese sustantivo. Al contrario, aunque yaitaiqui es el genitivo de posesion de los escritos, está en inflexion varonil de singular, con todo, como el nombre en si mismo es plural por la particula taiqui, y los antenatos son muchos, se pone el sustantivo macunumococa (escritos) en su inflexion varonil de plural ñacunomoco, que debia decir ñacunomocosma: si no sig(....) el varonil de plural unama, que suple, y entra en lugar dela particula final (...) un à todas las inflexiones varoniles de plural de todas las Declinaciones. Aetamaty es inflexion varonil de singular del adverbio ñetama, yo á solas aetama tu a solas: aetamaty, el á solas & y está en esa inflexion, porque debe concordar con Pedro, que es el que leyó a solas.

10 Eêzoro. 3a. De ñeêzoca. perderse (..) 5a pl. Omeêzoro. deto. aêzo. Pl. ameêzo (MsP-V) 11 Escribir, icunomococa, 2a. 3. no, r. to (MsM-V: 284) 12 Beldad, belleza: coñocos. Oximacas. Izoñoco; acoñoco, icoñocostii. Pl. Ocoñoco, zoizoñoco, & 1a./ ñoximaca, oximaca, oximacastii, &. masc.: oxima, fem.: omixima 5a. / coñotus. Oximatus (MsM-V:82) 13 Opinion, au niyaçaru, n‘açaru (MsM-G) 14 Sobre, de cosa llana, icuu. Pl. yocuu. V.g. icuu messas: sobre la mesa (MsM-V: 632) 15 Isucari = Adv. del lugar “delante de mi”, “en mi presencia” 1. Decl. (MsM-G: 184) LANGUAJE DE HOMBRES Y MUJERES EN CHIQUITANO 49 ipaquitosty es tambien inflexion var. de sing. del sustantivo ipaqui, mi madre; que debe concertar con Pedro masculino, porque este es su genitvo de posesion, que la madre es de Pedro. Fuera de esta dela inflexion, hai en el mismo nombre otras dos concordancias tambien masculinas: una que consiste en la sustancia misma del nombre, y otra que consiste en su incremento, que es la particula to. Por que si Pedro fuera muger, ò si en lugar de Pedro se pusiera Maria, entonces para decir delante de su madre, se habia de decir isucari n‘ipapas, con el nombre ipapa, que significa madre de la muger; y sin el incremento to, ni ta, que no se añaden á los nombres de parentezco, quando se refieren a muger, si esta no se nombra imediatamente. Quando la muger se nombra despues del nombre de parentezco, sela añade à este el incremento que le corresponde; y asi se dice ipapata Maria, madre de Maria: aito Francisca, hijo de Francisca, hijo de Franca.,- que es de zai, mi hijo. ñaquichonimataty es inflexion var. de sing. del modo respectivo del verbo absoluto ñaquichonimaca, que se pone en esa inflexion; porque la persona que hace o que lee es Pedro. Ni ñaquichonimataty yocuu. En esta frase, ó expresion, el ni que precede al verbo, junto con el ta inserto en él, y con la preposicion yocuu, forman un relativo de caso obliquo y de plural mugeril, que todo ello equivale al relativo Latino qua dela oracion qua legit Petrus. Y así el verbo que está en inflexion varonil, ò masculina, viene à estar precedido de articulo, seguido de preposicion, y aún penetrado de particula, todo femenino. El ni, como ya dixe, es articulo mugeril y quando precede à verbo, es articulo relativo, que aqui se refiere a los escritos leidos: mas de suyo es indiferente para singular ó plural y para ser relativo de caso recto, (hoc est Nominativo) ò de caso obliquo. El ta es la particula que forma el modo Respectivo de este verbo, por el qual modo se hade hablar siempre que la oracion es de relativo de caso obliquo, ó de relativo que en Latin debe estar en Genitivo, Dativo, ò en otro caso que no sea Nominativo. Y por esto dho ta determina al ni à ser relativo obliquo. Del yocuu ya dixe al principo, que es el caso que rige el verbo absoluto ñaquichonimaca, y que es inflexion de plural mugeril de la preposicion icuu; y por tanto ese yocuu es lo que determina al ni, y al ta à que siendo relativo mugeril obliquo, lo sean de plural mugeril y de tal determinado caso, que alli es Ablativo; porque icuu es preposicion de ablativo; y asi ñaquichonimaca icuu, vel yocuu, quiere decir propria, y verbalmente leo por sobre esos escritos: y ni ñaquichonimataty yocuu, quiere decir por sobre los cuales leyó. La razon de esto, y la explicacion se dará hablando de los verbos absolutos. au n‘asarusma Paica à juicio de los Padres = Aqui el au es preposicion mugeril de singular; porque su caso es el juicio femenino. La n es el articulo femenino ni con elicion de la i. El asaruma es inflexion varonil de plural del sustantivo yasaru, (..) juicio: y se pone en esa inflexion porque los Padres cuyo es el juicio son masculinos. isucari n‘ipaquitosty atoñe. De ipaquitosty ya dixe arriba que es inflexion varonil, y tres veces varonil; porque es en su inflexion, en su moremento, y en su sustancia. Con todo eso, como su significado que es la madre, es femenino; por eso le precede el articulo ni femenino, y la preposicion isucari en su inflexion mugeril de singular, y se le sigue atoñe inflexion tambien mugeril de adjetivo ñatoñe, yo mismo, vel yo proprio. 50 FALKINGER

Este es todo el artificio dela clausula, ó periodo arriba puesto, en el qual se puede haber observado la algarabia de concordancias tan varias; la alternativa, conque en fuerza dela precisa colocacion delas palabras se van entreverando inflexiones, ò voces varoniles y mugeriles, masculinas, y femeninas, de singular y plual; segun pide el romance: y finalmente aquella (a nuestro modo de concebir) extravagancia de estar v.g. ñacunomoco vel ñacunumoc‘ unama, los escritos de los, en inflexion varonil; y el verbo omeezoro, que concuerde con ese nominativo, en mugeril: y en mugeril tambien el pronombre maniño, y el articulo ni, de ese nombre: y el adjetivo omixima de ese sustantivo: y la preposision yocuu de ese que es su caso: y el relativo ni.....ta....yocuu de ese que es su antecedente. Lo mismo digo del estar unama, que el articulo varonil de plural acompañando al nombre yaitaiquity, que está en inflexion de singular: y la preposicion mugeril de singular au rigiendo à su caso n‘asarusma, que es inflexion de plural varonil &. Ahora pues que comparacion tienen en dificultad las concordancias Latinas con estas? Quanto sudarian desnudos en tiempo de nieves para hacerse cargo, y à entrar en este laberinto, los que sudaron algo en tiempo de caniculares para entender y acertar las concordancias en genero dela lengua Latina? Quanto trabajaria un maestro de niños para labrar su entendimiento y acomodarlo à las escabrosidades de esta alternativa de concordancias (Ms-Camaño:46f).”

Como se ve en la tabla 6, se establece un sistema de referencias a dos niveles entre el hablante y el protagonista Pedro, ambos hombres, y las personas ó cosas relacionadas a Pedro. Era un dolor de cabeza

“aquel estar el hombre mezclando inflexiones varoniles y mugeriles segun las personas de quien habla, y la muger ò diciendo, ó negando, aquello mismo con inflexiones todas mugeriles. Estan resonando aún en el oido las voces varoniles, quando entran por el otro las mugeriles, y como las siento diversos me parecen de diverso significado, siendo del mismo; (Ms-Camaño:48)”

La dificultad para los jesuitas al escuchar por ejemplo a una pareja, era distinguir las dos formas que aún refiriéndose al mismo tema sonaban diferentes. Su conclusión:

“Tiene pues la Lengua Chiquita como peculiar suya la dificultad grande que consiste en la diversidad de lenguages; pues aún sabiendo un hombre hab(lar) y aún porque sabe, le cuesta entender à la muger..... (Ms-Camaño:48).”

3. Comentario final Con lo que yo he podido entender hasta ahora a través de textos chiquitanos actuales, pienso que las observaciones de los jesuitas acerca de la diferencia en el manejo de la lengua entre mujeres y hombres, siguen válidas aún hoy en día en cuanto al uso de los artículos, pronombres y relativos en el lenguaje de los hombres y también en el manejo LANGUAJE DE HOMBRES Y MUJERES EN CHIQUITANO 51 de los nombres posesivos y de los verbos en la 3a persona singular y plural, como se puede ver en textos actuales. Hombres distinguen obligatoriamente una parte de los nombres, semánticamente definidos “Dios, ángeles, hombres y demonios” como categoría aparte. En esta categoría entran también préstamos del español como “apóstoles, santos, cristianos”. Esta categoría se distingue por la morfología de lo que llamo “género común”. En si no se justifica hablar de un “género femenino”, porque en boca de mujer está incluido todo, también hombres y Dios.

Denotación: Categoría: Ejemplo: Dios, hombre, demonio masculino, divino tupax, noñeix Dios, hombre ↑ ↓ Mujeres, bestias femenino, animado paix, tamokox cosas inanimadas inanimado ka’ax mujer, perro, piedra

Parece que el hablante masculino establece alguna relación entre él y sus semejantes. Aquí entraría también el tema de los nombres de parentesco y los nombres de animales y árboles y otros más, que los hombres distinguen fonológicamente. No se tratan aquí porque no exigen concordancia morfosintáctica.

LA CARACTERÍSTICA REFERENTE AL SEXO DE LOS PARTICIPANTES EN LA COMUNICACIÓN: • El sexo del hablante, si es masculino, le da dos opciones gramaticales. El hombre usa ambas formas entreveradas, depende de la categoría del objeto. • El sexo y/o la pertenencia a una clase semántica del objeto determina la variación lingüística, si el sexo del hablante es masculino. Como hombre obligadamente debe usarse la variación masculina distinguiendo las formas gramaticales al hablar de Dios, hombres y demonios. • Si el sexo del hablante es femenino se usa una sola forma gramatical, sin distinción. Podemos decir que se trata de un ‘género común’ ó de ningún género. La versión femenina singular es la mas corta y no marcada. La tercera persona singular es la base de la forma masculina y sirve como raíz para la derivación.

Considerando estas observaciones propongo un cambio en la perspectiva: La versión femenina debería tomarse como forma básica en la gramática de esta lengua y la versión masculina como desviación de la misma. 52 FALKINGER

Appendix

Pers. 1.Conj. 2.Conj. 3.Conjugacion 4.Conjugacion 5.Conjugacion Ex.1 Ex.2 Ex.1 Ex.2 Ex.1 Ex.2 1.sg ego i....ca i.....ca ya....ca ña...ca yu....ca ñu....ca z....ta ....ca ix....ca y....ca 2.sg tu a....ca ai.....ca a.....ca a.....ca au....ca au....ca ...... ca ...... ca 3.sg varonil ....=ty i.....=ty ba....=ty ma..=ty au....=ty au...=ty ....=ty .....=ty ille u.....=ty 3.sg mugeril .....= i....= ba....= ma...= au...= au...= .....= .....= illa u...= 1.pl incl. o....ca oi.....ca ba....ca ma....ca ou...ca ou...ca oz...ca oñ...ca nos u....ca uz...ca uñ...ca 1.pl excl. zoi...ca zopi..ca zupa...ca zupa..ca zopu...ca zopu.ca zob...ca zom..ca nos zub...ca zum..ca 2. pl. Vos au..ca api...ca apa...ca apa...ca apu...ca apu...ca ab...ca am...ca 3.pl.varonil ...=ma i....=ma ba..=ma ma...ma au...ma au...ma ...=ma ....=ma illi u...ma 3. pl.mugeril bo....= yopi... upa...= upa...= opu...= opu...= ob...= om...= illae bu....= ub...= um...= mo...= mu...= Tabla 1: Las cinco conjugaciones de los verbos. Sirve para toda especie de ellos. (Ms-Camaño:33). Nota: que en el lugar, que ocupan en cada casilla los puntitos, se pone el radical del verbo simple, y el radical junto con las particulas de composicion del verbo compuesto: y en el lugar notado con dos rayitas en esta forma (=) se pone la nota, ó particula de tercera persona, que al verbo segun su qualidad le corresponde.

Nom. 1Sg ñy yo 1plin oñy nosotros 1plex zomy 2sg y/ hy tu 2 pl año vosotros 3sg -ti el 3 pl -ma ellos ø ella -iño ellas Dat. 1sg iñemo para mi 1plin oemo para nosotros 1plex Zoiñemo 2sg aemo para ti 2pl Aume para vosotros 3sg motii para el 3pl moma para ellos imo para ella ñome / noe para ellas Abl. 1sg Zobi por mi 1plin ozoi por nosotros 1plex Zoboi 2sg obi por ti 2 pl Aboi por vosotros 3sg oitii por el 3 pl oima / oisma por ellos obis por ella oboi / oboix por ellas Tabla 2: Los pronombres y relativos. Los pronombres primitivos. Los pronombres primitivos son los que se siguen. Todos tienen solos tres casos distintos: Nominativo, Dativo, y Ablativo. Los demas casos no se distinguen del Nominativo. (MsM-G:30ff).

LANGUAJE DE HOMBRES Y MUJERES EN CHIQUITANO 53

1. 2. 3. Declinacion 4. Declinacion 5.Declinacion 1. mio i- i- ya- ña- ixa- iyu- ñu- ixu- z- ñ-ixo- y-ñ- 2. tuyo a- ai- a- a- a- au- au- au- - - - 3. suyo i-sty i-sty ya-sty ña-sty ya-sty au- au-styu- yu-sty -sty -sty -sty de el styyu- sty styu- sty 3. suyo i-s i-s ya-s ña-s ya-s au-syu- au-su-s yu-s -s -s -s de su-s ella 1. pl.in. o-u- oi- ba- ma- ba- ou- ou- ou- oz- oñ- oz-uz- uz- uñ- 1. pl.ex. zoi- zopi- zupa- zupa- zupa- zopu- zopuu- zopu- zob- zom- zub- zoixa- zub- zum- 2. pl. au- api- apa- apa- apa- apu- apuu- apu- ab- am ab- 3. suyo i-sma i-sma yas-ma ñas-ma yas-ma au- au- yu-sma -sma -sma -sma de smayu- smau- ellos smau- sma sma 3. suyo yo- yopi-s upa- ñupa- yupa- yopu- ñopuu- yopu-s ob- om- ub-s de syu- syupa-s supa-s sopu-s sopuu-s sub-s sum-s ellas sno- sñu-s Tabla 3: Las declinaciones de la lengua chiquita. Pos. sg. / Pos.pl. que inflecten los Nombres, y se incluyen en sus inflexiones (Ms-Camaño:5/6).

Singular Plural

usado por hombre y mujer na, manu, manio baa, maniño

usado por hombre naqui, manuqui, maniqui, bama, unama, manuma Tabla 4: Articulos - demostrativos – relativos (MsP-21).

usado por hombre y mujer n‘ / ni Tabla 5: Articulo invariable en numero (MsP-21).

Tabla 6 LANGUAJE DE HOMBRES Y MUJERES EN CHIQUITANO 55

Bibliografía Manuscritos inéditos Anónimo s.f. Gramatica de la lengua De los Yndios Chiquitos pertenecientes al Govierno de Chuquisaca En el Reyno Del Perù Doctrinados por los PP de la Extinta Compañia de De la Provincia Del Paraguay. Ms. Modena (Microfilm) (MsM-G). Anónimo s.f. Bocavulario De la Lengua de los Yndios Llamados Chiquitos. Escrita por un Misionero De la Compañia De Jesus. Ms. Modena (Microfilm) (MsMV). Anónimo 1718 Arte de la Lengua Chiquita. San Javier. Ms. Paris, Bibliothèque Nationale, Département des Manuscrits. Americain 19 (Microfilm) (MsP-19). Anónimo 1718? Bocabulario de la Lengua de los Chiquitos. San Javier. Ms. Paris, Bibliothèque Nationale, Département des Manuscrits. Americain 20 (Microfilm) (MsP-20). Anónimo s.f. Vocabulario De La Lengua chiquita Parte 2a. Chiquito-Español Del Pueblo San Xavier. Parte 3.a de los Raizes. Ms. Paris, Bibliothèque Nationale Département des Manuscrits. Americain 21 (Microfilm) (MsP-21). Camaño, Joaquin, SJ s.f. De la lengua Chiquita. Ms. Biblioteca Jagiellonska, Krakovia.

Referencias bibliográficas Corbett, Greville 1991 Gender, Cambridge: Cambridge University Press. Furlong, Guillermo 1946 Los Jesuitas y la Cultura Rioplatense, Buenos Aires. Hockett, C.F. 1958 A Course in modern Linguistics, New York: Macmillan. Jespersen, Otto 1998 ‘The Woman’, [1922], en: Deborah Cameron (ed.), The Feminist Critique of Language: A Reader, London/New York: Routledge. Key, Mary Ritchie 1975 Male/female language: With a comprehensive bibliography, Metuchen, N.J.: Scarecrow, [2nd edn. 1996 Lanham MD: Scarecrow Press]. Sapir, Edward 1990 ‘Male and Female Forms of Speech in Yana’, [1929], en: William Bright (ed.), The collected Works of Edward Sapir V. American Indian Languages 1, Berlin/New York: Mouton de Gruyter. 1991 ‘Abnormal types of speech in Nootka’, [1915], en: Victor Golla (ed.), The collected Works of Edward Sapir VI. American Indian Languages 2. Berlin/New York: Mouton de Gruyter.

MULTILINGUAL LITERACIES IN BOLIVIA

Utta von Gleich University of Hamburg

1. Introduction This paper offers some insights in our project “Multilingual literacies from a cross- cultural perspective”1, the everyday uses of literacy among bilinguals in Bolivia and Uganda focused from the perspective of the user. Due to the limited space available and the focus of the Seminar on Indigenous Bolivian and Rondónian Languages, held in Leiden, September 2000, we will concentrate on Bolivia. Our main interest is neither the acquisition of literacy in school settings nor the historical development of predominantly oral languages into written languages, but the importance and impact of literacy for the regular bi- or multilingual indigenous citizen in Bolivia. Literacy is seen not as a set of independent skills associated with reading and writing, but as the application of those particular skills for specific pur- poses in specific contexts, a socio-cultural practice. We should also emphasise that the analysis of multimedia communication events (oral and written) are gaining importance due to the cultural changes produced by global communication flows. Tannen (1987: 85) already stressed this fact in her seminal contribution “The Orality of Literature and the Literacy of Conversation” when she stated that “orality and literacy, speaking and writing, are not dichotomous but rather complex, overlapping and intertwined”. Von Gleich (2000) has elaborated on this medial development in the Aimara-Spanish social communication and called this type of literacy the ‘literacidad oculta’ (hidden or covert literacy) in the so-called new global orality. In what follows we will briefly present the outline of the project, the methodol- ogy developed for the empirical work and some evaluation procedures as well as some preliminary results obtained in Bolivia, concentrating on the Quechua-Spanish bilinguals in the department of Cochabamba.

1.1. Design of the project The project analyses literacy practices, i.e. “the social practices associated with the written word” (Barton 1994: 37) in multilingual contexts. In its first phase, the pro- ject focuses on those multilingual communities where languages differ significantly with regard to their social and functional recognition, their degree of standardisation and their literary tradition, as it is the case in large areas of Africa and Latin America.

1 The project was designed between 1998 and 1999, submitted for evaluation in January 1999 to the DFG (German Science Foundation) and approved in July 1999 as one of the projects in the package of the Son- derforschungsbereich 538, granted to the University of Hamburg under the direction of Prof. J. Meisel. 58 VON GLEICH

Due to the increased promotion of literacy in indigenous languages in countries of these two continents during the last two decades, these languages are expected to compete with European languages as written media. Our explorative study tries to find out whether literacy in more than one lan- guage might achieve importance in the near future in everyday life situations beyond the education system and beyond specialised professional groups such as interpreters, and to explore the nature of such new functions. On the basis and against the background of the “New Literacy Studies” ap- proach (e.g. Street 1982; Barton 1994; Prinsloo & Breier 1996; Hamel 1996; Horn- berger 1997; Hornberger & Skilton Sylvester 2000) the project analyses the use of literacy in various multilingual regions of Africa (first phase: Uganda) and in Latin America (first phase: Bolivia) applying qualitative and quantitative methods (obser- vation, interviews, questionnaires, biographical narration) in various communicative settings, family, work, church, public life, and administration among multilingual individuals and groups. The leading questions focus on:

1. Forms and ways of multilingual practices; 2. Factors that enhance or restrict the expansion and diffusion of specific literacy events; 3. Costs and benefits of multilingual literacy for individuals, groups and socie- ties.

The present and changing ecology (cf. Haugen 1972) of multilingual literacy will be investigated in case studies of communities selected according to their specific lin- guistic constellations including varying degrees of dominance in written and oral use. This user-oriented design will be complemented by studies of the selected popu- lation (groups and individuals), elaborating on attitudes towards literacy per se, own biographical experiences, and opinions about people who can be considered model speakers, always taking into account the reality of the respective setting, i.e. access to printed material, language policy, linguistic distribution, number of speakers, types of multilingualism, literacy tradition of the languages involved, etc. The sampling in Bolivia is based on the principles of the community profile method (Wölck 1975; von Gleich 1982; Moelleken 1997), in Uganda preferably on subsequent clustering as well as on elements of the communicative network ap- proach.

2. Methodology As already mentioned a package of different methods for the collection of the data is being used. Before going into the details it should be pointed out that the common methodology serves as a linking tool, a medium of comparison (tertium compara- tionis) between the two selected countries. MULTILINGUAL LITERACIES IN BOLIVIA 59

Let us begin with some explanations about the selection of the two countries or con- tinents. The two main researchers, Mechthild Reh and Utta von Gleich, have been working during the last twenty years in different sociolinguistic contexts of Africa and Latin America, not only as descriptive or contact linguists, but both being in- volved on various occasions in counseling governments on how to standardise local languages for schooling purposes, especially the introduction of mother tongue edu- cation in the framework of bilingual intercultural education and also in informal edu- cation system and in development projects. Moreover, our personal commitment to the protection and maintenance of vulnerable languages made us curious to find out what situations vernacular languages are facing when competing with the major writ- ten languages of the ex-colonial world, English, French and Spanish. Despite their geographic distance and many obvious differences Bolivia and Uganda share some socio-historical features:

- colonial administration - imported educational systems from Europe - high degree of illiteracy (Bolivia 1999: 20%, Uganda 47%) - a great variety, (number, size and linguistic properties) in up to forty indige- nous languages concentrated in different regions - an exogenous official language: Spanish and English - a sporadic experience of mother tongue literacy during the last 25 years in form of pilot projects - a high degree of poverty, both countries belong to the poorest of the respective continent (60-70% of the population below poverty line) - a very young population (more than 50% under the age of 20) - a big interest in development and modernisation along with a high degree of ethnic and linguistic loyalty claiming mother tongue education and civil rights - educational reforms are under way starting in the nineties and promoting bi- lingual education - economies based on primary sectors, mining, agriculture and trade in Bolivia, agriculture and trade in Uganda - rapidly growing urbanisation

Summarising, we can state that the everyday living conditions, on the micro and meso level and the socio-linguistic features look quite similar. However, comparing the societies of the two countries as a whole we find sig- nificant differences: Uganda’s population is almost 90% of African origin and only up to 10% are immigrants from Europe or other countries, whereas Bolivia is split up roughly into three different ethnic groups: the descendants of the original populations 42% (los pueblos originarios), the mixed Mestizo population, almost 30% and the so called white population 15%, descendants of the Spanish colonialists and the rest are diverse smaller immigrants groups. 60 VON GLEICH

The significance of bi- or multilingualism is similar at least in some regions, but there is an important difference in the language profile between the two countries: while in Bolivia only about 11% (Albó 1995: 23, on the basis of the census of 1992) are monolingual mother tongue speakers of American indigenous languages and 40% monolingual speakers of Spanish (first language = L1), the majority of speakers of an indigenous languages, almost 60%, are bilingual (indigenous mother tongue and Spanish, but increasingly Spanish as first language (L1) and an indigenous language as second language (L2)2; in Uganda almost 90% have an indigenous language as their mother tongue, being at the same time bilingual in another African language, and/or trilingual in English. Only a small portion shows English as a first language. The overall degree of individual multilingualism with normally three languages is higher in Uganda than in Bolivia where Spanish appears as a second language in ad- dition to a language of indigenous origin. In the case of Bolivia our selection of sub-samples according to certain charac- teristics of school level and individual multilingualism was facilitated by the excel- lent study of Albó (1995), a combined demographic-administrative and multi-lingual census prepared for the implementation of the Education Reform that is expected to extend intercultural bilingual primary education for all. The addressees of our study on reading and writing habits in literacy events are bilingual persons (indigenous languages + Spanish) with sufficient schooling (pri- mary completed). Monolingual illiterates are of course not in the centre of our inter- est; nevertheless their way of dealing with written elements in everyday life situa- tions will be observed in order to complete the study. Socially our informants belong to low-income groups and most of them have to be considered as poor. Up to the present only very few members of the indigenous population have obtained higher university degrees (there are some professionals in agriculture, some lawyers and mainly teachers). Therefore our study is limited to the socially under- privileged lower and middle class. An outstanding exception is the former vice- , a former teacher and lawyer, Victor Hugo Cárdenas, who en- forced the concept of interculturalism as a democratic attitude in a multiethnic soci- ety. Thanks to the help of our Bolivian colleagues in preparing the selection of our samples, we decided to work in three comparable smaller cities that show representa- tive situations for the three chosen languages:

- Aimara in Viacha (close to El Alto with 500,000 inhabitants the large satel- lite community next to the capital of La Paz), - Quechua in Sacaba near Cochabamba, and

2 The new national census of 2001 to be carried out in September will specify this question by more precise questions. The first question remains “¿ Qué idiomas o lenguas habla? (Which languages do you speak?), but will be complemented by ¿ En cuál de estas lenguas aprendió a hablar? (In which language did you learn to speak?). MULTILINGUAL LITERACIES IN BOLIVIA 61

- Guarani with permanent residents in Camiri in the Cordillera, and in various migrant settlements in Santa Cruz.

The original idea to work in El Alto or respectively in Cochabamba was given up due to the personal and financial limitations of the project to cover a representative por- tion of these huge populations and also because it was not easy to define a coherent social communication network that is needed for the application of an anthropologi- cally orientated community profile. In Viacha, Sacaba and Camiri, however, these conditions can be found, as will be shown later. As to the three main questions that guided the first phase of our joint research, it has to be emphasised that the first question: forms and ways of multilingual practices has received priority treatment in the first year. For the purpose of documentation of multilingual literacies we use two basic concepts: literacy practices and literacy events as defined by Barton (1994: 36-37) within the context of literacy:

“Literacy is a social activity and can be described in terms of people’s literacy practices which they draw upon in literacy events; the latter are all sorts of activi- ties in everyday life where the written word has a role. The second term that is use- ful is that of literacy practices. There are common patterns in using reading and writing in any situation and people bring their cultural knowledge to an activity. We refer to their ways of using literacy as literacy practices.”

Applying anthropological criteria for the selection of the mayor communicative situa- tions, we developed a common profile of potential everyday communicative lite-racy events structured in so-called domains (Annex 1). This type of communicative profile has been widely applied over the last 25 years in studies about language use in bilin- gual communities (Wölck 1975; von Gleich 1982; Moelleken 1997; and others) where full fledged data collection was either not possible or recommendable, as well as in literacy studies (Wagner 1993) in order to compartmentalise the universe in smaller relevant social units. The tentative list of domains in which to look for literacy events was discussed with our local colleagues. In the case of Uganda some minor adjustments were intro- duced later which albeit will not diminish the comparability with the Bolivian case. In order to systematise our data we developed observation sheets; one for the description of the visual environment (Annex 2) as we assume cultural differences in this field and impacts on the reading motivation; a second one for the recording and the categorisation of the literacy events (Annex 3) and a third (Annex 4) showing some essential data about the most active participants in the event that might be in- terviewed afterwards. This means that, wherever it was possible we added an infor- mal oral or tape-recorded interview with at least one participant of the event and the 62 VON GLEICH provider of the information in order to develop a coherent impression of the literacy event. A guideline containing a list of key questions was given to the interviewers to help them structure the interviews (von Gleich, Reh, and Glanz 2000). In addition to this systematic collection of information we had conversations with key persons about the values of writing and reading and observed carefully the ongoing public discussion on multilingualism and interculturalism in both countries. Key or focus persons were personalities who have some influence on the reading and writing cul- ture in the respective country (writers, politicians, professional, academics and espe- cially ethnic authorities). In the case of Bolivia we organised with the help of the local Goethe Institut a public roundtable discussion on ‘Aimara en la comunicación social’ where the local Aimara leadership as well as the professional academics ex- pressed and discussed their points of view. The second mayor question concerning factors that enhance or restrict the ex- pansion and diffusion of specific literacy practices as well as respective attitudes, beliefs, that have an impact on the peoples literacy practices will be studied and sys- tematised at a further stage of the project, complemented by tests to determine ver- nacular criteria of the so called perfect bilingual. Finally we want to mention that we made some videotapes about specific liter- acy events. For the sub-sample of Quechua bilinguals in Cochabamba we were fortu- nate to videotape a comprehensive interview (approx. 60 minutes) about the editing procedures of a bilingual newspaper that has a tradition of 15 years, and secondly we videotaped a working session with Quechua monolingual writers (also 60 minutes). For the Aimara sub-sample we have a video on the distribution of certificates by the Mallku (highest authority) as well as a video on a seminar about literacy among Ai- mara-Spanish bilinguals (60 minutes), and a wide range of snapshots of literacy events.

3. Biliteracy among Quechua-Spanish bilinguals 3.1. Short socio-economic profile of Sacaba Sacaba is a middle-sized town of about 36,000 inhabitants. It is situated at an altitude of 2,600m in a very fertile, well-irrigated valley with a mild climate, at 10 km from Cochabamba, the capital of the Department of Cochabamba. The administrative status as municipality, according to the new Constitution of 1994, has given consi- derable autonomy to the Council and the mayor of the town and it has been rated as one of the priority regions by the general developmental plan of the Republic (CORDECO). The educational level is still modest but improving: 30% of the population have accomplished basic education, 20% secondary school, only 2% teacher training seminars, 4% have some sort of University degree, and a medium of 10% are illiter- ate, a percentage which of course is higher in the local rural hinterland. MULTILINGUAL LITERACIES IN BOLIVIA 63

The linguistic composition is the following: of the population above 6 years 77% speak Spanish, 61% speak Quechua, and 4% Aimara, which indicates the high degree of bilingual Quechua speakers. More than half of the population is under 20 years. In the last years Sacaba has begun to benefit from the decentralisation law and the law of social participation. A major number of development projects for basic needs were established in order to attend the rural hinterland. In the urban areas the basic supply of water, electricity, health service, and schools is guaranteed but there is still a considerable deficit in rural areas. Cash incomes are modest, most people still supplement their budgets by agricul- tural and cattle raising activities. There is no predominant economic activity; instead we find people working in small commerce, administration, transportation and all kinds of handicraft activities. The three mayor markets, the Central market, the potato market (papa qancha) and the cereals market, which operate mostly during week- ends, are important elements for the local economy. Summarising, we could say the city of Sacaba is economically self-maintaining, although it is into-grated in the big- ger network of Cochabamba. This applies especially to the information and education network; newspapers come from Cochabamba, TV channels and also radio emis- sions, especially in Quechua for the farmers, are produced in Cochabamba. Due to the proximity, secondary and higher education is attended in Cochabamba. This diversity of local urban activities tied to agricultural activities in the rural districts nearby allowed us to identify a rather complex scenario of communications domains with pertinent literacy events. The pilot study was presented to the local authorities and welcomed.

3.2. Some results and conclusions 3.2.1. First question: visual environment, forms and ways of multilingual events and practices According to the notes in the observation sheets (Annex 2), the visual environment is almost monolingual in Spanish and includes commercial and professional propa- ganda, posters and slogans in all domains. Nevertheless, some elements such as proper names, names of streets, menus in restaurants, etc., integrate Quechua terms; in many cases they are completely hispanizised so that the normal user is not aware of the loan character. The written elements are stickers, flyers, price lists, selling ad- vertisements, professional propaganda, newspapers, books, postcards, and pur-chase conditions in shops and banks. Forms and ways of multilingual events have been observed according to a community profile of Sacaba and recorded in the observation sheets (Annex 3). The predominant configuration pattern of a literacy event is a medium-sized group in public and private institutions; this implies different clients waiting in line or sitting around, in general more users than providers, which induces automatically different flows of conversation. In the case of medical attention the groups are smaller and the central conversations are not public; one patient and two-three rela- 64 VON GLEICH tives vis-à-vis the doctor, accompanied by a nurse or other personal assistant. Almost all literacy transactions in private and public institutions are semi-public. The written elements used in the performance of the literacy event are letters of complaint, forms, documents, instructions, prescriptions, protocols, notes, bills, re- ceipts, etc., typical for the respective activity. Culturally they all belong to the written communication practices of the dominant Hispanic society. The language used in these written documents is mostly Spanish. The culturally different back-ground and the individual educational level are the main reasons for communication difficulties and not only the linguistic barrier as commonly suggested. We will illustrate this phenomenon on page 10 on the basis of the conversation between a doctor and a pa- tient. The domains that differ from this pattern are domain 5 (churches and ceremo- nies), domain 12 (mass media), and domain 9 (educational sector). In the religious domain, especially in protestant churches, the is available and read and explained in Quechua during the ceremony; books for the prayer and singing books, and the religious education recommendations are also in Quechua, but very few books are bilingual. In domain 12, the mass media, messages for the radio are, for instance, delivered in Quechua (afterwards promulgated in Spanish or Quechua according to the wish of the client). This also applies to the bilingual magazine Nawpaqman, in which half of the texts are in Quechua, some originally edited in Quechua, some translations or short versions of information relevant to the readers (clients). In the case of Naw- paqman we clearly see a symbiosis of the indigenous cultural contents shaped into a written form according to occidental norms (letters, protocols of meeting, list of top- ics of an discussion agenda, list of participants etc.). Since July 2000, Presencia a leading daily, privately financed newspaper (daily edition of 500,000), includes every day one full page with news in Quechua and Ai- mara. This is a surprisingly positive and very rare initiative. It is a clear proof for the linguistic flexibility and creative innovative capacity of indigenous languages, which means new cultural concepts and everyday events, political, societal, cultural and private topics that can be communicated in Quechua and also be written down and understood by the readers. Thus the story of the sunken Russian submarine or the law suit against general Pinochet in Chile and many other international as well as national topics were dealt with in this section. We still do not have a systematic analysis of the new readers but the fact that since July 2000 the journal is practically sold out one hour after its delivery, as well as many comments of the vendors support the specula- tion that the supplement in indigenous languages is attractive. In the educational sector (domain 9), all primary schools that belong to the trans- formation plan of the Educational Reform have sufficient teaching material in Quechua, in the classrooms, in the reading corner, in the library, as well as in the recreation areas. Furthermore, it is interesting to note that in the city of Cochabamba we find some secondary schools where Quechua is systematically taught as a second or foreign language, such as, for instance, at the schools of ‘Fe y Alegría’, at El Co- MULTILINGUAL LITERACIES IN BOLIVIA 65 legio San Rafael, or as a foreign language, at the Instituto Maryknoll, IDELCO, and other schools. Even at the local university Quechua is obligatory in the respective curricula for medicine, law, agricultural studies and linguistics, but the imple- mentation should be handled more strictly. For the first time, since August 2000, the School of Education of the University San Simon offers during weekends and holidays a ‘Licenciatura especial en Educa- ción Bilingüe Intercultural’ for teachers in service. The informal education institu- tions offered mostly by the non-governmental organisations (domain 14) are more inclined to communicate bilingually, in the oral and written mode, especially in adult literacy and professional training. The above mentioned domains (religion and educa- tion) tend to offer literacy events to a larger audience, open to everyone as in the case of churches, but in pre-established groups or on invitation (parents-teacher meeting, members of a club and so on) in the case of training and schooling. Summarising, we can state that the oral use of Quechua in private and public institutions is still alive, but that bilingual oral communication of course is pre- dominant. The written elements are mostly Spanish so that we could speak in terms of Fishman (1975) of a diglossia between written and oral modes of communication. But we should be aware that such a type of conflictive diglossia is still widely toler- ated as people in the street still take for granted the functional division: you talk in Quechua but you read and write (study) in Spanish. In some experimental interviews of on-air callers to radio stations in La Paz and Cochabamba we found out that most people still believe in this stereotyped prejudice. As a consequence of the educational reform only recently a process of awareness and questioning of these stereotyped beliefs is taking place publicly, initiated by the in- digenous speakers and their leading personalities. The Hispanic society’s inter- cultural behaviour, therefore, is increasingly challenged to change.

3.2.2. Communication flows in Sacaba If we look closer at a sequence of communicative acts among the local population in different situations –public and private institutions–, we can isolate six major linguis- tic communication patterns3 that may easily co-occur or intertwine in a single literacy event. As an illustration we present the following example: a medical consultation of a pregnant woman. The consultation of a bilingual family in a health centre, which I (Interviewer/ Observer 2: German/Spanish, Quechua) observed together with my local research assistant, a Quechua-Spanish bilingual teacher-trainer (=interviewer1), showed a rather complex communication configuration). The medical staff consisted of one administrative female assistant, a nurse (Quechua+Spanish) serving as a translator, one doctor and one assistant doctor (Spanish only), one technical assistant

3 Among our speakers we can find different types of bilingualism, those who have Quechua as a first language (L1) and Spanish as a second language (L2) or Spanish as first language and Quechua as L2. Therefore we indicate the sequence of acquisition in brackets. Simultaneous bilinguals are very rare among indigenous speakers in the Andean countries. 66 VON GLEICH

(Quechua+Spanish). The clients were the young mother (Quechua+Spanish), her husband (Quechua+Spanish) served as a translator and knowledge-mediator, and two children, 4 and 6 years old, respectively (Quechua only). The six major communication patterns are the following:

1. Spanish among Spanish monolinguals (doctor and assistant doctor); 2. Spanish between bilingual Spanish speakers and monolingual Spanish speakers (doctors and nurse); 3. Spanish among bilinguals with obvious linguistic variations, style variation or register according to the social class, e.g. the nurse talking to a bilingual peasant as compared to her colleague; 4. Bilinguals speaking with monolinguals in indigenous languages (nurse with children); 5. Speakers of the same indigenous language (children with mother, but also bilingual mother with our bilingual observer); 6. Foreign Spanish bilingual with local mono- and bilinguals (foreign observer in Spanish to the clinic staff and Quechua with the woman and children).

On this basis the development of a complex multilingual communicative event could be observed, which we will split up in episodes representing typical activity units, some with a written element.

Communicative bilingual episodes (E) in a routine medical consultancy:

E Participants Function Written element Languages Language used orally used 1a Interviewer 1 Introduction of visitors Personal card Sp.+Quechua Interviewer 2 to administration Off. letter of Spanish recommendation (Spanish) 1b Int.1 and Int.2 Introduction to the Spanish doctor 1c Client family Registration for Sequence coupon Spanish With administration consultation no. 1d Interviewers + clients Self-introduction to the client family Int.1 to female client Quechua Int.2 to male client Spanish 1e Female client + Explanation of the visit Quechua children Nurse to children Providing confidence Quechua 1f Male client to Interv.2 Explanation of the visit Spanish MULTILINGUAL LITERACIES IN BOLIVIA 67

2 Clients, doctors, nurses Routine consultation Folder (in Spanish) and interviewers Part 1 Diagnosis with data about Doctor to client Short interpretation of the patient Spanish situation Nurse to female client Translates into Quechua Female client to doctor doctor’s Spanish talk Husband to doctor Questions Quechua and Doctor to client Husband translates Spanish Nurse to client More questions into Quechua Female client Translates questions into Quechua Gives further informa- Quechua tion Nurse to Doctor Translates answers into Spanish Doctor to female client

Nurse to female client Part 2 Suggestion for treat- Demonstration ment/therapy cards (Spanish) Spanish Doctor to client (Caesarean section) Nurse to client translates into Quechua

Part 3 Doctor to clients Physical exam Ultrasound pictures Spanish Translates during exam into Quechua Nurse to clients Part. 4 Appointment for the Spanish last medical check Results of the exam into Quechua Translation of results 3a Clients to doctors and Thanks and goodbye Spanish nurse 3b Clients to administra- Appointment for the Writing down in Spanish/ tion last exam the calendar & Quechua little note to the client (Sp.) 4 Conversation between Additional information Spanish doctors & interviewers 5 Interviewers to nurse Thanks and new Leaving a note Spanish and appointment with the telephone Quechua number of Int.1 68 VON GLEICH

E1: Presentation between the actors 1a The presentation of the observers to the clinic administration starts with two writ- ten elements personal card and a recommendation letter from a professor of the Fac- ulty of Education of the University (in Spanish). 1b Followed by a brief presentation to the doctor (he reads the letter and the card and then talks to us in Spanish). 1c Registration. A family enters the waiting room and picks up a numbered coupon for the consultation. The local observer uses this opportunity to start a conversation in Spanish and Quechua with the family waiting for attendance (presentation of our research about multilingual language usage). 1d Self-introduction. Female patient explains her case in Quechua to the Quechua speaking interviewer; parallel to this the man talks in Spanish to the foreign ob- server/interviewer. 1e Instruction of the mother to the children in Quechua; the nurse also talks to them in Quechua, because she identifies them as Quechua mother tongue speakers (how: the pregnant woman wears traditional clothes and the children speak to each other in Quechua. 1f The husband is demonstrating his communicative competence in Spanish, speak- ing in a local variety to the foreign observer.

E2: Consultancy with the doctor Conversation with the doctors in the presence of the personal assistant and the inter- viewers/observers. Part 1: The doctor asks for the reason of the visit, reading silently in the folder the notes of previous visits and immediately asks the bilingual nurse to repeat what he had said (information) in Quechua to the woman for better and safer understanding. Then he asks again in Spanish for further information, the kind of problems she has or feels. The woman starts in Spanish, then switches into Quechua, looks at her hus- band who starts to translate and sometimes the nurse intervenes repeating with addi- tional explanations the information given by the woman and her husband (completing the previous knowledge about the patient’s condition). Part 2: The doctor tries to explain to the woman that they might have to perform a caesarean section due to overtime of the pregnancy. In this context he uses a set of demonstration cards and explains how the baby will be born/freed. Understandably this frightens the woman. She reacts shocked and silenced, a big susto, while the nurse is trying to give her confidence by giving a long explanation in Quechua. Part 3: Medical exam. Next step: the doctor tries to explain that he wants to make an ultrasound diagnosis in order to know the exact position and size of the baby. He shows two different ultrasound pictures and tries to explain to the woman that the baby has to turn head down for a normal birth. Again a long repetition follows in Quechua while the doctor also tries to calm the husband in Spanish. MULTILINGUAL LITERACIES IN BOLIVIA 69

Part 4: The medical check is done in another room. We stay with the husband and the children. E3a They comes back and the conversation ends with a long explanation to the hus- band in Quechua by the nurse and in Spanish by the doctor. The family leaves with respectful thanks to the doctor and an appointment for a last check within five days. During the whole consultation the doctor talks also to the foreign observer mention- ing the problem of translation into Quechua, but with no negative attitude. He regrets not to be able to speak Quechua, but fortunately he always finds a bilingual nurse. E3b The family expresses the gratitude for the good attendance and leaves. E4 We continue the conversation with the doctor who preferred not to be taped. He admits the necessity of written bilingual instruction material but points also to the fact of the high rate of illiterates and rather recommends visually understandable in- structions as he was using. E5 The interviewers express their gratitude to the nurse and administrative assistant and leave also.

The evolution of this conversation fits with some modification easily into the institu- tional pattern analysis of doctors and patients conversation, observed by Bührig, Dur- lanik and Meyer (2000), the so-called praxeogramm. It is culturally very similar due to the fact that both observations have their roots in an occidentalized classical clinic treatment. Compared to the clearly structured obligatory medical explanation talk in Germany (Meyer 2000), the Bolivian communicative practice of the doctor focuses directly on co-operation with the patient, overcoming fear and consolidating confi- dence. The explanations given by the doctor are therefore integrated in the part diag- nose and therapy proposal. The linguistic situation differs also: in Bolivia the bilin- gual nurse is the major communicative actor whereas in the German situation a fam- ily member or friend has to fulfil the interpretative job, i.e. in the case of the Bolivia the institution takes care of the interpretation and linguistic mediation, whereas in Germany the foreign client is on its own. In the case of Bolivia we can of course also identify an ethnic midwife model that is still in use in the countryside with a predominant oral communication in the local indigenous language, but in the case of feared complications bilingual clients nowadays approach with more confidence the institutions of the Hispanic culture. Interesting in our case is the continuous usage of both languages in changing speakers’ constellations and the interference of written elements in Spanish that are the origin of the vast explanation and translation procedures. Seen quantitatively the reduced amount of written elements in Spanish provokes a time consuming procedure of interpretation. But we have to be aware that the reason behind that is not just the lack of the translated instruction but the unfamiliarity with different cultural medical procedures and their understanding.

70 VON GLEICH

4. Conclusions Literacy practices among Quechua Spanish bilingual speakers in everyday communi- cation are highly dominated by the . All written elements, as shown in the example of medical routine consultation, are in Spanish. This is also the case in almost all other private and public institutions (cf. list of domains). The asymmetric relationship between Spanish and indigenous languages is obvious and confirms the assumption of a diglossic situation in terms of Fishman. Non-Spanish mother tongue speakers in Bolivia have to learn not only to speak but also to read and write in Span- ish if they want to have access to minimal basic services and civic rights. Although Art. 1 of the Constitution reads “Bolivia, libre, independiente, soberana, multiétnica y pluricultural, constituida en República unitaria, adopta para su gobierno la forma democrática representativa, fundada en la unión y la solidaridad de todos los bolivi- anos”, we miss the solidarity on the linguistic level. The educational sector is the only domain of the state where we can recognize serious efforts to democratise the unequal linguistic situation through the implemen- tation of the bilingual intercultural education model for all Bolivians. If we review the different oral monolingual and bilingual communication flows presented above, we can easily verify their combined and rich appearance in every- day multilingual literacy events and practices. This means that as long as Bolivia has such great numbers of bilingual speakers, biliteracy in the community and the society in general should be given a broader chance to become reality.

References Albó, Xavier 1995 Bolivia Plurilingüe: Una Guía para Planificadores y Educadores, 3 vols. + maps, La Paz (Bolivia): CIPCA (Centro de Investigación y Promoción del Campesinado) and UNICEF. Barton, D. 1994 Literacy. An Introduction to the Ecology of Written Language, Oxford: Blackwell. Bührig Kristin, Latif Durlanik and Bernd Meyer (Hrsg.) 2000 Dolmetschen und Übersetzen in medizinischen Institutionen, Arbeiten zur Mehrspachigkeit, No. 9 SFB 538, Universität Hamburg. Fishman, Joshua A. 1975 Soziologie der Sprache, Munich: Hueber Verlag. Gleich, Utta von 1982 ‘Soziale und kommunikative Bedeutung des Quechua und Spanischen bei Zweisprachigen in Peru (1968-1978)’, Dissertation, Universität Hamburg. 1987 Educación Primaria Bilingüe Intercultural en América Latina, Bd. 214, Schriftenreihe der GTZ, 2 edition 1989, Rossdorf . 2000 ‘Comunicacion y Literacidades entre Aymara-Castellano Bilingües’, Revista Lengua 10: 135-80, La Paz. MULTILINGUAL LITERACIES IN BOLIVIA 71

Gleich, Utta von, Mechthild Reh and Christine Glanz 2000 Mehrsprachige literale Praktiken im Kulturvergleich (Uganda/Boli-vien). Ausschnitte aus den Erhebungsmethoden, in: Bernd Meyer and Notis Toufexis (eds.), Text/Diskurs, Oralität/Literalität unter dem Aspekt mehr-sprachiger Kommunikation. Beiträge zum Workshop ‘Methodologie und Datenanalyse, Arbeiten zur Mehrsprachigkeit, Folge B (Nr. 11), Universität Hamburg. Hamel, Reiner Enrique 1996 ‘The inroads of literacy in the Hñáhñú communities in Central Mexico’, In- ternational Journal of the Sociology of Language: Vernacular Literacy in Nonmainstream Communities, 13-42. Haugen, Einar 1972 The Ecology of Language, Palo Alto, California: Stanford University Press. Hornberger, Nancy 1997 ‘ from the Bottom up’, in: Hornberger (ed.), Indigenous Literacies in the . Language Planning from the Bottom up, Ber- lin/New York: Mouton de Gruyter, 357-66. Hornberger, Nancy H. and Ellen Skilton Sylvester 2000 ‘Revisiting the Continua of Biliteracy, International Critical Perspectives’, Language and Education 14/2: 96-121. Meyer, Bernd 2000 Medizinische Aufklärungsgespräche. Struktur und Zwecksetzung aus diskurs- analytischer Sicht, Arbeiten zur Mehrsprachigkeit, AZB, Serie B, No. 8, Hamburg: Sonderforschungsbereich Mehrsprachigkeit. Moelleken Wolfgang 1997 ‘Kontaktlinguistische Forschung bei den mexikanischen Mennoniten’, in: Wolfgang Moelleken and Peter Weber (Hg.), Neue Forschungs-arbeiten zur Kontakt-linguistik, Plurilingue XIX, Dümmler, 365 – 376. Prinsloo, M. and M. Breier 1996 The Social Uses of Literacy. Theory and Practice in Contemporary South Africa, Amsterdam/Philadelphia: John Benjamins. Street, Brian 1982 Literacy in Theory and Practice, Cambridge: Cambridge University Press. Tannen, Deborah 1987 ‘The Orality of Literature and the Literacy of Conversation’, in: J. Langer (ed.), Language, literacy and culture, Norwood, N.J.: Ablex, 67-88. Wagner, Daniel A. 1993 Literacy, Culture, & Development. Becoming Literate in Morocco, Cam- bridge: Cambridge University Press. Wölck, Wolfgang 1975 ‘Community Profiles. An alternative to linguistic informant selection’, Inter- national Journal of the Sociology of Language 9: 43-57.

72 VON GLEICH

ANNEX 1 List of domains Corresponding to a tentative community profile of literacy events, discussed with colleagues in Bolivia, October-November 1999

Domain 1: Outdoor, (in the street, places, daily and weekly markets etc) Domain 2: Private transportation: taxis, (colectivos, micros, buses and trains) Domain 3: Health Centres Domain 4 Small shops, retailers and supermarkets Domain 5: Churches, religious and cultural centres Domain 6: Leisure-activities (sports, music, dancing, cinema and theatre) Domain 7: Cultural Ceremonies (birth, marriages, burials, funeral services, cele- brations, festivities) Domain 8: At home Domain 9: Education Sector (primary schools, non formal Education institutions, Public Education Administration dependencies) Domain 10: Public institutions and relevant basic services (registration and coun- seling, energy and housing supply) Domain 11: Private institutions (credit and finance, working place) Domain 12: Mass media (radio, films, TV, Newspapers + magazines, books, video-cassettes) Domain 13: Professional associations or organisations (taxi drivers, teachers + workers’ unions, etc.) Domain 14: Non-governmental social institutions and organisations (NGO’s) Domain 15: International organisations, UN-Dependencies, UNESCO, UNICEF, Human rights organisations, ILO etc.) MULTILINGUAL LITERACIES IN BOLIVIA 73

ANNEX 2 (example) Observation sheet for the documentation of Multilingual Literacy in the “Visual En- vironment”:

Observations

Observer: Julieta Date: 17/12 1999 in Sacaba

DOMAIN 3: Health Centres (clinic, doctors...) (Give details of type of location, signs and languages used)

Location: in the eastern area on the main road, Cochabamba to Santa Cruz On the building (outside): Hospital Mexico From far away a big visible announcement at the entrance of the building

In the building (inside) waiting area: Specifiers on the doors: Laboratory, x-rays, vaccinations A list of all doctors and their services, attendance hours A list of prices for the different services (birth attendance: free!) Walls full of information about health campaigns, preventive health programs, vacci- nation campaigns Some newspapers A television corner An altar for the local saint: San Juan de Dios

Reception desk: Distribution of numbered vouchers for attention Folders, entries in the cardex system

Doctor's room: Illustrative posters and instructive material for different health problems. Diplomas and awards, as for example for the first “vesícula extraction” in May 1999

All written elements in Spanish 74 VON GLEICH

ANNEX 3 LITERACY EVENT NO.______2.OBSERVER 3. DATE:______4. TIME:______5. OBSERVATION TYPE: natural ð experimental ð 6. Participatory: no ð or yes ð if YES introduced by:______7. Observer known to participants: No ð, yes to some ð, yes to all ð 8.PROPERTIES OF LOCALITY:______9. TOPIC of the literacy event______10. EVENT is: Spontaneous ð or organised ð 11. Type of Event: Irregular ð or regular ð, frequency:______12. PURPOSE/S of the event 13. TARGET AUDIENCE:______COMMUNICATION-CONFIGURATION: DATA CONCERNING THE WRITTEN ELEMENT: 14. Meeting ð, person-to-person ð 19. Type of written element (book, flyer...): ______15. Estimated no. of participants: ____ 20. Already produced ð or product of the event ð 16. Audience: invited ð, public ðor both ð 21. Physically present: yesð or no ð 17. Frontal ð, roundtable ð 22. Language/s: ______or informal grouping 18. Relationship: formal ð or informal ð 23. Function______24. Language/s of mediation in the event______

ANNEX 4: observation sheet for active participants during literacy events Date Observer: Participant Name: (F)em / (M)asc. Age Home/residence Occupation: Role or function in the event: Participant: behaviour during the event Activities during the event (type and with whom?)

Languages of mediation during the event: Written languages during the event: Nonverbal communication (gestures, mimics): LEBEN, EXPEDITIONEN, SAMMLUNGEN UND UNVERÖFFENTLICHTE WISSENSCHAFTLICHE TAGEBÜCHER VON DR. EMIL HEINRICH SNETHLAGE

Dr. Rotger Michael Snethlage

1. Einleitung Ich bitte um Nachsicht, daß ich als Jurist keine spezifischen ethnologischen Kenntnisse habe. Vielleicht erzähle ich Ihnen Dinge, die Ihnen selbstverständlich erscheinen. Ich werde kurz das Leben meines Vaters skizzieren, dann seine beiden Forschungsreisen vorstellen, dann die von ihm 1933/35 besuchten Indianerstämme vorstellen, berichten, was ich über den Verbleib seiner Sammlungen weiß, und kurz auf das eingehen was noch in seinem Nachlaß in meinem Besitz vorhanden ist. Mein Vater, Dr. Emil-Heinrich Snethlage, ist 1897 in Bremerhaven geboren und bereits 1939 mit 42 Jahren gestorben, an den Folgen einer im Kriegsmarinedienst erlit- tener Verletzung.1 Ich selbst war damals drei Jahre alt. Ich kann also nicht aus erster Hand über ihn berichten. Er war promovierter Ornithologe. Nach seiner ersten Forschungsreise 1923/26 wandte er sich aber aus Passion der Ethnologie zu. Zuletzt war er stellvertretender Lei- ter der südamerikanischen Abteilung am Völkerkunde-Museum in Berlin.

1 Albert Snethlage (1982:87 ff). 76 SNETHLAGE

2. Schul- und Studienzeit Seine Schulzeit war recht unruhig durch die verschiedenen Dienstorte seines Vaters in Pommern, in der Neumark, in Schleswig-Holstein und in Westfalen. 1917 wurde er zum Kriegsdienst bei der Kriegsmarine in Wilhelmshaven eingezogen. Nach seiner Entlassung 1919 und anschließendem Abitur studierte er Botanik, Zoologie und als Hauptfach Ornithologie. Vorbild war ihm seine Tante, die Ornithologin Dr. Emilie Snethlage. Sie war in Belém do Pará, Brasilien um Museum Goeldi tätig. Studienorte waren Freiburg, Kiel, zuletzt Berlin. Er promovierte dort 1923 zum Dr. phil. über ein zoologisches Thema aus dem südamerikanischen Bereich (‘Beiträge zur Kenntnis der Gattung Cecropia und ihrer Beziehungen zu den übrigen Conocephaloideen’, Friedrich- Wilhelms-Universität Berlin).

3. Forschungsreise Nordostbrasilien 1923-1926 Sofort nach seiner Promotion reiste er im März 1923 nach Brasilien ab, um dort auf einer ersten Forschungsreise mit seiner Tante Dr. Emilie Snethlage, gemeinsame orni- thologische Forschungen zu betreiben. Vom Februar 1924 an war er während seines Aufenthaltes in Nordostbrasilien auf sich alleine angewiesen und kam bald in nähere Berührung mit den und den Krân-Stämmen, über die er später publizierte.2

4. Museum für Völkerkunde Berlin Im Sommer 1926 traf Emil Heinrich Snethlage wieder in Deutschland ein und wurde im Frühjahr 1927 mit der Betreuung der südamerikanischen Sammlungen des Berliner Museums für Völkerkunde beauftragt. Überraschend schnell fand er sich in die Völker- kunde hinein. Passioniert widmete er dieser damals noch jungen Wissenschaft bis zu seinem vorzeitigen Tode seine Arbeit.3

5. 1933-1935: 2. Expedition nach Bolivien/Brasilien 1933/35 ging er mit Mitteln der Baeßler-Stiftung für das Berliner Museum auf seine zweite Expedition. Sie führte ihn in das brasilianisch-bolivianische Grenzgebiet am Rio Guaporé. Hier besuchte er nacheinander die Moré, die sich bis dahin allen Annähe- rungsversuchen unzugänglich erwiesen hatten, die Kumana, Abitana-Huanyam, Am- niapä und Guaratägaja, Makurap, Jabuti, Arikapu, Wayoro und andere Stämme. Von allen brachte er reiches Material in Gestalt von Sammlungen und Aufzeichnungen nach Hause. Zitat: “Diese Expedition wurde mit den einfachsten Hilfsmitteln ausgeführt und verdankte ihren vollen Erfolg vor allem der persönlichen Anspruchslosigkeit Emil Heinrich Snethlages und seiner menschlichen Güte, die ihm die Herzen der Indianer gewann”.4 Zitat Ende.

2 Nevermann (o.d.). 3 Nevermann (o.d.). 4 Nevermann (o.d.). LEBEN VON DR. EMIL HEINRICH SNETHLAGE 77

6. Veröffentlichungen Über seine Reise veröffentlichte er außer einigen kleineren Arbeiten 1937 sein popu- lärwissenschaftliches Buch “Atiko y” und in 1939 eine wissenschaftliche Studie über die Musikinstrumente im Guaporégebiete. Es sollten Monographien über die einzelnen Stämme folgen. Die über die Moré hatte er bereits in Angriff genommen. Ebenso war die Herausgabe seiner Wörterverzeichnisse geplant. Sein vorzeitiger Tod vereitelte die- se Arbeiten.5 In der Museumsarbeit beschäftigte er sich eingehend mit den alt- peruanischen Sammlungen des Museums, für dessen Webereien er eine neue praktische Systematik fand.6 Er veröffentlichte Studien über ein Ikat-Gewebe aus Peru sowie über Form und Ornamentik alt-peruanischer Spindeln.

7. Pläne Kurz vor seinem Tode wurde er noch am 5. Oktober 1939 mit der stellvertretenden Leitung der Amerikanischen Sammlungen des Museums für Völkerkunde betraut7. Zi- tat: “Vor seiner Einberufung war er auf dem Höhepunkt seines Schaffens angelangt und wandte sich weitergefaßten Themen zu: einer Arbeit über südamerikanische Symbolik, [...], einer Untersuchung über den Kulturwandel in Südamerika unter besonderer Be- rücksichtigung der Mischlingsfrage und Abhandlungen über das Wesen des Medizin- mannes und des Mana- und Seelenbegriffes der südamerikanischen Indianer. Noch auf seinem schweren Krankenlager sprach er immer wieder von diesen Arbeiten, unter de- nen ihm die letzten besonders am Herzen lagen, und hoffte auf die Zeit, in der er sich ganz der Ausarbeitung seiner vielen Pläne hingeben konnte.”8 Zitat Ende.

8. Forschungsreise 1923-1926 8.1. Mit der Ornithologin Dr. Emilie Snethlage nach S. Luiz Anfang 1923 reiste Dr. E.H. Snethlage nach Brasilien zu seiner Tante Dr. Emilie Snethlage, die am Museum Goeldi in Belém do Pará tätig war. Mit ihr zusammen unternahm er von Juli 1923 bis Februar 1924 eine ornithologische Forschungsreise nach Maranhão in Nordostbrasilien, das bis dahin ornithologisch noch kaum er- forscht war.9 Stationen waren San Luiz, San Bento, Tury-assú (damals noch Überfäl- le zivilisations-feindlicher Urubú-Indianer)10, Alto de Alegria, Insel Mangunça.

8.2. Als Ornithologe für das Field-Museum, Chicago, unterwegs Allein, mit Aufträgen des Field-Museums in Chicago versehen, bereiste er von März 1924 bis April 1926 das Landesinnere: von San Luiz, wegen Hochwassers unter er-

5 Nevermann (o.d.). 6 Nevermann (o.d.). 7 Der Generaldirektor der Staatlichen Museen (1939): “Für die Dauer der Abwesenheit des Kustos und Professors Dr. Krickeberg beauftrage ich Sie hiermit mit der stellvertretenden Leitung der Amerikanischen Sammlungen des Museums für Völkerkunde. (Unterschrift: Kummel [?]) An Herrn Dr. Snethlage hier.” 8 Nevermann (o.d.). 9 Snethlage (1927:482). 10 Snethlage (1927:456). 78 SNETHLAGE heblichen Schwierigkeiten abwechselnd mit der Eisenbahn und zu Schiff über Rosa- rio den Rio Itapicurú hinauf bis Codó und Cocos, damals nur eine Bahnstation mit einigen Hütten. Mit Maultieren weiter über Pedreiras am Mearim nach Barra do Corda. Zu den in der Umgebung wohnenden “zahmen”, aber noch wenig erforschten Indianern nahm er Kontakt auf, es waren die zum Sprachstamm der Gê gehörenden Krân- Stämme der Remkokamekrã und der Aponyekrã,11 beide Canellas, und die zu den Tupi gehörenden Guajajáras.12

8.3. Canellas Bei den Canellas im Remkokamekrã-Dorf Ponto, etwa 120 km südlich von Barra do Corda im Quellgebiet des Rio Corda gelegen, verbrachte er einige Zeit und gewann das Vertrauen der Indianer. Er berichtet von einer belustigenden Ansinnen der India- ner: Zitat: “Ich sollte bei Ihnen bleiben und ihr Häuptling werden. Hauptantrieb war wohl der Gedanke, sich mit Hilfe eines kriegerischen Deutschen und dessen Waffen an den verhassten, umwohnenden brasilianischen Ansiedlern rächen zu wollen. Ich nahm das zuerst nicht ernst. Als ich aber die Absicht zu erkennen gab, aufzubrechen, wurden mir Vorstellungen gemacht und zur Bekräftigung ein hübsches Indianermäd- chen, das höchstens 11 Jahre zählte, zugeführt. Meine Proteste halfen nichts. Ich saß richtig in der Klemme, da ich meinen Diener und meine Lasttiere nach Baara do Corda zurückgeschickt hatte. Erst als ich den Indianern den Wunsch aussprach, mei- ne Eltern doch noch einmal wiedersehen zu wollen, verstanden sie sich dazu mich fort zu bringen.” Zitat Ende.

8.4. Guajajára Von Barra do Corda ritt mein Vater im Oktober 1924 nach Grajahú und lernte dort die versteckt im Monsunwald gelegenen Dörfer der der Guajajára (Tupi), (damals noch ca 1.500 Seelen)13 kennen sowie später auf der Bootsfahrt den Rio Grajahú hin- unter die der unmittelbar benachbart siedelnden Kreapimkataye14 (Krân-Stamm15 der Tymbiras, Sprachstamm der Gê)16

8.5. Apinaye Anfang Dezember 1924 war er zurück in San Luiz, reiste aber unmittelbar anschlie- ßend weiter in den Staat Ceará um auftragsgemäß ornithologisch zu sammeln. Über Ipiapaba, Ceará und Arára zurück nach Parnaiba und von dort, nach einem Abstecher ins Landesinnere nach Deserto, den Rio Parnaiba ca 540 km flußaufwärts über Tere- sina, der damaligen ca 20.000 Einwohner zählenden Hauptstadt des Staates Piauí

11 Snethlage (1930:185). 12 Snethlage (1927:463). 13 Snethlage (1930:185). 14 Snethlage (1930:185). 15 Snethlage (1930:188). 16 Snethlage (1927:472). LEBEN VON DR. EMIL HEINRICH SNETHLAGE 79 nach Amarante/São Francisco auf der Grenze zwischen Piauí und Maranhão und nach kurzem Sammelaufenthalt weiter nach Floriano, Urussuhy, Inhumas und Tran- queira. Über Goiás und Certeza ging es dann im Oktober 1925 zunächst auf einem eigenen kleinen Floß, später auf größerem Handelsfloß den Rio hinunter über Pedro Afonso nach Carolina, das kurze Zeit später von Revolutionären, auf- ständischen brasilianischen Soldaten, besetzt wurde, die sich aber als sehr diszipli- niert erwiesen. Nach Abzug der Soldaten Ende November 1925 reiste er weiter in den Bereich der São Antonio-Fälle und kam dort in Kontakt mit den Apinagés17 (auch Apinaye)18, ein Krân-Stamm aber mit dem “g” statt des bei den Maranheser Horden gesprochenen “k”19, zur Sprachfamilie der Gês gehörend. Über die damals am Rio Gurupi noch wild lebenden Gaviões konnte er gute Auskünfte einziehen.20 Sich wiederholenden Malariaanfälle behinderten und unterbrachen hier seine Arbeit. 2 Monate mußte er auf den nächsten Dampfer warten bei schwer verdaulicher Kost ohne Chinin, ohne Pflege in einer Negerhütte, die durch Lagerung von Salzsäcken stets feucht war. Erst im März 1926 konnte er nach Carolina zurückgebracht werden, wo er von einem dort ansässigen Deutschen gesund gepflegt wurde. Mitte April 1926 kehrte er nach Parà zurück.

9. Forschungsreise 1933-1935 Die zweite Forschungsreise führte ihn 1933/35 in das Gebiet des Itenes/Rio Guaporé. Seine Nebenflüsse und teilweise auch seine Ufer boten damals ethnographisch noch vieles Neue. Denn weder der schwedische Forscher Erland Nordenskiöld noch der brasilianische General Rondon, noch die älteren Reisenden, Missionare und Wissen- schaftler haben alle Winkel dieses Landes durchstreifen können. So ist es erkärlich, daß mein Vater zahlreiche der Wissenschaft bis dahin noch unbekannte Stämme an- traf. Mitte Juli 1933 kam mein Vater in Pará an und war am 10. August in Porto Velho, damals Ausgangspunkt der Madeira-Mamoré-Eisenbahn. Mitte August be- sichtigte er die Steinzeichnungen bei Kilometer 151 dieser Eisenbahn. Anfang Sep- tember 1933 kam er am Ausgangspunkt seiner Expedition, im Campamento Koma- rek.21 Zur Begrüßung brannten die damals noch “wild” lebenden Moré in der Nacht seiner Ankunft das Werkstattgebäude der Farm nieder, aus Wut darüber, daß ein neuer Weißer angekommen war. Entsprechend zögerlich war die Kontaktaufnahme. Nach 4 Monaten Feldforschungen bei den Moré und Itoreauhip (inzwischen war endlich die immer wieder in Aussicht gestellte und dann doch wieder verzögerte Er- laubnis der brasilianischen Regierung zur Forschungsarbeit eingetroffen) brach er Weihnachten 1933 (am Tag des Todes seines in seiner Abwesenheit geborenen Töchterchens, von dem er aber erst drei Monate später etwas erfuhr) zum Rio Cauta-

17 Snethlage (1927:481). 18 Snethlage (1930:185). 19 Snethlage (1930:188). 20 Snethlage (1927:481). 21 Chronologischer Verlauf der Reise: Snethlage (1937b:8). 80 SNETHLAGE rio und den dort ansässigen Kumaná auf. Im Februar 1934 besuchte er die in Bella Vista, machte Ausgrabungen im Cafétal (Piso firme) und war Anfang März bei den schon akkulturierten Tschikitano in . Mitte März 1934 Aufbruch in die Dörfer der Makurap-Häuptlinge Uaikuri und Guata. Anfang April Weiterreise zur Serra de Allianza mit Ausgrabungen, Mitte April den Mequens aufwärts, zu den Amniapä im Dorfe Tapuawas. (Die Reisen meist allein im Canu, manchmal mit indianischen Ruderern und Wegweisern) Anfang Mai zu den Guaratägaja, dann zurück den Mequens abwärts und Mitte Juni den Rio Branco aufwärts zu den Arua in San Luiz. Ende Juni bis Anfang August Fußreise durch den Urwald mit Besuch der Maku- rap, Jabuti, Wayoro, Arikapu, und Tupari. Mitte August den Rio Branco abwärts und zurück zum Cautario, im Oktober wieder bei den Moré und Itoreauhip. Ende November Abreise vom Campamento Komarek aus heimwärts.

9.1. Moré und Itoreauhip (Tschapakura) Auf der bolivianischen Seite des unteren Itenes (Guaporé) haben die sprachlich den Tschapakura angehörenden Moré und Itoreauhip ihre Wohnsitze. Beide Stämme un- terschieden sich damals nur durch ihre Dialekte und durch die Haartracht: die Moré trugen ihre Haare offen auf die Schulter herabfallend, die Itoreauhip binden sie zu einem Knoten zusammen. Ihre Kleidung bestand aus meist gestreiften Rindenstoff- hemden, doch liefen die Männer gewöhnlich nackt herum. Zu Festen wurde Lippen- und Ohrschmuck, Reife und Federkronen, Federbänder in manigfachen Farben an Armen und Beinen getragen. Zahllos waren bemalte oder bastverzierte Bänder und Gürtel. Körperbemalung war selten, aber die Frauen rieben sich und die Angehörigen gern mit Urucu ein, einer in Palmöl gelösten roten Pflanzenfarbe.22

9.1.1. Wohnen In Großfamilien, etwa 15-70 Köpfe umfassend, lebten die Moré und Itoreauhip in mit Palmstroh bedeckten Giebelhütten. In der mückenreichen Zeit bezogen sie mit Patohu-Blättern völlig geschlossene Schlafhütten, die nur durch ein kleines, durch eine geflochtene Tür verschließbares Loch zugänglich waren. Ihre mit Bananen, Mais, Maiok, Inyame, Bataten, Ananas, Baumwolle und Urucu bepflanzen Rodun- gen und der Fischfang (Schießen mit Pfeilen von ihren Einbäumen aus, Reusen im Palisadenzaun, Giftliane) lieferten den Lebensunterhalt. Die Jagd mit Pfeil auf Säu- getiere und Vögel, bisweilen von bienenkorbartigen Jagdhütten aus, hatte nur Bedeu- tung, wenn die Moré zur Zeit der Fruchtreife ein Nomadenleben führten. Dann befe- stigten sie ihre baumwollenen Hängematten an die Stützen ihrer mit Patohu-Blätter bedeckten Unterschlupfe, die ihnen Schutz vor Regenschauern boten.

22 Auch für das Folgende: Snethlage (1937b:2 ff.). LEBEN VON DR. EMIL HEINRICH SNETHLAGE 81

9.1.2. Männerarbeit Als Angehörige eines Kriegervolkes legten die Männer großen Wert auf die Ausge- staltung der Rundbögen und Pfeile. Bögen wurden durch Bast- und Baumwollum- wicklung verstärkt. Die etwas mehr als meterhohen Pfeile waren mit zwei, drei oder vier Federhälften befiedert und trugen je nach Verwendungszweck für Krieg oder Jagd, Bambusmesser- Holzsäge-, Knochen- oder Rochenstachelspitzen. Außer Jagd und Fischfang besorgten die Männer die Hauptarbeit beim Hütten- bau und auf den Feldern. Sie holten den richtigen Bast aus dem Wald, klopften ihn weich, nähten die Rindenhemden, sie fertigten aus Holz die Boote, die Tröge, die Sitze, das Spielzeug für die Kinder. Sie fegen ihren Arbeitsplatz selbst sauber, oder halfen gelegentlich ihren Frauen, etwas Maniokmehl für die Chichabereitung zu zer- kauen.

9.1.3. Frauenarbeit Aufgabe der Frauen war in erster Linie die Hausarbeit: Mais zerstampfen in längli- chen Trögen mit Mahlstein oder Mahlholz, Maniokwurzeln mit einem Holzmesser schälen, an einem dornigen Stelzwurzelstück der Paxiubapalme zerreiben, die ge- wässerte Reibemasse durch die Stäbchenmatte drücken und dann auf dem Tonteller zu Farinha rösten, daraus wie auch aus Maismehl Fladen backen. Natürlich ist es ihre ständige Aufgabe für die Kinder zu sorgen, aber auch zu töpfern, Baumwolle zu zup- fen und zu verspinnen, und daraus Hängematten zu knüpfen, aus feineren Fäden auch Arm- und Beinschnüre auf einem einfachen Webrahmen zu flechten.

9.1.4. Spiele, Musik und Tanz Unter den Spielen sind das Maisballblattschlagen, das Drehen der Surrscheibe und die Fadenspiele besonders bermerkenswert. Die Moré und Itoreauhip besitzen eine Menge Musikinstrumente. Mit einem Pfiff auf einem einfachen Rohr hatte man sich schon vor jeder Wohnung bemerkbar zu machen. Diese “Pfeife” wurde in mannigfa- cher Weise ausgestaltet bis zur einfachen Längsflöte und Querpfeife. Wollte der Mo- ré verschiedene Töne vereinigen, nahm er zwei Flöten in den Mund, oder baute sich aus beliebig vielen Rohren eine Panflöte auf. Kürbistrompeten waren sehr beliebt, ebenso Rasseln aus mit Samen gefüllten Kalebassen oder aneinandergereihten klei- nen Kürbissen. Trommelschlag auf eine Palmblattscheide gab den Rhytmus eines Galopptanzes an; der “taran” Taktschläger. ein an einem Stab gleitender, aus einer Kalebasse bestehender Schallkörper de langsamen Rhytmus eines andern Tanzes. Andere Tänze wurden mit einer Heulkuye23, mit Flötenspiel oder Gesang begleitete. Alle Tänze stellten augenscheinlich Vorgänge aus legenden dar.

23 Snethlage (1937a:4, “Reib-Idiophon” in 1939:12). Wohl ein übersetzter portugiesischer Ausdruck. 82 SNETHLAGE

9.2. Tschapakura-Sprache Zur Sprachfamilie der Tschapakura (Chapakura) schrieb mein Vater:24 “Créqui- Montfort und Rivet haben 1913 im Journal Soc. Americ. Paris auf Grund des damals vorhandenen Materials die Sprachfamilie der Chapakura aufgestellt. Sie umfaßte damals die Stämme: Chapakura oder Huaci, die Kitemoka, die Pavumva (Huanyam), die Napeka, die Iten und mit Vorbehalten, Rokorona und Muré (oder Murä [Ni- muendaju]). Zu Grunde lagen vor allem die damals noch unveröffentlichten Vokabu- larien von d’Orbigny in der Bibliothèque Nationale in Paris und die bis dahin veröf- fentlichten Worte und Sätze von d’Orbigny, Cardús und Hasemann. Es war genü- gend, um die Aufstellung dieser Sprachfamilie zu rechtfertigen. Später brachte Nimuendaju noch Wortlisten heraus, die diese Gruppe um 3 Stämme vermehrte: die Tora, Jain und Urupá (Nimuendajú 1925). Das Gebiet wird also weithin nach Norden ausgedehnt. Nordenskiöld veröffentlichte leider nur ein- zelne Ausdrücke (Forschungen und Abenteuer), doch dürfte weiteres Material mit- gebracht worden sein. Die auf meiner Reise in das Guaporégebiet gemachte sprachliche Ausbeute er- laubt, noch die Kumaná und Kabixi-Huanyam hinzuzufügen. Gleichzeitig wurden die Wortlisten der Abitana-Huanyam am Miguel und die der Iten (in Wirklichkeit mindestens 2 Stämme: Moré und Itoreauhip) vergrößert, so daß es möglich ist, das von Créqui-Montfort und Rivet zusammengebrachte Material kritisch zu betrachten. Konnte ich auch nicht zusammenhängende Texte mitbringen, so dürfte doch das Ty- pische der Chapakura - oder wie ich sie nennen möchte - Huanyam- Sprachen schär- fer hervortreten.”

9.3. Kumaná Auch die Kumaná zwischen dem mittleren Cautario und dem Rio S. Domingos sind Tschapakura. In Kultureller Beziehung wichen sie aber damals beträchtlich von den Moré und Itoreauhip ab. Die Kumaná wohnen in großen ovalen Hütten. Nackt liefen bei ihnen nur die Frauen, bei Besuch und Festlichkiten zogen sie ihre mit bemer- kenswert schönen Mustern bemalten Rindenhemden an. Die Männer hatten ausge- sprochenes Schamgefühl. Sie zeigten sich nur in ihrem über der Hüfte hochgegürte- ten Bastkleid. Auch Bastjacken wurden angefertigt. Die einschnürenden Bänder an Armen und Beinen waren breit und mit Fransen versehen, der Schmuck war sorgfäl- tiger gemacht. Die Pfeile waren größer als bei den Moré, der Vogelpfeil hatte eine Spitze aus mit Wachs verbundenen Tapirzähnen. Die Bastklopfer waren rund statt kantig, die Sindeln hatten Wirtel aus Kalebassenschale, anstatt aus Früchten oder korkigem Holz, die Frauen zermahlten den Mais auf breiten Platten aus der Brett- wurzel eines Urwaldriesen anstatt in einem Mörser. Der Tanz ist ein Gänsemarsch im Kreise nach Gesang, Geräusch von Rassel und den tiefen Tönen einer Kürbistrompe- te.

24 Handschriftliche Notiz aus dem Nachlaß, im Besitz von Dr. Rotger Michael Snethlage Aachen. LEBEN VON DR. EMIL HEINRICH SNETHLAGE 83

9.4. Abitana-Huanyam Die Kumaná bilden kulturell schon den Übergang zu den ebenfalls Tschapakura sprechenden Abitana-Huanyam am Rio S. Miguel. Aber der Besitz von Giftpfeilen und Blasrohren unterscheidet diese doch wesentlich von jenen, zumal dazu noch zahlreiche Merkmale untergeordneter Art kommen. Jede Huanyamfrau trug damals noch trotz europäischer Kleidung den schweren Lippenpflock aus Quarz, der ihre Würde als verheiratete Frau anzeigte. Unter den Musikinstrumenten ist die aus den Oberschenkelknochen erschlagener Feinde gefertigte Trompete besonders zu erwäh- nen. Der Tanz bewegte sich auch bei den Huanyam im Kreise.

9.5. Makurap Im Gebiet des Rio Branco gab es keine Tschapakura mehr. Aber in sehr vielen Spra- chen traten Tupi-Elemente auf, doch so, daß diese Sprachen nach Auffassung meines Vaters nicht als miteinander nahe verwandt bezeichnet werden konnten. Unter ihnen beherrschste der tupoide Stamm der Makurap kulturell alle seine Nachbarn. Er war in vaterrechtliche Sippen aufgespalten, die sich nach Tieren oder Pflanzen nannten. Sie glaubten an zwei gute Götter und an einen schlechten, Tschoari, den Herrn der Gei- ster und Totenseelen. Dem Kult diente ein meist bemalter Mattenaltar, der in der Mit- te des hohen Kelhauses stand, der Eingangstür gerade gegenüber. Der Zauberer be- dienste sich bei verschiedenen Zeremonien und Krankenheilungen der Zauberrassel, des Zauberbrettes, verschiedener Heilpflanzen, des Schnupfrohres, der Zauberfeder und bisweilen auch anderer Geräte. Die Angehörigen beider Geschlechter gingen damals noch nackt bis auf den aus Samen bestehenden Hüftgürtel und Schmuck: breite auf einem runden Holz in pas- sender Größe gewebte, urucurot gefärbte Armbänder, aus Sämereien oder zurechtge- schliffenen Muscheln bestehendem Halsschmuck, dem aus Rohr oder einem mit Fe- dern beklebten und dann mit hellem Harz überzogenem Stäbchen bestehenden Na- senschmuck und dem aus Muschelschalen geschliffenen Ohrgehänge. Der Mann trägt außerdem immer den Penisstulp. Korb- und Mattenflechterei ist Männerarbeit, Frauen fertigen Kalebassen und Tragnetze aus Tukumfasern. Baumwolle wird auf Bakairiweise versponnen, spielt aber keine große Rolle. Bogen und Pfeilen entsprechen denen der von Nordenskiöld beschriebenen Huari. Der Tanz besteht aus schnellen Schritten hin und her nach dem Takt von Bambustrompeten oder einem Instrument, das aus 9 durch Wachs miteinander ver- bundenen Flöten besteht. Die Hand liegt dabei meist auf der Schulter des Vorder- mannes.

9.6. Arua Die Arua sind ebenfalls ein tupoider Stamm. Eine ihrer Horden war damals schon auf einer Indianerstation gesammelt worden und trug deshalb europäische Kleider. Ihre ursprünglich abweichende Kultur war von der der Makurap bereits überdeckt. 84 SNETHLAGE

9.7. Wayoro Auch die damals schon geringen Überreste der Wayoro, die sprachlich eine Mi- schung zwischen Makurap und anderen tupoiden Stämmen sind, hatten die Maku- rapkultur völlig übernommen.

9.8. Jabuti und Arikapu Sogar die Jabuti und Arikapu, deren völlig andere Sprachen zahlreiche Elemente der Oststämme Brasiliens, der Gê, enthalten, waren damals stark von den Makurap be- einflußt.

9.9. Tupari Ziemlich unabhängig waren damals noch die tupoiden Tupari. Sie besaßen keinen Mattenaltar, Dafür spielte das Schnupfen bei Zauberzeremonien eine viel größere Rolle als bei den übrigen Rio Branco-Stämmen. Von den Gegenständen der materiellen Kultur unterscheiden sich viele in ihrem Aussehen von denen der anderen Indianer: der Bogen, die Pfeile, der Sitz, die Schambekleidung, der größte Teil des Schmuckes, die Spindeln, die Musikinstru- mente. Hier gab es eine viergriffige Längsflöte aus Bambus. Auf den Rodungen wurden mehrere den übrigen Brancostämmen unbekannte Gemüsepflanzen gezogen, Käfer wurden in den dicken Rückständen der Chicha gezüchtet, um in den Larven einen wohlschmeckenden Leckerbissen und eine Beigabe zum Maniokbrot zu erhal- ten. Steinäxte waren noch im Gebrauch. Von den Tänzen ähnelt der Chichatanz dem der übrigen Brancobewohner. Der Flötentanz wurde dagegen nur von zwei sogenannten Häuptlingen, die ständig in gleicher Entfernung voneinander blieben, vorgeführt.

9.10. Amniapä und Guaratägaja Die Tupari nähern sich in ihrer materiellen und sozialen Kultur bereits den Amniapä (Mampiapä) und Guaratägaja im Gebiete des oberen Mequens und der auf demselben Höhenzug entspringenden Zuflüsse des Pimenta Bueno. Wie die Tupari aßen diese Indianer jedenfalls damals noch von Zeit zu Zeit Feinde und ungetreue Angehörige des eigenen Stammes. Das ganze Leben der Amniapä und Guaratägaja war von Ze- remonien ausgefüllt. Mit dem Pfeil im Bogen marschieren Gäste in ein immer aus mehreren bienenkorbartig aussehenden und um einen sauber gehaltenen Platz grup- pierten Kegelhütten bestehendes Dorf. Erst nach gemessenen Reden und dem Aus- trinken einer großen Kalebasse Chicha ist freierer Verkehr möglich. Groß ist dann auch die gewährte Gastfreundschaft. In einem Spiel mit einem aus Kautschuk beste- henden Ball, der nur mit dem Kopf berührt werden darf, wir die Geschicklichkeit erprobt. Pfeile oder Schmuck sind der Einsatz jeden Mitspielers; ausgezählt werden die einzelnen Spiele mit Maiskörnern. Kultisches Schnupfen im Männerkreis schließt sich an.; die Frauen haben wäh- rend dieser Zeit im Hintergrund der Hütte zu verbleiben. Erst wenn die Speisen ge- LEBEN VON DR. EMIL HEINRICH SNETHLAGE 85 segnet werden und die Krankenheilungen beginnen, dürfen sie wieder hervorkom- men. Der Tanz beginnt aber erst in den Abendstunden und dauert die Nacht hindurch an. Er endet bei Aufgang der Sonne. Kurze Zeit danach verabschieden sich die Gäste mit traurigen Reden und Tränen in den Augen. Während im Rio Branco-Gebiet Männer und Frauen fast den gleichen Schmuck tragen, ist er am Mequens bei beiden Geschlechtern sehr verschieden. Die Männer der Guaratägaja tragen außer dem Stulp, der nach seiner Machart die einzelnen Gruppen unterscheidet (die Amniapä tragen keinen) noch einen Schurz aus Buriti- palmfasern, dazu reichen Schmuck an Schulter- und Halsketten, Armbändern, Ge- sichtsschmuck. Die Frauen sind mit zahllosen Halsketten behangen, meist solchen aus Samen. Besonders charakteristisch sind Ketten aus ganzen Muschelschalen, dazu Nasen- und Ohrschmuck abweichend von denen der Männer. Bis auf die gewirkten Bänder machen sich die Männer ihre Sachen selber. Sie ritzen oder bemalen die Behälter ihres Schnupftabaks, von denen einige Muster auf- weisen, die denen des Xingu-Quelgebietes ähnlich sind. Sie fertigen auch die Mas- kenaufsätze, an denen die aus Palmfiedern geschnittenen Zähne auffallen. Als In- strumente dienen Piranhakiefer, Agutizahnmesser, Kieselsteine. Auch die Palmfie- derröcke werden von den Männern hergestellt, von ihnen werden auch die Körbe geflochten. Musikinstrumente sind Kürbistrompete, Panflöte, viergriffige Flöte, Knochenflöte, Rassel.

9.11. Pauserna-Guarayu Die Pauserna-Guarayu hatten ihre urprüngliche Kultur bis auf wenige Sachen völlig eingebüßt. Einige Gegenstände der Hauswirtschaft und des Mobiliars hatten sich erhalten. Die Hängematte z.B. wurde damals für den Verkauf hergestellt. Auch die Tongefäße wurden, wenn auch in sehr vereinfachter Form von den Frauen noch ge- töpfert. Die Pauserna mischen den durch Srampfen alter Tonscherben erhaltenen Staub mit frischer Tonerde, um widerstandsfähige Keramik zu erzielen. Der alte Schmuck war damals schon fast völlig verschwunden.

9.12. Tschikitano Die von den Jesuiten schon christianisierten Tschikitano hatten ihre ursprüngliche Kultur damals schon völlig verloren. Nur in wenigen Überlieferungen, vielleicht in einigen Tänzen lebte noch der alte Geist.

10. Sammlungen Die Hauptsammlung blieb in Brasilien, Ihr heutiger Verbleib ist mir nicht bekannt, vielleicht . Ein erhaltenes Übergabeverzeichnis vom 22. Februar 1935 nennt rund 130 Objekte und aus Ausgrabungen stammende Keramikscherben, die dem Museu Nacional in Rio de Janeiro übergeben wurden:25

25 Dr. E. Heinrich Snethlage: Relação dos objectos e cacos de ceramica do Rio Guaporé ofericidos ao Museu Nacional. (1935); Mit handschriftlicher Notiz: “Recebi o material ethnografico, constante da relaçõ acima, 86 SNETHLAGE

8 Objekte der Kumaná und 79 cacos de ceramica aus Canindé 21 Objekte der Abitana-Huanyam 19 Objekte der Arua 13 Objekte der Makurap 4 Objekte der Jabuti 11 Objekte der Wayoro 15 Objekte der Tupari 33 Objekte der Amnipä und Guaratägaja

Die Duplikat-Sammlung ging an das Völkerkunde Museum in Berlin

11. Wissenschaftliches Tagebuch und Wortlisten Die Durchschriften des wissenschaftlichen Tagebuchs meines Vaters von der For- schungsreise 1933/35 in das Guaporégebiet konnte meine Mutter mit Hilfe Professor Rivets, Paris und Professor Gusindes, Wien, über den Krieg hinweg retten, ebenso die Wortlisten. Erhalten sind mit geringen Lücken etwa 1000 Seiten Durchschriften des Tage- buchs, und die Wortlisten der Pauserna, Kumana, Abitana-Huanyam und Moré. Von den Filmen sind Abzüge nur erhalten von einen Filmen.

12. Veröffentlichungen (unvollständige Zusammenstellung)26 12.1. Dissertation 1923 Beiträge zur Kenntnis der Gattung Cecropia und ihrer Beziehungen zu den übrigen Conocephaloideen. Friedrich-Wilhelms-Universität Berlin.

12.2. Wissenschaftliche Zeitschriftenaufsätze o.d. ‘Neue Arten der Gattung Cecropia nebst Beiträgen zu ihrer Synonymik’. 1924 ‘Neue Cecropien aus Nordbrasilien’, in: Notizbl. Bot. Gart. u. Museum Dah- lem, Bd. IX (30.12.1924). 1927 ‘Meine Reise durch Nordostbrasilien’, in: Journal für Ornithologie, LXXV, Heft 3, S. 456 ff. 1930 Form und Ornamentik alt-peruanischer Spindeln, Baeßler-Archiv, Berlin: Dietrich Reimer/Andrews & Steiner. 1930 ‘Unter nordostbrasilianischen Indianern’, in: Ethnologischer Anzeiger, Bd. II, Heft 4, S. 185-188. 1931 ‘Unter nordostbrasilianischen Indianern (1924)’, in: Zeitschrift für Ethnolo- gie, LXII, S. 111-205. 1931 ‘Ein figürliches Ikat-Gewebe aus Peru’, in: Weltkreis.

que me foi entregue, para as coleções do Museu Nacional - Rio de Janeiro. Heloisa Alberto Tones Prof.-Chefe de Lecçad Rio, 22 de Fevereio de 1935” 26 Basiert auf Nevermann (o.d.). LEBEN VON DR. EMIL HEINRICH SNETHLAGE 87

1932 ‘Emilie Snethlage: Chipaya- und Curuaya-Wörter. Aus dem literarischen Nachlaß herausgegeben von E.H. Snethlage’, in: Anthropos, Bd 27, S. 65-93. 1932 ‘Worte und Texte der Tembé-Indianer’, Tucuman. 1936 ‘Nachrichten über die Pauserna-Guarayú, die Sirionó am Rio Baures und die S. Simonianes in der Nähe der Serra S. Simon’, in: Zeitschrift für Ethnologie, LXVII, S. 278-293. 1937 ‘Übersicht über die Indianerstämme des Guaporégebietes’, Tagungsberichte der Gesellschaft für Völkerkunde, (Bericht über die II. Tagung in Leipzig 1936), Leipzig, S. 172-180. 1937 Indianerkulturen aus dem Grenzgebiet Bolivien-Brasilien. Ergebnisse der Forschungsreise 1933-1935, Veröffentlichung der Reichsstelle für den unter- richtsfilm zu dem Archivfilm Nr. B 25. (ca. 1936/38), Berlin. 1939 Musikinstrumente der Indianer des Guaporégebietes, Baessler-Archiv, Bei- heft 10, Berlin: Dietrich Reimer/Andrews & Steiner. 1939 ‘Untersuchung über das Pferdchen-Spiel auf Java und Bali’, Berlin.

12.3. Populärwissenschaftliche Schriften und Zeitungs-Artikel 1927 ‘Bei den Indianern des nordostbrasilianischen Hochlandes’, in: Illustrierte Zeitung, Leipzig: Verlag J.J. Heber, 169. Band, Nr. 4298, 28.7.1927, S. 144. 1929 ‘Im Indianerdorf’, (Einführung zur Kinderstunde der Deutschen Welle, Rei- sen und Abenteuer, am Do. 10.1.1929, 14.30 Uhr), in: Deutsche WlleJgg, Nr. 1 (4.1.1929), S. 11. 1935 ‘Zwei Jahre in ewigen Wäldern’, in: Leipziger Neueste Nachrichten, Nr. 155, Di, 4.6.1935, 1. Beilage, S. 5. 1935 ‘Zwei Jahre in ewigen Wäldern’, in: Hannoverscher Kurier, Nr. 252 53, So. 2.Juni 1935. 1935 Indianerkulturen aus dem Grenzgebiet Bolivien-Brasilien. Ergebnisse der Forschungsreise Dr. E. Heinrich Snethlage 1933-1934, Führer durch die Ausstellung im Staatlichen Museum für Völkerkunde, Berlin. 1937 Atiko Y. Meine Erlebnisse bei den Indianern des Guaporé, Berlin: Klinkhardt & Biermann. [Rezension: Stig Rydén in: American Anthropologist, Vol. 40, No. 1, January-March 1938]

12.4. Nachrufe 1930 ‘Dr. Emilie Snethlage zum Gedächtnis’, in: Journal für Ornithologie, LXXVIII, Heft 1, S. 122-134. (mit Bild). 1932 ‘Erland Nordenskiöld (+ 5.7.1932)’, in: Zeitschrift für Ethnologie, 64. Jgg. S. 368. 1932 ‘Theodor Koch-Grünberg (+ 9.4.1932)’, in: Deutsche Allgemeine Zeitung, Do. 7.4.1932. o.d. ‘Robert Lehmann-Nitsche’, in: Archiv für Anthropologie, N.F. Bd. 24, Heft 3- 4, S. 275-278. 88 SNETHLAGE o.d. ‘Konrad Theodor Preuss’, in: Archiv für Anthropologie, N.F. Bd. 24, Heft 3- 4, S. 278-280.

12.5. Rezensionen o.d. Schauinsland, H.: Fragen und Rätsel, Bremen, 1931. o.d. Wassén, Henry: Original Documents from the Cuna Indians of San Blas, Pa- nama. o.d. Baldus, Herbert: Ensaios de Etnologia Brasileira, Rio de Janeiro, 1937 o.d. Rydén, Stig: Archaeological Researches in the Department of La Candelaria. Göteborg 1936. [o.d. zahlreiche weitere Besprechungen im Archiv für Anthropologie, N.F. und Eth- nologischer Anzeiger.]

13. Literaturverzeichnis Generaldirektor der Staatlichen Museen 1939 Tagungsberichte Nr. I 2114/39, Berlin C2, Am Lustgarten, den 5. Oktober 1939. Nevermann, H. o.d. ‘Emil Heinrich Snethlage zum Gedächtnis’, in: Archiv für Anthropologie. N.F. Bd. XXVI, Heft 1/2 (Sonderdruck ohne Seitenangabe). Nimuendaju, Curt 1925 ‘As tribus do Alto Madeira’, in: Journal de la Société des Américanistes, n.s.t. 17, S. 137-172. Nordenskiöld, Erland 1924 Forschungen und Abenteuer in Südamerika, Stuttgardt [Stockholm 1915]. Snethlage, Albert 1982 ‘[Genealogie van de familie] Snethlage [Duitse tak]’, in: De Nederlandsche Leeuw, XCIXe, S. 77-111. Snethlage, Emil Heinrich 1927 ‘Meine Reise durch Nordostbrasilien’, in: Journal für Ornithologie, LXXV, Heft 3, S. 456 ff. 1930 ‘Unter nordostbrasilianischen Indianern’, in: Ethnologischer Anzeiger, Bd. II, Heft 4, S. 185-188. 1937a Indianerkulturen aus dem Grenzgebiet Bolivien-Brasilien. Ergebnisse der Forschungsreise 1933-1935, Veröffentlichung der Reichsstelle für den unter- richtsfilm zu dem Archivfilm Nr. B 25. (ca. 1936/38), Berlin. 1937b Atiko Y. Meine Erlebnisse bei den Indianern des Guaporé, Berlin: Klinkhardt & Biermann. 1939 Musikinstrumente der Indianer des Guaporégebietes, Baessler-Archiv, Bei- heft 10, Berlin: Dietrich Reimer/Andrews & Steiner. LIFE, EXPEDITIONS, COLLECTIONS AND UNPUBLISHED FIELD NOTES OF DR. EMIL HEINRICH SNETHLAGE

Dr. Rotger Michael Snethlage

1. Introduction Let me first point out to the reader that I do not have any specialised ethnological knowledge, since I have been trained principally in Law. Maybe I will speak about things that are obvious to the specialist. Furthermore, I’m afraid that my own command of English is very limited and I’m very grateful to my wife for translating the original German text and to my cousin Andrew Galitzine for proof reading the final draft. In this article, I will shortly outline the course of my father’s life. Then I will in- troduce his two expeditions to South America. Thereafter, I will give a short overview of those Indian tribes that he visited during his last journey from 1933 to 1935. Finally, I’ll mention the whereabouts of his collections, and indicate which manuscripts are still in my possession.

2. Youth and studies My father, Dr. Emil Heinrich Snethlage, was born in Bremerhaven in 1897 on the 31st of August, and died in Potsdam in 1939 on the 25th of November at the early age of 42.1 He was injured whilst serving in the German Navy. I was only 3 years old at the time, and, therefore, I have no immediate recollection of his life, and only know him through all the stories told by my mother. My grandfather was a teacher who often changed towns because of his work. Therefore my father saw many schools between Pomerania and Westphalia in an- cient Prussia. During World War I in the years 1917 - 1919, he had to go to the Navy. After this he was discharged and when he had graduated from school in 1919, he started his university studies in botany, zoology and, as his main subject, ornithol- ogy. His aunt, Dr. Emilie Snethlage, also an ornithologist, inspired him. She worked in Belém, Pará, Brazil, at the Goeldi Museum. My father studied in Freiburg, Kiel and Berlin where he obtained in 1923 a doctorate in philosophy on a zoological- botanical theme.

3. Brazil and the passion for ethnology Immediately after his graduation my father went to Brazil in order to do his first orni- thological research, in cooperation with his aunt. From February 1924 onwards he con- tinued his investigations alone in the northeastern region of Brazil. Soon he came in contact with the Guajajára and the Krân-tribes, on whom he published later on. In the summer of 1926, he returned to Germany. In 1927 he was employed at the Völk-

1 See Mr. Albert Snethlage (1982), and H. Nevermann (n.d.). 90 SNETHLAGE erkunde Museum in Berlin, the Berlin Museum of Ethnology, in order to administer the South American collections. With passion he devoted himself to the then young science of ethnology. In 1933 he departed for his second exclusively ethnological expedition. It was sponsored by the Baessler-foundation in Berlin. The location of his research was the frontier region of Bolivia and Brazil, formed by the Itenez or Guaporé river. There he visited 13 unknown tribes and collected many of their cultural objects. Back in Ger- many, he published several small essays as well as his popular-scientific book “Atiko y” and the scientific study on the “Musical instruments of the Indians in the Guaporé- region”, the latter of which was done in association with his friend Mario Schneider. My father planned monographic studies about the different tribes, and had already started on a monograph of the Moré. He planned the publication of his vocabularies, started on essays on South American symbolism and on themes from the indigenous religions. World War II halted all of these plans in 1939. Shortly before his death he became vice-director of the South American collections of the Berlin Museum of Eth- nology.2

2 See Der Generaldirektor der Staatlichen Museen (1939).

LIFE OF DR. EMIL HEINRICH SNETHLAGE 91

4. The expedition of 1923-1926 Now a few words about the first expedition in 1923/26.3 In 1923 my father went to- gether with his aunt Dr. Emilie Snethlage on an ornithological expedition to Maranhão, north Brazil, a region which was at that time mainly terra incognita for ornithologists. They paused at São Luiz, São Bento, Tury-assú (where the Urubú-Indians attacked the non-indigenous people), Alto de Alegria and Mangunça Island. After his aunt had to return to Belém do Pará, he travelled alone in Brazil from March 1924 to April 1926, by order of the Field Museum in Chicago. He travelled from São Luiz, both by railway and by boat, with many difficulties caused by the floods, via Rosario up the Rio Itapicurú until Codó and Cocos, which at that time consisted of just a railway-station and a few huts of black people. With mules he travelled on via Pedreiras (close to Mearim) to Barra do Corda. There he contacted the so-called “tame” Indians, who had not been extensively investi- gated. They were the Remkokamekrã and Aponyekrã tribes that formed part of the Ca- nella peoples who spoke languages of the Gê family, and the Guajajáras, who spoke a .

4.1. Canellas My father spent some time with the Canellas in Ponto, a village of the Remkokamekrãs, situated about 120 km south of Barra do Corda in the source region of the Rio Corda. There he gained the confidence of the Indians. He reports about an amusing idea of the Indians: “They wanted me to stay with them and become their chief. The main reason must have been the thought, that they could have their revenge on the hated Brazilian settlers, with the help of a strong German and his weapons. In the beginning, I didn’t think they were serious. But when I started to leave, they wouldn’t allow it, and, to en- force their arguments, they brought me a pretty Indian girl who couldn’t have been more than 11 years old. They took no notice of my objections. I was in a real fix, be- cause I already had sent my servant and my beasts of burden back to Barra do Corda. Only when I told them about my wish to see my old parents just one more time, they agreed to let me go.”4

4.2. Guajajára In October 1924, my father went from Barra do Corda by horse to Grajahú and got to know the villages of the Guajajára Indians, a Tupi tribe of about 1.500 souls at that time, deep in the rainforest. Later, he sailed down the Rio Grajahú to the nearby Kreapimkataye, a Krãn-tribe of the Tymbiras, who spoke a language that belongs to the Gê linguistic family.

3 See Snethlage (1927). 4 See Snethlage (1930). 92 SNETHLAGE

4.2. Apinaye In early December 1924, he was back in São Luiz, only to travel on to the state of Ceará immediately to continue his work. Via Ipiapába, Ceará, and Arára he came back to Par- naíba. From there, after having made a short trip to Deserto, up the Rio Parnaíba - about 540 km – he went via Teresina, capital of the state of Piauí, which had about 20.000 inhabitants, to Amarante/São Francisco, situated on the border between Piauí and Ma- ranhão. Then, after a short stop to collect ornithological objects again, he went on to Floriano, Urrusuhy, Inhumas and Tranqueira. In October 1925, he went via Goiás and Certeza down the Rio Tocantins. Initially, he used a raft he had built himself, but later on he hitched a ride from some merchants who had a much bigger one. He passed Pedro Afonso and Carolina, which was occupied by revolutionary Brazilian soldiers soon after his arrival. They were, however, disciplined and well behaved towards him and after their departure, in November 1925, my father travelled on into the region of the São Antonio falls. There he came into contact with the Apinaye (or Apinagés), a Krãn-tribe of the Gê linguistic family, who used to pronounce a “g” instead of the “k” as pronounced by the Maranhese tribes. Because of a series of malarial attacks his work got interrupted. He had to wait two months for the next steam ship, with indigestible food and no quinine, no medical care, lying in an always humid native cabin between sacks filled with salt. Finally in March 1926, he was brought back to Carolina, where a German settler looked after him until he was healthy enough to return. In the middle of April 1926 he arrived again in Pará.

5. The expedition of 1933-1935 On his second expedition, from 1933 until 19355, my father travelled to the region of the Itenes/Guaporé river. Its tributaries and surrounding region still offered many op- portunities to obtain new ethnographical information. Neither the Swedish explorer Erland Nordenskiöld nor the Brazilian General Rondon, nor the very first travellers, missionaries and explorers had been able to visit every corner of this region. This was the reason that my father met many tribes who were not yet known to Western civilisa- tion. In the middle of July 1933, my father arrived in Pará and on the 10th of August he was in Porto Velho, which is the starting point of the Madeira-Mamoré railway. At km 151 of this railway he visited the stone drawings. In early September he reached the starting point of his expedition, called Campamento Komarek. The so-called “savage” Indians of the Moré tribe were so angry about the arrival of yet another white man, that they burnt down the farm workshop during the following night. Thereafter relationships were difficult, and my father needed an additional four weeks to contact the Moré.

5.1. Rio Cautario, Bella Vista, Rio Mequens From that time onwards my father spent three more months of exploration among the Moré and the Itoreauhip peoples on the Bolivian side of the border. When finally the

5 See Snethlage (1937b).

LIFE OF DR. EMIL HEINRICH SNETHLAGE 93 research permission from the Brazilian Government reached him, he was allowed to enter Brazilian territory. On Christmas day of 1933 he started off for the Rio Cautario to visit the Kumaná-Indians living on its headwaters. It was the day, - as he heard three months later -, that his baby daughter, who was born one day before, had died. In February 1934 he met the Pauserna in Bella Vista and conducted excavations in Caféthal (Piso firme). In early March he arrived at the Chiquitano-tribe in Pernambuco, an already westernised tribe. In the middle of March 1934 he set out for the villages of two chiefs of the Ma- kurap, called Uaikuri and Guata. About two weeks later he continued to Serra de Al- lianza, where he did excavations again. Thereafter he went up the Rio Mequens to the Amniapä tribe in the village of Tapuawa. He travelled mainly alone by canoe, some- times with Indian rowers and guides. In May 1934 he reached the tribe of the Guaratägaja and went back down the Rio Mequens. In the middle of June he travelled up the Rio Branco to visit the tribe of the Arua at São Luiz.

94 SNETHLAGE

5.2. Walking-tour across the rainforest In July my father set out on a walking tour across the rainforest and visited the Ma- kurap, Jabuti, Wayoro, Arikapu and Tupari tribes. In the middle of August he travelled down the Rio Branco and back to the Rio Cautario. In October, he was with the Moré and Itoreauhip Indians again. Finally, at the end of November, he left the Campamento Komarek in order to return home. In the following sections, something will be said about each of the different tribes6 mentioned here.

5.3. Moré and Itoreauhip The lower part of the Itenez or Guaporé river separates Brazil and Bolivia. The Moré and Itoreauhip lived on the Bolivian banks. Both tribes speak languages that belong to the Chapakura language family. In my father’s time they were distinguishable only by their different languages and different ways to wear their hair: the Moré let their hair hang down to their shoulders, the Itoreauhip bound it in a knot. They wore shirts made of bark, often striped, but the men were usually naked. In festive times they wore lip and ear ornaments, tiaras and feather-crowns, and feather ribbons in many colours on the arms and legs. They had innumerable ribbons and belts, which were painted or decorated with bark. They seldom painted their bodies, but the women loved to mas- sage themselves and their relatives with urucu, a red vegetable dye dissolved in palm oil.

5.3.1. Dwelling The Moré and Itoreauhip used to live in clans of 15 to 70 people in huts with gabled roofs, covered with palm straw. During the times when there were many mosquitos they went to sleep in special cabins which were completely shut off by leaves of patohu and which had an entry that was shut by an interwoven door. They planted bananas, corn, manioc, yams, batatas, , cotton and urucu. They fished by different methods: arrows, weir-baskets and poison vines. In the season when fruits were ripening and the Moré lived nomadically, they hunted for birds and other animals with arrows, which they sometimes shot from small hunting cabins that resembled beehives. They would attach their cotton hammocks to the supporters of their small huts. These huts were covered with patohu-leaves.

5.3.2. Men’s work As members of a warlike people, the Moré men attached great importance to the deco- ration of their round bows and arrows. Bows were strengthened with inner bark and cotton. The arrows of more than one metre length were trimmed with two, three or four half feathers. Depending on their function, the war or the hunt, the arrows had heads made of bamboo knives, wooden saws, bones and pricks taken from stingrays. Hunting and fishing, building the cabins and working in the fields was also mainly men’s work. Men had to look for the best inner bark from the forest, which had to be

6 See Snethlage (1937a).

LIFE OF DR. EMIL HEINRICH SNETHLAGE 95 softened by beating before it could be sewn into bark shirts. They made wooden boats, troughs, seats, and wooden toys for the children. They tidied their working place them- selves, and from time to time they helped the women to chew manioc flour for the pro- duction of chicha.

5.3.3. Women’s work The women’s first and foremost task was to work in the household. This included for example the pounding of corn using longish troughs and millstones or mill sticks, the peeling of manioc roots with wooden knives, grinding them down by rubbing with a thorny piece of a paxiúba palm root, washing the mash, filtering it through a stick mat, roasting it on an earthenware dish until it becomes “farinha”, and baking flat cakes out of it. Of course the women always had to take care of the children, as well as to make pottery, to pluck cotton, to spin and to knot hammocks. Thinner fibres were used to braid arm or leg ribbons on simple weaving frames.

5.3.4. Games, music, dance Some noteworthy games were: beating a shuttle made of corn leaf; playing a buzzing disk; string games. The Moré and Itoreauhip had many musical instruments. It was cus- tomary to blow on a simple reed whistle when passing by someone’s house. There were many different types of whistles and flutes: a simple long flute or a traverse flute, and a unit of two or more reeds for playing different notes. Trumpets made of calabash were quite popular, as well as rattles made of bottle gourds, filled with seeds or several small calabashes. Dances were accompanied by beating on palm leaves or (with a stick) on bottle gourds. Other dances were accompanied by a certain rubbing instrument7, flute music or singing. All dances seemed to represent events from legends.

5.4. Chapakura-language About the Chapakura language family my father wrote:8 “In 1913 Créqui-Montfort and Rivet posited in the Journal de la Société des Americanistes the existence of the Cha- pakura language family on the basis of the available material. At that time the languages of the following tribes belonged to it: Chapakura or Huaci, Kitemoka, Pavumva (Huan- yam), Napeka, Iten, and, with reservation, perhaps the Rokorona and Muré (or Murä [Nimuendaju 1925]). This was based on the (in my father’s time, R.S.) not yet published vocabularies by d’Orbigny at the Bibliothèque Nationale in Paris, and the already pub- lished vocabularies by Cardús and Hasemann. It was enough to hypothesise the exis- tence of this language family. Later on Nimuendaju published more vocabularies, which expanded this group with 3 more tribes: the Tora, Jain and Urupá (Nimundaju 1925). Consequently, the re- gion of this language family was extended far into the north. Erland Nordenskiöld, un-

7 Heulkuye, in Snethlage (1937a:4), which is probably based on a Portuguese expression. Also Reib-Idiophon in Snethlage (1939:12). 8 Handwritten note from the legacy of E.H. Snethlage, in the possession of Dr. Rotger Snethlage, Aachen. 96 SNETHLAGE fortunately, published only isolated expressions9, but it may be that he has collected more material. The linguistic knowledge gained on my voyage to the Guaporé region allows me to add the Kumaná and the Kabixi-Huanyam. At the same time, the vocabularies from the Abitana-Huanyam (on the Miguel) and from the Iten (who in reality represent at least 2 tribes: Moré and Itoreauhip) were enlarged. Therefore, it is now possible to evaluate critically the material collected by Créqui-Montfort and Rivet. Even though I didn’t obtain coherent texts, the typical character of the Chapakura languages - I would prefer to call them Huanyam languages – has become more clear.”

5.5. Kumaná Also the Kumaná, who live between the middle part of the Cautario river and Rio São Domingos, belong to the Chapakura tribes. However, culturally they were very differ- ent from the Moré and Itoreauhip. The Kumaná lived in big oval-shaped huts. Only the women were naked, but when they had visitors or festivities, they wore their bark shirts that were painted with remarkably beautiful patterns. The men were very shy and let themselves be seen only in their bark shirt wearing a belt. Even bark jackets were made. Broad ribbons of bark with fringes were worn around the arms and legs, and decora- tions of the face and head were very carefully applied. The arrows were longer than the arrows of the Moré. The heads of the bird arrows consisted of tapir teeth attached with wax. The bark beaters were round, without edges. The spindle’s flywheel was made of bottle-gourd skin instead of fruits or cork. The women did not use a mortar for pound- ing flour but used broad planks of wood cut from the root of a big tree from the rainfor- est. When they danced, they marched one behind the other in a circle, accompanied by singing, rattling sounds and the deep sound of a gourd trumpet.

5.6. Abitana-Huanyam In the cultural sense, the Kumaná were somewhat different from the Abitana-Huanyam, who lived close to the Rio São Miguel and who also spoke a Chapakura language. One of the main differences is the possession of poisoned arrows and blowguns, but there were also many differences of less importance. None of the Huanyam women in those days wore European clothing. A woman would wear a big lip-pin of quartz, which lent her the dignity of married status. About musical instruments I should mention especially the trumpet made out of the thighbones of enemies. The dance of the Huanyam was also the circle-dance.

5.7. Makurap In the Rio Branco region, there were no Chapakura-speaking tribes, but rather many tribes with Tupi-elements in their languages. According to my father, these were not closely related to one another. The Makurap, a tupoid tribe, dominated all its neighbours in a cultural sense. They were divided into clans of patriarchal law, and they

9 See E. Nordenskiöld (1924).

LIFE OF DR. EMIL HEINRICH SNETHLAGE 97 gave themselves the names of plants or animals. They believed in the existence of two good gods and a bad one, Choari, master of the ghosts and of the souls of the dead. For their cult, they used an altar of mats that were usually painted. The altar stood opposite the entrance in the big cone-shaped house. During the different ceremonies and healing sessions, the magician used a magic rattle, a magic plank, several medicinal herbs, a snuffing-reed, a magic feather, and from time to time other things. Both sexes used to be naked except for the hip belt made out of seeds. Other body ornaments were: bracelets painted red with urucu, woven in the appropriate size on a round piece of wood; necklaces made of seeds or polished shells; nose ornaments made of reed or of a feather-trimmed stick; and eardrops made of polished shells. Men always wore a penis-cover. Men had to weave baskets and mats, women had to manufacture calabashes and carrying nets of tucum fibres. Cotton was spun in a Bakairi manner, but it did not play a very important role in the material culture. Bows and arrows resembled those of the Huari, about which Erland Nordenskiöld reported. Dancing consisted of fast steps to and fro, in accord with the rhythm indicated by bamboo trumpets or by an instrument that consisted of nine different flutes in a row, attached to one another with wax. While dancing, the Indians often used to put a hand on the shoulder of the man ahead.

5.8. Arua The Arua are also a tupoid tribe. A number of them had already been rallied on an In- dian post and wore European clothes. Their original culture had already been over- whelmed by Makurap culture.

5.9. Wayoro The Wayoro, who were already then few in number, had also adopted the Makurap culture. Their language was a mixture of Makurap and other .

5.10. Jabuti and Arikapu Also the Jabuti and Arikapu were under strong cultural influence of the Makurap. Their completely different languages contain many elements from the languages of the Gê tribes of eastern Brazil.

5.11. Tupari At that time, the Tupari, also a tupoid tribe, were rather independent. They did not have a mat altar like the Makurap, but instead used snuff that played a much bigger part in their magic ceremonies than it did among the other Rio Branco tribes. In their material culture many things looked very different from those of the other Indians: bows, ar- rows, stools, the way to cover the genitals, most of the objects for personal decoration, the spindles and the musical instruments. Here there was a long bamboo flute with four finger holes. In the rainforest there were clearings where they cultivated several types of 98 SNETHLAGE vegetables unknown to other Rio Branco tribes. Beetles were bred in the pulp-mash of chicha, for their larvae were a well-appreciated delicacy and a welcome addition to the manioc-bread. Flint axes were used. The chicha dance was quite similar to the dances of the other Rio Branco inhabi- tants. The flute dance, however, was only done by two so-called chiefs who were al- ways at the same distance from each other.

5.12. Amniapä and Guaratägaja In their material and social culture, the Tupari were closer to the Amniapä (or Mam- piapä) and Guaratägaja from the upper Mequens region and the region of the Pi- menta Bueno tributaries, which have their sources on the same plateau. At least in my father’s time these Indians, like the Tupari, were cannibals from time to time and used to eat their enemies or disloyal members of their own tribe. The Amniapä’s and Guaratägaja’s whole life was filled with ceremonies. Guests walked into the village with their bows and arrows at the ready. The village consisted of several conically shaped huts resembling beehives that stood around a cleared yard. After ceremonial speeches and after having drunk a big calabash of chicha, the guests and the hosts were able to get together in a less formal way. The hosts were generous in their hos- pitality and competitive games were played. A game that required great skill in- volved a rubber ball that could only be touched with the head. Every participant con- tributed arrows or decorations. The games were counted with grains of corn. After the games, the men would use ceremonial snuff, while women had to stay in the back of the huts. Women were only allowed to appear again after the blessing of the food and at the beginning of healing ceremonies. Dancing, however, only began in the evening, lasted the whole night and ended at dawn. Soon thereafter, the guests would take their leave with sad speeches and tears in their eyes. In the Mequens region, men and women wore very different ornaments, in com- plete contrast to the people of the Rio Branco region. The Guaratägaja, but not the Am- niapä, wore a genital cover that was different among the diverse groups. In addition, they wore a loincloth of buriti palm fibres, as well as rich body ornaments consisting of shoulder chains, necklaces, bracelets, and facial ornaments. The women wore numerous necklaces, made mainly of seeds. Their chains made of entire shells were characteristic, and they wore nose and ear ornaments that were quite different from those of the men. Men used to manufacture their ornaments themselves, except for the ribbons. They scratched or painted on their snuffboxes - some of them had patterns like those in the region of the Xingu headwaters. They also produced mask tops provided with remark- able teeth cut out of palm fibre. The jawbones of , knives of agouti teeth, and pebbles were used as tools. Palm fibre skirts and baskets were also made by men. Their musical instruments were calabash trumpets, pan flutes, four-hole flutes, bone flutes and rattles.

LIFE OF DR. EMIL HEINRICH SNETHLAGE 99

5.13. Pauserna-Guarayu The Pauserna-Guarayu had completely lost their original culture, with the exception of a few of its aspects, such as certain pieces of household and furniture. Hammocks, for example, were made to sell. Also pottery, although in a much simplified form, was still made by the women. The Pauserna mixed pulverised pottery fragments with fresh clay for creating hard ceramics. The old ornaments were nearly all lost.

5.14. Chiquitano The Chiquitanos, christianised already by Jesuits, had completely lost their original culture. Only in a few traditions and maybe in some dances did the old spirit still per- sist.

6. Collections My father’s principal collection with all its unique objects was left behind in Brazil. Its present whereabouts are unknown to me, perhaps in Rio de Janeiro. An old register, which was dated February 1935 during the second expedition, counted about 130 ob- jects and ceramic fragments, which were given to the Museu Nacional in Rio de Ja- neiro. The register mentions objects from all the tribes discussed above and includes 79 ceramic pieces from Canindé. The duplicate collection was given to the Berlin Völk- erkunde Museum. Nearly 90% of it survived the Second World War intact. Only the objects from the excavations were completely lost to science. Of the 36 objects listed in 1928, which had been brought back to Berlin, only 8 examples are still in existence, all of which belong to the Canella tribe. The collection from the expedition of 1933 to 1935 consisted of 2.353 objects. Of these, 216 were lost during the war, 100 of which were objects from the excavations. The biggest part of the collection, with more than 800 objects, is from the Moré and Itoreauhip Indians. The Abitana-Huanyam are represented with 215 objects. Also the Makurap, Aruá, Amniapä, Wayoro and the Jabuti, are each represented with over a 100 objects. A further 80 objects belong to the Guaratägaja. More than 30 each are attributed to the Pauserna, Kumaná and Arikapu and about 20 each to the Chiquitano and Papamiän. The entire collection is in storage at Berlin-Dahlem. There is no complete exhibition of my father’s South American collections.

7. Scientific diary and word-list During the war, my mother, Dr. Anneliese Snethlage, saved the carbon copies of my father’s scientific diary kept during the expedition from 1933 to 1935 in the Guaporé region, and the vocabularies. After the war professor Rivet of Paris, and professor Gus- inde of Vienna assisted her. In my possession are 1.000 pages from the diary-copy and a few vocabularies with about 500 words each from the Pauserna, Kumaná and Abitana-Huanyam, and about 200 words from the Moré. The other lists are lost.

100 SNETHLAGE

8. Acknowledgements I would like to express thanks to the University of Leiden, professor Adelaar and Mrs. Brijnen, who encouraged my wife and me to present this talk in Warsaw. Perhaps it will be possible during my lifetime to publish the saved vocabularies and - if there is any interest - also the diary.

9. Publications of E.H. Snethlage (incomplete list partially based on Nevermann, n.d.) 9.1. Doctoral dissertation 1923 Beiträge zur Kenntnis der Gattung Cecropia und ihrer Beziehungen zu den übrigen Conocephaloideen. Friedrich-Wilhelms-Universität Berlin.

9.2. Scientific articles in journals n.d. ‘Neue Arten der Gattung Cecropia nebst Beiträgen zu ihrer Synonymik’. 1924 ‘Neue Cecropien aus Nordbrasilien’, in: Notizbl. Bot. Gart. u. Museum Dahlem, Bd. IX (30.12.1924). 1927 ‘Meine Reise durch Nordostbrasilien’, in: Journal für Ornithologie, LXXV, Heft 3, pp. 456 ff. 1930 Form und Ornamentik alt-peruanischer Spindeln, Baeßler-Archiv, Berlin: Dietrich Reimer/Andrews & Steiner. 1930 ‘Unter nordostbrasilianischen Indianern’, in: Ethnologischer Anzeiger, Bd. II, Heft 4, pp. 185-188. 1931 ‘Unter nordostbrasilianischen Indianern (1924)’, in: Zeitschrift für Ethnolo- gie, LXII, pp. 111-205. 1931 ‘Ein figürliches Ikat-Gewebe aus Peru’, in: Weltkreis. 1932 ‘Emilie Snethlage: Chipaya- und Curuaya-Wörter. Aus dem literarischen Nachlaß herausgegeben von E.H. Snethlage’, in: Anthropos, Bd 27, pp. 65-93. 1932 ‘Worte und Texte der Tembé-Indianer’, Tucuman. 1936 ‘Nachrichten über die Pauserna-Guarayú, die Sirionó am Rio Baures und die S. Simonianes in der Nähe der Serra S. Simon’, in: Zeitschrift für Ethnologie, LXVII, pp. 278-293. 1937 ‘Übersicht über die Indianerstämme des Guaporégebietes’, Tagungsberichte der Gesellschaft für Völkerkunde, (Bericht über die II. Tagung in Leipzig 1936), Leipzig, pp. 172-180. 1937 Indianerkulturen aus dem Grenzgebiet Bolivien-Brasilien. Ergebnisse der Forschungsreise 1933-1935, Veröffentlichung der Reichsstelle für den unter- richtsfilm zu dem Archivfilm Nr. B 25. (ca. 1936/38), Berlin. 1939 Musikinstrumente der Indianer des Guaporégebietes, Baessler-Archiv, Bei- heft 10, Berlin: Dietrich Reimer/Andrews & Steiner. 1939 ‘Untersuchung über das Pferdchen-Spiel auf Java und Bali’, Berlin.

LIFE OF DR. EMIL HEINRICH SNETHLAGE 101

9.3. Popular scientific writings and journal articles 1927 ‘Bei den Indianern des nordostbrasilianischen Hochlandes’, in: Illustrierte Zeitung, Leipzig: Verlag J.J. Heber, 169. Band, Nr. 4298, 28.7.1927, p. 144. 1929 ‘Im Indianerdorf’, (Einführung zur Kinderstunde der Deutschen Welle, Rei- sen und Abenteuer, am Do. 10.1.1929, 14.30 Uhr), in: Deutsche WlleJgg, Nr. 1 (4.1.1929), p. 11. 1935 ‘Zwei Jahre in ewigen Wäldern’, in: Leipziger Neueste Nachrichten, Nr. 155, Di, 4.6.1935, 1. Beilage, p. 5. 1935 ‘Zwei Jahre in ewigen Wäldern’, in: Hannoverscher Kurier, Nr. 252 53, So. 2.Juni 1935. 1935 Indianerkulturen aus dem Grenzgebiet Bolivien-Brasilien. Ergebnisse der Forschungsreise Dr. E. Heinrich Snethlage 1933-1934, Führer durch die Aus- stellung im Staatlichen Museum für Völkerkunde, Berlin. 1937 Atiko Y. Meine Erlebnisse bei den Indianern des Guaporé, Berlin: Klinkhardt & Biermann. [review by Stig Rydén in: American Anthropologist, Vol. 40, No. 1, January-March 1938]

9.4. Obituaries 1930 ‘Dr. Emilie Snethlage zum Gedächtnis’, in: Journal für Ornithologie, LXXVIII, Heft 1, pp. 122-134. (with photo). 1932 ‘Erland Nordenskiöld (+ 5.7.1932)’, in: Zeitschrift für Ethnologie, 64. Jgg. p. 368. 1932 ‘Theodor Koch-Grünberg (+ 9.4.1932)’, in: Deutsche Allgemeine Zeitung, Do. 7.4.1932. n.d. ‘Robert Lehmann-Nitsche’, in: Archiv für Anthropologie, N.F. Bd. 24, Heft 3-4, p. 275-278. n.d. ‘Konrad Theodor Preuss’, in: Archiv für Anthropologie, N.F. Bd. 24, Heft 3- 4, p. 278-280.

9.5. Reviews n.d. Schauinsland, H.: Fragen und Rätsel, Bremen, 1931. n.d. Wassén, Henry: Original Documents from the Cuna Indians of San Blas, Pa- nama. n.d. Baldus, Herbert: Ensaios de Etnologia Brasileira, Rio de Janeiro, 1937 n.d. Rydén, Stig: Archaeological Researches in the Department of La Candelaria. Göteborg 1936. [n.d. numerous other reviews in Archiv für Anthropologie, N.F. and Ethnologischer Anzeiger.]

102 SNETHLAGE

10. References Generaldirektor der Staatlichen Museen 1939 Tagungsberichte Nr. I 2114/39, Berlin C2, Am Lustgarten, den 5. Oktober 1939. Nevermann, H. n.d. ‘Emil Heinrich Snethlage zum Gedächtnis’, in: Archiv für Anthropologie. N.F. Bd. XXVI, Heft 1/2 (offprint without page numbers). Nimuendaju, Curt 1925 ‘As tribus do Alto Madeira’, in: Journal de la Société des Américanistes, n.s.t. 17, pp. 137-172. Nordenskiöld, Erland 1924 Forschungen und Abenteuer in Südamerika, Stuttgardt [Stockholm 1915]. Snethlage, Albert 1982 ‘[Genealogie van de familie] Snethlage [Duitse tak]’, in: De Nederlandsche Leeuw, XCIXe, pp. 77-111. Snethlage, Emil Heinrich 1927 ‘Meine Reise durch Nordostbrasilien’, in: Journal für Ornithologie, LXXV, Heft 3, pp. 456 ff. 1930 ‘Unter nordostbrasilianischen Indianern’, in: Ethnologischer Anzeiger, Bd. II, Heft 4, p. 185-188. 1937a Indianerkulturen aus dem Grenzgebiet Bolivien-Brasilien. Ergebnisse der Forschungsreise 1933-1935, Veröffentlichung der Reichsstelle für den unter- richtsfilm zu dem Archivfilm Nr. B 25. (ca. 1936/38), Berlin. 1937b Atiko Y. Meine Erlebnisse bei den Indianern des Guaporé, Berlin: Klinkhardt & Biermann. 1939 Musikinstrumente der Indianer des Guaporégebietes, Baessler-Archiv, Bei- heft 10, Berlin: Dietrich Reimer/Andrews & Steiner.

NOMINAL MORPHOPHONOLOGICAL PROCESSES OBSERVED IN PEDRO DE LA MATA’S ARTE DE LA LENGUA CHOLONA (1748)

Astrid Alexander - Bakkerus Leiden University

1. Introduction The Cholón language is an indigenous North-Peruvian language. The area in which this language was spoken is vast. It reached from the Huallaga River, a tributary of the Marañón River, until the eastern slopes of the Andes, and from Juanjui in the north until Tingo Maria in the south. Cholón formed one language family (Cholonan) together with Híbito. Both languages now seem to be extinct. Fortunately, there is a description of Cholón, the Arte de la Lengua Cholona, written by Fray Pedro de la Mata. Pedro de la Mata belonged to the Order of the Franciscans. He finished his Arte in 1748 in Truxillo, a city located on the north coast of Peru. In 1772, the was copied by another friar, Fray Gerónimo Clota. Clota wrote the copy in San Buenaventura del Valle, a mission on the Hual- laga River. This manuscript can nowadays be consulted in the British Library (De- partment Manuscripts, Shelfmark Additional 25322) in London. Pedro de la Mata’s grammar gives us an overall picture of Cholón. Structurally, Cholón is an agglutina- tive SOV language, and nominal and verbal forms can be composed of a stem and several affixes. Person markers are usually prefixed, and case markers, numeral clas- sifiers, aspect markers and auxiliaries are suffixed. In this paper we shall focus on the nominal prefixes (section 2) and on nominal morphophonological phenomena, such as assimilation (section 3), vowel sup- pression (section 4) and stem alternation (section 5). The grammatical data in this article are taken from Pedro de la Mata’s Arte de la Lengua Cholona. These data - the Cholón vowels, consonants and lexical items tran- scribed - is represented in bold characters. The original spelling, however, is not al- ways maintained. The symbols c, ch, g, i, j, ll, ng̃, qu, u, v, z have been replaced as follows:

c; qu > k c/_i, e; z > s ch > č g/_e; j > h i/$_V$, $V_$ > y ll > ¥ ng̃ > η tz > ts u/$_V$, $V_$; v > w 104 ALEXANDER-BAKKERUS

2. Nominal person markers The Cholón language has the following nominal affixes that mark possession:

1s: a- ‘my’ 2s.M: mi- ‘your’ 2s.F pi- ‘your’ 3s: Ø ‘his’, ‘her, ‘its’ 1p: ki- ‘our’ 2p: mi- ... -ha ‘your’ 3p: i-, či- ‘their’

These nominal person markers are prefixed to nouns, except for the plural marker - ha, which is suffixed. The forms which end in i are the neutral forms. This final i can assimilate with a stem vowel and change into e and u, as we shall see in section 3. Consequently, mi-, pi-, ki-, i- and či- can also appear as me-/mu-, pe-/pu-, ke-/ku-, e- /u- and ču- respectively. The form če- has not been found in nominal forms (cf. sec- tion 3). The second person singular presents an interesting case, because it makes a dis- tinction in gender. The form mi-/me-/mu- is used to indicate a male person, and the form pi-/pe-/pu- is used to designate a female person. This distinction in gender is restricted to the second person singular. The second person plural possessive marker is a discontinuous morpheme that consists of two elements: a second person marker mi-/me-/mu- and a plural marker - ha. This plural marker is suffixed to the stem of the noun. The third person plural has two forms: a vocalic form i-/e-/u- and a consonantal form či-/ču-. The former is used, if the stem begins with an alveolar t, s, n, l or with a palatal č, š, ñ, . The latter is employed before an initial p, k, m. In the examples below, the nouns pa ‘father’, kot ‘water’, pana ‘way’, taka ‘bone’ and šaš ‘armadillo’ are employed to illustrate the use of the person marker či- before p, k, m, and the use of the person marker i- before an alveolar, such as t, and before a palatal, such as š. The paradigms below will also show that the nouns pa, kot, and pana have a relational form1 which begins with a nasal. In the paradigms of pa and kot only the third person singular has a relational form. This form begins with the nasal η. The other persons keep the absolute form2 with the stem-initial p and k respectively. The third person singular form of pa is ηuč ‘his father’ (a irregular form), the third person singular form of kot is ηot ‘his water’. The relational form of pana is mana. This form can be reduced to mna. mana appears in the third person singular form, mna occurs in the other forms.

1 Relational form: the form which appears in the possessive paradigm of the noun, where it occurs after a prefixed person marker of after the zero-marked third person singular. 2 Absolute form: the citation form that could appear as a lemma in a dictionary. ARTE DE LA LENGUA CHOLONA 105

Absolute forms, relational forms and the irregular third person form ηuč, which is a case of stem suppletion, will be discussed in section 5. In the following table all pos- sessive forms of the Cholon nouns discussed above are listed:

pa kot pana taka šaš ‘father’ ‘water’ ‘way’ ‘bone’ ‘armadillo’

1s a-pa a-kot a-mna a-taka a-šaš 2s mi-pa mi-kot mi-mna mi-taka mi-šaš 3s Ø-ηuč Ø-ηot Ø-mana Ø-taka Ø-šaš 1p ki-pa ki-kot ki-mna ki-taka ki-šaš 2p mi-pa-ha mi-kot-ha mi-mna-ha mi-taka-ha mi-šaš-ha 3p či-pa či-kot či-mna i-taka i-šaš

3. Vowel assimilation From a structural point of view, a possessive person marker can be composed of a vowel (a-, i-, e-, u-), a consonant + a vowel: (mi-, pi-, ki-, či-) or is marked zero: Ø. Person markers composed of a consonant + a vowel (2s, 1p, 2p, 3p, including the third person plural which consists of a vowel only) have a neutral form in i (mi-/pi- (2s), ki- (1p), mi- ... -ha (2p), či- (3p), i- (3p)). This vowel can assimilate with the first vowel of a subsequent stem, provided that the first vowel of that stem is e or u. The neutral i does not harmonize with a and o. The first vowel of a stem can be sup- pressed (see section 4). If vowel elision occurs in a nominal stem, i can assimilate with a second stem vowel e or u. Assimilation of the vowel of a possessive person prefix with a stem vowel, how- ever, is not a compelling process. For instance, instead of the form *me-nen ‘your hand’, we find mi-nen in the Arte. Furthermore, assimilation does not take place in nominal forms, if the stem has an initial p, k and m followed by e, i.e. if the stem of a noun begins with one of the sequences pe, ke and me. Therefore, the third person plural marker či-, which only occurs before an initial p, k and m, cannot harmonize with an e-stem. As a consequence, či- cannot manifest itself as če- in nominal forms, like it does in verbal forms. It can, however, harmonize with a u-stem, and can thus appear as ču- as well. Since the vowel of a possessive marker assimilates with the vowel of a follow- ing stem, vowel harmonization is regressive in Cholón. Furthermore, it is also a “non-contact” harmonization. The vowel that harmonizes (the vowel of the person marker) is generally separated by a consonant from the vowel harmonized (the stem vowel), as a result of which there is no contact between (i.e. immediate adjacency of) both vowels. Even if the noun begins with a vowel, the vowel of person prefixes will not be adjacent to it. This is because both vowels are then separated through insertion of an epenthetic consonant n. In the case of el ‘yucca’ this n emerges even in the (zero-marked) third person (see below and see section 5). 106 ALEXANDER-BAKKERUS

The nouns el ‘yucca’, ¥u ‘peacock’ and pul ‘son’ below will serve as examples for vowel assimilation of the person prefixes with an e- and a u-stem. The paradigm of the noun kešum ‘chin’ shows that a possessive prefix does not harmonize with a noun that has a ke-sequence in initial position, but that it keeps its neutral form in i. The examples also give evidence that in Cholón assimilation is regressive and that it is a non-contact phenomenon.

el ¥u kešum pul ‘yucca’ ‘peacock’ ‘chin’ ‘son’

1s a-n-el a-¥u a-kešum a-pul 2s me-n-el mu-¥u mi-kešum mu-pul 3s Ø-n-el Ø-¥u Ø-ηešum Ø-mul 1p ke-n-el ku-¥u ki-kešum ku-pul 2p me-n-el-ha mu-¥u-ha mi-kešum-ha mu-pul-ha 3p e-n-el u-¥u či-kešum ču-pul

The possessive prefixes thus maintain their neutral form before a noun that begins with pe, ke and me; and before a noun that has an a- and an o-stem. This we have been seen in section 2. The neutral form is obviously also maintained before an i- stem, as can be seen in the paradigm of the noun ¥iš ‘ape’ below.

1s a-¥iš 2s mi-¥iš 3s Ø-¥iš 1p ki-¥iš 2p mi-¥iš-ha 3p i-¥iš

4. Vowel suppression Cholón nouns which consist of more than one syllable can be reduced by vowel sup- pression or elision (cf. the paradigm of the word pana ‘way’ in section 2). Vowel suppression can take place, if a multi-syllabic noun is preceded by a person marker. In such nominal structures, the vowel of the first syllable of the noun, i.e. the second syllable in the possessive nominal structure, can be elided. According to the data in the Arte, suppression generally occurs if the relational forms of the noun have m in initial position, and if the absolute form of the noun begins with the consonant n. Obviously, the vowel is not suppressed in a third person singular form, because this form is not preceded by an overt possessive marker. The first vowel of a multi- syllabic noun is not elided either, if the prefixed person marker only consists of one vowel, such as the third person plural marker i-/e-/u. Since this possessive marker occurs before an initial t, č, s, š, ts, n, ñ, l, (cf. section 2), the third person plural forms ARTE DE LA LENGUA CHOLONA 107 of nouns that have these consonants in initial position are not submitted to vowel elision. Like vowel assimilation, vowel suppression is not compulsory. In the paradigm of the noun pangala ‘turkey (species)’ the first vowel of the relational form mangala is not suppressed, notwithstanding the fact that the stem has m in initial position. The reason why the vowel is not elided could be that, in this case, vowel elision would produce a form with three contiguous consonants: *mngala, which is presumably not allowed in Cholón. Vowel elision can be found in the paradigms of the nouns pakup¥ew ‘passion fruit’ (relational form makup¥ew), puyup ‘bridge’ (relational form muyup), na¥o ‘pupil’ and sala ‘married woman’. The forms below represent evidence that the first vowel of a stem is suppressed, if the stem has m or n in initial position. They also show that the vowel is not suppressed after a mono-vocalic prefix, such as i-. The paradigm of the noun sala is remarkable, because vowel elision does usually not take place in a stem that begins with s, and because the form of the third person plural is irregular. The irregular form i-tala ‘their married woman/women’, instead of a regu- lar *i-sala, could suggest that the noun sala ‘married woman’ could be a relational form derived from an absolute form *yala ‘woman’. The fact is that nouns of which the absolute form begins with the semi-vowel y generally present a stem-initial s and ts in their relational or possessive forms (see section 5). *yala ‘woman’ would then become sala and tsala ‘somebody’s woman’, i.e. a ‘married woman’.

pakup¥ew puyup na¥o sala ‘passion fruit’ ‘bridge’ ‘pupil’ ‘woman’

1s a-mkup¥ew a-myup a-n¥o a-sla 2s mu-mkup¥ew mu-myup mi-n¥o mi-sla 3s Ø-makup¥ew Ø-muyup Ø-na¥o Ø-sala 1p ku-mkup¥ew ku-myup ki-n¥o ki-sla 2p mu-mku¥ew-ha mu-myup-ha mi-n¥o-ha mi-sla-ha s 3p ču-mkup¥ew cu-myup i-na¥o i-tala

5. Stem alternation A number of nouns make a distinction between an absolute form - the pure, free form that can for instance occur as a lemma in a dictionary - and a relational form. The relational form is the form that appears in the possessive paradigm of the noun, where it occurs after a prefixed person marker or after the zero-marked third person singular. Nouns that distinguish between an absolute form and a relational form are those of which the stem of the absolute form begins with a vowel or with one of the fol- lowing consonants: p, k, y, h. The former, the vowel-initial nouns, take an epenthetic n in their relational forms: #V > n-V/$_ , as we have seen in section 3. The latter, the 108 ALEXANDER-BAKKERUS p-initial, k-initial, y-initial and h-initial nouns alter the beginning of their stem. This alternation occurs as follows:

#p > m/$_ : nouns with a stem-initial p change p into the nasal m in their relational forms. The nouns pa ‘father’, pan ‘mother’, pul ‘son’ and puluč ‘husband’ are an exception. They keep the initial p in all forms, except in that of the third per- son singular. However, the third person singular form of pa and pan is ηuč ‘his father’ instead of *ma and ηeč ‘his mother’ instead of *man respectively. The third person singular forms of pul and puluč - mul ‘his son’(cf. section 3) and muluč ‘her husband’ respectively - have a regular relational form with m in ini- tial position.

#k > η/$_ : nouns that have k in initial position also have only one relational form, and this is also the third person singular form. In this form, k is also changed into a nasal, the nasal η (see kot ‘water’ > ηot ‘his water’, section 2). This k > η alternation in the third person singular form could explain the deviant forms ηuč ‘his father’ and ηeč ‘his mother’ in the paradigm of pa ‘father’ and pan ‘mother’ respectively. The fact is, Cholón could have borrowed the deviant forms from Híbito, because in Híbito the words for ‘father’ and ‘mother’ are Cotc and Queec respectively (Martínez Compañón, 1783). These words - transcribed as kotk and keek respectively by Adelaar & Muysken (to appear) - do have a stem- initial k which in Cholón could regularly change into η in the third person singu- lar form. The relational form of kotk would then be *ηotk, and that of keek would be *ηeek. Subsequently, while adopting these forms, Cholón could have changed the stem vowel o of kotk into u, and the final k of both kotk and keek into a palatal č.

#y > ts, s/$_ : nouns which have y- in initial position alter this consonant into s and ts in their relational forms. Note the paradigm of yuč ‘alfalfa’ below:

1s a-tsuč 2s mu-tsuč 3s suč 1p ku-suč 2p mu-suč-ha 3p u-tsuč

#h > s, /$_ : nouns with a stem-initial h change h into s in the third person singular form. Example: hil ‘word’ > sil ‘his word’.

These consonant alternations show that a modifying stem can present different grades: a continuant grade (y, h), a stop grade (k) and a nasal grade (η) (Anderson, ARTE DE LA LENGUA CHOLONA 109

1985: 168). I would like to add a (s) and an (ts) grade to this classi- fication. In Cholón, the continuant grade of the absolute form changes into a fricative and an affricate grade in the relational forms. Furthermore, the stop grade changes into a nasal grade in at least the relational possessive third person form. Note that Guarani also has nouns that distinguish an absolute form from rela- tional forms. These nouns also have a stem-initial alternation or oscillation. They generally have t- in initial position in the absolute form, r- in a relational form (a form preceded by a determiner), and h- in the relational third person form. Note the following examples from (Adelaar & Silva Lôpez, 1986: 25):

tera ‘name’ che-rera ‘my name’ hera ‘his name’, ‘her name’

6. Person markers and morphophonological processes tabled In conclusion, the following table will show the occurrence of the third person plural prefixes i-/e-/u-, či-/ču-, and of vowel suppression and stem alternation in relation with the consonants p, t, č, s, š, m, n, ñ, l, ¥, y. According to the data in the Arte, these are the consonants that can appear in initial position in nouns. Furthermore, these consonants can be preceded by a person marker or appear in a relational form, with the exception of the consonant y. This consonant does not appear stem-initially in a relational form, but it is relevant for the occurence of stem alternation.

p t č k s š m n ñ l ¥ y

i-/e-/u- - + + - + + - + + + + - či-/ču- + - - + - - + - - - - - VS ------+ + - - - - SA + - - + ------+

Table 1: stem-initial consonants and the occurrence of the person markers i-/e-/u, či-/ču-, and of the phenomena of vowel suppression (VS) and stem alternation (SA).

References Adelaar, W.F.H. and P.C. Muysken t.a. The Languages of the Andes, Cambridge: Cambridge University Press. Adelaar, W.F.H. and L. Silva Lôpez 1986 ‘Grammaticaal overzicht van het Guarani’, in: Wampum 7:11-59, Leiden. Anderson, S.R. 1985 ‘Inflectional ’, in: T. Shopen (ed.), Language Typology and Syn- tactic Description, Volume III, Cambridge: Cambridge University Press, pp. 150-201. 110 ALEXANDER-BAKKERUS

Martínez Compañón, B.J. 1978 Truxillo del Peru, facsimile [1783], Volume 11. Madrid: Ediciones Cultura Hispánica del Centro Iberoamericano de Cooperación. Mata, P. de la 1748 Arte de la Lengua Cholona, London, British Library, Ms. Additional 25322, Transcription: A. Alexander-Bakkerus, Leiden University, 1996.

Abbreviations and symbols

Ø zero 1 first person 2 second person 3 third person C consonant F feminine M masculine p plural s singular SA stem alternation V vowel VS vowel suppression # noun boundary $ syllable boundary - morpheme boundary * reconstructed form _ after, before / in the neighbourhood of LAS ESTRUCTURAS FONÉTICAS DE LA LENGUA EMBERA*

Esther Herrera Z. El Colegio de México

1. Introducción En este trabajo presento un estudio instrumental de la lengua embera; los datos provienen de la variante Chamí, hablada en Cristianía, y fueron proporcionados por Dálila Yagarí, a quien mucho agradezco1. El presente estudio tiene como objetivo indagar las pistas acústicas de las oposiciones fonológicas más notorias de la lengua; en particular, se centra en el análisis de las oclusivas y de las aproximantes, y en la determinación del timbre vocálico. El estudio instrumental de estos segmentos no resulta de una elección azarosa; la bibliografía sobre la lengua, consultada directa e indirectamente, presenta un panorama confuso al respecto. En las descripciones proporcionadas por autores como Loewen (1969b), Gralow (1987), Pardo (1988), o los tres estudios contenidos en Llerena (1995), la discrepancia no sólo tiene que ver con el número de segmentos, sino con parámetros como la sonoridad, la corriente de aire y la tensión, entre otros. Si bien la mayoría de los autores coinciden en distinguir tres series de oclusivas, Loewen (1969b), quien estudia el habla del Sambú, perteneciente al embera del Norte, propone, para el embera en general, una triple división que consta de oclusivas fortis, oclusivas lenis y sonoras (/p t k p t k b d /); Gralow (1987) [Apaud. Constenla y Margery (1991:142)] postula, para el Chamí, dos series: sorda-sonora (/p t k b d/); Aguirre Licht (1995), quien también estudia la variante Chamí, divide las oclusivas en sordas, sonoras e inyectivas (/p t k b d  /); Rito Llerena (1995) propone, para la variante occidental de Cristianía, una división que consta de oclusivas aspiradas, oclusivas no aspiradas y oclusivas glotalizadas (/ph h h t k p t k b d /). Por último, mencionaré lo que propone Edgar Hoyos (1995) para quien la triple distinción que encuentra en la variante de Quío consta de oclusivas tensas sordas, oclusivas tensas sonoras y oclusivas laxas (/p t k b d b’d’’/), mientras que la variante Chamí de Risaralda, según el mismo autor, tendría un contraste sordo-sonoro (/p t k b d/). Las divergencias que se desprenden de lo anterior, podrían interpretarse como un simple reflejo de la realidad lingüística de la lengua, esto es, las diferentes variantes del embera tienen distintos inventarios de segmentos oclusivos, situación perfectamente posible en algunas lenguas. Sin embargo, como más adelante veremos, en algunos casos los criterios proporcionados para las divisiones propuestas no

1 La recolección de los datos se realizó durante 1999-2000 en colaboración con Dálila Yagarí, hablante del embera y estudiante de la Maestría en Lingüística Indoamerica del Centro de Investigación y Estudios Superiores en Antropología Social de la ciudad de México. 112 HERRERA resultan del todo convincentes, y la discrepancia persiste para una misma variante, como la Chamí. En relación con los segmentos [-consonánticos], el panorama es igualmente incierto. A excepción de Pardo (1988), ninguno de los autores citados reconoce la vocal /ø/, ni a // como segmentos del inventario. En todos ellos, el inventario de vocales consta de /i e a o  u/ y la serie nasalizada respectiva. La presente investigación arroja, para la variante Chamí, resultados distintos. La evidencia fonológica, aunada al hecho acústico, muestra que en las oclusivas, la distinción pertinente es la de sordo, sonoro e ingresivo; tal distinción tiene que ver con el mecanismo de corriente de aire. Así mismo, permite postular que además de /w /, la lengua distingue la labiodental //, segmento que por su comportamiento fonológico, forma junto con // el grupo de aproximantes. Asimismo, el resultado del análisis acústico de las vocales, revela que el embera tiene dos vocales medias, pero contrariamente a lo que se ha propuesto, esas dos vocales son, la una [+anterior, - redondeada], y la otra [+ central, + redondeada]. Esto da como resultado que el espacio vocálico del embera presente un hueco en la posición media posterior. El presente trabajo está organizado como sigue: en el apartado 2 presento una breve descripción fonológica de la lengua, en el apartado 3 desarrollo el análisis acústico de la oclusivas, las aproximantes y la estructura acústica de las vocales, incluyo una posible explicación del hueco que presenta el inventario de vocales, mediante la teoría de la dispersión debida a Lindblom (1990). Para el análisis acústico se utilizó el programa CSL (Computer Speech Lab) fabricado por Kay Elemetrics Corp; los datos se digitalizaron a 11.025 Hz., con el fin de extraer información acústica hasta alrededor de los 5.000 Hz., y con ello poder calcular el LPC (Coeficiente de Predicción Lineal), y conocer la estructura formántica de las vocales, así como examinar los espectrogramas. La recolección de los datos se hizo en el laboratorio de El Colegio de México; los datos se grabaron directamente en el CSL usando un micrófono Shure.

2. Características generales del embera Siguiendo la clasificación de Loewen (1969b) y la que recientemente proporciona Mortensen (1999), la lengua embera pertenece a la familia lingüística del Chocó. Esta familia se localiza a lo largo del litoral pacífico, desde Panamá hasta Ecuador; consta de dos lenguas: el waunana, conocida también como noanama, y el embera. El embera a su vez se ha dividido en dos grupos: el del Norte y el del Sur, cada uno con sus dialectos respectivos; así el embera catío o dabeida y el embera del norte conforman el embera norte; el embera catío está en Colombia (en la cuenca del río Sinú, al sur del Departamento de Córdoba y en la cuenca del río Sucio, al norte del Departamento de Antioquia); mientras que el embera norte se localiza en Panamá y en el Norte de Colombia. Según datos de Mortensen (1999:2), el embera del norte cuenta con 25.000 hablantes aproximadamente; 10.000 viven en Panamá y el resto en Colombia. El embera catío cuenta con un número no mayor de 20.000 hablantes. El

ESTRUCTURAS FONÉTICAS EMBERA 113 embera del Sur incluye al saija (con 3.500 hablantes aproximadamente)2, al baudó (con 5.000), al tadó (con alrededor de 1.000) y al embera chamí (con 3.500), variante esta objeto del presente estudio3. En su fonología, los procesos segmentales más notorios incluyen un profuso proceso de armonía nasal desencadenado por la serie de vocales nasales. Este proceso ocurre en distribución complementaria en dos direcciones: de izquierda a derecha, una vocal nasal desarrolla una consonante nasal homorgánica cuando está antes de cualquier consonante [obstruyente]; de derecha a izquierda, cualquier vocal nasal nasaliza los segmentos que no son consonantes plenas, es decir, las vocales las semi-consonantes y las dos aproximantes / /. Presenta también un proceso de posteriorización en el cual /k h/ Æ [q ] desencadenado por las vocales /a ø/, y una armonía vocálica en ciertos sufijos. El sistema fonológico que propongo es el que aparece en 1.

(1) Consonantes

Lab Alveolar Pal Velar Laringea Plosivas p t k b d Implosivas   Africadas t Fricativas s h Aproximantes   Nasales m n (  ) Vibrantes r Semiconsonanes (w) (j )

Vocales orales y nasales

anterior central posterior [+alto][-bajo] i  u [-alto][-bajo] e ø [+bajo][-alto] a

Como se desprende de 1, con base en el mecanismo de corriente de aire, el embera establece una doble distinción en las oclusivas, esto es, oclusivas producidas con una

2 Las cifras para el embera del sur provienen del Leading Global Language Portal (http://www.your dictionary.com). 3 Una de las cuestiones que se han discutido es la posible relación de los grupos de la familia Chocó con las tribus brasileñas; así lo sugirió Nordenskiöld (1929). Tal suposición se motivó en la existencia de préstamos del español para referirse a ciertos animales marinos como el delfín y la ballena; en el hecho de que las canoas de los Chocó son para río y no para mar y en un parecido en motivos mitológicos. (Apaud. Loewen 1969a:239). Recientemente, Constenla y Margery (1991:172) proponen una relación con los Chibchas con base en un conjunto de cognados.

114 HERRERA corriente de aire pulmonar egresiva (plosivas), y sonidos oclusivos producidos con una corriente de aire glotálica ingresiva (implosivas). Asimismo, presenta una doble distinción respecto a la sonoridad. Las oclusivas sordas se distinguen en tres puntos de articulación; en el caso de las oclusivas sonoras e implosivas, en esta variante, el sistema no incluye segmentos con punto de articulación velar. En la serie de fricativas la lengua no opone sordas y sonoras. El sistema vocálico presenta una distinción oral-nasal con igual número de vocales en las dos series. Una de sus características es la notable ausencia de la vocal media posterior redondeada. Los segmentos que aparecen entre paréntesis en 1 tienen un estatus fonológico dudoso, ya que es posible derivar las dos semiconsonantes de las vocales correspondientes, cuando éstas no ocupan la posición nuclear de la sílaba tónica; la nasal palatal no contrasta y se puede interpretar como resultado de la nasalización de la vocal /i/. Respecto a los segmentos / / mostraré que se trata de aproximantes. Tal afirmación se apoya en el comportamiento que tienen en varios procesos fonológicos que veremos más adelante, así como en su estructura acústica. En 2 se dan ejemplos que ilustran los distintos contrastes4.

(2) Consonantes:

pana ‘puente’ ti ‘este’ bana ‘varios’ si- ‘golpear con la mano’ kea ‘tipo de árbol’ sø ‘pulmón’ pea ‘en diagonal’ hø ‘fruta’ kiu ‘carne con nervio mø ‘piedra’ tiu ‘apuntar con algo’ nø- ‘hacer el amor’ aba ‘amigo’ eda- ‘sacar, extraer’ aa ‘un(o)’ ea ‘dentro’ atau ‘cargado en la espalda’ puru ‘rojo’ adau ‘tomado con la mano’ puu ‘pueblo’ ra ‘manteca’ wa- ‘ofrecer’ ru ‘cola’ ba- ‘quemar’ ii ‘nuevo’ a- ‘salir, irse’ aaa ‘maduro’ ne ‘oro’ ekar ‘espalda’ ne- ‘llegar’ ban ‘agua’ apis ‘¡estate quieto!’

4 Las realizaciónes de las palabras con vocales nasales como /kiu/ ‘carne con nervio’, /tiu/ ‘apuntar con algo’ y /a-/ ‘salir, irse’ son [kiu ], [tiu], y [a] respectivamente. Véase en 3.1 el patrón que presenta la nasalización.

ESTRUCTURAS FONÉTICAS EMBERA 115

Vocales:

i ‘labio’ ta ‘semilla’ e ‘canasto’ u ‘cerbatana’ ta ‘sal’ t ‘piojo’ ø ‘camino’ tø- ‘lavar’  ‘tos ferina’ e ‘piel’ tø- ‘explotar’

De los datos anteriores podemos hacer algunas observaciones adicionales. El ítem de ‘oro’ y el de ‘llegar’ muestran que la distinción oral-nasal en las vocales se conserva después de consonante nasal; en la distribución de las vocales, en la secuencia CV, las vocales nasales no ocurren después de ninguna consonante implosiva; el contraste entre vibrantes sólo se da entre vocales, además no hay palabras que se inicien o se terminen con la vibrante simple (aproximante). Asimismo, los únicos segmentos consonánticos que pueden aparecer en coda son las coronales /r n s/. A final de palabra, esas consonantes ocupan la posición de coda debido a la elisión de uno o más segmentos. Así, [ekara] ‘espalda’ alterna con [ekar]; [ban] ‘agua’ alterna con [banja]; [apis] ‘¡estate quieto!’ alterna con [apise]...

3. Las consonantes oclusivas e implosivas Desde un punto de vista articulatorio una oclusiva como /p t k/ consiste en un momento de cierre, seguido por la soltura de ese cierre; en términos acústicos, la fase de oclusión, cuya energía es de cero, se traduce por un silencio, la soltura por una pequeña pero abrupta explosión. En una oclusiva sonora como /b d g/ la parte que corresponde al cierre presenta un número pequeño de componentes armónicos de baja frecuencia, mismos que conforman la llamada barra de sonoridad. A diferencia de las plosivas, los sonidos implosivos no abundan en las lenguas del mundo, aunque son comunes en algunas áreas geográficas como en África Occidental, particularmente en las lenguas de Nigeria, como el degema y el hausa, y el dahalo de la familia Cushitica (Costa Norte de Kenya). Según Maddieson (1984), estos sonidos se encuentran sólo en el 13% de las lenguas del mundo. Articulatoriamente, se producen gracias a la corriente de aire que se crea por un movimiento hacia abajo de la glotis. Comparando los factores aerodinámicos durante la producción de una oclusiva sonora y una implosiva, se observa que en las oclusivas sonoras la amplitud de la vibración de las cuerdas decrece o permanece estable durante la fase de cierre; mientras que en una implosiva sucede lo contrario: hay un incremento gradual de la amplitud. Este aumento es el correlato acústico del tipo de corriente de aire involucrado durante la producción de una implosiva. Lo anterior se observa claramente en el oscilograma del par de palabras emberas /aa/ ‘amigo’ y /aba/ ‘uno’ que aparecen en la figura 1.

116 HERRERA

Figura 1. Oscilograma de /aa/ “amigo” (izquierda) y de /aba/ “uno” (derecha).

Como se muestra en la zona marcada con las pequeñas flechas, la fase de cierre de la implosiva de “amigo” presenta un notorio aumento en su amplitud, mientras que en la oclusiva sonora de “uno” la amplitud decrece al final del cierre, justo antes del momento de la liberación del aire y el inicio de la vocal siguiente. Este aumento de amplitud durante la fase del cierre es una de las pistas acústicas que acompañan los sonidos implosivos. Así ocurre en la lengua degema analizada en Lindau (1984), y también en la lengua tsou reportada en Wright y Ladefoged (1994). La disminución de la amplitud en la oclusiva sonora es resultado de la disminución de la corriente de aire que pasa por la glotis. La corriente de aire disminuye debido a que al estar cerrada la salida de aire por la oclusión, la presión supralaríngea se incrementa. En las implosivas, el aumento de amplitud se debe al aumento de tamaño del tracto vocálico ya que al bajarse la laringe, se baja también el cuerpo de la lengua. Esta mayor cavidad impide que aumente el volumen de la presión y por ello es posible incrementar la amplitud a lo largo del cierre. El estudio acústico de las consonantes del embera revela tres situaciones de manera constante: para las oclusivas sordas hay cero amplitud; disminución o mantenimiento de la amplitud para las sonoras y aumento notorio en las implosivas. En la figura 2 tenemos un ejemplo con la bilabial de /sipe/ ‘cuchara’ que completa la caracterización acústica de la tres series. A la luz de esta evidencia, veamos más de cerca las distintas propuestas que se han hecho. Si la distinción fortis-lenis-sonora fuera aplicable para la variante Chamí, como lo plantea Loewen (1969), se esperaría encontrar algún correlato acústico relacionado con la tensión, como por ejemplo, una retención empecinada del cierre

ESTRUCTURAS FONÉTICAS EMBERA 117

Figura 2. Oscilograma que muestra la [p] en /sipe/ “cuchara”. específica en la producción de las fortis, o bien algún tipo de proceso como la sonorización de las lenis en contexto intervocálico, así sucede, por ejemplo en el coreano (Kagaya 1974, Arellanes 2001), en el danés (Fisher-Jørgensen 1969) y la lengua australiana jawon (Jaeger 1983) que han sido analizadas como lenguas con la distinción fortis-lenis en su sistema. Los datos de 2 revelan que en embera la distinción sordo-sonoro se presenta a principio de palabra y entre vocales. Aún más, una comparación somera entre las palabras que da Loewen y las de mi corpus, indica una correspondencia sistemática: los segmentos que él considera sonoros, corres- ponden a segmentos implosivos en mi corpus; los que transcribe como lenis, corresponden a sonoros y los fortis corresponden a sordos. Esta regularidad se observa en los datos de la siguiente tabla.

Loewen (1969b) Herrera Glosa

FORTIS SORDA p ak a paka ‘vaca’

LENIS SONORA to dø ‘río’ te de ‘casa’

SONORA IMPLOSIVA bi i ‘estómago’ puda bua ‘cabello’

Tabla 1

118 HERRERA

Loewen (1969b) argumenta su distinción en el hecho de que en waunana las fortis se distinguen de las lenis en que tienen una soltura aspirada, mientras que las lenis nunca se aspiran; las sonoras contrastan con las lenis en la “fuerza de la sonoridad”, nos dice este autor. En la variante de estudio, efectivamente las sordas se realizan con un componente de fricción en su soltura, pero se trata de un proceso que ocurre a principio de palabra y en sílaba tónica y es particularmente notorio ante vocales [- altas]. Se podría suponer que en la variante de estudio hay fonemas oclusivos sordos y sordos aspirados, sin embargo no se encontró ningún par de palabras que contrasten en este rasgo. Como vemos en la figura 3, la oclusiva bilabial se realiza sin ningún componente de aspiración en su soltura debido a que en esa palabra el acento recae en la sílaba anterior. Por el contrario en una forma como /øpøa/ ‘iguana’, la oclusiva forma parte de la sílaba tónica. En la figura 3 tenemos el oscilograma y el espectrograma correspondiente a la realización de ‘iguana’ y podemos apreciar claramente la aspiración de la bilabial.

Figura 3. Oscilograma y espectrograma de /øpøa/ ‘iguana’

Respecto a la sonoridad de las implosivas, los estudios tipológicos muestran que son normalmente sonoras (Greenberg 1970; Maddieson 1984). Por otro lado, si bien Aguirre Licht (1995:30) tiene el mérito de identificar que la corriente de aire es ingresiva, su descripción confunde: “Los sonidos inyectivos pueden ser preglotalizados, laringealizados e implosivos”. Efectivamente, Greenberg, de quien se inspira, aclara que las inyectivas pueden tener tres posibilidades fonéticas:

ESTRUCTURAS FONÉTICAS EMBERA 119 implosivas, preglotalizadas y laringealizadas. Esto quiere decir que en las lenguas dichos segmentos pueden tener una de estas tres realizaciones, pero no por ello son, fonológicamente hablando, ni preglotalizados, ni laringealizados. Por fortuna para la fonología, el término implosivo ganó terreno sobre inyectivo, éste último usado inicialmente por Catford (1939). Por su parte, Llerena (1995:210) da una caracterización por demás caprichosa de estos segmentos: “Aunque se ha interpretado que la corriente de aire en las [que él denomina - EH] oclusivas glotalizadas es ‘ingresiva’ en las distintas lenguas y variantes “epera”, la interpretación que aquí se hace es la de su caracter egresivo”. El problema esencial, más allá de usar un símbolo que indica preglotalización, mismo que usa Greenberg, es que no se conocen muchas lenguas que sólo tengan sonoras glotalizadas sin que a la vez tengan la serie de sordas correspondientes, y estas últimas no existen en la lengua. El componente de laringealización mencionado en Greenberg que acompaña la realización de las implosivas efectivamente lo encontramos en la lengua. En la figura 4 se da el espectrograma de la realización de /kua/ ‘cama’. En él se puede ver la laringealización que provoca la implosiva // sobre el inicio de la vocal siguiente; dicha laringealización aparece esporádicamente, la pista acústica constante es, como ya vimos, el aumento en la amplitud.

Figura 4. Espectrograma de /kua/ ‘cama’

Tengo la impresión de que la distinción de Hoyos (tensa sorda, tensa sonora y laxa) está fuertemente inspirada en algún aspecto de la descripción de Loewen; como se sabe, la oposición tenso-laxo es equivalente a fortis-lenis.

3.1. Las aproximantes En las descripciones de la lengua, se ha reconocido la vibrante simple //, aunque no se ha discutido su estatus en el sistema; respecto a la labiodental //, a excepción de Pardo (1988), quien distinguió un segmento distinto, la mayoría de los autores han interpretado este último segmento como /w/. Así lo señalan Constenla y Margery (1991:147) en su acuciosa recopilación y comparación de materiales: “Pardo (1988)

120 HERRERA suele transcribir [v], en Chamí, Catío y Sambú, en la mayor parte de los casos en que los otros autores escriben /w/”. En este apartado, mostraré que / / son segmentos distintos a las demás consonantes de la lengua; para ello me apoyo en la evidencia instrumental y en la fonológica, en particular en su estructura acústica y su comportamiento en dos procesos: la nasalización y la geminación. La nasalización es sin duda uno de los procesos más notorio. En los datos de 5a se ejemplifica uno de sus lados; en ellos las formas fonológicas de la izquierda muestran que una vocal nasal se realiza desarrollando una transición consonántica nasal homorgánica con la consonante si- guiente5.

(5) a apø --> apø ‘vómito’ ikua --> ikua ‘áspero’ øtøa --> øtøa ‘húmedo, para café’ hibusu --> hibusu ‘colibrí’ adau --> adau ‘falda’ aa --> aa ‘agua sucia’ kue --> kue ‘en la nariz’ itur --> itur ‘armadillo’ isisia --> isisia ’comezón’ pina --> *pina ‘ají’ tøra --> tøra ‘anciana’ tøa --> tøa ‘anciano’

Los ejemplos de ‘anciana’ y ‘anciano’ revelan que la vibrante múltiple y la simple tienen un comportamiento asimétrico respecto al proceso de nasalización: la múltiple provoca la aparición de la consonante nasal de transición, mientras que la vibrante simple se nasaliza; el caso de “ají” no dice que las consonantes nasales no permiten que se desarrolle la transición nasal. Los datos de 5b muestran la otra dirección del proceso de nasalización, consiste en la propagación, de derecha a izquierda, del rasgo nasal de la vocal sobre un conjunto de segmentos específicos. Al mismo tiempo nos dicen qué segmentos bloquean esa propagación y la extensión que tiene el proceso.

(5) b ø-a --> øa ‘caminos’ de-atø --> deatø ‘horcón de la casa’ kara-a --> karaa ‘raíces’ usa-a --> usaa ‘perros’

5 Como sabemos, durante la producción de una vocal nasal el velo se baja para permitir el paso del aire por la cavidad nasal; mientras que para una consonante se requiere una obstrucción oral y el velo levantado. En las secuencias de vocal nasal más consonante, el tiempo para efectuar dicho cambio de postura en los articuladores puede provocar que la obstrucción oral se realice antes de levantar el velo y con ello se genera una consonante de transición entre estos dos estados articulatorios. Véase Ohala (1983).

ESTRUCTURAS FONÉTICAS EMBERA 121

usa-a-ba --> usaaba ‘perros’ (sujeto) dua- esa --> duaesa ‘centro del patio’ pada-a --> padaa ‘plátanos’ ibana-a --> ibanaa ‘pájaros’ hua-a --> huaa ‘manos’

Como indican las formas después de las flechas, el conjunto de segmentos que se nasaliza incluye vocales, semivocales, la vibrante simple y la aproximante labiodental. Los segmentos que bloquean el proceso son los que a su vez permiten la creación la nasal de transición que vimos en 5a. Asimismo, la formas ‘horcón de la casa’ y ‘centro del patio’ ejemplifican las dos direcciones del proceso. La extensión no va más allá de la sílaba, ya a la derecha, ya a la izquierda. La evidencia anterior muestra que / / son segmentos que tienen un comportamiento diferente de los otros segmentos [-silábico]. Una prueba adicional de ello son los datos que aparecen en 6 en los que ilustra un proceso morfológico regular, en el cual se prolonga la consonante que está más a la derecha de la base adjetival. El morfema, que consiste en un rasgo [-continuo], agrega el significado de ‘muy’ a las bases.

(6) Formas base Formas sufijadas Glosa hipa --> hippa ‘recto, derecho’ ti-kka --> tikkka ‘reseco’ (piel) døa --> døa ‘transparente’ ti-tørøa --> titørrøa ‘blanco’ ti-døea --> tiddøea ‘torcido’ ti-paaa --> tippaaa ‘verde’ deseøa --> detseøa ‘largo’

Nuevamente vemos un comportamiento similar de / /, es decir su resistencia a geminarse, particularmente en los tres últimos ejemplos. Por otro lado, la comparación de la estructura acústica de // y de /w/ sugiere que se trata de segmentos distintos. En la figura 5a el espectrograma de [kukwara] ‘medio amarillo’ revela que se trata de un diptongo; uno de los indicios es la transición negativa y rápida que la [w] provoca sobre el el F2 de [a]; mientras que en [tiii] ‘nuevo’ la aproximante labiodental presenta zonas de resonancia bien definidas revelando una constricción menor a de una consonante y mayor a la de una vocal; además, como se observa en la parte marcada, el segmento presenta mayor estabilidad; la transición negativa que provoca en el F2 de las dos vocales altas que lo circundan revela que su punto de articulación es labiodental.

122 HERRERA

Figura 5a. Oscilograma y espectrograma de [kukwara]‘medio amarillo’

Figura 5b.Oscilograma y espectrograma de [tiii] ‘nuevo’.

ESTRUCTURAS FONÉTICAS EMBERA 123

3.2. Estructura formántica de las vocales. Si bien en el caso de las consonantes hay discrepancia entre los autores, para las vocales del embera parece haber consenso: en todas las descripciones está ausente la vocal central redondeada, en su lugar se menciona una /o/. Sin embargo, el estudio instrumental nos dice que la lengua tiene /ø/. Al igual que las demás vocales de la lengua, esta vocal se determinó con base en el estudio de su estructura acústica. Para ello, se calculó el LPC de los dos primeros formantes de cada vocal en su parte estable; se tomaron 30 casos de cada una de ellas en posición tónica; posteriormente se calculó la media de F1 y la de F2-F1 (Ladefoged y Maddieson 1990). Las medidas de F1 proporcionan información respecto a la altura; el resultado de F2-F1 nos indica anterioridad y posterioridad. Los resultados de esas mediciones se trasladaron a la carta de formantes de la gráfica 1. Las elipses indican la desviación estándar de cada vocal.

(F2-F1)

3000 2500 2000 1500 1000 500 200 Hz

(F1) i u u 300

400

ø e 500

600

700

a 800

900 900 600 300 0

Gráfica 1. Valores promedio de F1 (en la abscisa) y de F2-F1 (en la ordenada) y desviación estándar

Los sistemas como el anterior que presentan un hueco en la zona posterior media no son raros, se encuentran en noruego, wolof, y en otras lenguas documentadas por Maddieson (1984) que tienen /e/ sin tener la posterior correspondiente. Para una teoría como la de Lindblom (1990) que intenta predecir los sistemas vocálicos con base en la dispersión, el criterio importante es el máximo contraste perceptual en los sistemas. Un sistema con máxima dispersión acústica es el que ocupa los puntos más alejados del espacio vocálico. Esto da como resultado tres vocales periféricas /i u a/. A partir de esta máxima dispersión, si hay un número diferente de vocales anteriores que de posteriores, la tendencia es que haya más vocales anteriores que posteriores,

124 HERRERA como sucede en el embera. Según la base de datos de UPSID (Maddieson 1980), en que se retoman 317 lenguas de 20 familias, el 30% de los sistemas asimétricos tienen preferencia por las vocales anteriores. La teoría de Lindblom también predice que entre las vocales altas, por constituir una zona de máxima dispersión, es donde puede aparece una vocal adicional. Cuando se considera el espacio vocálico hacia abajo, se reduce la dispersión, por ello, las lenguas reducen las distinciones de las vocales no periféricas en la medida en que se va de las zonas altas a las bajas. El embera tiene una vocal baja, dos medias y tres altas. En la base de datos del UPSID, en el 50% de los casos con una vocal no periférica alta, hay una vocal alta central, el otro 50 % se reparte entre la /ü/ y la //. En este sentido, al tener la vocal central, el embera conserva sus zonas de dispersión bien diferenciadas. Las vocales centrales del embera reflejan una tendencia en las lenguas por privilegiar, para los contrastes, la zona media antes que la de los extremos.

4. Conclusión La evidencia fonológica, entrelazada con la evidencia acústica, ha permitido identificar las distinciones pertinentes de la lengua embera. Como se demostró, éstas tienen que ver con la corriente de aire y la sonoridad. El sistema vocálico presenta un hueco que encuentra una explicación con base en las tendencias acústicas hacia un aprovechamiento del espacio vocálico. Asimismo, la evidencia instrumental, aunada a la fonológica, ha permitido identificar una aproximante labiodental que no había sido identificada en la lengua. A partir de estos resultados, es posible abordar la riqueza fonológica que tiene en términos de las tendencia universales y los modelos teóricos actuales.

* Esta investigacíon se realizó con el apoyo de CONACyT Proyecto 27598 H.

Referencias Aguirre Licht, Daniel 1995 ‘Fonología del ebera-Chamí de Cristianía’, en: Rito Llerena Villalobos (Coord.), pp. 9-84. Arellanes, Arellanes Francisco 2001 ‘La Oposición fortis-lenis aspirado en las Consonantes obstruyentes del coreano’, en: Herrera Z. Esther (ed.), Temas de Fonética instrumental, El Colegio de México. Catford, J.C. 1939 ‘On the Classification of Stop Consonants’, Le Maître phonétique 65:2-5. Constenla, Umaña Adolfo y Enrique Margery 1991 ‘Elementos de Fonología comparada Chocó’, Filología y Lingüística, Universidad de Costa Rica, XVII:137-191. Fischer Jørgensen, E. 1969 ‘Voicing, Tenseness and Aspiration in Stop Consonants, with Special Reference to French and Danish’, ARIPUC 3:63-114.

ESTRUCTURAS FONÉTICAS EMBERA 125

Gralow, Frances L. 1987 ‘Fonología del Chamí’, Sistemas fonológicos de Idiomas Colombianos III:29- 42, ILV, Ministerio de Gobierno, Townsend, Lomalinda, Colombia. Greenberg, Joseph H. 1970 ‘Some Genaralizations Concerning Glottalic Consonants, Specially Implosives’, IJAL 36 :123-145. Hoyos, Edgar 1995 ‘Estudio embera: en Torno a la Comparación de algunas Variantes’, en: Rito Llerena Villalobos (Coord.), pp. 87-207. Jaeger, Jeri J. 1983 ‘The Fortis/Lenis Question, Evidence from Zapotec and Jawon’, Journal of Phonetics 11:177-189. Kagaya, Ryohei 1974 ‘A Fiberscopi and Acoustic Study of the Corean Stops, and Fricatives’, Journal of Phonetics 2: 161-180. Ladefoged, Peter y Ian Maddieson 1990 ‘Vowels of the World’s Languages’, Journal of Phonetics 18 : 93-122. Lindau, Mona 1984 ‘Phonetic Differences in Glottalic Consonants’, Journal of Phonetics 12: 147- 155. Lindblom, B. 1990 ‘Models of Phonetic Variation and Selection’, Phonetic Experimental Research, Institute of Linguistics, University of Stockholm, XI, 65-100. Llerena Villalobos, Rito 1995 ‘Fonología Comparada de las Lenguas epera de Occidente (Jaidukama) y Oriente (Cristianía Alto Andágueda)’, en: Rito Llerena Villalobos (Coord.), pp. 209-322. Llerena Villalobos, Rito (Coord.) 1995 Lenguas Aborígenes de Colombia Estudios Fonológicos del Grupo Chocó, CCELA-Uniades, Universidad de Los Andes, Colombia. Loewen, Jacob 1969a ‘Choco I: Introduction and Bibliography’, IJAL 29: 239-263. 1969b ‘Choco II: Phonological Problems’, IJAL 29: 357-371. Maddieson, Ian 1980 PSID: The UCLA Phonological Segment Inventory Database, Los Angeles: UCLA. 1984 Patterns of Sounds, Cambridge: Cambridge University Press. Mortensen, Charles A. 1999 A Reference Grammar of the Northern Embera Languages, Studies in Languages of Colombia 7, SIL International and the University of Texas at Arlington.

126 HERRERA

Nordenskiöld, E. 1929 ‘Les rapports entre l’art, la religion et la magie Chez les Indiens Cuna et Chocó’, Journal de la societé des Americanistes de Paris, Vol. XXI: 141-158. Ohala, John J. 1983 ‘The Origin of Sound Patterns in Vocal Tract Constraints’, en: Peter F. MacNeilage (ed.), The Production of Speech, Berlin/Heidelberg/Nueva York: Springer Verlag. Pardo, Mauricio 1988 ‘Indígenas del Chocó’, en: Correa, F. y X. Pachon, Introducción a la Colombia amerindia, Instituto Colombiano de Antropología, , pp. 251- 262. Wright, Richard y Peter Ladefoged 1994 ‘A Phonetic Study of Tsou’, en: Fieldwork Studies of Targeted Languages II,UCLA Working Papers Phonetics 87: 67-92.

LA FONOLOGÍA SUPRASEGMENTAL Y OTRAS PARTICULARIDADES DEL BANIVA DE MAROA, IDIOMA TONAL ARAWAK DEL RÍO NEGRO, VENEZUELA

Esteban Emilio Mosonyi Universidad Central de Venezuela

Dada nuestra defensa irrestricta de la sociodiversidad —y de sus componentes, la diversidad cultural y lingüística— hemos sido extremadamente críticos frente a planteamientos como la inevitabilidad e irreversibilidad de la llamada “muerte lingüística” (languague death) a partir de Nancy Dorian en adelante (1981). He sido muy cauteloso a la hora de lanzar una afirmación tan contundente. Me consta que ni esta autora ni otros que siguen sus lineamientos se pronuncian abiertamente a favor de la desaparición de los idiomas vulnerables del mundo. Me parece mucho más llamativo su modo impasible de hacer el diagnóstico correspondiente, sin asumir posición alguna ni sentirse tentados a formular una crítica siquiera tímida frente a la expansión, casi lineal, de las lenguas dominantes: ello se aprecia con bastante transparencia en el caso de Dorian ante la imposición del inglés sobre el gaélico. Parecieran sentirse a gusto en su papel de albaceas del fenómeno en cuestión. Debemos aquí conformarnos con una cita breve pero característica:

Such a shift is an aspect of sociocultural change, intimately linked to phenomena like urbanization, industrialization, and secularization, though —interestingly— not predictable from any of them. Increasingly, studies of linguistic persistence or replacement have focused on contexts of modernization and nationalization (Dorian, 1981: 4).

Nadie ignora que los idiomas minoritarios y minorizados —es decir, no tan minoritarios pero sí tratados con discriminación y cierto desprecio— son vulnerables y de hecho dejan de practicarse, con frecuencia acelerada, en sus respectivas comunidades. Lo que no quiero asumir es ver este fenómeno como algo natural y tal vez benéfico, ya que para mí y por fortuna para muchos, el valor irremplazable de una lengua —tanto intrínseco como extrínseco, vale decir, expresivo y comunicativo— constituye un dato axiomático. Ahora bien, si no puedo pensar en la extinción de una cultura sino como un caso de etnocidio, tampoco admito que la supresión de una lengua sea algo distinto de un lingüicidio, y como tal enormemente perverso; sobre todo, dada la frecuencia sin precedentes con que se viene dando el fenómeno bajo el manto de una globalización aplastante. Uno no puede ser conservacionista en lo ambiental y homogeneizante en 128 MOSONYI lo humano, sin presentar grandes contradicciones y fisuras en su personalidad. Igualmente, aceptar de manera aparentemente pasiva la supuesta fatalidad que va corroyendo los idiomas oprimidos es muestra —si no de irresponsabilidad y cinismo— de un cientificismo que sacrifica toda posición ética y política en aras de conocer y describir un determinado “objeto” de estudio. Involucra también cierto menosprecio hacia la creatividad y el conocimiento en cualquiera de sus formas, ya que todas pasan por el lenguaje, es decir, las lenguas del mundo. Además, frente a los representantes connotados de esta línea de investigación, aquellos que apuestan a la inevitabilidad del “language death”, sostengo que el proceso de erosión lingüística no es en modo alguno indetenible, salvo quizá en su fase final, y aun allí caben algunas alternativas de acción. Piénsese por ejemplo en los esfuerzos, si bien a menudo asistemáticos, que despliegan los movimientos y comunidades indígenas por conservar de manera dinámica su lengua y cultura, sin desligarse del contexto mundial contemporáneo. Sin embargo, mi condición de activista de los derechos lingüísticos y humanos en general no me ciega ante el hecho de que en la práctica estas lenguas, honrando excepciones, no solamente pierden fuerza sino que en su interior sufren procesos desestructurantes y desintegradores —bajo la presión de bilingüismos y multilingüismos orientados hacia la transición a los idiomas más poderosos— que las van empobreciendo y deteriorando gradualmente o incluso con cierta violencia. En tal sentido, sí es preciso reconocerles méritos a los investigadores mencionados, ya que ellos han diseñado esquemas descriptivos que diagnostican a cabalidad los fenómenos concretos que ocurren cuando una lengua comienza a dejar de utilizarse, ante todo en las generaciones jóvenes. Una experiencia de esta naturaleza me ha ocurrido con el idioma arawak baniva del Río Negro, también conocido en el Río Xié con el nombre de warekena o werekena, muy diferente por cierto del warekena hablado en Guzmán Blanco, estado Amazonas, Venezuela, estudiado y vivencialmente asumido por el lingüista venezolano Omar González Ñáñez (1997). En otros términos, el baniva del Río Negro —estudiado por nosotros— forma un sistema lingüístico o idioma único junto con el warekena o werekena hablado en el Xié, Brasil; vale decir, son dos variantes de una misma lengua. En cambio, el llamado warekena de Guzmán Blanco es bien diferente aunque pertenece a la misma familia (arawak); la distancia puede ser como la que media entre el lituano y el ruso. Tengo que confesar que no poseo datos de primera mano sobre la variedad baniva del Río Xié, si bien percibo —por los importantes trabajos de — que la diferencia es mínima, casi desestimable (cf. Aikhenvald, 1998). He permanecido en contacto con este baniva —no con el baniva-kurripako de Colombia (cf. Gómez-Imbert, 1996: 458 y )— desde los años sesenta (cf. Mosonyi, 1968a y b), aun cuando durante la mayor parte de este período no pude, lamentablemente, dedicarme a su estudio detallado y menos aún a su revitalización. FONOLOGÍA SUPRASEGMENTAL BANIVA 129

En efecto, el retroceso del baniva de Maroa empezó en la primera mitad del siglo XX, por motivo de la aculturación rápida e impetuosa de sus hablantes (cf. Mosonyi y Mosonyi, 2000: 184), mas también a consecuencia de una fuerte presión procedente de la escuela castellanizante, las instituciones de la República y la población criolla en general. El epicentro de esta etnia es el pueblo de Maroa, que ni siquiera en sus mejores días pudo llegar a los mil habitantes, pero que inclusive en el momento de la visita del sabio Humboldt (1950 [1799], IV: 202) presentaba ya ciertos visos de urbanización y un alto grado de aculturación: en el siglo XIX llegó inclusive a poseer una imprenta. Actualmente podría hablarse más bien de una suerte de vuelta a la ruralidad, pues en el presente su población está emigrando a Puerto Ayacucho y otras ciudades, aparentemente por falta de viabilidad económica, aunque hay muchas razones que no cabe aquí examinar. Cuando conocí al señor Hernán Camico, ya fallecido y probablemente el mejor hablante de baniva en los últimos tiempos, esta lengua se encontraba sumida en un deterioro muy profundo. Pero todavía muchos de los adultos maroeños lo utilizaban corrientemente y lo conocía relativamente bien un sector de la juventud. Hoy día la mayoría de los buenos hablantes —sería difícil cuantificarlos para la presente fecha, aunque el Censo Indígena (Oficina Central de Estadística e Informática, 1992) registra 418 conocedores del baniva de un total de 976, es decir, 43% entre hablantes y semihablantes— son ancianos y, sobre todo, ancianas. Aun así, justo es decirlo, hay en las últimas décadas un interés creciente por revitalizarlo y es probable que esto se logre algo tardíamente, dada la factibilidad institucional y financiera de inaugurar a fines del presente año al menos un nicho lingüístico baniva en Puerto Ayacucho y paulatinamente otros en Maroa y distintas poblaciones del estado Amazonas. La figura del nicho lingüístico (language nest) es una suerte de micro-institución revitalizadora, ya muy popular en Nueva Zelanda y el país de Gales, entre otros. Se trata, en principio, de hogares ampliados de índole preescolar donde los mejores hablantes —generalmente señoras ancianas— transmiten a los infantes de la comunidad, a menudo familiares suyos, el uso activo y coloquial de la lengua en peligro, con inclusión de contenidos discursivos y culturales concomitantes. Su informalidad y estructura flexible —moldeada sobre la familia tradiconal— constituyen sendas garantías de su éxito y difusión creciente. En el caso concreto de Nueva Zelanda, los resultados alcanzados pueden catalogarse de brillantes, por cuanto han logrado situar al idioma maorí —hace poco casi moribundo— en un estatus de lengua co-oficial con el inglés, eficazmente utilizado en muchas situaciones sociolingüísticas. Los mejores nichos lingüísticos son aquellos motorizados por las propias comunidades, y en ningún momento se oponen a otras estrategias didácticas como la enseñaza escolar y académica de las mismas lenguas. Ahora bien, sin renunciar al supuesto de que se forme y consolide una nueva generación de baniva-hablantes, hay una característica del idioma que difícilmente podrá retomarse: su sistema suprasegmental caracterizado por acentos tonales y la presencia de longitud vocálica. Después de la breve exploración que hiciéramos en

130 MOSONYI fecha reciente entre las familias banivas de Puerto Ayacucho, se corre el albur de que el señor Camico haya sido uno de los últimos “hablantes tonales” de este idioma, puesto que las mujeres de cierta edad que entrevistamos en esa oportunidad sólo presentan vestigios de tonos en el mejor de los casos. Para mayor precisión, ello significa una capacidad limitada de articular las alturas musicales fonológicamente pertinentes, a veces únicamente en un número reducido de palabras; mientras que en el resto del vocabulario tales hablantes solo pronuncian claramente los acentos de intensidad en una frecuencia tonal perfectamente predecible, al igual que en español, inglés o ruso. Para ellos ya no tiene validez el sistema tonal que propondremos en los párrafos siguientes. Debido a avatares inexplicables, no conservamos grabaciones del habla de Hernán Camico, por lo cual mi único referente se materializa en las notas de campo, mis recuerdos y publicaciones (cf. Mosonyi y Mosonyi, 2000: 184-223). En una conversación reciente que sostuve con el lingüista venezolano José Álvarez, éste me dio testimonio de su experiencia con otra hablante tonal, quien incluso presenta patrones melódicos muy similares si no coincidentes con los de Camico, ahora perpetuados mediante grabaciones de calidad profesional. Sería también prematuro descartar la accesibilidad de otros hablantes con las virtudes y el nivel de competencia de Hernán Camico, quienes servirían de colaboradores para un estudio pormenorizado de fonética instrumental sobre el sistema sonoro del idioma baniva. Dado que ya hemos descrito la mencionada tonalidad en otros contextos (cf. Mosonyi y Mosonyi, 2000: 188-189), esta vez tenemos que limitarnos a un breve resumen que enfatice solamente sus particularidades más distintivas. Existen tres acentos tonales fundamentales (conocemos por lo menos dos más, de poca ocurrencia), en cierta forma análogos a los que presentan el sueco y el noruego (del grupo germánico-escandinavo), así como distintas variedades del serbio-croata y el esloveno (grupo eslavo meridional). Sin entrar en una larga teorización, ciertamente muy delicada, trataremos de consignar lo más relevante de este tipo de sistemas suprasegmentales. Ello es particularmente importante, pues existen lingüistas que sienten cierta incomodidad —a mi modo de ver poco justificada— en relación con la categoría de acentos tonales, por cuanto creen se trataría de un cajón de sastre o de la no asunción plena de los tonos fonémicos como tales. Por el contrario, considero que esta categoría es pertinente y necesaria. Si preferimos un lenguaje puramente fonológico antes que fonético, pudiéramos hasta decir que en el fondo estamos frente a varios tipos cualitativamente distintos de acentos de intensidad en los idiomas citados y otros suprasegmentalmente análogos. En español, portugués, inglés, alemán, entre otros muchos, solo tenemos acentos de intensidad primarios y secundarios, ninguno de los cuales presenta especificaciones tonales —es decir de altura musical— que sean relevantes para su caracterización fonémica exhaustiva. En cambio, otros idiomas —incluido el baniva— comportan dos o más acentos fonémicos de intensidad, en la medida en que sus particularidades fonéticas son sistemáticamente diferenciables, pudiendo dar lugar a numerosos pares FONOLOGÍA SUPRASEGMENTAL BANIVA 131 mínimos y análogos. El sueco, por ejemplo, comporta —según la mayoría de sus gramáticos— dos acentos de intensidad diferentes. El primero se articula mediante el refuerzo y elevación melódica de la sílaba más dinámica, mientras que la sílaba siguiente, no acentuada, se pronuncia mucho más débil y melódicamente más baja: el resto de las sílabas no cuenta para el análisis. Por otra parte, el segundo acento conlleva no tan solo un refuerzo de la intensidad sino también la emisión de un descenso melódico en unos cinco o más semitonos, en tanto que la sílaba siguiente, no acentuada, se pronuncia a una altura musical algo superior a la parte final de la inflexión descrita. Esto genera cierto número de pares mínimos, una cantidad mucho mayor de pares análogos y una infinitud de pares potenciales no utilizados por la economía del idioma. Este ejemplo nos aclara que el llamado acento tonal constituye, ante todo, un acento de intensidad, en el cual los rasgos musicales son apenas marcas diacríticas para distinguirlo de otro u otros acentos de intensidad. La melodía, aun siendo relevante, está totalmente subordinada a la fuerza espiratoria. Vamos a terminar de comprobar este aserto acudiendo al idioma danés, pariente próximo del sueco. Esta lengua comporta también dos tipos de acentos de intensidad, pero la diferencia no radica en sus respectivas líneas melódicas, sino en el hecho de que el primer acento termina en saltillo u oclusión glotal, mientras que el segundo —fonémicamente diferente— no lleva esa coda glotalizada. Considero que valió la pena hacer esta digresión teórica para comprender cabalmente a qué nos queremos referir al tratar de describir los acentos tonales del idioma baniva. Tal como empezamos a desglosar más arriba, hay en esta lengua al menos tres acentos de intensidad que difieren fonéticamente entre sí por sus peculiares características musicales, de una forma muy análoga al ejemplo del sueco e incluso del danés. Es conveniente agregar de inmediato que muchos lexemas baniva contienen dos y a veces tres acentos tonales —en secuencia continua o discontinua—, si bien para efectos del presente artículo nos limitaremos mayormente a palabras que exhiban uno solo de estos acentos. El agudo (´) indica que la sílaba acentuada se pronuncia en un nivel alto y con bastante fuerza espiratoria, al tiempo que las inacentuadas que siguen, siempre débiles, mantienen exactamente el mismo nivel musical, e.g.: né:yawa (mujer), yáatsina (arena), yáatsipe (tierra). Las sílabas inacentuadas anteriores, si las hubiere, serían extramétricas, vale decir, inexistentes para este patrón, y para cualquiera de los patrones que se explicarán a continuación. Con el acento grave (`) la sílaba acentuada se pronuncia fuerte y en un nivel bajo, en tanto que las inacentuadas subsiguientes mantienen la misma altura o incluso descienden, e.g: wabù:pi (caño), numà (yo digo o hago), kerumài (es o está dulce). Las sílabas inacentuadas anteriores al acento grave son débiles y de altura musical media; aunque ya se señaló su naturaleza extramétrica. El tercer acento, llamado circunflejo (^), indica que la sílaba acentuada, siempre intensa, es alta y las inacentuadas, débiles, se desplazan al nivel medio o incluso bajo,

132 MOSONYI e.g:. ê:nami (hombre), abîda (báquiro), marîri (chamán). Para mayor transparencia fonética reiteramos que tanto las sílabas anteriores como las posteriores a la acentuada se pronuncian al menos cinco semitonos más bajos. Queremos precisar, de paso, que las consonantes “b”, “d” y “g” se pronuncian en baniva tensas, totalmente oclusivas y hasta alargadas. Es importantísimo señalar que el acento tonal desempeña papeles morfofonológicos muy precisos y perfectamente estructurados, como el marcaje de la tercera persona singular masculina, tanto en la morfología nominal como verbal, e.g: nunûma (mi boca), núma (su boca); nuyà:tsipere (mi tierra), yâ:tsipere (su tierra); numì:watà (yo juego), mí:watà (él juega); nutámà (yo bailo), támá (él baila); nú:runià (yo espero), ú:runiá (él espera). Sin entrar en detalles, en estos ejemplos la tercera persona singular masculina presenta siempre un patrón de mayor altura musical que la primera del singular y todas las demás personas en general. Recuérdese que el baniva permite la ocurrencia de dos o más acentos tonales en una misma palabra. En algunos casos la altura tonal constituye la única marca diferencial de persona: ú:nitá (él nada), ú:nità (nosotros nadamos); al igual que en el ejemplo señalado más arriba, ú:runiá (él espera), ú:runià (nosotros esperamos): se trata de verdaderos pares mínimos. Hay razones para suponer que en esta función la altura musical reemplaza una antigua vocal “i”, es decir, *iu:nita (él nada), *iu:runia (él espera), como ocurre en otras lenguas arawak, entre ellas el warekena de Guzmán Blanco y el piapoko. Tal hipótesis es razonable, dada la alta frecuencia acústica de la vocal “i”, la cual —mediante un proceso de cambio fonético— pudo haber desaparecido para ser compensada por una mayor altura musical en la configuración fónica adyacente. Inclusive, en algunas palabras del propio baniva, como en nú:tawapà (yo camino), í:tawapà (él camina), persiste la vocal “i” también el el plano sincrónico, con la interesante concomitancia de que al menos en estos casos la tercera persona del masculino singular no es musicalmente más alta que las demás personas. Nos parece que este hecho tiende a corroborar nuestro planteamiento en forma bastante significativa. Existen varios otros usos claramente establecidos del acento tonal como indicador —o más bien coindicador— de algunas categorías, por ejemplo el relativo: nutérukà (yo corto), terûkali (el que corta); nupú:liùtà (yo pienso), puliû:tali (el que piensa). Aclaramos de paso que en baniva la “l” es una ele-ere lateral vibrante, fonética y fonémicamente diferente de la “r” vibrante simple. Es también muy llamativo el contraste entre lo que nosotros denominamos verbos fuertes, es decir, los que conservan sus acentos tonales, y los verbos débiles, que los modifican notablemente (Mosonyi y Mosonyi, 2000: 199). Por ejemplo, la forma perfectiva del verbo nuwè (yo dejo) es nuwèmià (ya dejé), que conserva inalterado el acento grave en la sílaba -wè; en cambio el perfectivo de nubè (yo puedo) se forma cambiando el acento grave de -bè por uno agudo, y la palabra se cierra con el mismo morfo -mià del ejemplo anterior: nubémià (ya pude). FONOLOGÍA SUPRASEGMENTAL BANIVA 133

Aunque los ejemplos señalados corroboran el rol que juega el acento tonal en la estrategia morfofonológica del baniva, existen patrones mucho más complejos, igualmente consistentes, de esta índole; sospechamos que ya ni siquiera el habla de Hernán Camico reflejaba todas las oposiciones melódicas que estarían presentes en el baniva cuando éste se hallaba en pleno florecimiento. Otro rasgo fonológico, igualmente pertinente, que parece haberse perdido para siempre, es la longitud vocálica. Ya el señor Camico vacilaba algunas veces y su distinción fonética no revelaba propiamente vocales cortas acentuadas y largas acentuadas sino que establecía la oposición entre acentuadas largas y semi-largas, mientras que las no acentuadas siempre las articulaba cortas. Hacía muy pocos pares mínimos, siendo uno de los más conspicuos el que establecía bajo el contraste, claramente presente y consciente para el hablante Camico, entre ê:li (tabaco) y êli (tábano); el número de pares cuasi-mínimos es por supuesto mucho mayor. Ejemplos: yá:pa (cerro) y yâpa (lapa); á:srì (yuca) y ásri (fuego); amùsr:i (cadáver) y amûsri (sol). Observamos de paso que la grafía “sr” se pronuncia como una “s” retrofleja modificada por una brevísima vibración linguo-alveolar sorda. Las hablantes, de cierta edad, con quienes tuve contacto recientemente ya no hacen esta distinción relativa a la longitud vocálica, aun cuando hablen su lengua con mucha fluidez. Sin embargo, al llamárseles la atención expresamente sobre dicha oposición ellas la recuerdan en el habla de generaciones anteriores, al igual que los contrastes tonales. En consecuencia, lo que subsiste invariablemente del sistema suprasegmental baniva es la clara presencia de uno o más acentos de intensidad por palabra, vale decir acento sin características melódicas relevantes, en forma similar al español, portugués o inglés. El idioma yavitero, descrito por Jorge Mosonyi (1987), es el más semejante al baniva; curiosamente tal vez, éste también presenta un acento fonémico de intensidad, incluso uno solo por palabra, en labios de la señora Águeda Largo, colaboradora de dicho estudio y en ese momento la única hablante conocida de la lengua. Con todo, es probable que anteriormente el yavitero poseyera acentos tonales distintivos como el baniva. Para completar el panorama referente al colapsado sistema suprasegmental de este idioma, queremos añadir que en dos ocasiones logramos reunir grupos de baniva-hablantes, en Maroa (Venezuela) y Puerto Inírida (Colombia) respectivamente, cuyos miembros reconocían, asumían y corroboraban —aunque fuese por anuencia pasiva— los acentos, tonos y longitudes vocálicas realizados en la muy clara y firme articulación de Hernán Camico, quien para ello utilizó largas listas de palabras pertenecientes a diferentes categorías gramaticales. Todo confluye para atestiguar que el sistema suprasegmental de esta lengua sobrevive mucho más en la competencia pasiva que en la activa de sus hablantes y aun de sus semi-hablantes. Hemos revisado con admiración e interés el largo trabajo de Alexandra Aikhenvald (1998) sobre el warekena del Xié, prácticamente idéntico al baniva de Maroa, porque sus hablantes descienden de migrantes venezolanos llegados al Brasil a principios del siglo XX; pero nos resulta imposible establecer comparaciones entre

134 MOSONYI las dos variedades en lo atañente a las características suprasegmentales. Aikhenvald no marca en sus textos ninguno de los tres componentes del sistema: acento, tono y longitud vocálica. Presumimos la inexistencia o quizá la muy escasa perceptibilidad de los dos últimos factores, ya que para la presente fecha ni siquiera en Maroa están vigentes el tono y la longitud vocálica. Mas tenemos la impresión de que el acento de intensidad, sin las marcas tonales, tiene mucha más importancia de la que le asigna la mencionada autora. Ella, en efecto, trata el punto, pero en términos muy generales, más bien accesorios; y como señalamos arriba, en su larga colección de textos no marca ni un sólo acento. Ignoramos si posteriormente la autora ha publicado algún material de la misma lengua donde se evidencie un tratamiento más profundo de la fonología suprasegmental. Todo esto nos lleva a una reflexión que consideramos pertinente para la lingüística amerindia en general y las lenguas arawak en particular. Vemos con verdadera preocupación que durante estos últimos años, como ocurría a principios del siglo XX, la mayoría de los lingüistas —inclusive los más capaces y productivos— prestan cada vez menos atención a los sistemas fonológicos de las lenguas amerindias; a tal punto que realizan sus análisis gramaticales y semánticos, además de reunir textos a veces muy extensos, sin tomar en consideración, por ejemplo, los rasgos suprasegmentales, aun cuando tengan pertinencia fonológica. A veces la transcripción se ve sustituida por una mera aproximación, que utiliza signos alfabéticos de un idioma particular como el español o el inglés, tal como ocurría con los viajeros y cronistas en tiempos de Humboldt o antes. Pareciera haber cierto apresuramiento en recoger la mayor cantidad posible de material léxico y morfosintáctico, sobre todo al estar en presencia de lenguas seriamente amenazadas o al borde de la extinción. Tal vez se piense que sacrificando en cierto modo la fonología puedan documentarse otros aspectos de las lenguas señaladas, no sé si más relevantes para determinados tipos de estudios. Pensamos que tal forma de proceder es errónea, a más de manifiestamente superespecializada, y debe corregirse desde ahora y para el futuro, por múltiples razones. Hoy más que nunca, habida cuenta del valor intrínseco y patrimonial de las lenguas e incluso del deseo de conservarlas y recuperarlas, compartido por propios y extraños, estos sistemas lingüísticos minorizados deberían transcribirse y representarse con la mayor fidelidad posible. No nos resulta satisfactorio un análisis morfosintáctico, por más que parezca exacto, detallado y profundo, si éste se lleva a cabo sobre una base fonológica un tanto endeble y hasta insuficiente. Los idiomas adquieren su densidad y prestancia por la interacción de sus distintos niveles sistémicos, sin lo cual su configuración se desdibuja y pierde nitidez. Además, la fonología y la morfofonología suelen influir en la morfosintaxis de manera decisiva, como lo tratamos de comprobar más arriba con la descripción de algunos roles atribuibles a los acentos tonales del baniva. No existe ninguna razón para preferir los indicadores morfosintácticos a los morfofonológicos. Esto es válido tanto para la lingüística sincrónica como para la diacrónica, ya que la recuperación de las formas FONOLOGÍA SUPRASEGMENTAL BANIVA 135 históricamente anteriores pasa por la reconstrucción de los cambios fonológicos segmentales y suprasegmentales. No querría ser demasiado severo en mis críticas y apreciaciones, por cuanto me consta que han aparecido y siguen saliendo muchas publicaciones caracterizadas por excelentes transcripciones y un esmerado tratamiento de los niveles fonológico y morfofonológico. Por otra parte, es innegable y hasta obvio que un alto porcentaje de los trabajos descriptivos de lenguas amerindias fallan no solamente en lo fonológico sino que son deficitarios —igualmente o tal vez en mayor medida— en lo morfosintáctico, lexicosemántico y pragmático-discursivo. No obstante, en este artículo me siento en la obligación de conferirle un énfasis especial a la transcripción y al análisis fonológico en general. Considero que hay buenas razones para permitirme un sesgo de tal índole. Observemos tan solo lo siguiente. Cuando manipulamos un texto bien transcrito, como consecuencia de un estudio fonológico adecuado, invariablemente nos servirá de mucho, ya sea como insumo para otro tipo de investigaciones o para fines educativos o de lectura destinada a los hablantes nativos. En tal caso, ni siquiera parece demasiado grave si ese texto está deficientemente traducido o si el análisis gramatical es erróneo o simplemente inexistente. En cierta manera, todo esto se puede suplir en la medida en que nuevos investigadores contacten hablantes nativos dispuestos a colaborar y a perfeccionar esfuerzos anteriores. Sin embargo, ante la evidencia de un texto mal transcrito, la utilización ulterior de dicho material resultará siempre riesgoso, extremadamente complicado, a veces imposible como a menudo nos ha sucedido. Posiblemente estemos en presencia de esfuerzos irrecuperablemente perdidos. Por lo cual nuestro argumento no estriba en la mayor o menor importancia intrínseca de uno u otro subsistema lingüístico —discusión que nos parece además subjetiva e irrelevante— sino en cierta precedencia lógica que a nuestro modo de ver le corresponde al conjunto sonoro del idioma, al análisis fonológico en sus vertientes segmentales y suprasegmentales, antes de emprender el estudio pormenorizado de los demás niveles igualmente importantes y sustantivos. Los movimientos indígenas claman por estudios lingüísticos que les permitan la codificación y transmisión exacta de sus idiomas, sobre todo si se trata de revertir un proceso de erosión aún detenible. Y la docencia de cualquier lengua comienza por su recta escritura y pronunciación. Si la representación del habla ocurre mediante fórmulas cuasi-telegráficas, vale decir, grafías que sólo parcialmente reflejan la articulación sistémica, simplemente se paraliza la gestión educativa, con las gravísimas consecuencias que es innecesario detallar. Respetamos todas las prioridades investigativas, pero a estas alturas de la lucha por la sociodiversidad —y por la revitalización de las lenguas en particular— son las aspiraciones y metas de los pueblos en recuperación las que revisten mayor relevancia histórica.

136 MOSONYI

Referencias Bibliográficas Aikhenvald, A. 1998 ‘Warekena’, en: D. Derbyshire y G. Pullum (eds.), Handbook of Amazonian Languages, Volumen IV, Berlin: Mouton de Gruyter, pp. 225-433. Dorian, N. 1981 Language death, Philadelphia: University of Pennsylvania Press. Gómez-Imbert, E. 1996 ‘When animals become ‘rounded’ and ‘feminine’: conceptual categories and linguistic classification in a multilingual setting’, en: J. Gumperz y S. Levinson (eds.), Rethinking linguistic relativity, Cambridge: Cambridge University Press, pp. 438-469. González Ñáñez, O. 1997 Gramática de la lengua warekena (maipure-arawak): una aproximación tipológica-relacional, Tesis Doctoral, Caracas: Facultad de Ciencias Económicas y Sociales, Universidad Central de Venezuela. Humboldt, A. 1950[1799] Viaje a las regiones equinocciales del nuevo continente, Caracas: Dirección de Cultura y Bellas Artes, Ediciones del Ministerio de Educación. Mosonyi, E. 1968a ‘Elementos de lingüística arahuaca’, en: Economía y ciencias sociales, Año X, Nº 3, Caracas: Universidad Central de Venezuela, pp. 77-85. Mosonyi, E. 1968b ‘Introducción al análisis intraestructural del idioma baniva’, en: Economía y ciencias sociales, Año X, Nº 3, Caracas: Universidad Central de Venezuela, pp. 65-75. Mosonyi, J. 1987 El idioma yavitero: ensayo de gramática y diccionario, Tesis Doctoral, Caracas: Facultad de Ciencias Económicas y Sociales, Universidad Central de Venezuela. Mosonyi, E. y J. Mosonyi 2000 Manual de lenguas indígenas de Venezuela, II Tomos, Caracas: Fundación Bigott. Oficina Central de Estadística e Informática 1992 Censo indígena de Venezuela 1992, Tomo I, Caracas: Presidencia de la República. FONOLOGIA DEL PROTO-MATAGUAYO: LAS FRICATIVAS DORSALES

J. Pedro Viegas Barros CONICET-Universidad de Buenos Aires

0. Introducción La familia lingüística Mataguayo (también llamada Mataco, Mataco-Mataguayo, Mataco-Maká) incluye cuatro lenguas de la región chaqueña: Wichi (o Mataco), con unos 35.000 hablantes, en Argentina (provincias de Chaco, Formosa, Salta) y Bolivia (departamento Tarija), Chorote, con unos 2.000 hablantes en Argentina (provincia de Salta) y Paraguay (departamento Boquerón), Nivaklé (o Chulupí, o Ajlujlay), con entre 6.000 y 14.000 hablantes según distintas fuentes, también hablado en Paraguay (departamentos Boquerón y Presidente Hayes) y Argentina (provincia de Salta), y Maká, con unos 1.000 hablantes, de Paraguay (departamento Presidente Hayes). En este trabajo se analiza un aspecto de la fonología de la lengua originaria reconstruida para la familia lingüística Mataguaya las fricativas dorsales. Por “dorsales” me refiero aquí a aquellas consonantes en las que articulador es el dorso de la lengua, que incluyen a los puntos de articulación velar (incluyendo velar palatalizada) y posteriores (postvelar, uvular, laríngeo, etc.). Esta parte del sistema fonológico de las lenguas Mataguayas ha resultado una de las más problemáticas desde el punto de vista histórico-comparativo. En la reconstrucción de Najlis (1984) no se postulaba ninguna fricativa dorsal para el Proto-Mataguayo. En mi primera propuesta (Viegas Barros 1993) incluí una fricativa dorsal *h. En el presente trabajo discuto las reconstrucciones anteriores y presento evidencia de que el sistema consonántico del Proto-Mataguayo incluía no menos de dos fonemas fricativos dorsales no labializados.

1. Primera reconstruccción (Najlis 1984) En su reconstrucción de la fonología del Proto-Mataguayo, y trabajando sólo con tres de las cuatro lenguas Mataguayas, ya que no tuvo a su disposición materiales del Maká, Najlis (1984) propuso la siguiente reconstrucción de la consonantes del Proto- Mataguayo (remplazo algunas de las letras usadas por la autora), que no incluía ninguna fricativa dorsal:

*p *t *ts * kj *k *q *p’ *t’ *ts’ *kj’ *k’ *ph *th *tsh *w *l *s *j *hw *hl *hs *m *n *hm *hn

138 VIEGAS BARROS

Los fonemas h del Wichi, x del Chorote y x del Nivaklé eran interpretadas por esta autora como desarrollos secundarios, a partir del desplazamiento y fonologización de la aspiración que originariamente habría acompañado a ciertas oclusivas, africadas, “continuas” (fricativas, laterales, semivocales), y nasales del Proto-Mataguayo. Las “continuas” *hw y *hl habrían dado como resultado fricativas en casos esporádicos, por pérdida de su parte no aspirada1, como en:

(1) *-hwum (Najlis 1984: 10), Chorote -xum, Nivaklé -xum ‘estar borracho’, (2) *hl- (cf. Najlis 1984: 15 *hl-jae > Chorote xi-jaxe ‘tú bebes’), Chorote xi- ‘pre-fijo de segunda persona sujeto’.

Las demás aspiradas habrían producido fricativas dorsales por separación y fonologización de su porción aspirada, en una o en varias lenguas, de manera mucho más regular:

(3) *iphatha (Najlis 1984: 9), Wichi ihpat ~ sipot, Chorote ipjatax ‘maíz’, (4) *-itho (cf. Najlis 1984: 10, 40), Wichi -wit’oh, Nivaklé -t’ox ‘tía’, (5) *-lutshe ‘arco’, véase infra serie (34), (6) *-Ahse (Najlis 1984: 11), Wichi -øse, Chorote -axse, Nivaklé -øse ‘hija’, (7) *-hmi (Najlis 1984: 10), Wichi humin, Chorote emi ‘amar’, (8) *hnahn (Najlis 1984: 10), Wichi honah, Chorote axnax ‘noche’.

Además, Najlis supuso como reflejo de Proto-Mataguayo *q en posición intervocálica la secuencia bifonémica kx en Nivaklé.

(9) *snaqaj > Nivaklé snakxaj ‘niebla’, cf. infra, serie (82), (10) *slAqaj > Nivaklé sløkxaj ‘gato montés’, cf. infra serie (83).

En posición final, de acuerdo a esta reconstrucción, el mismo protofonema daba irregularmente en Wichi y en Nivaklé unas veces una oclusiva dorsal, otras veces una fricativa dorsal como resultado:

(11) *k’ataq > Wichi k’atak, Nivaklé …afkatax ‘mosca’, cf. infra serie (86), (12) *hnakotaq > Wichi nakwotak, Nivaklé šnakotax ‘abeja’, cf. infra serie (87), (13) *hwatsuq ~ *patsuq > Wichi xwatsuh, Nivaklé fatsuk ‘ciempiés’, cf. infra serie (88).

Finalmente, siempre según la citada autora, una fricativa posterior podía haberse originado de una oclusiva glotal automática antes de vocal inicial o tras vocal final Proto-Mataguaya:

1 El origen de ciertas fricativas x del Chorote por deslabialización a partir de *xw, y por deslateralización a partir de *… , ya habia sido propuesto por Gerzenstein (1983: 27-9, 1987).

PROTO-MATAGUAYO 139

(14) *jewu [*/jewu], Wichi hajewu ‘chamán’, véase infra, serie (47). j j (15) *k u [*k u/] (Najlis 1984: 12), Wichi c&uh ‘tener calor’.

2. Segunda reconstrucción (Viegas Barros 1993) Una primera dificultad para aceptar las explicaciones de Najlis provino de la descripción del Maká, lengua que, como mostró Gerzenstein (1989), posee en su inventario fonológico no uno sino tres fonemas fricativos dorsales distintos: velar x, uvular X, y laríngeo h. Suponiendo que uno de ellos realmente proviniera del desplazamiento de la aspiración de consonantes aspiradas originales, ¿cuál era el origen de los otros dos? Por otra parte, tanto en Maká como en Nivaklé las secuencias de oclusiva + fricativa dorsal en posición intervocálica son siempre heterosilábicas, es decir, claramente son grupos de dos fonemas. Y con respecto a las “continuas” y nasales no había mayor razón para la reconstrucción de aspiradas – salvo la simetría del sistema propuesto por Najlis–. La postulación para el Proto- Mataguayo de oclusivas y africadas aspiradas no parecía tener, por lo tanto, justificación en las lenguas Mataguayas históricamente documentadas. Todo esto me llevó (Viegas Barros 1993) a incluir un fonema fricativo dorsal no labializado *h en el sistema fonológico que entonces postulé para el Proto-Mataguayo, consecuentemente con la eliminación de las consonantes aspiradas (también aquí cambio algunas de las letras utilizadas originalmente):

*p *t *ts *kj *k *q *p’ *t’ *ts’ *kj’ *k’ *w *l *j w *x *… *s *h *m *n *M *N

En ese entonces me parecía decisivo –para la reconstrucción de un solo proto-fonema fricativo dorsal no labializado– el hecho de que tres de las cuatro lenguas Mataguayas tenían cada una un solo fonema de este tipo, según las descripciones disponibles para mí en ese entonces: Wichi h (Viñas Urquiza 1974)2, Chorote x (Gerzenstein 1978,

2 En la mayor parte de los dialectos Wichi, el fonema /h/ se realiza como fricativa laríngea [h] (generalmente nasalizada) en posición inicial de sílaba, y como fricativa velar [x] en final de sílaba. Pero en el dialecto Nocten, de Bolivia (Alvarsson 1984, Claesson 1994), existen algunas [h] en el postmargen silábico y algunas [x] en posición intervocálica (aunque solamente en limite de morfemas), de manera que /h/ y /x/ en estas posiciones se oponen y son -por lo tanto- fonemas distintos. Cf. en esta variedad:

/latax ‘caballo’, frente a /ije/lah ‘tapir’, /aqøxjax ‘tu alegría’, /ekhahjah ‘tu fe’, j j /ixk ah ‘¡padre!’, /ihk øh ‘permanece’, pitsaxeh ‘suena’, nisahes ‘zapatos’, taxah ‘este (tocado)’, tøhax ‘cuando (quien, que) reciente’.

Claesson (1994: 22) sugiere que uno de los orígenes (quizás el principal) de h en final de sílaba en esta

140 VIEGAS BARROS

1983), Nivaklé x (Stell 1972, [1989]). Aunque no lo explicité en ese momento, supuse que la situación “anormal” en cuanto al número de fonemas fricativos dorsales del Maká no debía ser originaria, y se explicaría, en parte, como resultado del presunto contacto del Maká con otras lenguas, a las que yo no podía identificar en ese momento, en parte –posiblemente– como resultado de la fonologización de presuntos alófonos originales de la fricativa dorsal originaria al fusionarse en un solo fonema vocales distintas en determinados contextos, como p. ej.:

(16) *hele [*xele] > Ma xili ‘suciedad’, cf. infra, serie (18), (17) *him [*him] > Ma him ‘coatí’, cf. infra, serie (50)3.

3. Una nueva reconstrucción La postulación de una fricativa dorsal para el Proto-Mataguayo parecía solucionar algunos problemas. Pero, al mismo tiempo, surgían otros. El principal, para el que yo no ofrecía respuesta en mi trabajo de 1993, consistía en la existencia de tres grupos importantes de “irregularidades” en las correspondencias de las fricativas dorsales:

(a) en posición inicial, *h daba como resultado en Wichi en ciertas ocasiones h, en otras O. (b) en posición final tras a, el resultado de *h en Chorote era unas veces x, otras O. (c) en contacto con las vocales *e, *i el Nivaklé tenía como reflejo de *h a veces x, a veces š.

Como ejemplos de (a), cf.:

*hino/ > Wichi hino/ ‘hombre’, cf. infra serie (51), *hupel > Wichi hupel ‘sombra’, cf. infra serie (53), frente a

*hinawøp > Wichi (i)nawop ‘primavera’, cf. infra serie (22), *huwela > Wichi iwela ‘luna’, cf. infra serie (27).

variedad Wichi, es la síncopa vocálica, como se ve en:

/no-wahn-ex ‘alguien se lo permite (hacer)’ < /no-wahin ‘alguien lo permite’. /no-jahn-ek ‘espejo’ < /no-jahin ‘alguien lo mira’. j j /o…a-/ik uju-ht-hen ‘ayunamos’ < /o…a-/ik uju-hat ‘yo ayuno’.

Es decir que en Proto-Wichi habría existido un solo fonema */h/ que en Nocten se habría escindido en dos /h/ y /x/. 3 De manera independiente, Campbell (1997: 405, nota 22) ha manifestado sus dudas con respecto a que *hw, *hl, *hs, *hm y *hn de Najlis fuesen es realidad monofonemáticos, es decir, que también según este autor *h debía ser un fonema ya en Proto-Mataguayo.

PROTO-MATAGUAYO 141

Como ejemplos de (b), cf.:

*hweh > Chorote fwax ‘hachar’, cf. infra serie (36), *øtah > Chorote atax ‘ser gordo’, cf. infra serie (31), frente a

*tøtsinah > Chorote taxsina ‘sapo’, cf. infra serie (61), */øh > Chorote -aa ‘voz’, cf. infra serie (68).

Como ejemplos de (c), cf.:

*him > Nivaklé xim ‘coatí’, cf. infra serie (50), *hokhajeh > Nivaklé xokxajex (cf. Najlis 1984: 44; Viegas Barros 1993, o w 202, item n . 107: Wichi x okjah, Chorote kaje, Maká xokhejaX) ‘pato silvestre’, frente a

*hitV- > Nivaklé šta- ‘1a. p. pl. inclusivo sujeto’, cf. infra serie (24), *lutseh > Nivaklé k•lutseš ‘arco’, cf. infra serie (34).

A fin de dar cuenta de estas “irregularidades”, propongo ahora –en remplazo de *h de Viegas Barros (1993)– dos fricativas dorsales no labializadas para el Proto- Mataguayo: una fricativa velar *x y una fricativa uvular *X. Los reflejos de ambos fonemas serían distintos siempre en Maká; en Chorote lo serían solamente en posición final tras a; en Nivaklé en contacto con las vocales *e, *i ; en Wichi, en posición inicial. En los demás contextos del Wichi, Chorote y Nivaklé, ambos proto- fonemas se habrían fusionado, dando un mismo resultado.

Wichi Chorote Nivaklé Maká Inicial demás inicial, inicial, demás Junto a demás inicial, demás contex- final tras i final tras i contex- i, e contex- entre contex tos y e y a tos tos vocales -tos *x O h O – x š x x x *X h h – O x x x h X

Como resultado de ello, ahora postulo el siguiente sistema consonántico para el Proto-Mataguayo:

142 VIEGAS BARROS

*p *t *ts *kj *k *q *p’ *t’ *ts’ *kj’ *k’ *w *l *j w *x *… *s *x *X *m *n

3.1. Proto-Mataguayo */x/ Proto-Mataguayo *x ha dado en Wichi O en posición inicial, h en los demás contextos (en Nocten, x en posición final); en Chorote, O en posición inicial y tras i y e en posición final, x en los demás contextos; en Nivaklé, š en contacto con e, i; en los demás contextos x; en Maká, siempre x La evidencia para la reconstrucción de Proto- Mataguayo *x incluye entre otras a las siguientes series cognadas (reformulo aquí varias de las reconstrucciones presentadas en Viegas Barros 1993):

(18) *xele, Ni4 šek•le, Ma xili ‘suciedad’. (19) *-xe(ne) (cf. Najlis 1984: 42), Wi -hen, Cho -xa, Ni -šane ‘plural del sujeto (sufijo verbal)’. (20) *-xetik (cf. Najlis 1984: 23, 34, 47), Wi -…etek, (con incorporación del prefijo …- posesivo 3a. p.’) Cho -xetek ~ -xitek, Ni -šatic& ‘cabeza’ (Ma -exkiti…a ~ -ixkiti…a ‘cerebro, seso’, con metátesis de t y k y composición con xi…a ‘cabeza’ como segundo miembro). (21) *(V)xejø/, Cho axeje, Ni šejø, Ma xaja/ ‘murciélago’. (22) *xinawøp (cf. Najlis 1984: 33; Viegas Barros 1993: 200, item no. 71), Wi(i)nawop, Cho nawop, Ni šnøwøp, Ma xinawap ‘primavera’. (23) *xina-, Wi (Nocten) /ina- ‘1a. p. inclusivo’, Ni šna-, Ma xinV- ‘3 p. sujeto / 1a. p. inclusivo objeto’. (24) *xita-, Ni šta-, Ma xitV- ‘1a. p. pl. inclusivo sujeto’. (25) *-xop (cf. Viegas Barros 1993: 198, item no. 30), Ni -xop, Ma -xup ‘al lado, alrededor’. (26) *xoxewuk, Ni xoxijuk (*wu > ju por disimilación), Ma xoxewuk ‘palo cruz’ (una planta). (27) *xuwe/la (cf. Najlis 1984: 35), Wi iwela, Cho we/la, Ni xiwe/k•la (Wichi y Nivaklé con disimilación *uw > iw5), Ma xuwel ‘luna’. (28) *xunxetek (cf. Najlis 1984: 47), Wi natek, Cho ixnjetak, Ni xunšatac&, Ma xunxetek ‘tusca’ (una planta). (29) *axa/, Cho axa ‘garza’, Ma exe/ ‘cigüeña’. (30) *…axi, Cho …axi, Ni …aši, Ma …exi ‘puerta’.

4 Utilizo las abreviaturas Cho = Chorote, Ma = Maká, Ni = Nivaklé, Wi = Wichi. Las formas citadas para cada lengua particular están extraídas de las fuentes mencionadas en las Referencias, p. ej.: Gerzenstein (1992, 1994, 1999), Messineo y Braunstein (1990), Seelwische (1975, 1979), Tovar (1981), Viñas Urquiza (1970). 5 Disimulación que habría ocurrido, obviamente, antes del cambio *x > š ante vocal anterior.

PROTO-MATAGUAYO 143

(31) *-øtax (cf. Najlis 1984: 44; Viegas Barros 1993, 201, item no. 81), Wi -atah, Cho -atax, Ni -øtex ‘ser gordo’. (32) *pøjtsax (Najlis 1984: 28, 49), Wi patsah, Cho pitsax, Ni pøtsex ‘cigüeña’. o (33) *-/øx (cf. Najlis 1984: 10, 19, 31; Viegas Barros 1993, 200, item n . 69), Wi -t’øh, Cho -ax ~ -ex, Ni -øx, Ma -/ax ‘piel’. (34) *-lutsex (cf. Najlis 1984: 11, 26), Wi -lutseh, Cho -luxsje, Ni -k•lutseš ‘arco’. (35) *sex, Ni sa/š, Ma …esex (con incorporación de …V- ‘3a. p. posesivo’) ‘hoja de árbol’. (36) *-xwex (cf. Najlis 1984: 29), Wi -xwah, Cho -fwax, Ni -faš ‘hachar’, Ma -fexinetki/ ‘hacha’. (37) *towex (cf. Najlis 1984: 27, 45), Wi toweh (Nocten towex) ‘agujero’, ‘olla’, Cho tjowe ‘agujero’, Ni tawaš ‘vagina’ (con asimilación regresiva de las vocales), Ma towxe-naX ‘aguja’. (38) *nøjix (cf. Najlis 1984: 10, 31, 48), Wi nøjih (Nocten /nøjix), Cho naji, Ni øjiš ‘camino’. (39) *(V)nix (cf. Najlis 1984: 31; Viegas Barros 1993, 199, item no. 60), Wi nih, Cho ni, Ni ni/š, Ma enjix ‘olor’. (40) *tix, Wi -tih-c&o, -tih-pa, Cho -tex-ej ~ -tix-i, Ni -tiš, Ma -tix-xu/ ~ -…ix-xu/ ‘lavar’. (41) *-xwejix, Cho -fweje, Ni -fajiš ‘izquierda’, Ma -fejix ‘derecha’. (42) *-tux (cf. Najlis 1984: 39), Wi -tuh ~ -tuxw, Cho -tux, Ma -tux ‘comer’.

Parece que Chorote -x se mantiene tras e en monosílabos, como muestra la forma de los dialectos Yowuwa y Manjuy nex “olor” frente a la correspondiente de la variedad Yofwaja ni (serie 39). También se mantiene sin caer h inicial en Wichi en un caso como el siguiente, donde el o los factores condicionantes pueden haber sido el monosilabismo y/o la glotalización de la última consonante :

(43) *xup’ (cf. Najlis 1984: 33), Wi hup’, Cho xupaj, Ma xup’el (Chorote y Maká con incorporación de sufijos de plural) ‘pasto’.

En Nivaklé, el cambio de *x a š en contacto con vocal anterior se ve bloqueado si coocurre un contacto con vocal posterior, cf. (26) y:

(44) *-xij, Wi -hi, Ni -ši (tras vocal no posterior), -xi (tras vocal posterior), Ma -xii ‘sufijo que indica lugar donde se realiza la acción’, (45) *ixø, Cho ixje-ti, Ni -jixø/-ši, Ma ixa ‘verdadero’.

3.2.Proto-Mataguayo */X/ Proto-Mataguayo*X se refleja como h en todos los contextos (incluso en posición inicial) en Wichi (en el dialecto Nocten, x en posición final, h en los demás contextos), en Chorote da O en posiciones inicial y final (en este caso sólo tras i y a),

144 VIEGAS BARROS x en los demás contextos; en Nivaklé siempre x; en Maká, h en posición inicial e intervocálica, X en los demás contextos. La evidencia para la reconstrucción de Proto-Mataguayo *X incluye entre otras a las siguientes series cognadas:

o (46) *Xa- (cf. Viegas Barros 1993, 196, item n . 1), Cho a-, Ni xa-, Ma hV- ‘1a. p. sujeto’. o (47) *Xaj- (cf. Viegas Barros 1993, 196, item n . 1), Ni xaj-, Ma hVj- ‘1a. p. sujeto’. o (48) *Xan-, (cf. Viegas Barros 1993, 196, item n . 1) Ni xan-, Ma hVn- ‘1a. p. sujeto’. (49) *Xajawu (cf. Najlis 1984: 41, 43, 47), Wi hajawu (Nocten hijawu/), Cho ajewu ‘chamán’. (50) *Xim, Ni xim, Ma him ‘coatí’. (51) *Xino/ (cf. Najlis 1984: 13, 16), Wi hino/ (Nocten hi/no/) ‘hombre’, Cho ixnjo/ ‘persona’. (52) *Xolo (cf. Najlis 1984: 33, 35), Wi holo, Cho xolo, Ni xo-søx ‘arena’. (53) *Xupel (cf. Najlis 1984: 10, 25, 28, 36), Wi hupel, Cho -peluk ~ -piluk, Ni -xpek ‘sombra’. (54) *tsøXøq, Wi tsøhøk (Nocten tsahaq), Cho saxak, Ma tsahaq ‘chajá’ (un ave). (55) *saXets, Ni saxec& (c& final irregular), Ma sehets ‘pescado (genérico); sábalo’. (56) *wiXelø, Ni wixak•lø, Ma wihil ‘nutria’. (57) *isøX, Cho isje, Ni xo-søx, Ma isaX arena’. (58) *itøX (cf. Najlis 1984: 16), Wi itøh (Nocten itøx), Cho (dialectos Yowuwa y Manjuy) ejtje, Ni itøx ‘fuego’. (59) *kine…itsaX (cf. Najlis 1984: 46), Cho kinjelisa, Ni c&in…itsex ‘garza rosada’. (60) *seulaX (cf. Najlis 1984: 50), Wi sulah (Nocten sulax), Cho seolje, Ni suk•lax ‘oso hormiguero’. (61) *tøtsinaX, Wi tøtnah (Nocten tatnax), Cho taxsina ‘sapo’. (62) *-taX, Wi -tah (Nocten -tax), Cho -tje, Ni -tax, Ma -taX ‘sufijo nominal’. (63) *tutsaX (cf. Najlis 1984: 43), Wi tutsah (Nocten tucax), Cho toxsa ‘humo’, Ni štutax ‘hollín’. w w w (64) *x etenaX (cf. Najlis 1984: 42), Wi x itanah, Cho f etina ‘luciérnaga’. w w w (65) *x oqotsaX (cf. Najlis 1984: 13), Wi x okotsah (Nocten x oqatsax), Ni fokotsex ‘mulita’ (esp. de armadillo). (66) *-ti…øX, Wi -ti…øh, Cho -texlje ~ -tixlje, Ni -ti…øx, Ma -ti…oX ~ -…i…oX ‘cavar’. (67) *wam(xa)…øX (cf. Najlis 1984: 42), Wi wøm…øh (Nocten wan…ax), Cho wan…a, Ni wanxa…øx, Ma waa…aX ‘ñandú’. (68) *-/øX, Cho -aa ~ -ee, Ni -øxøx (las formas Chorote y Nivaklé con reduplica- ción), Ma -/aX ‘voz’.

PROTO-MATAGUAYO 145

4. Algunos problemas remanentes La postulación de dos fricativas dorsales no labializadas para el Proto-Mataguayo parece ser un avance, ya que resuelve una serie de problemas. La mayor parte de las correspondencias pueden explicarse perfectamente a partir de estas dos fricativas. Sin embargo, quedan todavía algunos puntos oscuros en la reconstrucción de la fonología de esta protolengua, y no pocos de ellos se relacionan con las fricativas dorsales. Los principales problemas al respecto son los cuatro siguientes.

4.1. Maká X en posición intervocálica No ha sido aclarada la existencia de Maká X en posición intervocálica, que en la mayor parte de los casos no tiene correspondencia en las otras lenguas Mataguayas, como en los siguientes ejemplos:

(69) Ma qanaXa ‘charata’, (70) Ma tsiXaXa ‘tijereta (un pájaro)’, (71) Ma seXeXeX ‘maní’, (72) Ma aXem ‘no sé’, (73) Ma aXit ‘adulto’, (74) Ma -ajaXix ‘heredar algo’.

Sin embargo, Maká X intervocálica también ocurre en algunas formas con posibles cognados en las otras lenguas de la familia, como:

(75) Ma meXe ‘todavía’ : Wi (Nocten) mhe/ ‘reciente/visto/habitual’, (76) Ma seXesinaX, Cho esiniki ‘serpiente de cascabel’, (77) Ma tseeXe : Ni tsaxax ‘áspero’, w w (78) Ma (i)toXii : Wi toh e, Cho tjof e/, Ni toxej ‘lejos’, (79) Ma peXejek ‘batata’ : Ni pexaja ‘papa, batata’, quizás también Wichi (Nocten) pi/jok ‘mandioca’.

Por lo menos en el caso de (79) podemos decir que es muy probablemente un préstamo, ya que se trata de un verdadero “pan-chaqueñismo” léxico, cf. formas similares en lenguas de las familias Mascoy (Angaité pija ~ pihija, Lengua meridional piheji ‘batata’), Zamuco (Ayoré peheei ‘mandioca’) y Guaicurú (Pilagá pijok ‘mandioca’)6.

4.2. Correspondencias irregulares entre fricativas dorsales Ciertas correspondencias irregulares posiblemente puedan explicarse o como resultado de asimilaciones o disimilaciones de una fricativa dorsal con respecto al punto de articulación de otra consonante dorsal existente en el mismo término, o como préstamos, p. ej.:

6 Las formas de las lenguas de las familias Mascoy, Zamuco y Guaicurú están tomadas de Key (1997).

146 VIEGAS BARROS

(80) Ni -køjiš, Ma -qaweX ‘garganta’ (se esperaría o bien Nivaklé *-x, o bien Maká *-x), (81) Wi (Bazanero) hip’ehi, Ni kxopxi, Ma exupuj ‘tibio, cálido’ (aún explicando la segunda x del Nivaklé por influencia de una vocal posterior presuntamente caída, se esperaría Maká *exupuhi).

4.3. Correspondencia Maká /x/ : otras lenguas /s/ Existen también algunas correspondencias aun no bien establecidas de fricativas dorsales, como la que ocurre entre fricativas dorsales y dentales en los siguientes ejemplos (a los que hay que agregar la alternancia Wichi h ~ s en (3)):

(82) Cho injexkaj (pero en Najlis 1984 sinjaka), Ni snakxaj (variedad oriental šnakxaj), Ma xunkhaj ‘niebla’ (cf. Najlis 1984: 25, 38), (83) Wi silokaj, Cho siljaka, Ni sløkxaj (variedad oriental šløkxaj), Ma xunkhaj ‘gato montés’ (cf. Najlis 1984: 11, 37, 54),

De confirmarse esta correspondencia con otros ejemplos, habría que postular posiblemente un fonema Proto-Mataguayo fricativo palatal o dorsal palatalizado *xj.

4.4. Relación entre fricativas y oclusivas dorsales No se han esclarecido todavía las relaciones entre oclusivas y fricativas dorsales, que se manifiestan en correspondencias irregulares tanto en posición inicial, p. ej. en:

(84) Wi kalatu, Cho alatju, Ni xala/tu ‘granizo’ (posiblemente Proto-Mataguayo *qala/tu ~ *Xala/tu [o *xala/tu], cf. Najlis 1984: 16, Viegas Barros 1993: 200, item n° 73), (85) Wi hotoni, Ma k’ateni ‘especie de mono’ (posiblemente Proto- Mataguayo*k’øtøni ~ *Xøtøni), como en posición final, p. ej. en:

(86) Wi k’atak, Cho kataki, Ni …afkatax ‘mosca’ (cf. Najlis 1984: 23; Viegas Barros 1993, 202, item no. 115), (87) Wi nakwotak, Ni šnakotax ‘abeja’ (cf. Najlis 1984: 34, 42), (88) Wi xwatsuh, Cho impesjuk, Ni fatsuk ‘ciempiés’ (cf. Najlis 1984: 26, 46), (89) Cho siwalak, Ni siwøk•løk, Ma siwalaX ‘araña’ (cf. Najlis 1984: 41).

5. Conclusiones Todas las lenguas Mataguayas tienen por lo menos una fricativa dorsal, y la reconstrucción de la fonología del Proto-Mataguayo debe incluir también por lo menos una fricativa dorsal. Sin embargo, la reconstrucción de una sola fricativa

PROTO-MATAGUAYO 147 dorsal en Proto-Mataguayo no es suficiente para dar cuenta de las correspondencias documentadas en las lenguas Mataguayas. Dos fricativas dorsales originarias pueden explicarlas mucho mejor, aunque siempre parece quedar un residuo de correspondencias sin explicación. La solución satisfactoria de estos problemas se irá alcanzando, sin duda, a medida que vaya avanzando la investigación tanto descriptiva como comparativa de las lenguas Mataguayas

Referencias bibliográficas Alvarsson, J.-Å. 1984 Wenhayek Lhamet. Introducción al mundo de los matacos-noctenes de Bolivia, Cochabamba: Misión Sueca Libre en Bolivia, 2a. ed. revisada. Campbell, Lyle 1997 American Indian Languages. The historical linguistics of native America, (Oxford studies in anthropological linguistics), New York: Oxford University Press. Claesson, K. 1994 ‘A phonological outline of Mataco-Noctenes’, International Journal of American Linguistics, 60 (1): pp. 1-38. Gerzenstein, A. 1978-9 Lengua chorote, (Archivos de Lenguas Precolombinas, 3), 2 vol, Buenos Aires: Universidad de Buenos Aires, Facultad de Filosofía y Letras, Instituto de Lingüística. 1983 Lengua chorote. Variedad 2, (Archivos de Lenguas Precolombinas, 4), Buenos Aires: Universidad de Buenos Aires, Facultad de Filosofía y Letras, Instituto de Lingüística. 1987 ‘The Chorote language: three examples of internal reconstruction’, M. R. Key (ed.), Comparative Linguistics of South American Indian Languages, special issue of Language Sciences, 9 (1): pp. 11-15. 1989 Lengua maká: Aspectos de la fonología, Buenos Aires: Universidad de Buenos Aires, Facultad de Filosofía y Letras. 1992 ‘Una variedad oriental del mataco’, Algunos aspectos fonológicos y morfológicos, en: J. Braunstein (ed.), Hacia una nueva carta étnica del Gran Chaco III, Las Lomitas (Formosa, Argentina): Centro del Hombre Antiguo Chaqueño (CHACO), pp. 80-97. 1994 Lengua maká. Estudio descriptivo, (Colección “Nuestra América”. Serie: Archivo de Lenguas Indoamericanas), Buenos Aires: Universidad de Buenos Aires, Facultad de Filosofía y Letras, Instituto de Lingüística. 1999 Diccionario etnolingüístico maká-español. Índice español-maká, (Colección “Nuestra América”. Serie: Archivo de Lenguas Indoamericanas), Buenos Aires: Universidad de Buenos Aires, Facultad de Filosofía y Letras, Instituto de Lingüística.

148 VIEGAS BARROS

Key, M. R. (ed.) 1997 Intercontinental Dictionary Series. Vol. I. South American Indian Languages, Partes I y II, Provisional Unpublished Proofs. Messineo, C. y J. Braunstein 1990 ‘Variantes lingüísticas del mataco’, en: J. Braunstein (ed.), Hacia una nueva carta étnica del Gran Chaco I, Las Lomitas (Formosa, Argentina): Centro del Hombre Antiguo Chaqueño (CHACO), pp. 1-13. Najlis, E. L. 1984 Fonología de la protolengua mataguaya, (Cuadernos de Lingüística Indígena, 9), Buenos Aires: Universidad de Buenos Aires, Facultad de Filosofía y Letras, Instituto de Lingüística. Seelwische, J. 1975 Na Lhasinônash napi Nivacle. Gramática Nivacle, Asunción: “El Gráfico” S. R. L.. 1979 Diccionario Nivacle-Castellano, Mariscal Estigarribia: ASCIPM Filadelfia- Ed. El Gráfico. Stell, N. N. 1972 Fonología de la lengua axluxlaj, (Cuadernos de Lingüística Indígena, 8), Buenos Aires: Universidad de Buenos Aires, Facultad de Filosofía y Letras, Centro de Estudios Lingüísticos. 1989 Gramática descriptiva de la lengua niwaklé (chulupí), Tesis para aspirar al título de Doctora en Filosofía y Letras, orientación Letras, Presentada al Departamento de Graduados de la Facultad de Filosofía y Letras de Universidad de Buenos Aires, 2 vol, Ms. Tovar, A. 1981 Relatos y diálogos de los matacos, seguidos de una gramática de su lengua, Madrid: Instituto de Cooperación Iberoamericana. Viegas Barros, J. P. 1993 ‘¿Existe una relación genética entre las lenguas mataguayas y guaycurúes?’, en: J. Braunstein (ed.), Hacia una nueva carta étnica del Gran Chaco IV, Las Lomitas (Formosa, Argentina): Centro del Hombre Antiguo Chaqueño (CHACO), pp. 193-213. Viñas Urquiza, M. T. 1970 Fonología de la lengua mataca, (Cuadernos de Lingüística Indígena, 7), Buenos Aires: Universidad de Buenos Aires, Facultad de Filosofía y Letras, Centro de Estudios Lingüísticos. 1974 Lengua mataca, (Archivos de Lenguas Precolombinas, 2), 2 vol, Buenos Aires: Universidad de Buenos Aires, Facultad de Filosofía y Letras, Centro de Estudios Lingüísticos.

CASHINAHUA PERSONAL PRONOUNS IN GRAMMATICAL RELATIONS

Eliane Camargo* CELIA and NHII

1. The group The Cashinahua language belongs to the Panoan family, composed of nearly thirty languages spoken in the Amazonia Lowlands, on both sides of the Bolivian, Brazi- lian and Peruvian borders. Cashinahua is spoken by about 5.400 people (Ricardo 2001: 12) who live along the Brazilian-Peruvian border. The Cashinahua refer to themselves as huni kuin1 ‘kuin man’, translated as ‘real man’ (Kensinger 1994), and the name of their language is hanca kuin ‘kuin lan- guage’. The term kuin refers to part of the Cashinahua socio-cultural system of clas- sification (Deshayes & Keifenheim 1994, Erikson 1996, Lagrou 1998). Within the Cashinahua territory, they speak their language exclusively. Although they learn to read and write in their mother tongue, their whole school education is done in Portu- guese, for those who live in Brazil, and in Spanish, for those who live in Peru.

2. The language Cashinahua is an that uses almost exclusively suffixes and has no prefixes. The is verb-final and the basic sequence is SOV. In this language, the different classes of noun phrase have different morphological systems when they appear as S, A, or P of a clause. Thus, nouns (‘man’ huni-n A, huni-ø S/P) have an ergative-absolutive case marking system (1), pronouns (1st and 2nd person singular, and all non singular persons: ‘we’ nu-n A/S, nuku-ø P) have a marking system (2), and finally the 3rd person singular presents a neutral case system with only one form for all three functions (ha or ø as S/A/P) (3). This split ergativity system is shown in the table below:

* Member of Centre d'Études des Langues Indigènes d'Amérique (CELIA: American Indigenous Lan- guage Studies Center), and Nucleus of Indigenous History and Indigenous Studies- NHII/University of . A study on Cashinahua complementation (‘actancy’) was presented at the French research group about actancy (RIVALC - GDR 0749, CNRS). I thank Sergio Meira and Hein van der Voort for their comments on this paper, Vladmir Nedjalkov and Gilbert Lazard for the discussions we had about Cashinahua syntax. 1 This language has eighteen : (a) four vowels: /a/, /i/, /´/ (noted "i" in the examples), /u/, and (b) fourteen consonants: /p/, /t/, /c/ (regarded as an affricate /t/), /k/, /b/, /d/, /j/ (occlusive palatal sound ), /s/, /ß/ (a voiceless fricative retroflex), /h/, /ts/, /m/, /n/, /w/. The data in this study are transcribed in accor- dance with the Cashinahua phonological system. It is worthy to note that in coda position represents of the preceding vowel: = [kam] ‘dog’. 150 CAMARGO

System Univalent Bivalent verbs verbs ergative-absolutive nouns S-ø A-{a/i}n2 P3-ø

nominative-accusative pronouns SG (1st, 2nd) S-n A-n P-a (SG) PL (1st, 2nd, 3rd) P-ø (PL) neutral system pronoun ø ø ø SG (3rd) Table 1: The split ergative system

(1) a kaman -an baki -ø kiju -mis -ki dog -A child -P bite -HAB -ASS ‘I assert that the dog bites the child.’

b kaman -ø ußa -mis -ki dog -S sleep -HAB -ASS ‘I assert that the dog is always sleeping.’

(2) a mi -n i -a kiju -mis -ki 2SG -A 1 SG -P bite -HAB -ASS ‘I assert that you bite me.’

b mi -n ußa -mis -ki 2SG -S sleep -HAB -ASS ‘I assert that you are always sleeping.’

(3) a (ø) (ø) kiju -mis -ki (3SG.A) (3SG.P) bite -HAB -ASS ‘I assert that s/he bites her/him.’

b (ø) ußa -mis -ki (3SG.S) sleep -HAB -ASS ‘I assert that he/she is always sleeping.’

2 This case marker –n has two allomorphs –{a/i}n in different environments. These morphophonological changes occur only with nouns. Rule: (a) with final CV syllabes, -an occurs if the final vowel is /u/ : ainbu ‘woman’ > ainbu-n or ainbu-an. With the vowel /i/ it is only attested with the word badi ‘sun’ > badi-an. In all others contexts, the suffix –n appears; (b) in CVC final syllables, a rule of vowel harmony occurs with /i/: jaiß-in ‘armadillo’. With other vowels –an occurs: kaman-an ‘dog’, amin-an ‘capivara’ (currently the biggest rodent – Hydrochoerus hydrochoeris (L)), bikun-an ‘blind’. For a sketch of the split ergative system, see Camargo (2002). 3 The glosses: S(ubject) of an intransitive, A(gent), and P(atient) of transitive constructions, identify the three semantico-syntactic roles. CASHINAHUA PRONOUNS 151

This paper is concerned with the personal pronoun system and its irregular morphology. The irregularity depends on the case marking associated with the pronouns. Grammatical relations are indicated by different suffixes: (a) -n marks the nominative case on pronouns (and ergative on nouns4), and the (and the locative, and vocative cases5) on both nouns and pronouns. It is notable that the accusative and dative case marking show different morphological markers for nouns and singular and plural pronouns. A zero-morpheme, -ø, marks the nouns (huni-ø ‘man’ P) and the person plural (nuku ‘us’, matu ‘you’, hatu ‘them’ P), while in modern Cashinahua the 1st and 2nd person singular are marked by -a; (b) -win marks the the argument of reason and the 3rd person singular in a genitive construction (and the when associated with a noun); (c) -ki indicates the dative as subject of a state; (d) -bi (and -bitan) comitative; (e) -anu marks the allative, and (f) - anua the . These different syntactic functions and the representation of their semantic roles and pragmatic status are exemplified by the data collected in Peru during different field trips (1994-1999). In section 3., a sketch of the personal pronoun morphology and the different case markers is put forward. In section 4. I present some peculiar uses of the 1st and 2nd person singular, and a particular combination between these two persons, in a genitive construction. In section 5. the uses of the 3rd person either of the singular or of the plural are also shown, with special attention to the neutral system. In section 6. a summary of the morphology of the pronouns is presented, and in section 7. the neu- tral system is shown. In section 8. Dixon’s and Payne’s analysis of the Cashinahua split system is presented and discussed.

3. Personal pronouns The Cashinahua personal pronoun system is composed of seven forms, including two third-person: hatu and habu. The hatu form refers to a homogenous concept of per- son. The habu form relates to a heterogeneous concept of person, as described in section 5. Table I summarizes these forms, and presents the morphology of Cashina- hua pronouns in the three basic semantico-syntactic roles, termed S(ubject of intran- sitive), A(gent of transitive) and P(atient of transitive).

4 The suffix -{a/i}n marks the ergative: (i) huni-n nami-ø pi-mis (ii) huni-ø daja-mis man-A meat-P eat-HAB man-S work-HAB ‘The man usually eats meat’ ‘The man usually works’ 5 The vocative case marking appears on nouns and it is represented by the suffix -n. It is used on proper nouns madia-n 'hey Maria!' and especially on kinship terms: huci-n 'hey oldest brother!', cai-n 'hey bro- ther-in-law!'. 152 CAMARGO

free form P-a, ø (SG), -ø (PL) S/A-n Intr Trans SG 1. i-a i-a-n i-a i-n 2. mi-a mi-a-n mi-a mi-n 3. ha (*ha-a) ha ø ø PL 1. nuku-ø nuku-n nuku-ø nu-n 2. matu-ø matu-n matu-ø ma-n 3(ho). hatu-ø hatu-n hatu-ø (hatu-n)6 3(he). habu-ø habu-n habu-ø (habu-n) Table 2: Personal pronoun morphology as S/A and P

3.1. The topic forms The free pronominal form represents a left-dislocated, stressed participant. It occurs in the initial position of the sentence, cf. (4-7):

(4) nuku, nu -n ka -ai, ikis -dan 7 1PL, 1PL -S go -PROG, today -dan ‘We are going today.’ (lit. ‘we, we are going today’)

(5) ha inun ia, nu -n ka -ai, bai -anu -dan 3SG and 1SG, 1PL -S go -PROG slash and burn -ALL - dan ‘He and me, we are going to the plantation.’

In a transitive construction, the ergative marker -n is associated with the free form (6b-c), except in the 3rd person singular, is not expressed by overt morphology (7b).

(6) a ia -di, i -n ka -ai mi -bi daja -i -dan 1SG -also, 1SG -S go -PROG 2SG -SOC work -INF -dan ‘Me too, I am going to work with you.’

b ia -n -di, i -n bi -ai mabu -dan 1SG -A -also, 1SG -A take -PROG stuff -dan ‘Me too, I am buying things.’ (lit. ‘me too, I am taking things’)

c ia -n bisti, i -n ßinan -ai 1SG -A only, 1SG -A think -PROG ‘Only I am thinking about him. (lit. ‘only me, I am thinking’)

6 The parentheses indicate that the use of these forms as S/A is optional. 7 Morphemes with glosses in italics are still under study. CASHINAHUA PRONOUNS 153

(7) a ha, ø daja -paki -mis -ki 3SG, 3SG.S work -paki -HAB -ASS ‘He works continuously.’ (lit. ‘him, he works continuously’)

b ha (-dan), ø atsa -ø pi -mis -ki 3SG (-dan) 3SG.A manioc -P eat -HAB -ASS ‘He eats manioc.’ (lit. ‘him, he eats manioc’)

3.2. The nominative subsystem In the nominative subsystem, which concerns all pronouns (except 3rd person singu- lar), both S (2b, 4-5, 6a, 7a, 8) and A (2a, 6b-c, 7b, 9, 10), semantic roles are marked by the suffix -n. Nevertheless, in the , the P argument (2a, 7b, 9, 10) is not expressed by overt morphology.

(8) ußi badi -n ma -n hiki -ai moon sun -LOC 2 PL -S arrive -PROG ‘You are coming home in the moonlight.’

(9) mi -n nuku -ø bicipai haida -ai 2SG -A 1 PL -P like much -PROG ‘You like us very much.’

(10) a i -n matu -ø bicipai haida -ai 1SG -A 2 PL -P like much -PROG ‘I like you very much.’

b ma -n hatu -ø bicipai -ai 2PL -A 3 PL -P like -PROG ‘You like them.’

In the nominative case, the third person singular does not have to occur overtly, whe- ther it is S (11), A (12a, 13) or DAT (12b):

(11) ni midan ø ka -mis -ki jungle inside 3SG.S go -HAB -ASS ‘He always goes in the deep jungle’

(12) a ø atsa -ø pi -mis -ki 3SG.A manioc -P eat -HAB -ASS ‘He eats manioc.’

154 CAMARGO

b huni -n ø baka-ø inan -mis -ki man -A 3 SG.DAT fish -P give -HAB -ASS ‘The man is used to give (her/him) fish.’

(13) is extracted from the narrative about the character Basnen Pudu (BP). It tells the story of a woman who used to weave beautifully and quickly. Nobody could unders- tand how she could weave so quickly. When she got some cotton thread, she very quickly wove a hammock. BP’s sister-in-law was wondering where BP kept the thread8. This example shows the sequence 3rd person A and 1st person P. In this order, the 3rd person does not appear overtly either:

(13) hani ø i -a a -kain -minkain where 3SG.A 1SG -P range -MOT -INTER ‘I wonder where she could have put (my cotton)?’

Absence of an overt third person pronoun usually suggests that it is singular. If a plural is meant, then the 3rd-person plural forms have to be used: P (10b, 14), and optative in A (15) and S semantic roles:

(14) ikis, i -n hatu -ø nuku -ßu -ki today, 1SG -A 3 PL P meet -COMPL -ASS ‘I met them today.’

(15) hatu -n bisti bi -ßu -ki juinaka -dan, 3PL -A only take -COMPL -ASS game -dan,

ha bisti -tu -n pi -iki -ki 3SG.P only -PL -A eat -EVID -ASS ‘Only they took game, (it seems that) only they eat it.’

In (15), the 3rd person singular represented by ha refers to the game. It is an anapho- ric pronoun.

3.3. The dative case Trivalent propositions normally involve an A, a P and a DAT(ive). If the dative re- presents an argument of ‘reason’ of a state9, it is marked by a special suffix: -ki (16). However, if the dative is the actual receiver of the patient in a transfer-of-possession predicate, then it is not morphologically marked in the third porson (17):

8 Basnin Pudu's sister-in-law is wondering where BP put all the cotton thread she gave her to weave. She looks everywhere and is unable to find it, as BP put the cotton in the space stretching from her womb to her navel. As the sister-in-law mistrust BP’s skill of weaving, BP got ashamed. 9 Note Givón (1984: 88) about the dative argument of a state: “that state is most likely to be mental”. CASHINAHUA PRONOUNS 155

(16) a nu -n mi -ki dati -ai 1PL -S 2SG -DAT afraid -PROG ‘We are afraid of you.’

b i -n hatu -ki binima -mis -ki 1SG -S 3PL -DAT happy -HAB -ASS ‘I am happy with them.’

c i -n ha -ki nama -mis -ki 1SG -S 3SG -DAT dream -HAB -ASS ‘I usually dream of him/her.’

The examples below show that the pronouns marked by Dative (DAT) elements may have a Benefactive (BEN) or Indirect Object (IO) function.

(17) a i -n mi -a baci -ø inan -ai 1SG -A 2 SG -DAT dress -P give -PROG ‘I am giving you a dress.’

b i -n hatu -ø baci -ø inan -ai 1SG -A 3 PL -DAT dress -P give -PROG ‘I am giving them a dress.’

c Esperansa -anua, ø i -a baci -ø bi -ai Esperanza -all, 3SG.A 1SG -DAT dress -P bring -PROG ‘She is bringing me a dress from Esperanza.’

3.4. The comitative case The suffix -bi marks the case role with the meaning of ‘together with’:

(18) a i -n mi -bi tadi hanca -ai 1SG -S 2SG -COM private talk -PROG ‘I am talking in private with you.’

b bai -anu mi -n ha -bi ka -ai slash and burn -ALL 2 SG -S 3SG -COM go -PROG ‘You are going with her to the plantation.’

3.5. The genitive case In a genitive construction, all the personal pronouns are marked by -n (19), except the 3rd person singular which is marked by an instrumental morpheme -win (20):

156 CAMARGO

(19) a i -n hiwi 1SG -GEN house ‘My house.’

b mi -n hiwi 2SG -GEN house ‘Your house.’

c nuku -n hiwi 1PL -GEN house ‘Our house.’

d matu -n hiwi 2PL -GEN house ‘Your house.’

e hatu -n hiwi 3PL -GEN house ‘Their house.’

Comparing the case-marking morphology of (19a-b) and (2, 6, 9-10a, 14, 16b-c, 17a, 18), note identical personal pronoun forms for 1st and 2nd person singular (i-n ‘1SG’, mi-n ‘2SG’) in the genitive construction and in S/A semantic roles. The difference between genitive case-marking in singular and plural is that in the plural, number is specific for each person: -ku characterizes the 1st person nu-ku; -tu shows the 2nd person and the 3rd person homogeneous: ma-tu and ha-tu; and -bu marks the 3rd person heterogeneous: ha-bu, i.e. a generic plural. Plural pronouns in the genitive case have the same form as those in S/A semantic role. The 3rd person singular is represented by ha, and the morpheme –win is attached to it:

(20) a ha -win hiwi 3SG -GEN house ‘his/her house.’

b ha -win baki -bu 3SG -GEN child -PL ‘his/her children.’

c ha -win ibu huni piaja ka -ai -dan 3SG -GEN genitor man/male hunting go -PROG -dan ‘His father is going hunting.’

CASHINAHUA PRONOUNS 157

3.6. The genitive case with the function of ‘reason’ As shown above, the suffix -win functions as a genitive case marker only in the 3rd person singular. With the others persons, it indicates a reason adjunct ‘because (of)’:

(21) mi -win tai -a diti -nami -dabi -kan -iki -ki, 10 2SG -GEN begin -STAT hit -REC - DU -PL -EVID - ASS,

tomas inun jakobo -dan tomas and jakobo -dan ‘Tomas and Jacobo, they hit each other because of you.’ (lit. ‘having started with you, they hit each other, Tomas and Jacobo)

3.7. The allative and ablative cases The is marked by -anu, which, together with pronouns, indicates motion towards ‘the place where one lives’ (22)11. Pronouns with -anu can also have a loca- tive meaning referring to the ‘inside’ of a person (23):

(22) a mi -anu i -n hu -ßian -ki 2SG -ALL 1 SG -S arrive -PAST -ASS ‘I arrived at your place.’

b nuku -anu ka -ßan -ai 1PL -ALL go -PROSP -PROG ‘He is going to our place.’

(23) i -anu j ußin pipa, hiwi -a -ki 1SG -ALL “soul” good, live -STAT -ASS ‘In me lives a good soul.’

The suffix -anua marks the ablative case which indicates a motion ‘from’ (25). Example (24) comes from a dialogue, and means literally "no noise is heard from one’s place", i.e., "it is very calm at one’s place".

(24) mi -anua unan -uma haida -ki 2SG -ABL know -PRIV much -ASS ‘It is very calm where you live.’ (lit. ‘from your place, there is not much noise’)

10 A preliminary study on evidentials in Cashinahua is presented in Camargo (1996a). 11 It also occurs with nouns: (i) ikis bai-anu mi-n ka-ai today slash and burn-ALL 2sg-S go-PROG ‘Today, you are going to the plantation.’ 158 CAMARGO

(25) a mi -anua i -n hu -di -ai 2SG -ABL 1 SG -S arrive -quick -PROG ‘I’ve just com from your place.’

b bai -anua i -n hu -ai slash and burn -ABL 1 SG -S arrive -PROG ‘I am arriving from the plantation.’

4. Semantic uses of 1st and 2nd person singular I would now like to turn to some special situations involving 1st and 2nd person sin- gular pronouns. In many cases in dialogues, the pronoun in S/A function can be omitted (26-27). On the other hand, the P pronoun is obligatory. The relation bet- ween the narrator and his audience is marked only by the 2nd person P pronoun, like mia in the examples below, extracted from the story about Basnen Pudu:

(26) bi -wi ßuntiß -wi mi -a a -ßun -dan, ak -a bring -IMPER hull -IMPER 2SG -P make -OR -dan, say -STAT ‘Bring it! Do the hulling! (I’ll) weave it for you, she said.’

Example (27) shows a conventional formula used to start narratives:

(27) iska -ni -kiaki mi -a jui -nun, ninka -wi so -INDF.PAST -PAST.EVID 2SG -P tell -nun listen -IMPER ‘(lit.) That was so, (I) tell you, listen!’

As mentioned above, the basic order of the constituents is SOV, and the sequence i-n mi-a (I you) can usually be treated as transitive. However, in some cases, this se- quence has a different meaning. It refers to a possessive construction as shown in (28) by the sequence i-n mi-a ‘I you’, which in this case means ‘my possessed ele- ment/hammock made by you’. Extracted from the "BP" narrative, this sentence des- cribes how the hammocks, woven by Basnen Pudu, are very much appreciated by her sister-in-law, who asks her to weave one as soon as possible. This kind of possessive indicated by the sequence in mia expresses a distant family relationship, cf. (28), the literal reading of which seems to suggest ‘(for) you, I am a distant family’. However the sequence in min (29) refers to a close relationship, the literal interpretation of which suggests ‘my family is your family’.

(28) i -n mi -a inawai 1SG -GEN 2SG -P distant family ‘You are my distant family.’

CASHINAHUA PRONOUNS 159

(29) a i -n mi -n nabu kajabi 1SG -GEN 2SG -GEN family ‘real’ ‘You are my next of kin.’

b i -n mi -n caita kajabi -ki 1SG -GEN 2SG -GEN potential husband ‘real’ -ASS (lit.) ‘My potential husband is yours.

5. Semantic uses of the 3rd person pronouns: ha, hatu and habu. The third person form is ha. This form, representing the 3rd person singular, appears as a free form and is always placed at the head of the sentence:

(30) ha (-dan), atsa -ø pi -mis -ki 3SG (-dan) manioc -P eat -HAB -ASS ‘He, he usually eats manioc.’ (lit. Him, he eats manioc.’)

This form does not appear in the syntactic-semantic roles S/A and P, as shown above (7, 11-12). However, in any syntactic-semantic role where the 3rd person form ha appears, it is anaphoric:

(31) mai -n ha misti hu -mis -ki land -LOC 3 SG.S only come -HAB -ASS ‘The one [we are talking about] comes on foot by land.’

In contexts where it is used anaphorically, ha can refer to an A, as in (32a), or refer to a P, as in (32b-c) and in (15).

(32) a ia -ø ha a -mis -ki, hancawan i -n ain -nan -dan 1SG -P 3SG.A do -HAB -ASS speak.strongly 1SG -GENwife -nan -dan ‘She does so to me, my wife speaks (to me) strongly.’ (lit. She, whom we are talking about, does so to me)

b i -n ha a -mis -ki, hancawan i -n ain -dan 1SG -A 3 SG.P do -HAB -ASS speak.strongly 1SG -GEN wife -dan ‘I do so to her, (I) speak strongly to my wife.’

c mi -n ha haska wa -ma -i -dan, 2SG -A 3 SG.P so do -FAC - ASP -dan

jusin ninka -mis -ma -ki learn listen -HAB-NEG -ASS ‘You must do so, (you) don’t listen to learn.’ 160 CAMARGO

In the extract below, ha in (33) is used anaphorically as the Agent of the sentence. This myth tells the story of a man (Mana Dumeya - MD) who has to take care of his child (his married daughter) as his son-in-law is lazy. In order to get food, MD be- comes a when he goes hunting. In this extract, it is almost dawn, so MD tells himself that it is time to go home. In (33), ha refers to MD when he is (already) back home putting away his arrows :

(33) — ka -di -tan -nun, ißun, hu -iki -dan — go -quick -MOT -nun, he.thinks, come -EVID -dan

ha ha -win pia adu -bain -a -dan 3SG.A 3SG -INSTR arrow put.away -MOT -STAT -dan ‘— It’s time to go home quickly, he thinks. He’s back home, (where) he puts away his arrow.’

Concerning the 3rd person, in other such as Capanahua (Loos 1999), Shipibo (Valenzuela 1997), the 3rd person singular marker is ha (or haa), or a in Marubo (Costa 1997)12. It receives the case suffix -n, to mark the nominative (S/A). In modern Cashinahua, the nominative and the accusative case marking, -n and -a respectively, are linked to the 3rd person singular only in a specific combina- tion which appears sentence initially Both constructions express ‘also’. The nomina- tive case marking -n appears with the term tsidi ‘also’: ha-n tsidi (3sg-nom also) ‘he too’, cf. (34)13. The accusative marking -a appears in the combination with the im- minent suffix –di: ha-a-di ‘he too’, cf. (35). The data obtained with these forms show that with the nominative case marking, the first participant and/or agent is focused, and with the accusative, attention is focused on the benefactive.

(34) ha -n tsidi, (ø) bi -ai, mabu -dan (referring to A) 3SG -A too, (3sg) buy -PROG, stuff/thing -dan ‘Him too, he is buying things.’

(35) ha -a -di, (ø) bi -ai, mabu -dan (referring to P) 3SG -P -too, (3sg) buy -PROG, stuff -dan ‘(For) him too, he is buying stuff (for himself).’

12 Capanahua and Shipibo are spoken in Peru, and Marubo in Brazil. 13 This combination covers all the personal pronous: (i) mi-n tsidi pi-ju-wi 2sg-A also eat-ju-IMPER ‘So do you, eat!’ CASHINAHUA PRONOUNS 161

The sentences below show that in a transitive construction, the nominative case mar- ker -n is restricted to the 3rd person plural (36). This case marker, -n, does not in- volve the 3rd person singular (30) (repeated here for convenience):

(36) hatu -n (-dan), (haut -n) atsa -ø pi -mis -ki 3PL A (-dan), (3PL.ho -A) manioc -P eat -HAB -ASS ‘Them, they eat manioc.’

(30) ha (-dan), ø atsa -ø pi -mis -ki 3SG (-dan) 3SG.A manioc -P eat -HAB -ASS ‘Him, he eats manioc.’

The plural form is a combination between ha and plural values: ha-tu (homogenous plural) and ha-bu (heterogeneous plural). In fact, with hatu, the speaker refers to people he is in a close relationship with - for instance people from his village or his next of kin. With habu, the speaker refers to the people he is in a close relationship with (those he refers as hatu) and also with his/her distant kinship and/or with people from outside of his/her village. In some case, it can be interpreted as a collective morpheme. Contrary to the 3rd person singular, plural person markers are required in a P semantic role. In (37), for instance, the narrator refers to a cotton weaver who teaches people how to make weaving patterns. The pronoun habu indicates that the weaver taught everybody how to weave patterns.

(37) ø habu -ø uin -ma -a -ki 3SG.A 3pl (he) -P see -FAC - STAT -ASS ‘She showed them (how to make the weaving patterns).’

However it is optional in S/A semantico-syntactic roles. In this case, they appear to avoid ambiguities, as shown in (37). In a transitive construction, the 3rd person plu- ral receives the nominative case marker, -n:

(38) a hatu -n, tama bidu a -kan -iki -ki 3PL -A, peanut seed do -PL -EVID -ASS ‘Them, (it seems that) they peel peanut.’ (homogeneous plural) (The speaker refers to the people of his/her village)

b habu -n, tama bidu a -kan -iki -ki 3PL -A, peanut seed do -PL -EVID -ASS ‘Them, (it seems that) they peel peanut.’ (heterogeneous plural) (The speaker refers to the people in general, those from his/her village, and those from other Cashinahua villages)

162 CAMARGO

Normally, the plural (-bu or -kan) is associated with the predicate, as in (39b). In this case, the argument in the nominative case represented by the 3rd person plural is not required, cf. (38b). However, as mentioned before, the 3rd person plural in a S/A function appears to avoid ambiguities and to distinguish whom the speaker is refer- ring to: either members of his next of kin (38a) or persons who are not necessarily of his kin (38b). Comparing (39a-b), I note that in (39a) the argument, which is not ex- pressed morphologically, refers to a 3rd person singular. In (39b), the predicate is marked by the plural suffix -bu, which indicates that the subject argument is the 3rd person plural. In (39c), the predicate is not marked by the verbal plural suffix (-bu or -kan), but there we find a 3rd person plural pronoun:

(39) a ø tama bidu a -mis -ki 3SG peanut seed do -HAB -ASS ‘She usually peels peanuts.’

b tama bidu a -mis -bu -ki peanut seed do -HAB -PL -ASS ‘They usually peel peanuts.’

c hatu -n tama bidu a -mis -ki 3PL.ho -A peanut seed do -HAB -ASS ‘They usually peel peanuts.’

6. Summary of personal pronoun morphology In the examples presented above, morphological variations of the personal pronouns have been examined. Below, all the pronoun forms are shown in detail to indicate the loss of the final syllables. The two forms of the plural, hatu and habu, are not inclu- ded in this syllable reduction.

The 1st and 2nd person singular The 1st and the 2nd person singular root is: i- and mi- to which each different case- marking attaches.

SG S/A, gen Free form P, dat/ben dat(subj) soc gen abl/all -n -a (univalent) -a -ki -bi -win -anu(a) -an (bivalent) 1. i- 2. mi- Table 3: 1st and 2nd person singular

CASHINAHUA PRONOUNS 163

The 1st and 2nd person plural The 1st and 2nd person present a syllabic variation only in the nominative case: nuku → nu-, matu → ma- only in S/A case marking.

PL S/A Free form, gen dat(subj) soc gen abl/dir P, dat/ben -n -ø -n -ki -bi -win -anu(a) 1. nu- nuku- 2. ma- matu- Table 4 : 1st and 2nd person plural

The 3rd person singular and plural Especially in narrative, the 3rd person singular is represented formally in the three basic syntactic-semantic roles, S, A and P. In discourse, it is normally not represen- ted. It does, however, occur to mark an anaphoric relation, which refers either to A or to P. Elsewhere its form ha appears and receives the different case markers.

3sg S/A,P, dat/ben Free form gen dat(subj) soc gen abl/all -ø -ø -win -ki -bi -win -anu(a) speech - ha- situation narrative ha- situation Table 5 : 3rd person singular

In Loos’ overview on Pano languages (1999: 236), it is said that “All Pano langua- ges are characterized by a distinctive transitivity concord system”. His examples (25a-b), probably from Capanahua, cannot be compared to modern Cashinahua. In these examples, the 3rd person singular form is haa, which receives the ergative case marking –n: in transitive (25a) haa-n ta his-i-ki (3-A / decl / see-pres-fact) ‘He sees (it)’, but not in reflexive (25b) haa ta his-it-iQ-ki (3S / decl / see-refl-pres-fact) ‘He sees himself (in a mirror)’. In modern Cashinahua the forms ha-n and ha-a occur with the adverbial expression ‘him/her too’ as indicated above (34-35). Historically speaking, it may be the case that in earlier Cashinahua the 3rd person singular form was ha. In modern Cashinahua, it acts as an anaphoric or cataphoric pronoun in the context of storytelling, representing either S, A or P. However, similar to Cashina- hua, the form haa in Capanahua also refers to an anaphoric pronoun: “haa ‘that’, whatever has been referred to before in the context”, writes Loos (1999: 248). The 3rd person plural keeps its full form while marking all cases:

164 CAMARGO

Free form S/A, gen P, dat/ben dat(subj) soc gen abl/dir -ø (univalent) -n -ø -ki -bi -win -anu(a) -n (bivalent) 3ho hatu- 3he habu- Table 6: 3rd persons plural

7. The neutral system As mentioned before, Cashinahua has a split ergative system, characterized by the ergative-absolutive case marking system for noun phrases and by the nominative- accusative system for pronouns. In the examples presented, all the pronouns operate on a nominative-accusative basis, except the 3rd person singular as shown in Table 4. When the latter appears in its form ha it does not carry the marker of the ergative case in A role. Thus, the examples (31-32) show that ha in S (31), A (32a) and P (32b-c) is indeed morphologically unmarked. The three primitives (A, P and S) have the same form: “this is tantamount to lack of case marking for these relations” which is a neutral system “widespread in the languages of the world”, writes Comrie (1989:125). So, in the Cashinahua argument hierarchy diagram, the noun phrases are on the right and the pronouns on the left. In the middle of this hierarchy there is an overlap where the 3rd person singular A or P is unmarked. This diagram refers to the type of arguments which are more likely in A than in O function, as suggested by Dixon (1994: 85):

pronouns > nouns nominative > neutral > ergative 1st > 2nd > 3rd > Proper and common nouns: (including demonstratives) animate and some inanimate such as ‘wind’ and ‘sun’. Table 7: The Cashinahua argument system hierarchy diagram

In the Cashinahua system, an overlap in the split ergative system happens when the ergative marking stops at the place where the accusative begins. That is, the middle portion of the hierarchy will have the same form for all three of the core functions S, A, O. The third person singular shows the same case-marking form for all these dif- ferent syntactic-semantic roles.

8. Dixon’s and Payne’s analyses of the Cashinahua system The examples show that the full form for the singular pronouns are: i- 1st, mi- 2nd, ha 3rd. Dixon (1994: 85-87) analyses the form -a as a suffix which associates with all these persons in O function: i- → ia, mi- → mia, ha- → haa, which is valid only for the 1st and 2nd person, as I have shown above. The data show that the form haa does not correspond to the O function in modern Cashinahua. The 3rd person singu- CASHINAHUA PRONOUNS 165 lar, as an argument, does not have any phonological realization. The 3rd person ana- lysis as presented by Dixon (see Table 8) suggests that there are three different forms for each of the A, S, P semantic roles. The form habu as it is presented, may indicate the 3rd person singular form in A/S function, when it corresponds to one of the two 3rd persons plural. In fact, habu refers to the plural, while ha refers to the singular anaphorically. The A marking is not a nasalisation, but a -n, that on proper and commun nouns shows a morphophological variation -{a/i}n, depending on the final syllable of the lexeme to which is attached (see the rules in note 2). The nominative case is marked only in the 3rd person (plural), habun, in A function, but the morpheme -n is also attached to the 1st and 2nd persons, respectively i-n, mi-n. In short, all the persons in the singular and plural forms, except the 3rd person singular, are marked in the nominative case.

A -ø habu~ nasalisation S -ø habu -ø O -a haa -ø 1st and 2nd person 3rd person pronoun proper names and pronouns common nouns Table 8: Dixon’s sketch of the Cashinahua system (1994: 86)

R. Dixon’s analysis is pertinent when he says that “A and O markings overlap for some part of the middle of the hierarchy, rather than ergative marking stopping at the place where accusative begins” (Dixon 1994:87). However, the overlap occurs in the 3rd person singular which shows the same mopholology for S/A/O functions; thus, it is a neutral system and not a tripartite one. This same contradiction is shown in Payne’s analysis of the Cashinahua system. Payne (1997: 156) also analyses the Cashinahua 3rd person as “a tripartite system, in which each of the three primitives has distinct cases”. Like Dixon’s, Payne’s analysis refers to Cashinahua data (Table 6) obtained indirectly. Nevertheless these data do not correspond to modern Cashi- nahua14.

S A P 1, 2 pronoun -0 -0 -a nominative/accusative 3 pronoun -0 -~ -a tripartite Full NPs -0 - ~ -0 ergative/absolutive Table 9: Payne’s sketch of the Cashinahua system (1997: 156)

In Payne’s Cashinahua data, the 1st and 2nd persons as A are unmarked, but the 3rd person is marked by -n (noted as a nasal). According to our data, the only situation where ha-n can appear, is in the combination ha-n tsidi, as seen above. We can see

14 During the 20 months I was in the field the forms *ha-n as A and *ha-a as P were never attested. 166 CAMARGO this morpho-syntactic rule in the other Panoan languages as in Marubo (Costa 1997) where a ‘3sg’ receives -n → an in P function. In Capanahua, a double "haa" indicates the 3rd person (Loos 1999: 236). This discrepancy between grammatical relations and morphological case is marked only in the 3rd person, which I have indicated below as ø and ha. The latter appears more frequently in narratives than in discourse situations. Table 9 shows that the neutral system applies only to the 3rd person singular, so the morphology treats the three functions homogeneously. The complete data showing the three basic syn- tactic-semantic roles of the split system is proposed below.

1st and 2nd 3rd person 3rd person proper names person pronouns pronoun pronoun and common Singular Plural nouns A -n ø, ha hatu-n, habu-n -{-i, -a}n S -n ø, ha hatu-n, habu-n -ø P -a (SG) ø, ha hatu-ø, habu-ø -ø -ø (PL) Table 10: Camargo’s proposal for diagram of the Cashinahua split ergative system

9. Conclusion In this study the Cashinahua personal pronoun system was discussed, which contains seven forms, and its morphology varies depending on the case markers. I have tried to show that this language presents a split ergative system where the pronouns are characterized by the nominative-accusative system. The nominative case is marked by the -{a/i}n suffix. However, contrary to what has been reported in the literature (Dixon, Payne), the 3rd person singular form does not form a tripartite system, but rather a neutral one, with no morphological indication of the three semantic roles (A, P, S). They have the same form. However, example (34) shows ha- suffixed by -n in the combination with tsidi ‘also’ → han tsidi ‘he/she too’ or ‘so does he/she’. The accusative case marking appears with the suffix -di ‘imminent’, refering to the bene- factive: ha-a-di ‘(for) him too’. Therefore, it is possible that in ancient Cashinahua, the 3rd person singular in S/A semantic roles also received the nominative case marker suffix: ha-n (S/A) or a semantic role distinction, as data from the Capanahua language suggest. In this lan- guage, the 3rd person receives the same morphological treatment as proper and com- mon nouns: haa (S) and haa-n (A). It is possible that, in Cashinahua, ha is a relic from an earlier stage of the lan- guage when it had a nominative-accusative system for the 3rd person. Nowadays, the 3rd person singular is not marked as an argument in a discourse situation. The form ha appears indeed as anaphoric pronoun. And in this case it is unmarked in its se- mantic role of S, A or P. So, it overlaps the nominal hierarchy where, morphological- ly, the pronouns are characterized by the nominative-accusative case system and the CASHINAHUA PRONOUNS 167 nouns are marked by the ergative-absolutive case system. The 3rd person singular pronoun is outside this system. Another issue that was raised in this paper is the construction with the 1st and 2nd person singular with an embedded genitive: i-n mi-n nabu (1SG-GEN / 2SG-GEN / family), literally ‘your family is my family’. The context-dependent occurrence of this construction in the examples in (29), for instance, reflects sociocultural factors.

References Camargo, Eliane 2002 ‘Ergatividade cindida em caxinauá’, in : Línguas indígenas brasileiras: Fono- logia, gramática e história. Atas do I Encontro internacional do Grupo de Trabalho sobre Línguas Indígenas da ANPOLL, Belém : Gráfica Universitá- ria. 1996a ‘Valeurs médiatives en caxinaua’, in: Zl. Guentchéva (ed.) L’énonciation mé- diatisée, Paris: Éditions Peeters, pp. 271-284. 1996b ‘Des marqueurs modaux en caxinaua’, Amerindia, 21:1-20. Comrie, Bernard 1989 Language Universals and . Syntax and Morphology, Ox- ford : The University of Chicago Press. Costa, Raquel 1997 ‘Aspects of Ergativity in Marubo (Panoa)’, in: The Journal of Amazonian Languages, 1/2: 50-103. Deshayes, Patrick & Keifenheim, Barbara 1994 ‘Penser l’autre chez les Indiens de l’Amazonie’, Recherches et Documents, Amériques Latines, Paris: L’Harmattan. Dixon, Robert, W. 1994 Ergativity, [Cambridge Studies in Linguistics, Vol. 69], New York: Cam- bridge University Press. Erikson, Philippe 1996 La griffe des aïeux. Marquage du corps et démarquages ethniques chez les d’Amazonie, Paris: Éditions Peeters. Givón, T. 1984 Syntax. A Functional-Typological Introduction. Vol. I, Amsterdam, Philadel- phia: John Benjamins Publishing Company. Kensinger, Kenneth 1994 The Way Real People Ought to Live: Essays on the Peruvian Cashinahua, Prospect Heights, Illinois: Waveland Press. Lagrou, Elsje Maria 1998 Caminhos, Duplos e Corpos. Uma abordagem perspectivista da identidade e alteridade entre Kaxinawa, Ph.D thesis, FFLCH/USP. 168 CAMARGO

Loos, Eugene 1999 ‘Pano’, in: R.M.W. Dixon & A.Y. Aikhenvald (eds.) The Amazonian Lan- guages, [Cambridge Language Surveys], pp. 227-249. Payne, Thomas, E. 1997 Describing morphosyntax. A Guide for field linguistis, Cambridge: Cam- bridge University Press. Ricardo, Carlos Alberto (ed.) 2001 Povos Indígenas no Brasil, 1995-2000, São Paulo: ISA (Instituto Socioam- biental). Valenzuela, Pilar 2000 ‘Ergatividad escindida en Wariapano, Yaminawa y Shipibo-Konibo’, in: H. van der Voort en S. van de Kerke (eds.) Indigenous Languages of Lowland South America, [ILLA/1], Leiden: CNWS, pp. 111-128.

Abbreviations 1 first person LOC locative 2 second person MED mediative/evidential 3 third person MOD modal A agent MOT motion ABL ablative NEG negative ALL allative O object (of a transitive ANAPH anaphoric construction) ASP aspect OR valency increaser ASS assertive P patient BEN beneficiary PAST past (aspect) COM comitative PAST.EVID past evidential COMPL completive (aspect) PL (HE) heterogeneous plural DAT dative PL (HO) homogeneous plural DO direct object PL plural DU dual PREF preferential EVID evidential PRIV privative FAC factitive PROG progressive FOC focalisation PROSP prospective GEN genitive REC reciprocal HAB habitual (aspect) S subject HIST PAST historical past SG singular IMM imminent SOC sociative IMPER imper STAT state INDF PAST indefinite past TOP topicalisation INF infinitive VOC vocative INSTR instrumental V verb INTER interrogative ALGUNOS RASGOS TIPOLÓGICOS DEL KAMSÁ (VALLE DE SIBUNDOY, ALTO PUTUMAYO, SUDOESTE DE COLOMBIA) VISTOS DESDE UNA PERSPECTIVA AREAL∗

Alain Fabre Universidad Tecnológica de Tampere, Finlandia

1. Introducción El objetivo de este trabajo es presentar algunos rasgos tipológicos del kamsá, una lengua poco documentada, hablada en el valle de Sibundoy, Putumayo, sudoeste de Colombia, en una zona clave para las relaciones entre los Andes sur-colombianos y la selva amazónica. Empieza por un breve recuento de los estudios lingüísticos existentes sobre esta lengua, y por un resumen tipológico del kamsá. Luego se describen, con más detalles, dos fenómenos: incorporación nominal y clasificación nominal, vistos desde una perspectiva lingüística areal. El estudio considera, además del kamsá, lengua de partida para la comparación, un conjunto de once lenguas, distribuidas en tres zonas geográficas. Si bien el kamsá pertenece geográficamente a la zona andina, el área que ocupa cobra mucha importancia como punto de contacto entre dicha región y la selva amazónica. Como se echará de ver a lo largo de este trabajo, la lengua kamsá comparte rasgos tipológicos de ambas zonas. Las tres áreas y las lenguas consideradas son: (1) lenguas amazónicas: uitoto [familia -bora], waorani [lengua aislada], cofán [lengua aislada], andoke [lengua aislada], lenguas de la familia tucano occidental y oriental y lenguas de la familia záparo; (2) lenguas andinas: guambiano [lengua aislada], nasa yuwe o páez [lengua aislada]; (3) lenguas de la vertiente del Pacífico: emberá [familia Chocó] y awa pit o kwaiker [familia barbacoa]. El ingano (variedad del quechua) es una lengua con poca profundidad temporal en esta zona, razón por la cual se menciona pocas veces en el texto.

2. La lengua kamsá El kamsá (kamn.a) es una lengua aislada, hablada por unas cuatro mil personas en el valle de Sibundoy, situado en el extremo oeste del departamento de Putumayo, Colombia, hacia el este de la ciudad de Pasto. En el valle de Sibundoy viven, además de los kamsá, quechuahablantes (inganos) y mestizos castellanohablantes. Los kamsá han tenido contactos con los españoles y sus descendientes desde el año 1535, si bien sus contactos con la sociedad nacional han venido intensificándose recién desde

∗ Quisiera dar constancia de mi profundo agradecimiento a Sérgio Meira y Hein van der Voort para sus valiosos comentarios y sugerencias a la versión original de mi trabajo. Circunstancias imponderables hicieron que las cuidadosas notas editoriales de Sérgio y Hein, hechas varios meses atrás, no me alcanzaron hasta la semana prevista para dejar la versión definitiva del texto a la imprenta. Sin embargo, he tratado de aprovechar en la medida posible sus atinadas observaciones. Huelga decir que soy único responsable para cualquier error o inconsecuencia que aparezca en el texto. 170 FABRE principios del siglo XX (Bonilla 1972). Esta situación se ve reflejada en el fuerte influjo que ha tenido la lengua española sobre el kamsá. También la lengua quechua, llamada regionalmente ingano o inga, ha dejado su huella en el kamsá. La lengua kamsá, como las demás lenguas del sudoeste andino de Colombia, con excepción del ingano, está escasamente documentada. A continuación, hago un breve recuento cronológico de las fuentes lingüísticas modernas sobre el kamsá. El Instituto Lingüístico de Verano está presente entre los kamsá desde el año 1964.1 Los cuatro misionarios-lingüistas que se han dedicado a esta tarea se han interesado mucho más en el aspecto misional que en la descripción lingüística, lo que se desprende de lo poco que han publicado sobre la lengua. El número de títulos lingüísticos publicados por ellos en el lapso de 36 años suman cinco2, o en realidad tres, pues dos de ellos existen en dos versiones, inglesa y española3. Todos los demás títulos son ya elucubraciones misionales y/o cartillas de alfabetización.4 La compi- lación de vocabulario kamsá (Clough & al. 1992) se debe también al equipo de mi- sionarios del ILV. Cabe destacar que contiene varios errores que delatan un conoci- miento deficiente de la lengua, razón por la cual debe emplearse con mucha cautela.5 El trabajo de Howard sobre los tipos de párrafos, existe en dos versiones, cas- tellana (1977) e inglesa, esta última en el segundo volumen de un libro editado por Robert E. Longacre (1979). La primera es más amplia, la segunda acortada. Por lo que atañe a las partes comunes a las dos versiones, las mismas difieren solamente en puntos de detalle.6 Este estudio de los tipos de párrafos incluye algunas observacio- nes sueltas sobre la morfosintaxis del kamsá, sumergidas en una sarta de disparates y fórmulas tagmémicas, que casi nada revelan acerca de la estructura lingüística del kamsá. Además, los propios ejemplos del texto publicado en que se basa el artículo desmienten a menudo lo afirmado en el cuerpo del texto. El único lingüista pro- fesional del Instituto Lingüístico de Verano que ha trabajado en la región, Stephen H. Levinsohn, se ha dedicado exclusivamente al ingano, variedad del quechua. Fuera del Instituto Lingüístico de Verano, el investigador colombiano Monguí Sánchez publicó, en 1981, un libro titulado La lengua kame.ntza. Fonética -

1 Estas informaciones pueden ser consultadas en la página web del ILV/ Colombia: http://www.sil.org/americas/colombia/espn_index.htm 2 Howard (1967, 1972, 1977, 1979; Clough & al. 1992). 3 Se trata de los artículos de Howard (1967 [versión inglesa] y 1972 [versión española], que versan sobre la fonología del kamsá (ambos con 16 páginas) y Howard (1977 [versión española, más larga, con 66 páginas] y 1979 [versión inglesa acortada de 23 páginas]). Existe, además, una reseña sin publicar del libro de Monguí (Van Zyl & Solarte R. 1985). 4 Obsérvese, además, que los últimos títulos propiamente lingüísticos datan de 1977 (desconozco la fecha de compilación del vocabulario de Clough & al. (1992), mientras las fechas de publicación de los títulos restantes se extienden hasta el año 1999. La misma situación se da para las demás lenguas de la zona, con la salvedad del inga(no), en cuyo caso la presencia de un lingüista profesional (Levinsohn), influyó en el número relativamente elevado de estudios lingüísticos sobre esta lengua. 5 Cierta premura de tiempo para la publicación del vocabulario comparativo puede haber ocasionado tales errores e inexactitudes. Muy a menudo, resulta difícil extraer las raíces léxicas, especialmente cuando aparecen ligadas a morfemas irrelevantes. En otras ocasiones, el lexema resulta ser error de interpretación. 6 El texto español de las páginas 7-23 corresponde a las páginas 276-291 del texto inglés, mientras el texto español de las páginas 57/60-67 corresponde a las páginas 291-296 de la versión inglesa. RASGOS TIPOLÓGICOS KAMSÁ 171 fonología - textos (Bogotá, Publicaciones del Instituto Caro y Cuervo). En esta obra, tras una escueta, y en trechos poco clara, presentación de la fonética y fonología (pp. 3-25), el autor nos ofrece una compilación de textos con doble traducción, literal y contextual (pp. 29-222). Este trabajo de Monguí y el más reciente de McDowell (1994), representan el mayor caudal de textos kamsá que se haya publicado, por lo menos en forma accesible.7 Los textos kamsá de Monguí aparecen sin segmentación alguna, lo que dificulta el análisis lingüístico de los mismos. Es también importante destacar la labor del Centro Colombiano de Estudios de Lenguas Aborígenes, de la Universidad de los Andes, en Bogotá, que viene publicando estudios lingüísticos de alta calidad, también sobre el kamsá (Jamioy Muchavisoy 1989, 1992; Juajiboy Mutumbajoy 1995). Lamentablemente, por lo que a esta última lengua se refiere, no he podido, hasta la fecha, encontrar los materiales publicados por esta institución, razón por la cual no estoy en condición de comentarlos. De aquí en adelante, todos los ejemplos provienen de los materiales citados en la bibliografía. El análisis lingüístico y la segmentación, en cambio, son de mi entera responsabilidad. La transcripción se da según las normas de la Asociación Fonética Internacional (IPA) con las siguientes modificaciones: (a) la oclusiva palatal sorda (ch del español) se escribe /t/, para mantener el paralelismo con la palatal retrofleja /t./, mientras /c/ representa la africada alveolar del alemán (ts); (b) el acento, cuando su presencia ha podido colegirse en los textos, ha sido colocado directamente sobre la vocal acentuada. Para concluir este apartado, puede afirmarse que el kamsá sigue perteneciendo a aquella mancha de sombra del sur andino colombiano, bajo la cual se cobijan otras lenguas poco estudiadas como cofán, awa pit, nasa yuwe y guambiano que suelen permanecer fuera del alcance de la comunidad lingüística internacional.

3. Resumen tipológico del kamsá El kamsá es una lengua medianamente polisintética, con prefijos y sufijos gramaticales. Los prefijos gramaticales no son frecuentes en las lenguas del área considerada: además del kamsá, los hay solamente en el andoke y las lenguas záparo.8

3.1. Marcadores de personas El sintagma verbal puede llevar prefijos que corresponden al agente de primera persona y paciente de segunda (1a) y (1b), así como agente de segunda persona y paciente de primera (2). En (1b), llama la atención que el segundo actante remite al “objeto indirecto” (segunda persona de singular) y no al directo:9

7 Pude consultar los textos editados por McDowell recién después de la elaboración de esta ponencia. 8 El nasa yuwe (páez) tiene algunos prefijos empleados en la derivación lexical. No obstante, las relaciones gramaticales de esta lengua se concretan a través de sufijos. 9 Parte de las abreviaturas empleadas en las glosas de los verbos se basan en el cuadro (sumamente) parcial de los afijos del verbo (Howard 1977: 7). Estas interpretaciones deben ser tomadas con una fuerte dosis de cautela (v. comentarios en las notas a pie de página correspondientes a los ejemplos). Las demás 172 FABRE

(1) a ku-bu-ta-n-x-ua-nác 10 11 12 13 2SG-1SG -FUT-EV.VISTO.CON.PART-PRED -PRED -V -llevar ‘yo te llevaré’ (Monguí 1981: 89)

abreviaturas, así como el desglose correspondiente, cuando no expresamente indicado, son de mi responsabilidad y no son más que un intento de interpretación de los datos: ACT = actor; ASP = aspecto; AUM = aumentativo; BEN = benefactivo; CAR = caritativo; CL = clasificador; DIM = diminutivo; DEIC = deíctico; DIS = marcador de discurso; DU = dual; DUR = durativo; EVID = evidencial; EVID.VISTO.CON.PART = presenciado con participación del hablante; EVID.VISTO.SIN.PART = presenciado sin participación del hablante; FIN = finalidad/ meta; FUT = futuro; HAB = habitual; INF = infinitivo; ITER = iterativo; LOC = locativo (también progresivo con raíces verbales); PAS = pasado; PL = plural; PRED = predicativizador; RECIP = recíproco; SEPAR = separativo; SING = singular; SOC/INSTR = sociativo/instrumental; V = vocal temática. Algunos ejemplos, en las notas, van en la grafía original de Howard o Monguí, en cuyo caso se dan entre paréntesis angulares <...>. En el cuerpo del texto, la transcriptión (entre fonética y fonológica) es la de la A.F.I. El inventario fonémico del kamsá se presenta como sigue: (a) consonantes: /p, t, c (= ts), t., t, k, b, d, g, f, s, ., , x, m, n, , l, , r, w, j/; (b) vocales: /I, e, a, , u, o/. Los solapamientos de alófonos, sobre todo en las vocales, es un asunto que merecería un estudio detallado. Así, por ejemplo, a pesar de la existencia de un buen número de pares mínimos que permiten postular indudablemente /u, o, / como fonemas distintos, los textos publicados exhiben solapamientos de sus alófonos respectivos. En todos estos casos, he mantenido la vocal original de la fuente empleada. 10 En cuanto al morfema -bu, que podría ser llamado ‘dual’, he preferido glosarlo de acuerdo con la función que tiene en cada caso particular. Queda todavía por comprobar el valor ‘general’ de este morfema (v. también nota s 14 y 19). El morfema dual tiene los alomorfos siguientes: [ -bu, -bo, -b]. Los solapamientos alofónicos impiden a estas alturas del análisis decidir cual de estas formas debería ser escogida como forma básica. Por esta razón, he preferido mantener las vocales originales tales como aparecen en los textos publicados. Para la claridad de exposición, en vez de la glosa 2SG-DU para ku-bu- y DU-3SG para bu-(Ø), prefiero escribir 2SG-1SG y 3SG(-3SG). Las combinaciones de morfemas permiten generalmente interpretar las personas aludidas. Así /bu-/ es agente de primera persona de singular en ku- bu- ‘yo te…’, pero agente de tercera persona de dual en bu-(Ø) ‘ellos-dos…’. Se puede, tentativamente, también suponer cierta influencia de una jerarquía de personas en la interpretación final. 11 En cuanto a los morfemas verbales (al parecer evidenciales), que Howard (1977: 7) subsume bajo el rubro “participación”, la autora distingue dos: -n- (“presenciado y con participación del hablante”) y -k- (“presenciado, sin participación”); el morfema cero, en esta posición, indica que la acción no fue presenciada por el hablante. Los textos que manejo, parecen confirmar este análisis de Howard. 12 El morfema x- indica, según Howard (1977:8), ‘movimiento’. Ella lo explica como sigue: “El movimiento que sigue la línea del suceder del tiempo se marca con el morfema [grafía original] en tagmemas de desarrollo gradual, DG. Sin embargo, el morfema frecuentemente se obscurece por razón de los cambios morfofonémicos. También se da en los verbos genérico-históricos, y en el infinitivo”. Consi- dero aquí que hay dos morfemas diferentes, que aparecen en posiciones diferentes. Como primer prefijo a la izquierda del verbo, puede considerarse como marca de infinitivo (INF). En medio del verbo, parece señalar el carácter predicativo de la palabra, y lo llamo, en espera de un estudio más detallado, simplemente ‘predicativizador’ (PRED), aunque el empleo de este término tampoco me parece totalmente adecuado. 13 Hablando de las raíces verbales del kamsá, Howard (1977: 8) afirma que las mismas constan de una vocal de enfoque que indica que el sujeto es o agente o paciente. Como no aduce ejemplos del proceso, la existencia de una vocal de enfoque queda por comprobar. Sin embargo, encuentro en los textos de Monguí dos ejemplos que, por lo menos en estos casos, podrían confirmar la hipótesis de Howard. Se trata de los casos siguientes, casi pares mínimos, con la vocal de enfoque subrayada: i-o-x-o-f.éna (DIS-3SG-PRED-V- despertar) ‘él se despertó’, frente a i-o-x-a-f.éna-i (DIS-3SG-PRED-V-despertar-ACT) ‘él los despertó’ (Monguí 1977: 45 y 46). En la duda, la “vocal de enfoque” va glosada por mí con un signo de interrogación. El problema que dificulta el análisis de esta vocal, estriba en los solapamientos de alófonos, inclusive su desaparición total. RASGOS TIPOLÓGICOS KAMSÁ 173

b boko-je-tem k-b-ta-bo-prontá 14 chicha-CL-DIM 2 SG-1SG-FUT-BEN -alistar ‘la chichita te alistaré’ (Monguí 1981: 122)

(2) .-ko-n-x-on.á salbá 1SG-2SG-EV.VISTO.CON.PART-PRED-empezar salvar ‘tú me empezaste a salvar’ (Monguí 1981: 89)

Con un paciente de tercera persona, sin embargo, suele aparecer un solo morfema (3a), (3b), (3c), (3d) y (3e) . Como resalta de (3d) y (3e), este único morfema refiere al agente más que al paciente:

(3) a bu-x-ua-nác 3SG-PRED-V-llevar ‘él lo llevó’ (Monguí 1981: 89)

b t-mo-x-an-a-nác 15 16 DIS -3PL-PRED-PAS -V-llevar ‘ellos lo llevaron’ (Monguí 1981: 184)

c tarr--nga i maaxta ventador-a-ngá tarro-CL-PL y muchos aventador-CL-PL

i xasa-be-nga i-o-x-a-kxaje xocn--en.e 17 y calabazo-CL-PL DIS-3SG-PRED-V-poner cama-CL-LOC ‘puso en la cama tarros y muchos aventadores y calabazos’ (Howard 1977: 28)

14 McDowell (1990: 300), cuyo estudio no es propiamente dicho lingüístico, emplea el término “benéfico” para referirse al morfema -bo-, lo mismo que resulta ser idéntico a lo que, según la tabla de Howard (1977: 7) sería tercera persona de dual. Sea como fuere, en (1b), los dos morfemas se distinguen por su posición (es irrelevante el hecho de que la tercera persona de dual aparezca aquí como b-: como puede observarse con otros ejemplos, varía libremente con bo- (y bu-). 15 Existen, según Howard, dos “marcadores de discurso”: t-, que “se usa en las narraciones de hechos históricos y de hechos contemporáneos con raíz puntiliar (es decir, el hecho se ve como concluido en un momento dado)” e i- que “se da en verbos de narración de leyenda. También se puede dar en narraciones históricas y contemporáneas de acción previa no presenciada”. 16 Howard (1977: 8), dice que hay dos morfemas de pasado: -ec- “que se refiere a la acción (ya sea a la progresiva, o a la que se realizó en un momento dado) que ocurrió en el pasado reciente en verbos de acontecimiento contemporáneo o de leyenda” y -an- “pasado histórico” que es “obligatorio en verbos de acontecimiento en narraciones históricas y es optativo en verbos de acontecimiento contemporáneo o de leyenda. Este morfema coloca la acción más en el pasado”. Cabe observar, acerca de estas y otras afirmaciones gramaticales hechas por Howard, que en primer lugar, Howard nunca define los términos que emplea ni discute ejemplos en que se pueda comprobar la diferencia de significado y/o función entre los morfemas citados. 17 La glosa LOC subsume varios morfemas locativos: -en.e, -ok, -oi, -e (este último, al final de un verbo, es también marca del progresivo), y -ka (este último morfema aparece muy a menudo al final de toda clase de palabras. Muchavisoy (1997) lo traduce por ‘con’, por lo cual podría tener un valor de sociativo/ instrumental. Al final de verbos, podría simplemente tener un valor coordinativo ‘y’). 174 FABRE

d s-c-e-naná 1SG-ASP-V-golpear ‘yo lo golpeaba’ (Howard 1977: 45)

e lof-.e a.e s-njb-e-ate 18 pájaro-CL yo 1SG-ASP.HAB -V-coger ‘yo siempre cojo un pájaro’ (Howard 1977: 60)

3.2. Dual Debe notarse el uso del morfema bu-, que Howard (1977: 7) considera como tercera persona de dual, en casos involucrando a dos personas, cualesquiera que sean.19 Así el dual de los ejemplos (1a) y (1b) arriba se refiere a la primera del singular (agente), mientras en (3a) a la tercera del singular (agente también). En los ejemplos (4a) a (4e), el dual no se explica por el agente, que es de tercera del singular, sino por la acción en conjunto, que involucra a dos entidades animadas:

(4) a i-bo-x-c-a-tmonxa DIS-3DU-PRED-ASP-V-cortar ‘él le cortó (a otra persona)’

b t-bo-x-an-o-n.é x-em-pelia-na 20 21 DIS-3DU-PRED-PAS-V-empezar INF-V -pelear-DUR ‘él empezó a pelear (con otro)’ (Howard 1977: 18)

c t-bo-x-an-e-nataxt-ka t-tigre i DIS-3DU-PRED-PAS-V-voltear-SOC/INSTR DEIC-tigre y

t-bo-x-es-an-obá DIS-3DU-PRED-V-PAS-matar ‘él volteó al tigre y lo mató’ (Howard 1977: 18)

d bu-x-ua-nác 3DU-PRED-V-llevar ‘él se lo llevó a él’

18 Howard (1977: 7-8) llama “Aspecto 2" a los dos “marcadores de aspecto”: -nd- (“genérico contemporáneo”) y -njb-/-nd- (“habitual”). La misma autora añade que estos marcadores de aspecto “no manifiestan DG [desarrollo gradual]”. 19 El dual ocurre también con agente de tercera del singular y paciente en tercera del dual. En un trabajo en elaboración, analizo con más detenimiento las modalidades de uso de este morfema “dual” (Fabre, en preparación). 20 Talvez xe-n-: INF-EV.VISTO.CON.PART-, donde el primer morfema aparecería con una vocal de apoyo. 21 Muchavisoy (1997) trata el -n(a) final del verbo como un durativo. En muchos casos, sin embargo, parece tener un valor indeterminado que prefiero glosar como CON ‘conector’. RASGOS TIPOLÓGICOS KAMSÁ 175

e t-ora kanje tigre-fta-ka t-bo-x-an-e-nbe.e 22 DEIC-hora un tigre-CL-SOC/INSTR DISC-DU-PRED-PAS-V-encontrar ‘se encontró con un tigre’ (Howard 1977: 35)

La existencia del morfema bu- como tercera persona del dual queda, por otra parte, confirmada en el corpus. Ejemplos típicos son (5a) y (5b):

(5) a b-x-a-ta-t-ka 23 3DU-PRED-V-huir-DU-SOC/INSTR ‘huyeron los dos’ (Monguí 1981: 143)

b t-or bo-x-a-deskancá-t-ka DEIC-hora 3DU-PRED-V-descansar-DU-SOC/INSTR ‘luego descansaron los dos’ (Monguí 1981: 141)

La presencia de un dual, tanto en el predicado como en los sintagmas nominales, es poco frecuente en las lenguas del área considerada.24 En el área considerada en esta ponencia, la categoría de dual ha sido registrada, además del kamsá, en el awa pit, witoto y waorani. (6a) y (6b), son ejemplos del dual en sintagmas nominales:

(6) a bnda-t-be kontrát 1DU-DU-GEN contrato ‘nuestro contrato (de los dos)’ (Monguí 1981: 97)

b t-o-x-at-xaje t-uta señorita-ta DIS-3SG-PRED-ITER-preguntar DEIC-dos señorita-DU ‘él les preguntó a las dos señoritas’ (Howard 1977: 36)

3.3. Ausencia de marcación de casos Agente y Paciente en los nominales Llama la atención la coexistencia de los sintagmas nominales y los afijos verbales correspondientes, aunque en las lenguas amazónicas polisintéticas, la presencia de un afijo verbal de actante supone generalmente la ausencia del sintagma nominal correspondiente en el enunciado25. En este sentido, el kamsá se acerca más al esquema andino. Por otra parte, los sintagmas nominales agentes y pacientes del

22 Casi idéntico a la forma empleada por los quechuahablantes del valle de Sibundoy: chi-oras ‘a esa hora; entonces’. Ambas lenguas, por otra parte de estructuras muy diferentes y genéticamente no emparentadas, han tomado prestado la segunda parte de esta palabra del castellano (hora), precedida del deíctico quechua chi, que bien podría ser, en kamsá, préstamo del último. 23 El sufijo final –ka, llamado aquí SOC/INSTR puede tener también un valor de lazo entre oraciones. A veces, al modo del syfijo quechua –wan, incluso puede servir de conjunción (tipo ‘el oso y el tigre’). 24 Wise (1999: 318) señala que el dual es una categoría poco frecuente en las lenguas amazónicas en general. 25 De ahí, en muchas de estas lenguas, una diferencia muy significativa entre el número de verbos y de sintagmas nominales. 176 FABRE

kamsá no llevan ningún caso morfológico, como puede apreciarse en los ejemplos (7a), (7b) y (7c)26:

(7) a t-bo-x-an-e-natax -ka t-tigre i DIS-3DU-PRED-PAS-V-voltear-SOC/INSTR DEIC-tigre y

t-bo-x-es-an-obá DIS-3DU-PRED-PAS-matar ‘él volteó al tigre y lo mató’ (Howard 1977: 18)

b kánje bobóns bu-x-at-o-bianxe.í un jóven 3DU-PRED-ITER-V-atisbar ‘él atisbó a un joven’ (Monguí 1981: 97)

c i-bo-x-o-jije av-fxa DIS-3DU-PRED-V-entregar llave-CL ‘él le entregó la llave’ (Howard 1977: 89)

3.4. Orden de las palabras Acerca del orden de las palabras en el enunciado, se puede notar lo siguiente:

(1) el orden del objeto directo y del verbo está sujeto a variación: del desglose de un centenar de ejemplos, consta un leve predominio del orden V+O (55.6% del total) frente a O+V (44.3%):

(8) a kánje nená bo-x-a-twái-ka una señora 3DU-PRED-V-saludar-SOC/INSTR ‘a una señora (blanca) ellos dos saludaron’ (Monguí 1981: 36)

b t-mo-x-an-o-fxa inje-nga DIS-3PL-PRED-PAS-V-invitar otro-PL ‘ellos invitaron a otros’ (Howard 1977: 55)

(2) aunque el sujeto precede generalmente al verbo, no faltan ejemplos del orden contrario:

(9) a kanje salteador i-bo-x-ue-bwata.e-na un salteador DIS-3DU-PRED-V-salir-CON ‘un ladrón le salió (al encuentro)’ (Howard 1977: 73)

26 Lo mismo puede decirse del “objeto indirecto” (no locativo). RASGOS TIPOLÓGICOS KAMSÁ 177

b kánje sólid txán-.-én. mu-x-á-boka-n una solitaria montaña-CL-LOC 3PL-PRED-V-salir-DUR

únga .kena- .-ng tres blanco-CL-PL ‘del interior de una montaña solitaria salieron tres blancos’ (Monguí 1981: 32)

(3) los numerales siempre van antes del nombre (7b), (8b) y (9b);

(4) los adjetivos pueden aparecer tanto antes como después del nombre, con la particularidad notable que el adjetivo antepuesto no lleva clasificador, mientras el pospuesto sí lo lleva (véanse ejemplos más abajo).

El cuadro de la página siguiente presenta algunos rasgos tipológicos de las lenguas escogidas: (1) incorporación nominal; (2) presencia de clasificadores; (3) categorá de dual; (4) género masculino/ femenino); (5) prefijos gramaticales [mayormente actantes del verbo]. Algunas lenguas, como el nasa yuwe, hacen uso de algunos prefijos, pero solo con valor derivativo. Casos de este tipo no han sido tomados en cuenta en la elaboración del cuadro; (6) marca morfológica de paciente en el sintagma nominal; (7) orden preferencial “objeto”/verbo; (8) orden preferencial poseedor/nombre poseído; (9) orden preferencial adjetivo/nombre.

Notas al cuadro siguiente: (1) Los ejemplos en Aguirre (1999: 142) y Harms (1994: 86) muestran que la incorporación nominal existe en emberá, si bien su empleo parece restringirse a un número limitado de verbos transitivos. (2) Cook & Criswell (1993) no mencionan la posibilidad de incorporación nominal en coreguaje (tukano occidental), pero esta construcción queda atestada en otras lenguas de la rama oriental de esta familia lingüística. Gómez-Imbert (1988) y Strom (1992) dan ejemplos de incorporación nominal, para el y tatuyo (Gómez-Imbert) y el retuarã (Strom). (3) Obando (1992: 112-113) habla de cuatro clasificadores en awa pit, tres de los cuales figuraban ya en Jijón y Caamaño (1998 [1940], Tomo I: 160). Sin embargo, la segmentación arbitraria de palabras en “sílabas-morfemas” por parte de Jijón y Caamaño llevan a dudar de la existencia de clasificadores para esta lengua. Además, Curnow (1997) niega terminantemente la presencia de clasificadores en awá pit.27 (4) La presencia de una categoría de dual en awá pit está sujeta a variación dialectal: la variedad descrita por Curnow (1997) la desconoce mientras existe en los dialectos descritos por Obando (1992). (5) En nasa yuwe, los pronombres personales de primera y segunda persona singular, así como los sufijos verbales de segunda del singular, diferencian entre masculino y femenino (Rojas C. 1994, 1998; Rojas C., Nieves Oviedo & Yule Yatacué 1991; Slocum 1986). (6) En andoke, una diferencia de géneros entre masculino y femenino existe con referentes [+animado] (Landaburu 1979).

27 Agradezco a Curnow la gentileza de haberme enviado un ejemplar de su tesis doctoral. ZONAS GEOGRÁFICAS VERTIENTE DEL ANDES SELVA AMAZÓNICA (AL ESTE DE LOS ANDES PACÍFICO COLOMBIANOS/ ECUATORIANOS) Awá Emberá Nasa Guambiano Ingano Kamsá Cofán Coreguaje Uitoto Andoke Waorani Záparo pit yuwe / Tukano (páez) incorp. no sí (1) no no no sí no sí (2) no sí no no nom. clasif. no (3) sí no no no sí sí sí sí sí sí sí dual sí (4) no no no no sí no no sí no sí no género no no no/sí (5) no no no no sí sí sí/no (6) no/sí (7) no M/F prefijos no no no (8) no no sí no no no sí no sí marca de no no (9) sí sí (10) sí no sí no/sí (11) no no no no (12) paciente en el SN Orden O+V O+V O+V O+V O+V O+V O+V V+O (13) O+V O+V ? O+V O-V V+O V+O (14) V+O (15)

Orden POS+N POS+N POS+N N+POS POS+N POS+N POS+N POS+N POS+N POS+N POS+N POS+N POS-N (POS+N) (16) (17) Orden A+N N+A N+A A+N A+N A+N A+N N+A [A+N] [A+N] [A+N] A+N A-N N+A A+N (18) (19) (20) (21) N+A (22) RASGOS TIPOLÓGICOS KAMSÁ 179

(7) Los pronombres personales de tercera persona del singular, dual y plural, tienen, en waorani, una forma feminina honorífica (Peeke 1973; Pike & Saint 1988). (8) El nasa yuwe emplea algunos prefijos en la derivación lexical. Por lo demás, en las declinaciones y conjugaciones, esta lengua usa, exclusivamente, sufijos. (9) El emberá es una lengua ergativa. Ello significa que los pacientes de verbos transitivos aparecen con un caso cero (absolutivo). (10) Con la salvedad de que en guambiano, un objeto genérico no lleva sufijo de objeto directo (Long 1985: 15-16; Vásquez de Ruiz 1988, 1994). (11) El coreguaje marca solamente el paciente específico (Barnes 1997; Cook & Criswell 1993) (12) Wise (1999) señala que las lenguas záparo exhiben, en algunos casos, señales de ergatividad. Por otra parte, el paciente no lleva caso morfológico (V. también Eastman & Eastman 1963). (13) En coreguaje, el objeto va generalmente después del verbo. En otras lenguas tukano (por ejemplo en barasano y [tukano oriental], el objeto precede al verbo (Barnes 1999: 224). (14) La presencia de actantes nominales fuera de los afijos verbales no es necesaria. De aparecer los actantes primero y segundo este último se coloca más cercano al verbo. Landaburu (1979: 94) describe este fenómeno bajo el rubro de “proyección”. (15) El orden es O+V en arabela, pero en iquito y záparo, de la misma familia lingüística, el orden más usual es V+O (Wise 1999: 329). (16) Con un poseedor [-humano], el orden es en guambiano POS+N. Con un poseedor [+humano] el orden es ya N+POS (más usual), ya POS+N, cuando se quiere dar relieve al poseedor (Long 1985: 16). (17) La posesión nominal se marca, en andoke, mediante prefijos personales. A la izquierda de estos prefijos, puede aparecer el poseedor: pód ja-tí “luna su[clase 3.1] madre” (Landaburu 1979). De todos modos, el poseedor precede al nombre. (18) Los “adjetivos” del coreguaje son, en su mayoría, formas verbales. Cuando éstas aparecen como determinantes de nombres, Cook & Criswell (1993: 34) indican que suelen colocarse después del nombre, salvo cuando se desea realzar el efecto. (19) En lugar de adjetivos, el uitoto hace uso de verbos o nombres. Petersen señala sin embargo una muy reducida clase de adjetivos con estructura nominal, a la cual pertenecen cinco adjetivos calificativos. Éstos preceden al nombre (Petersen 1994: 71-76). (20) La categoría de adjetivo no existe como tal en andoke. Los ejemplos aducidos por Landaburu (1979: 176-7) muestran que la “determinación periférica” se sitúa a la izquierda de la cabeza del sintagma nominal. Ello podría justificar la inclusión de A+N para esta lengua en el cuadro. (21) Existen pocos ejemplos de palabras que equivalen a “adjetivos” en los textos waorani. Aparentemente, el orden más usual es A+N. Es imposible incorporar otros datos al respecto, pues los “adjetivos” se traducen por verbos o clasificadores y forman parte de una palabra compuesta. (22) Ambos ordenes existen en arabela (Wise 1999: 329-330).

4. Incorporación nominal en kamsá Entre las lenguas del sudoeste colombiano, el kamsá se distingue por el empleo de incorporación nominal, fenómeno registrado también en las lenguas tukano, andoke (Landaburu 1979: 183) y las lenguas de la familia chocó (Aguirre 1998: 142; Harms 180 FABRE

1994: 86).28 Resulta a menudo difícil distinguir entre incorporación nominal propiamente dicha e incorporación de clasificadores en el verbo, entre los cuales algunos pueden, en casos determinados, referir a entidades bastante concretas, tal como ocurre, por ejemplo, en waorani (v. textos publicados en Pike & Saint, eds. 1988). En su trabajo fundamental sobre la incorporación nominal, Mithun (1984) distingue cuatro tipos de incorporación nominal. Es importante señalar que cada una de estas estructuras puede existir en una misma lengua. Las estructuras de tipo I combinan un nombre con un verbo, ya mediante simple yuxtaposición, o mediante la formación de una palabra morfológica compuesta. El resultado típico es un nuevo predicado intransitivo, y el nombre incorporado queda despojado de índices particularizadores tales como deícticos, partículas o artículos definidos, número y caso. Desde un punto de vista semántico, el nombre puede ser paciente, locativo, o instrumental. La actividad o el estado involucrado se ven como un concepto unitario y poco saliente. El tipo II supone, además de lo mencionado para el tipo I, cierta “manipulación” de los casos, en la que la algún argumento oblicuo de la oración se ve promovido a la posición, ahora libre, del nombre incorporado. En el tipo III, a estos fenómenos se añade otro más, la “manipulación de la estructura del discurso”, a través de la cual la incorporación en el verbo de un argumento hace retroceder este último al fondo del discurso. Según Mithun, esta fase es típica de lenguas polisintéticas que traen afijos verbales para sujeto y objeto, y en las cuales los sintagmas nominales no se usan mucho, siendo el predicado portador de casi toda la información. En cuanto al último tipo, el IV, Mithun lo llama “incorporación nominal clasificadora”. Su particularidad es el empleo, además del nominal incorporado en el verbo, de un sintagma nominal exterior que refiere de una manera más específica, al argumento. Ello ocurre ya que el nombre incorporado suele adquirir una referencia más amplia, menos específica, a la par que muchas veces se ve fonéticamente reducido.29 Así, por ejemplo, una vez incorporada la palabra ‘ojo’, ésta puede referir a cualquier objeto pequeño y esférico. Basado en el análisis de más de cien lenguas diferentes, el estudio de Mithun muestra que los cuatro tipos forman una jerarquía implicacional: I > II > III > IV. Para los casos semánticos de los nombres incorporados, la investigación de Mithun hace resaltar lo siguiente: (1) cuando una lengua acepta un único caso semántico para los nombres incorporados, se trata del paciente de transitivo; (2) cuando dos casos semánticos distintos pueden ser incorporados, al paciente de

28 Para el andoke, Landaburu habla de “bases compuestas”, entre las cuales resaltan combinaciones de bases verbales con nombres de partes del cuerpo y clasificadores. Para el emberá, Aguirre, por su parte, habla de “composición nombre + verbo”, mientras Harms menciona verbos con incorporación de objeto. La incorporación nominal no ha sido registrada para las lenguas siguientes: awa pit, guambiano, nasa yuwe, cofán, waorani, witoto, y záparo. 29 La reducción puede llegar a los extremos de formas supletivas, en cuyo caso resulta a menudo imposible el deslinde entre incorporación de nombres e incorporación de clasificadores. RASGOS TIPOLÓGICOS KAMSÁ 181 transitivo del caso anterior se suma el sujeto de intransitivo30; (3) con tres casos semánticos susceptibles de ser incorporados, el tercero representa un instrumento y/o un locativo. Se trata, pues, de una segunda jerarquía implicacional: paciente de transitivo > sujeto de intransitivo > instrumental/locativo. En los textos publicados por Howard, constan algunos ejemplos de incorporación nominal, aunque Howard no ofrece ningún comentario al respecto:

(10) n.ema ..ona-tema recién nacido nene-DIM

i-n-ec-bta-t-xaxo-e DISC-EVID.VISTO.CON.PART-PAS-camino-CL-colocar-LOC ‘un nene recién nacido había sido puesto en el camino’

En (10) y (11), y quizás también en (15), el nombre incorporado conserva su(s) clasificador(es), pero ello no es necesario, como lo prueban (12b), (13) y (14), que no llevan ningún clasificador sufijado al nombre incorporado en el verbo. El ejemplo (10) puede ser contrapuesto al siguiente (11), que presenta una estructura sin incorporación:

(11) bta-t-xa-e beká wa-fawan-nga i-o-x-cá camino-CL-CL(?)-LOC muchas ?-fritada-PL DIS-3SG-PRED-ir

wa-ana-je ?-regando-? ‘iba regando muchas fritadas en el camino’ (Howard 1977:29)

(12a) es una estructura con nombre suelto, no incorporado, mientras en (12b), el nombre ‘cenizas’ queda incorporado:

(12) a t-ora tana i-o-x-o-cbaná inj-ok-ana 31 32 DEIC-hora ése DIS-3SG-PRED-V-levantarse ceniza -LOC-SEPAR ‘entonces él se levantó de las cenizas’ (Howard 1977: 37)

b i-o-x-o- inja-cbaná DIS-3SG-PRED-V-ceniza-levantarse ‘se levantó de las cenizas’ (Howard 1977: 37)

30 Refiero a Lazard (1994: 24-45) para una discusión de los problemas referentes al reconocimiento de entidades como ‘paciente’ y ‘sujeto’. También Baker (1988: 83) hace hincapié en las dificultades inherentes al uso del término ‘paciente’. 31 Cp. las semejanzas entre cbanán-oka ‘arriba’ (con -[o]ka = LOC) y -ana = SEPAR 32 El vocabulario de Clough & al. (1992) ofrece la forma xatinja para ‘ceniza’, mientras traduce ínje por ‘sol’. Sea cual fuere la traducción de esta palabra, el análisis morfológico sigue en pie. 182 FABRE

(13) kanje .nená i-n-ec-b.-tbema-e uno/a señora DIS-EVID.VISTO.CON.PART-PAS-puerta-sentar-LOC ‘una señora estaba sentada en la puerta’ (puerta-CL = b.á-.á) (Howard 1977: 58)

(14) mo tá tá-we-n-o-tamo-.ena-ka-ka cómo él FUT-?- EVID.VISTO.CON.PART -V-cuello-botar-SOC/INSTR ‘cómo debía amarrársela (la cadena) al cuello’ (tamo- a ‘cuello-CL) (Monguí 1981: 101)

(15) x-t-o-buakue-na-já-m-na 33 INF-ITER-V-mano-? -atar-FIN-DUR ‘para amarrársela (la trenza) en la mano’ (buakuá-.e ‘mano’) (Monguí 1981: 103)

Puede ser mera casualidad que en todos los ejemplos, los nombres incorporados sean locativos.34 En este caso, y si las predicciones de Mithun son ciertas, debería ser posible la incorporación de pacientes de transitivos y de sujetos de intransitivos, casos no atestados en mi corpus. En el caso de que ello no fuera posible, habría que modificar la jerarquía de implicancias propuesta por Mithun para los casos semánticos.

5. Clasificación nominal en kamsá A pesar de que ninguna de las fuentes de que yo tenga conocimiento haga referencia explícita a la existencia de clasificadores en Kamsá, llama la atención que el vocabu- lario de Clough et al. (1992) parece reconocer siete de ellos, en cuanto aparece, pero solamente en el caso de algunos nombres, una segmentación de los morfemas -bé, -xa, -.e, -.a, -.e, -fxa y -fxwa. Lo único que apunta, en dicho vocabulario, hacia un reconocimiento, si bien implícito, de un concepto de clasificador, es la traducción del adjetivo ‘redondo’ por el morfema -bé. Además, en varias entradas del vocabulario, los clasificadores antedichos aparecen ligados a la raíz, sin ninguna segmentación. A continuación, notaré otras terminaciones recurrentes en nombres, las que sin lugar a dudas son clasificadores. La lista está todavía abierta ya que las modificaciones mor- fofonémicas, sobre todo en las vocales finales, no siempre dejan claramente entrever si se trata o no de clasificadores distintos. Por las dudas, he reunido los clasificadores

33 Quizás el morfema -na- sea el (probable) clasificador de forma idéntica, aunque ‘brazo/ mano’ aparece normalmente con otro clasificador, -. e. 34 Después de la redacción de este trabajo, encontré once casos más de incorporación en los textos publicados por McDowell (1994), de los cuales destacan siete casos de locativos. De los cuatro restantes, todos parecen ser pacientes de transitivos, aunque no puede haber entera seguridad al respecto, debido a la traducción inglesa, que en trechos se aparta bastante del texto original. Los textos de McDowell aparecen en formato bilinear, con el original kamsá y la traducción (muy libre) al inglés, sin el paso intermediario de una traducción literal, lo que dificulta el análisis lingüístico. RASGOS TIPOLÓGICOS KAMSÁ 183 fonéticamente parecidos.35 Como se echará de ver, los clasifica-dores pueden unirse tanto a radicales nativos como a préstamos del castellano y del quechua.

1) -bé ‘esférico; redondo’: armadio-bé ‘armadillo’ (

35 La variación fonética, tanto para los fonemas vocálicos como para los consonánticos, no ha sido debidamente estudiada. Para los primeros, Monguí (1981) propone una lista de seis fonemas, /i, e,  , a, u, o/, identificando además algunas variaciones. Sin embargo, los numerosos solapamientos de alófonos justificarían un análisis más detallado. Así, por ejemplo, Monguí (1981) transcribe en la página 186 /bunguán ko-ticém/ y en la página siguiente /bunguán ko-ticém/ ( y , respectivamente), dejando entrever una variación [] / [i] que no implica cambio acentual. De la misma manera, [a] fluctúa con [] i con [i], en una misma palabra, incluso en préstamos del castellano. Cabe observar, además, que [], cualquiera que sea su posición, puede aparecer con o sin acento. 184 FABRE

8) -(a) : ataaj-a ‘atarraya’ (

36 -na aparece también con verbos, especialmente — pero no exclusivamente — en la forma nominal llamada ‘infinitivo’ por Howard. Llama la atención el número elevado de nombres de insectos que terminan en -na. Por otra parte, si se considera este -na como un clasificador, sólo un pequeño número de nombres (y “adjetivos”) queda sin clasificador. Además, al igual que el sufijo –ka, -na tiene funciones diversas, poco claras, entre las cuales destaca la de relacional. RASGOS TIPOLÓGICOS KAMSÁ 185

na ‘araña’ (talvez raíz quechua; cp. pakta ‘araña’ en quechua de San Martín, Perú); bcana-ná ‘anciano’; beó-na ‘pez; pescado’; biánga-na ‘venado’; bobinj-ná ‘anciano’; ctxó-na ‘nigua’; ijendó-na ‘ardilla’; jéb-na ‘casa’; mcá-na ‘piojo’; mn.é-na ‘carne’ (? mn-.é-na); ngó-na ‘humo’; ngnciá-na ‘picaflor’; .e.ó-na ‘criatura’; acbiá-na ‘mosca’ (? ac-biá-na); ákua-na ‘hierba’; ié-na ‘ratón’; txowá-na ‘abeja’; ta-ná ‘ñame’; utab-ná ‘jefe’; wabái-na ‘nombre’; wafté-na ‘lluvia’; xuá-na ‘hormiga’; xuakó-na ‘luna’; ai-ná ‘vivo’; bóte-na ’maduro’ (? bó-te-na); buánga-na ‘rojo’; buaá-na ‘seco’; bngnja-na ‘liso’; c.já-na ‘amarillo’; ft-ék-ua-na ‘mojado’; ngf-na ‘verde’; obo-ná ‘gordo’; ek-ba-na ‘frío’; xanguá-na ‘podrido’; xútxe-na ‘lleno’;

Algunos nombres pueden emplearse bien con uno u otro clasificador, bien con dos o más, sin cambio aparente de sentido:

(16) betá-.e ≈ betá-fxa ‘huevo’ (17) betí-je ≈ beti-jé- e ‘árbol’ (18) xuensa-bé ≈ xuenc-.a-.e ‘frente’37 (19) was-kwa- . ≈ was-kwa-. e-xwa ‘cola’ (20) jéb( )na ≈ jéb()na‘casa’ (21) binjia ≈ binjia-je ‘viento’

La lista siguiente representa los casos de acumulación de clasificadores (mi análisis es tentativo):

1) -kwa(.)e: bua-kwá-.e ‘brazo’; ku-kwa-.e ‘mano’; mac-kwa-.e ‘oreja’ (cp. wa-mac-axonxan-a ‘arete’, en que aparece un elemento común — mac— ‘oreja’, el verbo ‘colgar’ y el clasificador -a); e-kwá-.e ‘pie’; 2) -kwa-.e-xwa: was-kwa-.e-xwa ‘cola’ (también: was-kwa-.); 3) (?) -kwa-na: a-kwa-na ‘hierba’; ku-kwa-na ‘pava de monte’; 4) (?) -kwa-a: ku-kwa-a ‘mano; pata’ (cp. ku-kwa-.e ‘mano’); 5) (?) -kwa-xa: (?) bexn-gua-xa ‘pescuezo, cuello’; 6) (?) -kwa-je: m.-kwá-je ‘serpiente’; 7) (?) -xwa-(n)bé: a-xwan-bé ‘fruta’; 8) (?) -().a-xa: tamo-.a-xa ‘cuello’ (cp. tamo-.ton-bé ‘nuez’); st-.a-xa ‘espalda’; tan-.a-xa ‘hombro’; 9) (?) -ta-xa: ko-ta-xa ‘senos, pecho’; 10) (?) -.á-fxa: xn-.á-fxa ‘dedo del pie’ (cp. n.a-buá-fxa ‘dedo de la mano’; 11) (?) -bia (± -.a/ -.e); [quizás también -miá (± [-a] : wafs-biá ‘vientre’; ngéce- bia-.e ‘uña’; wajace-biá-.a ‘labio’; xwatng-miá-a ‘ala’; 12) (?) -fxá-.e: .-fxá-.e ‘pantorrilla’; 13) (?) -fxá-xa: .-fxá-xa ‘espinilla’;

37 El cambio en el radical puede ser simple condicionamiento fonético: [nc] : [] y [a] : [ ]. 186 FABRE

14) (?) -bua-fxa: n.a-buá-fxa ‘dedo (de la mano)’ (cp. x-n.a-fxa ‘dedo del pie’; quizás también ta-n.a-fxa ‘hombro’, aunque el elemento común puede guardar relación con el nombre en.á ‘persona’); 15) (?) -je-e: beti-jé-e ‘árbol’; bu-je-e ‘agua’; 16) (?) -xa-je: fá-xa-je ‘río’; bé-xa-je ‘agua’ (también bú-je-e); 17) (?) -a-xwa: wa[-]ngc-bobo--xwa ‘barba’ (cp. wa[-]já-.a ‘boca’, quizás tambien wa[-]mt-.a ‘ombligo’); kwa-a-xwa ‘bebida’; 18) (?) -á-xa: bobo-á-xa ‘pelo del cuerpo’; 19) (?) -a-e: xat-a-e ‘pantano’;

Pueden completar esta lista todavía el diminutivo -téma, el caritativo -xéma, así como los dos aumentativos -jéma y -kuem(a), morfemas cuyo empleo recuerda, por lo menos en algunos casos, el uso de los clasificadores como índices pragmático- discursivos. Así, en (23), -xéma parece retomar el -.e de (22) (Howard 1977: 20):

(22) t-ora kanje sapo- .e tamba sibia-ka entonces una rana-CL sanja orilla-LOC

i-nd-c-o-tbema-e DIS-ASP-PAS-V-sentar-LOC/PROG ‘entonces un sapo estaba sentado a la orilla de la zanja’

(23) s-n-x-i- ate sapo-xema 1SG-EV.VISTO.CON.PART-PRED-V-coger sapo-CAR ‘yo cogí la pobre rana’38

Sin embargo, prefiero distinguir, por dos razones diferentes, estos tres sufijos de los clasificadores propiamente dichos: en primer lugar, son bisilábicos en su forma básica frente a los clasificadores monosilábicos; en segundo lugar, su empleo parece ser independiente del de los clasificadores, si bien ambos pueden coexistir en una misma palabra:

(24) a kubaje-.-tem(-ka) caballo-CL-DIM-(SOC/INSTR) ‘El caballito gordito’ (Monguí 1981: 66)

b bóng-.e-tem -ng helecho-CL-DIM-PL ‘palos de helechitos’ (Monguí 1981: 105)

38 Howard traduce por ‘rana-vieja’. -xema es el sufijo caritativo, que puede indicar compasión, cariño y/o dulzura. RASGOS TIPOLÓGICOS KAMSÁ 187

El uso de más de un clasificador por palabra ha sido mencionado también en otras lenguas (amazónicas) de la zona, entre las cuales pueden citarse las tukanas. Para el koreguaje véase el estudio de Cook & Criswell (1993). En Waorani, Peeke (1973: 125) proporciona ejemplos de palabras formadas únicamente por clasificadores ensartados uno tras otro. Los sufijos, tanto de caso como número, se sufijan al clasificador:

(25) xocn--en.e ‘cama-CL-LOC’ ‘en la cama’

(26) pi-.-én. ‘peña (

(27) t-mn-be-ka ‘DEIC-huevo-CL-SOC/INSTR ‘con el (este) huevo’

El sufijo derivativo -mé, en cambio, y si es acortamiento de -téma/ -téme (diminutivo), aparece antes del clasificador en la palabra nd.-mé-e ’piedrecillas’. Compárense, sin embargo, los ejemplos siguientes que reflejan el orden usual raíz- CL-DIMIN(-CASO):

(28) boko-je-tem ‘chicha-CL-DIM’ ‘chichita’

(29) tuxua--tém-én.-ka ‘estera-CL-DIM-LOC-? ‘(en) donde la esterita’ (Monguí 1981: 45)

A veces, el clasificador no aparece, aunque este fenómeno puede deberse a la aplicación de reglas morfofonológicas:

(30) bes.á-.e ‘cabeza-CL’ : bes.á-n.e ‘cabeza-LOC’

El trabajo de Dixon (1986) hace hincapié en la distinción entre clases nominales, por una parte, y clasificación nominal, por otra: las primeras constituyen un sistema gramatical cerrado y obligatorio para casi todos los nombres de la lengua; además, los morfemas de clases nominales pueden aparecer como afijos. Los clasificadores nominales, en cambio, suelen distinguirse por su número, que puede llegar a ser bastante elevado. Por otra parte, subraya el mismo autor, los clasificadores son siempre lexemas libres, que nunca forman unidad morfológica con el nombre. Esta afirmación terminante se puede contrastar con la de Allan (1977), quien habló con más cautela de “morfemas” en vez de “lexemas libres”. Con referencia a la clasificación de Dixon (1986), el kamsá ostentaría rasgos de ambas categorías. En efecto, comparte con las clases nominales el hecho de poder afijarse al nombre, y con los clasificadores nominales el de tener un número elevado de miembros. Estos problemas con la clasificación de Dixon han sido señalados por Payne (1987) y Derbyshire & Payne (1990). Payne (1987), con referencia a los sistemas de 188 FABRE

clasificación nominal, agrupa las lenguas del oeste amazónico en cuatro grupos tipológico-areales. El grupo 1 comprende las lenguas tukano, peba-yagua y algunas de la familia witoto, y se distingue por la imposibilidad de incorporar los clasificadores al verbo. Además, muchos nombres pueden tomar más de un clasificador. También típico para estas lenguas es el hecho de que los clasificadores no sean formas libres sino afijos. Los clasificadores llevan rasgos característicos tanto de inflexión como de derivación. La marca de clase no aparece solamente en el nombre, sino también en determinantes del mismo. En la mayoría de los casos, los clasificadores deben aparecer con los numerales. Este último punto, sin embargo, no cuadra con los datos del kamsá.39 Otra diferencia, entre las lenguas del grupo 1 y el kamsá, es la presencia en la mayoría de las lenguas del grupo 1, según Payne, de un sistema de referencia cruzada en el verbo, que indica rasgos tales como [±animado] y género del sujeto y objeto. Con todo, es indudable que dentro del esquema propuesto por Payne, el kamsá queda mucho más próximo al grupo 1 que de los demás.40 Desde un punto de vista areal, es natural que el kamsá aparezca bien anclado en la red de las lenguas que corresponden a la parte norte de la zona estudiada por Payne. En su estudio sobre los diferentes sistemas de clasificación nominal en las lenguas amazónicas, Derbyshire & Payne (1990) distinguen siete tipos de lenguas:41 (1) sistemas con clasificación numeral, (2) sistemas con clasificación numeral y concordancia, (3) sistemas con clasificación numeral, concordancia e incorporación verbal, (4) sistemas con concordancia, (5) sistemas con clasificación numeral e incorporación verbal, (6) sistemas con incorporación verbal, y (7) sistemas con concordancia e incorporación verbal. Para situar la lengua kamsá en este contexto, cabe observar que esta lengua no usa clasificadores con los numerales:

(31) kanje bc trónkon-.-én. uno grande tronco-CL-LOC ‘en un gran troncazo’ (Monguí 1981: 31)

(32) tu-úta .kena-xemá-t DEIC-DU blanco-CAR-DU (concordancia de dual) ‘dos blanquitos’ (Monguí 1981: 172)

39 Tampoco es el caso en una lengua zápara como el arabela, que pertenece al grupo tipológico-areal 1 de Payne (1987). El arabela emplea solamente tres numerales nativos mientras el kamsá tiene muchos más. Sin embargo, los clasificadores del arabela, a diferencia de los del kamsá, van afijados a la raíz únicamente en casos de derivación, y nunca, según parece, con nombres básicos de la lengua. 40 Las lenguas del grupo 2 desconocen la clasificación nominal, así que no se da tampoco la incorporación de clasificadores en el verbo (omagua y cocama, ambos tupí-guaraní, más, quizás, algunas lenguas witoto). Al grupo 3 pertenecen lenguas como el waorani (lengua aislada), chayahuita (familia cahuapana) y las lenguas arawak pre-andinas. Este grupo se distingue por emplear, además del sistema clasificatorio del grupo 1, la incorporación en el verbo de los clasificadores. En cuanto al grupo 4, su rasgo distintivo es el empleo mayoritario de la incorporación verbal de los clasificadores (Harakmbet y algunas lenguas pano). 41 Este estudio parece tener su origen en Payne (1987), pero abarca un territorio mucho más ámplio a nivel amazónico. En otro artículo, Derbyshire & Payne (1990) han escogido, como lenguas representativas del grupo con concordancia como único rasgo de clasificación, las lenguas arawá y las maipure-arawak. RASGOS TIPOLÓGICOS KAMSÁ 189

(33) úta buakua-.-ka dos mano-CL-SOC/INSTR ‘dos manos’ (Monguí 1981: 195)

(34) únga .kena-.-ng bcca-.-ng-ka tres blanco-CL-PL grande-CL-PL-SOC/INSTR (concordancia de plural) ‘tres blancos grandotes’ (Monguí 1981: 32)

En cambio, los clasificadores aparecen, en determinados casos, con “adjetivos”42. Es significativo el hecho de que los “adjetivos” que ocurren con clasificadores lo hagan solamente cuando aparecen pospuestos (ejemplo 34 arriba y 35), pero no cuando preceden al nombre que modifican (ejemplo 31 arriba y 36). En estos casos, el clasificador aparece dos veces: primero sufijado al nombre y luego como copia en el “adjetivo”:

(35) kánje jébna- bien tangua--ka una casa-CL bien viejo-CL-SOC/INSTR ‘una casucha bien vieja’ (Monguí 1981: 192)

(36) bc jébna-.-ka i tanguá jébna-kuem-ka grande casa-CL-SOC/INSTR y vieja casa-AUM-SOC/INSTR ‘Grande casota y vieja casota’ (Monguí 1981: 193)

En (36), es posible que el sufijo aumentativo del adjetivo (-kuem) supla al clasificador del nombre. Además, como lo muestran los ejemplos (37) y (38), el clasificador del nombre y el del adjetivo pueden ser distintos:

(37) kánje jébna- bien tanguán-. , pero bcá-.-ka una casa-CL bien vieja-CL pero grande-CL-SOC/INSTR ‘una casucha bien viejota, pero grandota’ (Monguí 1981: 83)

(38) nje t-jébna- je t-. i-o-x-c-o-njai, ya DEIC-casa-CL ya sola-CL(?) DIS-3SG-PRED-ASP-V-mirar

pero tanguá-.-ka pero vieja-CL-SOC/INSTR ‘ya la casucha solitaria miraba, pero viejísima’ (Monguí 1981: 83)

42 Antes de hablar de una clase de adjetivos en el kamsá, cabría averiguar su existencia como categoría distinta en la lengua. De ahí el uso de las comillas. 190 FABRE

(39) úng kórent .kena-.-ng-ka tres fornido blanco-CL-PL-SOC/INSTR ‘tres blancos fornidos’ (Monguí 1981: 141)

(40) t-bobóns i-o-x-ua-bókn kánje bc DEIC-joven DIS-3SG-MOV-?-sacar un grande

líber-.a-ka, i kánje ku-.-ka libro-CL-SOC/INSTR y una aguja-CL-SOC/INSTR ‘el jóven sacó un gran libro y una agujota’ (Monguí 1981: 94-95)

(41) muás i-o-x-ua-cá. kánje kanden-.a bca-.a más DIS-3SG-PRED-V-caer una cadena-CL grande-CL ‘Además cayó una cadenota grandota’ (Monguí 1981: 101)

(42) faxáns íl-xa i fsng-x blanco hilo-CL y negro-CL ‘hilo blanco e (hilo) negro’ (Monguí 1981: 104)

Como el ejemplo (43) lo muestra, es también posible dejar el adjetivo pospuesto sin ningún clasificador, pero ello puede deberse a la ausencia del mismo en el nombre:

(43) ko-ta-n-xa kánje tx-ók bien solidi 2SG-FUT-EV.VISTO.CON.PART-ir una montaña-LOC bien solitaria ‘Irás a una montaña43 bien deshabitada’ (Monguí 1981: 93)

Las oraciones (44a) y (44b) se siguen en el mismo texto. El clasificador -be del ejemplo (44b), sufijado al “adjetivo”, es anafórico, y refiere a la ‘piedra’ del ejemplo (23a), que también lleva el mismo clasificador:

(44) a t-nd.-be kat-oka DEIC-piedra-CL-CL mismo-LOC

f-s-n-x-is-an-boxona O 44 PL-1 -EV.VISTO.CON.PART-PRED-? -PAS-dejar ‘nosotros dejamos la piedra en este mismo lugar’ (Howard 1977: 62)

43 El clasificador esperado no aparece aquí, ni tampoco en el adjetivo pospuesto. En el vocabulario de Clough & al. (1992), el equivalente para ‘monte/ bush’ es txá- e. 44 Este morfema aparece en la lista de McDowell (1994) glosado como ‘acción transitoria’. RASGOS TIPOLÓGICOS KAMSÁ 191

b n.á t tt-be x-tt-ijiba-ma cuán pesada-CL INF-ITER-traer-FIN ‘cuán pesada (era la piedra) para traerla de nuevo’ (Howard 1977: 62)

En cuanto a clasificadores incorporados en el predicado, mi análisis de los textos ha detectado dos ejemplos posibles (45 y 48). Como quiera que el análisis de los morfemas del verbo se encuentra en un estadio incipiente, es posible que algunos de estos morfemas tengan una forma diferente de los que aparecen unidos a lexemas pertenecientes al sintagma nominal. El sistema clasificador del kamsá podría, entonces, pertenecer al sistema (4) arriba citado, o sea a un sistema con clasificación nominal de concordancia. A continuación, paso a explicitar las modalidades propias de los clasificadores del kamsá, que difieren de las que ocurren en las lenguas de las familias lingüísticas arawá (dení, jamamadí, madija-culina y paumarí) y maipure (arawak pre-andino: apurinã y piro) que sirven de ejemplos para el estudio de Derbyshire & Payne. En efecto, las lenguas arawá emplean un sistema bifocal con clasificación nominal basada, por un lado, en el género (masculino/ femenino), y por otro lado, para una pequeña clase de nombres, en otras consideraciones. Las lenguas maipure (piro y apurinã), en contraposición, presentan un solo tipo de clasificación, basado en el género. En piro y apurinã, la concordancia con el referente exterior al verbo no aparece ligada al actante externo, sino queda incorporada al verbo como afijo de sujeto u objeto. Puede también aparecer con un demostrativo del sintagma nominal. En las lenguas arawá, la concordancia, tanto de género como de no-género, tampoco aparece unida al nombre sino en otras posiciones tales como demostrativos, adjetivos, interrogativos, marcadores de aspecto, imperativos, evidenciales, subordinadores etc. Lo que caracteriza a los clasificadores del kamsá, en cambio, son los hechos siguientes: (1) el kamsá no pertenece a ninguno de los dos subtipos de lenguas que emplean solamente una clasificación de concordancia, pues desconoce el género. Formaría pues un tercer subtipo de lenguas, ausente en el esquema de Derbyshire & Payne (1990); (2) algunos clasificadores, pero no me consta si todos, o solamente parte de ellos, señalan algún rasgo saliente del referido. Ello se conforma con los sistemas de clasificadores conocidos; (3) en la mayoría de los casos, estos clasificadores aparecen ligados directamente a los nombres, o a los adjetivos pospuestos, pero no, en cambio, a los antepuestos. Los clasificadores del kamsá no son lexemas aislados, ni tampoco clíticos, lo cual se desprende claramente y sin ambigüedad alguna del orden de los sufijos: RAÍZ-CL-CASO. Desde un punto de vista semántico, la expresión de un rasgo saliente del refe- rente, en el kamsá, es parecido a los sistemas de las lenguas del grupo tukano, witoto, andoke etc. (clase natural, forma, consistencia, uso etc.). Según Cook & Criswell 192 FABRE

(1993), en koreguaje (tukano occidental) los sustantivos con referente humano en el singular van seguidos por un sufijo de género (masculino y femenino). En el plural, hay que emplear palabras separadas para indicar el género de los referentes humanos. En cuanto a los inanimados, éstos se dividen en dos grupos: los que llevan clasifica- dor, que son unos treinta en total, y los demás. Una diferencia notable entre los siste- mas de clasificadores del kamsá y koreguaje, es la presencia sistemática de clasifica- dores con los numerales en la segunda lengua, rasgo típico de las lenguas tukano en general, frente a su ausencia en el kamsá. El género tampoco existe en kamsá, mien- tras se considera rasgo importante en lenguas como tukano y witoto45. Obsérvase el uso de los clasificadores en el texto siguiente (45-48) (Howard 1977: 20; 1979: 288):

(45) kanje rata-.e bca-.e s-n-x-i-ati-te una rata-CL grande-CL 1SG-EV.VISTO.CON.PART-PRED-V-coger-CL(?) ‘yo cogí una rata grande’46

(46) t-ora tramp-á inje soje entonces trampa-CL otra vez

t-o-k-x-i-soboc-ka DIS-3SG-EV.VISTO.SIN.PART-PRED-V-sonar-SOC/INSTR ‘entonces la trampa sonó otra vez’

(47) t-ora inje-.e t-oka s-n-x-i-ate entonces otro-CL DEIC-LOC 1SG-EV.VISTO.CON.PART-PRED-V-coger ‘entonces yo cogí allí otra (rata) grande’

(48) s-n-x-obá i t-en.e 1SG-EV.VISTO.CON.PART-PRED-matar y DEIC-LOC

45 Tanto en las lenguas tukano como en witoto, las distinciones de género MASCULINO/ FEMENINO aparecen, con unas pocas excepciones, con nombres de referentes animados, mientras los clasificadores se unen a nombres de entidades no-animadas. 46 Existe una diferencia entre las dos versiones. La versión inglesa es como sigue:

a rat-big big-big I-caught ‘I caught a big rat’

frente a la versión española:

una rata-grande yo-cogí ‘Yo cogí una rata grande’

La mayor diferencia estriba en que, en la versión inglesa, el verbo termina en un morferma que difícilmente podría ser otra cosa que el clasificador, mientras éste está ausente de la versión española. Quizás esta diferencia se deba al orden NOMBRE + ADJETIVO en el primer caso, y ADJETIVO + NOMBRE, en el segundo. RASGOS TIPOLÓGICOS KAMSÁ 193

s-n-x-a-ca-.e taxas-oi-ka 1SG-EV.VISTO.CON.PART-PRED-V-arrojar-CL patio-LOC-LOC ‘yo la maté y la arrojé al patio’47

Como se podrá apreciar, (45) presenta un actante nuevo, triplemente marcado con el clasificador -.e, en el nombre, en el adjetivo y luego en el verbo. En (47), aparece la palabra ‘otro/ otra’, a la cual ha sido añadido el clasificador -.e, que refiere a la misma categoría de referentes que ‘rata’ en (45). En la conclusión de su importante artículo sobre los sistemas de clasificación nominal de las lenguas amazónicas, Derbyshire & Payne (1990) aluden a la importancia de los clasificadores como índices conectivos pragmático-discursivos, hecho que se ve plenamente reflejado en el ejemplo precedente. Hay que señalar, no obstante, que una vez introducidos los actantes principales, el uso de los clasificadores se vuelve redundante, por lo menos en verbos y adjetivos. En cambio, los clasificadores parecen muy resistentes como morfemas ligados al sustantivo. En el ejemplo (49) se ve, nuevamente, que el clasificador - á del “adjetivo” retoma el del nombre que precede:

(49) t-karr-a-nga enan-á t-o-k-x-a-tnxna DEIC-carro-CL-PL vacío-CL DIS-3SG-EV.VISTO.SIN.PART-PRED-V-pasar ‘los carros vacíos pasaron’ (Howard 1977: 49)

Entre las lenguas del sudoeste colombiano y zonas aledañas, la existencia de clasificadores ha sido señalada para las lenguas siguientes: (1) cofán, lengua para la cual Borman (1976) menciona siete clasificadores48; (2) koreguaje y lenguas de la familia tukano en general, tanto en su rama occidental como en la oriental (Cook & Criswell 1993: 18-26); (3) Emberá, con cinco clasificadores identificados (Aguirre Licht 1999)49, (4) waorani (Peeke 1973: 125-128), (5) witoto, con unos 50 clasificadores (Petersen de Piñeros 1994: 38-39), y (6) andoke (Landaburu 1979: 133-134 y 322-323)50. En cambio, los clasificadores faltan en inga, nasa yuwe, awá pit y guambiano. No tengo los datos relevantes para el casi extinto carijona de la familia caribe, pero como estas lenguas no usan clasificadores, supongo que el carijona tampoco.51

47 La glosa original de Howard para la primera palabra reza: “yo-maté-la”, lo que presta a confusión, debido a que -obá es sin lugar a dudas la raíz del verbo ‘matar’. El texto de Howard está plagado de tales confusiones. 48 Cofán: - to ‘diminutivo o forma redonda’; -kho ‘aumentativo o forma angular’; -khu ‘ahuecado’; -fa ‘largo o alto’; -si ‘peludo’; -hiõ ‘alto como árbol’ y -he ‘plano’ (Borman 1976: 115-122). 49 Emberá: -xã ‘superficies planas’; -á ‘líquidos’; -pã ‘superioridad’; -dar ‘superficies cartilaginosas’; -mór ‘sustancias y órganos internos del cuerpo’ (Aguirre 1999: 75-76) 50 Para esta lengua, Landaburu habla de nombres clasificadores, los que se combinan obligatoriamente con determinantes pronominales. 51 El texto carijona publicado por Robayo (1989) parece comprobar la inexistencia de clasificadores en esta lengua. 194 FABRE

Para finalizar este apartado, debe mencionarse que algunos nombres, entre los cuales destacan varios nombres de vegetales y animales, no llevan ningún clasificador. Comoquiera que una terminación -na aparece con frecuencia, ésta, sin embargo, podría ser un clasificador, lo que reduciría drásticamente los nombres sin clasificador. En el caso contrario, la terminación -na debería ser tenida por el fruto de pura casualidad, y formaría entonces parte de la raíz.

6. Conclusiones La posición geográfica del valle de Sibundoy, así como el papel desempeñado por sus habitantes aborígenes, mediadores entre el mundo andino y la selva amazónica, son hechos que han dejado su huella tanto en la etnohistoria de la zona como en la estructura de la lengua kamsá, que comparte rasgos de las lenguas de estos dos mundos tan diferentes y sin embargo siempre comunicados como bien lo señalan Bonilla (1972) y Taussig (1985). Desde el punto de vista lingüístico areal, los lazos amazónicos (o por lo menos de las tierras bajas si se incluye a la vertiente del Pacífico) pueden detectarse en la presencia de incorporación nominal y de clasificadores, ambas categorías desconocidas en las lenguas andinas del sur colombiano. En cambio, la ausencia de género masculino/ femenino, categoría típica de las lenguas amazónicas adyacentes, con la salvedad del cofán, formaría un eslabón con las lenguas andinas. Como rasgos originales del kamsá, ausentes de las lenguas andinas y amazónicas del área, pueden citarse la presencia de dos actantes prefijados en el predicado, la categoría de dual y un subtipo de incorporación nominal, original por sus modalidades de empleo.

Referencias bibliográficas Aguirre Licht, Daniel 1999 Emberá. Munich: Lincom-Europa. Allan, Keith 1977 ‘Classifiers’, Language 53: 284-310. Baker, Mark C. 1988 Incorporation. A theory of grammatical function changing, Chicago: Chicago University Press. Barnes, Janet 1999 ‘Tucano’, en: R.M.W. Dixon & Alexandra Y. Aikhenvald (eds.), pp. 207-226. Bonilla, Víctor Daniel 1972 Serfs de Dieu et maîtres d’Indiens. Histoire d’une mission capucine en Amazonie, París: Fayard. Borman, M.B. 1976a Vocabulario cofán, Quito: ILV. 1976b ‘Apéndice gramatical’, en: M.B. Borman (1976a), pp. 115-122. 1979 ‘Cofán paragraph structure and function’, en: Robert E. Longacre (ed.), Part 3: 289-338. RASGOS TIPOLÓGICOS KAMSÁ 195

Clough, Alice, Linda Howard, Robert van Zyl & Katherine van Zyl 1992 ‘Kamsá’, en: Randal Huber & Robert B. Reed (eds.), Vocabulario comparativo: palabras selectas de lenguas indígenas de Colombia, Bogotá: ILV. Cook, Dorothy M. & Linda L. Criswell 1993 El idioma koreguaje (tucano occidental), Santafé de Bogotá: ILV. Colette Craig (ed.) 1986 Noun classes and categorization, Amsterdam: John Benjamins. Croft, William 1990 Typology and universals, Cambridge: CUP. Curnow, Timothy Jowan 1997 A grammar of Awa Pit (Cuaiquer): An indigenous language of south-western Colombia, Ph.D. diss., The Australian National University. Derbyshire, Desmond C. 1987 ‘Areal characteristics of Amazonian languages’, International Journal of American Linguistics 53: 311-326. Derbyshire, Desmond C. & Doris L. Payne 1990 ‘Noun classification systems of Amazonian languages’, en: Doris L.Payne (ed.), pp. 243-271. Dixon, R.M.W. 1986 ‘Noun classes and noun classification in typological perspective’, en: Colette Craig (ed.), pp. 105-112. Dixon, R.M.W. & Alexandra Y. Aikhenvald (eds.) 1999 The Amazonian languages. Cambridge: CUP. Eastman, Robert & Elizabeth Eastman 1963 ‘ Iquito syntax’, en: Benjamin F. Elton (ed.), Studies in Peruvian Indian languages, I, Norman: Summer Institute of Linguistics, pp. 145-192. Fabre, Alain 1998 Manual de las lenguas indígenas sudamericanas, 2 vol, Munich: Lincom Europa. e.p. a ‘Modalidades de uso del morfema -bu ‘tercera persona del dual etc.’ en kamsá (lengua aislada hablada en el valle de Sibundoy, Putumayo, Colombia)’. e.p. b Los estudios lingüísticos sobre la lengua kamsá. Resumen histórico y situación actual. Gómez-Imbert, Elsa 1988 ‘Construcción verbal en barasana y tatuyo’, Amerindia 13: 97-108. Harms, Phillip Lee 1994 Epena Pedee syntax. Arlington: SIL/ The University of Texas at Arlington. Howard, Linda 1967 ‘Camsá phonology’, en: Viola G. Waterhouse (ed.), Phonemic systems of Colombian languages, Norman: SIL/ University of Oklahoma, pp. , 73-87. 196 FABRE

1972 ‘Fonología del Camsá’, en: Viola G. Waterhouse (ed.), Sistemas fonológicos de idiomas colombianos, 1, Bogotá: ILV, pp. 77-92. [traducción de Howard 1967] 1977 ‘Esquema de los tipos de párrafo en Camsá’, en: Estudios en Camsá y Catío, Lomalinda: ILV, pp. 1-67. 1979 ‘Camsá: certain features of verb inflection related to paragraph types’, en: Robert E. Longacre (ed.), Part 2: 273-296. [traducción de Howard 1977] Jamioy Muchavisoy, José Narciso 1989 Estructura verbal Kamntsa, Tesis, Santafé de Bogotá: Universidad Nacional de los Andes, Centro Colombiano de Estudios en Lenguas Aborígenes. 1992 ‘Tiempo, aspecto y modo en Kamentsa’, Memorias del II Congreso del CCELA, Bogotá: Universidad Nacional de los Andes, Centro Colombiano de Estudios en Lenguas Aborígenes. 1997 ‘Experiencias vividas en la traducción de apartes de la Consitción política de Colombia a la lengua kamëntsa’, Amerindia. Revue d’Etnolinguistique Amérindienne, 22. Jijón y Caamaño, Jacinto 1998 El Ecuador interandino y occidental antes de la conquista castellana, Tomos I-IV, Edición facsimilar. Quito: Abya-Yala. Juajiboy Mutumbajoy, María Clara 1995 Actancia verbal en Kamntsa, Santafé de Bogotá: Universidad Nacional de los Andes, Centro Colombiano de Estudios en Lenguas Aborígenes. Landaburu, Jon 1979 La langue des Andoke (Amazonie colombienne), París: Société d’Études Linguistiques et Anthropologiques de France. 1994 ‘Deux types de prédication, avec ou sans sujet: quelques illustrations colombiennes’, en: Jon Landaburu (ed.), pp. 639-663. 1998 Clasificación de las lenguas indígenas de Colombia. Santafé de Bogotá. Landaburu, Jon (ed.) 1994 Estructuras sintácticas de la predicación: lenguas amerindias de Colombia, Número temático de Bulletin de l’Institut Français d’Études Andines, 23/3, Lima: Institut Français d’Études Andines. Lazard, Gilbert 1994 L’actance, París: PUF. Levinsohn, Stephen H. 1974 Una gramática pedagógica del inga (primera parte), Bogotá: Ministerio de Gobierno/ ILV. 1976a The Inga language, La Haya: Mouton. 1976b Una gramática pedagógica del inga (segunda parte), Bogotá: Ministerio de Gobierno/ ILV. RASGOS TIPOLÓGICOS KAMSÁ 197

Long, Violeta 1985 Estudio de la lengua guambiana, St. Andrews: University of St. Andrews, Centre for Latin American Linguistic Studies, Working Paper No. 20. Longacre, Robert E. (ed.) 1979 Discourse grammar. Studies in indigenous languages of Colombia, Panama, and Ecuador, Dallas: SIL. McDowell, John H. 1982 ‘Beyond iconicity: ostention in Kamsá mythic narrative’, Journal of the Folklore Institute, 19: 119-141. 1990 ‘Construir la comunidad kamsá por medio del lenguaje ritual’, en: Ellen Basso & Joel Sherzer (eds.), Las culturas nativas latinoamericanas a través de su discurso, Quito: Abya-Yala, pp. 297-318. 1994 “So wise were our elders”. Mythic narratives of the Kamsá, Lexington: University Press of Kentucky. Mithun, Marianne 1984 ‘The evolution of noun incorporation’, Language 60: 847-894. 1986a ‘The convergence of noun classification systems’, en: Colette Craig (ed.), pp. 379-397. 1986b ‘On the nature of noun incorporation’, Language 62: 32-37. Monguí Sánchez, José Raúl 1981 La lengua kame.ntzá. Fonética - fonología - textos, Bogotá: Publicaciones del Instituto Caro y Cuervo, LIX. Nieves Oviedo, Rocío 1994 ‘Morfología del predicado en nasa yuwe (lengua páez)’, en: Jon Landaburu (ed.), pp. 539-566. Obando Ordóñez, Pedro 1992 Awa-kwaiker: an outline grammar of a Colombian/Ecuadorian language, with a cultural sketch, Ph.D. diss., The University of Texas at Austin. Payne, David L. 1990 ‘Some widespread forms in South American languages’, en: Doris L. Payne (ed.), pp. 75-87. Payne, Doris L. 1987 ‘Noun classification in the Western Amazon’, Language Sciences 9: 21-44. 1990 ‘Morphological characteristics of Lowland South American languages’, en: Doris L. Payne (ed.), pp. 213-241. Payne, Doris L. (ed.) 1990 Amazonian linguistics. Studies in Lowland South American Languages, Austin: University of Texas Press. Peeke, M. Catherine 1973 Preliminary grammar of Auca, Norman: SIL/ University of Oklahoma. 198 FABRE

Petersen de Piñeros, Gabriele 1994 La lengua uitota en la obra de K.Th. Preuss, Bogotá: Universidad Nacional de Colombia. Pike, Evelyn G. & Rachel Saint (eds.) 1988 Workpapers concerning Waorani discourse features, Dallas: Summer Institute of Linguistics. Robayo, Camilo 1989 ‘“En faisant une rame”, texte carijona’, Amerindia. Revue d’Ethnolinguistique Amérindienne, 14: 189-199. Rojas Curieux, Tulio 1994 ‘Expresión de la categoría de tiempo gramatical en el predicado nasa yuwe (lengua páez)’, en: Jon Landaburu (ed.), pp. 567-600. 1998 La lengua páez. Una visión de su gramática, Bogotá: Ministerio de Cultura. Rojas Curieux, Tulio, Rocío Nieves Oviedo & Marcos Yule Yatacué 1991 Estudios gramaticales de la lengua Páez (Nasa Yuwe), Bogotá: Universidad Nacional de Colombia, Centro Colombiano de Estudios en Lenguas Aborígenes. Slocum, Marianne C. 1986 Gramática páez, Lomalinda: Editorial Townsend. Strom, Clay 1992 Retuarã syntax, Dallas: Summer Institute of Linguistics. Tandioy Chasoy, Domingo, Stephen H. Levinsohn & Alonso Maffla Bilbao 1978 Apuntes sobre la gramática inga, Lomalinda: Editorial Townsend. Taussig, Michael 1987 Shamanism, colonialism, and the wild man. A study in terror and healing, Chicago: The University of Chicago Press. Triviño Garzón, Lilia 1994 ‘Hacia una tipología de la predicación de la oración simple en la lengua guambiana’, en: Jon Landaburu (ed.), pp. 601-618. Van Zyl, Robert J. & Marco Freddy Solarte R. 1985 ‘Reseña al trabajo de José R. Monguí S. titulado “Fonética y fonología del kame.ntza”’ (MS.) Vásquez de Ruiz, Beatriz 1988 ‘La predicación en guambiano’, en: Lenguas aborígenes de Colombia, Serie Descripciones, vol. 2, Bogotá: Universidad Nacional de Colombia, Centro Colombiano de Estudios en Lenguas Aborígenes. 1994 ‘La oración compuesta en guambiano’, en: Jon Landaburu (ed.), pp. 619-637. Wise, Mary Ruth 1999 ‘Small language families and isolates in Peru’, en: R.M.W. Dixon & Alexandra Y. Aikhenvald (eds.), pp. 307-340. PERSON HIERARCHY AND ITS IMPLICATIONS: THE CASE OF AYMARA1

Elena Filimonova University of Konstanz, Germany

1. Introduction Hierarchies play a significant role in the organisation of phonological, grammatical, and lexical structure (to mention a few: Keenan & Comrie’s accessibility hierarchy (1977 [1972]), Berlin & Kay’s colour term hierarchy (1969), Foley’s syntactic bond- edness hierarchy (1980), Corbett’s hierarchy of semantic agreement targets (1979), hierarchy suggested by different scholars for various linguistics parameters – e.g. Smith-Stark (1974) for number marking, Silverstein (1976), Moravcsik (1978) and others for case marking distribution). In this paper, the person hierarchy will be at issue. Concentrating on Aymara, I will show in what way the ranking of persons can influence morphosyntactic patterns. I hypothesise that person hierarchy is not an “artificial” concept developed by linguists to describe different structural constraints but rather a cognitive parameter inherent to the worldview of the corresponding lan- guage society. According to their involvement in the speech act, one traditionally distinguishes three main characters, or persons: Speaker (S), Addressee (A), and Non-Locutor (or Non-participant of the speech act) (N). Since the time of the first grammars written by ancient Greeks, these persons are usually referred to by numbers, whereby 1st person refers to Speaker, 2nd to Addressee and 3rd to Non-locutor, where the order of enumeration clearly depicts the principle that the Speaker is presupposed to play the prominent role in interpersonal communication. In this paper, however, I will consider cases where the Addressee (the 2nd person) appears to be most salient. Starting with the analysis of morphosyntactic rules, I will argue that grammatical salience of the person spoken to has its functional motivation and is conditioned by the rules of interpersonal communication of the respective culture, which merits a special status not to the Speaker but to the person addressed. The paper will be structured as follows. In the next section, I will briefly intro- duce the main features of the person hierarchy. Section 3 will deal with Aymara data demonstrating the ordering of persons. Section 4 will discuss some principles of Ay- mara culture revealing the underlying conception of Addressee salience. Finally, the major findings of the paper will be summarised.

1 Acknowledgements: I am indebted to Martha Hardman for answering my numerous questions about Aymara structure and for providing me with additional materials. I am also grateful to Irina Nikolaeva and the editors for useful comments on an earlier version of the paper. Special acknowledgements go to the Spinoza project at Leiden University (Pieter Muysken and his colleagues) who invited me to participate at the Workshop on Bolivian and Rondonian Indigenous Languages. Thanks also to Arista Da Silva and Allison Wetterlin for polishing up my English. All mistakes are my own. 200 FILIMONOVA

2. Person hierarchy and its features Until recently the hierarchical ordering of persons has not received extensive analysis in linguistic theory. Proceeding from the traditional designation of persons by num- bers (and possibly supported by the egocentric nature of the Western worldview), the ordering 1>2>3, in our terms S>A>N, has been considered as the only possible and even axiomatic configuration. However, with the beginning of active studies of the indigenous languages in the last few decades, this stereotype was abandoned. Sur- veys of person marking morphology in Algonquian languages, which seemed to be clearly motivated by the opposite ranking of speech act participants, i.e. the hierarchy A>S>N, had an overwhelming effect in violating the assumption that the Speaker is universally more highly ranked than the Addressee. The only universal which seems to be absolute across languages is that speech act participants (S and A) always over- rank the Non-locutor(s): {S, A}>N. The person hierarchy is typically manifested in the polypersonal verb agreement which indicates more than one argument of the verb (e.g. subject and non-subject, subject and direct object, subject, direct and indirect object, etc.). Personal verb in- flection in itself is a two-faced phenomenon: On the one hand, agreement is triggered by verbal arguments and in this sense it signals their presence in the syntactic repre- sentation of the clause; on the other hand, agreement morphemes encode deictic characteristics of these arguments (e.g., Speaker, Addressee, Non-locutor). These two parameters are logically independent, but we may be able to speak about some proto- typical correlation between grammatical relations and the deictic properties they are associated with. It is generally assumed that locutors are more active than the Non-locutors, i.e. they are more likely to act and not to be acted upon, and therefore are typically con- sidered to be the best candidates for the subject position. The reverse is true for Non- locutors: They are typically considered to have less control over the situation than any of the speech act participants (SAPs) and in the clause structure are more likely to be expressed as one of the non-subject arguments. Technically this prototypical distribution of participants over grammatical relations is determined by a one-to-one correspondence between higher and lower ranked positions on two hierarchies: the person and relational hierarchies. The person hierarchy, as already mentioned above, generally orders SAPs above Non-locutors (whereby SAPs can in turn be differenti- ated as higher vs. lower ranked ones) and relational hierarchy ranks subject above other verbal arguments. When, however, the distribution of persons over grammatical relations does not fit the prototypical correlation, i.e. when the higher ranked person corresponds to a lower ranked argument and the lower ranked person occupies the more prominent argument position, a conflict situation arises. Languages which are sensitive to the combinability of deictic and argument statuses signal this conflict morphosyntactically. In such cases, the usual encoding of polypersonal situations does not hold, thus falling outside the frames of the standard morphological para- digm. A close inspection of the persons involved and the analysis of their distribution AYMARA PERSON HIERARCHY 201 over higher vs. lower ranked argument positions enables us to reconstruct their mu- tual ordering. The means of expressing the prominent status of a given person may considera- bly vary from language to language depending on morphonologic, syntactic and se- mantic constraints of the language in question. It would, therefore, be useless to search for a procedure generally applicable to all languages in order to determine the configuration of the person hierarchy (cf. Heath 1998). It should be noted, however, that cases of unambiguous salience of S or A in the hierarchy of persons occur much more rarely than those where the mutual ranking of SAPs is unclear or opaque. Heath, for example, in discussing the patterning of transi- tive constructions in Australian and Native American languages, argues that it is ex- tremely difficult to rank one SAP over another: “The 1Ù22 combinations are doubly dangerous because they not only contain the most pragmatically sensitive pronomi- nals they also combine them into a syntagmatic structure and thereby necessarily focus attention on the speaker-addressee relationship” (Heath 1991:86). In these situations, “structures that make the most sense cognitively or formally are actually avoided” (Heath 1998:102), instead, opaque and non-transparent structures which mitigate the reference to speech act participants are used. In these cases, the configu- ration of the person hierarchy may be defined as {S ≈A}>N, without any preference to anyone of the SAPs. Among the languages which do signal the higher rank of a given speech act par- ticipant, contrary to the expectations that languages with Speaker and Addressee sali- ence are equally distributed, those favouring the Speaker clearly prevail. The cases where the Addressee overrides the Speaker in the person hierarchy are very uncom- mon and therefore often tend to be ignored in cross-linguistic surveys. This paper investigates the phenomenon of person hierarchy in Aymara, a Jaqi language spoken by 2,200,000 people in Bolivia, Peru, Chile, and Argentina (data from Ethnologue, 14th edition). The goal of this study is to demonstrate deictic and grammatical salience of the person spoken to in Aymara grammar and to examine extralinguistic features possibly motivating this particular configuration of the person hierarchy (A>S>N). A special status of the Addressee in the Jaqi languages has already been pointed out by Martha Hardman. She considers the salience of the 2nd person (in our terms Addressee) to be one of the tenets, or ‘linguistic postulates’, as she calls them, real- ised in the language. In Hardman’s terms, linguistic postulates are “those recurrent categorisations in the language which are most directly and most tightly tied to the

2 Heath (1991) uses abbreviation ‘X=>Y’ to refer to the distribution of persons over the main arguments of the transitive verb. An element to the left from the arrow refers to the person occupying the subject position and an element to the right corresponds to direct object. An arrow (=>) denotes the transition of the action from one person to another. E.g. 1=>2 means that 1st person is a subject and 2nd person a direct object. Further, the bi-directional arrow (Ù) in abbreviations like 1Ù2 denotes the conjunction of person combinations 1=>2 and 2=>1. In the following discussion I adopt these abbreviations referring to person distribution over subject and non-subject arguments in the verbal inflectional markers.

202 FILIMONOVA perceptions of the speakers, those elements which, while language imposed, are so well imposed that speaker considers them just naturally part of the universe ...” (1978a:122-123). In other words, she believes that language specific categories con- tribute to a worldview in a speaker’s culture. Hardman’s theory of linguistic postu- lates clearly resembles (and was probably inspired by) the Sapir-Whorf hypothesis, according to which language determines, or less resolutely, has a tendency to influ- ence thought and cognition, thus presupposing correlations between different linguis- tic phenomena and variations in non-linguistic behaviour. I, however, inspired by the ideas presented by Lee 1959 and Plank 1985, suggest that it is not the language that shapes speakers’ worldview but rather the culture that imposes the language and gets manifested in its structure.

3. Aymara data In this section, I will mainly concentrate on Aymara data, demonstrating the salience of the person spoken to. This salience can be seen, first and foremost, in the verbal personal inflection. The Aymara verb is characterised by cumulative tense-person morphemes. Usually only two participants of the particular situation receive their reference in the person marker, one of them always being a subject participant and the other one of the non-subject arguments. In Jaqi languages, person-tense suffixes are normally treated as unit morphemes. Person and tense markers have amalgamated to an extent that it is difficult to distinguish (at least synchronically) any constitutive morphemes. Nevertheless, due to the comparative analysis of Jaqi languages and historical reconstruction (Hardman 1978b), some correlations can be stated as “dis- tinctive features” designating the four persons. These person features are not always fixed to a particular argument position and in some cases, the same marker can indi- cate both subject and non-subject complements, cf. Table 1.

Person Person distinctive features Speaker: -it (as subject and non-subject complement); -ta (as subject) Addressee: -ta in the simple tense; -m in the non-realised tenses and some other places, especially as non-subject complement, and long vowel3, particularly when joined to one of the other markers; Non- basically unmarked, either as subject or non-subject comple- Locutor: ment. -i in the simple tense is easily absorbed and/or omitted; -p or -pa in non-realised tenses; Inclusive: an interpolated -s- over the corresponding form for 1st person (as non-subject complement); a nasal (as subject), additionally: -ta for non-future realised tense, -sa for non-realised tenses Table 1: Person distinctive features in Aymara (adapted from Hardman-de Bautista, Vásquez & Yapita (from here on HVY) 1974:232-233)

3 This long vowel derives from a proto-Jaqi syllable *-ma, still occurring in some dialects of Aymara and in the other Jaqi languages. AYMARA PERSON HIERARCHY 203

Considering these person features and analysing the tense-person markers of different combinations of persons, we can observe that the distribution of the corresponding person markers is far from uniform. Consider table 2, where simple tense person markers are given:

Person Morphol. Marked person Marked argument combination marker 1=>2 -sma Addressee marked non-subj. 2=>1 -ista Speaker and Addressee marked4 subject & non-subj. 2=>3 -ta Addressee marked subject 3=>2 -tam Addressee marked twice non-subj. 1=>3 -ta Speaker marked subject 3=>1 -itu Speaker marked non-subj. Incl=>3 -tan Incl marked (Speaker subject subject marker + nasal) 3=>Incl -istu Incl marked (Speaker comple- non-subj. ment marker + interpolated -s-) Table 2: Simple tense person markers in Aymara (adapted from HVY 1974:232-233)

In most combinations, only one participant of a given situation appears to be marked. However, it is neither systematically the subject nor the non-subject. The person pat- terning in Aymara follows the hierarchy: A>S>N., i.e. each person combination is signalled by the participant which occurs higher on the person hierarchy, independ- ently of its syntactic status. However, in the situations where Addressee is involved, the distribution of participants over grammatical relations is also significant. Thus, according to the above-mentioned rule, the combination 3=>2 should be signalled by the Addressee marker, since it is the higher ranked person in this combination. How- ever, in this particular case, due to its non-subject position which is considered to be less prestigious that the subject one, the person spoken to additionally appears to be even doubly marked5. The distribution of person markers in “1Ù2” combinations, i.e. where both SAPs are involved, plays an essential role in determining the configuration of the

4 With 2nd person as subject, 1st person and Inclusive fall together in form, and Inclusive semantically falls out. However, the old -s- referring to Inclusive, appears intermittently in 2=>1 markers. 5 The multiple marking of the Addressee in its subordinate role is not unique for Aymara, it is character- istic of the Jaqi family in general (Hardman 1966, 1978b). Interestingly, a similar pattern of Addressee salience is found in (Weber 1986, 1989). There transitive verbs usually have two slots for both subject and direct object, whereby the person markers precisely refer to the participants involved. However, in 3=>2 and 3=>Incl combinations, the 3rd person subject marker undergoes a kind of shift and appears to be of the same person as the direct object marker. That is to say that, 3=>2 is indicated by mor- pheme combination 2.Object-2.Subject and 3=>Incl is labelled by Incl.Object-Incl.Subject and not by 2.Object-3.Subject and Incl.Object-3.Subject, respectively. Thus, the 2nd person and inclusive appear to be marked twice, whereas 3rd person receives no encoding at all. Aside from morphological similarity, however, the tendencies of Addressee salience in Jaqi and Huallaga Quechua seem to be historically independent.

204 FILIMONOVA person hierarchy. Here, if the Addressee occurs in a subject position (2=>1), both Speaker and Addressee get encoded in the combination marker. However, if Ad- dressee occurs in a low-ranked non-subject position (1=>2), it is the only person that receives an overt expression. I see this marking strategy as clearly indicating that the person spoken to overranks the Speaker in the person hierarchy. Another interesting aspect of Addressee salience concerns the rules of person marking when arguments with semantic roles other than patient/theme trigger agree- ment. In this case, the choice of the complement to be referred to in the person marker is presupposed by a kind of relational hierarchy, according to which a benefi- ciary/purposive complement (marked by the -taki- morpheme) always takes prece- dence over a directional complement (marked with -ru- or -ta-). This principle is illustrated in examples (1-3) and (4-6), where the first series shows the prominence of the beneficiary complement over the directional complement ‘to, toward’, and the second series, the beneficiary over the directional complement ‘from, of’:

(1) jupa-r ch’uq chura-m 3-to potato give-IMP.2=>3 ‘Give potatoes to her’ (HVY 1974:316)

(2) k’’iti-taki-s (jupa-r) (ch’uq) chura-rapi-:-xa ? who-BEN-SENT6 3-to potato give-VBEN-FUT.1=>3-POL ‘On whose behalf shall I give her potatoes?’ (HVY 1974:316)

(3) naya-taki-w (jupa-r) (ch’uq) chura-rap-ita:ta 1-BEN-SENT 3-to potato give-VBEN-IMP.2=>1 ‘Give them to her on my behalf’ (HVY 1974:316)

(4) jupa-t kis ala-ni-: 3-from cheese buy-Near-FUT.1=>3 ‘I’ll buy cheese from him’ (HVY 1974:316)

(5) k’’iti-taki-s (jupa-t) (kis) ala-rapi-:ta ? who-BEN-SENT 3-from cheese buy-VBEN-FUT.2=>3 ‘On whose behalf will you buy cheese from him?’ (HVY 1974:317)

(6) juma-taki-w (jupa-t) (kis) ala-rapi-:ma 2-BEN-SENT 3-from cheese buy-VBEN-FUT.1=>2 ‘I’ll buy cheese from him on your behalf’ (HVY 1974:316-317)

6 Information on SENT: -wa marks the sentence as affirmative and/or personal knowledge; -sa marks one of the two basic question types of Aymara. It ordinarily occurs with an interrogative, either directly on the interrogative or on a construction containing it. AYMARA PERSON HIERARCHY 205

However, when Addressee occurs in a position of directional complement it can override a beneficiary and gain its expression in the person-tense marker. Thus, in (7a), unlike the examples above, a directional complement is marked.

(7) a juma-r jupa-tak ch’uq alja-:ma 2-to 3-BEN potato sell-FUT.1=>2 ‘I will sell you potatoes for (you to deliver/transmit etc.to) him’ (HVY 1974:317)

b juma-r jupa-tak ch’uq alja-: 2-to 3-BEN potato sell-FUT.1=>3 ‘I will sell you potatoes for him’ (Hardman, p. c.)

In (7a), the verb person-tense marker -:ma signals the Addressee, which is a direc- tional complement, and does not encode the person of a beneficiary-marked constitu- ent. Based on examples (4-6), one would expect here a “1=>3” verbal marker. According to Hardman (p.c.), any time there is an Addressee, it may take prece- dence over any other person marking. Thus, the Addressee, be it a beneficiary or di- rectional complement, may, at the discretion of the speaker, be marked in the verb over any other person. Thus, example (7b) is also grammatical, but the preferred variant is probably (7a). The next aspect of Aymara grammar which is also most probably motivated by Addressee salience concerns the use of independent personal pronouns. Typically, free pronouns in Aymara are omitted but may occur for redundancy or explicitness, in direct reflection of the verb person (HVY 1974:209). One would suppose that free pronouns of all four persons are used equally. However, as McKay (1985) showed with his statistical calculations of the frequency of free personal pronouns in Aymara texts, this assumption is far from true, cf. Table 3. As we can see, the pronoun referring to the Addressee clearly appears to be the most frequently used (in 3/4 of cases where 2nd person has been involved), whereas 1st person and inclusive free pronominal forms are only used in half of the cases (each 50%). Moreover, diachronically the 2nd person pronoun also appears to be the most stable. Thus, in comparison to the 1st person pronoun, which shows much variation between Jaqi languages in general and Aymara dialects in particular, the 2nd person pronoun seems to be resistant to all kinds of morphophonemic changes. This diachronic stability of pronominal forms referring to the Addressee may also be attributed to the salienced status of the person spoken to (Hardman 1978b:435).

206 FILIMONOVA

Person Pronominal forms % Speaker nä, na, naya 50 Addressee juma 75 Non-Locutor upa, jupa, jup’’a 10 Inclusive jiwsa, jiwasa 50 Table 3: The frequency of free pronoun use in Aymara (in %) (McKay 1985, adapted from Hardman 1988/1989:127)

Finally, one should mention in particular the inclusive/exclusive dichotomy in the Aymara pronominal system which, as I have shown in Filimonova (1999), seems to correlate with the grammatical salience of the person spoken to7. Traditionally, the inclusive/exclusive opposition is regarded as a subdistinction within the 1st person category. Its meaning is usually interpreted as denoting the inclusion vs. exclusion of the Addressee into the referential group of the Speaker, whereby the Speaker — aside from the presence of the other participants (A(s) or N(s)) which actually appear to be involved in the action to no lesser extent than the Speaker — is considered to be the main person in the corresponding set. In languages characterised for an Addressee salience, however, the semantic focus of the inclusive form shifts towards the Ad- dressee, thus emphasising his/her involvement in the corresponding group. Thus, structurally the Aymara inclusive has nothing to do with the 1st person and would be better recognised as a separate fourth person. The exclusive, however, is distinguished as a 1st person form. This can be clearly seen from (8) where the four basic pronominal forms and the corresponding referential sets are presented. The forms are unspecified for number but can optionally take a plural marker -naka, thus yielding in: naya-naka, jiwasa-naka, juma-naka, and jupa-naka.

(8) naya / na {S}, {S+N+...+N} jiwasa {S+A+...+A}, {S+A+...+N...} juma {A}, {A+A+...+A}, {A+N+...+N} jupa {N}, {N+N+...+N}

Interestingly, the status of an inclusive as a separate person seems to be not only a structural but also a psychological reality, being perceived by the speakers as an “in- tegral part of Aymara thought and culture” (Hymes 1972:106 referring to Cole 19698). I believe that the distinct status of inclusive in Aymara is stipulated by the presence of the person spoken to in its referential set. This person, as I will show in the next section, forms a kind of a pivot in the interpersonal communication of Ay- maras.

7 The reverse is, however, not true. Languages possessing inclusive /exclusive opposition do not neces- sarily imply grammatical salience of the 2nd person. 8 His thesis remained unavailable for me. AYMARA PERSON HIERARCHY 207

In sum, all these features (Addressee salience in the person markers, the use of inde- pendent person pronouns and their historical development, the presence of inclu- sive/exclusive opposition and the structure of inclusive) clearly indicate the influence of the person hierarchy, particularly of the A>S>N kind, in patterning the morpho- syntactic structure of Aymara.

4. Aymara world view and tenets of interpersonal communication In the previous section I summarised the features that characterise the grammatical salience of the person spoken to. However, the special status of Addressee in Aymara is not limited to the language only. In this section, I will investigate the hypothesis that the morpho-syntactic salience of the person spoken to is initiated and determined by a certain conception of self in the Aymara culture. Following Shweder and Bourne (1984), I will call this sociocentric as opposed to egocentric. The splitting of the universe into two parts — society and self — is a phenome- non, which most likely occurs in every individual. However, the identification of these two parts, the definition of their boundaries and their interrelation can be com- pletely different. In the egocentric world (where undoubtfully the Western culture is classified), “society is imagined to have been created to serve the interests of ideal- ised ... autonomous”, while in the sociocentric world, “individual interests” are sub- ordinated “to the good of collectivity” (Shweder and Bourne 1984:190). In the socio- centric world, the value of a person depends on the position he/she occupies in the system of interperson or intergroup relationship and not on his/her personal achieve- ments and success in any given sphere of life. Thus open competition and forceful self-expression are missing in Aymara culture (Saavedra 1981:27). In such cultures, “... self-expression and self-fulfilment can find their fullest scope within the coopera- tive situation, in terms of helpfulness, of sharing, of respect for inviolability of the rights of the others” (Lee 1959:52). Saavedra observes that “Given the chance, the Aymara will generally judge a person, regardless of background, according to the individual qualities. This process is quite different from criteria of wealth, social class, education, and cultural background that the Hispanic [and other peoples of the Western culture – E.F.] use” (1981:25-26)9. Within this framework that Aymara cul- ture offers to its members, there seems to be no boundary between an individual and community, no clear-cut opposition between the individuals in the interpersonal communication. Hence in speech acts, the Speaker does not behave as a dominant person; rather s/he is always considerate of those in his/her presence, showing them respect and deference. Culturally the Aymara are very aware of who they are speaking to, always ac- knowledging the presence of another. make tentative requests and avoid command situations, i.e. situations where the Speaker is supposed to have con- trol over the actions of the other person. Lee (1959) calls this characteristic “the prin-

9 It should be mentioned that this observation is made by a Hispanic who grew up with a Hispanic atti- tude toward the Aymara and later studied Aymara people and language.

208 FILIMONOVA ciple of the inviolate integrity of the individual”, mentioning “that no personal orders can be given or taken without a violation of personal autonomy” as a corollary of this principle (1959:8). Possibly because of this tenet, all imperatives in Aymara are usu- ally softened by special ‘polite’ suffixes and particles. Aymara has a much more elaborated system of politeness than most European languages. Beside polite and courteous welcomings and farewells (address forms (lexical items and honorific forms)) and lexical politeness markers such as ‘please’, Aymara has an abundant system of derivational and inflectional suffixes and sen- tence markers that convey the presence or absence of politeness and courtesy. There- fore, unlike European languages, where politeness is of lexical and pragmatic nature, politeness in Aymara seems to be ‘primarily morphological’ (Briggs and England 1981:291). I will now give a few examples of Aymara polite morphology, a more detailed analysis, however, can be found in Briggs (1981). In general, Aymara roots are not in themselves polite or impolite but acquire overtones of politeness and impoliteness through the presence or absence of suffixes (Briggs 1981:90). For example, a verbal derivational suffix -t’a-, which usually has the semantics of a ‘momentaneous action’ also signals the courtesy of the corre- sponding verb. Thus, the verb jisk’’iña ‘to ask’ seems to have a corresponding polite form jisk’’it’aña which is then better translated as ‘to ask a specific action’. “Without -t’a- the verb has the meaning ‘to ask aimlessly, purposelessly, like a drunk” and “would typically produce a negative reaction in the person addressed” (Briggs 1981:92). The most obvious label of politeness in Aymara is a sentence marker -ya-, which can be attached to nouns or verbs. Its semantic colouring ranges from “mere courtesy to urging to emphasis” and frequently occurs with imperatives, where it is used as a softener or attenuator. Here are some examples demonstrating the usage of the -ya- suffix with imperatives:

Mother addressing an adult son:

(9) t’aqtara-pi-k-ita-ya look-be.sure.to-just-IMP.2=>1-POL ‘Please just look for it for me’ (concerning the hoe she had previously given to him but needs back) (Briggs 1981:102)

(10) juma-ki-y amuy-t’a-m 2-just-POL look.after-POL-IMP.2=>3 ‘You look after it’ (HVY 1974:417)

An adult or older sibling urging a child to do something:

AYMARA PERSON HIERARCHY 209

(11) iraqqasi-ma-y take-IMP.2=>3-POL ‘Take it!’ (Briggs 1981:104)

(12) jala-ma-y, jala-ma-y run-IMP.2=>310-POL run-IMP.2=>3-POL ‘Run, run’ (Briggs 1981:104)

(13) anaki-ma-y herd-IMP.2=>3-POL ‘You go on and herd’ (Briggs 1981:104)

Different polite suffixes in Aymara are easily combined with each other, expressing various shades of courtesy. Consider the following examples, 14a and b, where more than one polite suffix are used:

(14) a al- t’a- s- ita- y, kasiru buy-POL-REFL-IMP.2=>1-POL sir.POL ‘Please buy from me’ (because it’s the end of the day and there is still a lot to sell) (Briggs 1981:92)

b al-t’a-si-way-ita-y, kasiru buy-POL-REFL-POL-IMP.2=>1-POL sir.POL ‘Please, buy from me, sir’ (Briggs 1981:92)

An extensive usage of polite suffixes in Aymara also seems to penetrate into the speech of the native Spanish speakers in La Paz. Laprade (1981) mentions a great frequency of the particles pues, nomás, siempre, and pero in La Paz Spanish. Unlike standard Spanish, these particles often have the function of softener or attenuator for the preceding phrase. Interestingly, the morphological markedness of politeness in Aymara is usually accompanied by corresponding intonational contours. “In Aymara a subdued, almost whining intonation connotes courtesy in persuasion or in requesting a favour” (Saavedra 1981:27). These contours seem to be an integral part of expressing cour- tesy in Aymara and serves to avoid brusqueness. The Aymara specialists perpetually emphasise the necessity to consider correct address forms, politeness suffixes and intonational contours when teaching Aymara to . Not to be addressed in polite forms is interpreted by Aymara as a nega- tion of the presence of the human being, since only animals can be addressed without

10 According to the Aymara linguistic tradition, all verbs in Aymara are considered to be transitive and ‘intransitive verbs’ (in an Indo-European sense) imply a zero 3rd person complement (HVY 1974:210). Therefore in (12) and (13), where ‘intransitive’ verbs are used, -ma- is glossed as a morpheme denoting personal interaction.

210 FILIMONOVA the polite forms (Saavedra 1981:26). Otherwise the Aymara feel insulted, something the Hispanics seldom realise. Instead they misinterpret the sullen reaction of the Ay- mara as unfriendliness or unwillingness to communicate. In sum, the worldview of the Aymara is characterised by a special concept of ego, according to which and an individual, contrary to the Western conception, is de- emphasised and not contrasted to the other selves. Hence the centre of awareness is shifted from an individual to those in his/her presence. For interpersonal communica- tion this means that the role of Speaker becomes less authorised, relinquishing the foreground to the other SAP, i.e. person spoken to. This perception of world structure is presumably reflected in the language, which evolves particular grammatical con- straints in response to the specific demands of the society in which it is used.

5. Conclusion In this paper, I first argued that irregularities of person marking in Aymara are influ- enced by a special configuration of the person hierarchy, according to which the per- son spoken to overranks the Speaker: A>S>N. Secondly, I suggested that this layout of the person hierarchy is determined by a special conception of ego in the Aymara culture which, unlike Western perception, does not consider an individual to be the centre of the universe and hence shifts the emphasis from the speaker to those in his/her presence. Support for this hypothesis comes from other language families, which also demonstrate grammatical salience of the person spoken to (e.g. Algonquian, Sioux). As I have shown elsewhere (Filimonova 2000), anthropological data (descriptions of philosophy of life and everyday communication) which characterise these language societies also confirm the suggestion that the underlying conception of ego may in- fluence the language and shape its person marking. In addition, these tentative results clearly reveal the hidden potential of cooperative work between language specialists and anthropologists. Linguists can learn much in trying to describe the grammars of non-native languages within the framework of the corresponding logic rather than imposing their own, as if it were universal. Finally, independent of whether the link between the development of morpho- syntax and the culture is to be considered as casual or spurial, an important typologi- cal implication of the findings discussed in this paper is the fact that one-to-one cor- respondence between the referents (i.e. S, A, and N) and their morphosyntactic labels (i.e. 1st, 2nd, and 3rd persons) is not universal. In languages in which the morpho- syntactic structure provides examples of the salience of the 2nd person forms, this person actually appears to be the 1st person. Using the terms of the referential cate- gories (the Addressee is more highly ranked than the Speaker, or the Speaker is set lower than the Addressee) allows this disambiguity to be avoided.

AYMARA PERSON HIERARCHY 211

References Berlin, Brent and Paul Kay 1969 Basic Color Terms: Their Universality and Evolution, Berkeley: University of California Press. Briggs, Lucy 1981 ‘Politeness in and culture’, in: M. J. Hardman (ed.), pp. 90- 113. Briggs, Lucy and Nora C. England 1981 ‘Linguistics and Foreign Aid’, in: M. J. Hardman (ed.), pp. 282-294. Cole, John Tafel 1969 The Human Soul in the Aymara Culture of Pumasara: An Ethnographic Study in the Light of George Herbert Mead and Martin Buber. PhD thesis, Univer- sity of Pennsylvania. Corbett, Greville G. 1979 ‘The agreement hierarchy’, Journal of Linguistics 15:203-224. Filimonova, Elena 1999 Universal’nye anomalii licnyx mestoimenij [Universal Anomalies of Personal Pronouns]. Doctoral dissertation, State University. 2000 ‘Nature, Culture, and Compromise: Person Hierarchy as a Cognitive Parame- ter’, Paper held at the International Conference on Cognitive Typology, Ant- werpen, April 12-14. Foley, William A. 1980 ‘Toward a universal typology of the noun phrase’, Studies in Language 4:169- 199. Hardman, Martha J. 1966 Jaqaru: Outline of Phonological and Morphological Structure, The Hague: Mouton. 1978a ‘Linguistic postulates and applied anthropological linguistics’, in: V. Honsa & M. J. Hardman-de-Bautista (eds.), Papers on Linguistics and Child Language, Ruth Hirsch Weir Memorial Volume, The Hague: Mouton, pp.117-136. 1978b ‘Proto-Jaqi: Reconstuccion del sistema de personas grammaticales’, Revista del Museo Nacional [Lima] de la cultura Peruana 41:433-456. 1982 ‘The Mutual Influence of Spanish and the Andean Languages’, Word 33:143- 157. 1988/89 ‘Fuente de datos y persona gramatical en las lenguas Jaqi’, Dialogo Andino 7/8:123-133. Hardman-de-Bautista, Martha, Juana Vásquez, and Juan de Dios Yapita Moya 1974 Outline of Aymara Phonological and Grammatical Structure, Gainesville, Fla.: Aymara Language Materials Project, Department of Anthropology, Uni- versity of Florida.

212 FILIMONOVA

Hardman, Martha J. (ed.) 1981 The Aymara Language in Its Social and Cultural Context, Gainesville: Uni- versity Presses of Florida. Heath, Jeffrey 1991 ‘Pragmatic Disguise in Pronominal- Paradigms’, in: Frans Plank (ed.) Paradigms: The Economy of Inflection, Berlin: Mouton de Gruyter, pp. 76-89. 1998 ‘Pragmatic Skewing in 1Ù2 Pronominal Combinations in Native American Languages’, IJAL 64:83-104. Hymes, Dell 1972 ‘On personal pronouns: ‘Fourth’ person and phonesthematic aspects’, in: M. Estellie Smith (ed.), Studies in Linguistics in Honor of George L. Trager, The Hague: Mouton, pp.100-121. Keenan, Edward L. and Bernard Comrie 1977 [1972] ‘Noun phrase accessibility and universal grammar’ [Paper read at LSA Annual Meeting, Atlanta], Linguistic Inquiry 8:63-99. Laprade, Richard A. 1981 ‘Some Cases of Aymara Influence on La Paz Spanish’, in: M. J. Hardman (ed.), pp. 18-29. Lee, Dorothy 1959 Freedom and Culture, Englewood Cliffs: Prentice-Hall, Inc. Moravcsik, Edith A. 1978 ‘On the distribution of ergative and accusative patterns’, Lingua 45:233-279. McKay, James Tuell 1985 Language, Structure, Worldview and Culture Contact: Understanding Aymara Culture and History in a Bolivian Context, M.A. thesis. University of Florida, Gainesville. Plank, Frans 1985 ‘Die Ordnung der Personen’, Folia Linguistica 19:111-176. Saavedra, Carlos 1981 ‘Social and Cultural Context of the Aymara in Bolivia Today’, in: M. J. Hardman (ed.), pp. 207-227. Silverstein, Michael 1976 ‘Hierarchy of features and ergativity’, in: R. M. W. Dixon (ed.) Grammatical Categories in Australian Languages, Canberra: Australian Institute of Abo- riginal Studies, pp. 112-171. Smith-Stark, T. Cedric 1974 ‘The plurality split’, Chicago Linguistic Society 10:657-671. Shweder, Richard A. and Edmund J. Bourne 1984 ‘Does the Concept of the Person Vary Cross-Culturally?’, in: R. A. Shweder & R. A. LeVine (eds.), Culture Theory: Essays on Mind, Self, and Emotion, Cambridge: Cambridge University Press, pp. 158-199. AYMARA PERSON HIERARCHY 213

Weber, David J. 1986 ‘Huallaga Quechua Pronouns’, in: U. Wiesemann (ed.), Pronominal Systems, Tübingen: Narr, pp. 333-349. 1989 A Grammar of Huallaga (Huánuco) Quechua, Berkeley: University of - fornia Press.

Abbreviations BEN Benefactive FUT Future HON Honorific IMP Imperative INCL Inclusive NEAR Nearative POL Polite REFL Reflexive SENT Sentential marker SG Singular VBEN Verbal Benefactive 1 1st person 2 2nd person 3 3rd person

NOMINAL CLASSIFICATION IN MOVIMA

Colette Grinevald Université Lumière, Lyon 2

1. Introduction Nominal classification systems appear to be a common feature of South American lowland languages (Aikhenvald 2000, Payne 1987, Derbyshire and Payne 1990), and their analysis is often presented as a challenge for the typologies of nominal classifi- cation systems developed so far (Aikhenvald 1999, Allan 1977, Craig 1992, 1994, 1995, Dixon 1986, Grinevald 2000, to appear a, for instance). From a few systems that have been described in some detail (Aikhenvald 1994, 2000, Barnes 1990, Go- mez-Imbert 1982, Klein 1979, Payne 1986, Vidal 1995), it appears that many nomi- nal classification systems from this region of the world do not match very closely the types found elsewhere in the world (Asia, Africa, Australia, the Pacific and North America). One of their major characteristics is that they seem to overlap with two major types of systems: classifier systems on the one hand and noun class/gender systems on the other. It is in this context that any encounter with a system of nominal classification from South America constitutes a welcome opportunity to continue exploring our understanding of the parameters of variation of that linguistic phe- nomenon in the languages of the world. This paper is the result of the very brief but intense encounter between a linguist interested in the phenomenon of nominal classification and an Amazonian language that turned out to have an instance of such a phenomenon. The encounter took place in the fall of 1995, in the midst of a sweeping government sponsored alphabet project that aimed at “regularizing” writing systems for the benefit of future bilingual- bicultural education programs. The language in question was Movima, a little known and genetically unclassified language of the eastern lowland of Bolivia which is spo- ken today by a few thousand, mostly in the town of Santa Ana of the Province of Beni, and along the Yacuma, Matos and Apere rivers (Plaza and Carvajal 1985: 15). The linguist was then called Colette Craig and in the midst of developing a typology of the various types of classifier systems (Craig 1986b, 1992, 1994). She is now called Grinevald and has widened the scope of the study to the issue of nominal clas- sification systems in general, in great part due to more familiarity in recent years with the Amazonian descriptive challenge (Grinevald 2000, to appear a, Grinevald and Creissels 2000, Grinevald, Creissels and Seifart 2001). This brief sketch of the Movima nominal classification system aims at making three points. First, that this particular system might well be of the kind that is typi- cally found in that region of the world and, as such, challenges the morphosyntactic typologies of nominal classification systems conceived so far on the basis of data 216 GRINEVALD from other parts of the world. For that reason it clearly deserves further study. Sec- ond, that it is of particular interest because of its nature as a mixed system that com- bines phonologically and semantically based patterns of classification. And third, that the evolution of the various accounts of this system over the last decades can be used to illustrate the relation of linguistic description to the development of theoretical and typological frameworks. In this case, the analysis has evolved from an initial account by missionaries in the sixties to a partially erroneous analysis of it as a probable in- stance of a numeral classifier system by the present linguist. While the missionaries identified the existence of the phenomenon with descriptive limitations due to the vacuum of a theoretico-typologial context for interpreting it properly, the linguist was blinded by her preoccupation and interest limited at that moment to classifier sys- tems. This paper, however, provides a revised analysis of the system, considering it finally as a probable noun class system, in the context of on-going work on the vari- ety of Amazonian systems of nominal classification. The first part of this paper, Section 2, will therefore consist in the presentation of the information that was available, at the time of the 1995 field encounter, from the writings of Judy and Judy, the SIL missionaries who were faced with this system before the issue of nominal classification took shape in the world of general linguis- tics. Section 3, the second, and most extensive part of this paper, will present an up- dated analysis of this nominal classification system which is based on the data gath- ered for that purpose in 1995. This update will focus at this point on the description of a particular aspect of the system that had been seemingly missed by Judy and Judy: the coexistence of phonologically productive patterns of formation of classifi- catory morphemes and more familiar semantically based ones. Section 4, the third and final part, will then consider the issue of how and why this Movima classification system was first reframed as a numeral classifier system, and only later identified in fact as the noun class system it most likely is. The present description is therefore mostly meant to be an invitation to pursue the analysis of this system, for its own sake and for the sake of contributing to the ongoing debates about the nature of such nominal classification systems in that part of the world.

2. Early mention and initial re-interpretation of the system 2.1. The data in Judy and Judy (1962a) Some mention of the existence of a nominal classification system in Movima is to be found in the writings of Robert Judy and Judith Judy, of the Summer Institute of Lin- guistics, who studied the language in the early sixties and produced a word list (Judy and Judy 1962a), a pamphlet on the Movima phonemes (Judy and Judy 1962b) and a tagmemic grammar (Judy and Judy 1967).

2.1.1. “Descriptive-object pronouns” The existence of classifying morphemes is mentioned in the appendix of the 1962 Movima-Spanish Vocabulary entitled Notes on Movima Grammar, on pages 150 and MOVIMA NOMINAL CLASSIFICATION 217

151. At the time, Judy and Judy labeled these morphemes “pronombres objetos- descriptivos”, the translation of which is not obvious, being it literally something like “descriptive-object pronouns”. By way of introduction, Judy and Judy first state that “certain objects that are sufficiently clear in the context are referred to only by means of the descriptive- and appropriate object pronoun (sic). This [pronoun] is suffixed to a descriptive word or a noun (sic)”.

2.1.2. Inventory of these “pronouns” Since their presentation of this phenomenon is brief, and since the publication is not readily available, their entire account of such pronouns will be reproduced below. The list of some of these “pronouns”, with the objects they supposedly classify can be found on page 150 given below in (1) as it appears in their writings (in translation here and with new numbering, for ease of reference later in this paper):

(1) objects pronouns

1.1 round fruit, egg, nest, jug, etc. -ba 1.2 skin, paper, book, etc. -ben 1.3 seed, grain, star, viper, river, house, etc. -di 1.4 plantains, bird, insect, airplane, basket, tropical forest, etc. -mo 1.5 yuca, totora, etc. -pa 1.6 tooth, spoon, point, etc. -ła

A more extensive list of another 23 “other objects with their pronouns” is also given on page 151, reproduced here in the alphabetic order of the pronoun forms in which they appeared originally:

(2) ‘cane’ -as, ‘piece of pottery’ -bij, ‘plant’ -bo, ‘mud’ -bun, ‘toasted manioc’ -cho, ‘person’ -e, ‘location’ -huaj, ‘milk’ -lo, ‘thorn’ -lej, ‘straw’ -mas, ‘water’ -mi, ‘shell’ -moł, ‘wind’ -muj, ‘feather’ -mun’, ‘bundle’ -pi, ‘ani- mal’ -poi, ‘clothing’ -oj, ‘hat’ -to, ‘dust like’ -vas, ‘hoof’ -ve, ‘wood’ -vos, ‘ashes’ -vus, ‘arm’ -yin’, etc.

2.1.3. Their use Little is said about the use of such forms, although some information can be gleaned from direct or indirect statements. For instance, the second list given in (2) is pre- ceded by the following information (reproduced as is, the numbering being mine again, for later ease of reference):

218 GRINEVALD

(3) “examples of their use”

3.1 sota’ba (sota’ra ‘one’) 3.2 choesba (choesni ‘ugly’) 3.3 daujesben’ (daujes ‘deer’) 3.4 son’di (sonra ‘other’) 3.5 oidi (oira ‘two’) 3.6 mere’mo (mere’e ‘big’) 3.7 beumo (beuni ‘mature’) 3.8 tochi’pa (tochi’i ‘small’) 3.9 pacoła (paco ‘dog’)

No morphological analysis is provided, but it is at least clear from (3) that the phe- nomenon under question is a matter of substitutable or added last syllable of the word, and that it affects a number of lexical categories, including nouns (dog, deer), adjectives (ugly, big, mature, small), and number (one, two), as well as the word ‘other’ which is unspecified here as to its status as adjective or pronoun. Besides these examples, additional information about the use of these forms is contained in the following final statement: “these pronouns are also suffixed to the stem of verbs and identify the object of the verb”. This statement is then followed by the examples given in (4) below:

(4) a with the verb riłna ‘to soften’ riłaba ‘to soften a jug’ riła'oj ‘to iron’

b with the verb onarana ‘to know’ onamona ‘to know the bird’ onahuajna ‘to know the place’ onapoina ‘to know the animal’

Again, no morphological analysis or glossing was provided, but, at least with the verb onarana ‘to know’, one can recognize two of the three morphemes under scru- tiny: -huaj- associated with ‘location’, and -poi- associated with ‘animal’, already mentioned in the list in (2) above.

2.1.4 Conclusion Thus, in these original two pages of notes on the “descriptive-object pronouns” of Movima, Judy and Judy provided enough information to establish that Movima had a nominal classification system. It is on the basis of these intriguing few pages that specific elicitation was carried out in 1995 in order to confirm its existence and to clarify its status. MOVIMA NOMINAL CLASSIFICATION 219

2.2. The initial re-interpretation of the system as a ‘numeral classifier’ system (1995) Judy and Judy did not actually identify these “pronouns” as classifiers per se; this was, after all, 1962, and not much had been written on classifiers in the general lin- guistic literature by then; the first publications on the topic to be widely read not ap- pearing until a good decade after they had published their studies (such as Adams and Conklin 1973 and Allan 1977). The initial analysis done by the present author in 1995, before initiating fieldwork on the system, focused on the fact that it had much in common with numeral classifiers. Arguments that could be advanced to support this analysis were both of a morphosyntactic and semantic nature.

2.2.1. Morphosyntactic argument The morphosyntactic argument was basically to be found in the examples of such morphemes with numerals, as given in (3.1) ‘one’ and (3.5.) ‘two’ above, which are repeated below:

(5) 3.1 sota’ba (sota’ra ‘one’) 3.5 oidi (oira ‘two’)

2.2.2. Semantic arguments The semantic argument consisted in noticing that, on the basis of the initial list of apparent ‘classes’ of nouns given in (1) above, one could find similarities between the semantics of some identifiable Movima classes and the semantics of some of the classes commonly found in numeral classifier systems (see for instance the study of the semantics of classifiers in Adams and Conklin 1973, Adams 1986, and Allan 1977). The Movima classes given by Judy and Judy that seemed to clearly exhibit some semantic motivation and echo semantically motivated classes of numeral classi- fier systems are shown in (6). The numbering assigned is the same as the one given in (1) above, and the possible numeral classifier semantics are mentioned below the Movima instances:

(6) semantically motivated classes (1.1) round fruit, egg, nest and jug, could be a class of round objects, (1.2) skin, paper, book, could be the class of ‘flat and flexible’ objects, also commonly found in numeral classifier systems, (1.5) yuca and totora, could, maybe, be a plant class, (1.6) tooth, spoon, point, could be taken to be a class of ‘pointed objects’

220 GRINEVALD

For some of the classifying elements, the semantic motivation was therefore fairly straightforward and familiar (round, flat and flexible, long and pointed, plant). There were also classes which were more mixed, with no immediately obvious semantic motivation, but that could also be expected of a numeral classifier system. These semantically heterogeneous classes are given in (7), still with the numbering of the original (1) of this text:

(7) semantically heterogeneous classes (1.4) plantains, birds, insects, baskets and tropical forest, (1.3) seed/grain/star, but also viper, river, and house

At first glance these supposed classes seemed reminiscent of some famous heteroge- neous classes discussed in the literature on numeral classifiers. One such case is the case of a numeral classifier of Thai, tua (discussed in Carpenter 1986), which is said to have originally been used for ladle, spoon, fork, umbrella, for their parts that had a long and rigid shape, then to have come to also be used with rickshaw, also with long handles, but then with bicycle, plus bus and car, as an amalgamate of transportation means, and also with string instruments, at which point it started not applying any more to the original long and rigid objects. The most famous case of semantic het- erogeneity of a classification system is probably the case of Dyirbal (Australia) which was originally described by Dixon (1972) but later used as a case study of a particular type of semantic analysis of prototypes by Lakoff (1986, 1987). The title of Lakoff’s 1987 book Women, fire and dangerous things refers in fact to the apparent heterogeneity of a class of nouns taking the same class marker. It is worth noting in passing that the Dyirbal system is actually a system of noun classes and not of nu- meral classifiers. Lakoff (1986) contains also another case study of a very heterogeneous class, this time a real instance of numeral classifier: it is the class headed by the Japanese numeral classifier hon said to be used to count, at least, the following: sticks, canes, pencils, candles, trees etc; dead snakes and dried fish; martial arts contests; hits in baseball; shots in basketball…; judo matches; rolls of tapes, telephone calls; radio and TV programs; letters; movies; medical injections (Lakoff 1986: 25-26). This Japanese data was used by Lakoff to discuss the various means of extension of classes and to justify, on an a posteriori basis, the inclusion of those items in the same class. It is in this context, therefore, that the existence, in the Movima system, of ap- parently semantically heterogeneous classes did not appear particularly troubling. It seemed to call for the same kind of a posteriori reasoning to motivate, at least par- tially, the regroupings found in the heterogeneous classes.

MOVIMA NOMINAL CLASSIFICATION 221

2.2.3. ‘Unique classifiers’ and large systems The initial analysis of the system as a numeral classifier system could also possibly handle the list of the 23 “other objects and their pronouns” given in (2) above. The list seemed to make two claims: first that there were at least apparently 23 instances of so-called “descriptive-object pronouns” which applied to only one object, and, second, that the list was actually open, as indicated by the final “etc. ...”. Both char- acteristics found echo in what was known of some of the better known and best de- scribed numeral classifier systems, such as those of South East Asia. Those numeral classifier systems are known to include several types of classifi- ers, according to their level of generality or specificity (see Craig 1994, Grinevald to appear a). The most common type of classifiers are ‘specific classifiers’ which head more or less homogeneous classes. They may come in fairly large numbers (dozens if not hundreds). At the same time, it is common in large(r) systems of numeral classi- fiers, again such as those of Asian languages for instance, to have ‘general classifiers’ which may substitute the specific classifiers and are devoid of much semantic moti- vation (and could be translated as ‘thing’). Those systems also commonly include yet another type of classifiers, called ‘unique classifiers’ in that they head classes of only one item. These unique classifiers are usually idiosyncratic to the language and are taken to be indicative of the cultural importance of the elements they classify (some animals for instance, like the elephant or the tiger in some systems, see Craig 1986b for the case of the unique classifiers of Jakaltek-Maya noun classifiers). As for the mention of “etc. ...” at the end of the list, two interpretations were possible: either that this was just a sample of a larger but closed number of such mor- phemes, or that the list was open-ended. But such an “etc. …” at the end of the list was also not uncharacteristic of classifier systems: some such systems have been shown to have in the hundreds of classifiers, particularly when one takes into account all the registers of the language, from common language to most formal registers (see Berlin 1965 for Tzeltal, and Erbaugh 1986 for Chinese, for instance). Furthermore, large classifier systems often exhibit the phenomenon of ‘repeaters’, lexical nouns that occupy the classifier slot and start functioning as classifiers, leaving the door apparently open to create more classifiers.

2.2.4. Conclusion It is therefore on the basis of a preliminary re-analysis of the Judy and Judy material of the Movima system as a probable ‘numeral classifier’ system that the sessions of direct elicitation were carried out in 1995. Such analysis seemed viable at the time on both morphosyntactic and semantic grounds, as well as on the basis of general char- acteristics of such systems known then from studies from other parts of the world.

3. New data on the supposed ‘numeral classifiers’ of Movima. What follows is a reconstitution of the analysis carried out during and immediately following the fieldwork sessions led by the present author. The direct elicitation took 222 GRINEVALD place in the town of Trinidad (Department of Beni), in October 1995, with two speakers of Movima: Gilberto Machado Vega and Peregrina Cayu Mazaro. As ex- plained above, the elicitation carried out assumed at that time an analysis of the sys- tem as a system of ‘numeral classifiers’ and consequently concentrated on gathering data aimed at defining the specifics of that type of system for the Movima language.

3.1. Confirmation of the existence of supposedly ‘general’ and ‘specific’ ‘numeral classifiers’ Initial elicitation of numbers showed that Movima numerals are composed of a root and a suffix. The suffix -ra occurs when the numeral is used in simple enumeration and for general counting, as illustrated in (8) below:

(8) sota’-ra ‘one’ oy-ra ‘two’ tas-ra ‘three’ oyka-ra ‘four’

Judy and Judy had glossed -ra as ‘thing’, a gloss appropriate for a ‘general’, neutral classifier. The native counting system only goes up to the number ‘four’, after which Spanish loanwords are used, with the same system of suffixation:

(9) sinko-ra ‘five’, from Spanish ‘cinco’ seis-ra ‘six’, from Spanish ‘seis’ etc...

However, when objects are being counted, the most common construction is one in which the number takes any one of a number of other suffixes besides -ra. The same numeral therefore can appear with various suffixes, depending on the nature of the objects being counted, as illustrated below:

(10) a sota’-mo ‘one (said of banana)’ sota’-di ‘one (said of plate)’ b oy-b’a ‘two (said of oranges)’ oy–poy ‘two (said of dogs)’

The numeral suffixes did in fact behave in a ‘pronominal’ fashion, as the “pronom- bres objetos-descriptivos” mentioned by Judy and Judy, to the extent that the lexical noun corresponding to the object was indeed not given, even in direct elicitation. Those suffixes of numerals can actually also occur on adjectives, another fact that they seem to share with many numeral classifier systems in other parts of the world. The data on the adjectival use of classifiers in Movima remains very limited to MOVIMA NOMINAL CLASSIFICATION 223 date, although it is suggested, as illustrated by the variation in the suffix of the adjec- tive for ‘big’ given below:

(11) mere’-mo ‘big (said of a basket)’ mere’-roy ‘big (said of a house)’

Having confirmed the existence of the classifying elements mentioned by Judy and Judy, the novelty of the analysis of the data newly gathered consisted in the reanaly- sis of the apparently very heterogeneous classes reproduced from Judy and Judy in (1) above.

3.2. A semantic puzzle, with a phonological solution What follows is a reanalysis of the Movima data presented by Judy and Judy that shows that the apparently heterogeneous ‘classes’ of Movima suffixes are not actu- ally semantic classes at all but cases of homophonous suffixes resulting from the phonological truncation of source nouns. For instance, it is not the case that there is a class of nouns regrouped by their sharing the classifying suffix -d’i which would in- clude such disparate objects as grain/seeds and stars, as well as houses and snakes. Rather, it is a phonological accident that all those nouns end with the same syllable, which is the one preserved with the truncation rule that produces the numeral suf- fixes. What happens in Movima is that multiple cases of homophonous suffixes are due to the fact that if the simple syllable structure of Movima words (which, admit- tedly, remains to be studied in greater detail), is coupled with the fairly small phone- mic inventory of the language, the result is that many words in the language share similar last syllables. The apparent regrouping of such disparate words as grain, house and snake, is therefore an accident of morphology, due to the fact that all these nouns end in -d’i.

3.2.1. Monosyllabic suffixes The usual pattern of formation of the classifying suffix under study here is the trunca- tion of the source noun that leaves the last syllable of the noun to be suffixed. The examples below illustrate this principle of truncation.

(12) SUFFIX from NOUN -b’e hub’e ‘typical small river embarkation’ -d’o chad’o ‘plate’ -’o chon’o ‘tipoy’ -pi sukapi ‘belt’

The syllable retained for suffixation seems to be that of the root of the noun, although it may be that of a more complex noun stem. Much more work needs to be done on 224 GRINEVALD the noun morphology of Movima to establish the parameters of this truncation phe- nomenon, in view of probable extensive noun derivation and noun composition proc- esses in the language. For instance, the limited data gathered include an example of the truncation of the last syllable of an apparently fairly long word, which is probably the result of some composition process:

(13) -to chorankwanto ‘hat’ which contrasts with the fact that the truncation does not affect the last syllables of all the ‘words’, as the examples of the following words ending in -kwa show:

(14) SUFFIX from NOUN –moł momołkwa ‘shell’ –mun’ mumun’kwa ‘feather’

As it turns out -kwa is a fairly productive suffix in the language. Judy and Judy had identified it as a “partitive” suffix, although by its semantics and suffixal nature. It is reminiscent of a semantic class marker for round objects of the kind often found in noun classification systems in other parts of the world. Judy and Judy’s list of noun affixes (1967: 395) gives the following instances of -kwa:

(15) -kwa ‘partitive’ and/or ‘round’ class? chiraskwa tripes, intestines bałewkwa tail mo’mołkwa shell (of turtle) mumun’kwa feather tovenkwa shell (of fruits) bubunkwa mud hołkwa egg rułkwa tongue hułpakwa arrow barinkwa grinding handle d’ontokwa wild cat b’ab’akwa fruit b’akwa head woro’kwa neck d’id’inkwa seed, star lorankwa leaf dudułkwa root itilakwa man tobe’kwakwa skin risakwakwa hair MOVIMA NOMINAL CLASSIFICATION 225

The truncation process was not systematically checked on those nouns in the 1995 fieldwork sessions. This morpheme will deserve more attention. An exceptional form was collected, consisting of a disyllabic suffix from a word supposedly ending in the suffix kwa: -lakwa from itilakwa ‘man’, for instance. The same noun truncation principle accounts for the suffixation found on adjec- tives which was mentioned before:

(16) tun-to ‘black (said of hat)’, from choramkwanto ‘hat’

The origin of the suffix that Judy and Judy called “objective-descriptive pronouns” first reproduced in (1) and (2) above, and later considered as ‘numeral classifiers’ is therefore the truncated final syllable of a source noun.

3.2.2. Confirming the truncation analysis with loanwords Support for this phonological truncation analysis of the “descriptive pronouns” of Judy and Judy can be found in the treatment of loanwords. In the case of loanwords, the truncation process is somewhat different in that it results in disyllabic rather than monosyllabic suffixes, but this apparent exception to the rule of final syllable trunca- tion is actually taken to be a confirmation of the existence of a truncation process in the first place:

(17) DISYLLABIC SUFFIX from LOANWORD from Spanish -misa kamisa camisa ‘shirt’ -pato sapato zapato ‘shoe’ -reta kareta carreta ‘cart’

The analysis that, in the case of loanwords, the corresponding suffixes on numbers and adjectives must be two-syllable long is further confirmed by the situation that arises with simpler originally disyllabic loanwords. In this case, the disyllabic suffix is formed by reduplicating the last syllable, as illustrated below:

(18) CLASSIFIER from LOANWORD from Spanish -sasa mesa mesa ‘table’ -yaya siya silla ‘chair’

The choice of the process in those cases seems, furthermore, to strengthen the analysis of the truncation of the final syllable of native nouns. The indirect support for this analysis consists in the fact that the reduplication of the last syllable of such loanwords preserves the principle of truncation. Simply using the disyllabic loanword would fulfill the requirement of a disyllabic suffixation but not that of a truncation for the purpose of suffixation. 226 GRINEVALD

Counting shirts and chairs, which are both nouns borrowed from Spanish, is therefore done in the following way:

(19) a oy-misa ‘two shirts’ from kamisa (Spanish camisa) 2-SUF a’ *oy-sa 2-SUF b oy-yaya ‘two chairs’ from siya (Spanish ‘silla’) 2-SUF b’ *oy-ya. 2-SUF

This constraint on the length of the numeral suffix for loanwords could be seen as an example of a functional principle in the spirit of some of T. Givón’s proposals (1979, 2001), that longer forms are for less familiar material and shorter forms for more pre- dictable, or known material. The issue of loanwords is an interesting one in Movima, as there exist diverse combinations of loanwords and native Movima numerals and names of counted ob- jects, in which either the number or the number suffix (SUF) is (from) a loanword, in the following patterns:

(20) a Number loanword + SUF from Movima native noun sinko-mo ‘five (said of baskets, for instance)’ five-SUF

sinko-di ‘five (said of plates, for instance)’ five-SUF

b Number loanword + SUF from Spanish loanword noun sinko-reta ‘five (said of carts)’ five-SUF (from Spanish cinco carretas)

c Movima adjective + SUF from Spanish loanword noun baschim-pato ‘unmatched (said of shoe)’ unpaired-SUF (from Spanish zapato ‘shoe’)

The truncation principle that is at the origin of the Movima suffixes found on numer- als and other elements explains therefore the fact that several nouns would be repre- sented by similar looking suffixes because of their phonological composition, more specifically the similarity of their final syllables. In addition, this truncation principle accounts for the apparent ‘open–endedness’ of the system hinted at by Judy and Judy, since it appears to be very productive, with specific accommodations for loanwords. MOVIMA NOMINAL CLASSIFICATION 227

Finally, this truncation principle can explain why the vast majority of the classifying suffixes of Movima look like instances of ‘unique classifiers’, i.e. classification markers heading classes that are made of only one element. However it is worth noting here that, within a numeral classifier system perspec- tive, the presence of unique classifiers is fairly common. Those unique classifiers are usually in a small minority in the totality of the inventory of classifiers. This is not the situation in Movima where the apparent proliferation of what would be consid- ered unique classifiers raises the question of their actual function in the language. This and other elements of the initial analysis of these Movima suffixes as ‘numeral classifiers’ will be reconsidered later in Section 4, where the system will be argued to be a noun class system rather than a numeral classifier system, although a noun class system of the Amazonian kind. Although the truncation principle accounts for the majority of the cases of suf- fixation on numerals (and other elements as seen below), a complete description of the functioning of the counting system requires accounting for the few semantic classes that also exist in the Movima system.

3.2.3. Semantically based classification too As mentioned earlier, it is not the case that there is no instance of semantic classifica- tion in the Movima counting system. There are indeed a few semantically based ‘classes’ of nouns which manifest themselves, in the counting process for instance, and which were readily identified by the Movima speakers which took part in this study. Some of the Movima numeral suffixes do function like prototypical numeral classifiers in that they classify nouns into identifiable semantic classes, with seman- tics of the kind commonly found in numeral classifier systems. Four suffixes were readily identified as marking such semantic classes by the Movima speakers inter- viewed in 1995:

(21) -poy four-legged animals -mo two-legged animals -b’a fruits -b’o plants

The suffix -poy is used for counting animals, of the four-legged type. It applies to animals such as:

(22) pakonana’a fox zorro rulrul jaguar tigre d’ontokwa puma león b’abama giant anteater oso bandera waewae anteater oso hormiguero 228 GRINEVALD

lume’e paca jochi pintado punsi brown agouti jochi colorado si’do monkey mono d’ud’utus porcupine puerco espino saray gray brocket deer urina oma tapir anta yonali crocodile caimán karan alligator lagarto

The semantic class suffix -poy comes in fact, by truncation, from the generic noun popoykwa meaning ‘animal’, itself a complex word [po–poy–kwa] made of a redu- plication (as is the case for a number of animal names in that language and in many other South American languages), and the ‘partitive’suffix -kwa found in many words and already mentioned. The other animal suffix is -mo. It classifies two-legged animals, such as birds and fowls, of all sizes:

(23) toroware rooster gallo mataware hen gallina kokore duck pato d’anra piyu chimili hummingbird picaflor wonowo’o toucan tucán letos seagull gaviota toba heron garza worokoko owl buho

No source was identified for this biped animal class marker. Another clearly semantic class is that of fruits, marked by the class suffix -b’a. This class includes native nouns and loanwords from Spanish:

(24) maropa ‘papaya’ pa’di ‘guayaba nara’sa ‘orange’ sewoya ‘onion’ manka ‘mango’

Here again, a noun source can be found for the class suffix. -b’a is the result of the truncation of the generic noun for fruit: b’ab’akwa, which exhibits the same complex morphological structure as the source of the marker for four-legged animal (redupli- cation and suffixation of the now familiar suffix -kwa).1

1 The generic noun b’ab’akwa ‘fruit’ appears to be itself derived from the noun b’akwa ‘head’ (with no MOVIMA NOMINAL CLASSIFICATION 229

Next to the class of fruits is that of plants, which is marked with the class suffix -b’o. Included in this class are nouns like:

(25) pere ‘cooking banana’ ‘plátano’ talummo ‘sweet banana’ ‘guineo’ chuła ‘manioc’ ‘yuca’ nara’sa ‘orange’ ‘naranja’ manka-b’o ‘mango tree’ ‘mango’

No noun source could be elicited for this plant class suffix. The Movima system of numeral suffixes includes therefore some semantically motivated classes, two for animals (four- and two-legged) and two for the vegetable world (fruits and plants), the largest of all being the animal classes. It is noticeable that the semantic classification does not seem to include class(es) for humans, which are generally considered to be among the most basic ones and the first to appear in numeral classifier systems.

3.2.4. Semantic extension Some processes of semantic extension of these basic classes further confirm the analysis that some of the suffixes found on numerals are semantically motivated. The data collected contained two instances of the semantic extension of the suffix for the fruit class, to include rounded objects. For example, to count bottles and glasses, as well as eggs, one uses the class suffix for fruits -b’a, presumably because of their roundish shape:

(26) a oy-b’a ‘two (of bottles)’ for boteya ‘bottle’ two-class(fruit-round) b oy-b’a ‘two (of glasses)’ for waso ‘glass’ two-class(fruit-round) c oy-b’a ‘two (of eggs)’ for hołkwa ‘egg’ two-class(fruit-round)

It is interesting to note that the semantic classification can apply to loanwords too, as shown by the numeral suffixation for (26a) boteya from ‘botella’ and (26b) waso from ‘vaso’. One could have expected disyllabic suffixes specific to loanwords of the kind mentioned above, such as -teya or -soso, but as will be considered in the next section, there are three ways of counting objects in Movima, two of which have been mentioned so far (phonological truncation/ reduplication and semantically based), and the system is presently in flux. The literature on numeral classifiers is replete with examples of the extension of the basic classes defined originally as plant parts to extended classes that include ob- reduplication but suffixation of -kwa). 230 GRINEVALD jects that share some shape characteristics of plant parts. There are three basic cases of such semantic extension. One is the extension of ‘fruit/seed’ to ‘round and com- pact’, which is exemplified above. The others are the common extensions of the lexi- cal item ‘tree/trunk’ to ‘long and rigid’ objects, and that of ‘leaf’ to ‘flat and flexible’ ones. There may exist hints in Movima of such association in that Judy and Judy (1967: 394) give the “root citation form r” as follows: -as ‘sugar-cane’; -b’a as ‘roundness’ and -ben’ as ‘flatness’. However, no examples accompany this statement, and none could be elicited from the speakers interviewed in 1995, for either the no- tion of long (and rigid) or the notion of flat (and flexible).

3.2.5. Conclusion There is therefore a clear but limited process of semantic classification indicated by numeral suffixes at least. It co-exists with an apparently productive system of forma- tion of numeral suffixes by phonological truncation of the source noun being counted. The semantic classification may in fact be secondary to the extent that it might have arisen originally from the truncation of generic names, at least for two of the semantic classes identified, the four-legged animals and the fruits, for which syn- chronic evidence could be elicited.

3.3. A classifier system in flux One of the striking features of the elicitation sessions on the Movima system was the fact that apparently simple questions on how to count simple objects of the environ- ment provoked intense discussions between the two speakers. Besides the items of the four semantic classes noted earlier for animals and plants, many items were the topic of some discussion, including some heated disagreement between the speakers. It was surprising to watch how counting even familiar objects was apparently not as straightforward a matter in Movima as it is in European languages. The system seems to be in a state of flux, counting being done in one of three ways:

a. counting with a suffix of phonological origin by way of the truncation princi- ple. As noted earlier, this gives the impression of both heterogeneous classes and numerous ‘unique’ numeral classifiers, when all it is, is widespread ho- mophony of the last syllable of nouns; b. counting with a class suffix with semantic motivation, regrouping animals and plants in certain classes, with limited extension to objects sharing similar shapes; c. counting with the general classifier in a periphrastic construction.

All three alternatives are illustrated below:

(27) a oy-d’o ‘two (of plates)’ from chad’o ‘plate’ two–class(truncation) MOVIMA NOMINAL CLASSIFICATION 231

b oy-poy ‘two (of monkeys)’ from si’do ‘monkey’ two-class (four-legged animal)

c oy-ra di’ tirinchi ‘two forks’ two-class (general) of fork

Nouns seem to fall into different categories, according to the combinations of these three constructions the speakers permitted for counting them. There were those nouns that speakers agreed could be counted only one way, and those for which more than one way was possible. The examples of (27) above appeared each to be limited to the pattern shown (truncation for ‘plate’, semantic for ‘monkey’, general periphrastic for ‘fork’). The examples of (28) below further illustrate the exclusive use of the general periphrastic construction with ‘plank’ and ‘jugs’:

(28) a oy-ra di’ chakatorawa ‘two planks’ two-class(general) of plank

a’* oy-wa ‘two (of planks)’ two-class(truncation)

b oy-ra di’ lotob’a ‘two jugs’ two-class (general) of jug

b’* oy-b’a ‘two (of jugs)’ two-class(truncation)

There were nouns that speakers agreed without hesitation could be counted either one of two ways, such as ‘cats’, ‘pigs’ and ‘chairs’, as shown below:

(29) a oy-chichi ‘two (of cats and pigs)’ two-class (truncation) from michi ‘cat’ and kochi ‘pig’

a’ oy-poy ‘two (of cats and pigs)’ two-class(four-legged animal)

b oy-yaya ‘two (of chairs)’ two-class (truncation) from siya ‘chair’

b’ oy-ra di’ siya ‘two chairs’ two-class(general) of chair

232 GRINEVALD

There were also nouns that elicited much discussion and had no obvious established way of being counted. This was the case for instance for ‘benches’, an item that pro- voked heated discussion with no resolution. Not unexpectedly, the treatment of loan- words afforded interesting insights into the dynamics of the Movima classifier sys- tem.

4. The Movima system in an Amazonian perspective: a noun class system The second part (Section 3) has presented data that had been elicited with the view that the phenomenon at work in Movima was of the numeral classifier system general type. As already acknowledged, this particular interpretation of the nature of the sys- tem stemmed from the preoccupations of the field linguist at the time with building a typology of classifier systems. This typological bias blinded her to the evidence that could be found in Judy and Judy that the system was not strictly speaking a numeral classifier system per se. The present analysis is that Movima must have a noun classi- fier system of the concordial type now commonly identified in languages of that Amazonian region of the world.

4.1. Noun class systems of the Amazonian region As discussed in Grinevald and Creissels (2000), Grinevald, Creissels and Seifart (2001), and Grinevald (to appear b), many of the nominal classification systems of the Amazonian region which have been labelled in the literature in various ways, including as classifier systems, are probably better analyzed as noun class systems. A characteristic of these Amazonian noun class systems, however, and the one that pre- vented their being identified as such earlier, is that they are much less grammatical- ized systems than the systems of Niger Congo languages which have been the basis for the analysis of prototypical noun class systems. These typically less grammatical- ized Amazonian systems are more open-ended, more semantically motivated, more subject to speaker variation, as well as to discourse context and register level. Overall they tend to feel less stable and more lexical and to be more difficult for the field workers investigating them to capture in their diversity and their complexity. Many early descriptions of such systems exude a feeling of these systems being hard to circumscribe, a feeling justified as it were by their very nature of lexico-grammatical systems. It is typical of descriptions and discussions of lowland South American nominal classification systems to note that they cut across the various types of classifier sys- tems identified in other parts of the world. This is the main point of studies such as Aikhenvald (1994, 2000b), Derbyshire and Payne (1990), and Payne (1986). They all argue, for instance, that the current typologies that distinguish between numeral and verbal classifier systems do not apply to those languages where that distinction seems to be blurred. Within the context of more recent typological work on classifier sys- tems which emphasizes the need to distinguish between distinct subtypes of classifi- ers (Grinevald 2000 and earlier versions of it found under Craig 1992, 1994), more MOVIMA NOMINAL CLASSIFICATION 233 recent studies of Amazonian systems have tended to treat the descriptive challenge that those systems represented by considering them to be cases of yet unseen situa- tions of ‘multiple classifier systems’ within the same language. All along, people have also noticed that those systems appear to challenge the classical distinction between inflectional and derivational morphology in that, when one wants to describe the so-called classifiers of these languages, one must handle sets of morphemes that function like classifiers, which are generally considered to be part of the inflectional morphology of the language, but also like derivational affixes, in their capacity to create new lexical items. Finally, another specificity of these Amazonian systems of nominal classifica- tion, often mentioned in descriptions, is that they function primarily in an anaphoric way. What is often commented on by field linguists is the extent to which the use of full lexical nouns is quite limited in discourse, in such a way that these affixes fulfill the greatest part of the referent tracking process. This is discussed, for instance, in Barnes (1990) for Tuyuca and in Payne (1986) for Yagua; it is also mentioned in the overview article of Derbyshire and Payne (1990).

4.2. Movima noun class system There was indeed information provided by Judy and Judy (1962a) pointing to the existence of such a noun class system in Movima. Their listing of instances of what they had called “descriptive-object pronouns” included for instance a number of dif- ferent loci for these affixes, hinting at a concordial type of system typical of noun class systems. Their “examples of their use” already presented in the first part of this paper in (3) and (4) will be recalled here now to identify on which items such affixes were said to be found:

(3) “examples of their use”

3.1 sota’ba (sota’ra ‘one’) 3.2 choesba (choesni ‘ugly’) 3.3 daujesben’ (daujes ‘deer’) 3.4 son’di (sonra ‘other’) 3.5 oidi (oira ‘two’) 3.6 mere’mo (mere’e ‘big’) 3.7 beumo (beuni ‘mature’) 3.8 tochi’pa (tochi’i ‘small’) 3.9 pacoła (paco ‘dog’)

(4) a with the verb riłna ‘to soften’ riłaba ‘to soften a jug’ riła'oj ‘to iron’

234 GRINEVALD

b with the verb onarana ‘to know’ onamona ‘to know the bird’ onahuajna ‘to know the place’ onapoina ‘to know the animal’

From these examples one could assume that there was a concordial system, which affected elements of the noun phrases such as numerals (3.1 and 3.5), adjectives (3.2, 3.6, 3.8), and determiners at large (3.4). No specific information was given on two common categories of the noun phrase: possessives and demonstratives. It appeared also that there was concord with the verb, at least for the objects of some transitive verbs. Notice the element –poi– in the last example ‘to know the animal’ where one can identify the semantic suffix for animals mentioned earlier (ex. 21). The examples also included pointers to the existence of suffixes on nouns, which could correspond to class markers on head nouns of noun class systems (3.3, 3.9). The pattern of use of those suffixes suggested by the data from Judy and Judy’s material could therefore be sketched out as (30) below:

(30) [ N-SUFF Adj-SUFF Numeral-SUFF det-SUFF ] [ V-SUFF ]

Such a pattern could receive two possible interpretations at first sight. One approach consisted in considering it a matter of ‘multiple classifier systems’, with at least a numeral classifier system (which often includes markings of adjectives and determin- ers) and a verbal classifier system. This was the approach taken in the initial analysis, the one that led to the lack of interest in consciously checking the full pattern when eliciting new data in 1995. The other could have been to consider it as a noun class system from the start, with affixes on head nouns as well as concordial affixes of the same nature on different elements of the clause. This second noun class system analysis is imposing itself today, as much for Movima as for a number of other Amazonian languages. The elements that were there to suggest it a few years ago included: the multiple loci of affixation and the fact that similar suffixes seemed to be used in the different loci. The typology of clas- sifier systems relies, on the other hand, on the demonstration that distinct classifier systems have distinct semantic profiles and distinct sets of classifiers (see Grinevald 2000). In fact new data from 1995 included information that was not really compatible with the analysis of a numeral classifier system. It was the presence of suffixes on adjectives qualifying both count and mass nouns. Therefore, besides affixes on quali- fiers of count nouns such as the ones found in (31) below (from O’Connor, p.c. 1996):

MOVIMA NOMINAL CLASSIFICATION 235

(31) a mere’ ‘big’ mere’poy ‘the/a big one (animal)’ mere’mo ‘a big one (plantain)’ mere’ve ‘a big one (canoe)’ mere’mo ‘a big one (forest)’ there were examples involving mass nouns such as (32a, b):

(32) a adjective-SUFFIX from mass NOUN sokosoko’-mi ‘boiled (of water)’ tomi ‘water’ sokosoko’-lo ‘boiled (of milk)’ nonlo ‘milk’ dip-muj ‘strong (of wind)’ pawmuj ‘wind’

b sokosoko’-mi ‘boiling (water)’ palui-mi ‘cold (water)’ koyb’u-mi ‘muddy (water)’

As for the derivational properties of those suffixes, they were indirectly shown at least in one example of the almost undecipherable tagmemic grammar of Judy and Judy (1967: 393), which is reproduced below:

(33) kay–wan–ra ‘eat–nom–thing’ (food) where -ra can be identified as the generic suffix found also on numerals out of con- text. The point of belaboring how the initial analysis of the Movima system mistak- enly took it as a numeral classifier system, is to illustrate a case of partial blindness one might suffer in the process of collecting and analyzing data that is probably more common in field studies than admitted or identified. It is a reminder of the bind in which field linguists often find themselves: without some sort of theoretical frame- work, chances are that interesting linguistic phenomena would remain untreated be- cause they would easily go unnoticed, while too much of a theoretical bias can easily blind one to the nature of the data at hand. Good descriptive linguistics is an art and always a matter of balancing out being guided by but also guiding theoretical ad- vances for a better understanding of the wealth of the languages of the world. Today it seems clear that the analysis of the Movima nominal classification system has to be taken within the more general framework being developed presently for such systems of many languages of the Amazonian region, taken to be little grammaticalized noun class systems. All the data collected in 1995 on their manifestation with numerals therefore needs to be completed in the perspective of such an analysis to do it better justice.

236 GRINEVALD

4.3. Conclusion Movima has a productive nominal classification system. It would appear to be fairly typical of the type of not very grammaticalized noun class systems found in lan- guages of the Amazon region. It is a mixed system that functions mostly on the basis of morphophonological characteristics and partly on the basis of some semantic or- ganization. The basic principle of the truncation of the source nouns that was demonstrated is interesting in part for what it contributes to the study of one of the possible origins of classification systems, showing them to be secondary linguistic systems which develop from basic lexical material. The Movima case of truncation echoes other studies showing the same phonological process increasingly identified as a trait of the systems of the region, as well as in other parts of the world. It has been identified for a number of lowland languages of this region of the world, see for instance Barnes (1990) for Tuyuca, Payne (1986) for Yagua, and the other cases mentioned in Derby- shire and Payne (1990). The double process of truncation and reduplication in loan- words was taken to support the notion that the truncation, i.e. the process of dropping at least one syllable, was the primary operation. This was shown in the fact that, when there was a convergence of two constraints on the system, one being that a syl- lable be dropped, and the other being that the class suffix corresponding to loanwords be two-syllable long, the solution was the reduplication of the only syllable left once truncation had first applied, as the dominant process. The mixed nature of the system, which functions in some cases morphopho- nologically and in others semantically needs further study. The morphophonological truncation process points to a non-classificatory function of the suffix formation since it produces potentially innumerable ‘unique’ class markers. As was argued, the ap- parent classes presented by Judy and Judy which regrouped items supposedly sharing a same suffix were not really classes of items, but accidental regroupings which were the result of the application of the truncation principle. The multiplication of ho- mophonous suffixes was due in fact to the limited number of phonemes in the lan- guage and its simple inventory of basic syllable structures. The semantic classifica- tion of items seemed quite limited, but corresponds to what is known of classification systems of the world. As Denny (1976) argued, they are meant to highlight the rela- tion that humans have to the objects of their worlds, among which items of the ani- mal and plant world are indeed the most important for survival. The semantic classes mentioned are the most obvious ones, animals (four-legged and two-legged) and vegetal (fruits and plants); they were the easiest to elicit but it remains to see whether there might not be others at work in the language. The dimension of the use of the same suffix for what appear to be derivational and inflectional functions remains to be investigated. It would be a matter of system- atically observing the basic denomination of objects of the world, to see to what ex- tent the nominals of the language take themselves suffixes in their isolated form. In the grammaticalized Niger-Congo noun class systems, all nouns enter in the system MOVIMA NOMINAL CLASSIFICATION 237 and all are analyzed as having affixed class markers and participating in the system of concordial marking. In Movima the process of head marking does not seem very widespread but it is likely to exist to a greater extent than acknowledged here. As for the use of the suffixation, much remains to be done. On the one hand all loci need to be systematically checked, including demonstratives and possessives, and on the other hand the phenomenon of counting objects in itself may warrant its own study, for the richness of its constructions, and the apparent extensive variation in their use in the language today. It is to be hoped that such studies will contribute to justify in the eyes of the general linguists the study of the languages of Bolivia, in particular those that are genetic isolates. In this particular case of the study of a nominal classification system in an isolate language, it is worth considering how much of the phenomenon is in fact a reflection of an areal characteristic that is attracting increasing attention, as one of the characteristics of so-called Amazonian languages. Finally, beyond strictly aca- demic considerations of the kind just expressed, it needs to be said that the justifica- tion for such studies lies also largely in the expectation clearly expressed by inter- ested members of this community of speakers. They have been asking for linguists to come and work with them on their language so that they themselves learn to analyze the intricacies of their language and include it in the pedagogical material they want to produce. So may it be that this little essay on a fascinating aspect of their language impulse further study in which they will fully participate.2

References Adams, K. L. 1986 ‘Numeral classifiers in Austroasiatic’, in: C. G. Craig (ed.), 241-62. Adams, K. L. and N. F. Conklin 1973 ‘Toward a theory of natural classification’, in: Papers from the Annual Re- gional Meeting of the Chicago Linguistic Society 9: 1-10. Aikhenvald, A. 2000a ‘Unusual classifiers in Tariana’, in: G. Senft (ed.), 93-113. 2000b Classifiers: A Typology of Noun Categorization Devices, Oxford Studies in Typology and Linguistic theory, Oxford: Oxford University Press. Allan, K. 1977 ‘Classifiers’, Language 53: 285-310. Barnes, J. 1990 ‘Classifiers in Tuyuca’, in: D. L. Payne (ed.), 273-92. Berlin, B. 1965 Tzeltal numeral classifiers, The Hague: Mouton.

2 In particular the participants of the Preworkshop and Workshop for the normalization of the Movima alphabet in Trinidad and San Ignacio de Mojos of November 1995: Eligardo Chirimani Malue, Cristobal Agapito Cujuy Zeladi, Gilberto Machado Vega, Hipólito Mole Caumol; Melvin Rossell Yoqui and Pere- grina Cayu Mazaro. 238 GRINEVALD

Carpenter, C. 1986 ‘Productivity and pragmatics in Thai classifiers’, Proceedings of the Twelfth Annual Meeting of the Berkeley Linguistic Society 12: 14-25. Craig, C. G. 1986 ‘Jacaltec noun classifiers: A study in language and culture’, in: C. G. Craig (ed.), 263-93. 1992 ‘Classifiers in a functional perspective’, in: M. Fortescue, P. Harder, and L. Kristoffersen (eds.), Layered structure and reference in a functional per- spective, Amsterdam/Philadelphia: John Benjamins, 277-301. 1994 ‘Classifier languages’, in: R. E. Ascher (ed.), The Encyclopedia of Language and Linguistics, Oxford: Pergamon Press, 565-69. Craig, C. G. (ed.) 1986 Noun classification and categorization, Amsterdam/ Philadelphia: John Ben- jamins. Denny, J. P. 1976 ‘What are noun classifiers good for?’, Papers from the 12th Regional Meeting, Chicago Linguistic Society, 122-32. Derbyshire, D. C. and D. L. Payne 1990 ‘Noun classification systems of Amazonian languages’, in: D. L. Payne (ed.), 243-71. Dixon , R. M. W. 1972 The Dyirbal Language of North Qeensland, Cambridge: Cambridge Univer- sity Press. 1986 ‘Noun classes and noun classification in typological perspective’, in: C. G. Craig (ed.), 105-112. Erbaugh, M. S. 1986 ‘The development of Chinese noun classifiers historically and in young chil- dren’, in: C. G. Craig (ed.), 399-436. Givón, T. 1970 ‘Some historical changes in the noun-class system of Bantu, their probable causes and wider implications’, in: C.W. Kim and H. Stahlke (eds.), Papers in African Linguistics, Edmonton, Alberta: Linguistic Research, 33-54. 1979 On understanding grammar, New York/San Francisco: Academic Press. 2001 Syntax: An Introduction, 2 vols, Amsterdam/Philadelphia: John Benjamins. Gomez-Imbert, E. 1982 De la forme et du sens dans la classification nominale en Tatuyo (Langue Tukano Orientale d’Amazonie Colombienne), Thèse de 3ème cycle, Paris IV. Grinevald, C. 2000 ‘A morphosyntactic typology of classifiers’, in: G. Senft (ed.), 50-92. i.p. a ‘Making sense of noun classifiers: the grammaticalization variable’, in: E. Wischer and G. Diewald (eds.), New reflections on Grammaticalizations, 265- 82. MOVIMA NOMINAL CLASSIFICATION 239 i.p. b ‘Linguistics of classifiers’, in: B. Comrie (ed.), Encyclopedia of the Social and Behavioral Sciences. t.a. a ‘Classifiers’, in: C. Lehmann and J. Mugdan (eds), Handbuch der Morpho- logie, Berlin: Mouton de Gruyter, article 97. t.a. b ‘Classification nominale: le défi Amazonien’, in: Jon Landaburu (ed.), Lan- gues d’Amérique, Faits de Langues No 20. Grinevald, C. and D. Creissels 2000 ‘Typologie des systèmes de classification nominale: le défi amazonien’, Col- loque de Typologie, Université Paris III. Grinevald, C., D. Creissels and F. Seifart 2001 ‘Noun classes in African and Amazonian languages: towards a comparison’, Colloque international de Linguistique Africaine, Cologne. Judy, R. and J. Emerich de Judy 1962a Movima y Castellano, Vocabularios Bolivianos No1, Cochabamba, Bolivia: Instituto Lingüístico de Verano. 1962b Fonemas del Movima: con atención especial a la serie glotal, Notas lingüísti- cas de Bolivia No 5, Cochabamba, Bolivia: Instituto Lingüístico de Verano. Judy, R. and J. Judy 1967 ‘Movima’, in: Esther Matteson (ed.), Bolivian Indian Grammars 1, Summer Institute of Linguistics Publications in Linguistics and Related Fields 16, Norman: Summer Institute of Linguistics and the University of Oklahoma, 353–408. Klein, H. E. M. 1979 ‘Noun classifiers in Toba’, in: T. M. Mathiot (ed.), Ethnology: Boas, Sapir and Whorf revisited, The Hague: Mouton, 85-95. Lakoff, G. 1986 ‘Classifiers as a reflection of mind’, in: C. G. Craig (ed.), 13-51. 1987 Women, fire, and dangerous things: what categories reveal about the mind, Chicago: Chicago University Press. Payne, D. L. 1986 ‘Noun classification in Yagua’, in: C. G. Craig (ed.), 113-131. 1987 ‘Noun classification in the Western Amazon’, Language Sciences 9: 21-44. Payne, D. L. (ed.) 1990 Amazonian linguistics. Studies in Lowland South American languages, Austin: University of Texas Press. Plaza, P. and J. Carvajal 1985 Etnias y Lenguas de Bolivia, La Paz: Instituto Boliviano de Cultura. Senft, G. (ed.) 2000 Systems of Nominal Classification, Cambridge: Cambridge University Press. Vidal, A. 1995 ‘Noun classification in Pilaga (Guaykuruan)’, Master’s thesis, University of Oregon.

COMPLEX VERB FORMATION IN LEKO

Simon van de Kerke University of Leiden/WOTRO

1. Introduction1 Leko used to be spoken widely on the Andean eastern slopes of the province of La Paz in Bolivia. Today it is almost extinct. The last speakers have passed the age of 60. They did not speak the language for over 40 years and did not pass the language on to their children. The major published source on the language, apart from the small word lists by Lázaro de Ribera (in Palau & Saíz 1989), Cardús (1886), Brinton (1946) and Montaño Aragón (1987), is a Christian doctrine composed by the missionary Andrés Herrero in the beginning of the nineteenth century. The Leko doctrine was published by Lafone Quevedo (1905), who used it as the major source for a grammatical sketch of the language. His work has remained the only serious analysis of Leko, since no other linguistic study of the language was undertaken since then. In 1994, I was able to trace some of the last speakers of the language and during visits in the following years I gathered enough language data to extend the linguistic analysis presented in Lafone Quevedo and publish a reinterpreted version of the Leko doctrine (van de Kerke 1999). A first view of complex verb formation was presented in van de Kerke (1998) and the purpose of the current article is to pursue this line of investigation, since I have col- lected much more reliable data since then.

2. The Leko language In Grimes (1996) Leko, classified as an isolate, was reported to be extinct. However, Montaño Aragón (1987) reported a number of speakers in the region of Atén and Apolo in the province of Franz Tamayo and along the river Mapiri in the province of Larecaja, both in the Andean foothills region (the eastern slopes), to the north of La Paz, Bolivia. In 1994, responding to an appeal in Adelaar (1991) to investigate the possibility that speakers of the Leko language might still survive, I undertook a fieldwork trip to Bolivia. A thorough search in the region of Atén and Apolo was in vain, but I contacted some elderly men and women on the Mapiri river, who in their youth had learned to speak a language that, on the basis of the existing data, could be characterized as Leko. Since then, ongoing fieldwork has produced much more language data. One of the major problems is the fact that the Leko speakers do not feel free to enter into spontane- ous conversation with each other, so most of my data are the result of elicitation.

1 I would like to thank the Netherlands Foundation for the Advancement of Tropical Research (WOTRO), a subdivision of the Netherlands Organisation for Scientific Research (NWO), for financial support. H. van der Voort and Mily Crevels were so kind to comment on a draft version of this article. An explanation of the glosses that are used in the morphological analysis are given in a separate appendix. 242 VAN DE KERKE

Happily enough, in july 2001, a speaker2 whom I had contacted the year before, told me a good number of spontaneously produced stories. Wherever possible I will use exam- ples from these stories to illustrate the use of the different derivational suffixes. Tenta- tively, a number of ordering constraints of the suffixes will be discussed in section 5.

3. Complex verb formation in Leko Leko belongs to the class of so called ‘agglutinative’ languages in which a complex meaning may be expressed by means of a morphologically complex word. Complex words consist of a root and a number of clearly recognizable morphemes with a specific meaning. Apart from the addition of inflectional morphemes, which express Person, Number, and Case (in the case of nouns) and Tense (in the case of verbs), complex words may be formed by the addition of derivational morphemes and by means of the compounding of roots. Whether and how these word formation proc- esses may be distinguished is an unresolved question that has led to a number of highly theory-dependent and language-specific answers. The distinction between derivation and inflection is difficult to make when inflectional morphemes occur in between clearly derivational morphemes. The same holds for the distinction between derivation and compounding. The fact that a root is combined with another mor- pheme that may occur as a root in the language, is by itself no conclusive evidence for the combination to be a compound, especially when other morphemes occur in between them. Our knowledge of Leko is still so incomplete that it is premature to take a firm stand on the question whether some complex verbs are the result of derivation or of compounding. In the following I will present all the word formation processes as derivational, although I am well aware of the fact that in the end a number of them may be better analyzed as cases of compounding. Unlike the neighbouring highland languages Quechua and Aymara with which it shares a number of features, Leko has both suffixes and prefixes. Due to the fact that until this year almost all data from Leko were gathered by means of elicitation and due to the fact that the speakers had not spoken the language for over 40 years, the number of affixes collected so far probably form a small subset of the affixes that were used when the language was still in daily use. However, by scrutinizing the collected data closely from time to time a new affix is discovered, the meaning and position of which can be understood by subsequent elicitation. For example, apart from the object markers that have already been treated in van de Kerke (2000), we find in the set of prefixes the element ka- that expresses a subjunctive: it refers to unrealized but possible states of affairs. In the set of suffixes, we find a number of elements that may be described as derivational in nature. They are meaning and/or valency changing and are reminiscent of equivalent sets of elements, both in meaning and in position, in many other languages that rely on verbal derivation to form complex verbal expressions. As far as position in the verbal complex is concerned,

2 Cerilo Figueredo, who was born in Karura around 65 years ago and who still lives there.

COMPLEX VERBS IN LEKO 243 they can be identified as elements that occur in between the verb root and an inflec- tional element, which can be either verbal like tense or a nominalizing suffix. This leads to the possibly arbitrary decision that an element like -chi that mitigates an imperative, which forms part of the inflectional system, is treated as derivational. A suffix like -sin, an equivalent of untranslatable particles like English ‘then’ and Spanish ‘pues’, is not, since it follows the imperative suffix:

1. a. ka- yin- k’o -chi -a SUBJ 1 DAT eat MIT IMP ‘He should eat it on my behalf’

b. yin- k’o -a -sin 1DAT eat IMP ‘then’ ‘Eat it on my behalf, then’

Besides clearly inflectional elements like imperative -a, I will analyse -no as inflec- tional, too. In earlier papers on the language I have presented -no as an indicative marker, the suffix being in complementary distribution with the negative marker -in. However, I now tend to analyze it as a nominalizing element since it can be followed by a case marker. Note, for example the dative marker -ki in lulchanoki in (2), which is the head of the object phrase ber kallapote lulchanoki ‘the one passing on a raft’, the complement of the main verb dosoqchano ‘he was looking at’:

(2) kala -ra do- soq -cha -no (-te3) ber kallapo -te lul -cha -no -ki beach LOC 3OB look DUR NOM ( MT) one raft LOC pass DUR NOM DAT ‘From the beach he (yobora 'my friend') was looking at someone else passing on a raft’ (the beach story)

The same holds for the suffix -a, which I used to analyze as a part of the past tense marker, but that can better be analyzed as a nominalizing element: a past participle. There is still a lot we do not know about Leko, but on the basis of the data I collected during my fieldwork we can obtain a first impression of the role of the derivational suffixes in this hardly studied Bolivian ‘pie de monte’ language.

4. Derivational suffixes in Leko 4.1. –cho ‘feel like’ The suffix -cho is glossed as ‘desiderative’, but that is only one part of the meaning complex, the idea of which is better expressed by ‘to feel like’. As a lexical verb it

3 I used to analyze the suffix -te as the third person main tense marker. I now analyze it as a tense marker 'pur sang', the third person marker being zero. In colloquial speech -te is easily dropped. Therefore, one could also analyze dosoqchano as a nominalized clause in apposition to the noun yobora ‘my friend’. This would lead to the interpretation: ‘my friend, the one looking from the beach at the one passing on a raft’.

244 VAN DE KERKE expresses the idea ‘to hurt’. It must be realized immediately after the root and it is restricted to bases that express express mental or physical states: ‘be in pain’, ‘feel sad’, ‘be hungry’, etc.:

(4) sok’och chikala lais soncho -no hichis -cho -no food very well well smell NOM be hungry DESID NOM ‘The food smelled very nice, he felt hungry’ (the bear story)

The concept of mental or physical state has to be understood broadly since a verb like sisich ‘to sleep’ is also a target for suffixation:

(3) sis -cho -no -too sleep DESID NOM 1 SGMT ‘I feel like sleeping’

I encountered a great number of combinations of -cho with the adjective lais ‘good’. This suggests that the suffix can turn adjectives into verbs:

(5) lais -cho -no -te charki lais -cho -no utu -no on charki well DESID NOM MT dry meat well DESID NOM fat NOM that dry meat ‘It was very good, the dried meat, very good, fat it was, the dried meat’ (the wild boar story)

4.2. The ‘Motion’ modifiers Modification of motion verbs by means of suffixes is a well known characteristic of languages in the region, like Quechua and Aymara, and Leko is no exception. Motion verbs that are not inherently specified for the direction in which the motion takes place, are marked as such by means of motion modifiers. Frequently occuring are ‘upward’, ‘downward’, ‘inward’ and ‘outward’ motion. The first three of these are attested in spontaneous speech.

4.2.1. -su(ri) ‘upward motion’

(6) kawot do- to -suri -a has -rep above 3OB put UP IMP below ABL ‘Throw it from below’

(7) soboto hekor bat -tha -te chaka -su -no kawot -tha insect outside tree DIM LOC sit UP NOM above DIM ‘Outside in a little tree the insect sits high, a little bit high’ (the soboto story)

COMPLEX VERBS IN LEKO 245

4.2.2. -sa(ri) ‘downward motion’

(8) kawot -rep4 do- to -sari -a has above ABL 3OB put DOWN IMP below ‘Throw it down from above’

(9) ho cha wela -ra yin- to -sari -tan on moo lelwo this three arrow LOC 1DAT put DOWN OBL that monster ‘With these three arrows you have to put down for me that monster’ (the story of the lord of the animals)

4.2.3. -sori ‘inward motion’

(10) ho cha kura wet -a ho -ra wet -no ho- to -sori -ko dowa -ra this three priest die PP this LOC die NOM 3 OB put INTO EXH river LOC `These three priests have died, they died here, let’s throw them in the river’ (the story of the three fathers)

(11) kandado e -ri -m ha- amphas -sori -a on pilpuq -aya -ki sor se lock pull INC PRP 3OB push INTO PP that door PL DAT in side ‘Pulling out the lock, he pushed the doors inward’ (the story of the three fa- thers)

Although I tried to find the missing ‘outward motion’, I could not get my informants to form an acceptable utterance like ‘he walked out of the house’ other than by using the free verb ubusich ‘to leave’ in combination with the place adverb hekor ‘outside’:

(12) wonon -mo ubus -cha -no hekor walk PRP leave DUR NOM outside ‘He leaves walking outside’

Whilst searching for ‘outward motion’, I was confronted with some other motion modifiers, which give us a clear indication of how these suffixes may have emerged: via a process of verb incorporation. Contrasting pairs show that the same lexeme may be used either as a free verb or as a suffix. In (14a) and (15a) we find the independent verbs wari- ‘to climb a slope’ and hori- ‘to enter’ in combination with a motion verb in the present participle, in (14b) and (15b) we find the same motion verb in combi- nation with -wari and -hori, that now funcion as a suffix.

4 At least after nasals the case markers -ra and -rep change to -da and -dep, just like the inceptive marker -ri and the future marker -ra change to -di and -da All informants use -rep as the ablative marker, but Cerilo Figueredo also uses the marker -bet. This is an intriguing fact, since in van de Kerke(2000) I already noted the possible link of -rep with the ablative marker in Cholón -(llac)-tep. A similar link may be stipulated for -bet.

246 VAN DE KERKE

4.2.4. -wari ‘upward motion’ The difference with ‘upward motion’ -suri is a difference in grade. While -suri indicates a relatively vertical upward motion, -wari indicates upward motion on a slope.

(13) on wotha wari -m on -da tintatinta laq -a -te era -iki that hill move up PRP that LOC tree peel bark PP MT I DAT ‘Climbing the hill, I peeled the bark from the tintatinta (tree) for me’ (story of the tintatinta tree)

So, we obtain the contrast between two independent verbs or one derived com- pounded verb:

(14) a. wonon -mo wotha wari -no -te walk PRP hill UP NOM MT ‘Walking he climbs the hill’

b. wotha wonon -wari -no -te hill walk UP NOM MT ‘He walks the hill up’

4.2.5. -hori ‘inward motion’ The verb ho- means ‘to enter’. Used as a suffix in combination with ‘inceptive’ -ri, it marks ‘motion into’:

(15) a. on kachu ho -ri -no -te heboa -ra se -m that bird enter INC NOM MT nest LOC fly PRP `The bird enters the nest flying’

b. on kachu se -hori -no -te heboa -ra that bird fly INTO NOM MT nest LOC `The bird flies into the nest’

4.3. -ri ‘inceptive’ The ‘inceptive’ character of the suffix -ri may mark that an action is going to take place, cf. (16). In other cases it has related meanings, such as changing a stative verb like huchich ‘to be the end’ into a process verb huchirich ‘to come to an end’, as in (17):

(16) miis yo- yo -ki yin- hal -di -ra -no -te wes -ra tomorrow 1SG mother GEN 1DAT buy INC FUT NOM MT Huanay LOC ‘Tomorrow my mother will go and buy me (a new one) in Huanay.’

COMPLEX VERBS IN LEKO 247

(17) on saltaqwo -ne chika hel topaq ki- chinwa -ra that insect TOP very brilliant back 3SG nose LOC

huch -ri -a -ra hote -te siri -aya end INC PP LOC have MT eyes PL ‘That flying insect has a very brilliant back, where its nose ends it has its eyes’ (the saltaqwo story)

4.4. -hi ‘completive’ The completive aspect that is marked by the suffix -hi relates to different concepts of completedness. It indicates that a process is completed:

(18) yo- moki warsuch tiltil -hi -no -te 1 GEN trousers old CMPL NOM MT `My trousers are completely used.’

(19) on -ne vasias -a -te -am it -ki -hi -a -te kaldera that TOP empty PP MT 1PL full CAUS CMPL PP MT container ‘That one we have emptied, we have completely filled the container’ (the story of the sugarmill)

Completive aspect is also attested in the next example, but in a different way. Here it relates to the concept of completedness in respect to the set of elements that either performs the action (Agents) or on which the action is performed (Patients). It is quite logical that in such cases one also encounters the universal quantifier seneng ‘all’:

(20) seneng se -hi -a -ra decien se -ra -no -te on suma lito all fly CMPL PP LOC then fly FUT NOM MT thay big queen ‘When they have flown all, then that queen mother will fly (the ant story)

(21) ni ber noq- bes -in mo -no -te senen kin- k’o -hi -a polo not one PL? wake NEG say NOM MT all 3DAT eat CMPL PP puma ‘Not one (mule) did wake up, it is said, the puma(s) had eaten them all to his detriment’ (the bear story)

4.5. -har/-handa ‘again, to come to’ The semantics of -har (and its allomorph -handa) are far from clear, but elicited examples suggest the implication of ‘motion’ prior to action:

(22) do- woy -di -a sok’och men -cha -no -te k’o -har -ai 3OB call INC IMP food cool DUR NOM MT eat COME IMP ‘Go and call him, the food is getting cold, he must come to eat (it)’

248 VAN DE KERKE

In the text material I only encountered examples where -har was combined with the verb nech ‘to exist, to be’, in which cases it expresses that a former state of affairs has come into existence again. In the following example the speaker could have used wilkach ‘to return’, but that would have been different from ‘to be there again’:

(23) on yobas -ne contento ko- noko -aya ne -har -no -aya -te -s that man TOP happy 3 brother PL be COME NOM PL MT PL ‘That man felt happy, his brothers showed up again’ (the story of the lord of the animals)

4.6. –somo ‘at once’ The semantics of the suffix -somo appear to be very complex, but in a number of contexts it indicates that an action is performed immediately or in a hurry:

(24) hap k’eso do- hoq -somo -a ka -te thunu -ki viper (type) 3OB swallow IMM PP be MT rat DAT ‘The viper has swallowed the rat at once’

(25) chera samas -no -te -am perol primera to -somo -a khiri -te we rest NOM MT 1PL pan first put IMM PP oven LOC ‘We rest, (but) first we have put at once the pan on the fire’ (how to make chankaka)

4.7. –bats ‘almost’ The semantics of the suffix -bats is not yet clear, since it is a rarely used suffix. However, the few examples I obtained suggest that it may be used in contexts where English would use the adverbial element ‘almost’, As in English, we find that the scope of the adverbial element may be either on the object as in (26), or on the action expressed by the verb, as in (27):

(26) thunu -aya senen aros k'o -hi -bats -a ka -te -s rat PL all rice eat CMPL NEAR PP be MT PL 'The rats have eaten almost all the rice'

(27) polo -ki ber puñete di- ki -a di- kis -bats -a mo -no -te puma DAT one punch 3OB make PP 3 OB kill NEAR PP say NOM MT ‘He gave the puma a punch, it is said that he almost killed him’ (the bear story)

4.8. -mo ‘reciprocal’ The suffix -mo, homophonous with the present participle, combines with transitive action verbs, the thematic grid of which contains both an Agent and an animate Patient. It then indicates that these two animate entities perform one and the same

COMPLEX VERBS IN LEKO 249 action with respect to each other. In combination with ditransitive verbs with an animate third argument (the Goal), like kuch ‘to give’ and maytasich ‘to lend’, the reciprocal relation holds typically between the Agent and the Goal.

(28) yobas -aya yanapas -mo -no -aya -te ta balich -ki man PL help REC NOM PL MT corn plant DAT ‘The men help each other to plant corn’

With a number of verbs the reciprocal marker forms semi-lexicalized derivations:

(29) kelich ‘hit’ kelmoch ‘fight’ somduch ‘speak to’ somdumoch ‘chat’

In reciprocal constructions one usually finds the two (or more) participants in the reciprocal action as subject of the verb:

(30) abor -mo -a mo -no ka -te ber tropa wutili ber tropa ch’owe meet REC PP say NOM be MT one group wutili one group ch’owe ‘It is said that they met again, one group of wutili one group of chowe (two different species of monkey) (the monkey story)

However, especially with the semi-lexicalized reciprocal verbs, one of the actors can be realized as the subject while the other is realized as a comitative complement:

(31) kibi kel -mo -cha -ra k’eso -i, ni haka do- soq -in -kama -te he hit REC DUR LOC snake COM nobody 3OB look NEG be able MT ‘When he is fighting with the snake, nobody can look at him’ (the ‘soboto’ story)

In the next fragment we find an unexpected use of the reciprocal marker with the verb wasuch ‘to close’ which can be understood when we realize that doors in many cases consist of two parts, here explicitly expressed by mentioning on toi korwa ‘the two doors’:

(32) wasu -a mo -no won wasu -mo -a toi candado lewa -ki -a close PP say NOM house close REC PP two lock hang CAUS PP

toi candado toq -a yawis -a mo -no ka -te, on toi korwa two lock put PP key PP say NOM be MT that two door ‘He (the king), it is said, closed the house, he closed it (together), he hung two locks, he put two locks, he closed them with a key, it is said, the two doors (the story of the three fathers)

250 VAN DE KERKE

As I have argued in van de Kerke (1998), there is no reflexive marker in the lan- guage. When an object complement is not overtly expressed, the absence of an object marker on transitive verbs is understood as an implicit reflexive.

4.9. ‘Higher verbs’ It may be the result of influence from Spanish that speakers of Leko use the same ‘higher verb’ roots, either as infinitival complements or as derivational suffixes. So the string yisisich kich rainte ‘they do not want to let me sleep’ in (33), has almost the same meaning as yisiskirainte:

(33) on bel -aya era -iki lais yi- sis -ich ki -ch ra -in -te that bat PL I DAT well 1OB sleep INF make INF want NEG MT ‘Those bats do not want to let me sleep well’ (the story of the three fathers)

Note the, from the viewpoint of argument structure, unexpected placement of the object marker yi- on the lowest verb sisich ‘sleep’. As an argument of the verb kich ‘to make’, the object marker should have been realized on this verb. This is seen in the next example in which the speaker makes use of the Spanish loan dejar ‘to let’ as a higher verb, on which the object marker is realized:

(34) ho -ra bel -aya -te ye- dejas -in -te sis -ich -ne this LOC bat PL MT 1OB let NEG MT sleep INF TOP ‘There are bats here, they do not let me sleep’ (the story of the three fathers)

4.9.1. -ki ‘’ The lexeme ki- ‘to make’ or ‘to let’ can be used as a free verb selecting a nominal or verbal complement, as in (33), or as a causative verbal suffix with the same meaning:

(35) o- botha -tha -ki do- ko -ki -a 2SG brother DIM DAT 3OB drink CAUS IMP ‘Make your little brother drink.’

(36) ... hasta wakia -ra ho -m noka ubus -ki -no -te -aska until hole LOC enter PRP how leave CAUS NOM MT UNCER

ber -ki ber -ki he- hepka -m hi- kis -no -te senen one DAT one DAT 3OB grab PRP 3OB kill NOM MT all ‘(that soboto kills) even entering the hole, how would he make them get out, grabbing them by for one, he kills them all’ (the ‘soboto’ story)

COMPLEX VERBS IN LEKO 251

4.9.2. -ra ‘future’ So far a special inflectional suffix to mark future tense has not been encountered in Leko. To refer to future events speakers make use of the derivational suffix -ra which I suppose to be related to the free verb dach ‘to want’. In many instances it is realized as rach, and it can also have a future connotation, the meaning ‘to want’ being linked to the idea of ‘unrealized’. As a free verb dach takes either a nominal or an infinitival complement. So as equivalent phrases are accepted:

(37) a miis on yobas yo- won -ki -ra sis -ra -no -te tomorrow that man 1 house GEN LOC sleep FUT NOM MT ‘That man will sleep tomorrow in my house’

b miis on yobas yo- won -ki -ra sis -ich da -no -te tomorrow that man 1 house GEN LOC sleep INF want NOM MT ‘That man wants to sleep/will sleep tomorrow in my house’

The following two sentences are produced in the context of storytelling and clearly show the future/unrealized character of -ra:

(38) es -ra -no -te cha wison -da es -ra -no -te rain FUT NOM MT three day LOC rain FUT NOM MT ‘It will rain, within three days it will rain’ (the ant story)

(39) dingu -i era din- somdu -ra -no -too cuento kulew -moki cuento gringo COM I 3OB speak FUT NOM 1 MT story vulture GEN story ‘I will tell a story to the gringo, the story of the vulture’ (vulture story)

4.10. -cha ‘durative’ The fact that actions or processes are ongoing is expressed by means of durative cha5:

(40) wachi ne -a o- yo -ki sis -cha -no -te silent be IMP 2 mother GEN sleep DUR IND MT ‘Shut your mouth, your mother is sleeping’

(41) yo -moki choswai moa mo -ki -a nohal -cha -no -te warapu 1 GEN woman fire fire CAUS PP boil DUR NOM MT cane juice ‘My wife has made fire, she is boiling the sugar cane juice’ (how to make chankaka)

5 There is an obvious link with the durative marker in Quechua which in various dialects is realized as : -chka, -cha, -ša, or -sa. Note that the nominal diminutive marker -cha in Quechua is realized in Leko as -tha.

252 VAN DE KERKE

4.11. -chi ‘politeness’ When speakers want to soften the imperative character of their requests, they can use the suffix -chi:

(42) lulaq yin- ura -a ka -te yi- siri -ra de- e -chi -a bee 1DAT enter PP be MT 1 eye LOC 3OB pull MIT IMP ‘A bee has entered my eye, please pull it out’

4.12. Mood The syntax and semantics of the following two suffixes is still badly understood. I assume that they form part of the ‘mood’ system of the language; normally an extremely complex matter. The impossibility of combining them with the nominalizer -no is a quite striking, although not understood, phenomenon: -kama-te or -bibi-te and never *-kama-no-te or *-bibi-no-te.

4.12.1. -kama6 ‘be able’ As one might expect the verbal suffix -kama has also a free variant with the same meaning. The derived form lamkas-in-kama-te-am in (43) may also be realized as lamkas-ich kama-in-te-am.

(43) chika es -cha -no -te lamkas -in -kama -te -am very rain DUR NOM MT work NEG BE ABLE MT 1PL ‘It’s raining cats and dogs, we can’t work’

(44) on -ka te -in -kama -te ni te -in -kama -too on -ka that COMP live NEG BE ABLE MT not live NEG BE ABLE 1MT that COMP

heka te -in -kama -te -noq heka -asne lais te -ch -moki you live NEG BE ABLE MT 2PL you too well live INF GEN ‘Like that one cannot live, nor I can live, like that you cannot live you also (should choose) to live right (the story of the monkeys)

Another example can be found in (31).

4.12.2. –bibi ‘almost’ I assume that the suffix -bibi is a mood marker. Almost always my informant added the Spanish adverbial casi ‘almost’ to underline the uncertainty of the expression. Apart from that, -bibi may occur in a paratactic construction. It is realized as a main verb, while the sentence which it modifies is realized independently, cf. (45b):

6 It is an interesting question whether there exists a link between the root kama as a verb and a verbal suffix, and the case marker -kama that indicates a limit in space or time.

COMPLEX VERBS IN LEKO 253

(45) a bat -te waqa -m (casi) kel -bibi -te tree LOC climb PRP (almost) fall almost MT ‘Climbing into the tree, he might fall/almost fell’

b bibi -te kaldera lelda -cha -no moa -te almost MT pan burn DUR NOM fire LOC ‘It is almost/may be the case that the pan is burning on the fire’

The little information I have on this suffix is all the result of elicitation, so care has to be taken with these data. Much more attention has to be paid to the complex nature of mood in the language.

5. The order of the verbal suffixes In the present state of knowledge of the language, it is dangerous to make explicit statements on the order of the verbal suffixes. My knowledge of Leko is only based on elicited information from a small number of informants and on the analysis of spontaneous speech of not more than two informants. Furthermore we have to keep in mind that all informants have to rely mainly on memories from their childhood. Apart from that, we have to realize that the use of the verbal modifiers may be restricted by the fact that all speakers are using Spanish as the language for daily use, and that Spanish is not a language in which verbal derivation occurs frequently. It should also be mentioned that other languages, which are of the same linguistic type as Leko, do not make use of verbal suffixes all the time either. In my research of Bolivian Quechua (van de Kerke1996) I found that of all the verbal types in a quite large transcribed Quechua corpus some 35% consisted only of a verb root, 45% of a verb root with one derivational suffix, leaving 20% for the combinations of a verb root with two or more derivational suffixes. So, by nature, affix combinations seem te be restricted and it is to be expected that it is one of the language features that is strongly affected by the process of language loss. However, in a number of cases, grammatical judgements on the impossibility of certain affix combinations were very outspoken, while a number of other affix combinations is well attested in my data. Of course, the latter all involve combina- tions of suffixes that are regularly used. Happily enough this group is relatively large. The following suffixes that have been presented in the present article are regularly used: -cho, -ri, -ra, -ki, -mo, -hi, -cha, -kama and -chi. The other suffixes:-somo, -bats, -bibi, -har/-handa, and the direction markers show up infrequently. It has already been noted that ‘desiderative’ -cho must be realized immediately after the root. It is followed by ‘inceptive’ -ri. In between is a position for the direc- tion markers, which all involve the suffix -ri, as we have seen. As regards the ‘aspec- tual’ markers, at least -hi, and possibly -somo, -bats and -har, are followed by ‘reciprocal’ -mo. This last suffix can be encountered at either side of the ‘higher verbs’, ‘future’ -ra and ‘causative’ -ki. The mood markers -kama and -bibi do show up at the end of the suffix string. This also holds for the ‘politeness’ marker -chi.

254 VAN DE KERKE

References Adelaar, W.F.H. 1991 ‘The Endangered Languages Problem: South America’, in: R.H. Robins and E.M. Uhlenbeck (eds.) Endangered Languages, Oxford: Berg Publishers. Brinton, D.G. 1946 The American Race, Biblioteca Americanista, Buenos Aires: Editorial Nova, Imprenta López. Cardús, Fr. José OFM 1886 Las Misiones Fransiscanas entre los Infieles de Bolivia: descripción del estado de ellas en 1883 y 1884, Barcelona: Librería de la Inmaculada Concepcíon. Grimes, B.F. 1996 Ethnologue: Languages of the World, 11th edn., Dallas: SIL. Kerke, S. van de 1996 Affix Order and Interpretation in Bolivian Quechua, Doct. Diss. University of Amsterdam. 1998 ‘Verb Formation in Leko: , Reflexives, and Reciprocals’, in: L. Kulikov and H. Vater (eds.) Typology of Verbal Categories, Linguistische Ar- beitsberichte 382, Tübingen: Max Niemeyer Verlag, pp. 195-203. 1999 ‘A 19th Century Doctrine in the Leko Language’, in: S. Dedenbach-Salazar and L. Crickmay (eds.) The Language of Christianisation in Latin America: Catechi- sation and Instruction in Amerindian Languages, BAS 20/CIASE 29, Markt Schwaben: Saurwein, pp. 115-150. 2000 ‘Case Marking in the Leko Language’, in: H. van der Voort and S. van de Kerke (eds.) Indigenous Languages of Lowland South America, ILLA 1, Leiden: CNWS, pp. 25-39. Lafone Quevedo, S.A. 1905 ‘La Lengua Leca’, Anales de la Sociedad Científica Argentina, tomo 60. Montaño Aragón, M. 1987 Guía Etnográfica Lingüística de Bolivia, La Paz: Don Bosco. Palau, M. and B. Saíz 1989 Moxos, descripciones exactas e historia fiel de los indios, animales y plantas de la provincia de Moxos en el virreinato del Perú por Lázaro de Ribera 1786- 1794, Ministerio de Agricultura y Alimentación, La Paz: El Viso.

Appendix: Glosses used in the text

1,2,3 Person marker DIM Diminutive IMM Immediate PL Plural ABL Ablative DUR Durative INTO Inward motion PP Past Participle BE ABLE be able EXH Exhortative LOC Location PRP Pres. Participle CAUS Causative FUT Future MIT Mitigator REC Reciprocal CMPL Completive GEN Genitive MT Main Tense SG Singular COM Comitative HORT Exhortative OBL Obligative SUBJ Subjunctive COMP Comparison IMP Imperative NEAR Almost TOP Topic DAT Direction INC Inceptive NEG Negation UNCER Uncertainty DESID Desiderative INF Infinitive NOM Nominalizer UP Upward motion

A FIRST COMPARISON OF PRONOMINAL AND DEMONSTRATIVE SYSTEMS IN THE CARIBAN LANGUAGE FAMILY*

Sérgio Meira Max Planck Institute for Psycholinguistics, Nijmegen, The Netherlands

1. Introduction The Cariban language family is composed of approximately 25 languages (numbers ranging from 20 to 50, depending on different researchers’ opinions about which varieties are dialects and which are independent languages), spoken by approximately 100,000 people in lowland South America, from south-eastern Colombia (where Karihona is spoken) to the Oiapoque river in Brazil (Karinya), from the coast of (Karinya) down to the southern Xingu area in central Brazil (Bakairí). The field of comparative Cariban studies was initiated more than two hundred years ago, when the relationship between a number of Cariban languages was first noticed by Filippo Salvadore Gilij (1782). Unfortunately, the historical-comparative method has been only very rarely applied to Cariban languages, for two main reasons: (1) most of the languages are, to this day, poorly known, which means that there is very little reliable material to compare; (2) most people who compared Cariban languages were not trained comparativists. Girard (1971) remains the only methodical attempt at reconstructing Proto-Cariban lexical items and proposing a classification (unfortunately based on a still very poor data base). In the area of morphosyntax, Gildea (1998) presents the first reconstruction of the person-marking and tense-aspect-mood (TAM) systems of Proto-Cariban and their syntactic properties. The present work attempts to contribute to the development of historical studies in the Cariban family by presenting a first preliminary reconstruction of the pronominal system of Proto-Cariban (including non-third-person and third-person, i.e. anaphoric and demonstrative, pronouns). For this purpose, the available sources (cf. Table 1 below) were scanned in search of pronouns, which were then sorted in cognate sets (Tables 2-4), according to what is known about the sound correspondences between Cariban languages (taking Girard 1971 as a guide), so as to draw conclusions on the historical evolution of these forms. In Section 2 below, the sources and standardized transcription are introduced. Section 3 has a summary discussion of pronouns in Cariban languages, which is the main background for the rest of the paper. The actual reconstructions are discussed in

* This paper is a revised version of a short comparison of Cariban demonstrative systems that was presented at the 50th International Congress of Americanists in Warsaw. It includes a significant amount of new material, and discusses also non-third-person pronouns. I wish to thank Hein van der Voort and Mily Crevels for comments on an earlier version, and Ana Carla Bruno and Bruna Franchetto for sharing their data. Any remaining mistakes are, of course, my own. 256 MEIRA

Sections 3.1 (non-third-person pronouns) and 3.2 (third-person pronouns). In Section 4, a summary table presents the reconstructed forms, followed by some further speculative comments on the relationships between these forms.

2. Sources and transcription Sources on Cariban languages, as is the case with most other language families in lowland South America, are very different in their level of reliability, accuracy, and breadth of coverage. For some languages, the best available sources are word lists from the last century; for others, there are recently published high-quality grammatical descriptions and occasionally even dictionaries. In view of that, the actual availability of data was a factor of importance in the selection of the languages to compare. Table 1 contains a list of the languages and sources selected for this study. Data from the best sources (marked with ‘++’ in Table 1) is assumed to be good in all respects; missing pronouns from these sources will thus be considered as non-existent. The less good sources (marked with ‘+’ in Table 1), and especially the worst sources (unmarked), are less reliable, and need to be handled with care. Mis- transcriptions and inadequate phonological analyses are a real danger; missing pronouns may in many cases actually result from gaps in the data.

Languages Sources Languages Sources

Tiriyó Meira 1999, 2000; fn ++ Tamanaku Gilij 1965[1782] Akuriyó fn ++ Cumanagoto Yangues 1683, Ruiz Blanco 1690 Karihona Robayo 1987, 2000a; fn ++ Chayma Tauste 1680 Hixkaryana Derbyshire 1979, 1985 ++ Pemón Armellada & Olza 1994 + Waiwai Hawkins 1998; fn ++ Taurepán Koch-Grünberg 1916 Katxuyana fn; Gildea’s fn + Makushí Abbott 1991, Amodio & Pira 1996 ++ Karinya Hoff 1968, Mosonyi 1978 ++ Ingarikó Koch-Grünberg 1916 Apalaí Koehn & Koehn 1986; fn ++ Arekuna Edwards 1977; Koch-Grünberg 1916 + Jackson 1972; fn ++ Akawayo Edwards 1977; fn + De’kwana Hall 1988 + Panare Muller 1994 ++ Yawarana Méndez-Arocha 1959 + Pacheco 1997; Gildea’s fn + Yukpa Vengamián 19781 Souza 1992 Waimirí- Bruno 1996; Bruno’s fn Bakairí Steinen 1892, Wheatley 1973, 1978 + Atroarí Kuhikuru Franchetto’s fn ++ Table 1. Cariban languages and sources considered in this work. he ‘+’ signs mark the best sources; fn = field notes (Meira’s if unidentified). The dotted lines identify probable lower-level genetic subgroups.2

The different transcription systems of the various sources have been standardized to facilitate comparisons. Certain details have been ignored (e.g. Koch-Grünberg’s

1 A couple of forms also from Robayo (2000b). 2 Tiriyó, Akuriyó, and Karihona were classified together already in Girard (1971) and in Kaufman (1994); Meira (2000) proposed the name Taranoan for this subgroup and reconstructed part of the grammar and vocabulary. Gildea (pers. comm.; cf. also 1998:91-96) considers Hixkaryana, Waiwai, and Katxuyana a subgroup, which he named Parukotoan. The other groupings, present in Girard (1971) and Kaufman (1994), contain very closely related languages or dialects. CARIBAN PRONOMINAL AND DEMONSTRATIVE SYSTEMS 257 attempt at marking non-phonemic distinctions between [e], [E] and [o], [ç], here tran- scribed simply as e, o; his marks — á, é, etc. — were also left out). The symbols in need of explanation are: ï = IPA [ˆ], ë = [´], j = [j] (a palatal glide, English y), χ = [x], γ = [ƒ] (velar fricatives), ’ = [/], ñ = [¯], x = [S], tx = [tS]. Sequences of identical vowels (aa, ee, etc.) are phonetically long. The language names, which often vary from source to source, have been respelled here for clarity, avoiding language-specific uses (‘Karihona’ instead of the Spanish ‘Carijona’). Names with final stress have a mark (e.g. Makushí, Apalaí); the others have penultimate stress (e.g. Arekuna, Akawayo = Arekúna, Akawáyo).

3. Cariban pronominal systems A typical Cariban system has pronouns for: first person (1), second person (2), first 3 person dual inclusive (1+2), first person exclusive (1+3), and third person (3). In terms of number, Cariban languages oppose ‘collective’ (i.e. focus on a group) vs. ‘non-collective’ (i.e. focus on less than a group, but not necessarily a single individual). Usually, there is a second person collective pronoun (2Col), based on the non-collective form plus a collective ending, a first person inclusive collective (1+2Col), based on the first person dual inclusive form, and third-person collectives; the first person exclusive form (1+3) is unmarked for number. Note that the first- person pronoun does not have a collective form; semantically, the 1+3 and 1+2Col forms play this role. The third-person pronouns form a relatively complicated system, including ana- phoric and demonstrative (proximal, medial, distal) forms which, along with number (collective vs. non-collective), distinguish also animacy. Derbyshire (1999:53-54) gives a first comparative overview of Cariban pronominal systems. For the sake of convenience, we shall follow his system of separating non-third-person from third-person pronouns as two subsystems, discussed in Sections 3.1 and 3.2, respectively.

3.1. Non-third-person forms The pronominal forms to be compared, from the sources in Table 1, are listed in Table 2 below. As can be seen, they seem to form good cognate sets. The first-person forms can be first divided in those that end in ro or rë, and those that do not. The same syllable is present at the end of other pronouns in many other languages. Considering its frequency (14 occurrences), one might feel tempted to reconstruct it, at least to some intermediate level. However, for the following reasons, this is not a good idea: (i) this syllable has no clear cognates in the other languages (the final wï found in several languages cannot be compared to ro ~ rë,

3 Syntactically, the 1+3 form is treated as a third person (e.g. verbs agree with it as if it were a third person pronoun); one wonders if it could have been an old non-possessible noun (cf. e.g. a gente ‘we’, literally ‘the people’). 258 MEIRA

since there is no regular w : r correspondence in the Cariban family); (ii) it has an obvious source in the ‘emphatic’ particle ro or rë, synchronically attested in most Cariban languages (e.g. Tiriyó wïï ‘I’, wïï rë ‘really me’, ‘yes, that’s me’; cf. Hoff 1990:508 for ro in Karinya [Carib of ], Derbyshire 1985:250 for ro in Hixkaryana). It seems best to assume that the endings ro and rë result from the reanalysis of the emphatic particle as part of the stem (much like otros in Spanish nosotros, vosotros).4

Language 1 2 1+2 1+3 1+2Col 2Col I you you and I we excl. all of us all of you Tiriyó wïï ëmë kïmë anja kïmënjamo ëmënjamo Akuriyó wï ëmë kïmë anja kï(më)njamo ëmënjamo Karihona ëwï ëmërë kïmërë aña kïñamoro añamoro Hixkaryana uro omoro kïwro amna kïwjamo omñamo Waiwai owï amoro kïïwï amna kïwjam amjamro Katxuyana owï omoro kïmoro amna kïmjarï omjarï Karinya-Hf5 au amooro kïχko a’na kïχkaaro amïijaro Karinya-Ms aau amooro kïm wooro na’na kïmwoññaaro amoññaaro Apalaí ïwï omoro kïmoro ïna kïmarokomo amarokomo Wayana ïu ëmë kunmë emna kunmëramkom ëmëramkom De’kwana ïwï ëmëdë kïwï ññaa kïnwanno ënwanno Yawarana wïrë mëërë ehnë Yukpa awï, awë amo, amor nana amora, amoja Waimirí awï, aa amï, amïrï kïkï a’a (amïrïtï) Tamanaku ure amare kiwe jumna kikemo amñamoro Cumanagoto ure amuere amna amia(mo)rkom Chayma ure, utxe amuere kutxe amna kutxekon amiamorkon Pemón (j)ure amare (j)uureto in(n)a juurenokon amarenokon Taurepán jïurë amaarï iná jïurïnïkon amaarïnïkon Makushí uurï amïrï uurï’kon anna uurï(’)nïkon amïrï(’)nïkon Akawayo urë amë(rë) ina, nja urë’nogon amërë’nogon Ingarikó ïure amëëre kiulenïkon tïmïïlïnïkon Arekuna jurë amërë inna jurëtokon amërë(k)nokon Panare ju amën juto, juta ana jutakon amënton Ikpeng uro omro, omo ug(u)ro tximna ugromo omromo Arara uro ugoro tximna ugoromo Bakairí urë ëmë kurë (x)ina (a)mareemo Kuhikuru uγe e(e)γe kukuγe tisuγe amaγo Table 2. Cariban non-third-person pronouns. Elements in parentheses did not occur consistently.

4 Interestingly, in Hixkaryana, the new first-person pronoun uro has given rise to a new first-person prefix ro-, r- (e.g. ro-jïmï ‘my father’), which has replaced an earlier Proto-Cariban *u- (cf. Gildea 1998). Note also that the final syllable ro, rë occurs as rï in Makushí and Arekuna (cf. below for Makushí ë > ï), and as γe in Kuhikuru (for which γ : r and e : ë are also regular correspondences: cf. Kuhikuru uγu ‘manioc bread’, tehu ‘stone’, Tiriyó uru, tëpu). The cases of re (Pemón, Ingarikó, Tamanaku, Cumanagoto, Chayma) are certainly mistranscriptions of rë. 5 Hoff (1968) and Mosonyi (1978) describe mutually intelligible dialects of the same language (‘Carib’ for Hoff, ‘Cariña’ for Mosonyi), here labeled ‘Karinya’. They are here treated independently (Hoff’s as Karinya-Hf, Mosonyi’s as Karinya-Ms) because their pronouns differ in form. CARIBAN PRONOMINAL AND DEMONSTRATIVE SYSTEMS 259

If we ignore the final ro or rë, all first-person forms seem to contain a w, or a reflex of it in the form of the vowel u; the longer forms contain a preceding and a following vowel (*VwV). The second vowel can be reconstructed as ï, and the cases of u can be seen as the result of vowel loss and syllable reduction (*Vwï > Vu > u). The first vowel, however, is a more difficult case: there are forms beginning with a, o, ë and ï. Such problematic vowel correspondences are not infrequent in Cariban languages, due to (often irregular) vowel assimilation (cf. Girard 1971:79). A final reconstruction must wait for better lower-level comparisons. For a tentative reconstruction, consider that: (a) ï is often the result of the weakening of an earlier vowel in Cariban languages, possibly as a first step in the process of syllable reduction and loss (Gildea, pers. comm.; cf. Gildea 1995 on Cariban syllable reduction); (b) ë and o seem to be diachronically related (cf. the second-person pronouns in Table 2), so that the ë- and o-initial forms are probably not independent. Taking (a) and (b) into account, *a is the best tentative reconstruction: with a following w, an *a > o assimilation would be much more natural than *o > a (cf. Gildea 1998:83-84 for a similar argument concerning the reconstruction of the second-person prefix *a(j)-). Tentatively, one could suggest a protoform *awï. Three problematic details remain, for which some suggestions are presented here. (1) Tiriyó ïï is probably the result of a metrical reanalysis of pronoun–clitic sequences: e.g. Pre-Tiriyó *ëwï rë ‘really me’ would go from [´VÆ@:}´] to [VÆ@:}´] by losing the initial vowel, at which point the surface long vowel would be reanalyzed as underlying ïï (or else it would become short — [Vˆ}´] —, as in all CVCV words; cf. Meira 1998, 1999 on the stress system), thus yielding wïï rë. (2) The initial j in Panare, Pemón, Makushí, and Taurepán may result from the resyllabification of an earlier *Vw > *ïw sequence (e.g. *ïwï rë > *ïu rë > juurï).6 (3) Chayma txe is rather puzzling; one might suggest that an element txe was added to an earlier *u (still attested in u-re), maybe by analogy with the 1+2 form kutxe (but note that the txe in kutxe is also of mysterious origin). In the second-person forms, one can again exclude the final syllables that reflect the particle ro ~ rë: ro, rë, rï, re, γe, and also Waimirí ra and De’kwana dë (De’kwana d often corresponds to r in other languages: cf. jïwïïdï ‘tapir’, Tiriyó ïwïrï). Panare n is also a likely reflex of an earlier rë; cf. Panare tunkë ‘horsefly’, akuñ ‘agouti’, Tiriyó turëkë, akuri. The cases of long vowels in the second syllable (Karinya, Taurepán, Ingarikó, Arekuna) are probably phonetic effects of the rhythmic stress system (cf. Meira 1998); Yawarana mëërë, on the other hand, may represent a case of underlying ëë resulting from the loss of the initial vowel, like Tiriyó wïï (cf. above). Looking at what remains, the second consonant m is almost always present (except in Kuhikuru; cf. below) and can safely be reconstructed, together with two

6 Taurepán jïu looks like an attempt at transcribing what could have been an intermediate stage (something like e.g. ïú:). Cf. the case of Portuguese eu [ew] and Spanish yo [jo], which have stressed different parts of an earlier *eo < Lat. ego. 260 MEIRA adjacent vowels: *VmV. The exact nature of the vowels is less clear; one can only make tentative suggestions. For the first vowel, one has the possibilities a, o, ë. As was mentioned above, ë and o may not be independent, which would reduce the choice to a vs. ë/o. Considering that the second vowel was probably ë/o, the ë/o cases in the first vowel could be the consequence of assimilation (*a > o, possibly made easier by the intervening labial *m), whereas the a cases are harder to derive from *o. The best hypothesis is thus *a. For the second vowel, one basically has ë/o: the cases of ue (Cumanagoto, Chayma) and a (Tamanaku, Taurepán) are probably mistranscriptions of ë, and the cases of ï (Waimirí, Makushí, Arekuna) look like reflexes of ë (ë : ï is attested in Makushí, as in e.g. sikï ‘flea, chigger’, Tiriyó sikë; Waimirí has no phonemic ë). As was mentioned above, ë and o are clearly related; there are numerous exemples of the ë : o correspondence (e.g. Tiriyó sikë ‘flea, chigger’, Apalaí xiko). Gildea (pers. comm.) considers ë to be always a reflex of Proto-Cariban *o, which is quite plausible phonetically. There are, however, o : o correspondences without apparent conditioning factors (e.g. Tiriyó okomo ‘wasp’, Apalaí okomo; cf. the second-person forms in Table 2). The question of whether o : o and ë : o are independent correspondences has not yet been settled. Taking a conservative stance, *o will be 7 reconstructed for o : o, and *o2 for ë : o. One thus ends up with a tentative proto- form *amo2. The last problem is the unexpected Kuhikuru form e(e)-γe (long ee attested in the author’s [Meira’s] field notes; short e attested in Franchetto’s field notes). One idea could be intervocalic m-loss: Pre-Kuhikuru *eme-γe > e(e)-γe. However, all attested cases of m-loss in Kuhikuru are word-initial, not word-internal (e.g. Kuhikuru oto ‘worm’, Tiriyó moto). It seems thus better to suppose that the initial *e was lost first: *eme-γe > *me(e)-γe > *e(e)-γe. (The long ee, in case it is not a transcription mistake, might result again from the influence of an earlier rhythmic stress system, as in the case of Tiriyó wïï.) The first-person dual inclusive (1+2) forms show more complex patterns. After eliminating the reflexes of the particle ro ~ rë, there are two major groups: (a) forms that contain the intial element ku, kï, ki, and (b) forms that contain an initial element ju, u (Panare, Pemón, Makushí; presumably, the other languages of the same group also have similar forms, unfortunately unattested). The best idea seems to be, since there is no initial k loss rule for the (b) languages, that these two groups of forms are not cognate. In fact, the (j)u-initial forms all seem to be based on the first-person plus a final element to, ta, kon, all reminiscent of number (collective) markers (e.g. Tiriyó ton, kon, Apalaí tomo, komo, etc.; cf. below the discussion of collective forms). This would imply a path of evolution whereby an original 1+2 form was lost and replaced

7 Note that the o : o and ë : o correspondences have distinct reflexes in Kuhikuru: e.g. tehu ‘stone’, Tiriyó tëpu, and oti ‘field, grass’, Tiriyó oi; cf. also Kuhikuru okõ ‘wasp’. Thus, Pre-Kuhikuru apparently had *o and *o2. Considering the number of (not obviously closely related) languages that have *ë, it is not impossible that Proto-Cariban *o2 was actually *ë. Not much, however, can be said without a detailed study of the distribution of o : o and ë : o in the family. CARIBAN PRONOMINAL AND DEMONSTRATIVE SYSTEMS 261 with an analytical 1 + Col form. One may further suggest that this form had originally collective, not simply dual, meaning, and that the collective forms (which have additional collective suffixes) may have originally been more emphatic synonyms. The Ingarikó and Arekuna forms would thus represent — in case they are not simple mistranscriptions — a retention of original k forms. The k-initial languages all share an initial syllable reconstructible as *kV. Given the overwhelming majority of cases of kï, the first idea is to reconstruct *kï. The cases of ku, however, give food for thought. First, ku occurs in Wayana and in the Southern languages (Kuhikuru, Bakairí, Arara, Ikpeng), which are as far away from each other as is possible within the family. One may consider also the earlier mentioned tendency for vowels to ‘weaken to ï’, and also the fact that k is not an obvious environment for (*ï > u). On the other hand, the possibility of deriving ku from an earlier *kïwï, at least for some languages (cf. below), must be borne in mind. All in all, reconstructing *kï seems to be still the best tentative hypothesis. The second syllable of the k-initial forms, however, varies quite wildly; it does not seem possible to view më (mo, mwo, nmë), wï (wi, we), txe, χko, ku (gu, go) as all cognate. Rather, it would seem that an initial element *ku (probably the same as the 1+2 prefix that Gildea (1998:92, 114) reconstructs as Proto-Cariban *k(ï)-) was added to several independent elements (maybe old possessible nouns) to make 1+2 pronouns; even dialects may end up with different forms (e.g. Karinya: kïχko [Hf], w kïm ooro [Ms]). The various forms can be separated in several groups, which correspond only imperfectly to proposed subgroupings (e.g. in Kaufman 1994): the më group (Tiriyó, Akuriyó, Karihona [= Meira’s Taranoan], Wayana, Apalaí, Katxu- yana, Karinya-Ms; most of Kaufman’s Guianan branch plus two Central branch lan- 8 guages; tentative reconstruction *kï-nmo2), the wï-group (Hixkaryana, Waiwai [= Gildea’s Parukotoan without Katxuyana], De’kwana, Tamanaku; two Guianan and 9 two Central branch languages; tentative reconstruction *kïwï), and the ku-group (Kuhikuru, Arara, Ikpeng, Waimirí; the Southern branch without Bakairí, plus one North Amazonian language; tentative reconstruction *kuku). Bakairí might be added to the ku-group (so that it includes all of Kaufman’s Southern branch) by assuming that ku-rë actually results from *kuku ro2 (which would also yield Arara-Ikpeng ug(u)ro, ugoro if one assumes the loss of the initial k). It is not unthinkable that Kari- nya-Hf kïχko is related to the ku-group: *kuku + *ko could yield present-day kïχko, but not *kïnmo2 + *ko or *kïwï + *ko (cf. Gildea 1995 on syllable reduction). Chayma kitxe remains isolated.

8 Notice that mwo instead of mo in kïmwooro represents no problem, since Mosonyi’s (Venezuelan) Ka- rinya has rules of palatalization and labialization of consonants depending on the quality of the adjacent vowels; the long oo results from the rhythmic stress system. The n in Wayana kunmë is less readily explained; it is tentatively reconstructed, despite the rather strange absence of its reflexes in the other më languages. 9 The long ïï in Waiwai kïïwï results from an idiosyncratic change (probably related to the stress system) that lengthened the first vowel of all CVCV words. 262 MEIRA

Going farther than this means going into the realm of speculation, which, all in all, is not a bad source of ideas. One first notices that *kïwï is not implausible as a source for the *ku forms (e.g. Wayana kunmë < *kïu nmo2 < *kïwï nmo2; for the *ku group, one might have e.g. *kïwï ro2 > *kïu ro2 > Bakairí kurë). However, this leaves the ‘double-ku’ forms (Kuhikuru kuku, Arara ugo, Ikpeng ug(u)) unexplained, and also Karinya kïχko; the lack of any reflex of the syllable wï in Tiriyó, Akuriyó, Karihona, Apalaí, and Katxuyana (one would expect at least a long vowel) is a further difficulty. One might also suspect that a simple *ku could have been the original source of both the 1+2 pronouns and the 1+2 person-marking prefix; it may even have been an independent element at some point (maybe still preserved in

Bakairí kurë < *ku-ro2), and would later on have blended with other elements (erstwhile independent nouns). However, the evidence for this element as an independent word in Proto-Cariban is very scant (Bakairí, the only apparent case of retention, could also result from *kuku + *ro2 with syllable reduction). Thus, in view of the variety of forms, it does not seem possible to reconstruct the form of a 1+2 pronoun to Proto-Cariban. Notice that it must have existed, since there are 1+2 pronouns in all languages (even those who lost the *ku-forms innovated new 1+2 pronouns) and the 1+2-marking prefix can be reconstructed; its form, however, must remain unreconstructed. This fact will be represented with the formula *kïCV for the presumed Proto-Cariban 1+2 pronoun. The first person exclusive (1+3) forms, like the first person dual inclusive forms, are also all apparently partially, but not completely, cognate. All forms end in na (Yawarana ehnë possibly explained by weakening, and Waimirí a’a maybe from an earlier *a’na), so that a final syllable *na can be reconstructed. However, the initial syllables, like the final syllables in 1+2 forms, clearly do not form a single cognate set. One can separate the attested forms into: an n-ñ or palato-alveolar group (a-nj, a-ñ, a-nn, i-nn, i-n, ñ: Tiriyó, Akuriyó, Karihona [Meira’s Taranoan], De’kwana, Yukpa,10 Pemón, Taurepán, Makushí, Akawayo, Panare, and probably also Apalaí and Bakairí;11 there are members of Kaumfan’s Guiana, North Amazonian, and Central branches; tentative reconstruction, *a-in(n)a); an m or labial group (a-m: Hixkaryana, Waiwai, Katxuyana [Gildea’s Parukotoan], Cumanagoto, Chayma, and probably also Wayana e-m, Tamanaku ju-m, and Ikpeng- Arara txi-m; tentative reconstruction, *a-m-na); and an ’ or glottal group ((n)a-’, e-: Karinya, Waimirí, Yawarana; two Central branch languages, one isolate; tentative reconstruction, *a/e-h/’-na). At this point, one may speculate further. It would seem that the three groups could be unified if one presupposes an initial element *ap which, in contact with an original *ina, could then: (a) nasalize to *am and yield amna with the loss of the vowel i, or emna without this loss (e.g. via *aimna < *am-ina), and further

10 It may be that the Yukpa form is missing a (na’na), in which case it would be transferred to the glottal group. 11 For Apalaí, one may suggest ï < *i (weakening-to-ï); for Bakairí, the initial x- may be a later addition: notice that xina is found in only one of the two dialects, the other having ina. CARIBAN PRONOMINAL AND DEMONSTRATIVE SYSTEMS 263 assimilate to the n, creating anna, inna, *ainna > anja, aña, ana, ïna (the last form with ‘weakening-to-ï’); or (b) reduce to a glottal segment, yielding Karinya-Hf a’na with loss of i (and also n for Waimirí a’a) and Yawarana ehnë without loss (via e.g. *ai-p-na > *e-χ-na). This would suggest reconstructing *ap-ina or *apina to Proto- Cariban, as depicted in Fig. 1 below. However, the in-/ñ-initial form could as easily be reflexes of a simpler *ina, without *ap; and Tamanaku jumna, Ikpeng and Arara tximna suggest that initial elements other than *ap could also occur (though their final m does suggest some relation to *ap). The formula *(ap)ina will be adopted here to stress the tentative status of the reconstruction of the initial element *ap.

amna *apna a’na a’a *apina *epna ehnë *aipna emna *ainna anja aña anna ana inna ina ïna Figure 1. A speculation on the evolution of *(ap)ina ‘1+3’. The remaining initial elements n, ju, x, tx are not included.12

The collective forms (1+2Col, 2Col) all seem to be derived from the respective non- collective forms with the help of the collective suffixes -njamo, -jamo, -jarï, -jaro, -aro, -wanno, -komo, -kemo, -ton, -nï, -no often more than one and not in the same order as other languages (though, with a few exceptions — Tamanaku, Panare, Arekuna —, every language uses the same suffixes in the same order for its 1+2Col and 2Col forms); the ‘emphatic’ particle ro, rë (< *ro2) often occurs, sometimes between suffixes. The best hypothesis seems to be the reconstruction of three collective markers, *jamo, *komo, *tomo (all still attested synchronically as such in several languages), and maybe also *no. The various collective forms would then be derived as follows:

Tiriyó kïmë-njamo, ëmë-njamo < *jamo Akuriyó kï(më)-njamo, ëmë-njamo < *jamo

Karihona kï-ñamoro, a-ñamoro < *jamo ro2 Hixkaryana kïw-jamo, om-ñamo < *jamo

Waiwai kïw-jamo, om-ñam-ro < *jamo (ro2)

Katxuyana kïm-jarï, om-jarï < *jamo ro2 (?) 13 Karinya-Hf kïχk-aaro, amïi-jaro < *jamo ro2

12 It is also possible to derive *apna from *aipna, rather than directly from *apina; in this case, *aipna would be Proto-Carib, and *apina either unnecessary, or maybe pre-Proto-Cariban. 13 Long aa < *ïja (as in amïjaro; note the short a here). Note that Karinya reduces nasal syllables to zero, even in synchronic morphophonology, so that *-jamo ro2 > (j)aro is not surprising (cf. awoomï ‘to get up’, aj-aawo-ja ‘I am getting up’). 264 MEIRA

w Karinya-Ms kïm o-ññaro, amo-ññaaro < *jamo ro2 Apalaí kïm-arokomo, am-arokomo < *jamo ro2 komo

Wayana kunmë-ramkom, ëmë-ramkom < *ro2 jamo komo

De’kwana kï-nwanno, ë-nwanno < *jamo ro2 Yukpa amo-ra, amo-ja < *(ro2) jamo

Waimirí amï-rïtï < *ro2 tomo

Tamanaku ki-kemo, am-ñamoro < *komo; *jamo ro2 Cumanagoto am-ia(mo)rkom < *jamo ro2 komo

Chayma kutxe-kon, am-iamorkom < *komo; *jamo ro2 komo Pemón (j)ure-nokon, amare-nokon < *no komo Taurepán jïurï-nïkon, amaarï-nïkon < *no komo Makushí uurï-(’)nïkon, amïrï-(’)nïkon < *no komo Akawayo urë-’nogon, amërë-’nogon < *no komo Ingarikó14 kiule-nïkon, tïmïï-lïnïkon < *no komo Arekuna jurëtokon, amërë(k)-nokon < *tomo komo; *no komo Panare juta-kon, amën-ton < *komo; *tomo Ikpeng ugro-mo, om-romo < *(ro2) komo Arara ugoro-mo < *komo Bakairí (a)ma-reemo < *ro2 jamo Kuhikuru am-aγo < *jamo ro2

Some suggestions for the problematic details are listed below.

(i) For the suffix -njamo in Tiriyó and Akuriyó, Meira 2000:59 suggests that it results from the reinterpretation, in an earlier collective pronoun, of the*n-jamo sequence as *-njamo, followed by the forming of new collective pronouns with *-njamo. He suggests the following steps for the 1+2Col form: *kïmë + jamo > *kïn-jamo > *kï-njamo, *kïmë + -njamo > kïmënjamo. For Hixkaryana -ñamo,

the obvious answer is nasalization by the preceding m (*om-jamo > omñamo). A similar explanation for the ñ in Karinya-Ms was not found thus far, but it probably exists.

(ii) Karinya-Hf -jaro < *jamo ro2, without nasal reflex, is not surprising: Karinya

loses NV syllables, even in synchronic morphophonology (cf. awoomï ‘get up’, aj-

aawoi-ja ‘I am getting up’). The long aa in the 1+2Col form probably results from syllable fusion (*kïχko-jaro > kïχkaaro; cf. 2Col amïijaro, with a short a); for the Karinya-Ms forms, however, no obvious explanation was found. NV loss also oc- curs synchronically in Apalaí and Kuhikuru. Katxuyana -jarï is surprising, both

because there usually is no NV loss in this language, and also because *ro2 should

occur as ro, not rï. Nevertheless, an irregular evolution of *jamo ro2 still seems

14 The initial k, t in Ingarikó (cf. also, from Koch-Grünberg 1916, Arekuna kuulïnïkon ‘we’) are probably mistranscriptions (but the k’s might also be remnants of earlier k-initial forms). CARIBAN PRONOMINAL AND DEMONSTRATIVE SYSTEMS 265

less strange than a whole new collective element *rï or *jarï without additional evidence. (iii) The nw in De’kwana nwanno is hard to explain (it is not related to Karinya-Ms mw; cf. fn. 7). There are some correspondences between j and w in the Cariban family (e.g. Tiriyó ë-jomi ‘your language’, Wayana ë-womi), so that it may still

be derivable from *jamo ro2 (> *jan-no, with nasalization of *r). The preceding n, however, remains unexplained (though it may indicate a connection with

*kïnmo2 languages). (iv) The *no found in the Pemón group languages (reconstructed as *no, rather than *nï, because of its frequency, including in the best documented languages: Aka- wayo, Arekuna, Pemón) is a surprising element, without obvious equivalent in the other languages. Its origin remains unknown (though one may compare it to the ‘postposition collective’ -:ne, -’ne, which sometimes occurs on nouns; cf. Tamanaku jeje ‘tree’, jeje-’ne ‘trees’; notice that a similar marker occurs in Apalaí on inanimate demonstratives: moro ‘that (medial)’, moro-’ne ‘those (medial)’.

Some languages seem to lack collective forms. Kuikuru (Franchetto, pers. comm.) has no 1+2Col pronoun and uses the simple 1+2 kukuγe in all contexts. Waimirí (Bruno, pers. comm.) has no 1+2Col form, and the 2Col form amïrïtï is often replaced by the simple form amï(rï). Some of the gaps in Table 2 may also indicate actual non-existent forms, and even some of the attested forms may be ad hoc, non- lexicalized formations (maybe Panare jutakon, amënkon, and Chayma kutxekon). One therefore wonders if collective forms should be reconstructed to Proto-Cariban at all. If one looks only at *komo and *tomo, the answer is probably ‘no’; but *jamo, which is apparently older than *komo (it is always closer to the stem when the two co-occur) is so frequent that it seems at least equally possible that Proto-Cariban *jamo forms were lost in the languages that lack them (the Pemón group, Panare, and Ikpeng-Arara). Collective *jamo forms are thus tentatively reconstructed here as

*kïC-jamo and *am-jamo (not *kïCV-jamo and *amo2-jamo, since there are no reflexes of the final vowel in any of the languages, except for Tiriyó and Akuriyó, in which it results from analogy — cf. (i) above — and Wayana, in which the final vowel was protected by the following *ro2).

3.2. Third-person forms Cariban languages usually distinguish animate from inanimate forms (the only ex- ception being apparently Waimirí; cf. below). For the sake of convenience, these two sets will be examined separately, in Tables 3 (inanimate forms) and 4 (animate forms). Note that only animate pronouns have lexicalized collective forms. A general classification (cf. Derbyshire 1999:54) recognizes anaphoric (or referential) and de- monstrative (proximal, medial, and distal) forms; though not all languages fit exactly 266 MEIRA into these categories, they are still frequent enough to be useful for comparative purposes. The anaphoric pronoun is not attested in most of the Venezuelan languages (Chayma, Cumanagoto, the Pemón group, Yawarana, Yukpa, Panare). In some cases, this may be due to gaps in the data; however, even the languages with the best sources (e.g. Makushí, Panare) do not mention special anaphoric terms. It is also absent in the Southern languages Arara and Ikpeng, but this is possibly a spurious gap, given the very poor available sources on these languages. If they are not taken into account, the languages without anaphoric pronouns form a geographically contiguous area, and may be more closely related to each other, while those with anaphoric pronouns occupy a larger area and do not seem to form any subgroup within the family. Based on this pattern, an anaphoric term may be reconstructed for Proto-Cariban.

Language Anaphoric Proximal Medial Distal Tiriyó15 irë senï, serë mërë ooni, mënï Akuriyó irë txenï, txerë mërë o’ni Karihona irë enï, ërë mërë mënï Hixkaryana ïro onï moro monï Waiwai ero on, tan moro mïnï Katxuyana ijo soro moro monï Karinya-Hf iiro eenï, eero mooro moonï Karinya-Ms ijjo eero mooro moonï Apalaí ïro senï, sero moro monï Wayana irë herë, sin mërë mïn De’kwana iijë(ë) ee’dë mënë Yawarana (s)eeni mënni Yukpa16 Waimirí irï (h)anji, kanji mïrï mo’o, mïmo Tamanaku txene more Cumanagota (tx)en muere muen Chayma (tx)en muere muen Pemón sene(k), sere txinek, muere Taurepán seene(k), sïlë mërï mënïg Makushí se(e)ni, sïrïrï siini, mïrïrï Akawayo se(e)rë mërë Ingarikó Arekuna seenïi(g) mërë Panare sï(h) mën, ëmë mu~mï’ Ikpeng nen mun Arara Bakairí ilë xirë mërë (awërë) Kuhikuru ïle iγe eγe Table 3. Cariban third-person pronouns: inanimate forms.

15 The terms senï and mënï usually occur in their reduced forms sen and mën, except in contexts that preserve the final ï (a following C(CV)-initial clitic or suffix). 16 The Yukpa sources contain a wealth of terms, all very poorly analyzed (e.g. Spanish ‘ese’: obsek, opse, okano, otka, maa, orko). Although some of them may be cognate with terms in Tables 3 and 4 (e.g. mari, maari, mas ‘this’), it seems wiser not to take them into account and wait for better data to become available. CARIBAN PRONOMINAL AND DEMONSTRATIVE SYSTEMS 267

The final vowel of this pronoun was clearly o2, given the ë : o correspondences (the long ëë in De’kwana remains unexplained). The first vowel is somewhat more dif- ficult to determine; *i looks like the best reconstruction, since an *i > ï change in Apalaí, Hixkaryana and Kuhikuru is more likely (‘weakening’) than the reverse *ï > i, without any clear conditioning environment (for Apalaí, note that the 1+3 pronoun ïna also has an ï where an i or a might be expected; for Hixkaryana, consider that *i > ï is elsewhere attested — e.g. the third-person prefix is ï-, not i-). The length (ii) in Karinya-Hf and De’kwana is probably the result of the stress system and should not be reconstructed. The intermediate consonant is usually an r, but (a) there are reflexes as j, and (b) also as l in Kuhikuru, a language in which *r > γ (cf. fn. 5). One possible explanation for this pattern would be a different proto-segment (e.g. *rj, or maybe a cluster *rj). However, two of the languages with j reflexes, De’kwana and Karinya, have synchronic morphophonological rules that change r into j in the vicinity of i in at least some cases (cf. e.g. the De’kwana possessive suffix -rï, which has an allomorph -jï used on stems that end in i); the possibility that this might also happen in Katxuyana cannot be excluded. For Kuhikuru, it may be argued that the same *i (which later became ï) was the environment conditioning the l reflex instead of γ. In the absence of more detailed comparisons, it seems better not to postulate a new segment for Proto-Cariban. The anaphoric pronoun is thus reconstructed as *iro2. The proximal terms seem to belong either to a *ro2 or to a *nï series, often with both terms co-existing in the same language (e.g. Tiriyó serë, senï). The fact that many languages do not have both terms raises the question of whether they should be both reconstructed to Proto-Cariban. More work on the actual distribution of those terms, their semantic value,17 and their diachronic relations to each other is clearly necessary. For the time being, considering that many absences may actually be gaps in the data, that there are some indications of occasional loss of a term (e.g. the

Waiwai *ro2 term seems to have taken up the anaphoric role, being replaced by the non-cognate tan as a proximal), and that languages with one term sometimes have one and sometimes the other (e.g. De’kwana and Katxuyana have *ro2 forms, while Yawarana and Hixkaryana have *nï forms), it seems best to reconstruct two proximals. The reconstruction of their form presents two problems: (1) the fricative initial element s, tx, h present in some languages but not in others; given that even closely related languages may disagree (e.g. Tiriyó and Karihona), it seems best not to reconstruct it;18 and (2) their initial vowel, which occurs almost always as e, but as a in Waimirí, ë in Karihona, and o in Hixkaryana, Waiwai, and Katxuyana (the Parukotoan languages). Waimirí is a very divergent language, so that the a might still simply be an idiosyncrasy; but Parukotoan o : e elsewhere is a correspondence attested

17 The semantic distinction between the two terms is still unsettled. Hoff (1968:272-273) argues that Karinya eero and mooro are the proximal and distal terms of a speaker-based subsystem, opposed to the speaker-and-addressee-based subsystem of eenï and monï. Meira, in a preliminary corpus study (to ap.-b), suggests that the difference is ‘newness’: serë refers to ‘new’, ‘recently introduced’ objects, while senï refers to previously known objects. 18 Ikpeng initial n is probably not cognate with this element; its origin remains unknown. 268 MEIRA also in other words (e.g. Hixkaryana jo ‘tooth’, Tiriyó je). This correspondence is probably related to ë : o, here represented as o2, a problem that can only be solved with more comparative work. Here, e : o is simply represented as o3. The reconstructed 19 forms are thus *o3ro2 and *o3nï. The medial and distal forms are easier to reconstruct, as *mo2ro2 and *mo2nï, respectively. Further comments: (1) Tiriyó ooni, Akuriyó o’ni, the actual distal terms (mënï is used for referents which are hearable but not visible; about the noise made by a non-visible motor, for instance, a Tiriyó speaker might ask: atï mën? ‘what’s that?’), have no clear origin. They do not correspond to the other terms in this series (Tiriyó has no m : ∅ correspondences word-initially; there is no source for length in the other words — notice that the stress system in Tiriyó does not automatically lengthen the first vowel in CVCV words —; and the final vowel does not correspond to the expected ï); they must have some other, yet unknown, origin. (2) The same can be said for Makushí siini, Pemón txinek, which are reminiscent of the proximal terms. (3) Panare mën seems to be the true cognate (with n < *ro2; cf. the discussion of second-person forms in the previous section); the origin of ëmë, and how its meaning differs from the meaning of mën, remain unknown. (4) Kuhikuru eγe exemplifies initial m loss, a normal feature of the language (cf. e.g. Tiriyó moto ‘worm’, Kuhikuru oto); one wonders whether it has become homophonous with the second-person pronoun or not; they might provide a minimal pair for length (in case Meira’s ee is not a mistake). (5) Ikpeng u is surprising; it is not known if this is a normal reflex. (For additional details, cf., mutatis mutandis, fn. 19). The animate anaphoric pronoun, as was the case with its inanimate counterpart, is mostly not attested in Venezuela (but notice Tamanaku nare). Again, since it exists in most other branches, it should be reconstructed to Proto-Cariban. The languages are more or less evenly divided into those with an initial vowel (i or ï), and those without it; it is not clear whether or not it should be reconstructed (note, in passing, that Apalaí again has ï where other languages have i, as was the case for the 1+3 inanimate anaphoric pronouns).20 It can be tentatively added to the final reconstructed form: *(i)no2ro2. (The a’s in Bakairí and in Tamanaku are probably mistranscriptions; the glottal stop ’ in De’kwana remains unexplained.) Note Kuhikuru l instead of r: the idea that the original *i ‘palatalizes’ the *r and keeps it from becoming γ, though still possible, becomes less plausible, since the *r is separated from the i by one syllable. The possibility that the intermediate consonant should be reconstructed as having a palatal element (*rj, or *rj) cannot be ruled out.

19 On remaining details: note that the final rï in Makushí sïrïrï probably stems from the emphatic particle *ro2, that the long vowels in Karinya (and probably in Yawarana and Taurepán) are due to the rhythmic stress system, and that the final g’s and k’s are probably mistranscriptions. 20 There are some indications that an earlier i-form may have existed in Tiriyó. The particle inëërë ‘that’s the one!’, which follows pronouns (as in e.g. mërë inëërë ‘it’s that one!’), looks related to nërë. Consider also the occurrence of nëërë, synchronically equivalent to nërë + rë (the emphatic particle), but maybe diachronically related to the i-initial forms. Akuriyó nëërë seems to be the same (although it is not known if it has the same nërë + rë meaning). CARIBAN PRONOMINAL AND DEMONSTRATIVE SYSTEMS 269

Language Anaphoric Proximal Medial Distal N-Col. Col. N-Col. Col. N-Col. Col. N-Col. Col. Tiriyó21 nërë namo mëe mëesamo mëërë mëëjamo mëkï mëkïjamo ohkï ohkïjan Akuriyó22 nëërë namoro më(’)e mëtxamo mëkïrë më(ë)txamo Karihona nërë namoro mëhe mëkë mëkamoro Hixkaryana noro ñamoro mosonï moxamo mokro mokjamo mokï mokjamo Waiwai noro ñexamro moso moxam mïkro mïkjam mïkï mïkjam Katxuyana noro mosoro mokoro mokï Karinya-Hf inooro inaaro moose moojan moχko moχkaaro mookï moχkan Karinya-Ms ñooro ñoorokon mohse mohseekon mohko mohkaaro mookï mookïkon Apalaí ïnoro ïnaro mose moxiamo mokïro mokaro mokï mokamo Wayana inërë inamoro mëi, mëhe mëham mëkrë mëk mëkjam De’kwana23 në’dë nñanno më’dë kanno më’kï ma’kamo Yawarana Yukpa24 Waimirí25 mïkï Tamanaku nare motxe txamoro krere kiamoro make mukiamo Cumanagot metxe metxamo muekrere muek mukiamo Chayma metxe metxam(o) muekere muek mukiam(o) Pemón mesere itxamo(re) muere Taurepán mëserï, mësenï mësëmonan maarï Makushí mïserï insemoro mïïkïrï inkamoro Akawayo kïrë Ingarikó mëk(ï)re mïkamoro Arekuna mïserë mëitxamorï mïkrërë Panare më(i)’ mëhtxanton kën kamonton muku mukukon Ikpeng oren wam ugun ugjam Arara Bakairí inëra asaemo (awëkë) mëkë akaemo Kuhikuru ïele ese ekise Table 4. Cariban third-person pronouns: animate forms.

The animate proximal terms all seem to form a good cognate set. The initial con- sonant is clearly *m (which is, as expected, lost in Kuhikuru, and maybe also in

Bakairí, judging by the collective form). The second vowel is *o2, and the final vowel

*o3, given their different correspondences (ë : o and e : o). The intermediate consonant is a fricative, probably *tx (cf. *c in Girard 1971); notice, however, that Karihona h is an unexpected reflex (h in this language is supposed to come from *p; cf. Meira 2000). The Karinya-Ms form suggests the reconstruction of a *hs (or *htx)

21 The Tiriyó collective forms usually occur as mëesan, mëëjan, mëkïjan, ohkïjan (cf. fn. 15). 22 Meira (2000:60) listed më’etxamo, më’jamo, mëkïjamo as Akuriyó collective pronouns. More recent data (presented here) shows that these forms were mistaken (probably Tiriyó influence). 23 De’kwana në’dë is described as a distal form; the anaphoric pronoun is tïwï, a non-cognate. 24 Cf. fn. 16. 25 Bruno (pers. comm.) describes Waimiri as (surprisingly) lacking an animacy distinction. Irï (cf. Table 3) is also used to refer to people; and mïkï ‘that’ to inanimate objects. 270 MEIRA cluster, which is not a bad hypothesis; however, there is no evidence yet of a Proto- Cariban *h. The possibility of reconstructing two proximals, suggested by the two Wayana forms mëi and mëhe (for which no good semantic description is yet available), seems less likely: no other language has two forms, and, except for the Karihona mëhe (which cannot be cognate with Wayana mëhe, since Karihona h < *p and Wayana h < *tx; cf. Girard 1971), all forms look cognate (i.e. there do not seem to be two sets, but only one). For these reasons, the proximal form is here reconstructed as *mo2txo3, and the Karihona h is left unexplained. (The two Wayana fborms might come from combinations with non-deictic elements, e.g. particles; this is certainly the explanation for the ro, rï in Katxuyana, Taurepán, and Makushí, and possibly also for the nï in Hixkaryana and Taurepán). The animate medial and distal pronouns share suggestive similarities. Looking at cases such as Apalaí mokïro vs. mokï, Chayma muekere vs. muek, Waiwai mïkro vs. mïkï, etc., one has the impression that the distal terms are simply combinations of the medial term with a reflex of the emphatic particle *ro2. This is probably true diachronically, but it even may be true synchronically for some languages.26 For instance, it is not so hard to imagine Apalaí as having a single distal term mokï that, when co-occurring with the emphatic particle ro, is used for closer referents: the ‘closer range’ may be an effect of the semantics of the particle. The two plural forms mokamo (for mokï) and mokaro (for mokïro) are also as expected: with the total reduction of the final syllable mo, one would expect mokamo + ro > mokaro (though the failure of the vowel a to nasalize is unexpected); cf. also Karinya-Hf moχkan and moχkaaro. In Waiwai, there even is only one collective form mokjam corresponding to both the medial and the distal pronouns. All of this strongly suggests that Proto- Cariban did not have two non-proximal pronouns, but only one: all forms in the medial and distal columns of Table 4 would then belong to one cognate set. (The only problematic case is Karinya: moχko does not look like mokï + ro. One wonders if there could be a connection with the 1+2 pronoun kïχko). The form of this animate distal pronoun presents relatively few problems:

*mo2kï seems to be the best hypothesis. Almost all languages have an initial syllable mo, më, mue (e in Kuhikuru); it is easier to assume that Panare, Tamanaku and Ikpeng lost it. The final syllable kï, or clear reflexes of it (e.g. De’kwana ’, Tiriyó long ëë) are also overwhelming. The few problematic cases are: (1) Bakairí awëkë, which is not a clear cognate; (2) De’kwana më’kï ‘distal’, with an unexpected ’ (glottal stop); (3) Ikpeng oren, which may not be cognate; ugun, with loss of initial m, looks like a better canditate. Note than Panare kën comes from *mo2kï-ro2, with *ro2 > n (cf. the inanimate medial mën and the second-person pronoun amën above). The animate collective forms are also, as was the case with the non-third-person pronouns, formed with reflexes of the collective elements *jamo, *komo, and the

26 This is not, of course, true for all languages. In Tiriyó, mëërë and mëkï are semantically very distinct; they are clearly two lexical items (cf. Meira to ap-a). Panare kën and muku (also attested as mïkï) also seem to be clearly independent, at least formally. CARIBAN PRONOMINAL AND DEMONSTRATIVE SYSTEMS 271

particle *ro2. The anaphoric collective can be reconstructed as resulting from *(i)n + *jamo (without the final *ro2, since it does not occur in Tiriyó). It is not clear whether the palatalization in Hixkaryana ñamoro, Waiwai ñexamrom comes from the preceding *i (in which case one could reconstruct *(i)namo), or from the following *j (in which case one could reconstruct *(i)njamo). To keep both possibilities in mind, the formula *(i)n(j)amo will be used. The other collectives are again derivable from the non-collective stem plus a combination of collective markers and *ro2 (e.g. Panare kamonton < *jamo ro2 tomo). In Taurepán mësëmonan, there seems to be a new collective element nan (< *jamo nan). The problematic cases are: (1) Ikpeng wam, which might simply be the element *jamo, without any original stem, or else non-cognate; (2) Makushí insemoro, inkamoro with an initial unexplained i- (perhaps related to the (i)- in the anaphoric forms

(i)no2ro2, (i)n(j)amo). As was the case for the non-third-person pronouns, collective forms with *jamo possibly existed; they can be reconstructed as *mo2k-jamo and

*mo2tx-amo. They are reduced, since almost all reflexes are reduced; Tiriyó mëkïjamo may have been analogically rebuilt, apparently a frequent phenomenon in Tiriyó collectives — cf. the 1+2Col and 2Col forms. In fact, Tiriyó mëesamo also looks like an analogically rebuilt word, given the fact that it conserves an intervocalic reflex of *tx (cf. Meira 2000:31, 54 for the loss of intervocalic *tx in Tiriyó). A hypo- thesis would be: *mo2txamo > *mëtxamo; at this point *-txamo is reanalyzed as a suffix, while *mo2txo3 > mëe; then *mëe-txamo > mëesamo. As a final observation, it is interesting to note that, apparently, the most complicated Cariban demonstrative systems are found in the Guiana area (from Tiriyó to De’-kwana in the tables). As one moves away from this area, the systems become simpler: there may be no anaphoric term, and often only two distance terms (distal vs. proximal, without medial; e.g. Makushí, Kuhikuru).

4. Conclusion The Proto-Cariban pronouns reconstructed in the preceding two sections are summarized in Table 5 below.

Pers. N-Col. Col. Categ. Inanimate Animate N-Col Col

1 *awï Anaph *iro2 *(i)no2ro2 *(i)n(j)amo

2 *amo2 *am-jamo Prox-1 *o3ro2 *mo2txo3 *mo2tx-jamo

1+2 *kïCV *kïC-jamo Prox-2 *o2nï

1+3 *(ap)ina Med *mo2ro2 *mo2kï *mo2k-jamo

Dist *mo2nï Table 5. Proto-Cariban pronominal and demonstrative system.

The non-third-person pronouns form a typical Cariban system, with all categories duly represented. They correspond to the set of person-marking prefixes reconstructed by Gildea (1998:114) as *u- ‘1’, *a- ‘2’, and *k- ‘1+2’. 272 MEIRA

The set of third-person pronouns is also typical, despite the absence of a medial-distal distinction (which may be less frequent than the available descriptions suggest). Since most semantic analyses of demonstratives in Cariban languages are not very sophisticated, the meanings of the reconstructed terms are very approximative. In fact, the cognate sets were determined by how well their members fit the known correspondences in the family, rather than by putting together terms with the same gloss; especially for the older sources, glosses such as ‘this’, ‘that’, ‘este’, ‘ese’, ‘aquel’ etc. are not very trustworthy. The elements in Table 5 display certain recurrent similarities that lead to some speculative ideas. Certain elements can be identified — *ro2, *nï, *mo2, *kï — which suggest that the third-person pronouns are actually old combinations of yet older pro- nouns. The anaphoric *iro2 could be a combination of a third-person marker *i-

(from Gildea’s *jï-) with the element ro2, which could be the emphatic particle — i.e. ‘really third-person’. (This presupposes that the third-person prefix would have been an independent element in the past, so that it could be followed by the particle *ro2).

The element *mo2, also found in combination with *ro2, might be compared to the ‘evidential’ mo or më that, in some languages, occurs with the third-person prefix to indicate certain evidential values (e.g. Wayana nï-të-jai ‘he is going’, më-n-të-jai ‘he is going (but I do not see him)’; Hixkaryana mo-n-eweh-no ‘he took a bath (out of sight)’). More comparative research should help decide how much truth there is in such speculations.

References Abbott, Miriam 1991 ‘’, in Desmond C. Derbyshire & Geoffrey K. Pullum (eds.), Handbook of Amazonian languages, vol. 3, Berlin: Mouton de Gruyter, pp. 23-160. Amodio, Emanuele, & Vicente Pira 1996 Língua Makuxi — Makusi Maimu. Guia para a aprendizagem e dicionário da língua makuxi, Boa Vista (Brazil): Diocese de & Missionários de Scarboro (Canada). Armellada, Cesáreo de, & Jesús Olza 1994 Gramática de la lengua pemón (morfosintaxis), San Cristóbal (Venezuela): Universidad Católica del Táchira. Bruno, Ana Carla dos Santos 1996 Guia de aprendizagem da língua Waimirí-atroari, Manaus: Programa Waimirí-Atroari, Convênio FUNAI / Eletronorte. Derbyshire, Desmond C. 1979 Hixkaryana, Lingua Descriptive Studies, vol. 1. Amsterdam: North-Holland. 1985 Hixkaryana and Linguistic Typology, Dallas: Summer Institute of Linguistics and the University of Texas at Arlington. CARIBAN PRONOMINAL AND DEMONSTRATIVE SYSTEMS 273

1999 ‘Carib’, in R. M. W. Dixon & Alexandra Y. Aikhenvald (eds.), The Amazonian languages, Cambridge Language Surveys, Cambridge: Cambridge University Press, pp. 23-64. Edwards, Walter F. 1977 An introduction to the Akawayo and Arekuna peoples, Amerindian Languages Project, Georgetown: University of . Gildea, Spike 1995 ‘A comparative description of syllable reduction in the Cariban language fam- ily’, International Journal of American Linguistics 61:62-102. 1998 On reconstructing grammar: comparative Cariban morphosyntax. Oxford Studies in Anthropological Linguistics, vol. 18. Oxford: Oxford University Press. Gilij, Felipe [Filippo] S. 1965 Ensayo de historia americana, 3 vols, transl. by Antonio Tovar, Caracas: Aca- demia Nacional de la Historia (original 1782, Saggio di Storia Americana, 4 vols., Rome). Girard, Victor James 1971 Proto-Carib phonology, Berkeley (CA): University of California Ph.D. disser- tation. Hall, Katherine L. 1988 The morphosyntax of discourse in De’kwana Carib, Saint Louis: Washington University Ph.D. dissertation. Hawkins, Robert E. 1998 ‘Waiwai’, in Desmond C. Derbyshire & Geoffrey K. Pullum (eds.), Handbook of Amazonian languages, vol. 4, Berlin: Mouton de Gruyter, pp. 25-224. Hoff, Berend J. 1968 The , The Hague: Martinus Nijhoff. 1990 ‘The non-modal particles of the Carib language of Surinam and their influence on constituent order’, in Doris L. Payne (ed.), Studies in Lowland South Ame- rican Languages, Austin: University of Texas Press, pp. 495-541. Jackson, Walter S. 1972 ‘A Wayana grammar’, in Joseph E. Grimes (ed.), Languages of the Guianas, Norman: Summer Institute of Linguistics and University of Oklahoma Press, pp. 47-77. Kaufman, Terrence K. 1994 ‘The native languages of South America’, in Christopher Moseley and R. E. Asher (eds.), Atlas of the world’s languages, London, New York: Routledge, pp. 46-76. Koch-Grünberg, Theodor 1916 Vom Roroima zum Orinoco. Ergebnisse einer Reise in Nordbrasilien und Venezuela in den Jahren 1911-1913, vol. 4: Sprachen, Stuttgart: Strecker u. Schröder. 274 MEIRA

Koehn, E., & S. Koehn. 1986 ‘Apalai’, in Desmond C. Derbyshire & Geoffrey K. Pullum (eds.), Handbook of Amazonian languages, vol. 1, Berlin: Mouton de Gruyter, pp. 33-127. Meira, Sérgio 1998 ‘Rhythmic stress in Tiriyó (Cariban)’, International Journal of American Lin- guistics 64:352-378. 1999 A grammar of Tiriyó, Houston (USA): Rice University Ph.D. dissertation (to appear in the Mouton grammar series). 2000 A reconstruction of Proto-Taranoan: phonology and morphology, München: LINCOM Europa. to ap-a ‘Non-contrastive exophoric use of Tiriyó demonstratives’, in Michael Dunn & Sérgio Meira (eds.), Demonstratives in Cross-Linguistic Perspective. to ap-b ‘Les démonstratifs proximaux non-animés de la langue tiriyo (caribe): une étude de corpus’, Amérindia. Méndez-Arocha, Alberto 1959 ‘Un vocabulario yabarana con apuntes fonémicos’, Antropológica (Venezuela) no. 7:65-84. Mosonyi, Jorge C. 1978 Diccionario básico del idioma cariña, Thesis, Caracas: Universidad Central de Venezuela. Muller, Marie-Claude Mattei 1994 Diccionario ilustrado panare-español, español-panare: un aporte al estudio de los Panares-E’ñepa, Caracas: Comisión Quinto Centenario, Gráficas Armitano. Pacheco, Frantomé Bezerra 1997 Aspectos da gramática ikpeng (karíb). M.A. Thesis, Campinas (Brazil): Uni- versidade de Campinas (UNICAMP), Instituto de Estudos da Linguagem. Robayo, Camilo 1987 Le système des personnes de la langue carijona, DEA Thesis, Paris: Université de Paris VII - Jussieu. 2000a ‘Avance sobre morfología carijona’, in María Pérez & María Montes (eds.), Lenguas indígenas de Colombia: una visión descriptiva, Bogotá: Instituto Caro y Cuervo, pp. 171-180. 2000b ‘Introducción al estudio de la lengua yupo o yukpa’, in María Pérez & María Montes (eds.), Lenguas indígenas de Colombia: una visión descriptiva, Bogotá: Instituto Caro y Cuervo, pp. 709-717. Ruiz Blanco, Matias 1690 Arte y tesoro de la lengua cumanagota, Madrid (facsimilar edition by Julio Platzmann, Leipzig: B. G. Teubner, 1888). Souza, Isaac de 1983 Arara - Karíb. Relatório para a FUNAI. (Includes unpublished material — hand-outs — from presentations at two conferences, dated 1992). CARIBAN PRONOMINAL AND DEMONSTRATIVE SYSTEMS 275

Steinen, Karl von den. 1892 Die Bakaïrí-Sprache: Wörterverzeichnis, Sätze, Sagen, Grammatik (mit Bei- trägen zu einer Lautlehre der Karaïbischen Grundsprache), Leipzig: K. F. Koehler’s Antiquarium. Tauste, Francisco de 1680 Arte y bocabulario de la lengua de los indios chaymas, cumanagotos, cores, parias, y otros diversos de la provincia de Cumana, o Nueva Andalucia, Madrid (facsimilar edition by Julio Platzmann, Leipzig: B. G. Teubner, 1888). Vegamián, Felix M. 1978 Diccionario ilustrado yupa-español, español-yupa (con onomástica y apuntes gramaticales), Caracas: Formateca C.A. Wheatley, James 1973 ‘Pronouns and nominal elements in Bacairi discourse’, Linguistics 104:105- 115. 1978 Vocabulário Bakairí-Português, Português-Bakairí (edição experimental), Brasília: Summer Institute of Linguistics. Yangues, Manuel de 1683 Principios y reglas de la lengua cummanagota, Burgos (facsimilar edition by Julio Platzmann, Leipzig: B. G. Teubner, 1888).

LE PARFAIT EN KALI’NA

Odile Renault-Lescure Institut de Recherche pour le Développement

1. Introduction Les Kali’na de la Guyane française, ou Galibi comme ils y ont été désignés jusqu’à une époque récente, parlent une langue de la famille caribe (ou karib). Cette famille comprend une trentaine de langues parlées dans de vastes régions au sud et au nord de l’Amazone, plutôt dans la partie orientale de l’Amazonie, bien qu’une des langues, le carijona, soit parlé à l’ouest, en Colombie et une autre, le yupka, à la frontière nord entre Colombie et Venezuela. Dans le nord du bassin amazonien, elles s’étendent jusqu’à la côte de la mer caraïbe. Au sud, elles sont limitées à la vallée du Xingu, affluent de l’Amazone. Parmi toutes ces langues, le kali’na est aujourd’hui encore probablement celle qui a le plus de locuteurs1, et celle qui a la plus large extension géographique, depuis les savanes nord-orientales du Venezuela jusqu’au nord de l’Amapá, au Brésil. Les données qui sont à la base de cet article ont été recueillies en Guyane française et représentent la variété dialectale la plus orientale de la langue.2

2. Construction du verbe fini Le verbe en fonction prédicative est formé d’une base verbale préfixée d’un indice personnel et suffixée d’une marque de temps, mode ou aspect. La valence des verbes 3 est à 1 ou 2, les verbes intransitifs sont à un actant, les verbes transitifs à deux .

2.1. L’énoncé transitif L’énoncé prototypique transitif est formé d’un verbe, d’un indice actanciel préfixé et d’une marque de temps, aspect ou mode suffixée:4

(1) ni-melo-i 3P-décorer-PARFT ‘Il (elle) l’a décoré(e).’

1 Les données dont on dispose permettent d’indiquer une population en Guyane française dont le nombre se situe entre 2.000 et 4.000 locuteurs, et une population globale estimée à 25.000 personnes. Mes remer- ciements vont à Jean Appolinaire pour son travail d’informateur. 2 C'est essentiellement sur la variété décrite par Hoff au Surinam, appelée carib, que s'appuient, par exemple, les travaux de comparaison et de reconstruction de Gildea (1998). 3 En raison des pressions exercées par les rôles sémantiques et les statuts pragmatiques, nous adopterons une terminologie qui souligne les rôles sémantiques. Pour une description des relations grammaticales en kali'na de Guyane française, voir Renault-Lescure (sous-presse). 4 AC = Accompli; ATT = Attention; DEM.ANIM = Démonstratif animé ; DEM.INAN = Démonstratif inanimé; FACT = Factitif; FUT = Futur; IMP = Impératif; INF = Infinitif; INT = Intensif; INTER = Interrogatif; INTERJ = Interjection; NEG = Négation; PARFT = Parfait; PART = Participe; PLUR = Pluriel; PRES = Présent; REL = Relateur; 1(2,3) = 1ère (2ème, 3ème) personne actant unique; 1(2,3)A = 1ère (2ème, 3ème) personne agent; 3P = 3ème personne patient. 278 RENAULT-LESCURE

L’indice actanciel 3P réfère sémantiquement à l’agent et au patient; morphologique- ment, seul le patient est marqué. Les indices personnels préfixés aux verbes bi- actanciels marquent explicitement l’un des actants. Dans l’énoncé suivant:

(2) s-eyuku-i 1A-inviter-PARFT ‘Je l’ai invité(e).’ eyuku est le radical verbal, -i la marque de parfait et s- la marque de personne. Le verbe est bivalent et comporte deux actants dont un seul présente une expression grammaticale. Comme on pourra le vérifier plus loin, s- est une marque de première personne. En effet, le choix s’opère suivant une hiérarchie des personnes qui réalise une distinction fondamentale entre les personnes de l’intralocution et la personne de l’extralocution: les personnes de l’intralocution sont situées plus haut sur l’échelle des personnes que les personnes extralocutives:

1, 2, 1+2 3

Cela se traduit par l’effacement systématique des 3èmes personnes lorsqu’elles sont en compétition avec une autre personne 1ère , 2ème ou 1ère inclusive, quel que soit le rôle sémantique de la personne. La hiérarchie des personnes l’emporte ici sur celle qui serait liée aux rôles d’agent et de patient.

Personne Marque d’Agent (A) Marque de Patient (P) 1 s- y- 2 m- ay- 1+2 kVs-5 k- Tableau des indices de personne des verbes biactanciels.6

(3) eyuku ‘inviter quelqu’un’

Marque d’Agent: Marque de Patient:

s-eyuku-i ‘je l’ai invité(e)’ y-eyuku-i ‘il/elle m’a invité(e)’ m-eyuku-i ‘tu l’as invité(e)’ ay-eyuku-i ‘il/elle t’a invité(e)’ kes-eyuku-i ‘toi et moi, nous k-ayuku-i7 ‘il/elle nous (toi et moi) l’avons invité(e)’ a invité(e)s’

5 V indique une assimilation de la voyelle du préfixe à la voyelle initiale du radical verbal. 6 Je n’indique pas les variantes morphophonologiques qui ne sont pas pertinentes ici. 7 Le passage de la voyelle e à a est une variante morphophonologique régulière. PARFAIT EN KALIN’A 279

Aucune hiérarchisation n’apparaît par contre entre les personnes de l’intralocution, c’est à dire la première personne et la deuxième personne:

Personne Neutralisation de l’opposition A/P 1 > 2, 2 > 1 k-

(4) k-ayuku-i 1A2P/1P2A-inviter-PARFT ‘je t’ai invité(e)’, ‘tu m’as invité(e)’

Lorsque les deux participants sont de troisième personne, une marque spécifique est préfixée qui renvoie au patient:

Personne P 3 kVn- /n-

(5) n-eyuku-i 3P-inviter-PARFT ‘il/elle l’a invité(e)’

(6) ken-eyu’-san8 3P-inviter-PRES ‘il/elle l’invite’

Cette marque est phonologiquement vide lorsque le patient est explicité lexicalement mais présente lorsque l’agent est instancié par un nom (pronom, SN):

(7) tanpoko dudi eyuku-i vieux Dudi inviter-PARFT ‘Il/elle a invité le vieux Dudi.’

(8) tanpoko dudi n-eyuku-i vieux Dudi 3A-inviter-PARFT ‘Le vieux Dudi l’a invité(e).’

2.2. L’énoncé intransitif L’énoncé prototypique intransitif est formé d’une base verbale, d’un indice actanciel préfixé qui réfère à l’actant unique et d’une marque de temps, aspect ou mode suf- fixée.9

8 La chute de la syllabe finale du radical verbal devant la marque du présent est indiquée othographique- ment par l'. 280 RENAULT-LESCURE

10 (9) n-opˆ-i 3-arriver-PARFT ‘Il est arrivé.’

Deux séries d’indices actanciels servent à marquer l’actant unique de ces verbes:

Personne Actant unique (A) 1 O- y- 2 m- ay- 1+2 KVt- k- 3 n- /kVn- Tableau des indices de personne des verbes uniactanciels

(10) waimoikˆ ‘embarquer’ awa ‘rire’

∅-waimokˆ-i ‘j’ai embarqué’ y-awa-i ‘j’ai ri’ m-aimokˆ-i ‘tu as embarqué’ ay-awa-i ‘tu as ri’ kat-aimokˆ-i ‘toi et moi, nous k-awa-i ‘toi et moi, avons embarqué’ nous avons ri’ n-aimokˆ-i ‘il/elle a embarqué’ n-awa-i ‘il/elle a ri’

L’une des séries présente des similitudes avec le paradigme des marques de personne agent des verbes biactanciels, l’autre série est identique aux marques de personne patient du verbe bi-actanciel; les deux séries présentent cependant une marque unique de troisième personne:

Personne Agent Actant unique Patient Actant unique 1 s- O- y- y- 2 m- m- ay- ay- 1+2 kVs- kVt- k- k- 3 kVn-/n- Tableau comparatif

9 Certaines marques de TAM varient (a) suivant le type de radical verbal auquel elles sont associées, (b) uivant les personnes indiciées, en fonction de l’intra- ou extralocution, (c) en fonction de la valence: (a) s-alo-ya ‘Je l’emporte.’ (radical alo) (b) s-ema-e ‘Je le jette.’ s-ema-e ‘Je le jette.’ (radical ema) ken-ema-no ‘Il le jette.’ s-uku’-sa ‘Je le connais.’ (radical ukutˆ) (c) s-uku’-sa ‘Je le jette.’ (valence 2) si-poi-ya ‘Je le plante.’ (radical pomˆ) ∅-wˆ’-sa ‘J’y vais.’ (valence 1) 10 La transcription utilise la proposition graphique retenue par les Kali’na. On notera essentiellement que le phénomène de la palatalisation affecte presque toutes les consonnes (les occlusives, les nasales et les semi-consonnes) qui suivent une voyelle i (ou la précèdent pour s), que les nasales finales ont une pronon- ciation affaiblie, la nasalité se propageant sur les voyelles précédentes, que l’apostrophe désigne soit une glottale ou un allongement vocalique, comme dans na’na ‘nous (exclusif)’, soit la chute syllabique d’une syllabe finale d’un radical verbal (pˆ, tˆ, kˆ, ku / _#). La sonorisation des occlusives sourdes est soit liée à la prosodie, soit en variation libre. PARFAIT EN KALIN’A 281

3. Le parfait 3.1. Morphologie A la différence des autres marques de temps et d’aspect,11 la marque de parfait ne présente aucune variation de forme:

Radical verbal Morphème Tous les radicaux -i

3.2. Valeurs temporelles Le parfait illustre bien la combinaison de valeurs aspectuelles et temporelles. On pourrait le qualifier comme Feuillet (1988: 92) de ‘temps-charnière dans la mesure où il est un accompli (et exprime par conséquent un procès achevé au moment de l’énonciation) tout en appartenant à la sphère de la non-distanciation. Il est tiraillé entre le passé pur et simple et le présent.’ Les valeurs du passé exprimé dans le par- fait du kali’na sont celles de passés récents.12 L’exemple suivant décrit une situation dans laquelle une personne qui vient d’arriver dans une maison se voit poser la question:

(11) m-opˆ-i 2-arriver-PARFT ‘Tu viens d’arriver?’

Le parfait a ici une valeur temporelle de passé immédiat. Dans l’exemple suivant, la valeur temporelle, renforcée par l’emploi d’un circonstant, indique un passé récent:

(12) n -opˆ -i tela koinalo 3-arriver-PARFT déjà hier ‘Tu es déjà venu hier?’

Dans tous les cas, le passé ne pourra être antérieur au temps de vie du locuteur. L’évocation au parfait ne peut dépasser les limites de l’expérience personnelle. Dans l’exemple ci-dessous, le narrateur évoque un souvenir d’enfance:

(13) kaleta ta s-ene-i newala pˆime nokan livre dans 1A-voir-PARFT comment beaucoup animaux

t-ainka-po i-’wa man PART-rentrer-FACT 3PERS-à 3+être+PRES ‘Je l’ai vu dans les livres, comment il [Noé] a fait rentrer beaucoup d’animaux sauvages.’

11 Voir Hoff (1968) et Renault-Lescure (1999). 12 C'est la valeur que lui assigne Courtz (s.d.); Hoff (1968) ne lui donne aucune valeur temporelle. 282 RENAULT-LESCURE

3.3.Valeurs aspectuelles La valeur fondamentale du parfait est la résultativité telle que définie par différents auteurs (Comrie 1976, Desclés 1989, Guentcheva 1990). Il dénote une situation dans laquelle un état est le résultat d’un processus antérieur qui a été arrêté. Il sera repré- senté figurativement par le diagramme suivant: un processus achevé indiqué par l’intervalle de gauche, fermé à gauche et à droite, suivi d’un intervalle hachuré ouvert à droite, qui est la zone de l’état résultant:

------[------]/////////////////////////////[To

Observons les exemples ci-dessous:

(14) mo’ko mˆlekoko n-opˆ-i DEM.ANIM garçon 3-arriver-PARFT ‘Le garçon est arrivé.’

(15) penalo kolomelu n-emopotˆ-i ene-ko te tout à l’heure tonnerre 3-gronder-PARFT voir-IMP donc

moe tˆkalaiye moe kˆn-ˆ’-san osa po là-bas noir là-bas 3-aller-PRES Laussat à ‘Tout à l’heure le tonnerre a grondé, vois donc ! c’est noir là-bas, il pleut sur Laussat.’

(16) yaya kˆn-ka-no ‘oto ko mai ko isuwi’ mon frère 3-dire-PRES quoi INTER 2+être+PARFT ATT sœur

kˆn-ka-no ‘yaya na’na n-opekˆ-i yaya’ ∅-ka-e 3-dire-PRES mon frère nous.EXCL 3-couler-PARFT mon frère 1-dire-PRES ‘Eh bien, que t’est-il arrivé ma sœur? a demandé mon frère. On a coulé mon frère! ai-je dit.’

Le parfait y décrit un processus achevé dont le résultat est observable au moment de l’énonciation. En (14) le garçon est arrivé et il est encore là, on peut l’observer, en (15) après l’orage, on peut voir le ciel noir et la pluie qui tombe, en (16) la pirogue a coulé et la suite du récit en explicite le résultat visible: les rescapés sont mouillés, leurs affaires sont trempées, la farine de manioc imbibée d’eau… Outre la valeur fondamentale décrite ci-dessus, des nuances variées vont appa- raître au fil d’autres énoncés:

PARFAIT EN KALIN’A 283

(17) ‘i-to-’pa ∅-wai-take’ kˆn-ka-no ‘pasque samedi dimanche NEG-aller-NEG 1-être-FUT 3-dire-PRES parce que samedi dimanche

dimanche n-emamina-i da kolopo pêcher poko tela dimanche 3-travailler-PARFT alors demain pêcher occupé à déjà

t-ˆto-lˆ se man 3-aller-INF avec la volonté de 3+être ‘Je n’irai pas, dit-il. Comme il a travaillé samedi et dimanche, alors, de- main, il veut aller à la pêche.’

De cet exemple se dégage une valeur ‘avoir fini de’. L’homme dont il est question a fini de travailler et souhaite passer à un autre type d’activité. L’accent est ici porté sur le processus achevé. Dans l’exemple suivant, le résultat montre qu’un objectif visé est atteint:

(18) ilonpo ka’pa lo y-asalˆ-kon ∅-wˆto-n lo ene alors vraiment INT 1-ami-PLUR 1-aller-PARFT INT voir

molo wonumekatopo-npo si-kapˆ-i DEM.INAN rêve-ancien 1A-réaliser-PARFT ‘C’est alors, mes amis, que je suis allé les voir [les pièces de musée], j’ai réalisé ce vieux rêve.’

Contrairement à ce que nous avons vu en (17), le parfait met ici l’accent sur le résul- tat atteint. Procédons maintenant à l’examen des exemples suivants:

(19) olosi telapa si-melo-i montre déjà 1A-dessiner-PARFT ‘J’ai déjà dessiné la montre.’

(20) kaleta s-epekatˆ-i cahier 1A-acheter-PARFT ‘J’ai acheté un cahier.’

(21) otˆ ko mi-tanpo-i quoi INTER 2A-renverser-PARFT ‘Qu’est-ce que tu as renversé?’

Le parfait y décrit bien un processus achevé dont le résultat est observable au mo- ment de l’énonciation. A la différence des exemples précédents, on y observe un pro- cessus de transformation mené jusqu’à son terme. L’état qui en est issu affecte un 284 RENAULT-LESCURE patient à la suite de la transformation opérée par un agent. Cette valeur du parfait correspond à ce que Guentcheva (1990) appelle, dans sa classification terminologique des valeurs du parfait, un ‘état acquis’. Une autre valeur fondamentale du parfait est représentée par la valeur d’expérience. Ce parfait d’expérience (Comrie 1976) se réfère à une situation dans laquelle un événement ou plusieurs, interrompus par le dernier de la série, aboutissent à un état résultant dont la propriété est l’expérience nouvelle acquise par le sujet: l’expérience des évènements vécus par lui-même (Desclés 1989). Le diagramme qui le représente se compose d’un intervalle fermé à gauche et à droite, dans lequel appa- raît la séquence d’évènements, et une zone hachurée, ouverte à droite, correspondant à l’expérience acquise:

------[x--x--x------x]/////////////////////////[------To

(22) kaleta ta s-ene-i newala pˆime nokan livre dans 1A-voir-PARFT comment beaucoup animaux

t-ainka-po i-’wa man PART-rentrer-FACT 3PERS-à 3+être+PRES ‘Je l’ai vu dans les livres, comment il [Noé] a fait rentrer beaucoup d’animaux sauvages.’

(23) molo-kon y-asakalˆ-kon s-ene-i ase’ke y-enu-lu ke DEM.ANIM-PLUR 1-ami-PLUR 1A-voir-PARFT propre 1-œil-REL avec ‘Mes amis, je les ai vues [les pièces de musée] de mes propres yeux.’

(24) ∅-ene-’po melo y-asakalˆ-kon po’tome imelo ∅-wonumenka-i 3P-voir-AC après 1-ami-PLUR grand très 1-avoir nostalgie-PARFT ‘Après les avoir vues [les pièces de musée], j’ai ressenti une très grande nostalgie.’

Dans chacune des occurences ci-dessus, la manière dont l’expérience est acquise est explicitée. En (22) un savoir a été acquis par la lecture de livres d’images, en (23) l’expérience a été acquise directement, de façon perceptible. On pourrait donner ici une valeur épistémique au parfait, car nous sommes dans le domaine du certain, du constaté. En (24) l’expérience est provoquée par la vue d’objets traditionnels. Dans ce dernier énoncé, la valeur d’expérience ne semble guère se différencier de la valeur résultante. Les deux notions se recouvrent souvent. Le contexte de l’énonciation est alors parfois à même de permettre l’une ou l’autre des interprétations. C’est ce qui se passe dans l’exemple qui suit:

PARFAIT EN KALIN’A 285

(25) ˆ’me s-utaka-i mon fils 1A-perdre-PARFT ‘J’ai perdu mon fils.’

Le contexte dans lequel cet exemple a été relevé indique qu’il s’agit d’un parfait d’expérience: une femme parle de la perte de son fils, de cette douloureuse expé- rience dans sa vie. Mais cet énoncé, proféré dans un autre contexte, pourrait indiquer typiquement un état résultant et se référer à l’état de deuil dans lequel se trouverait une mère à la suite du décès de son enfant. Une autre valeur du parfait se trouve illustrée en (26). Un homme, visitant son abattis, y découvre déjà mûres les bananes qu’il n’y a pourtant que plantées la veille. Le parfait, renforcé par l’interjection, dans ce cas, exprime la surprise:

(26) kosi koinalo loten si-pomˆ-i palulu INTERJ hier seulement 1A-planter-PARFT banane ‘Ça alors! Les bananes, je ne les ai plantées qu’hier!’

C’est aussi le cas de l’exemple suivant qui présente une situation inattendue. Dans une course entre le vent et la tortue, celle-ci, malgré sa lenteur proverbiale, arrive la première:

(27) ∅-tunta-i te iseme t-ˆka man wayamˆ 1-arriver-PARFT mais pourtant PART-dire 3+être tortue ‘Et pourtant, je suis arrivée, dit la tortue.’

3.4 Valeur modale Observons les deux énoncés qui suivent:

(28) s-apo-i 1A-goûter-PARFT ‘Je veux le goûter!’

(29) kaleta s-epekatˆ-i – ne ko s-ene-i cahier 1A-acheter-PARFT – justement ATT 1A-voir-PARFT ‘J’ai acheté un cahier!’ – ‘Je veux justement le voir!’

La valeur du parfait qui s’en dégage relève du domaine du modal.13 Il exprime une intention, une visée à atteindre très fortement marquée. Une traduction peut-être plus adéquate y ajouterait une valeur injonctive ‘Fais-moi goûter !’, ‘Fais-moi voir !’.14

13 Décrit comme un optatif par Hoff (1968 et 1986). 14 Dans la mesure où cette valeur modale est très différente des autres valeurs du parfait, il est peut-être intéressant de se poser la question d'une possible forme indépendante homophone. Cette valeur modale ne semble apparaître qu'à la première personne et pourraît alors correspondre à un impératif de première 286 RENAULT-LESCURE

4. Conclusion Comme dans bien d’autres langues, le parfait recèle en kali’na une grande variété de valeurs. Les valeurs temporelles s’articulent autour d’un temps plus ou moins récent mais qui est toujours celui de l’expérience vécue. Une des valeurs aspectuelles fon- damentales est celle que l’on observe dans le parfait d’expérience. Quant à la valeur modale, elle peut se concevoir comme une expérience à atteindre. On pourra dès lors se demander si cette notion d’expérience qui traverse le système TAM du parfait n’en forme pas la pierre angulaire, le lien qui permet de passer du temporel à l’aspectuel puis au modal.

Références bibliographiques Comrie, B. 1976 Aspect: an introduction to the study of verbal aspect and related problems, Cambridge: Cambridge University Press. Courtz, H. s.d. De Karaïbse Taal. Ms. Desclés, J.P. 1989 ‘State, event, process, and typology’, General Linguistics, vol. 29, 3: 159-200. Feuillet, J. 1988 Introduction à l’analyse morphosyntaxique, Paris: Presses Universitaires de France. Gildea, S.

1998 On reconstructing grammar: comparative Cariban morphosyntax, New York / Oxford : Oxford University Press. Guentcheva, Z. 1990 Temps et aspect: l’exemple du bulgare contemporain, Paris: Editions du CNRS. Hoff, B.J. 1968 The Carib language: phonology, morphonology, morphology, texts and word index, The Hague: Martinus Nijhoff. 1986 ‘ in Carib: particles, affixes, and a variant of Wackernagel’s law’, Lingua 69: 49-103. Renault-Lescure, O. 1999 ‘Le dispositif aspecto-temporel des verbes finis en kali’na oriental (langue caribe de Guyane française)’, Actances 10: 163-176. s.p. ‘Dynamique des relations actancielles en kali'na de Guyane française (ou gali- bi)’, Langues de Guyane, Goury L. (ed.), Amerindia 26.

personne marqué /s- -i/ (suggestion de S. Meira). Elle se rapprocherait de la forme du vétatif /kˆs- -i/, exemple s-ene-i "Ne le regarde pas!". GENDER AGREEMENT IN MOSETÉN

Jeanette Sakel MPI Leipzig

1. Introduction1 Mosetén (Chimane, Tsimane’) is a from eastern Bolivia. The lan- guage consists of three dialects, Mosetén de Covendo, Mosetén de Santa Ana, and Chimane.2 There are about 5000 speakers, of which the majority speaks Chimane. My research is based on the dialect of Mosetén de Covendo, which has only about 500 speakers.3 Derbyshire and Payne (1990) discuss noun classification systems as a common structure in Amazonian languages. They distinguish three different kinds of classifi- ers: numeral classifiers, concordial classifiers and verb-incorporated classifiers. Gen- der systems are treated under the heading of concordial classifiers. While most classi- fier systems have a semantic base for assignment, such as classification according to size, shape, function or the like, gender assignment may in some cases not be com- pletely semantically based. Mosetén has a gender system, which is rather arbitrarily assigned. There are two genders: masculine and feminine. Female humans are in the feminine gender and male humans are in the masculine gender. The rest of gender assignment does not seem to follow any semantic rule.4 In most languages with a gender system, there is some kind of gender agreement in the noun phrase. Often, determiners and modifiers agree in gender with the head noun. In some languages, gender agreement is furthermore found outside the noun phrase, as for example in the Caucasian languages Lak and Abkhaz, discussed by Corbett (1991). In Mosetén, as well, gender agreement affects many different ele- ments, from now on called targets, in the clause. As gender agreement is very fre- quent, it is exceptional for a clause in Mosetén not to have any marking for gender at all.

1 I want to thank Mily Crevels, Simon van de Kerke and Hein van der Voort for their review of this pa- per. 2 Often, Mosetén and Chimane are considered to be different languages, forming a small language fam- ily. The self-consciousness of the Mosetenes and the Chimanes as separate groups supports this argument. Linguistically, however, Mosetén de Santa Ana and Chimane are more closely related than the two Mo- setén dialects. Furthermore, all three dialects are mutually intelligible. I therefore choose to consider these as three dialects of one language. 3 The data presented in this linguistic work are from my own fieldwork in Bolivia 1999-2000. I want to thank my consultants Juan Huasna, Cleto Tahe, Adrian Topepe, Lidia Misange, as well as those who con- tributed to this linguistics work by telling stories. 4 In some cases, phonology and certain nominalization indicate specific gender assignment. For example all nouns ending in the nominalizer –dye’ are feminine. 288 SAKEL

I will look at the different forms of gender agreement, starting with the noun phrase, the prototypical locus for gender. I will then gradually move on to structures where we do not usually expect to find this form of agreement.

2. Nouns As in many other languages, gender in Mosetén is an inherent feature of nouns, with- out these usually having special gender marking themselves. Corbett (1991) calls this a covert gender system, as no phonological form or marker on the noun itself predicts the gender of this noun. There are, however, some nouns in Mosetén usually referring to humans or ani- mals which have different forms for each gender. These nouns have certain traits in common: the feminine forms often end in –si’ and the masculine forms in –tyi’. Some of these nouns might also have a shorter variant, so nanasi’ ‘girl’ below can also be nanas, and nanatyi’ ‘boy’ nanaty:

(1)e nana-si’ girl (2)e nana-tyi’ boy (3)e min-si’ female human being (4)e min-tyi’ masculine human being

Other nouns have suppletive forms for feminine and masculine gender:

(5)e phen woman, wife (6)e soñi’ man, husband

3. Pronouns Personal pronouns in the 3rd person are marked for gender, while 1st and 2nd person pronouns are not. The 3rd person masculine pronoun is mi’ and the feminine pronoun is mo’. In the plural, the clitic in, which is the general marker of plurality in Mosetén, is added to these pronouns, and thus gender is also apparent in the plural:

(7)e mo’ (3F) (8)e mi’ (3M) (9)e mo’-in (3F-PL) (10)e mi’-in (3M-PL)

Demonstrative pronouns also distinguish between feminine and masculine gender. These are the suppletive forms öij for the feminine and its for the masculine. Again, these can be pluralized with the clitic in:

(11)e öij (DEM.F) öij-in (DEM.F-PL) (12)e its (DEM.M) its-in (DEM.M-PL) MOSETÉN GENDER AGREEMENT 289

The possessive pronouns are formed by adding the gender markers –tyi’ (M) and -si’ (F) to the personal pronoun. They show gender agreement to the possessed element:

(13)e yäe-si’ phen 1SG-F woman ‘my wife’

(14)e yäe-tyi’ mama’ 1SG-M father ‘my father’

When the possessor is a 3rd person pronoun, there are two kinds of gender agreement in the possessive pronoun: agreement with the possessor, which is apparent in the personal pronoun-part of the possessive pronoun and agreement with the possessum in the added gender marker:

(15)e mi’-si’ phen 3M.SG-F woman ‘his wife’

There is a way of marking possession other than by the possessive pronoun, namely by cliticizing a personal pronoun to the possessum as a postclitic. This clitic usually agrees in gender with the possessor:

(16)e Mo’ jiri-s-tom aka’, jiri-ty waemtyi’-mo’ 3F.SG one-F-COM house one-M man-3F.SG

jiri-ty äwä’ soñi’-tyi’. Me’ momo’. one-M child man-M so only.F ‘She has a house, a husband (is hers), a masculine child. That’s all.’

Usually, such a person clitic agrees in gender with the possessor, but there are exam- ples where it agrees with the possessum. This is a phenomenon that most consultants deny in elicitation, although it appears in the use of their language. Most of all, it seems to be used by younger speakers, which could mean that this has to do with language change. I also have few examples, however, where this appears in the speech of an older speaker. Usually, when asked again, the informants wonder about the forms, try the masculine and the feminine form and are not sure which form to use. In free speech, the form agreeing with the possessum often simply occurs. The following example is given by a young speaker of Mosetén:

290 SAKEL

(17)e Öij phen faraj-ye-i mo’ lotej. Khin’ DEM.F woman leave-VB-3SG.F.O 3 F.SG plantation.E now

äwä’-mi’ soñi-tyi’ mi’-mimi’ karij-tye-’. child-3M.SG man-M 3M.SG-only.M work-APPL-3SG.F.O ‘This woman has left her plantation. Now only her masculine child (i.e. her son) works (on) it.’

In what most Mosetenes would consider the correct form, the person clitic should be –mo’ here (i.e. in the feminine) and not -mi’ in the masculine gender, i.e. it ‘should’ agree with the possessor and not with the possessum. Since I also have examples of older people using this form, it might not be a recent change in the language, even if it is rather infrequent as opposed to the cases where this clitic agrees with the possessor. In the following example, a 63-year-old speaker uses this form:

(18)t …yi-katyi’-khä mi’ aj jen’-mi’ dyaba-tyi’ say.M.S-EVID-EVID 3M.SG ASP father-3M.SG peanut(F)-M

jam aj tye-te aj-win. NEG ASP give-3SG.M.O ASP-before ‘… said the father of the peanut, he did not give (it) to him (any more).’

Another pronoun with different gender forms is the reference tracking pronoun ‘other’. This has the form yoktyi’ in the masculine gender and yoksi’ in the feminine gender. Again, the gender marking morphemes are -tyi’ (M) and –si’ (F):

(19)t Kiwïj yok-si’ pheya-k-dye’ yäe se’we-’ dyiñae-ij again other-F say-PASS-NOM 1SG hear-3SG.F.O trap-VB.M.S

jedye’ dyash dyiñae’. thing QUE trap ‘Another word I heard: ‘dyiñaeij’ – what is ‘dyinae’?’

(20)e yok-tyi’ soñi’ other-M man ‘another man’

The morphemes –tyi’ in the masculine gender and –si’ in the feminine gender resem- ble those used in the possessive construction. Thus, when then the yoksi’/yoktyi’ itself appears in a possessive construction, it may appear to have two gender markers of the same kind. This gives rise to double gender agreement on the same word. However, MOSETÉN GENDER AGREEMENT 291 agreeing with different antecedents in gender, one of these might be in the feminine gender, while the other one is in the masculine gender: other itself agrees with the noun it represents, while the possession marker agrees with the possessum of this possessive construction:

(21)e Fan chhïï-ye-’ äwä’ Alfredo-si’ Juan know-VB-3SG.F.O child Alfredo-F

jam-ki-kij chhïï-ye-’ yok-tyi-si’ Mo’ momo’. NEG-and-and know-VB-3SG.F.O other-M-F 3F.SG only.F ‘Juan knows the daughter of Alfredo. But (and) he does not know others (i.e. other man’s daughters). Only her.’

4. Determiners In the function of determiners in the noun phrase, usually 3rd person personal pro- nouns and demonstrative pronouns are used in Mosetén. These are marked for gender (see section 2. above) agreeing with the gender of the head noun:

(22)e mi’ mintyi’ 3M.SG man ‘the man’

(23)e mo’ minsi’ 3F.SG woman ‘the woman’

5. Numerals Of the cardinal numerals, only the numeral ‘one’ is marked for gender: jiris in the feminine and jirity in the masculine gender. All other numerals have the same form for both genders. When counting ‘one, two, three’, and in other cases where gender is not specified, the feminine form jiris is used. In the noun phrase, the numeral one agrees in gender with the head noun:

(24)e jiri-s son daer-si’ 5 one-F tree big-F ‘one big tree (feminine)’

(25)e jiri-ty kojti daer-tyi’ one-M heart big-M ‘a big heart (masculine)’

5 An explanation of the long adjectival form is given below. 292 SAKEL

All other cardinal numerals are not marked for gender. This even includes forms that are compounded with one, as twenty-one in (26). This numeral does not agree in gender with the head noun, but remains in its unmarked feminine form:

(26)e paerae’ -ki’ tyak jiri-s jiyi’ soñi-in two-times ten one-F NUM man-PL ‘21 men’

(27)e paerae’-ki’ tyak jiri-s jiyi’ phen-in two-times ten one-F NUM woman-PL ‘21 women’

The ordinal numerals, on the other hand, have both feminine and masculine forms of all numerals due to their derivation. The derivation of ordinal numerals is the addi- tion of -yi’si’ to the cardinal numeral in the feminine gender and -yityi’ in the mascu- line gender. The -yi- appears to be derived from the verb yi, ‘to say’, which in the feminine gender is marked by a glottal and in the masculine gender remains un- marked. To this inflected verbal form, the gender marker -tyi (M) and -si’ (F) is added and agrees with the head noun:

(28)e Chhibin’-yi-’-si’ mayedye’ khin’-dyem-ra’ karij-tya-ki-tsin three-ORD-F.S-F day now-yet-POT work-APPL-INTR.M.S-1PL ‘The third day we already work.’

(29)e Chhibin’-yi-tyi’ soñi’ tsin kawe-te. three-ORD.M.S-M man 1PL see-3SG.M.O ‘The third man we have seen.’

6. Adjectives Adjectives have two forms in Mosetén, from now on called ‘long’ and ‘short’ forms. While the short forms might be used as adjectives or adverbs, the long forms are strictly adjectival forms. The long form consist of the short form plus the gender marker -tyi’ (M) and -si’ (F). Thus, adjectives in the long form agree in gender with their antecedent, while adjectives in the short form have no marking for gender. In the following table, I summarise the use of the long and the short adjectival forms in Mosetén:

short from long form Adverb yes no Adjective used predicatively yes yes Adjective used attributively no yes Table 1: the long and the short adjectival forms in Mosetén MOSETÉN GENDER AGREEMENT 293

The long adjectival form appears with adjectives used predicatively and attributively, while the short form, apart from some exceptions (see below), is only used with ad- jectives in predicative position. The adjective marking is similar to the marking in Russian, which also has a long and a short adjectival form (Wade 1992) with the same formal distribution as in Mosetén. As in Russian, in Mosetén the distribution of long and short forms of the predicatively used adjective depends on various semantic factors. The following ex- amples show the attributive use of adjectives, where the adjective is usually marked by the gender marker:

(30)e Mi’ mintyi’ daer-tyi’. 3M.SG man big-M ‘The man is big.’ or ‘The big man.’

(31)e Tsin ya’ij its kasko öij-dye-tyi’ nana-si’ ïchäeke-si’. 1PL buy.M.S DEM.M boat.E DEM.F-BEN-M girl-F little-F ‘We buy this boat for that little girl.’

Some adjectives are exceptions and do not always take the genitive in attributive po- sition:

(32)e dyam’ momo’ phen-in few only.F woman-PL ‘only few women’

The following examples show the predicative use of adjectives. In this case, the ad- jectives may be in the long or in the short form. When in the long form, the adjectives agree in gender with their antecedent, which in this case is the subject of the clause:6

(33)e Mo’ aka’ daer-si’. 3F.SG house big-F ‘The house is big/the big house.’7

(34)e Mi’ soñi’ daer-tyi’. 3M.SG man big-M ‘The man is big/the big man.’8

6 In the case of an object predicative the agreement would be with the object. 7 This has also another translation (see ‘Relative clauses’ below): ‘the house that is big’ 8 This also has another translation (see ‘Relative clauses’ below): ‘the man who is big’ 294 SAKEL

(35)e Mo’ aka’ daer. 3F.SG house big ‘The house is big.’

7. Other nominal relationships Apart from the adjectival marking discussed above, there are a number of structures in Mosetén that express nominal relationships and mark for gender: the possessive construction, which has already been discussed with the possessive pronouns, the benefactive/purposive case and participant-nominalization. The consists of the benefactive marker and the gender marker -tyi’ (M) and -si’ (F). It expresses benefactive and purposive meanings. Gender agreement is with the subject of an intransitive clause:

(36)t Jikej katyi’ mo’-in ish-mo’ PST EVID 3 F-PL mother.in.law-3F.SG

phan’-ye-j-ki-’ mo’-chhe’ ijme-dye-si’. feather-VB-DIR-DIR-F.S 3F.SG-SUP arrow-BEN-F ‘Then her mother-in-law (went) up there to get feather(s) for the arrow.’

In transitive clauses, gender agreement in the benefactive case suffix is with the object of the clause:9

(37)e Yij tsin jiri-s jedye’ its-dye-si’. say.M.S 1 PL one-F thing DEM.M-BEN-F ‘We say one thing to (‘for’) him.’

(38)e Tsin ya’ij mi’ o’sho’ mo’-dye-tyi’. 1PL buy.M.S 3M.SG clothes 3F.SG-BEN-M ‘We buy these clothes for her.’

One type of nominalization marking involves the morphemes –tyi’ (M) and –si’ (F) and thus shows gender agreement. This nominalizer primarily derives verbs. The nominalization focuses on the subject participant of the action, so that the nominal- ized element denotes a person. Gender agreement is with the subject of this verb, i.e. with the person(s) that is / are expressed:

9 In other structures, to be discussed below, the antecedent of gender agreement can be an established topic in the situation. Also the benefactive case marking might agree in gender with an established topic, even if I have no clear examples of this yet. MOSETÉN GENDER AGREEMENT 295

(39)t Mi’-in-nä’ aj ro’ya-ki-tyi’-in aj äejai saeksi-in. 3M-PL-and ASP dance-INTR-M-PL ASP stop.M.S eat.M.S-PL ‘And the dancers (M) stop eating.’

8. Verbs There is gender marking on all intransitive verbs. Transitive verbs are marked for gender with 3rd person objects and also 1st person inclusive plural subjects. The other person forms in the transitive cross-reference system do not mark for gender.

8.1. Intransitive verbs Intransitive verbs agree with the gender of the subject. This agreement marking is apparent in all persons. The feminine gender is expressed by a glottal suffix, the mas- culine gender is expressed by the lack of this glottal and when ending in a vowel, it is often followed by a slight aspiration:

(40)e Yäe karij-tya-kij. 1SG hard-APPL-INTR.M.S ‘I (M) work.’

(41)e Yäe karij-tya-ki-’. 1SG hard-APPL-INTR-F.S ‘I (F) work.’

The inflection of intransitive verbs, with one exception, shows gender marking for the subject. The exception is the 1st person plural inclusive subject, which behaves differently from the other persons in the cross-reference paradigm, and which has a separate form with intransitive verbs. The intransitive 1st person plural inclusive in- flection has the same form for both genders:

(42)e Saekse-ja’. eat.INTR-1PL.in ‘We (incl.) eat.’ (masculine, feminine, mixed)10

Mosetén also has a verb with suppletive stem forms for each gender: atsij, ‘he goes’ and ayij ‘she goes’. Usually, the feminine ayij has no glottal at the end to indicate the feminine agreement, even if I have few examples where speakers use the form ayi’ with a glottal. An explanation for this could be that since the gender is already speci- fied by different verbal stems, there is no need for a special ending for feminine gen- der agreement:

10 In Chimane and Mosetén de Santa Ana, the same form –ja’ is also used with transitive verbs with a 1st person plural inclusive subject, agreeing in gender with the object. The affix -ja is used with a masculine object and –ja’ with a feminine object. Thus, the form –ja’, which is used in the intransitive cross-reference ending in all three dialects, seems originally to be a feminine form. 296 SAKEL

(43)e Mo’ ayij yäe-ya’. 3F.SG come.F.S 1SG-AD ‘She comes to me.’

(44)t Yoshropai-mij khin atsij i-khan’ mij thank-2SG now come.M.S M-IN 2SG ‘Thanks that you now came here.’

When affixes are added to the verb atsij/ayij, however, the glottal for the feminine gender form and the lack of the glottal on the masculine form become evident. The following examples show the directional marker –joij (M)/-jo’ (F) with the alterna- tive form –joi’ (F), which frequently is added to this verb and which takes typical verbal gender agreement forms:11

(45)t Me’-nä-yäe atsij-jo-ij jäe’maj atsij-jo-ij DM-‘and’-1SG come.M.S-DIR-M.S DM come.M.S-DIR-M.S

yäe viaje-ij chhata’-yäe 1SG travel.E-VB.M.S truly-1SG

jäe’maj äwä-tom-yäe jäe’maj certificado-dye-si’. DM son-COM-1SG DM certificate.E-BEN-F12 ‘And I came here, I truly traveled here with my son for the certificate.’

(46)t Ayij maj ojñi’ ayij-jo-i’13 mo’ Köwë’döj. come.F.S much water come.F.S-DIR-F.S 3F.SG Covendo (rio C.) ‘Much water came (here) in the river ‘Covendo’.’

Another verb that has two different stems, one for each gender, is the verb iyayekij, ‘get better.M’, oyayeki’, ‘get better.F’. With these stems, the development of differ- ent gender forms in the stem is synchronically clear, as this verb derives from a place adverb with comparative meaning, which itself derives from a pronoun (see the sec- tion below on place adverbs):

(47)t Yidyej san’ mimi’ ji’-tye-tej mi’-mimi’ pure mentisan.herb only.M CAUS-drink-3SG.M.O 3M.SG-only.M

11 Only the affixes –jo’/-joi’ and –ban’ are used with this verb (in my data) and they have marking for the feminine gender. 12 The Spanish word certificado ‘certificate’ seems to act as a feminine noun in Mosetén, even if its gender assignment is masculine in Spanish. 13 The feminine form in Mosetén de Covendo is –jo’, even if some speaker use –joi’, which is the form of this morpheme in the other dialects Mosetén de Santa Ana and Chimane. MOSETÉN GENDER AGREEMENT 297

mi’-mimi’ jö’dyë’yä’ aj iyaye-k-han. 3M.SG-only.M and already get.better.M-INTR-again.M.S ‘We made him drink purely San’ (a special herb) and he became better.’

(48)t Kaechen’ öyaye-ki-’ jikej. go.on get.better.F-INTR-F.S PST ‘She was getting better.’

8.2. Transitive verbs The complicated cross-reference system of transitive verbs contains some forms agreeing in gender with an argument, while other forms do not show any such agreement. All 3rd person object forms show gender agreement with a number split in respect to their antecedent: singular 3rd person object forms inflect for the gender of the object, 3rd person plural object forms agree with the gender of the subject of the clause. Also 1st person inclusive plural subject forms (of transitive verbs) show gen- der agreement. The following table summarises the forms of the cross-reference markers in the 3rd person object forms and the 1st person plural inclusive forms in order to show the gender agreement structure in transitive cross-reference. Only forms that show gender agreement are given in the table:

1SG.O.M 1SG.O.F 3SG.O.M 3SG.O.F 3PL.O.M 3PL.O.F 2SG.O.M 2SG.O.F 1PL.O.M 1PL.O.F 2PL.O.M 2PL.O.F 1SG.S.M -te -(V)’ -ksi -ksi 1SG.S.F -te -(V)’ -ksi’ -ksi’ 2SG.S.M -te -(V)’ -ksi -ksi 2SG.S.F -te -(V)’ -ksi’ -ksi’ 3SG.S.M -te -(V)’ -ksi -ksi 3SG.S.F -te -(V)’ -ksi’ -ksi’ 1PL.in.S.M -tij -tij -tij -tij -kseja -kseja 1PL.in.S.F -ti’ -ti’ -ti’ -ti’ -kseja’ -kseja’ 1PL.ex.S.M -te -(V)’ -ksi -ksi 1PL.ex.S.F -te -(V)’ -ksi’ -ksi’ 2PL.M -te -(V)’ -ksi -ksi 2PL.F -te -(V)’ -ksi’ -ksi’ 3PL.M -te -(V)’ -ksi -ksi 3PL.F -te -(V)’ -ksi’ -ksi’ Table 2: Transitive cross-reference markers that show gender agreement

298 SAKEL

8.2.1. The forms of the 3rd person singular object These forms are marked for the gender of the object, but they are independent of the gender of the subject of the clause. Apart from the special person marking forms of the 1st person plural inclusive subject marking, the forms of the 3rd person singular masculine object are –te and for the 3rd person singular feminine object they are -V’ (or variably -Vi or -Vi’)14. The gender agreement is with this object form (and not with e.g. the subject):

(49)t “Jäe’nä’ abi’ mi jij-tiij” ye-te-kätyï’ mo’ soñi’; where QUE 2SG come-DIR.M.S say-3SG.M.O-EVID 3F.SG man

me’kij mi’ yij soñi’ ye-’ phen … thus 3M.SG say.M.S man say-3SG.F.O woman ‘“Where did you come from”, she said to the man; thus the man said to her…’

8.2.2. The forms of the 3rd person plural object In the cross-reference forms for the 3rd person plural object, the gender agreement is with the subject, as opposed to the singular, where the gender agreement is with the object. In all cases, apart from the special case of where the subject of the clause is the 1st person plural inclusive, the form is -ksi for 3rd person plural object and a mas- culine subject and -ksi’ for a feminine subject.15

(50)e Mo’ phoch-ya-ksi-’ mi’-in 3F.SG inject-VB-3PL.O-F.S 3M-PL ‘She injects them.’

(51)e Its Fan ji’-karij-tya-ksi yidyej phen’-in. DEM.M Juan CAUS-work-APPL-3PL.O.M.S pure woman-PL ‘This Juan makes purely women work.’

8.2.3. The forms of the 1st person plural subject The forms of the transitive 1st person plural inclusive subject differ from other transi- tive cross-reference forms, as do intransitive cross-reference forms for the 1st person plural inclusive subject. All forms of the 1st person inclusive subject agree in gender with the subject:

14 V means vowel. In this case the vowel is often e. 15 The forms of the 3rd person plural objects differ from the other cross-reference forms in various other ways. Many derivational affixes that usually appear before the cross-reference ending occur after the marker for the 3rd person plural object. In many ways this form functions like an intransitive derivation, also in its gender agreement to the subject of the clause. MOSETÉN GENDER AGREEMENT 299

(52)t Pa’-ki-ti-ra’ tyäkä’-dyerä’ ija-tij hit-INTR-1PL.in.S.M-POT always-must.be kill-1PL.in.S.M

otej-nä-rä’ jikej not.let.disappear.again-and-POT PST ‘We will hit and thus kill it, and we (M) wouldn’t let it disappear again.’

(53)t Me’-kij mo’ yi-’ “jam yäe raise-’ ka-i’ thus-that.M 3F.SG say-F.S NEG 1SG want-3SG.F.O bring-3SG.F.O

öij käedäej jedye-dye-si’-dye’-ki-ra’ ka-ti-’ ”. DEM.F baby thing-BEN-F-NOM-that-POT bring-1PL.in-F.S ‘Thus she says “I do not want to bring this baby, why should we (F) bring it?”’

8.3. Copular verbs Gender differences also occur in copular verbs. There is no obligatory copular verb in affirmative clauses and the optional copular verbs do not show gender agreement. In negative clauses, however, the copular verb itsi’ (F)/itsij (M) must be present. Gender agreement is with the subject:

(54)e Jebe-’ shiish-yäe, yodye’ itsi-’-ya’ phen-yäe. eat.(trans)-3SG.F.O meat-1SG when not.be-F.S-when wife-1SG ‘I eat meat when my wife is not (at home).’

(55)e Itsij aka’-khan’. not.be.M house-IN ‘The (man) is not in the house (i.e. at home).’

9. Place adverbs In Mosetén, even place adverbs agree in gender. The form is diachronically derived from personal pronouns, in which this gender difference also is apparent (compare section 2).16 The gender marking of the place adverb in transitive and intransitive clauses agrees with the gender of the subject of the clause. However, as I will show below, the antecedent of the agreement marking may change due to pragmatic factors.

(56)e Mo’ Jeanette mo-wej jadyi-ki-’. 3F.SG Jeanette F-DR go-DIR-F.S ‘Jeanette went there.’

16 There are various deictic differences in the place adverbs: mo’wej (far), mowej (very far, cannot be seen) and owej (close) differ in distance to the speaker. 300 SAKEL

(57)e Yäe mäen’ja’ mi-wej jadyi-ki. 1SG(M) yesterday M-DR go-DIR.M.S ‘I (M) went there yesterday.’

The two examples above show intransitive clauses. In the first example the subject of the clause is feminine and the place adverb ‘over there’ is in the feminine gender and in the second example the subject is masculine, leading to a masculine agreement marking in the place adverb. The following example shows a transitive clause, where the gender agreement of the place adverb is also with the subject of the clause:

(58)e Alfredo nash tyaj-k-ei-rä’ Maria i-ya’. Alfredo EMP meet-DIR-3SG.F.O-POT Maria M-AD ‘Alfredo will go to meet Maria here.’

The subject of the clause is masculine, the object feminine and the place adverb car- ries masculine agreement, agreeing with the subject. The examples above are all elicitation examples, i.e. clauses outside of a dis- course context. Looking at discourse context, however, the picture of gender agree- ment is different. In the following example a topic is established in the first clause, which then is the trigger of gender agreement in the second clause:

(59)t a Jikej mo’ anik-si-si’ jïjka-baj-te tya’kaj-ye-’ aj PST 3F.SG sure-RD-F follow-again-3SG.M.O leave-VB-3SG.F.O ASP

äwä’ tyäjä’-wë-dyërä’ wen-jo-’ o’yi-si’-wej. child edge.of.plantation-DR-must.be ‘move’-DIR-F.S yuca-F-DR ‘She surely followed him, left (her) child and went to the edge of the yuca field.’

(59)t b Mo’-wej aj phij-ye-’ jam-ra’ mo’ aj soñi’-ra’ 3F.SG-DR ASP blow-VB-3SG.F.O NEG-POT 3F.SG ASP man-POT

mi’ aj-win sino ïtsïkïj chhi-mo’ aj ïtsïkïj. 3M.SG ASP-before but.rather.E jaguar also-3F.SG ASP jaguar. ‘There he blew her (in her ear), (and from then on) she no longer (saw) the man (he was) before, but rather a jaguar, and also she was a jaguar.’

In (59b) the place adverb is in the feminine gender, agreeing with the established topic of the preceding clause, whereas the subject of (59b) is masculine in gender. Thus here, the agreement of gender does not follow a strictly syntactic pattern, but reacts to pragmatic factors. MOSETÉN GENDER AGREEMENT 301

10. Particles Remarkably, gender can also be marked on some particles.

10.1. Only, just The first particle I want to discuss is ‘only, just’. This is expressed by mimi’ (mascu- line) and momo’ (feminine). Diachronically, these seem to be reduplicated forms of the personal pronouns mi’, ‘he’ and mo’, ‘she’. There are two kinds of uses, where mimi’/momo’ appears. On the one hand this particle can be attached to a noun phrase, meaning ‘only’. On the other hand, the particle can be used in a clausal context, meaning ‘just’. When this particle is used attached to the noun phrase with the meaning ‘only’, the agreement in gender is with the head noun:

(60)t Jäe’ma mimi’ äwä’-yäe jö’dye’ya’ tyiñe’-yäe DM only.M son-1SG and son.in.law-1SG

mi’-in mimi’ karijtyakij mi’-we-in 3M-PL only.M work.M.S 3M.SG-DR-PL ‘Well, only my son and my son in law, only they are working in there.’

In this example, ‘only’ is part of the noun phrase and agrees in gender with the mas- culine head noun. When the meaning is ‘just’, the element in the scope of this particle is the whole clause, and the gender agreement is with the subject of that clause. Example (61) shows an intransitive clause involving agreement with the subject, (62) is a transitive clause, where the agreement also is with the subject of the clause:17

(61)t Yok-min’-tyi’ katyi’-in jam jäe’maj jam’ jam-dyërä’ other-DIS-M EVID-PL NEG DM NEG NEG-must.be

chhaekh-dyi’-dyiij-in yok-min’-tyi’-in me’-mimi’ jiyij-in jam careful-DIR-PROG-PL other-DIS-M-PL so-only.M pass.M-PL NEG

jäe’maj kawa-kij-k-hoi-in me’-katyi’ mimi’ jiyij-in DM look-INTR-DIR-DIR.M.S-PL so-EVID just.M pass.M-PL ‘Others were not going carefully, they just passed without looking (watching out), in this way they just passed.’

17 There might again, as with the place adverbs, be a broader, pragmatically based pattern of gender agreement. This pattern would then take an established topic as its antecedent. However, the lack of corre- spondence between subjecthood and topichood remains to be investigated. 302 SAKEL

(62)e Maria chhïï-ye-te äwä’-mi’ Martin-tyi’. Me’-momo’ Maria know-VB-3SG.M.O child-3M.SG Martin-M thus-only.F ‘Maria knows the child of Martin. That is all (about this).’

10.2. Question particles and other particles Several particles are diachronically derived from the pronouns i and o, which in Chimane seem to be free particles (Gill 1999). In Mosetén de Covendo, they always occur in fixed combinations with particles or affixes, as for example together with the unproductive suffix –ka’, which is used in rhetorical questions.18 Thus, diachroni- cally, these two now unproductive morphemes have developed into the productively used question particles ika’ (M) and oka’ (F). They show agreement in gender with the subject:

(63)e Tyi’-ra’ ö-ka’ mo’ tï’-ï-’. person-POT F-QUE 3F.SG name-VB-F.S ‘What was her name?’ (I knew it once, but can’t remember it right now)

(64)t Tyi’-ra’ i-ka’ ti’-ij mi’ wïyä’-[in] tyashi-tyi’ person-POT M-QUE name-VB.M.S 3M.SG old.man-[PL] approach-M ‘How was this old guy called who approached (the certain place).’

11. Discussion and conclusion There are most probably several other targets showing gender agreement in addition to those discussed here. Further research may reveal these.

11.1. The feminine as the unmarked gender Formally, there is no clear difference in the markedness of the different genders, even if with verbs the feminine gender often appears longer, as it involves a glottal where there is no marking in the masculine form. Functionally, however, the feminine gen- der is usually the unmarked form in Mosetén as opposed to the masculine gender. Thus, when a verb, such as the modal verb ‘want’, is followed by a clause as its for- mal object, this verb always gets feminine object marking:

(65)t Mo’ jam raise-’ jij-ka-baj-te 3F.SG NEG want-3SG.F.O go-DIR-again-3SG.M.O ‘She does not want to follow him.’

When males and females constitute one group, usually the feminine form is used to refer to them:

18 In the same way as the pronouns i and o are unproductive in Mosetén, the suffix ka’ is also unproduc- tive in Mosetén and productive in Chimane. It only exists in some combinations in Mosetén, together with other question particles such as dyaj or dyash, which themselves are productive in Mosetén. MOSETÉN GENDER AGREEMENT 303

(66)e Mo’-in yi-’-in atsi-jo-ij katyi’ äwä’-mi’. 3F-PL say-F.S-PL come-DIR-M.S EVID child-3M.SG ‘They (father and mother) said that their son came.’

(67)e Elena y Fan, mo’-in käijedye’-tom San Jose-chhe’-in. Elena and.E Juan 3F-PL plantation-COM San Jose-SUP-PL ‘Elena and Juan, they have a plantation in San Jose.’

Sometimes, however, the masculine gender form is used in this case, especially in the speech of young people, surely due to the influence of Spanish.

11.2. Individual differences I have encountered variation in the speech of consultants with respect to the formal expression of gender. One of my female consultants differentiates between feminine and masculine in the plural clitic in. She frequently uses on in feminine forms. Thus, mo’in, ‘they, F’ becomes mo’on. She might be re-interpreting the i in this plural clitic as a masculine form, as opposed to o being the feminine form. This difference be- comes apparent in the personal pronouns, where mi’ is the 3rd person singular mascu- line and mo’ the 3rd person singular feminine. This seems to be a personal invention of my consultant, since I have not yet found any other speakers who use this form, nor any other data that could lead me to the conclusion that the form on for the plural feminine is a general form.

11.3. Conclusion In Mosetén, gender agreement is not restricted to ‘typical’ targets such as adjectives, personal pronouns and verbal cross-reference, but it is also found in place adverbs, case marking, relative clause markers, and particles. In many cases, it is rather un- clear from a functional perspective why these targets show gender agreement. The explanation lies in the development of these structures. The forms of gender agree- ment to some extent resemble each other. The gender forms in Mosetén can be di- vided into four formal classes:

1) –si’ (F) and –tyi’ (M) 2) o (F) and i (M) 3) forms ending in a glottal (F) and forms not ending in a glottal (M) 4) suppletion differences

I will briefly discuss these:

1) Gender agreement by the forms –si’ (F)/–tyi’ (M) and the short form –s (F)/-ty (M) is frequently found inside the noun phrase. This is a phenomenon usually called mac- rofunctionality of morphemes (Gil 2001), i.e. the same morpheme is used for posses- 304 SAKEL sion structures, adjectival marking, relative clause marking, nominalizations and other functions. In Mosetén this macrofunctional morpheme furthermore marks some pronouns and numerals. In addition to that, it appears in the benefactive case mark- ers. As this morpheme shows gender differences, all its uses have these differences. The development of macrofunctional morphemes as in Mosetén may have proceeded in several ways. In the case of Mosetén, the development of macrofunctionality might have started with the use of a morpheme in only one structure for which gender agreement was functionally important. One such structure could have been the geni- tive case. Then, the morpheme gained a broader use, bringing its gender agreement forms into other structures, which one would not usually expect to show gender agreement.

2) The differentiation between feminine and masculine gender by o (F) and i (M) is found with pronouns, place adverbs and certain particles. Diachronically, these place adverbs and particles appear to have developed from the personal pronouns mo’ (F), mi’ (M) and the particles o (F), i (M), the path of grammaticalization still being rather clear in the synchronic forms. In this way, place adverbs consist of a 3rd person pro- noun such as mo’ and a local case marker such as –wej: mo’wej, ‘there’. In the same way, the particle momo’ (F), mimi’ (M) has most probably developed by a reduplica- tion of the personal pronoun. The place adverb ikhan’, ‘here.M’/okhan’ here.F’ and the question particles oka’ (F), ika’ (M) have developed by combining the particle o, i with a case marker or a question particle.

3) The marking of the feminine gender by a glottal and the lack of this glottal in the masculine gender is found in verbs. Both transitive and intransitive verbs show this marking, even if the transitive 3rd person singular masculine object has another agreement form

4) Gender differences marked by suppletion are found in a few forms, among them one verb, several nouns denoting females/males and the demonstrative pronouns öij (F) and its (M).

I have shown that gender agreement morphemes are not separate forms in Mosetén, but that there are few forms that are found in various different targets of gender agreement. This seems in many cases to have to do with diachronic developments in the language. Since personal pronouns carry gender marking in Mosetén and since these pronouns have played a role in creating new forms, gender agreement had a way to expand. In the same way, a certain morpheme that has developed into a mac- rofunctional morpheme by chance carried gender agreement, and thus brought gender agreement to targets where we usually would not expect to find it.

MOSETÉN GENDER AGREEMENT 305

References Corbett, Greville 1991 Gender, Cambridge: Cambridge University Press. Derbyshire, Desmond C. & Doris L. Payne 1990 ‘Noun Classification Systems of Amazonian Languages’, in: Doris Payne (ed.) Amazonian Linguistics: Studies in Lowland South American Languages, Austin: University of Texan Press. Gil, David 2001 ‘Escaping Eurocentrism: Fieldwork as a Process of Unlearning’, in: P. New- man & M. Ratcliff (eds.) Linguistic Fieldwork, Cambridge: Cambridge Uni- versity Press. Gill, Wayne 1999 A Pedagogical Grammar of the Chimane (Tsimane’) Language, revised edi- tion, unpublished manuscript (Cochabamba: New Tribes Mission, Bolivia). Payne, Doris L. 1990 ‘Morphological Characteristics of Lowland South American Languages’, in: Doris Payne (ed.) Amazonian Linguistics: Studies in Lowland South American Languages, Austin: University of Texan Press. Wade, Terence 1992 A Comprehensive Russian Grammar, Oxford: Blackwell.

Abbreviations 1PL.in first plural inclusive E Spanish loan POT potential 1PL.ex first plural exclusive EMP emphatic marker PROG progressive aspect AD adessive relation F feminine gender PST past tense marker APPL applicative IN inessive relation QUE question marker ASP aspect INTR intranisitivizer RD reduplication BEN benefactive case M masculine gender S subject CAUS causative NEG negation SG singular COM comitative NOM nominalization SUP superessive relation DEM demonstrative pronoun NUM numeral marker V vowel DIR directional O object VB verbalization DIS distributive ORD ordinal numeral ()e elicitation example DM discourse marker PASS passive ()t text example DR “downriver” relation PL plural

THE QUOTATIVE CONSTRUCTION IN KWAZA AND ITS (DE-)GRAMMATICALISATION

Hein van der Voort Katholieke Universiteit Nijmegen/WOTRO

1. Introduction Kwaza is spoken in Southern Rondônia, Brazil, by about 25 people, most of whom live among the Aikanã of the indigenous reserve Tubarão-Latundê.1 Kwaza, or in the litera- ture: Koaiá (e.g. Rodrigues 1986), is an unclassified language which has often been considered as isolated, just like its neighbours Aikanã and Kanoê. Almost all the other languages by which it is traditionally surrounded belong to the Tupi and linguistic stocks. The basic grammatical categories of Kwaza are verbs, nouns and adverbs. Kwaza is a morphologically complex language. Most of the grammar of the language is con- tained in derivational and inflexional verbal suffixes. Categories like classifiers, direc- tionals, valency changing suffixes and tense, modality and aspect are in Kwaza best regarded as derivational. Main word stress in Kwaza is basically on the last syllable of the (extended) root. The extended root may include derivational morphemes, but in principle no inflexional suffixes. As will be shown in Section 1.1 below, the morpho- logical categories of subject and mood are obligatory and should be considered as in- flexional. In addition to person inflexion, corresponding pronouns may be used for em- phasis. Word order is relatively free, but SVO is the most frequent order. In the present article I want to discuss the grammatical characteristics of the most common way in which speech is quoted in Kwaza. I will also discuss other modality- like constructions that show striking similarities to the quotative construction. Further- more, I will try to provide an explanation for the origin of both the quotative construc- tion and certain modalities. I will not discuss indirect quotation as sometimes attested in nominal argument clauses.

1.1. Argument agreement The following examples show how the matrix verb in Kwaza sentences is obligatorily inflected for subject person, and optionally for object person or a combination. Note that the third person subject is not expressed. The default value of absent person inflex-

1 This article is based on data from linguistic fieldwork conducted among the speakers of Kwaza during the years 1995-1998. I am especially indebted to my teacher Kyikãu Mãdε who is also known as Mário. I am furthermore very grateful to the inhabitants of the Área Indígena Tubarão-Latundê for their hospitality. The Netherlands Organisation for Scientific Research (NWO) has generously financed the entire descrip- tive project of the Kwaza language under grant nr. 300-72-021. Finally, I want to thank Mily Crevels, Simon van de Kerke, Sérgio Meira and Pieter Muysken for their highly valued comments. It should go without saying that none of these people necessarily shares the views expressed here and that all errors are mine. 308 VAN DER VOORT ion is either third person subject, or, in imperative sentences, second person. For con- venience sake this is symbolised with a -Ø-. Consider these examples2:

(1) ´hã ´kui-da-ki (2) ´hã ´kui-xa-ki water drink-1S-DEC water drink-2 -DEC ‘I drank/am drinking water’ ‘you drank/are drinking water’

(3) ´žwãu ´mε-hata-Ø-ki xyitsε-´wã João hit-3S2O-3-DEC you.PL-AO ‘it was João who beat you’

Kwaza has no copular verbs. Instead, nouns may receive person and mood marking without any intervening morphology. In that case an operation that can be characterised as zero-verbalisation has taken place, and the result conveys the meaning “to be/to want N”. Nominalised verbs, pronouns and adverbs can also be zero-verbalised. Person marking in Kwaza is to be considered as bound inflexional morphology for the following reasons: a) there is a clear formal difference between free personal pro- nouns and bound person morphemes; b) bound person morphemes cannot be separated from their verbal host by either a change in the order of elements or; c), by the insertion of free forms; d) bound person morphemes are part of the word as an accentual unit; e) the presence of these bound morphemes is necessary for verb formation; f) bound per- son morphemes do not occur in isolation. The exceptions to criteria c), e) and f) will be discussed in Section 1.5 below. Table I represents the evidence for criterium a):

PERSON PRONOUNS MARKERS USE 1 ´si -da- I 2 ´xyi -xa-/-Ø- you 1+2 txa´na -a- we [inclusive] 1+3 tsi´tsε -a-xa- we [exclusive] 2+3 xyi´tsε -xa-(xa)- you (PL) 3 ´ĩ -Ø- he, she, it, they IS -wa- they, people, it Table I: Personal pronouns and subject cross-reference

Note that there is no number marking in Kwaza. In the pronominal system there are three basic persons and two “associated” persons. According to this analysis, the dis- tinction between first person plural inclusive and exclusive is explained by the associa- tion of a second and a third person respectively. Second person plural is represented in

2 In the Kwaza examples in this article I have indicated main word stress by an inverted comma [ ´ ] preced- ing the syllable that receives stress (or by a [ “ ] when another syllable gets secondary stress). Apart from ex- plicit stress marking, the examples are in phonemic notation. The following symbols are used in a special way: /a/ represents IPA [a]; /ε/ = [ε]; /y/ = []; /j/ = [j]; /b/ = []; /d/ = []; /D/ = [d]; /x/ = [s]; /c/ = [t]; /ž/ = []; /~/ = [nasality].

QUOTATION IN KWAZA 309

Kwaza by a second person associated with a third person (may be singular or plural). With respect to criterium d), main word stress tends to occur on the last syllable of the root. If the root is extended with derivational morphemes, main word stress is on the last syllable of the last extension. Main word stress is never on the subject cross- reference morphemes, but it does occur on mood morphemes under specific circum- stances, which will be mentioned in Section 1.2.

1.2. Matrix clauses: mood or speech acts The examples above are all in the Declarative mood. Besides the declarative, there are seven other matrix clause moods: the Interrogative, which combines with the entire paradigm of person markers; the Imperative, Volitive and Exhortative, which combine to constitute the Persuasive mood paradigm; the Negative Imperative, Negative Exhor- tative and Monitory, which together form the Prohibitive mood. The following exam- ples show the morphological expression of these moods:

(4) ka´wε kui-´nã-xa-re (5) ´mĩu ´kui-a-xa-mỹ coffee drink-FUT-2-INT chicha drink-1P-AS-VOL ‘will/do you want to drink coffee?’ ‘we’re going/want to drink chicha!’

(6) ´hã ´kui-a-ni (7) ka´wε kui-Ø-´ni water drink-1P-EXH coffee drink-3-EXH ‘let us drink water!’ ‘let him have coffee!’

(8) ehỹ-´he-a-ni (9) ay-´hỹ kui-´he-Ø-ky do-NEG-1P-NEE that-NOM drink-NEG-2-NEI ‘let’s not do that!’ ‘don’t drink that!’

(10) kui-Ø-´tsi (11) kui-Ø -´ra drink-3-MON drink-2-IMP ‘don’t let him drink!’ ‘drink!’

Remember that in the (negative) imperative mood cross-reference to the second person subject is not expressed and in the other moods third person subject cross-reference is not expressed. Note that certain morphemes, such as negative -he- and future -nã-, and certain moods, such as the imperative, usually attract word stress.

1.3. Subordinated clauses: adverbial clauses Besides mood marking in matrix clauses, there are also certain mood-like morphemes that indicate the specific adverbial status of subordinated clauses: Conditional, Conces- sive, Additive, Absolutive/Manner, Temporal, Nominal and Contemporative. The fol- lowing example shows the expression of a concessive adverbial clause:

310 VAN DER VOORT

(12) awy-´hỹ-da-lete ´ba-da-ki cold-NOM-1S-CONC cut-1S-DEC ‘although I had become cold, I did clear a field’

The next example shows that the element -tja (a special variant of cosubordinative -ta, that will be further discussed below in Section 1.4) may also indicate that the clause functions as an absolutive/manner adverbial:

(13) txu´hũi-Ø-tja hũnũ-´dy-da-ki small-3-CSO scorch-CAU-1S-DEC ‘I burnt the food (just) a little’

1.4. Cosubordinated clauses: medial clauses and switch reference There is a type of clauses in Kwaza that can be linked to the matrix clause. Such clauses can form long chains and their illocutionary value is often determined by the sentence- final matrix clause. Semantically they are coordinated with the matrix clause, but for- mally they are subordinated to it. Such constructions are known in the literature (e.g. Foley & Van Valin 1984) as cosubordinated or medial clauses. The following examples show cosubordinated clauses that have the same subject as the matrix clauses:

(14) bilo´tswa e-´he-da-ta okja-´he-da-ki shotgun have-NEG-1S-CSO hunt-NEG-1S-DEC ‘I can’t hunt because I have no shotgun’

(15) ca´ri-da-ta ´jo-da-mỹ shoot-1S-CSO devour-1S-VOL ‘I’m going to shoot and devour him!’

(16) ´tswa-wã ´mε-Ø-ta e´mã-Ø-ki man-AO beat-3-CSO cry-3-DEC ‘(the woman) beat the man and (she) cried’

When the subject differs from that of the next clause in the chain this is morphologi- cally indicated by either a different subject marker or, if the cosubordinated clause has a non-third person subject, a special switch-reference morpheme that occurs in a position normally reserved for mood inflexion. This is shown in the following examples:

(17) e´tay tswa-´wã ´mε -dy-ta e´mã-Ø-ki ´tswa woman man-AO beat-DS-CSO cry-3-DEC man ‘the woman beat the man and the man cried’

QUOTATION IN KWAZA 311

(18) ´kwε-da-si ho´Bεto atxitxi-´nũ wa´dy-ta-Ø -ki enter-1S-SWR Roberto maize-porridge give-1O-3-DEC ‘I entered and Roberto gave me maize porridge’

A narrative text may consist of one long chain of cosubordinated clauses, terminated by a matrix mood clause, which then functions as a marker of the end of the story.

1.5. Morphological ellipsis and typological characterisation It is a well-known fact that different criteria for the definition of a word may not al- ways define the same entity in a language (see e.g. Bauer 2000). In Section 1.1 a number of criteria are listed for considering person and mood marking in Kwaza as (inflexional) bound morphemes. There are, however, exceptions to criteria e) and f). Note that under certain specific circumstances verb roots can be omitted in Kwaza:

(19) [Q:] ´ja-xa-re [A:] da-´ki [or, complete:] ´ja-da-ki eat-2-INT 1S-DEC eat-1S-DEC ‘are you eating?’ ‘(yes) I am’ ‘(yes) I’m eating’

In the small dialogue in (19), the verb root of the answer ´jadaki “I’m eating” can be omitted because the (discourse) context does not leave any doubt as to which verb root is intended. In the result, daki “I am”, the root ja- “to eat” is ‘understood’, as it were, by all speech participants. In Kwaza, not only roots can be omitted, but inflex- ions as well.3 In van der Voort (2000) I have referred to these phenomena as ‘mor- phological ellipsis’. They form exceptions to the claims that e) these bound mor- phemes are indispensable for verb formation and f) that bound person morphemes do not occur in isolation. Furthermore both types of ellipsis may work together to pro- duce exceptions to criterium c); the uninterruptability of the predicate and bound inflexional suffixes. Finally, cross-linguistically speaking, criterium a) may not be tenable since there are languages such as French, which has cross-reference agree- ment, free anaphoric particles and free pronouns (as in e.g. moi, je parle vs. toi, tu parles), and languages such as Karo (Tupi-Ramarama), which has both cross- reference clitics and free pronouns (see Gabas Jr. 1999). This means that there are at least three morphological criteria on the basis of which one could claim that rather than bound inflexional suffixes, Kwaza has free morphemes that function as auxiliary verbs, particles or clitics. Nevertheless, on the basis of the other criteria mentioned in Section 1.1 one could maintain that Kwaza per- son and mood are bound morphemes. In the latter analysis, Kwaza is a morphologically complex (largely agglutinative) and in the former analysis, Kwaza is a morphologically simple, isolating language. At the present, both analyses seem to be possible.

3 Kwaza is unlike the neighbouring Tupi languages in that it does not have a special closed category of particle verbs that cannot be inflected (see Moore 2002).

312 VAN DER VOORT

For the time being, I have chosen for a synthetic approach. One of the reasons for this choice is that it results in a relatively transparent and consistent analysis of the lan- guage. The present article deals precisely with phenomena that are exceptional in synthetic languages. Furthermore, exceptions to criteria such as e) and f) are not im- possible in other, well-established synthetic languages either, such as the Tarramiut dialect of Eastern Canadian Inuktitut (see Allen 1996: 15; 27 n.13; 153 n.12). Finally, my provisional impression is that if the Kwaza person and mood combination were considered as a free particle, the entire grammar of Kwaza would have to be regarded as syntax rather than as morphology. In such an approach, many derivational mor- phemes should also be seen as free morphemes, and stress placement and word order rules would be greatly complicated. However, for a thorough understanding of the consequences of an isolating approach more research is required.

2. Quoted speech Kwaza syntax is relatively simple. Word order is rather free, and there are neither con- junctions nor complementisers. Not even quotation of speech involves such grammatical devices. In Kwaza, speech is quoted by repeating literally what was said. The quoted utterance is then embedded in an extra layer of person and mood inflexions that refer to the quoting subject. Compare the following two examples:

(20) kukui´hỹ-da-ki ill-1S-DEC ‘I am ill’

(21) kukuihỹ-da-´ki-da-ki ill-1S-DEC-1S-DEC ‘I said I am ill’

The Portuguese translation offered by the informants is usually in the form of an indirect speech quotation. However, when analysing the construction, its literal meaning repre- sents a quotation of direct speech. This suggests that in Kwaza, no formal distinction is made between direct and indirect speech. Note the following examples:

(22) kukuihỹ-da-´ki-Ø-tsε4 ill-1S-DEC-3-DEC

‘shei says shei is ill’ (lit. ‘she says ‘I’m ill’’)

4 Note that the alternative form of the declarative mood marker, -tsε instead of -ki, seems to occur only in the third person and because of independent reasons. These reasons, however, are not well understood and speculations about the alternative occurrence of the two forms are beyond the scope of this article.

QUOTATION IN KWAZA 313

(23) kukuihỹ-Ø-´ki-Ø-tsε ill-3-DEC-3-DEC

‘shei says shej is ill’ (lit. ‘she says ‘she is ill’’)

(24) maga´riDa kukuihỹ-xa-´ki-Ø-tsε Margarida ill-2-DEC-3-DEC ‘Margarida says you’re ill’

(25) maga´riDa kukuihỹ-Ø-´ki-xa-ki Margarida ill-3-DEC-2-DEC ‘you say Margarida is ill’

With respect to the notion ‘quotative construction’, the following remark is in order. It will appear from many examples in the rest of the article that this construction does not strictly express quotation of (direct or indirect) speech. In fact, with the term ‘quotative construction’ I refer first and foremost to a specific morphological construction in Kwaza in which a predicate occurs with double person marking.

2.1. Analysis of the structure of quotations The following example represents an analysis of the structure of the verb in example (25). It shows how extra inflexions are attached to a verb that is already inflected for the same categories (i.e. person and mood):

(26) [ [ kukuihỹ-Ø-´ki ] -xa-ki ] [ [ ill -3-DEC ] -2-DEC ] [ [ quoted utterance ] event of quoting ] [ [ predicate subject: ‘he’ ] matrix subject: ‘you’] lit: [ ‘ [ ‘he is ill’ ] you say’ ] ‘you say he is ill’

As we have seen in the above examples, the layer of inflexions that is closest to the verb root, the primary layer, cross-refers to the subject of the quoted utterance. The secondary layer of inflexions, i.e. the one that follows the primary layer, cross-refers to the subject of the event of quoting. Compare the structural analyses of (24) and (25), in the first of which magariDa is the quoting subject (24b), and in the second of which magariDa is the quoted subject (25b):

(24) b [ maga´riDa [ kukuihỹ-xa-´ki ] -Ø-tsε ] [ Margarida [ ill -2-DEC ] -3-DEC ] [ [ quoted utterance ] event of quoting ] [ [ predicate subject: ‘you’ ] matrix subject: ‘Margarida’ ] lit: [ ‘Margarida [ ‘you’re ill’ ] she says’ ] ‘Margarida says you’re ill’

314 VAN DER VOORT

(25) b [ [ maga´riDa kukuihỹ-Ø-´ki ] -xa-ki ] [ [ Margarida ill -3-DEC ] -2-DEC ] [ [ quoted utterance ] event of quoting ] [ [ predicate subject: ‘Margarida’ ] matrix subject: ‘you’] lit: [ ‘ [ ‘Margarida is ill’ ] you say’ ] ‘you say Margarida is ill’

One could argue that the attachment of an extra layer of inflexions has come about through cliticisation. In that case, one may just as well argue that both layers of inflex- ions are cliticised, which would require a less synthetic analysis of Kwaza (see 1.5). This is because these inflexional layers are represented by the normal inflexional morphemes of Kwaza, rather than by special clitic elements. In the first place, the forms in the pri- mary and secondary layers of inflexions are identical, i.e. the very same person and mood markers are employed. In the second place, in quotative constructions main word stress is on the last syllable of the primary layer of inflexions. This can be seen in all relevant examples. It suggests that, from the perspective of the secondary layer of inflexions, the primary layer is to be considered as part of the extended root of the verb. In the light of the observations that follow example (3) in Section 1.1, it may be best to suppose that the inflected verb inside a quotative construction is zero-derived as a verbal stem.

2.2. Other moods in the quotative construction The quoted speech construction is fully productive in Kwaza. It can occur with all dif- ferent persons and in all existing moods. In fact, entire discourse units can be quoted. In the following example an exhortative utterance is quoted:

(27) kui-a-´ni-Ø-tsε drink-1P-EXH-3-DEC ‘he wants us to drink together’ (lit.: ‘he says ‘let’s drink!’’)5

Example (28) shows that quoted utterances can be embedded in cosubordinated clauses:

(28) pεrε´jã-tja-a-´ni-da-ta oja´nỹ-da-ki speak-TRA-1P-EXH-1S-CSO arrive-1S-DEC ‘I came for us to talk’ (lit.: ‘I arrive, me saying: ‘let’s talk!’’)

The illocutionary status of the quoting event can vary. In the next example it is impera- tive. The utterance does not report on something that was said, but it orders something to be said:

5 The first person plural has a default interpretation of inclusive. If it were exclusive this would have been marked formally by the associated person marker -xa-.

QUOTATION IN KWAZA 315

(29) kwε-da-´mỹ-ca-Ø -ra enter-1S-VOL-EMP-2-IMP ‘say ‘I will enter’!’

Note that in this example, the emphatic imperative element -ca- should not be seen as forming a part of the quoted utterance. It belongs to the secondary layer of morphemes.

3. A possible origin of the quotative construction We have seen that the quotative construction is characterised by the attachment of a secondary layer of inflexions to (an already inflected) quoted utterance. One of the pos- sible origins of this secondary layer is that it has remained after omission of the root of a verb of speech. The following examples consist of a quoted utterance embedded in a matrix clause headed by a verb of speech:

(30) u´te-Ø-ta kukui´hỹ-xa-ki ´ta-ta-Ø-ki6 notify-3-CSO ill-2-DEC talk-1O-3-DEC ‘she said (to me) that I am ill’ (lit.: ‘‘you are ill’, she talked to me, notifying’)

(31) ´ta koto´rε-le tsũ´hũ xare´ja-xa-´re ta-Ø -ta CSO toad-only what search-2-INT talk-3-CSO ‘then the toad said ‘what are you looking for?’’

However, the juxtaposition of several matrix clause moods within one sentence is very rare in Kwaza and makes sense probably only when speech is quoted. I assume that if the root of the verb of speech ta- “talk” is omitted7 in these examples, the remaining combination of person and mood inflexions could become cliticised to the previous word. The result could be a quotative construction of the kind we have seen in the pre- vious sections. However, it is not at all certain that the quotative construction necessar- ily emerged from the omission of ta- “to say”. Since there are no constructions in Kwaza that have exactly the same form and the same function as the quotative con- struction, a verb with double layers of inflexions could probably be considered as a quotative construction by default.

4. The special use of the quoted interrogative Even though the quotative construction does not seem to be created for anything else in Kwaza than to quote speech, it can be used in a different way that seems to be meta- phorically related to its normal function. Note that the interpretation of the examples (27) and (28) is not strictly quotative. Their literal reading may be quotative, but they have a desiderative or purposive connotation. The purposive sense of (28) is partly an independent property which is inherent to the cosubordination of clauses. However, the

6 Note that the preceding cosubordinated clause is not part of the quoted utterance. 7 The omission of verb roots was discussed in Section 1.5.

316 VAN DER VOORT desiderative connotation of (27) and (28) may result from a merger of semantic and pragmatic aspects of the exhortative morpheme and the event of quoting. Because the pragmatic differences at this level, if they they can be investigated at all, are rather subtle, it may be helpful to look at somewhat clearer cases of metaphori- cal use of the quotative construction. The following example concerns the quotation of an interrogative utterance:

(32) (ku´kui) ja-xa-´re-da-hỹ-ki wow! eat-2-INT-1S-NOM-DEC ‘my, did you eat much!’ / ‘I say, you did eat!’

The literal meaning of ´jaxare “did you eat?” is interrogative. However, (32) has an exclamative connotation of surprise, or of being impressed. It is true that the interroga- tive itself can be used in a non-interrogative and emphatic manner, as in:

(33) kawa´pe ku´kui nỹ-´hỹ-re cockroach wow! big-NOM-INT ‘my, that cockroach is big!’

But in such expressions, the expletive element ku´kui “wow!”, “gosh!”, “damned!” is required, whereas it is optional in (32). So it seems that in (32) the quotative nature of the utterance is used as an additional strategy for emphasis. Its alternative translation shows that Kwaza is not the only language that can use a ‘verb’ of speech to create em- phasis. From the contrast between the following examples it appears that the quoted inter- rogative may also have a reflective or ruminative connotation:

(34) warañỹ-´e-da-tsy-re work-again-1S-POT-INT ‘am I (going) to work again?’

(35) warañỹ-e-da-tsy-´re-da-ki work-again-1S-POT-INT-1S-DEC ‘I think I’m going to work again’ (lit.: ‘I say ‘am I going to work again?’’)

One could suppose on the one hand that the origin of this ‘reflective’ construction does not involve the hypothetical omission of a verb of speech, but rather of some other verb of cognition, say, “to think”. However, Kwaza does not have a verb “to think”, and the verb which approaches the sense of “to think” most closely is the verb tutunita´hỹ- “to worry”. Therefore it may be better to regard the quotative construction as being quota- tive, but then only in a symbolic way. Since this construction appears to have other uses besides quotation and since it does not involve an overt verb of speech one could proba- bly also say that an ‘abstract’ verb of cognition is ‘understood’. This may become an

QUOTATION IN KWAZA 317 even more likely analysis for the phenomena discussed in the following section. The fact that the verb ta- “to say” cannot be used to refer to other mental processes8 further corroborates this approach. This fact suggests that it is not the actual verb ta- “to say” that was omitted, or gets ‘zero’ expression in the quotative construction, but something more abstract. It may be, however, that the most probable analysis of these construc- tions does not involve the omission of anything at all. Such an analysis was already hinted at at the end of Section 3.

5. Non-quotative use of the quotative construction There are two derivational modal morphemes in Kwaza that require double person marking: purposive -te- and desiderative -heta-. Consider the following examples and their translations:

(36) tso´roi-da-”te-da-ki run-1S-PURP-1S-DEC ‘I’m determined to run’

(37) tso´roi-da-”te-Ø-tsy-hỹ-ki run-1S-PURP-3-POT-NOM-DEC ‘he is supposed to run’

The purposive morpheme is preceded and followed by a person marker. However, there is no formal relation of the element -te- with one of the aforementioned mood markers. It is probably derivational because it adds a semantic value of purposive modality to the predicate. Furthermore, it occurs exclusively in this position inside the predicate, and cannot take the verb-final position of a normal mood marker in the matrix sentence. The following expression was not attested9:

(38) *tsoroi-da-te run-1S-PURP

Nevertheless, the utterances in (36) and (37) obey the structure of the quotative con- struction as if -te- were an embedded mood marker. When compared with the respec-

8 Such a phenomenon is reported to exist in several Andean languages by Adelaar (1990) and in by de Vries (1990) where an explicit quotative element is used to refer not only to actual speech but also to thoughts, wishes, intentions and other kinds of “inner speech”. In certain African languages, such as Shona (Bantu), the use of the quotative element -ti- is not even limited to “inner speech” (Güldemann t.a.). In Kwaza, such an element does not exist, but the absence (or zero-expression) of such an element in the charac- teristic ‘quotative construction’ could be regarded as having a similar function. If one would analyse Kwaza alternatively as an isolating language with auxiliary verbs, these could be regarded as semantically relatively abstract. Then the meaning of daki would depending on its context range between “I do”, “I am”, “I want”, “I say” and “I think”. Karo (Tupi-Ramarama, see Gabas Jr. 1999) has an auxiliary verbal root -e- that could be translated as “to do”, “to say”. In Kwaza, however, the auxiliary verb would not have an identifiable root. 9 But see the remarks concerning examples (46) and (47).

318 VAN DER VOORT tive quotative examples (21) and (22) above, it is clear that (36) and (37) display the same productive cross-reference properties. It is almost as if the omission of an abstract verb of cognition has led to this construction. Another explanation could be that the purposive construction is modelled after the quotative construction by analogy. One could imagine the quotative and purposive examples to share this structure:

(39) verb.stem-PERSON-MOOD/MODALITY (zero.verb.of.cognition)-PERSON-MOOD

For the following purposive examples, a literal translation involving an omitted verb of speech can be contrived:

(40) eromũtsa-da-´te-xa-ta ´nãi-xa-re10 wrist-1S-PURP-2-CSO like-2-INT ‘is it for you to wear on your wrist?’

(lit.: ‘is it for youi to (say) ‘Ii wear it on the wrist’?’)

(41) eromũtsa-xa-´te-xa-ta ´nãi-xa-re wrist-2-PURP-2-CSO like-2-INT ‘is it for you to put on my wrist?’

(lit.: ‘is it for youi to (say) ‘youj wear it on the wrist’?’)

The same observations can be made with respect to the morpheme -heta-, which is like -te- also considered as derivational on grounds of its distributional properties. Consider the following examples:

(42) cari-da-he´ta-da-ki shoot-1S-DESI-1S-DEC

‘I wanted to kill’ (lit.: ‘Ii want ‘Ii kill’’)

(43) ´ĩ cari-da-he´ta-Ø-tsε he shoot-1S-DESI-3-DEC

‘he wanted to kill’ (lit.: ‘hei want ‘Ii kill’’)

(44) ´txa kui-da-he´ta-xa-re tea drink-1S-DESI-2-INT

‘would you like to drink tea?’ (lit.: ‘do youi want ‘Ii drink tea’?’)

10 Note that the construction itself is embedded as a cosubordinated clause under a dummy matrix predicate nãi- “to be like”. Constructions involving nãi- usually have an explicative value and cosubordination itself may add up to the purposive meaning of the sentence. However, the following example (created by myself, but based on similar examples) would convey more or less the same meaning as (40):

(40) b eromũtsa-da-´te-xa-ki (HYPOTHETICAL EXAMPLE) wrist-1S-PURP-2-DEC ‘(it is) for you to put on your wrist’

QUOTATION IN KWAZA 319

(45) pεrεjã-Ø-he´ta-da-(le)-ki speak-3-DESI-1S-FRUST-DEC

‘I would like him to talk’ (lit.: ‘Ii want ‘hej talk’’)

Again, the structure is typically quotative, and the embedded person markers remain productive. The alternative person marking in (40) vs. (41) and in (44) vs. (45) go with a predictable change of perspective. Both in example (40) and (44) the first person marker -da- and the second person marker -xa- refer to the same subject. The situation is reversed in example (41) where the double marking of second person -xa- cross- refers literally to different subjects. Unfortunately, a double second person version of (44), which would represent a desiderative structural equivalent of (41), was not re- corded.11 The change of perspective in (45) is especially clear when it is placed contrast with (43). Neither -te- nor -heta- can occur in any other position than in between person markers inside the predicate. The only attested exceptions to this were elliptic (see Sec- tion 1.5):

(46) wa´dy-hata-Ø-heta give-3S2O-3-DESI ‘would he gave to you!’, ‘if only he gave (it) to you’

(47) dilε-´wã oi´tsi-da-heta who-AO sex-1S-DESI ‘(I) would like to make love with someone’

Although a purposive utterance like example (38) was never attested, I suppose that it may in principle occur as a result of morphological ellipsis. Just as in the case of (46) and (47) it will be interpretable in an appropriate context, in this case: “(I’m/you’re/he is) determined to run”. So in spite of the quotative structures, the quotative interpretations of the pur- posive and desiderative modalities are somewhat contrived. If omission of a verb root of cognition led to these constructions, it is unclear which root. It seems more likely that the morphemes -te- and -heta- are inserted via a process of analogy into a grammatical- ised and fixed quotative template.

6. The development of new modality morphemes out of mood inflexions It was explained in Section 2 that the productive quotative construction in Kwaza in- volves a double layer of person and mood inflexions. Its semantic content is also quota- tive. In Section 5, the purposive and desiderative modality morphemes were shown to

11 It would certainly be grammatical:

(44) b ´txa kui-xa-he´ta-xa-re (HYPOTHETICAL EXAMPLE) tea drink-2-DESI-2-INT

‘would you like me to drink tea?’ (lit.: ‘do youi want ‘youj drink tea’?’)

320 VAN DER VOORT occur in a very similar construction. The difference, however, is that a quotative inter- pretation is hardly possible and that the purposive and desiderative elements should probably not be regarded as inflexions. Apart from these two types of ‘quotative’ constructions, there is a third kind of construction which holds a position somewhere in between. It involves derivational modal elements which have apparently developed from person and mood inflexions. It does not involve, however, productive word-internal person inflexion. Compare the following example to example (7) above:

(48) ´ja kui-´nĩ-da-ki already drink-CAUS-1S-DEC ‘I already let (him) drink’ (lit., ±: ‘I already said: ‘let him drink!’’)

The causational12 element -nĩ- in example (48) bears a strong phonetic resemblance to the exhortative mood marker -ni and is probably derived from it. I suppose it could have emerged from a process of change of grammatical status of the exhortative ele- ment embedded in a quotative construction. With only a little effort, (48) can be read as a quoted exhortative. This, however, is less easy in the following examples:

(49) kuraku´ra ja-´dy-da-ki ũi-´nĩ-da-ta chicken eat-CAU-1S-DEC lie-CAUS-1S-CSO ‘I feed the chickens so that they can go to sleep’

(50) hadai-´nĩ-da-ki hack-CAUS-1S-DEC ‘I cut myself (by accident)’13

Whereas (48) is still somewhat quotative, (49) and (50) can hardly be interpreted as quotative on semantic grounds. These latter two examples suggest that the original in- flexional element -ni is on its way to become a new derivational element -nĩ-. Of course the possibility cannot be excluded that both elements are not etymologically related at all, and that their near homophonousness is a matter of coincidence. However, the deri- vational causational is not the only modality suffix that has a similar (both formally and semantically) inflexional mood counterpart in Kwaza. Compare the following example to the monitory example (10) in the introduction:

(51) a-´wy wotsu-´tsi-da-ta Ø-time skinny-MON-1S-CSO ‘I do it (feeding the cattle) before they become emaciated’

12 The causational modality is different from the valency-increasing causative morpheme -dy-. 13 This example is not reflexive. In the reflexive version of (50) the morpheme -nỹ- would have occurred in the place of the causational morpheme, and the subject would have cut himself on purpose.

QUOTATION IN KWAZA 321

Here a quotative reading is difficult to conceive. It may be that a derivational modal morpheme with a ‘preventive’ semantic content, “lest”, is developing from the moni- tory mood inflexion. The volitive mood seems to have made it even further down the road to ‘lexicalisa- tion’. In a fossilised combination with the first person singular cross-reference marker -da-, the volitive mood morpheme -mỹ is attested as a derivational morpheme and a verb root (-)damỹ-, meaning “to want”. Compare the following examples:

(52) ε-da-´mỹ14 vs. ´ε-da-ki go-1S-VOL go-1S-DEC ‘I’m going!’ ‘I went’, ‘I am going’

(53) ε-da-´mỹ-xa-re vs. ´ε-xa-re go-1S-VOL-2-INT go-2-INT ‘are you going?’ ‘are you going?’

The first example in (53) could still be regarded as a quoted speech construction in which the volitive and the first person morphemes work together to produce the pre- dictable literal meaning of: “do youi say ‘I’mi going!’?”. In the following example, no such quotative reading is likely anymore:

(54) tãlo-da´mỹ-Ø-tsε or: tãlo-´nã-Ø-tsε angry-want-3-DEC angry-FUT-3-DEC ‘it is becoming angry!’ (so watch out for that dog)

If (54) is to be interpreted quotatively, it would be interesting to know more about the linguistic abilities of the dog, at least if the example is supposed to contain direct speech. But also an indirect quotation of speech would be difficult to conceive. An in- terpretation that involves an (omitted) semantically abstract verb of cognition would then be more likely: “iti thinks/feels/growls ‘Ii will become angry!’”. Now consider the following example:

(55) bwa-da´mỹ-Ø-tsε finish-want-3-DEC ‘it is about to run out’ (the gas of the cigarette lighter)

In example (55) a strictly quotative interpretation is impossible. But even a more ab- stract cognitive interpretation would not make sense. Maybe these examples involve quotation in a metaphorical manner. This is also attested in other languages, where in- animate beings may be said to ‘say’ something if they are likely to produce a sound. In colloquial Dutch, for example, a firecracker may ‘say ‘bang’’. In (55), however, there is

14 Note that the stress pattern ´εdamỹ “I’m going!” was also attested.

322 VAN DER VOORT no auditive connotation at all. Furthermore, the volitive mood morpheme can only be applied to ‘controlled’ verbs. This means that the volitive mood -mỹ is anomalous on verbs where the subject has no ‘control’ over the event, such as e.g. the event of getting a fever. This restriction does not apply, however, to the petrified volitional combination -damỹ, as is also shown by example (55). So it is perhaps better to say that the first per- son singular volitive set of inflexions -da-mỹ has developed into a different, deriva- tional morpheme that has a general meaning that ranges over intentional or volitional modality and ingressive aspect.15 An additional reason to see -damỹ as a separate, new morpheme is that the first person marker that it contains is fossilised. In the word- internal positions in which it occurs in the above examples, -da- could never be substi- tuted for another person marker, whereas it can in its normal matrix clause use, as illus- trated by example (5) in the introduction. The next examples show that damỹ- is even used as a lexeme, a verb root that means “to want, to intend, to be going to” or a particle that means “yes! (I will)”:

(56) da´mỹ-xa-xa-ki want-2-AS-DEC ‘you (pl) are going to do (it)’

(57) da´mỹ-Ø-tsε want-3-DEC ‘he is going (or wanting) to do (it)’, ‘he says ‘yes (I will)!’’

The latter example can still be seen as quotative. According to that interpretation, as many as two verb roots may have been ‘omitted’: a verb of speech and another verb root that is ‘understood’ from the specific discourse context of the utterance, as in (19).

7. Comparable developments in Quechua and Eskimo The phenomenon of inflexional morphemes developing into derivational morphemes as sketched in Section 6 may not be unique to Kwaza. According to Muysken (1977:105- 107) a similar process is likely to have led to new aspectual morphemes in Ecuadorian Quechua. The following example contrasts an older pattern, involving a nominalised verb and an inflected auxiliary verb, with a more recent construction, involving a (deri- vational) morpheme:

(58) miku-k ri-ni > miku-gri-ni eat-NOM go-1SG eat-INC-1SG ‘I am going to eat’ ‘I am going to eat’

15 Note that in many languages of the world, the expression of future or ingressive aspect of inanimate and non-controlled events may involve cognitive modal verbs or suffixes with a meaning like “to want”. In languages as diverse as Indo-European, Eskimo-Aleut (see Johns 1999 for Inuktitut) and Kwaza one can say things like “it wants to rain”. The peculiar fact of Kwaza is that this modal element includes a fossil- ised first person marker.

QUOTATION IN KWAZA 323

Muysken assumes that the word boundary between the two verbs has disappeared, and that the (inflexional) nominaliser -k of the first word has merged with the root ri- “to go” of the second word, which has led to a new inchoative morpheme -gri-. One of the motivations for this historical development could have been the necessity to differenti- ate between emphatic and non-emphatic forms. Also in the Eskimo languages, comparable developments seem to have taken place. In Kalaallisut, the polysynthetic suffixing language of the West Greenlandic Inuit, nominal oblique case markers may have developed into verbal directional morphemes.16 Consider the following ablative example (from Olsen & Hertling 1988:38):

(59) Nuum-mit vs. Nuum-meer-poq Nuuk-SG.ABL Nuuk-be.from-IND.3SG ‘from Godthåb’ ‘he comes from Godthåb’

In this type of construction inflexional aspects such as person and number marking remain productive, as is shown by the following allative and prosecutive examples (from Kristoffersen 1991:16 and Bergsland 1955:92):

(60) illu-anut vs. illu-anu-kar-poq house-3SG.POS.ALL house-3SG.POS.ALL-go.to-IND.3SG ‘to her house’ ‘he went to her house’

(61) nuna-p ilu-atigu-le-raa-ngatta land-REL.SG inner-3SG.POS.PROS-start-every.time-CAU.1PL ‘always when we start to go through the land’

These constructions form an exception to the standard approach of Greenlandic as a suffixing language, according to which obligatory word structure and morpheme order is root-derivation-inflexion. The elements that follow the case marker in the above ex- amples are not considered as clitics, but as derivations, that are followed again by obligatory inflexions. This phenomenon occurs also in the Canadian, Alaskan and Sibe- rian Eskimo languages, and it is usually referred to as ‘postinflectional morphology’ (e.g. De Reuse 1994:170-230; Sadock 1991:174-175).17 In Eskimo, postinflexional constructions based on inflected verbs are also attested, however seldomly. The following Central Alaskan Yupik example from Sadock (1991:175) involves the “very rare postinflexional clitic-like affix” -Vr- “to say”:

16 See the following entries in Fortescue et al. (1994:403;411;412 respectively): -k(k)u(C)ar- “go via”, -mt- “be in or at”, -muaq- “go to(wards)”. 17 In the Danish literature it may be known as indre bøjning “internal inflexion”, (Berthelsen et al. 1998:128).

324 VAN DER VOORT

(62) Liisaq-una tai-gu-ur-tuq Lisa(ABS)-this.one(ABS) come-IND.3SG-say-IND.3SG ‘Lisa said this one is coming’

The same affix is attested in West Greenlandic, where it has the connotation of “to shout”. Consider the following example from Schultz-Lorentzen (1967:350):

(63) palasi agger-po-or-pata priest come-IND.3SG-say-COND.3PL ‘when they shout ‘the priest is coming!’’

The contrast between the following Greenlandic examples, from Fortescue (1984:3) and Kristoffersen (p.c.) respectively, clearly demonstrates productive internal plural inflexion:

(64) umiar-sua-ar-poq vs. umiar-sue-er-put boat-big-say-IND.3SG boat-big.PL-say-IND.3PL ‘he shouted ‘a ship!’’ ‘they shouted ‘ships!’’

According to its treatment in Fortescue et al. (1994:423), the morpheme -Vr- “to say” is considered as a bound derivational morpheme. Because it is obligatorily followed by inflexional suffixes it does not belong to the category of clitic affixes in Eskimo, even though its attachment behaviour is similar.

8. Conclusion In this article I have described the quotative construction in Kwaza as a specific gram- matical construction that is characterized by word-internal person inflexion. It has a limited and very specific use in Kwaza and does not occur elsewhere in the language. The following table summarises the different appearances and semantic values (“func- tions”) of the quotative construction. The parameter ‘quotativity’ concerns the possibil- ity of a quotative interpretation:

SECTION STRUCTURES AND FORMS FUNCTIONS QUOTATIVITY EXAMPLES 2 -person-mood-person-mood quotation yes (21) 5 -person-te-person-mood purpose no (40) 5 -person-heta-person-mood wish no (44) 6 -1S.volitional-person-mood intention yes/no (55) 6 -causational-person-mood causation yes/no (50) 6 -preventive-person-mood warning yes/no (51) Table II: The different functions and forms of the quotative construction

Do the phenomena described and discussed here represent a kind of lexicalisation or grammaticalisation? When a morpheme becomes lexicalised, it turns into a lexeme or it

QUOTATION IN KWAZA 325 becomes a fossilised part of another morpheme or lexeme. This seems to be indeed the case in damỹ-, at least as a new lexical verb root “to want”, in which the first person marker -da- has become petrified together with the volitive marker -mỹ, has lost its original first person sense and cannot be substituted for another person marker. But lexicalisation is not the only process of grammatical change in Kwaza. It seems also that certain inflexional elements have become derivational. This was also shown in Sec- tion 6. The phenomenon that grammatical elements may develop out of lexical (or con- tent) words is called grammaticalisation. As pointed out in Hopper and Traugott (1993:7), most linguists agree that such developments in a language are subject to a ‘cline of grammaticality’ that goes in the following direction:

content item > grammatical word > clitic > inflexional affix

It is a much recurring phenomenon in grammaticalisation changes that the original form, on which a certain grammaticalised form is based, does not disappear, but contin- ues to exist with (traces of) its original meaning. The new form does not make the old one obsolete. Rather, new distributional possibilities are opened for the original form. This characteristic of grammaticalisation is referred to as ‘divergence’ by Hopper and Traugott (1993:117). If Kwaza were to be analysed as a language that is syntactically complex rather than morphologically complex one could regard the person and mood combination as an independent auxiliary verb. Such auxiliaries are then grammatical- ised in the ‘quotative construction’, where they are cliticised. As explained in Section 1.5, I have chosen not to adopt this approach. Only a few linguists assume the existence of processes that go in the opposite di- rection of the cline of grammaticalisation. In Norde (1997, 2001) data from Swedish are presented to argue for the existence of such counterdirectional developments. Norde considers the change of certain inflexional morphemes into derivational morphemes in the historical development of Swedish as an instance of ‘degrammaticalisation’. Unlike Hopper and Traugott (1993), she and several other linguists assume that derivational morphology has a lower grammatical status than inflexion, and that it should be placed to its left on the cline of grammaticality. An important aspect of the definition as em- ployed by Norde is that degrammaticalisation differs from lexicalisation in that the for- mer is a gradual phenomenon. Furthermore, degrammaticalisation cannot be regarded as the reverse of grammaticalisation; it is a phenomenon of linguistic change in its own right. Finally Norde appears to assume that ‘deflexion’ is a prerequisite of degrammati- calisation, i.e. degrammaticalisation crucially involves loss of original grammatical categories. In the present analysis of Kwaza as a morphologically complex language, counter- directional developments can be argued to have taken place. The data presented here call for an analysis in terms of degrammaticalisation. A conspicuous aspect of degram- maticalisation in Kwaza is that the original inflexional elements, from which deriva-

326 VAN DER VOORT tional and lexical elements have derived, did not undergo ‘deflexion’ themselves. That is, the volitive, exhortative and monitory moods are still productive inflexional mor- phemes, in spite of the fact that volitional, causational and preventive derivational mor- phemes seem to have developed from them. Contrary to what the Swedish data may suggest, the data from Kwaza indicate that the abovementioned phenomenon of ‘diver- gence’ can also be observed in degrammaticalisation changes. With respect to pur- posive -te- and desiderative -heta-, deflexion may have taken place. It is a pity, how- ever, that no earlier stages of the language have been documented. The true quotative expressions in Section 2 are not considered as degrammatical- ised in Kwaza. They are completely transparent and productive, and in this they resem- ble the Eskimo constructions with word-internal inflexion. The difference is that in Es- kimo the embedded number, possessor and oblique case inflexions and the embedded person and mood inflexions are followed by an overt derivational morpheme. They are not zero-derived as in Kwaza.

References Adelaar, Willem F.H. 1990 ‘The role of quotations in Andean discourse’, in: Pinkster & Genee (eds.), pp 1- 12. Allen, Shanley E.M. 1996 Aspects of argument structure in Inuktitut, Language aquisition & language dis- orders vol. 13, Amsterdam/Philadelphia: John Benjamins Publishing Company. Bauer, Laurie 2000 ‘Word’, in: Geert Booij, Christian Lehmann, Joachim Mugdan et al. (eds.), Morphologie: Ein internationales Handbuch zur Flexion und Wortbildung, Handbücher zur Sprach- und Kommunikationswissenschaft vol 17.1, Berlin/New York: Walter de Gruyter, pp. 247-257. Bergsland, Knut 1955 A grammatical outline of the Eskimo language of West Greenland, [typoscript], Oslo: Skrivemaskinstua. Berthelsen, C., B. Jacobsen, R. Petersen, I. Kleivan & J. Rischel (eds.) 1998 Oqaatsinut Tapiliussaq / Oqaatsit Supplementsbind. Nuuk: Atuakkiorfik Ilin- niusiorfik. Foley, William A. and Robert D. Van Valin 1984 Functional Syntax and Universal Grammar, Cambridge Studies in Linguistics 38, Cambridge: Cambridge University Press. Fortescue, Michael 1984 West Greenlandic, London etc.: Croom Helm. Fortescue, Michael, Steven Jacobson & Lawrence Kaplan 1994 Comparative Eskimo dictionary, with Aleut cognates, Fairbanks: Alaska Native Language Center.

QUOTATION IN KWAZA 327

Gabas Junior, Nilson 1999 A grammar of Karo, Tupí (Brazil), [PhD thesis], Santa Barbara: University of California. Güldemann, Tom t.a. ‘When ‘say’ is not say: The functional versatility of the Bantu quotative marker ti with special reference to Shona’, in: Tom Güldemann & Manfred von Roncador (eds.), Reported discourse as a meeting ground for different linguistic domains, Typological studies in language, Amsterdam/Philadelphia: John Benjamins Pub- lishing Company. Hopper, Paul J. & Elizabeth Closs Traugott 1993 Grammaticalization, Cambridge Textbooks in Linguistics, Cambridge: Cam- bridge University Press. Johns, Alana 1999 ‘On the lexical semantics of affixal ‘want’ in Inuktitut’, in: International Jour- nal of American Linguistics, 65/2:176-200. Kristoffersen, Lars 1991 Verbal derivation and inflection in a functional grammar of West Greenlandic, [master’s dissertation], Copenhagen: Københavns Universitet. Moore, Denny 2002 ‘Verbos sem flexão’, to appear in: Atas do I encontro internacional do Grupo de Trabalho de Linguas Indígenas, Brazil. Muysken, Pieter 1977 Syntactic developments in the verb phrase of Ecuadorian Quechua, [doctoral dissertation], Lisse: The Peter de Ridder Press. Norde, Muriel 1997 The history of the genitive in Swedish: A case study in degrammaticalization, [doctoral dissertation], Amsterdam: Universiteit van Amsterdam, Vakgroep Skandinavische Taal- en Letterkunde. 2001 ‘Deflexion as a counterdirectional factor in grammatical change’, in: Language Sciences, [special issue edited by Lyle Campbell: Grammaticalization: a critical assessment], 23: 231-264. Olsen, Lise Lennert & Birgitte Hertling 1988 Grønlandsk tilhængsliste, Pilersuiffik. Pinkster, Harm & Inge Genee (eds.) 1990 Unity in Diversity: Papers presented to Simon Dik on his 50th birthday, Dordrecht/Providence R.I.: Foris Publications. Reuse, Willem J. de 1994 Siberian Yupik Eskimo: The language and its contacts with Chukchi, Salt Lake City: University of Utah Press. Rodrigues, Aryon Dall´Igna 1986 Línguas Brasileiras: Para o Conhecimento das Línguas Indígenas, São Paulo: Edições Loyola.

328 VAN DER VOORT

Sadock, Jerrold 1991 Autolexical syntax: A theory of parallel grammatical representations, Chi- cago/London: The University of Chicago Press. Schultz-Lorentzen, Christian W. 1967 Den grønlandske Ordbog: Grønlandsk - Dansk, [facsimile edition of 1926], S.L. Møller. Voort, Hein van der 2000 A grammar of Kwaza: A description of an endangered and unclassified indigenous language of Southern Rondônia, Brazil, [doctoral dissertation], Leiden: Universiteit Leiden. Vries, Lourens J. de 1990 ‘Some remarks on direct quotation in Kombai’, in: Pinkster & Genee (eds.), pp. 291-309.

Abbreviations A: answer NEG negative ABL ablative case NOM nominaliser ABS absolutive case PL plural ALL allative case POS possessive AO animate object POT potential AS associated person PROS prosecutive case C consonant PURP purposive CAU causative (a mood in Q: question Eskimo examples) REL relative/genitive case CAUS causational SG singular CONC concessive SWR switch reference mood COND conditional mood TRA transitiviser CSO cosubordinating mood V vowel DEC declarative mood VOL volitive mood DESI desiderative 1O 1st person object DS different subject 1P 1st person plural EMP emphatic 1PL 1st person plural EXH exhortative mood 1S 1st person singular FRUST frustrative 2 2nd person singular FUT future 3 3rd person IMP imperative mood 3SG 3rd person singular INC inchoative 3S2O 3rd person subject, 2nd IND indicative mood person object INT interrogative mood - morphemic boundary IS indefinite subject . separates semantic units in MON monitory mood a portmanteau morpheme NEE negative exhortative = composition or clitic NEI negative imperative boundary

APPLICATIVE AFFIXES IN PERUVIAN AMAZONIAN LANGUAGES

Mary Ruth Wise SIL International

1. Introduction Affixes, which promote oblique noun phrases to arguments of the verb, are found in the verbal morphology of Peruvian Amazonian languages representing at least six language families.1 “The semantic content [of the affixes] falls within the range gen- erally attributed to so-called oblique relations” (Craig and Hale 1988:312). Since the most common functions of these affixes are valence-increasing, “applicative” is the cover term used for them in this paper. Nevertheless, the term is not completely satis- factory because in some cases they do not change the transitivity, or they may even decrease it. Applicatives were first described, as far as I can ascertain, by Carochi in 1645 for . According to him they serve to designate an “order[ing of] the action of the verb towards another person or thing, connecting it to him (atribuyéndosela) by way of damage or benefit, taking it off him, or putting it on him, or relating it to him (refiriéndosela) in any way whatever” (1645:166; quoted from Tuggy 1981:413). A given “applicative”—as the term is used here—may define the semantic role of the argument; or, a general applicative, may simply indicate that the argument is some- how involved in the action of the verb. Data from the following languages are presented to illustrate the range of func- tions of the applicatives in Peruvian Amazonian languages: Chayahuita (Cahua- panan), Arabela (Zaparoan), Yagua (Peba-Yaguan), Yaminahua (Panoan), Yanesha (Maipuran Arawakan), Nomatsiguenga (Pre-Andine Maipuran Arawakan).2 Ashéninka (Pre-Andine Maipuran Arawakan) and Iquito (Zaparoan) are mentioned briefly. The data are then summarized in order to look for correlates with other grammatical characteristics.

2. Data The basic word orders of the languages discussed are SOV (Chayahuita, Arabela, and Yaminahua), SVO (Iquito), and VSO (Yagua, Yanesha, Nomatsiguenga, and Ashen- inka). The affixes discussed are all suffixes; one applicative suffix in Noma- tsiguenga appears to be cognate with a causative prefix. All of the languages dis- cussed have morphological causatives.

1 I am grateful to Simon van de Kerke and Mily Crevels for helpful comments on an earlier version of this paper. 2 The term “Pre-Andine” is used by David Payne (2000) to refer to the group of Maipuran Arawakan languages formerly classified “Campa.” It does not include Yanesha (Amuesha) and Piro, which were listed as Pre-Andine in some previous descriptions such as Wise (1986). 330 WISE

2.1. Chayahuita (Cahuapanan) “In Chayahuita the applicative suffix -të/-ta may verbalize (1a); transitivize (1b); change impersonal verbs to intransitive (1c); detransitivize [i.e., change transitive to intransitive] (1d); and change transitive to ditransitive (1e)” (Wise 1999:327).3 (In Chayahuita the third singular object is zero; -r is ‘indicative’, -in is ‘third singular agent or subject’.)

(1) a ira ‘trail’ (noun) ira-të-r-in ‘he/she walks’ b ama-r-in ‘he/she bathes’ ama-të-r-in ‘he/she bathes him/her’ c tashi-r-in ‘becomes night’ tashi-të-r-in ‘it becomes night in the (lit. ‘it nights’) place where he/she is’ (lit. ‘it nights on him/her’) d nati-r-in ‘he/she obeys nati-të-r-in ‘he/she obeys’ him/her’ e apa-r-in ‘he/she sends it’ apa-të-r-in ‘he/she sends it to someone’ (Hart 1988:269-270)

Chayahuita is a nominative–accusative SOV language with switch-reference suffixes in subordinate verbs. The desiderative, reflexive, and causative morphemes are pre- fixes and there is a set of bound roots, which occur in compound verb stems. These bound roots could be considered either prefixes or noun incorporation. Otherwise, verbal affixes are suffixes, including those which cross-reference the arguments of the verb.

2.2. Arabela and Iquito (Zaparoan) In Arabela the suffix -ta-/-tia is not always valence-changing. It may imply that the subject is complex or plural, e.g., a monkey and its baby although the baby is left implicit. Some of the meanings added to the basic verb by this suffix, which is glossed ‘applicative’ for convenience, include: compassion on the part of the speaker; passive accompaniment (2); abnormal condition (3); subject or object contained or containing something, e.g., carrying something as in (4);4 the subject is sick or old or wounded, as in (5); the verb has to do with a goal or reason.

(2) napa naana -akua tiuu -tia -a guacamayo tree in perch APPL CONT ‘A guacamayo is perched (with it’s mate) in the tree.’ (Rich 1999:55)

3 Since the functions of the Chayahuita suffix glossed ‘applicative’ are so diverse, Bill Henning (p.c.) suggests that -të/-ta might be more appropriately glossed ‘reverse default valence’. Thus, nati- ‘to obey’ is by default transitive while ama- ‘to bathe’ is by default intransitive. The symbol  in Chayahuita and Yanesha examples is used for a glottal stop which is part of the syllable nucleus. 4 Alternatively, -tia could be considered transitivizing in (4), that is, ‘I stumbled my manioc.’ APPLICATIVE AFFIXES 331

(3) kia maka -ta -re tee pa kia -nu -taniya -ni 2SG climb APPL IRR where 1PL.INCL go INF FUT SCL ‘Climb (the tree in order to see) where we should go (having lost the trail).’ (Rich 1999:56)

(4) kua morehaka tiurii -tia -ree -nihia 1SG manioc stumble APPL COMPL 1SG ‘I stumbled while carrying my load of manioc.’ (Rich 1999:431)

(5) haniya -ri nu -koko -hi kua shikiorta -ashi 1SG S trail by of 1SG hurt CL

roshi -yo -ko -ho -ta -rikio -wa -ni drag CONT CONTACT MULTIPLE APPL IMPERV REPET 1R ‘With my wounded [leg] paining me at each step, ‘I dragged myself along the trail and returned [home].’ (Rich 1999:56)

An additional function unique to Arabela (and other Zaparoan languages?) is that the applicative occurs in the main verb when the subject of the infinitival complement is different from that of the main clause. Compare, for example, (6) where the subject of the verbal complement is the same as that of the main clause with (7) where the subjects are different. (Example 3 might also be a case of this construction.)

(6) haniya kia -ta kia -nu pani -ya -ni 1SG 2SG COM go INF want CONT 1R ‘I want to go with you.’ (Rich 1999:91)

(7) haniya kia pani -tia -a kia -nu -ni 1SG 2SG want APPL CONT go INF 1R ‘I want you to go.’ (Rich 1999:91)

Note that the applicative -ta/-tia in (7) and the sociative or comitative postposition -ta in (6) appear to be cognate. (This was first pointed out by Doris Payne 1984.) The suffix -ta/-tia is apparently the only applicative in Arabela. Some pronomi- nal forms are proclitics and others are enclitics; they are not cross-referencing. Verbal affixes are all suffixes. Arabela appears to be a nominative–accusative SOV language although there are some traits which may indicate it is a split-ergative language. Iquito, another Zaparoan language, also has the applicative affix -ta. Eastman and Eastman (1961) describe it as both transitivizing and intransitivizing, i.e. reflex- ive. (R. Rich, p.c., suggests that this may not be the correct analysis. Recall, however, that Chayahuita -të/-ta has both functions also.) Iquito word order is SVO.

332 WISE

2.3. Yagua (Peba-Yaguan) In Yagua, also, there is a valence-increasing applicative suffix -ta/-tya. T. Payne (1997:187) states that this suffix indicates that the promoted argument is a locative (8b), an instrument (9b), or comitative. Compare (9b) with the postposition -tya in (9a). In (8b) -ta occurs with an intransitive verb and in (9b) with a transitive. (As in Arabela, the forms -ta and -tya are allomorphs.) In Yagua the clitic cross-referencing the direct object always immediately precedes it. In (9a) -ra/-rya shows that meat is the direct object while in (9b) ‘knife’ is.

(8) a sa- duu rá -viimú 3SG blow INAN into ‘He blows into it.’

b sa- duu -tá -ra 3SG blow APPL INAN.OBJ ‘He blows it.’ (T. Payne 1997:187)

(9) a si- ichití -rya javanu quiichi -tya 3SG poke INAN.OBJ meat knife INST ‘He poked the meat with the/a knife.’

b si- ichití -tya -ra quiichiy 3SG poke APPL INAN.OBJ knife ‘He poked something with the knife.’ (T. Payne 1997:187)

Yagua is a nominative–accusative VSO language with cross-referencing pronominal clitics; the cross-reference to the subject is a verbal proclitic while the cross-reference to the object is encliticized to the word preceding the object, as in (9b). Apart from the pronominal proclitics all affixes are suffixes. Classifiers occur in nouns and ad- jectives but are not incorporated into the verb except to derive a noun.

2.4. Yaminahua (Panoan) In Yaminahua there are three applicative affixes each of which increases transitivity. The promoted noun phrase has the form of an object—absolutive in the case of a noun but accusative in the case of a first or second person pronoun. The applicative suffixes are -xon ‘benefactive’, as in (10); -ni/Ø (following n) ‘malefactive’ (11); and -kin ‘comitative’ (12b).

(10) en mia waka we -xon -non I you water bring BEN FUT ‘I will bring you water/I will bring water for you.’ (Faust and Loos, fc)

APPLICATIVE AFFIXES 333

(11) oa noko wenen min awara mia win(Ø) -a this person man (ERG) you thing you rob(MAL) COMP ‘This man has robbed you of something/This man has robbed something from you.’ (Faust and Loos, fc)

(12) a man wake kawasan -i already child walk PROG ‘The child is already walking.’

b kawasan -kin -we min exto walk COM IMPV your brother ‘Accompany your brother/Go with your brother.’ (Faust and Loos, fc)

Yaminahua is an ergative–absolutive language except for first and second person pronouns, which are nominative–accusative. (If the benefactive and malefactive ob- jects -mia in (10) and (11) were nouns, the form would be absolutive rather than ac- cusative.) Word order is SOV (AOV) with switch-reference suffixes in the verb which also indicate the temporal/logical relation of the subordinate clause to its ma- trix clause and whether or not the verb of the matrix clause is transitive or intransi- tive.5 All affixation is suffixal except for bound forms of nouns which may be incor- porated as verbal prefixes. There is no cross-referencing of arguments in the verb.

2.5. Yanesha (Maipuran Arawakan) In Yanesha (Amuesha) there are three applicative affixes; the first two -n/-on/-nan/-ñan ‘benefactive/malefactive’, and -apr ‘sociative/comitative’, always increase transitivity, as seen in (13) and (14) where en- ‘look for’ transitive is ditran- sitivized, (15) where w-‘come’ intransitive is transitivized. A verb with the comita- tive -apr may be detransitivized, i.e., intransitivized, by the addition of -ann ‘recipro- cal’, as shown in (16).

(13) p- en -apr -et -eerr -en ne - roor -eer 2SG look.for COM EP REP 1SG 1SG flower POSS ‘Help me look for my flower/Look for my flower with me.’ (Duff-Tripp 1997:99)

(14) p- en -err -n -an n- akosh 2SG look.for REP BEN 1SG 1SG needle ‘Look for my needle for me.’ (Duff-Tripp 1997:173)

5 See Valenzuela (2000) for discussion of a possible diachronic relationship between switch-reference (same subject) markers and applicatives in Shipibo-Konibo (closely related to Yaminahua). 334 WISE

(15) Ø- w -ahp -on -ay -a 3SG come ARR BEN 1PL REFL ‘It arrived on (against) us (an illness).’ (Wise, field notes)

(16) Ø- rr -apr -et -ann -at -eet 3SG eat COM EP RECIP EP 3 PL.REFL ‘They ate together/They accompanied one another in eating.’ (Duff-Tripp 1997:99)

The functions of the other applicative suffix -amypy/-apy are more varied, as in the cases of -të/-ta in Chayahuita and -ta/-tia in Arabela. The applicative -amypy/ -apy usually adds another argument, as in (17). In some cases it simply changes the meaning of the verb stem, as shown in (18).6

y y (17) Ø- e- t om -amp -s -as s- aneets- er 3SG CAUS burn APPL 2 PL EP 2PL village GEN ‘He/she burned your village (to your detriment).’

(18) a Ø- kow -een -aan chesha -tyoll 3SG look CONT OBJ.FOLLOWS child DIM ‘He/she is looking at the small child.’

b Ø- kow -amypy -een -aan chesha -tyoll 3sg look APPL CONT OBJ.FOLLOWS child DIM ‘He/she is caring for the small child.’ (Duff-Tripp 1997:99)

The suffix can be ‘benefactive/malefactive’, as shown in (19a). An impersonal verb (20), or one in which the subject is normally only third person (21), can be changed into a regularly conjugated intransitive verb, as in (20) and (21). Thus, an almost ava- lent verb becomes monovalent when suffixed by -amypy.

(19) a ye- mayoch -amypy -een -s -a 1PL pray APPL CONT 2PL REFL ‘We pray for you (pl) (along with others).’

b ye- mayoch -een -s -a 1PL pray CONT 2PL REFL ‘We pray to you (pl).’ (Duff-Tripp 1997:100)

6 Les Bruce (p.c.) suggests that (18b) could be considered to increase transitivity in the sense that the actor is more involved. APPLICATIVE AFFIXES 335

(20) a ye- chekmy -et -amypy -es tyooñ -o 1PL to.night EP APPL EP trail LOC ‘The darkness overtook us (while we were still) on the trail.’

b Ø- chekmy -et -een tyooñ -o 3SG to.night EP CONT trail LOC ‘It’s dark/night along the trail.’ (Duff-Tripp 1997:100)

(21) a y- ahn -om -amypy -s -a 1PL to.fly COMP APPL EP REFL ‘We flew (in the plane).’

b Ø- ahn -om -a 3SG to.fly COMP REFL ‘It (the bird) flew.’ (Duff-Tripp 1997:100)

The applicative may also indicate that the subject (21a) or object (22) is in a con- tainer or that the object is somehow included in the action (23).

(22) y- atyoor -amypy -een -y ony 1PL to.fan APPL CONT 3SG water ‘We fan the fire to heat the water.’ (lit. ‘we blow/fan the water contained’) (Duff-Tripp 1997:100)

(23) Ø- anos -amypy -s -aan -a rreera po- choy -oor 3SG climb APPL EP OBJ.FOLLOWS REFL hawk 3SG child POS ‘He climbed up to where the hawk’s chicks were.’ (Duff-Tripp 1997:100)

Yanesha is a nominative–accusative VSO language. The subject cross-referencing morphemes and one of the causatives are prefixes; all other verbal affixes are suf- fixes. Nouns and classifiers may be incorporated in the verb.

2.6 Nomatsiguenga (Pre-Andine Maipuran Arawakan) Nomatsiguenga and the other Pre-Andine Maipuran Arawakan languages are consid- ered by Thomas Payne as probably having “the most highly developed systems of morphologically distinct applicative operations on earth” (T. Payne 1997:190). One verb can contain several applicative markers. Example (24) is from the closely re- lated Pajonal Asheninka. (Both applicative affixes, -ako and -imo, occur in Nomatsiguenga.)

336 WISE

(24) no- p -ako -ts -imo -tsi -ro -ri Irena Irokarto paño 1SG give APPL EP PRES ASP 3f 3m Irene Richard scarf ‘I gave Richard the head scarf in Irene’s presence.’ (Shaler 1971:45)

The general applicative -ako/-oko/-ko, illustrated in (25) and (26), has a whole range of meanings which can be summarized as “the action is somehow in reference to the object or the object is somehow involved.”7

(25) i- samë -ko -k -e -ro i- gisere 3m to.sleep APPL ASP NF 3f 3m comb ‘He went to sleep with reference to his comb (e.g., he was making it and dropped it’).’ (Wise 1971:50)

(26) i- komo -t -oko -k -e -ri pabati otsegoha 3m dam.stream EP APPL ASP NF 3m father stream ‘He dammed the stream with reference to father (father was the leader of the project).’ (Wise 1971:50)

Other applicatives are listed and exemplified in (27)-(33). Although T. Payne (1997:190-191) describes them as applicatives, some analysts might consider them to be simply valence-changing affixes. (See Aikhenvald 1999:90-92 for discussion of these affixes in “applicative derivations”.) The instrumental suffix -an/-ant in (30), for example, is listed as an applicative since it promotes an oblique noun phrase to object. Without -ant in the verb, the object suffix -ro ‘3f’ could not occur and abio ‘plane’ would be suffixed by -kë ‘in, at, by, with’.

(27) -bi/-birí ‘because, for, why, because of’ paíró pi- á -birí -k -e Iríma -kë why 2 go REAS ASP NF Lima LOC ‘Why did you go to Lima?’ (Shaver 1996:47)

(28) -así ‘purposive (action done with some purpose in view), for’ ni- pok -así -k -i -mi 1 come PURP ASP F 2 ‘I came for you (to take you).’ (Shaver 1996:47)

(29) -pí ‘with respect to, in relation to’ i- sigo -pí -t -ë -na 3m run RESP EP NF 1 ‘He ran away from me.’ (Shaver 1996:47)

7 Will Kindberg (p.c.) suggests that -ako in the closely related Ashaninka language may be derived from the noun root ako ‘hand’, that is, -ako is an incorporated noun which has been grammaticalized. APPLICATIVE AFFIXES 337

(30) -an/-ant ‘instrumental’ i- ken -ant -ak -a -ro abió 3m travel INST PERF REFL.NF 3f plane ‘He traveled by plane.’ (Shaver 1996:48)

(31) -ben/-bin ‘for, benefactive’ no- pën -a -ben -k -e -ri 1 pay EP BEN ASP NF 3m ‘I paid it for him.’ (Shaver 1996:48)

(32) -té ‘towards, against’ i- hok -a -té -t -abé -k -a -ri 3m throw EP TOWARDS EP FRUS ASP REFL.NF 3m ‘He threw it at him but missed/it didn’t hurt him.’ (Shaver 1996:48)

(33) -ak/-akag ‘comitative/sociative causative’ i- komo -t -ak -ak -e -ri Pablo otsegoha 3m dam.stream EP COM PERF NF 3m Paul stream ‘He dammed the stream with Paul/he caused Paul to dam the stream working along with him.’ (Wise 1971:107)

Passive constructions are rare in Nomatsiguenga and are limited to third person. Ex- ample (34) uses the suffix -imo/-omo—illustrated for Pajonal Ashéninka in (25)—in a passive construction.

(34) Pablo ir- ia -t -omo -t -I -nani Pablo 3m go EP PRES EP ASP PASS ‘Someone went to where Paul was’ (lit. Paul was gone in the presence of)

Nomatsiguenga is basically a nominative–accusative language although in stative intransitive verbs the pronominal form cross-referencing the subject is a suffix (ob- ject position), rather than a prefix. Compare the position of the first person marker in the active (35a) and stative (35b) examples.

(35) a no- pok -e 1 come NF ‘I arrive.’

b pok -ak -i -na Come PERF F 1 ‘I arrive (said upon arrival).’ (Shaver 1996:121)

338 WISE

The basic word order in Nomatsiguenga is VSO. Apart from pronominal forms cross- referencing the subject, one of the causatives, and the future/irrealis prefix; all verbal morphology is by suffixation. Nouns and classifiers may be incorporated in the verb.

3. Summary of data and discussion 3.1 Summary of data The chart summarizes the data for six of the eight languages mentioned in section 1. Notice that all six have applicative suffixes rather than prefixes, although applicative prefixes seem to be more common than suffixes in other parts of the world. (See sec- tion 2.2.) Most verb morphology is by suffixation in all of the languages, irrespective of basic word order. Although not indicated in the chart, all of the languages are head marking; as indicated, all have morphological causatives. Comparison of the lan- guages discussed here with languages which do not have applicatives, but do have morphological causatives, e.g., Bora (see Thiesen 1996), leads to the tentative hy- pothesis that the presence of an applicative affix implies the presence of a causative affix, but not vice versa. Clear correlations with other grammatical features such as nominative–accusative, ergative–absolutive, or presence of cross-referencing affixes are not apparent. Incorporated nouns in two of the languages are prefixed and suffixed in two others. Two of the languages do not incorporate nouns. In two of the languages the origin of at least one of the applicatives can be shown to be a postposition; further investigation is needed in this area. In all of the languages applicative suffixes increase valence most of the time but in those where one of the applicatives has many functions (Chayahuita, Arabela, Ya- nesha, and Nomatsiguenga), some of the functions do not increase valence. Only in Yagua is the object demoted in an applicative construction. (See example 9 in which the direct object is implicit and the instrument is promoted to the object position.)

APPLICATIVE AFFIXES 339

Chayahuita Arabela Yagua Yaminahua Yanesha’ Nomatsiguenga Applicative prefix ------Applicative suffix + + + + + + Causative prefix + - - - + + Causative suffix - + + + + + Has only suffixes - + + - - - SOV order + + - + - - VSO order - - + - + + Noun incorporation by prefix + - - + - - Noun incorporation by suffix - - - - + + Nom–acc + + + - + + Some split-erg traits or split-S marking - ? - + + + S/O cross-ref affixes + - + - + + Applicative cognate with a postposition - + + - - - Applicative usually increases valence + + + + + + Applicative has many functions + + - - + + Case markers on core NPs + - - + - - Object demoted - - + - - -

3.2. Contrasts with other languages In many languages of the world, applicatives are prefixes. Craig and Hale (1988), for example, describe “relational preverbs” in Warlpiri (of Central Australia), Rama (Chibchan of Nicaragua), and Nadëb (Maku of Brazil). O’Herin (1995) describes a set of five applicative prefixes in Abaza (a Caucasian language of Russia). Abaza has SOV order and is an ergative–absolutive language. Several applicative prefixes may 340 WISE co-occur but, in contrast to the languages of Peruvian Amazonia, neither they nor the causative prefix increase transitivity. There is always one, and only one, cross- referencing agreement prefix from the absolutive set in the verb and only one abso- lutive noun phrase. (There are a few “inverted verbs” in which no absolutive occurs.) The agreement prefixes, which cross-reference the “applied objects,” are from the ergative set and the noun phrases which they cross-reference are likewise in the erga- tive case. Abaza shares with Nomatsiguenga and the other Pre-Andine languages the trait that one verb can have several applicative markers. O’Herin (1995) suggests that the noun phrase must be definite in order for the postposition to be incorporated as an applicative. This is probably true also in those Amazonian languages where alternative constructions are available, e.g. an applica- tive affix or oblique case-marking on the noun phrase, as in Yagua (example 9). The correlations between definiteness and use of an applicative affix in a given construc- tion need to be investigated. Note that David Payne (2000) states that in Ashéninka the occurrence of a pronominal suffix cross-referencing an overt syntactic object “in- dicates a difference between highly referential, highly topical objects [36a] versus non-referential, non-topical objects [36b].” (Ashéninka is closely related to Nomat- siguenga; the applicative suffixes are cognate with those of Nomatsiguenga.)

(36) a r- etsiya -t -ako -t -aka -a -ye -t -ak -e -ri eentsi 3M be.well EP APPL EP CAUS EP DIST EP PERF MODE 3M child ‘He healed (caused to be well) the children (the ones being referred to in the prior dialogue).’

b osheki mantsiyari many sick r- etsiya -t -ako -t -aka -a -ye -t -ak -e 3M be.well EP APPL EP CAUS EP DIST EP PERF MODE ‘He healed many sick people.’ (David Payne 2000)

Perhaps the same is true in the choice of an applicative affix rather than a postposi- tional phrase where that alternative is available. It is clear from the examples in (36), however, that the use of the applicative -ako does not depend on definiteness or ref- erentiality. Pragmatic functions, such as coding thematically peripheral participants as pragmatically salient arguments in the flow of discourse, is also a topic for further study (see Zavala 2000).

3.3. Applicatives as adposition incorporation and relation to causatives Baker (1988:229) proposes that preposition [read adposition] incorporation is the source of the grammatical function changing processes called “applicative” and “da- tive shift.” Some of the applicatives discussed in this paper are clearly cases of post- position incorporation, e.g., Arabela -ta/-tia and Yagua -ta/-tya. Others are clearly APPLICATIVE AFFIXES 341 cognate with postpositions in other languages of the same family. For example, the ‘benefactive/malefactive’ -n/-on in Yanesha’ has a postposition counterpart -(V)na in Piro, -ni in , etc. The origin of most of the applicatives, however, is not yet clear. The similarity between causative and applicative constructions (indicated by Baker 1988:233) shows up in Pre-Andine Maipuran Arawakan languages even in the origin of some of the applicatives. In most of those languages there is a causative prefix im-, which appears to be cognate with the applicative suffix -imo ‘in the pres- ence of’. In Wise (1990) I showed that the Pre-Andine inflectional suffix -akag/-aka ‘causative/comitative’ is clearly cognate with a reciprocal verbal suffix in the broader Maipuran Arawakan family. David Payne (2000) posits a scenario in which the “original reciprocal sense developed into a broader sociative sense (which it still re- tains with verbs of physical activity in Pre-Andine languages), and from there to a causative sense.”8

3.4. Applicative suffixes as an Is the similarity in form of the applicative suffixes -të/-ta in Chayahuita, -ta/-tia in Arabela, -ta/-tya in Yagua happenstance, indicative of a possible genetic relationship, or an areal feature? Doris Payne (1984 and 1985), on the basis of the similarity in the Yagua and Arabela applicatives and the fact that both have postpositions of the same form, and on the basis of other apparently cognate morphemes, suggested a possible Yagua-Zaparoan connection, but left the exact nature of the connection as a topic requiring further study. The fact that the causative in Huitoto is -ta, and many - wakan languages have a causative suffix of the form -da/-ta and/or an epenthetic, verbalizing, or transitivizing suffix which includes a -t, leads me to propose (as I did in 1993) that we are dealing with a wide-spread grammatical form (see David Payne 1990) or an areal feature rather than a genetic connection. If it is an areal feature, what is the focal area? What other areal features should be added to those listed by Dixon and Aikhenvald (1999:8-9)? Should this be listed as a western Amazon feature only (cf. Dixon and Aikhenvald 1999:10)? Or should it be considered a northwestern Amazon feature since Panoan languages do not share it? Perhaps it is one indicator that the hypothesis is correct which assumes that Panoan languages originated in the southern Amazon tributaries (Bolivia?) and migrated only as far north as the Amazon itself while the Arawakan languages originated around Manaus (or up the Rio Negro) and dispersed in all directions from there. I leave these questions open pending fur- ther research.

8 T. Payne (1997:190-191) notes that in some languages “the causative and [instrumental] applicative are the same morpheme…The only real difference…is the animacy of the ‘causee.’ In both cases a causer acts on something or someone to accomplish some action.” Tuggy (1981:449) describes a variety of Nahuatl as “a halfway or hybrid case in which non-prototypical causatives exhibit a trait associated with applicatives,” i.e., they are overlapping categories. Valenzuela (2000) asserts that in Shipibo (closely related to Yamina- hua) “the associative applicative construction also encodes ‘sociative causation’.” 342 WISE

References Aikhenvald, Alexandra Y. 1999 ‘The family’, in: Dixon and Aikhenvald (eds.), pp. 65-106. Baker, Mark C. 1988 Incorporation: A theory of grammatical function changing, Chicago: The University of Chicago Press. Carochi, Horacio 1645 Arte de la lengua mexicana, Reprinted 1892, Mexico: Museo Nacional. Craig, Colette and Ken Hale 1988 ‘Relational preverbs in some languages of the Americas: Typological and his- torical perspectives’, Language 64:312-344. Dixon, R. M. W. and Alexandra Y. Aikhenvald 1999 ‘Introduction’, in: Dixon and Aikhenvald (eds.), pp. 1-21. Dixon, R. M. W. and Alexandra Y. Aikhenvald, eds. 1999 The Amazonian languages, Cambridge Language Surveys, Cambridge: Cam- bridge University Press. Duff-Tripp, Martha 1997 Gramática del idioma yanesha (amuesha), Serie Lingüística Peruana 43, Lima, Peru: Instituto Lingüístico de Verano. Eastman, Robert and Elizabeth Eastman 1961 (ms.) Iquito grammar outline, Información de Campo 108 (microfiche), Lima, Peru: Instituto Lingüístico de Verano. Faust, Norma and Eugene E. Loos fc Gramática del idioma yaminahua, Serie Lingüística Peruana 51, Lima, Peru: Instituto Lingüístico de Verano. Hart, Helen 1988 Diccionario chayahuita—castellano, Serie Lingüística Peruana 29, Pucallpa, Peru: Instituto Lingüístico de Verano. O’Herin, Brian fc Case and agreement in Abaza, SIL International and the University of Texas at Arlington Publications in Linguistics 138. Payne, David 1990 ‘Some widespread grammatical forms in South American languages’, in: Doris L. Payne (ed.), pp. 75-87. 2000 ‘Causatives in Asheninka: The case for a sociative source,’ paper presented at the Eighth Biennial Rice University Symposium on Linguistics: Causation and Interpersonal Manipulation in Languages of Central and South America, Houston, April 2000. Forthcoming in the Proceedings of the Symposium. Payne, Doris L. 1984 ‘Evidence for a Yaguan-Zaparoan connection’, Work Papers of the SIL Uni- versity of North Dakota 28:131-156. APPLICATIVE AFFIXES 343

Payne, Doris L. 1985 ‘-ta in Zaparoan and Peba-Yaguan’, International Journal of American Lin- guistics 51:529-531. Payne, Doris L., ed. 1990 Amazonian linguistics: Studies in lowland South American languages, Austin: University Press of Texas. Payne, Thomas 1997 Describing morphosyntax: A guide for field linguists, Cambridge: Cambridge University Press. Rich, Rolland G. 1999 Diccionario arabela—castellano, Serie Lingüística Peruana 49, Lima, Peru: Instituto Lingüístico de Verano. Shaler, Dorothy 1971 (ms.) Clause types and participant roles in Pajonal Campa, Información de Campo 56 (microfiche), Lima, Peru: Instituto Lingüístico de Verano. Shaver, Harold 1996 Diccionario nomatsiguenga—castellano, castellano—nomatsiguenga, Serie Lingüística Peruana 41, Pucallpa, Peru: Instituto Lingüístico de Verano. Thiesen, Wesley 1996 Gramática del idioma bora, Serie Lingüística Peruana 38, Pucallpa, Peru: Instituto Lingüístico de Verano. Tuggy, David 1981 The transitivity-related morphology of Tetelcingo Nahuatl: An exploration in space grammar, Ph.D. dissertation, University of California at San Diego. Valenzuela, Pilar 2000 ‘Applicativization and its relationship to switch-reference in Shipibo-Konibo’, paper presented at the SSILA-American Anthropological Association, Annual Meeting, San Francisco. Wise, Mary Ruth 1971 Identification of participants in discourse: A study of aspects of form and meaning in Nomatsiguenga, Summer Institute of Linguistics Publications in Linguistics and Related Fields 28, Norman, OK: Summer Institute of Linguis- tics of the University of Oklahoma. 1986 ‘Grammatical characteristics of PreAndine Arawakan ’, in: Handbook of Amazonian languages, Desmond Derbyshire and Geoffrey Pul- lum (eds.), Berlin: Mouton de Gruyter, pp. 566-642. 1990 ‘Valence-changing affixes in Maipuran Arawakan languages’, in: Doris L. Payne (ed.), pp. 8-116. 1993 ‘Algunas interrogantes en la clasificación de las lenguas indígenas sud- americanas’, paper presented at the 10o Congreso Internacional de la Asocia- ción de Lingüística y Filología de la América Latina, Veracruz. 344 WISE

1999 ‘Small language families and isolates in Peru’, in: Dixon and Aikhenvald (eds.), pp. 307-340. Zavala, Roberto 2000 ‘Simple, complex, and multiple applicatives in Olutec (Mixean)’, paper pre- sented at the SSILA-American Anthropological Association, Annual Meeting, San Francisco.

Abbreviations ACC = accusative MAL = malefactive ANTIPAS = antipassive NF = nonfuture APPL = applicative NOM = nominative ARR = upon arrival O/OBJ = object ASP = aspect PASS = passive BEN = benefactive POSS = possessive CAUS = causative PRES = in the presence of CL = classifier PROG = progressive COM = sociative or comitative RECIP = reciprocal COMPL = completive REFL = reflexive CONT = continuative REP/REPET = repetitive DIST = distributive RES = resolved perfective EP = epenthetic REAS = reason ERG/ERG = ergative RESP = with respect to FRUS = frustrative S = subject F/FUT = future SCL = subordinate clause marker GEN = genetive 1 = first person HAB = habitual 1R = related to first person IMPERV = imperfective 1 SG = first person singular IMPV = imperative 1 PL = first person plural INAN = inanimate 2 = second person INCL = inclusive 2 SG = second person singular INF = infinitive 3F = third person feminine INST = instrument 3M = third person masculine IRR = irrealis Ø = zero

This book is freely available for download at http://www.etnolinguistica.org/illa