Learning Indigenous Languages: Child Language Acquisition in Mesoamerica 9783110923148, 9783110195590

Citation preview

Learning Indigenous Languages: Child Language Acquisition in Mesoamerica

≥

Studies on Language Acquisition 33

Editor Peter Jordens

Mouton de Gruyter Berlin · New York

Learning Indigenous Languages: Child Language Acquisition in Mesoamerica edited by Barbara Pfeiler

Mouton de Gruyter Berlin · New York

Mouton de Gruyter (formerly Mouton, The Hague) is a Division of Walter de Gruyter GmbH & Co. KG, Berlin.

앝 Printed on acid-free paper which falls within the guidelines 앪 of the ANSI to ensure permanence and durability.

Library of Congress Cataloging-in-Publication Data Learning indigenous languages : child language acquisition in Mesoamerica / edited by Barbara Pfeiler. p. cm. ⫺ (Studies on language acquisition ; 33) Includes bibliographical references and index. ISBN 978-3-11-019559-0 (cloth : alk. paper) 1. Language acquisition. 2. Indians of Central America ⫺ Languages. I. Pfeiler, Barbara, 1952⫺ P118.L38986 2007 4011.93⫺dc22 2007024455

Bibliographic information published by the Deutsche Nationalbibliothek The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data are available in the Internet at http://dnb.d-nb.de.

ISBN 978-3-11-019559-0 ISSN 1861-4248 쑔 Copyright 2007 by Walter de Gruyter GmbH & Co. KG, D-10785 Berlin. All rights reserved, including those of translation into foreign languages. No part of this book may be reproduced in any form or by any means, electronic or mechanical, including photocopy, recording, or any information storage and retrieval system, without permission in writing from the publisher. Cover design: Sigurd Wendland, Berlin. Printed in Germany.

Contents

Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii Introduction: The view from Mesoamerica . . . . . . . . . . . . . . . . . . . . . . . . Barbara Pfeiler

1

Crosslinguistic study Roots or Edges? Explaining variation in children’s early verb forms across five Mayan languages . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15 Clifton Pye, Barbara Pfeiler, Lourdes de León, Penelope Brown, Pedro Mateo Acquisition of ergative Mayan languages Explaining Ergativity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47 Clifton Pye Early acquisition of the Split Intransitive System in Yukatek . . . . . . . . . . 69 Carlos Carrillo Carreón Acquisition of the early lexicon A preliminary view at Ch’ol (Mayan) early lexicon: The role of language and cultural context . . . . . . . . . . . . . . . . . . . . . . . . . . . 85 Lourdes de León Acquisition of referential and relational words in Huichol: from 16 to 24 months of age . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103 Paula Gómez López Semantic development Culture-specific influences on semantic development: Learning the Tzeltal ‘benefactive’ construction . . . . . . . . . . . . . . . . . . . . . 119 Penelope Brown Bcuaa quiang – I stepped HEAD it! The acquisition of Zapotec bodypart locatives . . . . . . . . . . . . . . . . . . . . . . 155 Kristine Jensen de López

vi Contents Language Socialization “Lo oye, lo repite y lo piensa.” The contribution of prompting to the socialization and language acquisition in Yukatek Maya toddlers . . . . . 183 Barbara Pfeiler List of contributors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203 Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205

Acknowledgements

I wish to thank the following collegues for their help in thoroughly reviewing the chapters of this book: Eve Clark, Soonja Choi, Suzanne Gaskins, Jürgen Bohnemeyer, Enrique Palancar, Clifton Pye. This publication is financially supported by the National Council for Science and Technology as part of the project CONACYT # 27893-H to Barbara Blaha Pfeiler. My warmest thanks also go to Ursula Kleinhenz and Frank Benno Junghanns for their patient and meticulous editorial and formatting work throughout the process of producing this book.

Introduction: The view from Mesoamerica Barbara Pfeiler

The large number of vernacular languages in Mesoamerica1 and the fact that most are threatened with extinction makes the study of language acquisition in them urgent. Language acquisition data for non-Indo-European languages constitutes a valuable contribution to linguistics, and Mesoamerica is a rich analytical field. Interest in Mesoamerican languages is rooted in their formal and typological linguistic differences. They are also promising for exploration of the acquisition process in sociocultural contexts other than the Western middle classes to which most children studied to date have belonged. Language acquisition research in Mesoamerican languages is a valuable contribution to linguistics, but can also advance other disciplines such as ethnolinguistics and social anthropology. Acquisition research in indigenous languages, in this case in Mesoamerica, is a challenge for psycholinguistic analysis since some of these languages still lack linguistic descriptions and even fewer studies have been done of their many dialects. Their distinct sociocultural context also poses a problem because, in addition to a deep understanding of the language and its grammatical rules, a large quantity of ethnographic data is needed to allow an understanding of the non-Western discursive practices that occur during early socialization. To date, the acquisition process has been studied in very few Mesoamerican languages. Five languages in the Maya family are the most thoroughly studied: K’iche’ (Pye 1983, 1986, 1990, 1992; Pye & Rekart 1990) and K’anjob’al (Mateo 2005), both spoken in Guatemala; and Tzeltal (Brown 1997, 1998a, 1998b, 1998c, 2001), Tzotzil (de León Pasquel 1994, 1997, 1998a, 1998b, 1999a, 1999b, 2001a, 2001b, 2005; Rojas Nieto & de León Pasquel 2001) and Yukatek (Pfeiler 1998, 2002, 2003, 2006; Pfeiler & Martín Briceño 1997, 1998; Blaha Pfeiler & Carrillo Carreón 2001), spoken in southeast Mexico. Paula Gómez López (1998), José Luis Iturrioz L. (1998), Iturrioz & Gómez López (1998) have studied acquisition in the Huichol language, spoken in central Mexico, and Jensen de López (2002a, 2002b) has studied acquisition in the indigenous Zapotec language in the

2 Barbara Pfeiler state of Oaxaca. Although the theory underlying these studies varies, all have produced interesting data on the development of language morphology, semantics and socialization that have begun to significantly affect language acquisition theory and linguistics in general. These studies are based on methodologies of longitudinal and some transversal studies, as well as elicitation tests. The present edited collection is intended to offer an overall panorama of acquisition research in Amerindian languages and the methodological, theoretical and empirical perspectives it encompasses. The collection includes a comparative study of five Mayan languages; a theoretical discussion of ergativity; an empirical study on mixed ergativity; and five individual theoretical-empirical studies. Each of the latter has a specific subject, such as name and verb acquisition, spatial dimension acquisition or language socialization. “Roots or Edges? Explaining variation in children’s early verb forms across five Mayan languages” is a novel methodological approach applied by Clifton Pye, Barbara Pfeiler, Lourdes de León, Penelope Brown and Pedro Mateo. In it, five related languages are compared using the same systematization and analysis procedures. The crosslinguistic study of language acquisition is a critical tool that is needed in order to test acquisition theories in other languages. However, crosslinguistic studies are a relativiely crude instrument of cognitive science in the absence of a method for establishing equivalence between linguistic categories in different languages. The comparative method is used to establish equivalence between linguistic categories in the five Mayan languages. Historical linguistics relies on the comparative method to establish genetic relationships between languages, and a significant result of the comparative method is fundamental understanding of category equivalence in related languages. Reconstruction of linguistic divergence establishes crosslinguistic relationships between categories at all linguistic levels and leads to better control of crosslinguistic variation. This contribution to language acquisition illustrates application of the comparative method to the study of early verb forms in children in these languages. The basic research question about Mayan language acquisition in this study is, what motivated children to produce bare verb roots? The verb can appear at the extreme right of the verb if no suffixes are present, or at the extreme left if there are no prefixes. The goal was to examine the degree to which common factors in the input in the five languages influence children to make them produce bare verb forms. Three input factors were analyzed that could affect production of bare verb roots in children:

Introduction: The view from Mesoamerica

3

the frequency with which the verb root occurs at the verb’s extreme left; the number of times the root occurs at the beginning of a sentence; and the number of positive imperatives in the input. None of these factors was found to be related to production of bare roots in children, indicating that children ignore the extreme left of verbs and sentences when extracting verb roots. An additional five input factors were analyzed that could focus children on the extreme right of verbs and sentences: the frequency with which the verb roots are produced at the extreme right; the frequency with which they occur at the end of sentences; the frequency of derivational suffixes; the frequency of status suffixes; and the frequency of CVC roots. The frequencies at which verb roots occur at the extreme right of the verb and sentence was found to positively correlate with production of bare verb roots by children, whereas the frequency of derivational and status suffixes in the input was negatively correlated with the children’s production of bare verb roots. The frequency with which adults produce verbal roots at the extreme right of words and sentences influences the frequency with which children produce bare verb roots in their expressions, while production of verb roots at the extreme left do not. In addition, the position the verb occupies at the beginning of a sentence affects neither the ability of children to extract verb roots nor the degree to which the verb bears inflectional prefixes. Most of the languages addressed in this collection are ergative, a language type that has long proved enigmatic for linguistic theory. No apparent justification has yet been forwarded for the distinction between the grammatical relationships between agent and subject that occur in ergative languages. Research on ergativity has exposed the limitations of the basic theoretical assumption that syntax and semantics can be isolated in separate grammatical modules. The basic grammatical category of subject has been questioned within this context. Many definitions of the subject category have been designed based on European languages, which are largely nominative-accusative. In terms of semantic roles, ergative marking represents a different dimension of conceptual and linguistic organization (Langacker, 2000: 35–36). Because of this, ergative and mixed ergative languages pose interesting challenges in language acquisition research. Acquisition of ergative languages proves problematic for theories such as Parameter setting (Chomsky 1981; Roeper & Williams 1987) and Semantic bootstrapping (Pinker 1984, 1989, 1996) due to these languages’ different marking of the subject category. Both Ochs (1988) and Pye (this volume) state that the subject category is acquired using different strategies in different language

4 Barbara Pfeiler types. In nominative-accusative languages, children learn the subject expressed equally as agent or experiencer, whereas in ergative languages children learn to grammatically distinguish between agent and experiencer. Only a few, recent pioneering studies on acquisition in ergative languages have been done to date. These were done to explore language acquisition in grammatically different languages and thus fill the void left by the recursivity of language acquisition theory design based on nominativeaccusative languages such as English (Slobin 1992: 1–2; van Valin 1992: 15). Ergative language acquisition research has been done with the Kaluli, Samoan, Georgian, Western Greenland Eskimo, Inuktitut and Euskera languages, as well as the K’iche’, K’anjob’al, Tzeltal, Tzotzil and Yukatek Mayan languages. Nonetheless, the only studies specifically on acquisition of the ergative system are those of Samoan (Ochs 1988), comparative studies between K’iche’, Tzeltal, Tzotzil and Yukatek Maya (Pye, Brown, de León & Pfeiler 2002, 2004; Carrillo Carreón 2005), and a study of agent case marking in Hindi (Narasimhan 2005, in press; Narasimhan et al. 2005). This collection contains one theoretical study of ergativity and another empirical. In his article “Explaining Ergativity” Clifton Pye critiques the four approaches that have attempted to integrate ergative languages into structuralist syntactic theory. According to Pye, these approaches manage to acknowledge certain aspects of ergativity but not of mixed ergativity, specifically the different types of mixed ergativity in the Mayan languages. These range from total ergativity (K’iche’), to mixed ergativity conditioned by person (Mochó), aspect (Yukatek subgroup), clause type (Kanjobalano subgroup) or focus (Mamean subgroup). He argues that a theory that encompasses different features of mixed ergative languages must consider the interaction between concordance and discourse (e.g. Dixon 1994; Du Bois 1990). Carlos Carrillo’s empirical study “Early Acquisition of the Split Intransitive System in Yukatek Maya” shows early development of the aspectconditioned split intransitive system in Yukatek Maya based on a longitudinal study of children between the ages of 2;0 and 3;0. He demonstrates that the frequency and morphological transparency of the person marker produced by children has a role that leads them towards acquisition of split intransitivity. The person marker/verb class/aspect relationship that characterizes Yukatek Maya apparently offers clues for early use of the pronominal system. Division of the subject of intransitive verbs was observed from the outset, but, even at 3;0, there was no clear evidence of complete acquisition of the split intransitive system or of the total person markers series.

Introduction: The view from Mesoamerica

5

A large part of the overall language acquisition discussion has been taken up by early lexicon. Nouns have been said to be a predominant category in early development, regardless of the language (Gentner 1982; Markman 1989). Research on non-European languages, however, has shown that linguistic structure and socialization type play a central role in the preference for verbs over nouns (Tardif 1996; Choi 1998). Studies of acquisition in non-European languages (e.g. Mandarin, Tzotzil and Tzeltal, among others) highlight the influence of linguistic and cultural factors on the preference for, and early morphological development of, verbs over nouns in children’s early vocabulary. There are two studies of Mesoamerican languages in this collection that demonstrate the preference for verbs over nouns at young ages. In “A preliminary view at Ch’ol (Mayan) early lexicon: The role of language and cultural context”, Lourdes de León analyzes the role played by frequency and type of adult sentences in noun and verb acquisition. The results show that verbs are more frequent than nouns in children’s speech, and that this is also the case in speech directed toward children. De León argues that this is due to the structure of the Ch’ol language, which allows a high elision of nominal arguments (the result of its cross-referenced pronoun system) that makes the verb prominent in sentences. The type of interaction between caregivers and the informant also show the high incidence of directives formed as imperatives. The data for Ch’ol are comparable to those for Tzotzil (de León 1999a, 1999b, 2001b). In “Acquisition of referential and relational words in Huichol: from 16 to 24 months of age”, Paula Gómez thoroughly discusses the structural differences in the Huichol language that allow acquisition of verbs before nouns. She uses data from a girl between 16 and 24 months of age (Gómez López 1998) and the criteria considered for nouns and verbs in Huichol. This language has a series of properties that create conditions that allow the use of predicative constructions in general, and verbs in particular. These can be more frequent than the use of nouns and referential constructions. The criteria of morphosyntactic context, and adult situation and interpretation were established for words and roots with referential and relational functions. In summary, the two studies mentioned above highlight the importance of structural and discursive factors in favouring verbs over nouns. These factors converge with cultural language learning contexts in which activities and references to actions are contained in the foundation of the language of socialization.

6 Barbara Pfeiler Other studies in this collection emphasize semantic properties as a determinant factor in early acquisition. According to Bowerman & Choi (2001), certain types of relational words with semantic properties within a specific language, such as verbs and adpositions, can appear in early acquisition. In “Culture-specific influences on semantic development: Learning the Tzeltal ‘benefactive’ construction”, Penelope Brown states that little evidence exists for the sensitivity of children to constructional meanings in specific language. Her research on the “benefactive” in Tzeltal shows that considerable cultural support exists in caregiver speech and behaviour for learning the construction of this grammatical category in this language. She analyzes how three Tzeltal-speaking children around the ages of 2;0, 2;6 and 3;0 acquire the “benefactive”. This is a ditransitive construction with three arguments and is the basic way of expressing three arguments in Tzeltal verbs. The “benefactive” suffix is standard for transference verbs and certain speech verbs. Tzeltal children are shown to acquire the “benefactive” construction quite early. They begin to use it in the two-word stage and make productive use of it by the age of 2;6, long before other arguments like agent and possessor. “Benefactive” suffixes are used as frozen forms in the one- and two-word stages but quickly appear with person contrasts and in different verbs. Kristine Jensen’s study “Bcuaa quiang – I stepped HEAD orange peel! The acquisition of Zapotec body part locatives” addresses the acquisition of body part nouns in spatial use. The data are for a Zapotec-speaking boy between 15 and 33 months of age, and show that four body part nouns were used as spatial locatives: láani ‘stomach’; quia ‘head’; lo ‘face’; and dets ‘back’. In particular, lo and quia were used in different syntactic constructions and with polysemic meanings. Nouns making an actual reference to body parts were not exactly the same as those used as locatives: nii ‘foot’; lo ‘face’; naa ‘hand’; and dets ‘back’. Compared to acquisition in languages with prepositions, body part terms in Zapotec emerge later and are not used in reference to the child’s own body. Jensen argues, however, that Zapotec-speaking children progress more rapidly in their domination of the spatial relations system than children learning prepositions. There are clear differences in acquisition of the Mesoamerican languages addressed in this collection compared to European languages. These differences are not limited to morphological or semantic acquisition, but extend to the area of discursive practices. Use of a sociocultural approach in language acquisition research on non-European languages has revealed that discursive practices during early socialization vary from those reported

Introduction: The view from Mesoamerica

7

for Western cultures. Mesoamerican languages include more frequent use of dialogic conversation between adults and small children, and caregiver directed speech includes more directives and imperatives. One discursive practice with a socialization function is prompting, in which adults formulate what the child must say using a “say…” framework. This has been documented in languages as varied as Kaluli (Schieffelin 1979), Chicano (Eisenberg 1982), Afro-American English (Ward 1971), K’iche’ (Pye 1986) and Tzeltal (Brown 1998c). Some of these languages exhibit specific linguistic forms for this type of prompting. Other documented strategies include repetition and expansion. Classic studies of acquisition in European languages have treated the use of simplified language (i.e. “baby talk”) with children as a way of teaching the language, as well as a manifestation of socialization of children in nuclear families, where dyadic interaction between child and mother (or caregiver) predominates. Children in Western nuclear family contexts have few opportunities to interact with other adults or children of different ages. Anthropological research in cultures such as Samoan, Basotho, Inuit and Tzotzil has shown that child development in these cultures occurs in extended family homes. These children grow up in close and constant contact with more than two siblings, and with relatives of different ages, including babies, adolescents and the elderly, which produces interaction with multiple parties. The final article in this collection, “‘Lo oye, lo repite y lo piensa.’ The contribution of prompting to the socialization and language acquisition of Yukatek Maya toddlers in Yucatan, Mexico” by Barbara Pfeiler, shows the contribution of prompting to socialization and language acquisition in Yukatek Maya toddlers. Pfeiler analyzes parents’ beliefs about acquisition of the language as the mother tongue and determines if these beliefs coincide with spontaneous interactions between adults and children. In an analysis of three longitudinal studies Pfeiler shows the frequency and types of prompts, and analyzes the socialization objectives of each: a’l-ti’ ‘say to him’; t’aan-eh, ‘call him’; and the dyadic and triadic quotatives kech ‘say this’ and kech-ti’ ‘say it like this to him’. The kech-ti’ form in particular is a frequent practice with a number of functions, such as instruction of proper behaviour, lexical teaching and grammatical correction. Another aspect that contributes to acquisition are the evidential inputs ki ‘s.o. says’ and bin ‘it is said’, which serve to reformulate children’s confused messages. Pfeiler also gives examples of grammatical corrections that do not include specific linguistic forms, but are less frequent than prompts. Overall, this article demonstrates that prompting routines provide small children with a model of speech as such,

8 Barbara Pfeiler but in most cases this has a more social purpose rather than an explicit function of language instruction. Altogether, the studies in this book contribute a substantial addition to research on the acquisition of non-European languages and provide vital new insights into the acquisition process. These studies demonstrate that much more information remains to be recorded about the acquisition of the world’s languages before we will be in a position to offer firm accounts of the resources children employ to learn their first languages. While investigators have studied the acquisition of several Mayan languages, acquisition data from the majority languages in this family remain unrecorded. It is disheartening to think that acquisition data has yet to be recorded from the majority of the other language families in Mesoamerica. This work must be accomplished before the children of Mesoamerica cease to acquire these languages. Note 1. The Mesoamerican languages are spoken in an area covering central and southern Mexico, Guatemala, Belize, El Salvador and portions of Honduras and Nicaragua.

References Blaha Pfeiler, Barbara & Carlos Carrillo Carreón 2001 La adquisición del maya yucateco: el número. In La adquisición de la lengua materna. Español, lenguas mayas, euskera, Cecilia Rojas Nieto & Lourdes de León (eds.), 75–97. México: Centro de Investigaciones y Estudios Superiores en Antropología Social/Universidad Nacional Autónoma de México. Bowerman, Melissa & Soonja Choi 2001 Shaping meanings for language: universal and language-specific in the acquisition of spatial semantic categories. In Language acquisition and conceptual development, Melissa Bowerman & Stephen C. Levinson (eds.), 475–511. Cambridge: Cambridge University Press. Brown, Penelope 1997 Isolating the CVC root in Tzeltal Mayan: A study of children’s first verbs. In The Proceedings of the 28th Annual Child Language Research Forum, Eve V. Clark (ed.), 41–52. Stanford, CA: CSLI / University of Chicago Press.

Introduction: The view from Mesoamerica 1998a

9

Early Tzeltal verbs: Argument structure and argument representation. In Proceedings of the 29th Annual Child Language Research Forum, Eve V. Clark (ed.), 129–140. Stanford, CA: CSLI/ University of Chicago Press. 1998b Children’s first verbs in Tzeltal: Evidence for an early verb category. In Special edition of Linguistics 36 (4), Elena Lieven (ed.), 713–53. 1998c Conversational structure and language acquisition: The role of repetition in Tzeltal adult and child speech. Journal of Linguistic Anthropology 8(2): 197–221. 2001 Learning to talk about motion UP and DOWN in Tzeltal: Is there a language-specific bias for verb learning? In Language acquisition and conceptual development, Melissa Bowerman & Stephen C. Levinson (eds.), 512–543. Cambridge: Cambridge University Press. Brown, Penelope, Lourdes de León, Barbara Pfeiler & Clifton Pye 2002 The acquisition of agreement in Maya. Symposium presented at 9th IASCL Conference, Madison, WI, 16–21 July, 2002. 2004 Mayan language acquisition studies. Symposium presented at the 78th Annual Meeting /Linguistic Society of America /Society for the Study of the Indigenous Languages of the Americas; Boston, January 8 –11, 2004. Carrillo Carreón, Carlos 2005 La adquisición de los pronombres de referencia cruzada en el maya yucateco. Estudio de caso. Masters Thesis, Universidad Autónoma de Yucatán. Choi, Soonja 1998 Verbs in early lexical and syntactic development in Korean. Linguistics 36: 755–780. Chomsky, Noam 1981 Lectures on Government and Binding: The Pisa Lectures. Berlin / NewYork: Mouton de Gruyter. de León, Lourdes 1994 Exploration in the acquisition of geocentric location by Tzotzil children. Linguistics 32 (4/5): 857–885. 1997 Vertical path in Tzotzil (Mayan) acquisition: Cognitive vs. linguistic determinants. In The Proceedings of the 28th Child Language Research Forum, Eve Clark (ed.), 183–197. Stanford, CA: CSLI /University of Chicago Press. 1998a The emergent participant: Interactive patterns of socialization of Tzotzil Mayan infants. Journal of Linguistic Anthropology 8 (2): 1– 31. 1998b Raíces verbales tempranas en tzotzil: Input materno vs. restricciones cognoscitivas. Función 18: 147–174.

10 Barbara Pfeiler 1999a

Verb roots and caregiver speech in early Tzotzil (Mayan) acquisition. In Cognition, discourse and function, Barbara Fox, Dan Juravsky & Laura Michaelis (eds.), 99–119. Stanford, CA: CSLI /University of Chicago Press. 1999b Verbs in Tzotzil early syntactic development. International Journal of Bilingualism 3 (2/3): 219–240. 2001a Why Tzotzil children prefer verbs over nouns: the role of linguistic, cognitive, and cultural factors. In Selected papers of the Proceedings of the VIII International Congress for the Study of Child Language, Margareta Almgren, Andoni Barreña, María-José Ezeizabarrena, Itziar Idiazabal & Brian Mac Whinney (eds.), 947–969. Cascadilla Press. CD Rom Edition. 2001b Finding the richest path: the role of language and cognition in the acquisition of verticality in Tzotzil (Mayan). In Language acquisition and conceptual development, Melissa Bowerman & Stephen Levinson (eds.), 564–565. Cambridge: Cambridge University Press. 2005 La llegada del alma: lenguaje, infancia y socialización entre los mayas de Zinacantán. México: Centro de Investigaciones y Estudios Superiores en Antropología Social: Instituto Nacional de Antropología e Historia. Dixon, Robert M. W. 1994 Ergativity. Cambridge Studies in Linguistics 69. Cambridge: Cambridge University Press. Du Bois, John W. 1990 El cero absolutivo: adaptación paradigmática en Sakapulteko. In Lecturas sobre la lingüística maya, Nora C. England & Stephen R. Elliot (eds.), 473–485. La Antigua, Guatemala: Centro de Investigaciones Regionales de Mesoamérica. Eisenberg, Ann R. 1982 Language acquisition in cultural perspective: talk in three Mexicano homes. PhD dissertation, University of California. Gentner, Dedre 1982 Why Nouns are learned before Verbs: Linguistic Relativity versus Natural Partitioning. In Language Development, Vol. 2, Stan Kuczaj (ed.), 301–334. (Language, Thought and Culture.) Hillsdale, New Jersey. Gómez López, Paula 1998 Factores perceptuales y semánticos en la adquisición de la morfología en huichol. Función 18: 175–204. Iturrioz Leza, José Luis 1998 Acoplamiento estructural y adquisición del huichol como lengua materna. Función 17: 1–140.

Introduction: The view from Mesoamerica

11

Iturrioz Leza, José Luis & Paula Gómez López (eds.) 1998 Estudios sobre adquisición de algunas lenguas de México, Función 17–18. Jensen de López, Kristine 2002a Baskets and Body-Parts: A cross-cultural and crosslinguistic investigation of children’s development of spatial cognition and language. PhD dissertation. Psykologisk Institut, Aarhus Universitet, Denmark. 2002b Language-specific patterns in Danish and Zapotec children’s comprehension of spatial grams. In The 31st Stanford Child Language Forum, Eve Clark (ed.), 50–69. Stanford, CA: CSLI /University of Chicago Press. Langacker, Ronald W. 2000 Grammar and Conceptualization. Cognitive Linguistics Research 14. Berlin /NewYork: Mouton de Gruyter. Markman, Ellen M. 1989 Categorization and Naming in Children: Problems of Induction. Cambridge, MA: MIT Press. Mateo, Pedro 2005 The Acquisition of Inflection in Q’anjob’al. MA thesis. The University of Kansas. Narasimhan, Bhuvana 2005 Splitting the notion of ‘agent’: Case-Marking in early child Hindi. Journal of Child Language 32: 787–803. in press Case-Marking and transitivity in early child Hindi. In Language and Linguistics in South Asia: Selected Papers from South Asian Language Analysis 24 (provisional title), K. K. Sridhar & S. N. Sridhar (eds.). New Delhi: Manohar Publications. Narasimhan, Bhuvana, Nancy Budwig & Lalita Murty 2005 Argument realization in Hindi caregiver-child discourse. Journal of Pragmatics 37: 461–495. Ochs, Elinor 1988 Culture and language development. Language acquisition and language socialization in a Samoan village. Studies in the Social and Cultural Foundations of Language 6. Cambridge: Cambridge University Press. Pfeiler, Barbara 1998 Acquisition of Number in Yucatec Maya. In Studies in the Acquisition of Number and Diminutive Marking, Steven Gillis (ed.), 77–95. Antwerp Papers in Linguistics Vol. 95, Universiteit Anwerpen. 2002 Noun and verb acquisition in Yucatec Maya. In Pre- and Protomorphology: Early phases of morphological development in nouns and verbs, María D. Voeikova & Wolfgang U. Dressler (eds.), 75–82. Lincom Studies in Theoretical Linguistics 29. München: Lincom.

12 Barbara Pfeiler 2003

Early acquisition of the verbal complex in Yucatec Maya. In Development of Verb Inflections on First Language Acquisition. A Cross-Linguistic Perspective, Dagmar Bittner, Wolfgang U. Dressler & Marianne Kilani-Schoch (eds.), 379–399. Berlin/New York: Mouton de Gruyter. 2006 Polyvalence in the acquisition of early lexicon in Yucatec Maya. In Lexical Categories and Root Classes in Amerindian languages, X. Lois & V. Vapnarsky (eds.), 319–341. Frankfurt: Peter Lang. Pfeiler, Barbara & Enrique Martín-Briceño 1997 Early verb inflection in Yucatec Maya. Papers and Studies in Contrastive Linguistics 33: 117–125. 1998 La adquisición de los verbos transitivos en el maya yucateco. Función 17–18: 97–120. Pinker, Steven 1984 Language learnability and language development. Cambridge, MA: Harvard University Press. 1989 Learnability and cognition. Cambridge, MA/London: The MIT Press. 1996 Language Learnability and Language Development with new commentary by the author. Cambridge, MA/London: Harvard University Press. Pye, Clifton 1983 Mayan telegraphese: intonational determinants of inflectional development in Quiché Mayan. Language 59: 583–604. 1986 Quiché Mayan speech to children. Journal of Child Language 13: 85–100. 1990 The Acquisition of Ergative Languages. Linguistics 28: 1291–1330. 1992 The Acquisition of K’iche’ (Maya). In The Crosslinguistic Study of Language Acquisition, Vol. 3, Dan I. Slobin (ed.), 221–308. Hillsdale, NJ: Erlbaum. Pye, Clifton & Debora Rekart 1990 La adquisición del K’iche’. In Lecturas sobre la Lingüística Maya, Nora C. England & Stephen R. Elliott (eds.), 115–126. La Antigua, Guatemala: Centro de Investigaciones Regionales de Mesoamérica. Roeper, Thomas & Edwin Williams (eds.) 1987 Parameter Setting. Dordrecht: Reidel. Rojas Nieto, Cecilia & Lourdes de León (eds.) 2001 La adquisición de la lengua materna. Español, lenguas mayas y euskera. México: Centro de Investigaciones y Estudios Superiores en Antropología Social/Universidad Nacional Autónoma de México. Schieffelin, Bambi 1979 How Kaluli children learn what to say, what to do, and how to feel: an ethnographic study of the development of communicative competence. Unpublished PhD dissertation, Columbia University.

Introduction: The view from Mesoamerica

13

Slobin, Dan I. 1992 Introduction. In The Crosslinguistic Study of Language Acquisition, Vol. 3, Dan I. Slobin (ed.), 1–13. Hillsdale, NJ: Erlbaum. Tardif, Twila 1996 Nouns are not always learned before verbs: Evidence from Mandarin speakers’ early vocabularies. Developmental Psychology 32: 492–504. van Valin, Jr., Robert D. 1992 An overview of ergative phenomena and their implications for language acquisition. In The Crosslinguistic Study of Language Acquisition, Vol. 3, Dan I. Slobin (ed.), 15–37. Hillsdale, NJ: Erlbaum. Ward, Martha Coonfield 1971 Them Children: A Study in Language Learning. New York: Holt, Rinehart & Winston.

Roots or Edges? Explaining variation in children’s early verb forms across five Mayan languages Clifton Pye, Barbara Pfeiler, Lourdes de León, Penelope Brown and Pedro Mateo *

Introduction Children learning K’iche’, Tzeltal, Tzotzil, Q’anjob’al and Yukatek produce different forms of their first verbs (Brown 1997, 1998; de León 1999a; Mateo 2005; Pfeiler & Martín Briceño 1998; Pye 1983). Children acquiring Tzeltal and Tzotzil initially produce a high proportion of bare verb roots, while children learning K’iche’, Q’anjob’al and Yukatek produce many more combinations of a verb root plus suffix. Brown (1997: 45) found that the first 35 verbs produced by one Tzeltal boy were all bare roots. A Tzeltal girl produced a greater variety of verb forms, but bare roots still constitute over seventy percent of her first 35 verb forms. De León (1999a, 1999b) documents similar early use of verb roots by children learning Tzotzil. In contrast, Pye (1983: 592) showed that children learning K’iche’ at a comparable age preferred verb forms containing a suffix over bare roots by better than a four to one ratio. Our basic research question in this paper is this: what motivates Mayan children’s production of bare verb roots? We use the term ‘edge’ to refer to the beginning or end of a verb. The verb root may appear at the right edge of the verb if there are no suffixes, and at the left edge of the verb if there are no prefixes (1). Our goal is to examine the degree to which common factors in the input influence children’s production of bare verb forms across a set of five Mayan languages. (1)

Roots and Edges in Mayan Verbs Prefix + [Root] + Derivation + Suffix | | | | Left Left Right Right word Root word edge edges edge

16 Clifton Pye, Barbara Pfeiler, Lourdes de León, Penelope Brown & Pedro Mateo Children’s verb forms have assumed a significant theoretical importance in recent years since the introduction of the Root Infinitive Hypothesis (Wexler 1994), which claims that children optionally use non-finite verb forms in simple clauses. Constructionist theorists, on the other hand, claim that the form of children’s verbs reflects structural features of the adult language (Tomasello 2003). Data from the Mayan languages are ideally suited to resolving this debate as the languages feature a rich system of verbal inflection with separate clitics or affixes for aspect, subject, object and transitivity. In the following section, we first introduce the structure of verbs in the adult languages and discuss the morphemes that occur on the left and right edges of Mayan verbs. We next discuss the verb forms that children produced in five Mayan languages. We then proceed to test nine different left edge and right edge structural features in the adult language that could account for the children’s production of bare verb forms. We end with a summary of our findings and a discussion of the implications the findings have for language acquisition theories. Mayan Verbal Inflection Verbs in Mayan languages have a predominantly agglutinative morphology. The canonical root form of verbs is CVC. Mayan languages are verbinitial, head marked languages with separate agreement inflections for the subject and object on transitive verbs. Mayan verbs are also inflected for aspect and mood. In addition, the verbs carry what Terrence Kaufman (1990) labels a ‘status suffix’. The status suffix differs considerably across the Mayan languages, but generally marks verb transitivity and mood. A general morphological template for the Mayan languages is shown in (2). The absolutive comes after the aspectual prefix in K’iche’ and Q’anjob’al (Absolutive1), and after the status suffix in Tzeltal and Yukatek (Absolutive2). It is found in both positions in Tzotzil. (2)

Mayan transitive verb template1 fn abbreviation conventions1 Aspect + Absolutive1 + Ergative + Verb_Stem + Status + Absolutive2 + Plural K’iche’: k-at-k-il-o INC-2ABS-3ERG.PL-see-STATV-INC

‘They see you’ Yukatek: k-uy il-ik-eč-o’ob’ INC-3ERG see-STATV-INC-2ABS-PL ‘They see you’

Roots or Edges?

17

Mayan languages use a set of ergative cross-reference prefixes to mark the subjects of transitive verbs and absolutive cross-reference affixes to mark both the subjects of intransitive verbs and the objects of transitive verbs (3). The absolutive markers are prefixes in Q’anjob’al and K’iche’, and suffixes in Yukatek and Tzeltal. Tzotzil has absolutive prefixes and suffixes. The third person absolutive marker is a zero morpheme in these five Mayan languages, which provides a context in which the verb root might occur on the right or left edge of the verb. Q’anjob’al and Yukatek extend ergative markers to the subjects of intransitive verbs in specific contexts. (3)

Ergative Agreement

Absolutive Agreement

Subject of transitive verbs Nominal possessors

Subject of intransitive verbs Object of transitive verbs

Left Edge Contexts Aspect marking differs considerably across the Mayan languages and affects the contexts in which the verb roots appear at the left edge of the verb word. These differences are evident in the aspectual paradigms shown in (4). This variation occurs in both the range of obligatory contexts for aspectual prefixes as well as the degree of fusion between the verbal complex and the verbal prefixes. K’iche’ is clearly different from the other languages in requiring an overt prefix for positive imperatives. Negative imperatives (e.g., ‘don’t hit your sister’) all have a marker for negation that precedes the verb in our languages. Tzotzil has distinct incompletive and completive aspectual prefixes for transitive and intransitive verbs as well as for person. First and third person transitive verbs in Tzotzil have the incompletive prefix ta-, while second person incompletive transitive verbs have the prefix ch-. (4)

Mayan Aspect Paradigms2

INC COM IMP SUBJ POT

Yukatek TV IV

Tzeltal TV IV

Tzotzil

k- kt- h/∅∅ ∅ káa/sáan

ya ya x- ta-(1,3) ch-(2) ch-(1,2) ta-(3) chla ∅ i-(1,3) ∅ (2) l-(1,2) i-(3) max∅ ∅ ∅ ∅ ∅ ∅ ∅ ∅ ∅ ∅ hoq

TV

IV

Q’anjob’al TV IV

K’iche’ TV IV

chkkmax- xx∅ ch-/k- ch-/k∅ ch-/k- ch-/khoq kk-

18 Clifton Pye, Barbara Pfeiler, Lourdes de León, Penelope Brown & Pedro Mateo K’iche’ adults use an imperative prefix as well as the regular agreement prefix for second person in their imperatives, both positive and negative. However, K’iche’ child directed speech contains some examples where the adults omit the prefixes in imperatives. The negative imperatives in the other languages retain the PM imperative system with agreement markers and even aspectual prefixes. Tzotzil does not mark completive aspect on transitive verbs with second person subjects. When an adverb follows the verb, Tzotzil permits the omission of the completive aspect prefix for third person intransitive verbs so the verb root occurs at the left word edge in this context. Yukatek, Tzeltal, Tzotzil and Q’anjob’al allow adverbs to occur between the aspect markers and the verb roots, hence the aspect markers have a different morphological status in these languages. The following example for Q’anjob’al shows a transitive verb with an adverb. In this dialect the third person ergative marking for consonant-initial verbs is zero. In this example, the transitive verb is a complement of an auxiliary verb, and in the third person the verb root can occur at the left verb edge. The Yukatek example shows a case where an adverb is inserted between the ergative subject marker and the verb root. In this case, the verb root does not appear at the left word edge because of the overt ergative subject marking. (5)

Adverb Particles a. Q’anjob’al max-ach wal ∅-tayne-j COM-2ABS INTENS 3ERG-look_after-STATV-DER ‘He/she really looked_after you.’ b. Yukatek k-in hach il-ik INC-1ERG INTENS see-STATV-INC ‘I am focussing on him/her/it.’

Q’anjob’al allows intransitive complements of the progressive to occur with the verb root at the left word edge. Table 1 lists the contexts for our languages in which the verb root can occur at the left word edge. Note that we distinguish between transitive and intransitive verbs since the contexts where the verb root can occur at the left word edge differ by the transitivity of the verb.

Roots or Edges?

19

Table 1. Left Word Edge Contexts Language

Transitivity

Left Word Edge Contexts

Yukatek

Intransitive Verb Transitive Verb

Positive Imperative; COM 1–6 Positive Imperative

Tzeltal

Intransitive Verb Transitive Verb

Positive Imperative; COM 1–6 Positive Imperative

Tzotzil

Intransitive Verb Transitive Verb

Positive Imperative; (COM 3 and 6) Positive Imperative

Q’anjob’al

Intransitive Verb Transitive Verb

Positive Imperative; IV verb complement Positive Imperative; IV verb complement

K’iche’

Intransitive Verb

Imperative (suppletive forms)

Table 1 shows the necessity of distinguishing between transitive and intransitive verbs for the purpose of analyzing the contexts in which the verb root may appear at the left word edge. We provide examples of verb roots at the left word edge in (6). (6)

Examples of Left Word Edge Contexts Yukatek – IVCOM2 (h) lúub-ech (COM) fall-2ABS ‘You fell.’ Tzeltal – Positive Imperative jajch-an get_up-STAIV-IMP ‘Get up!’ Tzotzil – IVCOM3 (i)-∅-bat xa (COM)-3ABS-go now ‘She’s gone now.’ Q’anjob’al – IVverb complement q-∅-xew ∅-qajab’-i POT-3ABS-finish 3ERG-talk-STAIV-INC ‘When he/she finishes talking.’

20 Clifton Pye, Barbara Pfeiler, Lourdes de León, Penelope Brown & Pedro Mateo K’iche’ – Suppletive Imperative saj come ‘Come!’ Right Edge Contexts The status suffixes occur to the right of the verb root in Mayan languages, however they exhibit considerable formal variation across the languages as shown in (7). Tzeltal and Tzotzil lost the ProtoMayan (PM) status suffixes for verbs in the incompletive and completive aspects (Kaufman & Norman 1984), while Yukatek has two forms of status suffixes for the two classes of intransitive verbs. The Q’anjob’al and K’iche’ ‘plain’ status suffixes vary in form according to whether the verb occurs at the end of the sentence. The forms shown in parentheses only appear when the verb is in sentence final position. K’iche’ and Q’anjob’al have a separate class of status suffixes that appear on derived transitive verbs in all sentence positions. (7)

Mayan status paradigms Yukatek INC COMP PERF IMP SUBJ

TV -ik -ah -m-ah ∅/-eh ∅/-eh

IV ∅/-Vl ∅/-ih -ah-a’an -(n)en -Vk/-nak

Tzeltal

Tzotzil

Q’anjob’al

TV IV ∅ ∅ ∅ ∅ -oj -em -a -an -ok/-uk

TV ∅ ∅ -oj -o -uk

TV (-V’) (-V’)

IV (-i) (-i)

(-V’)

-oq

IV ∅ ∅ -em -an

K’iche’ TV (-o) (-o) -V:m -V

IV (-ik)/-a (-ik)/-a -inaq -a

In Yukatek all the status suffixes appear when the verb occurs at the end of the sentence. The status suffixes also appear on transitive verbs in sentence-medial position when the verb is in the incompletive and completive aspects. In verbs in the completive aspect, an optional phonological contraction process deletes the status suffix (8). (8)

Yukatek Contraction a. Before contraction t-a mach-ah le che’=o’ COM-2ERG grab-STATV-COM DET wood=DIST ‘You grabbed this stick.’

Roots or Edges?

21

b. After contraction t-a mache che’=o’ COM-2ERG grab=DET wood=DIST ‘You grabbed this stick.’ Mayan languages commonly employ contraction processes at the surface phonetic level. We defined the “right edge” of the word at the syntactic level before such elisions occur. The verb root in the first Yukatek example (8a) was not counted as occurring at the right edge since it contains a status suffix while the verb in 8b does not have an overt status suffix, but at the morpho-syntactic level it does (i.e. before the contraction -ah + le > e) and so was not counted as a case where the root occurs at the right edge. Several other details must be recognized. The position of absolutive marking on the verb also interacts with the use of a status suffix. The absolutive is a prefix in K’iche’ and Q’anjob’al, but a suffix in Tzeltal and Yukatek. Consequently, the completive status suffix for intransitive verbs in Yukatek only occurs for third person absolutive subjects that have a zero form. The imperative status suffix for transitive verbs in Tzeltal and Tzotzil also fails to surface when the verb has the applicative suffix -be. Table 2 lists the contexts in which the verb root can surface without an overt suffix – i.e. at the right edge of the word – in our five languages. Table 2. Right Word Edge Contexts Language

Transitivity

Right Word Edge Contexts

Yukatek

Intransitive Active Verb Transitive Verb

INC1–3 IMPSM

Tzeltal

Intransitive Verb Transitive Verb

INC3; COM3; INC3; COM3; IMPNEG

Tzotzil

Intransitive Verb Transitive Verb

INC1–3; COM1–3; IMPNEG INC1–3; COM1–3; IMPNEG

Q’anjob’al

Intransitive Verb Transitive Verb

INC1–3SM; COM1–3SM; POT1–3SM INC1–3SM; COM1–3SM; POT1–3SM

K’iche’

Intransitive Verb Transitive Verb

INC1–3SM; COM1–3SM INC1–3SM; COM1–3SM

This table specifies distinct contexts for transitive and intransitive verbs, since verb transitivity changes the contexts where the verb root can occur at

22 Clifton Pye, Barbara Pfeiler, Lourdes de León, Penelope Brown & Pedro Mateo the right edge of the verb word in some of the languages. Yukatek adds a further restriction in the form of the ‘active’ intransitive verbs. Table 2 only includes nonderived verbs since most derivation adds a suffix to the right edge of all verb stems in all of the languages. The verb root can occur at the right edge of the verb word in the incompletive and completive aspects in all of the languages. However, there is a restriction to third person contexts for transitive verbs in sentence-medial position in Yukatek where the absolutive cross-reference morpheme for the object is zero (9a). Yukatek, Q’anjob’al and K’iche’ further restrict the verbs to cases where the verb occurs in sentence-medial position (9b and c). There are a set of more varied contexts shown in the final column of Table 2. Tzotzil and Tzeltal have an irrealis verb form in which the root can occur at the right edge of the verb word (9d). This is also the case for the potential in Q’anjob’al (9e). Once again, we highlight the verb root in each example in italics. (9)

Examples of Right Word Edge Contexts a. Yukatek – TV Taas] u=láak’ silla. bring:IMP 3ERG=other chair. ‘Bring another chair!’ b. Yukatek – IVactive táan in=meyah] k’iiwik] PROG 1ERG-work] in the plaza ‘I’m working in the plaza.’ c. K’iche’ x-in-pet] iwir] COM-1ABS-come] yesterday] ‘I came yesterday.’ d. Tzotzil mu x-a-pik-∅] li vaj=e] NEG INC-2ERG-touch-3ABS] DET tortilla=PTDET] ‘Don’t touch the tortilla.’ e. Q’anjob’al hoq-∅ hin-tx’aj] an q’apej] POT-3ABS 1ERG-wash] CL clothes] ‘I will wash the clothes.’

Roots or Edges?

23

Sentence Edge Contexts Tzeltal and Tzotzil lack status suffixes in the incompletive and completive aspects, so in these languages, the verb root can appear simultaneously at the right edge of the verb stem and the right edge of the sentence. We refer to verb roots at the right edge of the verb as occurring at the right word edge. We refer to the condition when the verb root occurs at the end of the sentence as occurring at the right syntactic edge. Verb roots can only occur at the right syntactic edge if they are already at the right word edge. By distinguishing between the word and syntactic contexts we can determine the degree to which children’s verb forms reflect features of the lexical and/or syntactic environment of the input language. Q’anjob’al and K’iche’, and to a lesser extent Yukatek, have status suffixes that appear at the right sentence edge, so in these three languages, the verb root only occurs at the right edge of the verb stem but not at the end of the sentence. By distinguishing between medial and final contexts across each language, we can assess the degree to which the structurally licensed use of the verb root at the right edge of the verb or at the right edge of the sentence contributes to the children’s productions of bare verb roots. Mayan verb roots also appear at left sentence edges when the sentence begins with a verb that lacks any morphology to the left of the root. This commonly occurs in positive imperative sentences. This review of Mayan morphology suggests various factors that determine when a Mayan verb root appears at the left or right edge of the verb word in the adult language. We divide these factors into left edge and right edge factors. The left edge factors are: 1. overt aspect prefixes, and 2. overt cross-referencing prefixes. The right edge factors are: 1. overt derivational suffixes, 2. overt status suffixes, 3. overt absolutive and plural suffixes, and 4. whether the verb appears in sentence final position. We have divided what follows into four sections. The next section presents the data we used in our study. The third section presents our tests of the effects of the left edge and right edge factors on the children’s production of bare verb forms. The final section discusses the implications of our findings for current theories of verb development in child language. Subjects & Data Our data is drawn from corpora of naturally occurring speech in family contexts for each of the five languages.3 For this study we extracted data on

24 Clifton Pye, Barbara Pfeiler, Lourdes de León, Penelope Brown & Pedro Mateo verb production from a single child for each language. We used samples with two-year-old children at the one- to two-word stage. The following table provides information on the children’s age, and number of sentences for each language. Table 3. Age, number of sentences, and number of verbs for the child language samples

Yukatek Tzeltal Tzotzil Q’anjob’al K’iche’

Child

Child’s Age

No. of Sentences

No. of Verbs

ARM XAN MAL NIK TIY

2;0 2;2 2;2 2;3 2;1

265 557 270 772 157

71 176 47 29 37

We tested the verb realization factors by analyzing samples of child directed speech in each language on the assumption that child directed speech would reflect the structural constraints in each language. We looked at a sample of adult speech in the vicinity of a child around the age of 2;0 in each language. Due to the nature of our recordings, all of which have lots of input speech from other children (siblings, cousins, neighbors, etc.), we agreed to count children over ten years old as adults for the purpose of analyzing the input speech. These samples are approximately the same duration with the exception of Q’anjob’al. Table 4 provides background information on each sample. Table 4. Samples of Mayan Input Speech

Yukatek Tzeltal Tzotzil Q’anjob’al K’iche’

Child’s Age

No. of speakers

Hours Taped

2;0 1;10 –2;10 1;9 2;3 2;1

3 2 1 1 5

~4 ~6 ~4 ~1 5

No. of verbal utterances 245 ~256 ~186 296 807

It is worth noting that even when we tried to find samples that were similar across the languages, some variables were beyond our control. One striking difference concerns the number of speakers who were present during the

Roots or Edges?

25

recording sessions. The speech of any given adult was restricted by the number and social status of other speakers who were present at each recording session. Mayan Children’s Early Verb Forms Since verbs in the adult Mayan languages have similar inflectional templates it is not clear what factors determine the form of children’s early verb productions. We provide examples of the children’s early verb forms in (10). We highlight the verb root in each example in italics. The examples in (10a) are all cases where the children produced the verb root minus obligatory inflections for aspect, agreement and status. Adult Mayan speakers frequently produce sentences that only contain a verb, but the adult verbs contain the inflections for subject and object as well as for aspect and status. The children’s productions in (10a) lack these obligatory inflectional features. The examples in (10b) illustrate another common verb form for children – the use of the verb root with the status suffix. The status suffix is the most semantically complex affix in the Mayan languages, and yet we find abundant evidence of its early appearance in children’s speech (Brown 1998; de León 1999b; Mateo 2005; Pfeiler 2003; Pye 1983). The remaining examples in (10c and d) illustrate more idiosyncratic productions which have a more irregular distribution across the Mayan languages. The applicative suffixes shown in (10d) frequently appears in Tzeltal, Tzotzil and Yukatek child and adult speech, while it is infrequent in K’iche’ and absent in Q’anjob’al (Pye 2007). (10) Mayan Children’s Early Verb Forms a. Bare Root Yukatek pax ‘play/music’ Tzeltal muk ‘cover’ Tzotzil k’an ‘want’ Q’anjob’al tantu ‘take care’ K’iche’ loq’ ech wa’ ‘buy ours then’

(Pfeiler 2003) (SAN 1;9) (Brown 1997) (MIK 1;10) (de León 1999a) (MAL 1;5) (Mateo 2005) (B 2;7) (Pye 1983) (CHA 2;9)

26 Clifton Pye, Barbara Pfeiler, Lourdes de León, Penelope Brown & Pedro Mateo b. Root+Status Yukatek bin-ih ‘go-STAIV-COM’ Tzeltal poch’-em ‘peel-STAINTR-PERF’ Tzotzil jam-o ‘open(it)-STATR-IMP’ Q’anjob’al man-a’ ‘buy-STATV-IMP’ K’iche’ e-k ‘go-STAIV-COM’

(Pfeiler 2003) (SAN 1;9) (Brown 1997) (XAN 2;1) (de León 1999b) (MAL 1;1) (Mateo 2005) (B 2;7) (Pye 1983) (TIY 2;1)

c. Root+Applicative Yukatek ts’iib’-t-∅-eh ‘write-APL-3ABS-STATV-IMP’ Tzotzil ak’-b-∅-o ‘give-APL-3ABS-STATV-IMP’ d. Ergative+Root Q’anjob’al ko-ten aj ‘4ERG-touch up’

(Mateo 2005) (N 2;3)

Figure 1 provides quantitative data on the frequency of bare verb roots in our child data. In this figure, ‘root’ indicates the pattern illustrated in (10a). We analyzed the productions of transitive (TV) and intransitive (IV) verbs separately in each language since these verb types have different affixes, which affected the rate of bare root production across the languages. Figure 1 shows large differences between the languages in the proportion of bare verb roots the children produce. Bare verb roots are the most frequent form produced by Tzeltal and Tzotzil children, while K’iche’ and Yukatekan children produce both bare roots and root-suffix combinations. Q’anjob’al children display an intermediate position. They produce a mix of root plus status forms for intransitive verbs, but concentrate on verb roots alone for transitive verbs. In fact, Figure 1 shows an interesting difference in the frequency in the verb roots of the children’s transitive and intransitive verbs across these five Mayan languages. In all except Q’anjob’al, the children produce more bare roots for intransitive verbs than for transitive verbs. The children also produce more root plus status combinations for transitive verbs than for intransitive verbs. We will show that

Roots or Edges?

27

Yukatek

Figure 1. Percentage of Bare Verb Roots in Mayan Children’s Speech

this variation provides further insight into the factors that affect children’s use of bare verb roots. In the next section, we test the left edge factors that we think might influence the children’s use of bare verb roots. Testing Left Edge Factors An obvious hypothesis is that variation in the linguistic structure of the adult languages dictates the forms of children’s verbs. One possibility is that Mayan children are more likely to produce bare verb roots if adults frequently produce verbs without prefixes. If the left word edge is a significant factor in the children’s use of bare verb roots, we predict a positive correlation between the frequency of roots at the left edge of verbs in the input and the children’s production of verb roots. The input frequencies of verb roots occurring at the left word edge of the verb stem are shown in Figure 2. We used the Spearman rank order correlation to compare the rank orders of the children and adults (Siegel 1956). This comparison reveals factors that work in common across the five languages, but does not show whether a factor has a stronger effect within the individual languages. We compare the frequency of left word edge verb roots in the input with the children’s production of bare verb roots in (11). This comparison is not

28 Clifton Pye, Barbara Pfeiler, Lourdes de León, Penelope Brown & Pedro Mateo

Yukatek

Figure 2. Frequency of Left Word Edge Verb Roots in Child Directed Speech

significant (r = 0.139 ns). We conclude that the frequency at which Mayan children produce bare verb roots is not tied to the frequency with which verb roots appear at the left edge of the verb in the input. (11) Comparison of Children’s Bare Verbs and Input Frequency of Left Word Edge Roots Child Frequency of Bare Verb Roots 1 Tzotzil IV 2 Tzeltal IV 3 Tzotzil TV 4 Q’anjob’al TV 5 K’iche’ IV 6 Tzeltal TV 7 Q’anjob’al IV 8 Yukatek IV 9 Yukatek TV 10 K’iche’ TV

Input Frequency of Left Word Edge Roots 1 Tzeltal IV 2 Yukatek IV 3 Tzotzil TV 4 Yukatek TV 5 K’iche’ IV 6 Tzeltal TV 7 Tzotzil IV 8 K’iche’ TV 9 Q’anjob’al IV 10 Q’anjob’al TV

Roots or Edges?

29

Yukatek

Figure 3. Frequency of Sentence-Initial Verb Roots in Input Speech

Sentence-Initial Position We also analyzed the frequency with which the speakers produced verb roots in sentence-initial position in their child directed speech. Verbs commonly occur at the beginnings of sentences in all Mayan languages, so Mayan children might be drawn to the extraction of verbs and verb roots by the frequency of verb-initial sentences in their input. Figure 3 shows the input frequencies at which verb roots appear in sentence initial position in the five languages. We compare these results with the children’s production of verb roots in (12). This comparison also turns out to be non-significant ( r = –0.219 ns). (12) Comparison of Children’s and Input Frequency Rank Order for Sentence-Initial Roots Child Frequency of Bare Verb Roots 1 Tzotzil IV 2 Tzeltal IV 3 Tzotzil TV 4 Q’anjob’al TV

Input Frequency of Left Sentence Edge Roots 1 Yukatek IV 2 Tzeltal IV 3 Yukatek TV 4 K’iche’ IV

30 Clifton Pye, Barbara Pfeiler, Lourdes de León, Penelope Brown & Pedro Mateo 5 6 7 8 9 10

K’iche’ IV Tzeltal TV Q’anjob’al IV Yukatek IV Yukatek TV K’iche’ TV

5 6 7 8 9.5 9.5

Tzeltal TV K’iche’ TV Tzotzil IV Tzotzil TV Q’anjob’al IV Q’anjob’al TV

Imperatives In addition to word and sentence edges, we explored the possibility that the number of imperative verb forms in the input might affect the children’s production of verb roots. Our thinking is that positive imperative verb forms are less inflected than verb forms in other aspects and modalities. In all the languages except K’iche’ the positive imperatives lack prefixes for modality and subject agreement. Imperatives in all Mayan languages have a status suffix that should help children to isolate the verb and the verb root. We provide examples of imperative verb forms in (13). The frequency of positive imperatives as a percentage of verbal utterance in input speech is shown in Figure 4. (13) Examples of Mayan Imperative Verb Forms Tzeltal jajch-an get_up-STAINTR-IMP Get up! Tzotzil lik-an get_up-STAINTR-IMP Get up! Yukatek líik’-en get_up-STAINTR-IMP Get up! K’iche’ ch-at-pakal-oq IMP-2ABS-get_up-STAINTR-IMP Get up!

Roots or Edges?

31

Yukatek

Figure 4. Frequency of Positive Imperatives in Input Speech

We compare the frequency of imperatives in the input with the children’s use of bare verb roots in (14). The correlation is not significant (r = 0.042 ns). We conclude that the number of imperatives in the input is not responsible for the number of bare verb roots we find in the children’s production. (14) Comparison of Positive Imperatives in the Input and Children’s Verb Root Production Child Frequency of Bare Verb Roots 1 Tzotzil IV 2 Tzeltal IV 3 Tzotzil TV 4 Q’anjob’al TV 5 K’iche’ IV 6 Tzeltal TV 7 Q’anjob’al IV 8 Yukatek IV 9 Yukatek TV 10 K’iche’ TV

Input Frequency of Positive Imperatives 1 Tzotzil IV 2 K’iche’ IV 3 K’iche’ TV 4 Yukatek IV 5 Tzotzil TV 6 Yukatek TV 7 Tzeltal TV 8 Tzeltal IV 9 Q’anjob’al TV 10 Q’anjob’al IV

We have tested three left edge factors in the input that might influence the Mayan children’s production of bare verb roots: 1. the frequency at which

32 Clifton Pye, Barbara Pfeiler, Lourdes de León, Penelope Brown & Pedro Mateo the root occurs at the left edge of the verb, 2. the number of times the root occurs at the beginning of a sentence, and 3. the number of positive imperatives in the input. None of these factors correlates with the children’s production of bare verb roots; hence, we conclude that the children ignored the left edges of verbs and sentences in extracting verb roots. In the next section, we test the effect of right edge factors on the children’s production of bare verb roots. Testing Right Edge Factors Another possibility is that Mayan children are more likely to produce bare verb roots if adults frequently produce verbs without suffixes. In other words, the frequent occurrence of verb roots at the right edge of the verb stem will increase the likelihood of children producing just the verb root. If the right word edge is a significant factor in the children’s use of bare verb roots, we predict a positive correlation between the frequency of roots at the right word edge in the adult input and the children’s production of bare verb roots. The input frequencies of verb roots occurring at the right edge of the verb word in our data are shown in Figure 5.

Yukatek

Figure 5. Frequency of Right Word Edge Verb Roots in the Input

Roots or Edges?

33

We next compared the frequency at which verb roots in the input were produced at the right word edge in Figure 5 with the children’s production of bare verb roots presented in Figure 1. Recall that Figure 1 shows that bare verb roots are the most frequent form produced by the Tzeltal and Tzotzil children, while the K’iche’ and Yukatekan children produce both bare roots and root-suffix combinations. If the frequency of adult verbs with roots at the right word edge influences the form of the children’s verbs we should find a positive correlation between these two measures. The frequency rank orders for the children’s production of bare verb roots and adult verbs with roots at the right word edge are shown in (15). (15) Comparison of Children’s Bare Verbs and Adult Frequency of Right Word Edge Roots Child Frequency of Bare Verb Roots 1 Tzotzil IV 2 Tzeltal IV 3 Tzotzil TV 4 Q’anjob’al TV 5 K’iche’ IV 6 Tzeltal TV 7 Q’anjob’al IV 8 Yukatek IV 9 Yukatek TV 10 K’iche’ TV

Adult Frequency of Right Word Edge Roots 1 Q’anjob’al TV 2 Tzotzil TV 3 Tzeltal IV 4 Tzeltal TV 5 Q’anjob’al IV 6 K’iche’ IV 7 Tzotzil IV 8 K’iche’ TV 9 Yukatek IV 10 Yukatek TV

This comparison reveals a significant positive correlation between the frequency of verb roots occurring at the right word edge in input speech and the children’s production of bare verb roots (r = .624 p = .05). The verb forms that adult Mayan speakers produce influence the frequency at which children produce bare verb roots. Even though these languages share a common inflectional template for verbs, small differences in the adult inflectional paradigms lead to significant differences in the verb forms that children produce. The same comparison can be made between the children’s production of verb roots and the frequency of verb roots at the right sentence edge in adult speech. As we stated previously, only Tzeltal and Tzotzil, as shown in the examples (7), place the verb root at the right edge of the sentence (when

34 Clifton Pye, Barbara Pfeiler, Lourdes de León, Penelope Brown & Pedro Mateo the verb occurs at the end of the sentence). Yukatek, Q’anjob’al and K’iche’ have obligatory status suffixes that come between the verb root and the right edge of the sentence. Figure 6 shows the frequency of verb roots at the right edge of the sentence in our samples of input speech.

Yukatek

Figure 6. Frequency of Sentence-Final Verb Roots in Child Directed Speech

Figure 6 confirms our impression that adults produce verb roots at the end of sentences most frequently in Tzeltal and Tzotzil. Although verb roots frequently occur at the end of the verb stem in Q’anjob’al, they never appear at the ends of sentences. If the input frequency of verbs with roots at the right sentence edge influences the form of the children’s verbs we should find a correlation between these two measures. We compare the frequency of right sentence edge verb roots in the input speech with the frequency of verb roots in the children’s productions in (16). This comparison is also significant (r = 0.596; p = .05). Our comparisons show that the children’s production of bare verb roots is significantly correlated with the frequency with which the verb root appears at the right word edge as well as the right sentence edge in the input.

Roots or Edges?

35

(16) Comparison of Children’s Bare Verbs and Input Frequency of Right Sentence Edge Roots Child Frequency of Bare Verb Roots 1 Tzotzil IV 2 Tzeltal IV 3 Tzotzil TV 4 Q’anjob’al TV 5 K’iche’ IV 6 Tzeltal TV 7 Q’anjob’al IV 8 Yukatek IV 9 Yukatek TV 10 K’iche’ TV

Input Frequency of Right Sentence Edge Roots 1 Tzotzil TV 2 Tzotzil IV 3 Tzeltal IV 4 Yukatek IV 5 Tzeltal TV 6 Yukatek TV 7 K’iche’ IV 9 K’iche’ TV 9 Q’anjob’al IV 9 Q’anjob’al TV

Tzeltal and Tzotzil lack status suffixes for the plain aspects, a fact which increases the frequency with which the verb roots appear in sentence-final position in these languages. Yukatek, Q’anjob’al and K’iche’, on the other hand, have status suffixes which intrude between the verb roots and the right sentence edge. Despite this structural difference, our result shows the variation in the children’s use of bare verb roots correlates with the proportion of verb roots at the right edge in the input. Tzeltal and Tzotzil merely lie at one extreme of a continuum.

Verb Derivation We also explored the hypothesis that a high proportion of derived verb forms in adult speech will have a negative correlation with the children’s production of bare verb roots. It seems reasonable to suppose that, since verb derivation is expressed through suffixes in Mayan languages, the number of derived verbs in the input should interfere with children’s use of bare verb roots. Another possibility is that the more derived verbs children find in the input the more likely they are to produce a form that consists of a verb root and a derivational suffix. We provide examples of derived verbs in (17). (Roots are italicized; derivational suffixes are underlined.) These examples illustrate various derivational processes that occur in Mayan languages. All of the languages have a form of the ProtoMayan causative suf-

36 Clifton Pye, Barbara Pfeiler, Lourdes de León, Penelope Brown & Pedro Mateo fix similar to that found in Yukatek. Tzeltal and Tzotzil have a productive applicative suffix that is not used as frequently in the other languages. Q’anjob’al and K’iche’ maintain a reflex of the PM status suffix that distinguishes a class of derived transitive verbs. The frequency of each of these derivational processes differs dramatically from language to language. The proportion of nonderived verbs in adult speech is shown in Figure 7. (17) Examples of Derived Verbs Yukatek – Causative li’i(k)-s-eh get_up-CAUS-STATR-IMP ‘Get someone up!’

Tzotzil – Applicative y-ak’-oj-be-on 3ERG-give-PERF-APPL-1ABS ‘He/she gave it to me.’

Tzeltal – Applicative k-ak’-oj-be-at 1ERG-give-PERF-APPL-2ABS ‘I have given it to you’

K’iche’ – ‘Derived’ Verbs k-0-inw-aa-j INC-3ABS-1ERG-want-STATV-DER ‘I want it’

Yukatek

Figure 7. Proportion of Nonderived Verbs in Input Speech

We compare the proportion of nonderived verbs in the input with the children’s production of bare verb roots in (18). This time we find a significant

Roots or Edges?

37

correlation between the proportion of nonderived verbs in adult speech and the children’s production of bare verb roots (r = 0.588; p = .05). This result supports our earlier finding that the children are sensitive to the right word edge. Both inflection and derivation affect the right word edge and influence the children’s production of verb roots. (18) Comparison of Adult Nonderived Verbs and Children’s Bare Verb Root Production Child Frequency of Bare Verb Roots 1 Tzotzil IV 2 Tzeltal IV 3 Tzotzil TV 4 Q’anjob’al TV 5 K’iche’ IV 6 Tzeltal TV 7 Q’anjob’al IV 8 Yukatek IV 9 Yukatek TV 10 K’iche’ TV

Input Frequency of Nonderived Verbs 1 Tzotzil IV 2 Tzeltal IV 3 Yukatek TV 4 Tzotzil TV 5 Q’anjob’al TV 6 Tzeltal TV 7 Q’anjob’al IV 8 K’iche’ TV 9 Yukatek IV 10 K’iche’ IV

Status Suffixes There is also a common set of verb inflections we can examine for its effect on childreńs bare root production – the status suffixes. The status suffix appears just after the verb root and any derivation (see the Mayan verb template in (2) above). The status suffix also lies at the heart of the propositional structure since the suffix expresses aspect, modality and sentence status. We can press the status suffix into service as an independent test of the right edge effect. We expect children to produce more bare roots to the extent that the input lacks overt status suffixes. Figure 8 shows the frequency of status suffixes in the input. We tested whether the use of status suffixes in the input was inversely related to the children’s production of bare verb roots, and this is precisely what we find (19). This comparison produced a significant negative correlation (r = –0.806; p = .01). It also turns out that the use of status suffixes in the input positively correlates with the children’s use of status suffixes (r = 0.894; p = .01), as presented in (20).

38 Clifton Pye, Barbara Pfeiler, Lourdes de León, Penelope Brown & Pedro Mateo

Yukatek

Figure 8. Status Suffix Frequency in the Input

(19) Comparison of Child and Input Status Suffix Usage Children 1 Yukatek TV 2 K’iche’ TV 3 K’iche’ IV 4 Yukatek IV 5 Q’anjob’al TV 6 Tzotzil TV 7 Q’anjob’al IV 8 Tzeltal TV 9.5 Tzeltal IV 9.5 Tzotzil IV

Input 1 K’iche’ TV 2 Yukatek TV 3 K’iche’ IV 4 Q’anjob’al TV 5 Yukatek IV 6 Q’anjob’al IV 7 Tzeltal TV 8 Tzeltal IV 9 Tzotzil TV 10 Tzotzil IV

(20) Comparison of Input Status Suffix Usage with the Children’s Production of Bare Verb Roots Child Frequency of Bare Verb Roots 1 Tzotzil IV 2 Tzeltal IV 3 Tzotzil TV

Input Frequency of Status Suffixes 1 K’iche’ TV 2 Yukatek TV 3 K’iche’ IV

Roots or Edges?

4 5 6 7 8 9 10

Q’anjob’al TV K’iche’ IV Tzeltal TV Q’anjob’al IV Yukatek IV Yukatek TV K’iche’ TV

4 5 6 7 8 9 10

39

Q’anjob’al TV Yukatek IV Q’anjob’al IV Tzeltal TV Tzeltal IV Tzotzil TV Tzotzil IV

Verb Root Structure The final factor we explored was the shape of the verb root. Tzeltal and Tzotzil contain a high proportion of CVC verb roots, while K’iche’ verb roots display more variation. Examples of two different types of K’iche’ verb roots are shown in (21). Figure 9 shows the proportion of CVC verb roots in the Mayan input. (21) Examples of K’iche’ Verb Root Structure k-0-pet-ik INC-3ABS-come-STAIV-INC He/She is coming.

k-0-xojow-ik INC-3ABS-dance-STAIV-INC He/She is dancing.

Yukatek

Figure 9. Frequency of CVC Verb Roots in the Input

40 Clifton Pye, Barbara Pfeiler, Lourdes de León, Penelope Brown & Pedro Mateo We compare the frequency of CVC Verb Roots in the input with the children’s production of bare verb roots in (22). This factor also turns out not to be significant in predicting the children’s use of bare verb roots (r = 0.541 ns). (22) Comparison of Input CVC Verb Root Frquency and Children’s Verb Root Production Child Frequency of Bare Verb Roots 1 Tzotzil IV 2 Tzeltal IV 3 Tzotzil TV 4 Q’anjob’al TV 5 K’iche’ IV 6 Tzeltal TV 7 Q’anjob’al IV 8 Yukatek IV 9 Yukatek TV 10 K’iche’ TV

Input Frequency of CVC Verb Roots 1 Tzeltal TV 2.5 Tzotzil IV 2.5 Tzotzil TV 4 Tzeltal IV 5 Yukatek IV 6 Q’anjob’al TV 7 Yukatek TV 8 Q’anjob’al IV 9 K’iche’ TV 10 K’iche’ IV

In sum, we tested five right edge factors in this section: 1. the frequency at which verb roots are produced at the right lexical edge, 2. the frequency at which verb roots occur at the ends of sentences, 3. the frequency of derivational suffixes, 4. the frequency of status suffixes, 5. the frequency of CVC verb roots. We found that the frequency at which verb roots occur at the right edge of the verb and sentence in the input was positively correlated with the children’s production of bare verb roots, while the frequency of derivation and the status suffixes in the input was negatively correlated with the children’s production of bare verb roots. We conclude that the frequency with which adults produce verb roots at the right edge of words and sentences influences the frequency at which children produce bare verb roots in their productions, but the left edge does not. The position that the verb usually occupies in the sentence (initial position) does not affect children’s ability to extract the verb root, nor does the degree to which the verb carries inflectional prefixes. The present analysis shows the predictive power of right edge factors even without accounting for other possible factors.4 We can add that this right edge influence is general across the five languages in our study, despite significant differences in their inflectional paradigms.

Roots or Edges?

41

Discussion In the Mayan languages, children must determine where the division between root and affix occurs. This necessity pits the basic semantic kernel against inflectional elements such as person and modality. If we just looked at the children’s productions in Tzeltal and Tzotzil, we would think the children were drawn to the semantic kernels of verbs. However, the children’s productions in Yukatek, Q’anjob’al and K’iche’ highlight the degree to which the structure of the input in Tzetal and Tzotzil favors the extraction of verb roots. The structure of the input language, not the semantic kernal, influences the form of children’s verbs. One of the more interesting variables across our languages is the degree to which adult speech employs derived verb forms. This is not a variable one normally thinks of when making crosslinguistic comparisons. Most derivations in Mayan languages affect the right edge of the verbal complex and we have shown that derivation in adult speech interferes with the childreńs production of bare verb roots. Mayan children will produce derived verbs to the extent that they appear in the input. We conclude that children produce bare verb roots to the extent that adult speech features verb roots at the right edge of the word and sentence. When derivation and status inflections are added to the right edge, children also incorporate these elements into their early productions. If the children are repeating what they hear in their input, it is natural to ask whether there is any indication that they understand every element of the verb they produce. There are two indications that they do. The first is from children acquiring Yukatek: some go through an early phase in which they extend the transitive status suffix on imperatives to intransitive imperative verbs (Pfeiler 2003). They do not hear these forms in the input, so once phonological development has been eliminated as a source of this error the only other possibility is that two-year old Yukatek children recognize the status suffix as separable from the verb base and extend it to intransitive verbs, good evidence that they have successfully processed the status suffix at some level. A second indication that Mayan children actively process the status suffixes comes from the alternation between phrase-medial and phrase-final forms of the status suffixes in Yukatek, Q’anjob’al and K’iche’. If the children simply produced the status suffixes as rote memorizations in these languages, we would not expect them to differentiate between these forms. Instead we find they vary their production of the status suffixes by the posi-

42 Clifton Pye, Barbara Pfeiler, Lourdes de León, Penelope Brown & Pedro Mateo tion of the verb. While it is true that they occasionally overgeneralize the phrase-final forms to phrase-medial contexts, they do not do this constantly (Mateo 2005; Pfeiler 2003; Pye 1983). Throughout our analysis, we have pursued a surface account of Mayan children’s verb forms. We have examined several surface features of verbs in the adult input that might determine the children’s initial verb forms. Our findings have major implications for theoretical accounts of language acquisition. Wexler and others have suggested that children alternate between finite and non-finite verb forms in the initial period of language development (Radford 1996; Rizzi 1993/1994; Wexler 1994). These accounts link uninflected forms of children’s verbs to limitations in children’s syntactic structures. All of these approaches assume that the non-finite verb forms children produce are appropriate in some adult sentence contexts, e.g. following negation. The Mayan children’s verb forms summarized in Figure 1 above violate this assumption, suggesting that attempts to state syntactic conditions on the forms of children’s early verbs are mistaken. At first glance, usage-based theories of acquisition (Maratsos & Chalkley 1980; Tomasello 2003) appear to provide a better account of the Mayan children’s early verb forms. These accounts place a great deal of emphasis on the effects of input frequency and semantic complexity. Our findings create some difficulties for these theories as well. Mayan children do not simply produce a copy of the adult verbs. Instead they initially produce only parts of the verb. Frequency alone cannot account for the parts of the verb that the children produce, since we have shown that an asymmetry exists between the left and right edges of the verb. The status suffixes present severe difficulties for theories based on semantic complexity. Mayan children are far more sensitive to features of the right verb edge (derivation, status) than to features of the left verb edge (aspect, agreement). The examples of inflectional overgeneralization we noted above also argue against a simple usage-based model of inflectional acquisition in the Mayan languages. Mayan children display a remarkable sensitivity to each part of the Mayan verbal complex. The parts of the verbal complex they produce are attuned to the structure of the input which displays considerable variation across the languages. The historically constrained variation in the forms of the verbal complex across our languages allows us to see how a single developmental process unfolds from five different perspectives. The results we have presented to this point are very preliminary. For the purposes of this presentation we examined the correlation between lan-

Roots or Edges?

43

guage features from a single adult and child in each language. Our next task is clearly to add other pairs of speakers to the analysis for each language so that we can better ascertain the variation that exists within the languages as well as across the languages. We are encouraged that the crude statistical analyses we performed on this data were sufficient to reveal common features of the acquisition process across the languages. Our study represents the first large-scale analysis of acquisition data from related languages.5 This unique data set allows analyses that would not be possible with acquisition data from unrelated languages. Inclusion of acquisition data from English and Italian would introduce too many extraneous variables, as both the morphology and syntax of the Indo-European languages differ markedly from the morphology and syntax of the Mayan languages. The restriction to this set of related Mayan languages permits us to analyze the variation in children’s verb productions in great detail. We can also exploit evidence from some of the languages to reach conclusions about the acquisition of all the languages. Evidence in some of the languages that Mayan children actively process the status suffixes allows us to conclude that children acquiring all Mayan languages actively process status suffixes. Documenting the acquisition of Q’anjob’al leads to a deeper understanding of the acquisition process in K’iche’ and vice versa. Working within a single language it is all too easy to assume that features such as aspect cohere semantically as well as in acquisition. It is only after comparing K’iche’ with Tzotzil and Yukatek that we recognized the need to make aspectual distinctions between transitive and intransitive verbs. There were literally a hundred other comparisons that we needed to make as a group in order to produce this analysis. Crosslinguistic comparison of unrelated languages cannot obtain this level of precision. Notes *

This paper is a revised version of a presentation we made at the International Association for the Study of Child Language in Berlin in July 2005. The order of authors reflects the fact that the first author took responsibility for writing various drafts. The research is a product of various years of joint collaboration by the authors. We thank Elena Lieven, Dan Slobin, Ann Peters, Richard Weist, Nancy Budwig and Joan Bybee for their comments in that venue. We owe a special thanks to Jürgen Bohnemeyer for his critical reading of a preliminary version of this paper. We accept responsibility for any remaining errors in our analysis.

44 Clifton Pye, Barbara Pfeiler, Lourdes de León, Penelope Brown & Pedro Mateo

1.

2.

3.

4.

5.

The research on Tzotzil was supported by CONACYT Mexico proyecto 42585-H to de León; the research on K’iche’ and Q’anjob’al was supported by National Science Foundation grant BCS-0515120 to Pye. K’iche’ and Q’anjob’al are the official spellings adopted in Guatemala during the 1980s. All Mayan examples are shown in the practical orthography developed by the Proyecto Lingüístico Francisco Marroquín (Kaufman 1976) with a single exception: we use rather than for the glottal stop. The other orthographic symbols have their standard IPA values except: = /ts/, = /t∫ /, = /ɓ/, = /ts’/, = /t∫ ’/, = /∫ / (except in Q’anjob’al where = /ʂ/), = /x/, and = high tone. We use the following abbreviations throughout the article: 1 first person singular ABS absolutive cross-reference 2 second person singular ERG ergative cross-reference 3 third person singular STA status suffix 4 first person plural V vowel COMPL completive aspect NEG negation INC incompletive aspect SM sentence medial PL plural DIST distant PRF perfect INTENS intensifier APPL applicative suffix PASS passive suffix ANTIP antipassive suffix POT potential AUX auxiliary verb SBJV subjunctive IMP imperative PT particle TR transitive verb DET determiner INTR intransitive verb DER derived CLF noun classifier PROX proximate PROG progressive CAUS causative The aspectual prefixes coordinate with the Mayan status suffixes discussed below. Mayan languages use the combination of aspectual prefix and status suffix to distinguish between different aspects and moods. The data for Yukatek was collected in Yalcobá, Yucatan, the data for Tzeltal in Majosik’, Tenejapa, Chiapas, the Tzotzil data in Nabenchauk, Zinacantán, Chiapas, the data for Q’anjob’al in Santa Eulalia, Guatemala, and the data for K’iche’ in Zunil, Guatemala. Our analysis neglects stress placement in the languages, which we assume also influences the children’s verb forms (Pye 1981). An analysis of the interaction of the right edge factors with stress remains for a future study. Plunkett & Strömqvist (1992: 540) cite a personal communication from Melissa Bowerman, who discusses the strengths and weaknesses of two approaches to crosslinguistic research. The first approach compares languages that differ considerably from one another. This approach can ‘refute gross overgeneralizations of universalist claims’, but cannot reveal details of the acquisition process

Roots or Edges?

45

‘precisely because of the fine-grained nature of the differences between the linguistic systems under investigation.’ The second approach compares languages ‘that differ only marginally across broad and detailed dimension’ (sic). ‘Given that two closely related languages share many properties within a given linguistic domain but differ on just one or two dimensions, the causes and ramifications of these differences can be more carefully explored.’ The second approach comes close to our application of the comparative method, but does not implement the comparative method’s techniques.

References Brown, Penelope 1997 Isolating the CVC root in Tzeltal Mayan: a study of children’s first verbs. In The Proceedings of the 28th Annual Child Language Research Forum, Eve V. Clark (ed.), 41–52. Stanford, CA: CSLI /University of Chicago Press. 1998 Children’s first verbs in Tzeltal: evidence for an early verb category. In Elena Lieven (ed.), Special edition of Linguistics 36 (4): 713–53. de León, Lourdes 1999a Verb roots and caregiver speech in early Tzotzil (Mayan) acquisition. In Cognition, discourse and function, Barbara Fox, Dan Juravsky & Laura Michaelis (eds.), 99–119. Stanford, CA: CSLI /University of Chicago Press. 1999b Verbs in Tzotzil early syntactic development. International Journal of Bilingualism 3 (2/3): 219–240. Kaufman, Terrence 1976 El proyecto de alfabetos y ortografías para escribir las lenguas mayances. Guatemala: Proyecto Lingüístico Francisco Marroquín & Ministerio de Educación. 1990 Algunos rasgos estructurales de los idiomas mayances con referencia especial al K’iche’. In Lecturas sobre la Lingüística Maya, Nora C. England & Stephen R. Elliott (eds.), 59–114. Guatemala: Centro de Investigaciones Regionales de Mesoamérica. Kaufman, Terrence S. & Norman, William M. 1984 An outline of proto-Cholan phonology, morphology, and vocabulary. In Phoneticism in Mayan Hieroglyphic Writing, John S. Justeson & Lyle Campbell (eds.), 77–166. Albany, NY: Institute for Mesoamerican Studies, State University of New York at Albany. Maratsos, Michael & Mary Anne Chalkley 1980 The internal language of children’s syntax: the ontogenesis and representation of syntactic categories. In Children’s Language, Vol. 2, Keith E. Nelson (ed.). New York: Gardner Press.

46 Clifton Pye, Barbara Pfeiler, Lourdes de León, Penelope Brown & Pedro Mateo Mateo, Pedro 2005 The Acquisition of Verb Inflection in Q’anjob’al. Unpublished MA thesis, Lawrence, KS: The University of Kansas. Pfeiler, Barbara 2003 Early acquisition of the verbal complex in Yucatec Maya. In Development of Verb Inflection in First Language Acquisition: A Cross-Linguistic Perspective, Dagmar Bittner, Wolfgang U. Dressler & Marianne Kilani-Schoch (eds.), 379–399. Berlin/New York: Mouton de Gruyter. Pfeiler, Barbara & Enrique Martín-Briceño 1998 La adquisición de los verbos transitivos en el maya yucateco. Función 17–18: 97–120. Plunkett, Kim & Sven Strömqvist 1992 The acquisition of Scandinavian languages. In The Crosslinguistic Study of Language Acquisition, Vol. 3, Dan I. Slobin (ed.), 457–556. Hillsdale, NJ: Erlbaum. Pye, Clifton 1983 Mayan telegraphese: intonational determinants of inflectional development in Quiche’ Mayan. Language 59: 583–604. Ms. The comparative context for Mayan status suffix acquisition. Kansas Working Papers in Linguistics. 2007 The genetic matrix of Mayan three-place predicates. In Bhuvana Narasimhan, Sonja Eisenbeiss & Penelope Brown (eds.), Special Edition of Linguistics 45 (3): 653–681. Radford, Andrew 1996 Toward a structure-building model of language acquisition. In Generative Perspectives on Language Acquisition, Harold Clahsen (ed.). Amsterdam: John Benjamins. Rizzi, Luigi 1993/94 Some notes on linguistic theory and language development: the case of root infinitives. Language Acquisition 3: 371–393. Robertson, John S. 1992 The History of Tense/Aspect/Mood/Voice in the Mayan Verbal Complex. Austin, TX: University of Texas Press. Siegel, Signey 1956 Nonparametric Statistics for the Behavioral Sciences. New York: McGraw-Hill. Tomasello, Michael 2003 Constructing a Language: A Usage-Based Theory of Language Acquisition. Cambridge, MA: Harvard University Press. Wexler, Kenneth 1994 Optional infinitives, head movement and economy of derivation. In Verb Movement, Norbert Hornstein & David Lightfoot (eds). Cambridge: Cambridge University Press.

Explaining Ergativity Clifton Pye 1

Ergative phenomena pose a formidable challenge to syntactic theory. Ergativity splits the core syntactic role of subject along lines of verb transitivity; the subjects of transitive verbs (A) take an ergative marking that distinguishes them morphologically from the subjects of intransitive verbs (Dixon 1979, 1994). The subjects of intransitive verbs (S) have an absolutive marking (often zero) that is identical with the marking of objects of transitive verbs (O). Any theory which assumes a primary syntactic relation of subject is forced to add some mechanism specifically tailored to such ergative features. Such mechanisms, be they an ergative “parameter” (Murasugi 1992) or an “active” Agreement element (Chomsky 1995) are patently ad hoc, failing to provide a satisfactory explanation of why ergative languages insist on morphologically dividing a core syntactic relation. The shortcomings of these approaches to ergativity become obvious when one considers various forms of split ergativity. Split ergativity occurs when the intransitive subjects that normally take the absolutive marking split into two groups, one with ergative marking and the other with absolutive marking. The ad hoc mechanisms of current theory do not predict where ergative splits can occur or place any constraints on ergative splitting. An investigation of ergative phenomena reveals the degree to which syntactic theories based on an Anglo-centric set of core syntactic relations fail to explain agreement facts in a significant proportion of the world’s languages. In this paper I demonstrate an approach to synchronic theory that uses the comparative method. Linguists commonly appeal to data from one or more unrelated languages to illustrate or support their theories without imposing any constraints on the languages they select. The lack of any systematic constraint on the selection process preserves linguistic theories from scientific testing, reducing them to little more than anecdotes. Linguists can use the comparative method to impose constraints on synchronic linguistic theory by using three or more historically related languages to test their theories. Working within a family of related languages rather than with several unrelated languages helps prevent theorists from selecting a few languages that happen to meet their expectations.

48 Clifton Pye I have selected the Mayan languages as a linguistic laboratory to expose the limitations of current syntactic approaches to agreement. These are languages that I have worked with for many years and languages that also have an extensive body of linguistic description (c.f., England & Elliott 1990). Mayan languages make use of a predominantly ergative agreement morphology and display diverse forms of ergative splits. In this paper I explore where current syntactic theories stand in providing an explanation for the ergative features of Mayan languages. I begin with a basic description of ergativity and the ergative features in Mayan languages that will constitute the focus of my investigation. I then survey several recent accounts of ergativity and identify their weaknesses. The failure of these theories to predict how Mayan agreement interacts with other functional categories provides the primary evidence that such theories fall well short of accounting for the basic structure of human language. The theories primarily fail to account for the interaction between agreement and discourse observed in Mayan and other languages. Such an interaction calls for a theory that integrates structure and discourse within a single framework rather than relegating discourse to a separate module. Ergativity in the Mayan Languages The Mayan languages typically display a head-initial syntactic order with variation for pragmatic reasons. The languages display an ergative-absolutive cross-referencing pattern on the morphological level. Although several authors have pointed out ergative features at the syntactic level (Foley & van Valin 1984; Larsen 1987; Manning 1996; Norman & Campbell 1978), alternative accounts of the syntactic constraints have been proposed (Aissen 1999). The ergative pattern of argument cross-referencing is clearly seen in the following examples from K’iche’, an Eastern Mayan language: (1)

K’iche’ examples a. x-in-r-il-oh COMP-1ABS-3ERG-see-STATUS ‘She/he/it saw me.’ b. x-0/-inw-il-oh COMP-3ABS-1ERG-see-STATUS ‘I saw him/her/it.’

Explaining Ergativity

49

c. x-in-pet-ik COMP-1ABS-come-STATUS ‘I came.’ Note that the example with the transitive verb in (1a) uses the same form of the absolutive cross-reference marker for the object as the intransitive verb example in (1c) uses for the subject. The first person absolutive is different from the first person ergative shown in (1b). The accusative structure of English assigns accusative case to objects and the nominative case to subjects. Mayan languages, in general, display an ergative pattern of cross-reference marking on finite verbs. The languages do not case-mark the subject and direct object arguments, which are subject to pro-drop in most clauses. Dixon and other researchers distinguish morphological ergativity such as that shown in (1) from syntactic ergativity (Dixon 1979, 1994). Morphologically ergative languages display the typical ergative morphological patterns, but preserve the basic accusative syntactic relations. Syntactically ergative languages extend the ergative pattern to their underlying syntactic structure. In such languages the object of a transitive verb displays a syntactic behavior like the subject of an intransitive verb. For example, the subject of transitive verbs is less accessible to wh-questions and relativization in K’iche’ than the subject of intransitive verbs and the direct object (Larsen 1987). This difference in accessibility could be accounted for if the direct object had a higher structural position than the subject (see 2). (The precise designation of the XP (i.e., whether it is a VP or some other functional projection) does not affect the argument that the object is in a higher structural position than the subject.) (2)

IP ei XP Object ei Subject X

I focus in this paper on structural accounts of morphological ergativity rather than syntactic ergativity because the ergative patterns of agreement have to be explained regardless of the language’s underlying syntax and because the concerns I will raise apply to morphologically and syntactically ergative languages alike. The issue of syntactic ergativity turns out to be orthogonal to an explanation of morphological ergativity, and thus need not

50 Clifton Pye be considered when constructing a theory of morphological ergativity. This will become clear in the third section which reviews several recent accounts of morphological ergativity. Split Ergativity in Mayan Languages Split ergativity takes many forms in the world’s languages (Dixon 1994). The Mayan languages demonstrate a typologically unusual pattern of split ergativity in which the intransitive subjects that normally take the absolutive marking split into two groups, one with ergative marking and the other with absolutive marking (Larson & Norman 1979). The phenomenon of split ergativity in Mayan clarifies one aspect of morphological ergativity that is otherwise obscure in fully ergative systems. The morphological feature of ergativity appears to be a primary attribute of the intransitive subject relation (S) rather than a feature of either the transitive verb subject or object. In most ergative languages, ergative splits occur with transitive subjects, e.g., Hindi, Dyirbal. The Mayan form of split ergativity shifts the focus of attention to the morphological realization of the S relation. The Mayan pattern implies that theoretical accounts of ergativity which treat S and O as underlying nominatives are mistaken. The O relation of transitive clauses in the Mayan languages retains its subordinate relation with respect to the A, as shown by the existence of passive voice constructions in the languages. Mayan languages provide a wealth of examples of different ergative splits. The Mayan language family contains some 30 different languages with an historical depth of roughly 6,000 years (Kaufman 1990). The Mayan languages fall into four main subdivisions: a. Wastekan, b. Yukatekan, c. Western and d. Eastern. The first three branches are located primarily in Mexico while the Eastern branch is located in Guatemala and Honduras. Larsen & Norman (1979) provide an excellent overview of Mayan ergativity.2 Mochó, a member of the Kanjobalan subgroup of Mayan languages, has a split based on person. The ergative prefixes are used to cross-reference first and second person subjects on both transitive and intransitive verbs, as shown in (3):

Explaining Ergativity

(3)

51

(example (11) in Larsen & Norman 1979) a. ii-muq-u-0 1ERG-bury-status-3ABS ‘I buried it’ b. ii-maaq-i 1ERG-go_up-status ‘I went up’

In these examples, an ergative prefix marks the subject of both a transitive (3a) and intransitive verb (3b). With third persons, however, Mochó displays a typical ergative pattern in requiring an absolutive prefix for the subject of intransitive verbs (4): (4)

(example (12) in Larsen & Norman 1979) a. x-muq-u-0 3ERG-bury-status-3ABS ‘He buried it’ b. maaq-i-0 go_up-status-3ABS ‘He went up’

Thus, in Mochó, the ergative split is controlled by person. Third person intransitive subjects take absolutive agreement while first and second person intransitive subjects require ergative agreement. The split is restricted to the subject of intransitive verbs. Silverstein (1976) pointed out that split ergativity is frequently conditioned by an animacy hierarchy with the first and second person intransitive subjects frequently showing ergative agreement and thus conditioning the split. Mopan, like other languages in the Yukatekan branch, has a split based on aspect: (5) (example (13) in Larsen & Norman 1979) a-lox-aj-en 2ERG-hit-status-1ABS ‘You hit me’ (6) (example (14) in Larsen & Norman 1979) lub’-eech Fall-2ABS ‘You fell’

52 Clifton Pye (7)

(example (15) in Larsen & Norman 1979) in-lox-aj-ech 1ERG-hit-status-2ABS ‘I hit you’

Mopan intransitive verbs in the completive aspect display the expected pattern of ergative agreement. Mopan marks the subject of intransitive verbs and the object of transitive verbs with the absolutive set. The progressive aspect, however, creates the ergative split: (8)

(example (16) in Larsen & Norman 1979) tan

a-lox-ik-en 2ERG-hit-status-1ABS ‘You are hitting me’ PROG

(9)

(example (17) in Larsen & Norman 1979) tan

a-lub’-ul 2ERG-fall-status ‘You are falling’ PROG

(10) (example (18) in Larsen & Norman 1979) tan

in-lox-ik-ech PROG 1ERG-hit-status-2ABS ‘I am hitting you’ Ch’ol, Ixil and Pokomam display a similar type of ergative split in which the ergative cross-reference set is extended to the subjects of intransitive verbs in the incompletive aspect. Some languages in the Kanjobalan subgroup display an ergative split in some types of subordinate clauses. For example, Jakaltek main clauses use the regular pattern of ergative cross-referencing (Craig 1977:118), while aspectless complement clauses create a split: (11) (example (19) in Larsen & Norman 1979) x-0-w-ilwe hach hin-kol-ni COMP-3ABS-1ERG-try 2ABS 1ERG-help-status ‘I tried to help you’

Explaining Ergativity

53

(12) (example (20) in Larsen & Norman 1979) sab’ ichi ha-munlayi Early start 2ERG-work ‘You started to work early’ In (11), the subject of the transitive verb in the subordinate clause has ergative agreement while in (12), the subject of the intransitive verb in the subordinate clause also has ergative agreement rather than the expected absolutive agreement form. Larsen & Norman note that all of these forms of split ergativity have been documented in other language families. However, the Mamean subgroup displays a unique type of ergative split based on the presence of certain focused constituents. The unmarked word order in these languages is VSO, and verb agreement is one way the languages signal when a focused constituent is preposed to the verb. Larsen & Norman take the following Ixil examples from Ayres (1977: 612): (13) (example (21) in Larsen & Norman 1979) a. i-b’an q’oon kuxhtu? 3ERG-do slowly just ‘He did it slowly’ b. q’oon kuxh i-b’an-ata? slowly just 3ERG-do-status ‘He did it SLOWLY’ (14) (example (22) in Larsen & Norman 1979) a. wat’-oʔ jojli Sleep-4ABS face_down ‘We slept face down’ b. jojli ku-wat-eʔ face_down 4ERG-sleep-status ‘We sleep FACE DOWN’ The sentences in (13) show that preposing the adverb has no effect on the ergative agreement with transitive verbs, whereas the adverbial focus in (14) creates an ergative split with an intransitive verb. The subject of the intransitive sentence in (14b) has ergative agreement (ku-) rather than the expected absolutive agreement (-oʔ). In both (13b) and (14b), the preposed

54 Clifton Pye focused constituent triggers the use of a dependent status suffix. These dependent verb forms would be expected if the verbs were part of a subordinate clause rather than a main clause. If the focused adverbs actually belong to a higher clause, then the Mamean examples of split ergativity would be similar to the Kanjobalan form triggered by subordinate clauses. To conclude, the Mayan language family features a full range of ergative morphological phenomena from the full ergative paradigm of K’iche’ to ergative splits conditioned by person (Mochó), aspect (the Yukatekan subgroup), clause type (the Kanjobalan subgroup), and focus (the Mamean subgroup). A theoretical account of ergativity should not only account for the canonical ergative paradigm, but the forms of split-ergativity as well. The different patterns of split ergativity found within a single family of languages suggests that an ergative cross-referencing morphology has a strong proclivity to split along natural fracture zones. The challenge to linguistic theory is to develop a motivated account for just where such fracture zones can be expected. It would be a mistake to dismiss split ergative phenomena completely by relegating the languages to the dustbin of lexically motivated exceptions. Mayan languages with split ergativity cannot be analyzed as instances of a lexically assigned inherent Case since split ergativity results when different agreement forms are assigned to the same verb stem. Thus, split ergative agreement systems in Mayan properly belong to the class of structurally assigned agreement rather than quirky or lexically specified agreement. In the following section I will review several recent accounts of ergativity to determine the extent to which they offer a satisfactory account of Mayan ergativity. Ergativity in Theory In this section I review several recent attempts to incorporate ergative features into generative theory. I include Chomsky (1995), Ritter & Rosen (2000), Woolford (2000), and Bittner & Hale (1989) as typical approaches to ergative phenomena within the generative framework. For each approach, I discuss the mechanisms that were added to account for ergativity as well as the potential for these mechanisms to predict the types of ergative splits I discussed in the preceding section. I have adopted a modified version of syntactic tree structures for my discussion to illustrate the essential features of each approach. I list the functional projections in a left to right fashion that presupposes the usual hierarchical relations. Thus, the notation in (15) is equivalent to the tree structure in (16).

Explaining Ergativity

55

(15) An example of functional notation AgrSP Spec AgrS NP ϕ

AgrOP Spec AgrO NP ϕ

(16) An equivalent tree structure AgrSP qp Spec AgrS’ | ei NP AgrS AgrOP ϕ ei Spec AgrO’ | | NP AgrO ϕ

Chomsky 1995 In an admittedly preliminary sketch of his Minimalist Theory Chomsky (1995) includes a few observations on ergative syntactic structures. 3 Chomsky assumes that both agreement and structural Case are manifestations of the Spec-head relation (NP, Agr). Under his proposal both accusative and ergative languages would have the same structure for transitive verbs, as shown in (17). (17) Transitive verb structure (Chomsky 1995) AgrSP agr T AgrS Nom ϕ

TP tT

AgrOP agr V AgrO Acc ϕ

VP tV

Tense (T) raises to AgrS (e.g., [agr T AgrS]), and V raises to AgrO, (e.g., [agr V AgrO]). This complex includes the φ-features of Agreement and the Case feature provided by T and V. In both positions the φ-features of the Agr head of the Agr complex determine agreement. The element

56 Clifton Pye that adjoins to Agr (T or V) determines Case. The NP in the [Spec, head] relation to this Agr complex (after raising from the VP) bears the associated Case and agreement features. Chomsky (1998) relies on feature checking by the head through c-command rather than head raising, but my criticisms apply to both mechanisms. To extend this account to ergative forms of agreement Chomsky assumes that if the VP only contains one NP, one of the two Agr elements will be “active” (the other being inert or missing). If the AgrS is active: the single NP will have the properties of the subject of a transitive clause, i.e., Nominative. If the AgrO is active: the single NP has the properties of the object of a transitive clause, i.e., Absolutive. The “active” element typically assigns a less-marked Case to its Spec, which is also higher on the extractibility hierarchy. I provide an illustration of this difference in (18). (18) Intransitive verb structures (Chomsky 1995) Accusative AgrSP TP AgrOP Active tT Inactive [agr T AgrS] Nom

Ergative AgrSP TP AgrOP Inactive T Active [agr V AgrO] Abs

Chomsky’s approach has the virtue of localizing ergative phenomena on the subject of intransitive verbs (where the VP only contains one NP). One technical problem this approach must answer, however, is how the tense features are checked in the ergative structure. The story is coherent for the accusative intransitive structure since the tense features can be checked against the features of the subject NP. However, the situation is quite different for the ergative intransitive structure where the tense projection is stranded by the inactive subject agreement projection. It makes no sense to claim that tense is “inactive” in intransitive sentences as opposed to the transitive sentences since all sentences in ergative languages have some type of finiteness marking just like all sentences in accusative languages. One possibility would be to let the subject NP check its Case in T after checking agreement in AgrO. However, this possibility leads to a Case conflict between the Case features of T and V. The real problem for Chomsky’s approach is that it does not explain why one or another of the agreement phrases would become active. Thus, it fails to predict when the variable “active” applies to AgrS and AgrO in languages with split ergativity. Furthermore, the model does not admit any

Explaining Ergativity

57

principled means of deriving such predictions. Consider the Mochó form of split ergativity by person, and recall that intransitive verbs take the ergative cross-referencing prefixes with first and second person subjects, but take the absolutive cross-referencing suffixes with third person subjects. Person is one of the complex of φ-features of the agreement head, so Chomsky could describe the Mochó form of split ergativity by saying that the first and second person φ-features were responsible for “activating” nominative agreement. However, this line of analysis could just as easily predict that the third person φ-features would activate nominative agreement, a split not found in the world’s languages. The mechanism can describe the split, but not motivate it. Next, consider the role of aspect in the split ergativity of Mopan and Yukatek. It must be admitted at this stage that Chomsky never reconciled the place of aspect in the inflectional structure of his model. Versions of the model that include an aspect projection generally place it below the object agreement projection. The interaction between aspect and agreement in the Mayan languages forces us to consider the possibility that Aspect, like Tense, can check Case and Agreement features. Thus, a transitive structure that includes aspect, but not tense might look something like (19) in Chomsky’s model. The intransitive structure for an ergative language is shown in (20). (19) Transitive structure with Aspect AgrSP agr A AgrS Nom

AgrOP agr V AgrO Acc

AspectP

(20) Ergative intransitive structure with Aspect AgrSP Inactive

AgrOP agr V AgrO Acc

AspectP A

This structure offers the potential for the Aspect head to move to either AgrS or AgrO, although the technical problem of how the Aspect head interacts with the Verb in (20) would still need to be resolved. If Aspect moves to AgrS, it could “activate” the AgrS agreement features and generate an accusative structure, whereas if Aspect moves to AgrO it could “activate” the AgrO agreement features and generate an ergative structure. The

58 Clifton Pye problem is that the model does not offer any means of constraining Aspect movement other than stipulation, and thus, the model fails to constrain the movement of incompletive Aspect to AgrS that generates the ergative splits observed in the world’s languages. Finally, consider the case of ergative splits in subordinate clauses of the type found in the Kanjobalan subgroup of Mayan languages. This type of split reveals the real inadequacy of Chomsky’s approach. I assume that clause types, if distinguished in a model such as Chomsky’s, would have distinct Complementizer projections. Such projections are universally taken to dominate all other clausal projections as shown in (21). (21) Transitive structure with a Complementizer projection CP Comp

AgrSP agr A AgrS Nom

AgrOP agr V AgrO Acc

Chomsky’s model makes no provision whatsoever for the effect of Complementizer type on ergative agreement. All the essential operations of agreement checking in the model take place below the level of the Complementizer. Hence, if the model makes any prediction at all, it would predict that the type of Complementizer could not affect agreement. Thus, Chomsky’s account of ergative agreement, and by extension the model’s account of accusative agreement, must be rejected on empirical grounds. Ritter & Rosen Ritter & Rosen (2000) sketch a version of Chomsky’s model that adds a semantic interpretation to the subject and object agreement projections. They claim that Chomsky’s subject agreement head provides information about the initiation of events, and the object agreement head provides information about the delimitation (completion) of events. In their words ‘when the Spec position of an eventive functional projection is filled, that FP is “activated” and enters into the composition of the event interpretation’ (p. 193–194). They make a distinction between languages that focus on the initiation of events determined by the presence or absence of an initial event boundary (‘I-languages’), and languages that focus on the delimitation of events determined by the presence or absence of a terminal event boundary (‘D-languages’). I-languages have an active (or ‘specified’) ini-

Explaining Ergativity

59

tiation projection (equivalent to Chomsky’s AgrS) in event predicates, while D-languages have an active delimitation projection (equivalent to Chomsky’s AgrO) in event predicates. I-languages grammaticalize properties of subjects (e.g., nominative Case, agentivity), while D-languages are sensitive to properties of the direct object (e.g., Case distinctions, definitness). Ritter & Rosen claim that I-languages are subject to ergative splits based on person features since the initiator of an event is typically higher on an animacy scale than the patient. In their model the initiator NP moves to [Spec Init] (i.e., Spec AgrS) thereby ‘specifying’ the event initiation. This enables the model to assign nominative Case to the subject and accusative Case to the object (in [Spec Delim]). Third person NPs or NPs low in animacy, on the other hand, may remain in the VP and receive inherent Case. If the inherent Case is ergative, the model generates the form of split ergative Case marking found in Dyirbal (Dixon 1994), see (22). Third person subjects receive ergative Case in the VP, while objects receive a “default” nominative Case (= absolutive) in the topic phrase. (22) Ergative split by person (after Ritter & Rosen (72)) TopP Obj [nom]

FP-init

FP-delim

VP Subj [erg]

Ritter & Rosen also propose that D-languages are subject to ergative splits based on tense or aspect since many languages limit accusative Case marking of objects to the perfective aspect that is commonly associated with event delimitation. The delimiting object NP should only move to [Spec Delim] (i.e., Spec AgrO) in perfective aspect producing the usual nominative/accusative agreement pattern. In the imperfective aspect, the object NP would remain in the VP and potentially generate ergative agreement in a structure like that of (22). The non-delimiting object would force the subject NP to remain in the VP and take an inherent (= ergative) Case. Ritter & Rosen admit that this prediction is contradicted by the usual pattern of tense/aspect induced split ergativity. In Hindi, for example, the ergative pattern of Case marking occurs with the perfective aspect rather than the imperfective. Ritter & Rosen suggest that the perfective forms actually denote resultant states rather than events, thus inducing the ergative pattern of Case marking. While their model provides an ingenious method of adding semantic features to Chomsky’s minimalist structure, it has several flaws. The model

60 Clifton Pye does not distinguish between the marked nominative Case it assigns in the [Spec AgrS] position from the default nominative Case it assigns in the topic phrase. Their model places some subject NPs in the [Spec AgrS] and others in the [Spec Top] position without any syntactic evidence. The primary division between I-languages and D-languages relies on an unconstrained list of initiator and delimiter features that make it impossible to categorize any language with certainty. Their model describes the syntactic structure and Case marking of Dyirbal, but requires an extra assumption for Hindi. Most importantly, their model does not provide an accurate description of Case and Agreement in intransitive clauses. Ritter & Rosen focus their attention on ergativity in transitive clauses. They claim that intransitive clauses such as activities (e.g., run, sleep; c.f. Vendler 1967; Dowty 1979) form a natural class with accomplishments (e.g., build, draw) in I-languages, while result verbs (e.g., break, arrive) would group together with accomplishments in D-languages. Their model, then, generates the intransitive (and transitive) configurations shown in (23). (23) Prototype configurations in I and D-languages I-language transitive (delim)

D-language

FP-init

FP-delim

FP-init

FP-delim

NOM

ACC

NOM

ACC

intransitive (activity) FP-init

VP

NOM

intransitive (result)

VP ERG

or

ERG

or

Topic

FP-delim

NOM

ACC

Topic NOM

Their model generates nominative agreement for subjects of activities in I-languages, but predicts that subjects of activities in a D-language like English, where a delimiter is not present, would either receive nominative agreement in the specifier to the Topic projection or take inherent or ergative agreement. In either case, the intransitive subject would have a syntactic position that is different from the position of transitive subjects in the language. Their model also predicts the subjects of result verbs receive Case and Agreement differently than the subjects of activity verbs. In particular, their model predicts that the subjects of result verbs in English (a D-language) should have accusative Case.

Explaining Ergativity

61

Their model does not predict the particular constellation of ergative agreement systems found in Mayan languages. Many Mayan languages use word order to distinguish between definite and indefinite objects, so the languages would be classed as D-languages according to Ritter & Rosen’s criteria. The ergative split based on person in Mochó, a D-language, is unexpected in their model. If Mochó were classified as an I-language, their model would not predict the ergative agreement pattern found in transitive clauses. Their model would also not explain why the Mochó ergative split only occurs in intransitive clauses. The prototype for I-languages in (23) does not assign accusative agreement to the subject of an intransitive verb of any sort. Ritter & Rosen’s model predicts splits in both transitive and intransitive clauses. The model also fails to predict the Mopan pattern of ergative split based on aspect. The main difficulty is that their model fails to capture the intransitive constraint on the Mopan split. Finally, their model fails to predict the ergative split that occurs in subordinate clauses in the Kanjobalan Mayan languages. It lacks a mechanism that would allow the complementizer to influence Agreement and Case assignments. Bittner & Hale Bittner & Hale (1996) propose a complex structural account of Case and Agreement. The heart of their proposal centers around three distinct Cases, a structurally unmarked Nominative Case, and two marked Cases, the Ergative and Accusative Cases. The marked Cases arise in the presence of a “Case competitor.” For the Ergative Case, the Case competitor is the direct object. In ergative languages, the direct object lacks its own Case competitor, and thus functions as a Case competitor for the subject. The unmarked nominative object must satisfy a Case Filter by raising to [Spec, IP] or through transparency. For the Accusative Case, Bittner & Hale postulate that an abstract D element adjoined to the verb serves as the Case competitor for the direct object and licenses Accusative Case assignment. The unmarked nominative subject satisfies the Case Filter by raising to [Spec, IP] or through transparency. Bittner & Hale’s assumption of three distinct Cases allows them to immediately account for languages such as Nez Perce and Pitta-Pitta which make a three-way Case distinction. Bittner & Hale treat pronominal agreement as a relation between an argument chain and a functional head, the complementizer (C) or INFL (I) in their model. In the raising type of ergative language, the nominative object agrees with C and the ergative subject agrees with I. In languages like

62 Clifton Pye Warlpiri, with ergative Case and accusative Agreement, Bittner & Hale claim that base-generated binding accounts for the discrepancy between Case and Agreement. In these languages the nominative object agrees with I rather than with C. Given their extensive survey of Case and Agreement in the world’s languages it is surprising that Bittner & Hale do not discuss any languages with split ergative systems like Mayan.4 Split ergative languages create an immediate challenge for their proposal since Bittner & Hale assume that transitive clauses are the locus of ergativity. In Mayan split ergative systems, the intransitive verb can also appear with ergative agreement. Bittner & Hale’s model has no way to derive ergative agreement for intransitive clauses. Their analysis of Basque (a language with an ergative active Case system) requires the presence of an unincorporated object NP to generate an ergative Case for the subject of unergative intransitive verbs such as speak (hitz egin, lit. ‘do word’). This analysis cannot be extended to the types of ergative splits found in Mayan languages, since in Mayan the split applies to all intransitive verbs whether unergative or unaccusative. One possibility for dealing with the Mayan forms of agreement in their model would be to extend the marked Accusative Agreement to the subjects of intransitive clauses. The subjects and objects of transitive clauses would display the typical nominative-accusative pattern of agreement and the whole system would appear to have an ergative agreement pattern. In the Mayan languages with split ergativity, some intransitive subjects would then take nominative agreement (= ergative) while others would take accusative agreement (= absolutive), thereby producing a split ergative system. This suggestion would contradict Bittner & Hale’s claim that monadic verbs do not have Case competitors and therefore cannot assign marked structural Case, but it is the only real means of generating a split ergative system of the kind found in the Mayan languages. Even with such an extension, however, their model fails to predict the observed constraints on split ergative systems in Mayan. Their model has no means of limiting the splitting process to intransitive clauses. If the splitting process applied to a transitive clause, it would generate a second marked agreement (ergative) assigned to the subjects of some transitive verbs. There is no evidence of a three-way agreement system in any Mayan language. Their model also offers no plausible explanation why the ergative splits would be limited to the same person, aspect and clause conditions in language after language. Such a failure constitutes a serious flaw in their account, and makes their approach untenable.

Explaining Ergativity

63

Woolford Woolford distinguishes two types of ergative morphologies. Her ‘Hindi type’ of ergative morphology occurs when the subject has inherent Case (= ergative) and the object has nominative Case and agreement. Woolford claims that this type of morphology only occurs in languages with morphologically overt ergative Case. Her ‘Jacaltec type’ of ergative morphology occurs when the O is cross-referenced by a clitic. An S is cross-referenced by the same clitic set rather than by subject agreement, and ergative agreement is simply a nominative form of subject agreement as opposed to clitic cross-referencing. Subjects and objects in Jakaltek-type languages occupy the same positions as they do in accusative languages and the crossreferencing elements (subject agreement and clitics) work in a similar fashion. Woolford further assumes that the Jakaltek type of agreement results from a limit of one clitic and one agreement per clause, and that clitics have preference over subject agreement in intransitive clauses. Since the Mayan languages do not assign overt ergative Case, I will confine my analysis of Woolford’s model to a discussion of the Jakaltek type. Woolford’s model has several attractive features, not the least of which is the distinction between two different types of ergative structures. She accounts for languages such as Warlpiri by linking Case to ergative and absolutive grammatical relations, while controlling agreement with subject and object grammatical relations. Nevertheless, applying the model to the Mayan languages reveals certain inherent limitations. Like Chomsky’s model, Woolford’s has its own set of technical limitations. Woolford states that the object clitic has a higher attachment in the tree structure than AgrS. In a transitive structure such as that shown in (24), the higher attachment for the object clitic is unmotivated. In K’iche’, absolutive object agreement comes after the aspect prefix, and is morphologically and phonologically integrated into the verb. Woolford does not offer a solution for checking Tense features in intransitive clauses which presumably lack an agreement projection. (24) Transitive structure of the Jacaltec type (Woolford 2000) ClObj [agr V AgrO] Acc

AgrS [agr T AgrS] Nom

T

Like Chomsky, Woolford does not consider the full implications of split ergative systems for her model. To account for the Mochó type of person

64 Clifton Pye driven ergative split, Woolford would be forced to stipulate that the preference for clitic marking in intransitive clauses can be relaxed for first and second person subjects. The model does not supply any explanation why the clitic preference would be relaxed for first persons rather than third persons. Thus, Woolford’s model overgenerates ergative structures of the Jakaltek type. Once again, the Jakaltek type of clause complement ergative split is the most revealing of the model’s fundamental limitations. To account for the split form of Jakaltek ergativity, Woolford would be forced to add a mechanism that allows the Complementizer to relax the preference for clitic marking in intransitive clauses. Her model offers no hint of how this could be accomplished let alone why the restriction would be limited to subordinate clauses. Hence, Woolford’s model, like all the others, proves to be empirically inadequate. Explaining Ergative Splits In the previous section I explored four accounts of ergative agreement and rejected all of them on empirical grounds. The central problem for structurally-based models of agreement is providing an explanation for an interaction between ergative agreement and other inflectional features such as person, aspect and especially clause type. Of these limitations, the clause type of ergative split appears to be the most serious since any distinction of clause type in structural models occurs at a higher structural level than either subject or object agreement. Under these approaches, agreement checking is complete before the type of clause has a chance to influence the outcome. Interestingly, Larsen & Norman suggest deriving most of the Mayan types of split ergativity from the clause type of split. The aspectual split in the Yukatekan languages is driven by aspectual forms that can be traced historically to independent verbs. Assuming the progressive particle in Mopan is covertly an independent verb would align its structure with that of Jakaltek. Likewise, Larsen & Norman observe that the focus constructions in the Mamean languages also induce subordinate forms of the status suffixes on the verb. This morphology suggests that such focus structures likewise include covert subordinate clauses. Thus, an account of the clause type of ergative split would at once account for most forms of ergative splits in the Mayan languages as well as provide a solid empirical base for theoretical models of agreement.

Explaining Ergativity

65

It is important to remember that although clause complementation seems to have a powerful effect on ergative agreement systems, this effect is constrained in systematic ways. Across the world’s languages ergative splits occur in subordinate clauses rather than in root clauses, and only in incompletive aspects and non-past tenses. Thus, any model which hopes to explain split ergativity must also impose the correct set of constraints on where the ergative splits can occur. Dixon (1994) and Du Bois (1987) offer discourse-based accounts that tie together the types of splits I describe in this paper. Dixon’s primary claim is that where linguistic forces converge on the action of human agents languages will tend to tie the A and S relations together in an accusative pattern. Human agents are more likely to be involved in first and second person initiated events, and thus, an ergative split is expected between the first two persons and the third. Dixon also claims that human agents are usually the focus of non-past, incompletive events, just where the ergative split occurs in the Yukatekan languages. Finally, Dixon argues that human agents are likely to be the subjects of purposive clauses (“She covered the pot to hide the chocolate”), and thus an ergative split may be expected in some types of dependent clauses. Structural approaches to Case and Agreement miss this level of generalization because they lack a principled basis for incorporating semantically motivated distinctions into their models. All of the models that I discuss in this paper have this defect although Ritter & Rosen’s proposal goes some way toward incorporating semantic features via functional projections. Structurally based accounts lack a means of recognizing the systematic influence that discourse features and semantics inevitably exert on morphological structure. The study of split ergativity exposes a fundamental limitation of such approaches. Any explanation of the ergative agreement patterns found in Mayan languages must specify how the complementizer (the locus for clausal features), tense/aspect (the locus of temporal features) and transitivity (the locus of the A/S distinction) interact in assigning agreement to the S. We currently lack a principled theory that would predict how transitivity interacts with the complementizer, tense/aspect and person features. Dixon’s (1994) review of ergative languages suggests that the interaction is not limited to two factors at a time (transitivity, clause type and person), but there is no theoretical restriction on an interaction between all four factors in a single language.

66 Clifton Pye Conclusion Ergative languages have long constituted a major conundrum for theoretical linguistics. On the surface, there appears to be no justification for the distinction between the A and S grammatical relations found in ergative languages. Recent generative accounts of ergativity have attempted to produce ergative features through structural approaches. All of these accounts fail to consider the implications of the forms of split ergativity found in the Mayan languages for structural approaches to agreement. Such approaches predict many varieties of ergativity that are not attested in the world’s languages. This failure reveals a fundamental error in structural approaches to agreement. The error stems from a failure to incorporate a mechanism that recognizes the role that discourse and semantic factors play in organizing agreement morphology across ergative and accusative languages alike. The study of ergativity reveals a fundamental limitation in the core theoretical assumption that syntax and semantics can be isolated in separate modules of grammar. Notes 1. I wish to thank Jürgen Bohnemeyer and David F. Mora-Marín for their insights on Mayan grammar. I am solely responsible for any of the remaining defects in this paper. 2. Mayanists eschew the labels ergative and absolutive in favor of the terms ‘Set A’ and ‘Set B’ to refer to the cross-referencing morphology. In a simple ergative language, the Set A morphemes correspond to the traditional ergative set while the Set B morphemes correspond to the absolutive set. In split ergative languages, however, the ergative set of morphemes are no longer used exclusively with the subjects of transitive verbs. In these languages the terms Set A and Set B do not carry any expectations about an underlying ergative structure for the languages. I have retained the labels ergative and absolutive in this paper because they more accessible for non-Mayanists and because I feel a deeper understanding of ergative agreement includes an expectation of split ergativity. I have adopted the current spelling system for the names of the Mayan languages in this paper, which differs in some instances from older, Spanish-derived spellings. 3. Chomsky continues to revise major aspects of his conception of grammar as seen in more recent publications, e.g., Chomsky (2001). I have focused on his 1995 model since it the only one that recognizes the need to account for ergative systems of Case marking and agreement.

Explaining Ergativity

67

4. Section 12 of their paper discusses the split system of Austronesian. These types of split are actually alternative realizations of the same predicate argument arrays. The split Case and Agreement systems of Austronesian languages do not have any constraints by person, aspect or clause-type and occur in transitive clauses rather than in intransitive clauses.

References Bittner, Maria & Kenneth Hale 1996 The structural determination of Case and Agreement. Linguistic Inquiry 27: 1–68. Chomsky, Noam 1995 The Minimalist Program. Cambridge, MA: MIT Press. 1998 Minimalist Inquiries: The Framework. MIT Ocassional Papers in Linguistics 15. Cambridge, MA: MIT Press. 2001 Derivation by phase. In Ken Hale: A Life in Language, Michael Kenstowicz (ed.), 1–52. Cambridge, MA: MIT Press. Craig, Colette G. 1977 The structure of Jacaltec. Austin: University of Texas Press. Dixon, Robert M. W. 1979 Ergativity. Language 55: 59–138. 1994 Ergativity. Cambridge: Cambridge University Press. Du Bois, John 1987 The discourse basis of ergativity. Language 63: 805–855. England, Nora C. & Stephen R. Elliott (eds.) 1990 Lecturas sobre la Lingüística Maya. La Antigua, Guatemala: Centro de Investigaciones Regionales de Mesoamérica. Kaufman, Terrence 1990 Algunos rasgos estructurales de los idiomas Mayances con referencia especial al K’iche. In Lecturas sobre la Lingüística Maya, Nora England & Stephen R. Elliott (eds.), 59–114. La Antigua, Guatemala: Centro de Investigaciones Regionales de Mesoamérica. Larsen, Thomas W. 1987 The syntactic status of ergativity in Quiche. Lingua 7: 133–159. Larsen, Thomas W. & William M. Norman 1979 Correlates of ergativity in Mayan Grammar. In Ergativity: Towards a Theory of Grammatical Relations, Frans Plank (ed.), 347–370. London: Academic Press. Murasugi, Kumiko 1992 Crossing and Nested Paths: NP-Movement in Accusative and Ergative Languages. Unpublished PhD dissertation, MIT.

68 Clifton Pye Ritter, Elizabeth & Sara Rosen 2000 Event structure and ergativity. In Events as Grammatical Objects, Carol Tenny & James Pustejovsky (eds.), 187–238. Stanford, CA: CSLI Publications. Silverstein, Michael 1976 Hierarchy of features and ergativity. In Grammatical Categories in Australian Languages, Robert M. W. Dixon (ed.), 112–171. Linguistic Series 22. Canberra: Australian Institute of Aboriginal Studies. Woolford, Ellen 2000 Ergative agreement systems. The University of Maryland Working Papers in Linguistics 10: 157–191.

Early acquisition of the Split Intransitive System in Yukatek Carlos Carrillo Carreón

1. Introduction Research on language acquisition in non-European languages shows that children use a number of different strategies when learning grammar. This is the case with the Mayan languages, in which subject and object agreement is morphologically marked in the verb by two sets of cross-reference markers: Set A and Set B. Set A marks the agent, and Set B both the experiencer and the patient; it is thus characterized by an ergative pattern of argument marking. A comparative study of acquisition in four Mayan languages (Brown, de León, Pfeiler & Pye 2002, 2004) showed that crossreference markers do not begin developing until 2;6 years of age. Yukatek is different in this sense because Set A markers are developed in both transitive and intransitive verbs, which can be explained by the split intransitive system governing the language. Early verbal morphological development in Yukatek consists of imperative and status suffixes (Pfeiler & Martín Briceño 1997, 1998). Even at 2;4 years data on cross-reference markers indicate that these forms have not yet been acquired (Pfeiler 2003). To specifically analyze the development of cross-reference markers a study was done of a child aged 2;0 to 3;0 showing that at 3;0 verbal argument markers have not been totally acquired, and thus neither has the split intransitive system. Cross-reference markers are only productive with a small group of verbs compared to all the verb types, in the sense of Verb Islands (Tomasello 1992). Set A develops beginning at 2;6 in both transitive and intransitive verbs, but at this age it is possible to find clear distinction of intransitive verb classes. Set A markers are restricted to active intransitives, and Set B is exclusively used with inactive intransitives. The present study is aimed at demonstrating the role of frequency and morphological transparency in development of cross-reference markers in Yukatek and understanding the trajectory by which this split intransitive system is acquired.1

70 Carlos Carrillo Carreón 2. Split intransitive system in Yukatek Recent studies have restated the verbal argument marking pattern for Yukatek. Pustet (1992) questioned whether it is ergative, accusative or active. Krämer & Wunderlich (1999) and Bohnemeyer (2004) showed that Yukatek cannot be characterized as an active-inactive language since the subject of all intransitive verbs, be they active or inactive, can be marked as an agent or experiencer. In accordance with Bohnemeyer (2004) the Yukatek pattern differs from the ‘split-S’ system and ‘fluid-S’ system described by Dixon (1994: 71–78, 78–83) because it is neither conditioned by lexical semantics, nor by clause-level semantic construal of participant-structure factors of volitionality or control. Argument marking in intransitive clauses depends on aspectmood marking. Yukatek has two sets of cross-reference markers for verbal concordance of the subject and object: Set A (Table 1); and Set B (Table 2). In verbal inflection these forms are obligatory and independent of other (pro)noun phrases functioning as subject and object. Table 1. Set A Person 1

st

2nd 3rd

Singular

Plural

iN(w)= a(w)= u(y)=

k= in(w)= …-o?n2 a(w)= …-e?š u(y)= …-o?b

Singular

Plural

-en -eč -Ø, -ih3

-o?n -e?š -o?b

Table 2. Set B Person st

1 2nd 3rd

Set A forms are used to express the subject of transitive verbs (agent)(example 1), and the subject of intransitive verbs (experiencer) (example 2). (1)

k-

inw= il-ik-eč see-INCP-B.2SG ‘I see you’ IPFV A.1SG

Early acquisition of the Split Intransitive System in Yukatek

(2)

k-

a=

IPFV A.2SG

71

lúub-ul fall-INCP

‘you fall’ For the perfective aspect of intransitive verbs, the subject (experiencer) is marked with Set B (example 3): (3)

h= lúub-0/-eč PFV fall-COMP-B.2SG ‘you fell’

Note that the object of transitive verbs (patient) (as in example 1) is marked by Set B (except for -ih), the same set used for subjects of intransitive verbs (experiencer) in the perfective (example 3). There are semantically-motivated verbal classes that are based on the status4 inflection and form classes (Bohnemeyer 1998, 2002: 167). The intransitive verb class exhibits inactive, inchoative and positional intransitive verbal bases that lexicalize state changes. Active intransitives are derived from transitives by means of the antipassive voice. When Set A is the subject of an active intransitive (transitive in antipassive voice), it corresponds to an agent whose patient is omitted. In this sense, the subject (marked by Set A) in the active intransitive verbs does not have exactly the same experiencer roles as the subject of inactive intransitive verbs; it functions more as an initiator of incomplete processes. In my description of the Yukatek pronominal system I claim that Set B has a degree of morphological transparency, as in examples like No. 3, but Set A does not because it forms amalgams with a variety of aspect and mode auxiliaries. According to the Theory of Natural Morphology (Dressler 1985, 1999), affixes have morphological transparency to the degree that it is possible to morphotactically decompose and semantically recognize the root and affixes. This naturality parameter allows affixes to be the earliest morphological techniques in acquisition. Following is an analysis of how the grammatical structures described above occur in acquisition of Yukatek. 3. The data For this study I analyzed 51 recordings of spontaneous speech from the child Armando who lives in Yalcobá, in the east of the state of Yucatan.

72 Carlos Carrillo Carreón Table 3. Analyzed material Armando

Input

Number of recordings*

Age

Emissions

Verbs**

Emissions

Verbs**

5 4 4 4 4 4 3 3 4 4 4 4 4

2;0 2;1 2;2 2;3 2;4 2;5 2;6 2;7 2;8 2;9 2;10 2;11 3;0

330 434 416 470 491 531 416 357 446 501 326 690 504

30/109 75/191 28/49 75/72 60/106 89/196 73/153 58/94 99/145 111/157 86/119 149/222 127/199

362 338 353 259 185 133 169 52 107 207 111 111 152

197/253 168/203 102/139 87/114 80/85 63/72 94/106 28/36 48/50 92/121 52/57 39/45 89/106

* Each analyzed month consisted of 180’ of audio recording. ** (types/tokens)

4. Development of cross-reference markers The analysis was intended to understand the development of cross-reference markers by analyzing type and token frequency in all verbs, verb classes and the most frequent verbs.5 Cross-reference markers exhibited a different development in Set A and B that is supported by use frequency during the observation period and the use of verb classes. During the studied ages, Set A markers were generally more frequent than the overtly marked Set B markers, however, in the earlier stages Set A was almost nonexistent whereas Set B was present (Table 5).

Early acquisition of the Split Intransitive System in Yukatek

73

Table 4. Frequency of cross-reference markers Age

Armando

Input

Set A

Set B

Set A

Set B

2;0 2;1 2;2 2;3 2;4 2;5 2;6 2;7 2;8 2;9 2;10 2;11 3;0

3/3 3/3 7/7 9/12 12/13 27/38 26/32 33/39 55/65 74/81 60/67 89/94 90/117

8/15 9/15 1/2 5/7 8/9 9/12 6/9 10/14 13/16 9/12 6/8 13/14 20/23

125/139 109/117 71/93 51/59 52/52 40/40 57/63 18/22 29/29 54/63 34/35 21/21 63/67

12/17 17/19 11/15 7/7 5/5 7/7 12/12 4/4 1/1 7/7 2/2 1/1 6/7

Total

488/571

117/156

724/800

92/104

During the first two months, Set B markers occurred more frequently than those in Set A, with Set A markers increasing in frequency beginning at 2;5. Despite its early frequency versus Set A, Series B markers did not increase in frequency during the observation period. Only overtly marked forms were considered for Set B in this analysis. The most frequent forms from Set A were in(w)= (first person singular) and u(y)= (third person singular), and for Set B they were -en (first person singular), -ih (third person singular) and -o?ob (third person plural). In contrast to acquisition, Set A in the input (Figure 1) had a higher frequency than Set B from the beginning of the analyzed period. The most frequent forms in the input were a(w)= and u(y)= for Set A and –eč, -ih and -o?ob for Set B.

74 Carlos Carrillo Carreón

Figure 1. Frequency of cross-reference markers in acquisition and input.

Use of cross-reference markers in acquisition is different in each verbal class (Table 5), and these differences originate in use of Set A or B. Table 5. Armando (2;0–3;0). Use of cross-reference markers in transitive and intransitive verbs Age Transitives Set A Set B 3S-Ø Intransitives Set A Intransitives Set B 3S-Ø 2;0 2;1 2;2 2;3 2;4 2;5 2;6 2;7 2;8 2;9 2;10 2;11 3;0

16 24 3 11 12 27 32 22 41 37 42 71 77

1 6 3 2 4 20 22 14 32 31 36 54 74

3 4 0 0 2 3 15 1 5 1 1 2 2

13 20 3 11 10 24 27 21 36 36 41 69 75

7 17 5 7 6 23 22 15 27 34 22 28 30

0 1 1 7 5 20 19 15 27 32 21 25 30

16 18 2 7 19 20 16 16 10 9 11 20 26

10 16 2 5 13 18 10 9 9 4 10 9 9

6 2 0 2 6 2 6 7 1 5 1 11 17

Early acquisition of the Split Intransitive System in Yukatek

75

Table 6 shows the use of Set A in verbal classes. The numbers (#/#) correspond to the number of times a Set A marker is used over the number of obligatory contexts. Table 6. Armando (2;0–3;0). Development of Set A in verbal classes. Age

Transitives

Intransitives Actives

2;0 2;1 2;2 2;3 2;4 2;5 2;6 2;7 2;8 2;9 2;10 2;11 3;0

PFV

IPFV

SUBJV

1/5 2/11

0/7 2/11

0/4 2/3

1/4 0/2 9/10 3/6 3/8 4/4 3/5 2/2 11/11 16/16

0/6 3/8 7/12 13/17 9/11 15/20 12/14 18/19 18/27 25/27

0/1 2/3 5/8 1/2 8/12 14/16 12/14 19/27 20/21

NEG

IPFV

3/3 1/1 1/1 2/2 1/1 1/1 5/5 2/2 4/4 6/6 13/13

0/5 1/9 0/2 3/3 3/4 13/16 15/15 10/10 24/24 23/24 17/17 15/15 20/20

NEG

2/2 4/4 1/1

1/1

Inactives IPFV

0/2 0/8 1/3 3/3 1/1 1/1 4/7 2/2 3/3 7/8 3/4 8/11 9/9

NEG

1/1

Positionals IPFV

NEG

0/1 1/1

1/1 1/1 1/1

1/1 1/1

1/1 1/1

2/2

Productive use of Set A occurs in the transitive verbs class. In this context, Set A pronouns are used in the perfective, imperfective, subjunctive and negative. Among the intransitive classes, use of Set A was systematic in the actives, variable in the inactives and sporadic in the positionals; inchoatives did not occur in the recordings used for this study. Set A was not used during the first four months (i.e. 2;0 to 2;4), when many of the produced verbs consisted of the root and status inflection. Compared to the adult model these verbal bases in acquisition lack Set A markers and the aspect and mode auxiliaries. From 2;5 onward Set A was used systematically in both transitive and active intransitive verbs. With intransitives, Set A was most consistently used in the active. From the first recorded instances Set A markers were used with aspect and mode auxiliaries. Table 7 shows the Set B markers in verbal classes quantified by types and expressions; the use of verbs in third person singular context is shown in parenthesis (-0/).

76 Carlos Carrillo Carreón Table 7. Armando (2;0–3;0). Development of Set B in verbal classes. Age

Transitives

Intransitives Actives

2;0 2;1 2;2 2;3 2;4 2;5 2;6 2;7 2;8 2;9 2;10 2;11 3;0

PFV

IPFV

SUBJV

(5/5) (11/11)

(7/7) 4/4 (7/7)

3/3 (1/1) 1/1 (2/2)

(4/4) (2/2) 1/1 (9/9) (6/6) (8/8) (4/4) (5/5) (2/2) (11/11) 1/1 (15/15)

NEG

SUBJV

Inactives PFV

6/6 (1/1) 12/12 (1/1) (3/3) 2/2 (6/6) (1/1) (1/1) 5/5 (1/1) 2/2 (6/6) (1/1) (1/1) 11/11 (1/1) 2/2 (10/10) (3/3) (2/2) 6/6 (2/2) (18/18) (8/8) (1/1) 7/7 (2/2) 1/1(10/10) (2/2) (1/1) 9/9 5/5(15/15) (12/13) (5/5) 8/8 (14/14) 1/1 (16/16) (2/2) 2/2 (1/1) 2/2 (2/2) (19/19) 1/1 (13/13) (4/4) 2/2 3/3 (1/1) (27/27) 1/1 (26/26) 1/1 (5/5) (1/1) 5/5 (3/3) 1/1 (27/27) (21/21) (13/13) (3/3) 8/8 (7/7)

SUBJV

4/4 (5/5) 4/4 (1/1)

2/2 (5/5) 12/12 3/3 (4/4) (7/7) 1/1 (1/1) (2/2) 5/5 4/4 (7/7) 1/1 (7/7)

Productive use of overt marked Set B markers was found only in inactive intransitive verbs (Table 7), in which these forms occur in perfective and subjunctive. They occurred sporadically in transitive verbs and no examples were obtained in active intransitives. Transitive verbs are frequent in third person singular (–Ø). Up to this point the use of cross-reference markers between 2;0 and 3;0 is characterized by use of Set A in transitive and intransitive active verbs and by use of Set B in inactive intransitives in perfective and subjunctive. Development of verbal argument marking in these verbal classes, however, is only valid for a small group of verbs. Armando used cross-reference markers in 160 different verbal roots, but in this verb group person markers can be seen to develop in relation to the agent, experiencer and patient roles. These verbs are: Transitives

il-ik ‘see’, taas-ik ‘bring’, ¢?ah-ik ‘give’, č?a?-ik ‘give’, haant-ik ‘eat’, pul-ik ‘throw’

Intransitives bin-Ø ‘go’, taal-Ø ‘come’, lúub-ul ‘fall’, wen-el ‘sleep’, t?aan-Ø ‘speak’

Early acquisition of the Split Intransitive System in Yukatek

77

Examples of the morphological development of these verbs are included in the appendix (Tables 8, 9, 10 and 11). For the intransitive verbs, frequent and productive development occurred almost exclusively in the verbs bin (Table 9), taal (Table 10) and lúub-ul (Table 11). In adult language the verb lúub-ul is inactive intransitive and corresponds to state change. Bin and taal are intransitive class exceptions, inflecting as inactive intransitives in perfective and as active intransitives in imperfective. In the input, cross-reference markers were frequent in 15 verbs among 109 different verb roots, and these 15 also had the same frequent lexical entries in acquisition. Figures 2 and 3 allow comparison of the development of cross-reference markers in relation to the agent, experiencer and patient roles in acquisition versus input via verbs with frequent verbal argument marking. Set B refers only to overt marking forms.

Figure 2. Cross-reference markers in the input

78 Carlos Carrillo Carreón

Figure 3. Armando’s acquisition of cross-reference markers

In the input, Set A is used more frequently as agent than as experiencer, whereas in acquisition it fluctuates between development of agent and experiencer. Development of Set A use apparently begins at 2;6 and enters the most active stage at 2;8. From 2;8 to 2;10 the profile of Set A roles still differs from that of the input since the experiencer role is more developed. It should also be taken into account that fewer intransitive than transitive verbs occur in the recordings for the sampled age range. By 2;11 the agent and experiencer roles in the acquisition match that of the input. Use of Set B in the acquisition and input was essentially equal, although it was very occasionally used as object in the acquisition, but not in the input. Set B was used as experiencer in both the input and acquisition, but some examples of its use as patient did occur in the acquisition. 5. Conclusions These data demonstrate the role of frequency and morphological transparency in Yukatek language acquisition between the ages of 2;0 and 3;0. Frequency is related to the lexicon in which morphology develops. The verbs in which cross-reference markers were used frequently and produc-

Early acquisition of the Split Intransitive System in Yukatek

79

tively corresponded to the frequent verbs in the input. This, however, does not explain why Set B was more frequently and productively used than Set A during the first four months of observation. In this case, morphological development probably cannot be explained by frequency of use since Set A was in continual use in the input, but almost absent in the acquisition. The morphological transparency (Dressler 1985, 1999) of the Set B suffixes in perfective aspect inactive intransitive verbs may explain why these forms have an early use in acquisition. Set A may have occurred later than Set B because of its lack of morphological transparency, a result of its amalgamating with aspect auxiliaries. The relationship between the role of the verbal subject and its marking in accordance with verbal classes (Bohnemeyer 2004) apparently supports development of split intransitive system acquisition strategies. Children begin using Set B markers with inactive intransitive verbs in perfective aspect, and Set A markers enter into use months later with transitive verbs and active intransitive verbs in imperfective aspect. This delay may originate in the fact that the experiencer role is linked to Set B and the agent role to Set A, considering that the subject of active intransitive verbs in the imperfective does not play exactly the same experiencer role as in the inactives. The few examples of Set A use in inactive intransitive verbs in the imperfective aspect indicate that the experiencer marked as agent (split intransitive system) is still developing. Armando will have acquired the split intransitive system when, from 3;0 and onward, he develops use of Set A with inactive intransitive verbs in imperfective aspect and has used Set B with active intransitive verbs in the perfective aspect; this will also require acquisition of the derivative morphology of the antipassive voice. In summary, the data demonstrate that in acquisition of Yukatek Set A markers develop in both transitive and intransitive verbs through partial matching of the split intransitive system beginning at 2;6. There is a difference, however, in cross-reference markers frequency and use among the intransitive classes, even when all of them can have the subject expressed by Set A or B. In inactive verbs only Set B is frequent and productive, instead of the active verbs in which only Set A is present. Therefore, the difference between experiencer and agent is clearly marked at an early age, which is to be expected in acquisition of an ergative language.

80 Carlos Carrillo Carreón Notes 1. I am grateful to Barbara Pfeiler for all the facilities with which she provided me in order to use her database in Child Maya. I am also very thankful for all the comments and help to Barbara Pfeiler and Clifton Pye on this paper. 2. Variant of the first person plural used in the Valladolid region, where these data were collected. 3. According to Bricker (1998), the suffix -ih corresponds to aspect marking, whereas Bohnemeyer (1998) states that it corresponds to the third person singular. Carrillo Carreón (2005) included -ih in his analysis of cross-reference markers acquisition to determine if this form is used to mark person during acquisition. Research is still needed, however, on acquisition of -ih that specifically compares aspect and person development. If children use it as person marking, errors can be expected in transitive verb object utterances, for example, *k-a= mach-ik-ih ‘*you grasp it it’. If it is used as aspect marking, errors like *lúub-ih-en ‘*I did fell’ would be observed. Errors of this type were not observed in the data analyzed here or in other relevant analyses (i.e. Pfeiler 2003). 4. Kaufman (1990: 71). 5. Immediate imitations of the input, self repetitions, frozen forms and forms incoherent with the context are excluded.

References Andrade, Manuel 1955 A Grammar of Modern Yucatec. Microfilm Collection of Manuscripts on Middle American Cultural Anthropology. No. 41. Chicago: University of Chicago Library. Ayres, Glenn & Barbara Pfeiler 1997 Los verbos mayas. La conjugación en el maya yucateco moderno. Mérida: Ediciones de la Universidad Autónoma de Yucatán. Bohnemeyer, Jürgen 1998 Time relations in discourse. Evidence from a comparative approach to Yukatek Maya. Nijmegen: Eigenverlag. 2002 The grammar of time reference in Yukatek Maya. LINCOM Studies in Native American Linguistics 44. Alemania: LINCOM. 2004 Split intransitivity, linking, and lexical representation: the case of Yukatek Maya. Linguistics 42 (1): 67–107. Bricker, Victoria, Eleuterio Poot Yah & Ofelia Dzul de Poot 1998 A Dictionary of The Maya Language As Spoken in Hocabá, Yucatán. Salt Lake City: The University of UTAH Press. Brown, Penelope, Lourdes de León, Barbara Pfeiler & Clifton Pye 2002 The acquisition of agreement in Maya. Symposium presented at 9th IASCL Conference, Madison, WI, 16–21 July, 2002.

Early acquisition of the Split Intransitive System in Yukatek 2004

81

Mayan language acquisition studies. Symposium presented at the 78th Annual Meeting/Linguistic Society of America/Society for the Study of the Indigenous Languages of the Americas; Boston, January 8 –11, 2004. Carrillo Carreón, Carlos 2005 La adquisición de los pronombres de referencia cruzada en el maya yucateco. Estudio de caso. Tesis de maestría, Universidad Autónoma de Yucatán. Dixon, Robert M. W. 1994 Ergativity. Cambrige: Cambridge University Press. Dressler, Wolfgang 1985 Introducción a la Morfología Natural. Núcleo 2: 2–19. 1999 What is Natural in Natural Morphology (NM)? In Travaux du Cercle Linguistique de Prague, Vol. 3, Eva Hajicová (ed.), 135–144. Amsterdam /Philadelphia: John Benjamins. 2001 The Emergence of Morphology. A Constructivist Approach. Croatian Review of Rehabilitation Research 37 (1): 23–36. Dressler, Wolfgang U. & Lavinia Merlini Barbaresi 1994 Morphopragmatics. Berlin / New York: Mouton de Gruyter. Kaufman, Terrence 1990 Algunos rasgos estructurales de los idiomas mayances con referencia especial al k’iche’. In Lecturas sobre la Lingüística Maya I, Nora England & Stephen Elliot (eds.), 115–127. Guatemala: Centro de Investigaciones Regionales de Mesoamérica. Krämer, Martin & Dieter Wunderlich 1999 Transitivity alternations in Yucatec, and the correlation between aspect and argument roles. Linguistics 37: 431–480. Pfeiler, Barbara 2003 Early acquisition of the verbal complex in Yucatec Maya. In Development of Verb Inflections on First Language Acquisition. A Cross-Linguistic Perspective, Dagmar Bittner, Wolfgang U. Dressler & Marianne Kilani-Schoch (eds.), 379–399. Berlin /New York: Mouton de Gruyter. 2005 Polyvalence in the acquisition of early lexicon in Yucatec Maya. In Lexical Categories and Root Classes in Amerindian languages, Ximena Lois & Valentina Vapnarsky (eds), 319–341. Frankfurt: Peter Lang. Pfeiler, Barbara & Enrique Martín Briceño 1997 Early verb inflection in Yucatec Maya. Papers and Studies in Contrastive Linguistics 33: 117–125. Pustet, Regina 1992 Das Partizipationssystem des Yukatekischen: Nominativ-akkusativisch, ergativisch oder aktivisch? Zeitschrift für Phonetik, Sprachwissenschaft und Kommunikationsforschung 45 (2): 189–199.

82 Carlos Carrillo Carreón Tomasello, Michael 1992 First verbs: A case study of early grammatical development. Cambridge: Cambridge University Press. Tomasello, Michael 2000 First steps toward a usage-based theory of language acquisition. Cognitive Linguistics 11 (1/2): 61–82.

Appendix Table 8. Development of the verb il ‘to see’ Age 2;0 2;1 2;3 2;6 2;7 2;8

Auxiliaries

t- (PFV) ¢?o?ok (TERM) k- (IPFV) táan (DUR) ka?h (SUBJV)

2;11 3;0

Set A

Status

Set B

w=(1S)

-ik (INCP) -ah (COMP)

-eč (2S)

inw= (1S) aw- (2S) *(u-!w-)

-eh (SUBJV) -o?n (1P)

-o?b (3P) uy=(3S)

Table 9. Development of the verb bin ‘to go’ Age

Auxiliaries

Set A

2;1 2;3 2;4 2;5

2;6 2;8 2;11 3;0

Set B -eč (2S) -ih (3S)

he?l (ASEG) táan (DUR) k- (IPFV) táak (DESID) ¢?o?k (TERM)

in= (1S) !u= (3S) U= (3S)

-e?š (2P) -o?n (1P) -en (1S)

yan (OBLIG) A= (2S)

Early acquisition of the Split Intransitive System in Yukatek Table 10. Development of the verb taal ‘to come’ Age 2;0 2;1 2;4 2;5 2;6 2;7 2;8 2;9 2;11

Auxiliaries

k- (IPFV)

Set A

Status

*u=…-o?b (3P) in= (1S) u=…-o?b (3P) a= (2S) u= (3S)

Set B -en (1S)

he?l (ASEG) táan (DUR) -ih (3S) ka?h (SUBJV) ¢?o?k (TERM) yan (OBLIG)

-ak (SUBJV)

Table 11. Development of the verb lúub ‘to fall’ Age

Auxiliaries

Set A

Status

2;0

-en (1S) -eč (2S) -o?ob (3P) -ih (2S)

2;1 -uk (SUBJV) -ul (INCP)

2;5 2;7

k- (IPFV) san (SUBJV)

2;9 2;10 2;11 3;0

Set B

*(u … -!e?š) in= (1S)

ken (SUBJV) táan (DUR) óolak (SUBJV)

83

A preliminary view at Ch’ol (Mayan) early lexicon: The role of language and cultural context Lourdes de León

1. Introduction In recent years, I have been exploring the early lexical development of children learning Tzotzil Mayan. I have shown that linguistic and cultural factors influence the early preference and morphological development of verbs over nouns (de León 1999a, 2001a). The present paper is inspired, on the one hand, by my findings in Tzotzil acquisition and, on the other, by a typological and anthropological interest in comparing language acquisition in Mayan languages. I present here a preliminary view of the early vocabulary of a Ch’ol learner at the age of 1; 10,1 at the one word and early combination utterance period. In consistence to findings presented for Tzotzil, this young boy shows a striking preference of verbs over nouns, preference that resonates with features of (i) the structure of the native language, (ii) the adult input, and of (iii) the cultural context of acquisition. In previous studies I have explained Tzotzil early lexical development by a convergence of linguistic and contextual factors. At the linguistic level, I have argued that cross referenced arguments and a high deletion of nominal arguments in Tzotzil allows the verb to be salient in the utterance. At the cultural level, children are immersed in everyday activities where talk is about what is being done, rather than about objects per se. Here we talk about a peasant culture where toys and books are secondary in children’s socialization. The Ch’ol boy of my study is exposed to a similar learning situation to that of the Tzotzil learners, and shows a similar learning pattern of nouns and verbs in his early lexicon. I will start by giving a theoretical background, then I will give a general description of the Ch’ol verb complex. Finally I will present the production data of the boy of the study and the input data of his caregivers. I will show that findings from this Ch’ol child show many similarities to findings for early Tzotzil acquisition.

86 Lourdes de León 2. Theoretical background Studies in the acquisition of Chinese (Tardif 1996; Tardif, Shatz & Naigles 1997; Tardif, Gelman & Xu 1999), Huichol (Gómez, this vol.), Korean (Choi, 1998), Tzotzil Mayan (de León 1998a, 1999a, 1999b, 2001a), Tzeltal Mayan (Brown 1998a, 1998b), Otomí (Martinez-Casas & Alcaraz, 2003), Yukatek and Croatian (Kovačević, Pfeiler & Palmović 2002), among others, have shown that language-based explanations, namely factors of input (e.g., frequency, CDS, discourse patterns), language-structure, and socializing factors play a central role in the early preference for verbs over nouns, in the acquisition of these languages. Research done with English speaking children has also shown that kinds of activities in which children are participating can also have an effect into shaping their lexical production and comprehension (Goldfield 1993, 1998). In this paper I argue that properties of the native language, together with contextual factors orient the Ch’ol child of the study into using verbs more predominantly than nouns before his second year of life. I explore the following issues: Properties of the input: Studies of early lexical development have claimed that there is a positive relation between the input and the composition of early vocabularies (Goldfield 1993, 1998; Tardif et al. 1997; Pine, et al. 1997; Choi & Gopnik 1995; Choi 1998). Along this line, I argue that Ch’ol’s structural and discourse factors orient the child of the study into using more verbs than nouns. Types of interaction in which children participate: Some scholars note that the context and the cultural settings in which children participate can have an effect in shaping early vocabularies (Martinez Casas & Alcaraz 2003; Choi 1998; Fernald & Morikawa 1993; Goldfield 1993, 1998; Gelman & Tardif 1998). It has been argued that nouns predominate crosslinguistically in early acquisition due to a cognitive predisposition in preverbal children to name discrete, perceptually bounded entities. This approach leaves event naming as a supposedly later acquisition task because of its ‘language-specific nature’ (Gentner 1982; Gentner & Boroditsky 2001; see Tomasello 1992 and Tomasello & Merriman 1995: 1–18 for discussion). This proposal is consistent with theories that argue that children assume that early words refer to whole objects (Markman 1989). These scholars have claimed that, despite variation from language to language, the noun bias remains as a language-

Ch’ol (Mayan) early lexicon: The role of language and cultural context

87

independent fact, with probable cognitive bases. The study presented here does not intend to disclaim the cognitive argument about noun priority since I am not looking at the young boy’s earliest productions. However, it shows that at the early time of lexical development early productions are consistent with structural and contextual factors. I have argued elsewhere that these specific factors orient Tzotzil Mayan children in their very earliest preference of verbs over nouns, it remains to be shown whether this is the case for Ch’ol children’s at their most earliest productions. However, given the child’s age, and consistency of factors with the Tzotzil case study, I would dare to predict that this is the case of Ch’ol learners too. 3. The study: subjects and data Ch’ol is a Mayan language with about 140,000 speakers. Research for this study was carried out in the hamlet of Belisario Dominguez, Salto de Agua, Chiapas, México, a village with some 1200 inhabitants. In this community, Ch’ol is spoken by most of the population, and it is basically learned as a first language in monolingual homes. Other languages like Spanish or Tzeltal are spoken in some homes. The data come from longitudinal linguistic and ethnographic research based on a pilot case study of one boy raised in a Ch’ol speaking home from the end of his one-morpheme utterance period (MLU 1) and at the onset of his early morpheme combination period (roughly MLU 2.75). The boy, Tobias, is the third child of the family. Production data for this paper cover four sessions at the age of 1;10. Input data come from 100 utterances obtained in one session. The boy was audio and video recorded biweekly by his main caregiver in natural interaction with members of the family. Complementary data was obtained by the researcher in periodic visits. Transcriptions were made by the main caregiver, and checked for reliability by the researcher with the child’s father. Tobias’ main caregiver was his aunt. She received training from a linguist native speaker to transcribe the language.2 4. Coding In this study I describe Tobias’ word classes with the terms ‘verb’ and ‘noun’ in alternation to ‘object-name’ and ‘event-name’ only if they approximate the adults’ forms, and if they also refer to objects and events respectively.

88 Lourdes de León My main guideline is not only semantic but also morpho-syntactic criteria in coding the child’s grammatical categories (Gentner 1982; Choi & Gopnik 1995; Tardif 1996). At a semantic level, nouns are identified if they refer to an object. Proper nouns and kin terms are counted separately because they were used not referentially but pragmatically, for requests, as vocatives, etc. Gentner has recently argued (Gentner & Boroditsky 2001) that proper names and kin terms should be included as object names. However, such terms refer not to classes of objects, but to specific individuals. Studies on English lexical acquisition exclude kinship terms from their noun measures for the same reason (Lieven, Pine & Dresner Barnes 1992; Bloom, Tinker & Margulis 1993). Although studies include onomatopoeic nouns if recognized as CDS vocabulary, this work does not include them because the child of the study did not produce them. Demonstrative pronouns were excluded. Verbs were identified if they referred to actions or activities. At the one word-period both nouns and verbs were identified in terms of their approximation to conventional adult forms. At the time of early combinations, verbs are distinguished from nouns on the basis of morphosyntactic criteria. 5.

The language facts: the verb in Ch’ol

From the perspective of learning theory, the linguistic facts of the Ch’ol language present the young Ch’ol learner with a verb-friendly language. Here I think of morphologic and discursive factors that highlight the verb in complementary ways. 5.1. Structural facts Ch’ol is a morphologically split ergative language. It behaves like a nominative-accusative language in imperfective, progressive, exhortative, and in some subordinate constructions (Vázquez Alvarez 2001: 19). In the imperfective, the subject of intransitive verbs, is marked with the ergative pronoun. (1)

mi

k-wäy-el

IMPFV A1-sleep-SEII

‘I sleep’3

(Vázquez Alvarez 2001: 16)

Ch’ol (Mayan) early lexicon: The role of language and cultural context

89

In the perfective, the subject of intransitive verbs is marked with the absolutive: (2)

tyi

wäy-i-y-oñ sleep-VTI-EPN-B1 ‘I slept’

(Vázquez Alvarez 2001: 16)

PERFV

The agent of the transitive verb is marked with the ergative pronoun (Set A) and the patient is marked with the absolutive (Set B). (3)

tyi

i-mek’-e-y-ety

(Vázquez Alvarez 2001: 16)

PERFV A3-hug-VTT-EPN-B2

‘He hugged you’ Ch’ol has basic VOS word order. Verb stems are inflected for aspect and person. As is the case for other Mayan languages, Ch’ol is a head marking language. Lexical arguments are crossreferenced in the verb through pronominal affixes. This fact allows for the non obligatory mention of the lexical argument in the sentence. Thus, a verb can often stand alone as a wellformed sentence if its arguments can be established from context. Examples (4) and (5) have the expressed arguments, (‘woman’ and ‘child’). In examples (4a), and (5a) these arguments are omitted but remain cross-referenced in the verb phrase by absolutive and ergative affixes. (4)

a. tyi

i-mek’-e-ø

aläl x-ixik boy CL-woman ‘The woman hugged the child.’ (Vázquez Alvarez 2001: 16) PERFV A3-hug-VTT-EPN-B3

b. tyi

i-mek’-e-ø

PERFV A3-hug-VTT-SUF-B3

‘She hugged him’ (5)

a. tyi

y-il-ä-ø

aläl x-ixik boy CL-woman ‘The woman saw the child’ PERFV A3-see-SUF-B3

b. tyi

y-il-ä-ø

PERFV A3-see-SUF-B3

‘She saw him’

90 Lourdes de León 5.2. Discourse facts My preliminary review of Ch’ol texts, based on a sample of 100 utterances shows that Ch’ol conforms to the ‘preferred argument structure’ (PAS) pattern found in many languages (Clancy 1980, 2003), among them the Mayan languages (DuBois 1985, 1987; Givon 1983; Robinson 1996). Under PAS in a wide range of discourse contexts at least one argument per clause is overtly expressed. The overt argument is typically the object of a transitive verb or the subject of an intransitive verb. Overt arguments have high discourse prominence by contrast to deleted arguments, which encode given/ old information. An expected consequence of the high rate of argument elision is a relatively high frequency of verbs over nouns. In sum, morphological facts (Ch’ol as a head-marking language) together with discursive facts (Ch’ol with PAS pattern) favor a higher use of verbs over nouns in the input children are hearing. Our data confirm this tendency, as shown in Figure 1.

Figure 1. Nouns and verbs in CDS in 100 utterances

Figure 1 shows the rate of verbs and nouns (types and tokens) produced by three caregivers that participated in the interactions with the child (mother, sister, and aunt). Overall, verbs types and tokens are more frequent than noun types and tokens. They occur in a ratio of over 2 to 1. In section 5 I will compare this data with the production data of the child of the study.

Ch’ol (Mayan) early lexicon: The role of language and cultural context

91

6. Tobias’ noun/verb production at 1;10 Production data from Tobias at 1;10 shows that he produced verb types and tokens in a ratio over 2 to 1 in a very similar fashion to that of his caregivers (Figure 2).

Figure 2. Ch’ol child noun /verb production in 100 utterances

This data is consistent with the profile shown in CDS as shown in Figure 3 (next page). Here the ratio of verb type and verb token is over 2 to 1 both in child and caregivers as shown in table 1. Table 1. Noun /verb type/token ratio in caregiver and child Noun/verb type ratio Child Caregiver

2.4 2.2

Noun/verb token ratio 2 2.3

The production data is thus highly consistent with the input data. Before age two this Ch’ol boy is behaving in a very similar way to the pattern of verb/nominal production in the language he is learning. This is also the case for the Tzotzil children of my study.

92 Lourdes de León

Figure 3. Noun /verb type and token produced by caregivers and child

7. Verbal interaction with children and the nature of Tzotzil and Ch’ol CDS Elsewhere, I explored the factors that lead Tzotzil learners to prefer verbs over nouns in their early lexical development. I argued that linguistic factors and the cultural context of language learning play a role too. In this paper I am not drawing correlations in a controlled manner between context, input, structure and production as Tardif et al. (1999) did in her work about early lexical development in Mandarin. In my study about CDS and language socialization in Tzotzil learners, I argued that these children are socialized to participate in activities defined by such different interactional goals as: (i) attending to events (as a result of both verbal and nonverbal cues), (ii) prompting behavioral response (through requests, demands, warnings), (iii) discontinuing behavior potentially dangerous or harmful (through directives formed mainly with imperatives), (iv) participating in formal interactions (de León 1998b, 1999a: 108–109). The study I present here shows also that the interactions in which Tobias is participating are basically activities. These activities are very consistent

Ch’ol (Mayan) early lexicon: The role of language and cultural context

93

with activities from (i) to (iii) in the Zinacantec context. I have not found a comparable situation of that of (iv) where children are immersed as participants in formal interactions. There is a lot of ethnographic work to be done in the Ch’ol communities about discourse genres, registers, etc. My own experience in the Ch’ol hamlet of my study, is that children are not socialized explicitely to participate in formal interactions of the Zinacantec kind. In my case study, I have observed a more distant relationship between grandparents and children, strikingly different from the one I have documented in Zinacantan. However, I remind the reader that the study presented here is a case study of an extended family. Table 2 shows the rate of speech act types and their corresponding sentence type in Ch’ol CDS. Directives are by far the most frequent ones. These are expressed with imperatives and are mainly to control or prompt behavior (65%). 4 They consist of a verb nucleus without expressed lexical arguments. The utterance form thus combines with the interactive situation to provide this child with cues about the mapping of verbs into actions. Table 2. Frequency of speech act types in Ch’ol CDS in 100 utterances Speech Act type

Sentence type

Threat/Warning

Declarative

Request Directive

Tzotzil example

kläptyek’ety ka ayajlel Interrogative chuki awom? Positive Imperative ok'en Negative Imperative mach atyäk’lan

Total

Translation

%

I am going to kick you 8% You are going to fall down What do you want? 18 % Get out! 65 % Don’t bother him 100 %

(Child age: 1;10 months old)

Table 2 also shows the kind of learning context Tobias is immersed in. The sample is obtained from an everyday normal interaction. In a one hour segment, we find a collection of events that involve playing, fighting, eating, changing clothes. In one event Tobias is playing in the yard with his siblings. He is fighting for a hammock (that is when his sister threatens to “kick him”- kläptyek’ety 'I am going to kick you'), he gets into the hammock but is about to fall down so his aunt warns (ka ayajlel ‘you are going to fall down’), his sister yells again lok’en! ‘get out!,’ the aunt intervenes and tells her mach atyäk’lan ‘don’t bother him!’ The hammock event contains basically no nominals since all the action is focused on getting in and out of it, and there is no mention of the referent since it is pragmatically

94 Lourdes de León available. Other events as playing or eating introduce some new mentioned nominals (peanuts, tortilla, ball), but consist mainly on verbs. The sample together with ethnographic work done with the family shows that the child is participating in activities referred to by action words. A bulk of these activities are monitoried or prompted by the caregiver as the speech act types indicate in Table 2. No specific focus on objects per se is observed. The learning situation is thus consistent with the CDS: diverse activities referred to by a high rate of verbs, many of them imperatives. The rate of nouns/verbs has therefore a relation with the kinds of activities children are taking part of. This has been reported even in other studies for English which look at the pragmatics of the interactions children participate. Goldfield examined communicative contexts between mothers and children and overall she found a higher rate of nouns in speech acts that describe the child’s behavior, and a higher rate of verbs in speech acts that prompt the child to behave in a certain way. She concludes that in the latter contexts verb comprehension is favored over production (Goldfield 1998). The English learner thus hears both nouns and verbs in different activities that privilege one or the other word class in association with verbal or behavioral action. Given the structural and discursive facts, the Tzotzil and the Ch’ol learner hears an input with a higher rate of verbs across all contexts, and almost always inscribed in activities that prompt, monitor or orient behavior; they are likewise focused on what to do and not on what to say. Both in the Tzotzil and Ch’ol culture children are socialized in a what to do modality. Conclusions There has been much debate about the so-called noun-bias in the last decade. Tardif et al. (1999) have shown that there is an effect in early lexical acquisition in English and Mandarin depending on the languages, individuals, situations, measurement methods, acquisition ages, and in age of onset. In a laboratory study, she explored the input of Mandarin and English speaking mothers engaging in different activities: book reading, regular toy play, and mechanical toy play (Tardif et al. 1999). The case study I present here is based on natural spontaneous production data from a Ch’ol child, and draws relations with the learning situation in Tzotzil. In both cases we find a combination of factors leading the children of the study to use more verbs than nouns. In these studies both structural and contextual factors interact to produce the verb bias. Tzeltal (Brown 1998a, 1998b) and Yukatek early

Ch’ol (Mayan) early lexicon: The role of language and cultural context

95

lexical acquisition point into the same direction (Kovačević, Pfeiler & Palmović 2002). At a typological level we could argue that these Mayan languages share the features of being head-marking, have a preferred argument structure, and the learning situation of these children are very similar. They are socialized in an agricultural small-scale culture where they are mostly participating in everyday activities. In an earlier paper, I have also argued that semantic specificity may play a role into facilitating the use of more verbs in Tzotzil early language. I especially focused on spatial verbs which correspond to prepositions in English (de León 2001b). Brown has argued that in Tzeltal semantic specificity also provides a packet where object reference is implied by classificatory verbs (1998b). Tardif has recently argued in favor of semantic specificity in Mandarin early verbs (Tardif 2006), and relates this factor to the higher use of verbs over nouns in Mandarin acquisition. In the present study I have not included the semantic specificity argument since more investigation needs to be done about Ch’ol verb semantics. As other Mayan languages, this language has specific verbs for eating, cutting, or holding, among others as the Tzeltalan languages (Berlin 1967). However, Ch’ol also has a large set of auxiliaries or light verbs which are used frequently (Vázquez Alvarez 2001). In fact, Tobias' earliest verb production contains a good number of light verbs over specific verbs, in contrast to the Tzotzil children of my study. It would be interesting to compare the Ch’ol data with the Yukatek data given the fact that these languages are closer to each other. This study remains to be done. In sum, from a typological perspective structural and discursive factors play a role in favouring verbs over nouns in learning Ch’ol and Tzotzil. Similar findings have been presented for Tzeltal (Brown 1998a, 1998b) and Yukatek (Kovačević, Pfeiler & Palmović 2002). These factors converge with the cultural contexts of language learning where activities and reference to actions are at the core of language socialization. Notes 1. Age is given in years and months (year;month). 2. I thank my colleague Mtro. Domingo Meneses for training Rosa S. M. to write Ch’ol. I also thank him for collaborating as a consultant in the project. I acknowledge the help and enthusiastic collaboration of Tobias’ father (whose identity will remain anonymous) for collecting, transcribing and helping with

96 Lourdes de León data analysis. The main subject’s name and his family identity is changed for reasons of confidentiality. 3. Examples 1–3 come from Vázquez Alvarez (2001). I keep the same conventions used by him in his glosses. Examples 4–5 come from L. de León's elicited data, but have the same glosses used by the cited author in order to help the reader identify the morphemes present from 1 to 5. Conventions used are presented in an appendix at the end of the article. 4. Martinez-Casas and Alcaraz (2003) found that a large number of utterances directed to Otomi children have imperatives. They argue that they correspond to the conative and phatic functions, based on a Jakobsonian model. My data show that a considerable number of utterances directed to Tobias’ are in fact, imperatives. This feature of child directed speech has been found also as a feature of speech addressed to children in other studies.

References Berlin, Brent 1967 Categories of eating in Tzeltal and Navajo. International Journal of American Linguistics 33 (1): 1–6. Berman, Ruth A. & Dan I. Slobin 1994 Relating events in narrative: A crosslinguistic developmental study. Hillsdale, NJ: Lawrence Erlbaum. Bloom, Lois, Erin Tinker & Cheryl Margulis 1993 The words that children learn: Evidence against a noun bias in early vocabularies. Cognitive Development 8: 431–450. Bowerman, Melissa 1996 The origins of children’s spatial semantic categories: Cognitive versus linguistic determinants. In Rethinking linguistic relativity, John Gumperz & Stephen Levinson (eds.). Cambridge, MA: Cambridge University Press. Bowerman, Melissa, Lourdes de León & Soonja Choi 1995 Verbs, particles, and spatial semantics: Learning to talk about spatial actions in typologically different languages. In The Proceedings of the 27th Annual Child Language Research Forum, Eve V. Clark (ed.), 101–111. Stanford, CA: CSLI/University of Chicago Press. Brown, Penelope 1997 Isolating the CVC root in Tzeltal Mayan: A study of children’s first verbs. In The Proceedings of the 28th Annual Child Language Research Forum, Eve V. Clark (ed.), 41–52. Stanford, CA: CSLI /University of Chicago Press.

Ch’ol (Mayan) early lexicon: The role of language and cultural context 1998a

1998b

97

Early Tzeltal verbs: Argument structure and argument representation. In Proceedings of the 29th Annual Child Language Research Forum, Eve V. Clark (ed.), 129–140. Stanford, CA: CSLI /University of Chicago Press. Children's first verbs in Tzeltal: Evidence for an early verb category. In Elena Lieven (ed.), Special edition of Linguistics 36 (4): 713–753.

Choi, Soonja 1998 Caregiver input and the acquisition of nouns and verbs in English and Korean. Paper delivered at the 15th Biennial Meetings of the International Society for the Study of Behavioral Development, July 1– 4, 1998, Bern, Switzerland. 1998 Verbs in early lexical and syntactic development in Korean. Linguistics 36: 755–780. 1999 Early development of verb structures and caregiver input in Korean: Two case studies. International Journal of Bilingualism 3 (2/3): 241– 265. Choi, Soonja & Melissa Bowerman 1991 Learning to express motion events in English and Korean: The influence of language specific lexicalization patterns. Cognition 4: 83–121. Choi, Soonja & Alison Gopnik 1995 Early acquisition of verbs in Korean: A cross-linguistic study. Journal of Child Language 22 (3): 497–531. Clancy, Patricia M. 1980 Referential choice in English and Japanese narrative discourse. In The Pear Stories: Cognitive, cultural, and linguistic aspects of narrative production, Wallace Chafe (ed.), 127–202. Norwood, NJ: Ablex. 2003 The lexicon in interaction: developmental origins of preferred argument structure in Korean. In Preferred argument structure: grammar as architecture for function, John W. Du Bois, Lorraine Kumpf & William Ashby (eds.), 81–108. Amsterdam: John Benjamins. Clark, Eve 1993 The lexicon in acquisition. Cambridge, MA: Cambridge University Press. de León, Lourdes 1997a Vertical path in Tzotzil (Mayan) acquisition: Cognitive vs. linguistic determinants. In The Proceedings of the 28th Annual Child Language Research Forum, Eve V. Clark (ed.), 183–197. Stanford: CSLI /University of Chicago Press. 1998a Raíces verbales tempranas en tzotzil: Input materno vs. restricciones cognoscitivas. Función 17–18: 147–174. 1998b The emergent participant: Interactive patterns of socialization of Tzotzil Mayan infants. Journal of Linguistic Anthropology 8(2): 1– 31.

98 Lourdes de León 1999a

Verb roots and caregiver speech in early Tzotzil (Mayan) acquisition. In Cognition, discourse and function, Barbara Fox, Dan Juravsky & Laura Michaelis (eds.), 99–119. Stanford: CSLI /University of Chicago Press. 1999b Verbs in Tzotzil early syntactic development. International Journal of Bilingualism 3 (2/3): 219–240. 2001a Why Tzotzil children prefer verbs over nouns: the role of linguistic, cognitive, and cultural factors. In Selected papers of the Proceedings of the 8th International Congress for the Study of Child Language, Margareta Almgren, Andoni Barreña, María-José Ezeizabarrena, Itziar Idiazabal & Brian Mac Whinney (eds.), 947–969. Cascadilla Press. CD Rom Edition. 2001b Finding the richest path: the role of language and cognition in the acquisition of verticality in Tzotzil (Mayan). In Language acquisition and conceptual development, Melissa Bowerman & Stephen Levinson (eds.), 564–565. Cambridge, MA: Cambridge University Press. Du Bois, John W. 1985 Competing motivations. In Iconicity in syntax, John Haiman (ed.), 343–365. Amsterdam: John Benjamins. 1987 The discourse basis of ergativity. Language 63: 805–855. Fernald, Anne & Haruko Morikawa 1993 Common themes and cultural variations in Japanese and American mothers’ speech to infants. Child Development 64: 637–656. Gallaway, Clare & Brian Richards 1994 Input and interaction in language acquisition. Cambridge University Press: Cambridge. Gelman, Susan & Twila Tardif 1998 Acquisition of nouns and verbs in Mandarin and English. In The Proceedings of the 29th Annual Child Language Research Forum, Eve V. Clark (ed.), 27–36. Stanford: CSLI/University of Chicago Press. Gentner, Dedre 1982 Why nouns are learnt before verbs: Linguistic relativity versus natural partitioning. In Language development, Vol. 2: Language, thought and culture, Stan A. Kuczaj (ed.), 301–304. Hillsdale, NJ: Erlbaum. Gentner, Dedre & Lisa Boroditsky 2001 Individuation, relativity, and early word learning. In Language acquisition and conceptual development, Melissa Bowerman & Stephen Levinson (eds.), 215–256. Cambridge, MA: Cambridge University Press. Givon, Talmy 1983 Topic continuity in discourse: A quantitative crosslanguage study. Amsterdam: John Benjamins.

Ch’ol (Mayan) early lexicon: The role of language and cultural context

99

Goldfield, Beverly 1993 Noun bias in maternal speech to one-year-olds. Journal of Child Language 20: 85–100. Goldfield, Beverly 1998 Why nouns before verbs? A view from pragmatics. In Annabel Greenhill, Mary Hughes, Heather Littlefield, Hugh Walsh (eds.), BUCLD 22: Proceedings of the 22nd Boston University Conference on Language development: 281–292. Gómez López, Paula this vol. Acquisition of referential and relational words in Huichol: From 16 to 24 Months of Age. Haviland, John B. 1994 ‘Ta xetel xulem’ [The buzzards were circling]: Categories of verbal roots in (Zinacantec) Tzotzil. Linguistics 32: 791–743. Imai, Mitsumi & Dedre Gentner 1997 A crosslinguistic study of early word meaning: Universal ontology and linguistic influence. Cognition 62: 169–200. Kovačević, Melita, Barbara Pfeiler & Marijan Palmović 2002 Red thread of Croatian and Yukatek Maya: What could be common in the two languages? Paper delivered at the International Congress for the Study of Child Language. Madison, Wisconsin, July 2002. Laughlin, Robert M. 1975 The great Tzotzil dictionary of San Lorenzo Zinacantán. Washington, DC: Smithsonian Institution Press. Lieven, Elena, Julian Pine & Barnes H. Dresner 1992 Individual differences in early vocabulary development: Redefining the referential-expressive distinction. Journal of Child Language 19: 287–310. Markman, Elena 1989 Categorization and naming in children: Problems of induction. Cambridge, MA: MIT Press. Martínez-Casas, Regina & Victor Manuel Alcaraz 2003 El desarrollo del español en migrantes indígenas: la adquisición del conflicto. In Introducción al estudio del español desde una perspectiva multidisciplinaria, Esmeralda Matute & Fernando Leal (eds.) 571–505. Guadalajara: Universidad de Guadalajara. Ochs, Elinor & Bambi Schieffelin 1984 Language acquisition and socialization: Three developmental stories and their implications. In Culture theory: Essays on mind, self, and emotion, Richard Schweder & Robert A. Levine (eds.), 276–320. Cambridge: Cambridge University Press. Pfeiler, Barbara 2002 Noun and verb acquisition in Yucatek Maya. In Pre- and protomorphology: early phases of morphological development in nouns and

100 Lourdes de León verbs, Maria Voeikova & Wolfgang U. Dressler (eds.), 75–83. München: Lincom Studies in Theoretical Linguistics. Pine, Julian, Elena Lieven & Caroline Rowland 1997 Stylistic variation at the ‘single-word’ stage: Relations between maternal speech characteristics and children’s vocabulary composition and usage. Child Development 68: 807–819. Pye, Clifton 1986 Quiche’ Mayan speech to children. Journal of Child Language 13: 85–100. 1992 The acquisition of K’iché. In The crosslinguistic study of language acquisition, Dan Isaac Slobin (ed.). Hillsdale, NJ: Erlbaum. Robinson, Stuart 1996 The -van antipassive in Tzotzil: A discourse perspective. Unpublished Honors Thesis. Port, Oregon: Reed College, Linguistics Department. Slobin, Dan I. 1973 Cognitive prerequisites for the development of grammar. In Studies of child language development, Charles A. Ferguson & Dan Isaac Slobin (eds.). New York: Holt, Rinehart & Winston. 2001 Form-function relations: How do children find out what they are? In Language acquisition and conceptual development, Melissa Bowerman & Stephen Levinson (eds.), 406–449. Cambridge, MA: Cambridge University Press. Snow, Catherine 1984 Parent-child interaction and the development of communicative ability. In The acquisition of communicative competence, Richard Schiefelbusch & Joanne Pickar (eds.), 69–107. Baltimore, MD: University Park Press. Sperber, Dan & Dedre Wilson 1996 Relevance. Cambridge, MA: Harvard University Press. Tardif, Twila 1996 Nouns are not always learned before verbs: Evidence from Mandarin speakers’ early vocabularies. Developmental Psychology 32: 492–504. 2006 But are they really verbs? Mandarin words for action. In Action meets word: How children learn verbs, K. Hirsh-Pasek & Roberta M. Golinkoff (eds.). Oxford: Oxford University Press. Tardif, Twila, Susan A. Gelman & Fan Xu 1999 Putting the ‘noun bias’ in context: A comparison of English and Mandarin. Child Development 70: 620–635. Tardif, Twila, Marylin Shatz & Letitia Naigles 1997 Caregiver speech and children’s use of nouns vs. verbs: a comparison of English, Italian, and Mandarin. Journal of Child Language 24: 535–565.

Ch’ol (Mayan) early lexicon: The role of language and cultural context

101

Tomasello, Michael 1986 Learning to use prepositions: a case study. Journal of Child Language 14: 79–98. 1992 First verbs: a case study of early grammatical development. Cambridge, MA: Cambridge University Press. 1995 Pragmatic contexts for early verb learning. In Beyond names for things: Young children’s acquisition of verbs, Michael Tomasello & Willliam Merriman (eds.). Hillsdale, NJ: Erlbaum. 1998 The return of constructions. Journal of Child Language 25: 431–442. 2001 Perceiving intentions and learning words in the second year of life. In Language acquisition and conceptual development, Melissa Bowerman & Stephen Levinson (eds.), 132–158. Cambridge, MA: Cambridge University Press. Tomasello, Michael & William Merriman 1995 Beyond names for things: young children’s acquisition of verbs. Hillsdale, NJ: Erlbaum. Vázquez Alvarez, Juan Jesús 2001 Morfología del verbo chol. Tesis de Maestría en Lingüística Indoamericana. México: CIESAS. Vogt, Evon Z. 1969 Zinacantán: A Maya community in the Highlands of Chiapas. Cambridge, MA: Harvard University Press.

Appendix Transcription conventions (Vázquez Alvarez 2001: 332) 1 2 3 A B CL IMPFV EPN SEII PERFV SUF VTI VTT

First person Second person Third person Set A (Ergative) Set B (Absolutive) Classifier imperfective epenthesis status suffix for intransitive verbs in imperfective perfective status suffix thematic vowel for intransitives thematic vowel for transitives

Acquisition of referential and relational words in Huichol: from 16 to 24 months of age Paula Gómez López

1. Introduction Based on the longitudinal data of one girl, I present the results of a study of lexicon acquisition in Huichol from 16 to 24 months of age. Wixarika or Huichol is a Yutoaztec language spoken in Mexico, mainly in the states of Jalisco, Nayarit and Zacatecas. It is calculated that there are around 40,000 speakers. Although it is one of the minor languages in the country as a hole, it possesses great vitality. In the Huichol communities, children are still learning their parents’ language, although the latter can also speak Spanish. Children learn Spanish when they are in school, or later on when they leave their communities. I am going to talk fundamentally about nouns and verbs. These categories comprise mainly the classes of referential words and relational words respectively. According to numerous pieces of work, especially in English and other Indo-European languages, children show a clear preference for the production of nouns compared to that of verbs in the first acquisition stages1. According to some linguistic experts2, the above said is a projection of the position of the investigators who find that beginning a study of the acquisition of a noun is easier than beginning one of verbs. For different reasons, the study of this word class was not taken into account in the investigation about the acquisition of English and other related languages. The extension of studies about the language acquisition to languages typologically different from English has stimulated the investigation about verb acquisition3. There are different reasons that explain why nouns are produced in greater quantity than verbs in the process of English acquisition. The most important thing is that the existence of a universal tendency in the process of lexicon acquisition is associated with such a phenomenon. However, investigations about other languages have proved that results obtained in English cannot be generalized, which questions the validity of the proposed

104 Paula Gómez López universal restrictions4. Among the arguments against this universality, the influence of the structural properties of the language and the maternal input have been mentioned. In a comparative study between English and Korean, Gopnik and Choi (1995) present a complex chart in which the linguistic differences and the input interact with cognitive factors. In the Tzotzil language, de León (1998: 150) found that the structural properties of the learnt language and the maternal interactive styles played an important role in vocabulary acquisition in early stages5. More recently, Gentner and Borodistky (2001: 233) point out that a series of features of the input in English could explain the advantage of the nouns in children’s production without the necessity of invoking cognitive factors. In this study, I do not present quantitative data about the maternal input. I am going to refer principally to the structural factor, i.e. the specific properties of the Huichol language. Considering that the multiple evidences in studies in languages different from English have shown the influence of typological properties of the language in the acquisition process, the results of this exploration in the Huichol language are expected to differ from those about English. The Huichol language has a series of properties that create the conditions that allow the use of predicative constructions in general and verbs in particular to be more frequent than the use of nouns and referential constructions. 2.

Properties that favor verbal production and predicative constructions

2.1. The order of the constituents in the simple sentence The order of the sentence constituents, SOV, sets the verbs in the final position, i.e. prominent. This facilitates its perception: (1)

Paapá ne-p-eu-kwai-miki tortilla 1SG-AS-GLB-to.eat-DES ‘I want to eat tortilla.’

2.2. Names can function as a nucleus of the predicate Huichol nouns, and other categories such as pronouns and adverbs, can occupy the position of the nucleus of the predicate without derivation.

Acquisition of referential and relational words in Huichol

105

On the other hand, verbal lexemes must be derived in order to function as nuclei of the nominal syntagm.6 Therefore, in Huichol, identifying the function of a verbal root will be easier than that of a nominal root, since names can assume predicative functions very easily. The following examples show constructions in which the position of the nucleus of the predicate is occupied by nominal roots (in boldface). (2)

a. Pi-ne-xaki AS-1SG.POSS-plate ‘It’s my plate.’ b. Me-pi-mitsuri 3PL.SBJ-AS-cats ‘They’re cats.’ c. P-ana-xupureru AS-LOC-hat ‘He is wearing the hat.’

2.3. Properties and characteristics Another factor of the high production of verbal constructions are the words that express properties and characteristics, since they belong to the syntactic category of stative verbs. This implies that its morphosyntactic environment is that of verbs. (3)

a. Tsi-mi-pe DIM-REL-size ‘The little one.’ (Lit. ‘the one that is little’) b. P-eu-yiwi AS-GLB-black ‘It’s black.’

2.4. Polysynthesis Finally, the Huichol language represents a very highlighted characteristic, typical of polysynthetic languages: the verb holds other information, which in other languages, such as Spanish, is necessarily expressed through nouns or words belonging to other categories. In Huichol constructions with certain verb prefixes can describe especial schemes associated, in the first

106 Paula Gómez López place, with certain parts of the body or objects. This makes the presence of the corresponding subject unnecessary, and sometimes, impossible (like in 4.a and c due to the inexistence of the required noun). (4)

a. P-a-ye-xeta AS-LOC-LOC-red ‘His face is red.’ b. ?axa p-eu-?ane-ne dirty AS-LOC-to.be~PL ‘His hands/feet are dirty.’ c. P-a-xawa. AS-LOC-to.have.holes ‘It has a hole.’

In the Huichol language we can also find the lexemes that P. Brown called “heavy verbs” in Tzeltal, like “to eat tortilla”, which contains both the action of eating and the object tortilla. They are specific verbs since they express properties of their arguments (cited in Genter & Boroditsky 2001: 240). In the following examples, the collocation verbs are specific of objects with certain shapes. (5)

a. Mexa-tsie p-a-ka. Table-in AS-LOC-put ‘It’s on the table.’ (object with volume, e.g., a cup) b. Mexa-tsie p-a-ma. Table-in AS-LOC-put ‘It’s on the table.’ (flat object, e.g., a plate)

The information about the shape of an object given by many verbs in Huichol makes possible not to mention the respective nominal syntagm. This piece of information united to the situational context makes the use of nouns unnecessary, since this facilitates the listener’s identification of the discussed object. I assume that there is a relation between the tendency of this language to the production of predicative constructions and the early development of relational words more than to that of referential words. This is the main point of the study. Nevertheless, there is a decisive issue I must address: the setting of criteria for the identification of words with relational and referential function. In this stage, the category distinction Name-Verb starts when

Acquisition of referential and relational words in Huichol

107

-ya appears in nouns and -ni with verbs. Undoubtedly, these suffixes constitute a linguistic clue to identify both types of words. However, most of the times only roots appear. The category distinction is far from being established. For lack of morphosyntactic clues, the situation of the communication and the reinforcement of the adults become the only support to decide where the girl of the case focuses on: on the event or on some of the objects that participate in it. Thus, we can decide whether the word has a referential or a relational nature7. 3. A case study Because it is a case study, it is worth talking about the individual differences that can appear in this process. According to C. Shore (1995: 13–14), among English-speaking children, it has been found from the case of a thirteen-month-old girl who produced 34 words, 80% were nouns, to the case of a twenty-month-old one who produced only 4 words which she used to refer to activities: ‘out’, ‘down’, ‘uh-oh’ and ‘fall down’, besides ‘mama’, ‘no’ and ‘uh uh’ used for different aims. In the Huichol learning process, Maura, the Huichol girl of the study, does not look like any of the other cases of English acquisition before mentioned. Her verbal production is not particularly abundant, but it is far from the second case, although it is similar to it in its predilection for expressions referring to actions comprised by verbs and particles. At least in the quantitative aspect, Maura does not seem to be an extreme case; I expect she is not a rare case in the qualitative plane compared to most of the Huichol children. 4. Lexicon acquisition from 16 to 24 months of age Lexicon acquisition in the first stages (until before 24 months) presents special problems. A series of decisions about the categories we are looking for and the way in which we are going to identify them among children expressions must be made. 4.1. What categories are we looking for? M. Tomasello (1992: 9) gives a list of terms which have been used to refer to verbs or non-nominal words in the studies about early learning acquisition: verbs, words for actions, relational words, etc.

108 Paula Gómez López If the results are comparable to the results in other languages, the use of the terms noun and verb is not advisable, since the properties of names and verbs can vary from one language to the other a lot. In this work I use the term ‘referential words’ or ‘words for objects’ and relational words’ or ‘words for events’. I am trying to approach a phenomenon that belongs both to the cognitive and to the linguistic field, and answer how the concepts related to the reference to events and objects are organized and focused on their linguistic expression in the first stages of language development. 4.2. How can we identify referential words and relational words? Relational words are, in this work, the expressions that a girl uses to make reference to a state or process, i.e. an event. In this class I found words that correspond to different lexical classes in adult language: verbs, particles, nouns and roots shared by verbs and names. Referential words are used to designate objects. This word class is exclusively comprised by nouns, except for the word kaaka which is the root of the verb “to be sweet”, and that the girl uses to refer to sweet food such as fruits. Whatever terms are used, it is necessary to delimit the category very well. Tomasello (1992: 11), who opts to use the term “verb”, considers ‘a verb’ “any word that the child uses to predicate a process of something regardless of that word’s status in adult language”. Naturally, grouping mechanically the roots that are nouns in a target language in one group and the roots that are verbs in another one is not the idea. However, we have to establish criteria to make a decision about considering a given word as relational or referential, specially when it does not seem to show the same function than in the adult language. One of the most important criteria is the situational context in which the word is used; another one is the morphosyntactic context. Besides the situation, Tomasello takes into account the complete construction in which the word in the adult language appears. For instance, ‘Hat off’ counts as a verb, since the girl uses it to refer to a process (‘take the hat off’); but the reference of “hat off” becomes easier due to the use of the particle ‘off’. How would the expression be considered if the girl only said ‘hat’? For the analysis of Maura’s lexicon I took the following considerations: first of all, there must be take into account that in this stage of acquisition, words for objects and those for events do not always coincide with nouns

Acquisition of referential and relational words in Huichol

109

and verbs of the target language. Gospnik & Choi (1995: 64) point out that even English-speaking children frequently use words to codify events and relationships, although these words were rarely verbs. It has been proved that children can codify verbs since a very early age. They do this not only with verbs, but with a variety of words (Smiley & Huttenlocher, 1995). This shows that in the acquisition of certain languages, the production of verbs, but not the codification of events, can be difficult. This probably happens because of structural and input factors. What it is certain is that in some languages structurally different from English, the production of verbs does not seem to be a problem for children. Therefore, I believe that, actually, two issues have been addressed simultaneously: on the one hand, the verb production; on the other, the production of words for events and words for objects. Farther on, I will also discuss these two aspects of Maura’s lexical production. In the Huichol data, I consider the following five particles used by the girl to refer to actions as relational words8. (6)

Particles that refer to events kena ki akuxi mana kaami

‘give me’ ‘here you are’ ‘wait’ ‘go away’ ‘It’s ready’

Although they refer to events, these particles are invariable shapes that do not present syntactic properties of verbs or of any other lexical class. There is another kind of words and roots that can be used to make reference both to an object and to an event. We find this word class in the lexicon exclusive of children, as well as in the adult lexicon. (7)

Words and roots with referential and relational function a) Children lexicon: mamu ‘to eat/food’ washe ‘to write/result of the action of writing’ b) Adult lexicon: tsitsi ‘breast/to suckle’ kwita ‘poop/defecate’ kaakai ‘huarache/put on huaraches’

110 Paula Gómez López All these words are very frequent in Maura’s lexicon. Because of their importance and the difficulty they present, I established the following criteria to decide which was the corresponding function in each occurrence: a) The morphosyntactic context (nominal, relational) in which they appear b) The situation and interpretation made by adults The predicative and nominal morphosyntactic contexts in which these words appear are the following: (8)

predicative contexts

nominal contexts

tsitsi-tia ‘breastfeed’ suckle-CAUS

tsitsi-e-ya ‘her breasts/udder’ breast-EPT-3SG.POSS

tsitsi-miki ‘to want to suckle’ to suckle-DES kwita-ma poop-ORN

‘stained with poop’

kwita-ya ‘its poop’ poop-3SG.POSS

kwita-miki ‘to want to defecate’ to defecate-DES yu-kwita ‘to defecate’ RFL-to defecate kaakai-tsie neuketi ‘put on your h.’ huarache-in he steps

kaakai-ya ‘his huaraches’ huarache-3SG.POSS

kakai-tsie we-ti ‘with the huarache on’ huarache-in to.step-ADV In Maura’s lexicon we can already find the nominal possession category, specially the suffix of third person singular -ya, with which the girl starts to mark the verbal-nominal opposition. Of the five words that both functions present, only kakai- receives this nominal mark: kakai-ya in the girl’s production. In this context, I took into account that it is about a word for an object. However, the constructions kakai-tsie count as relational words, not because they are so in adult language, but because it is the expression of the event “to put on/wear huaraches” in the girl. Kakai-tsie is a part of multiple constructions related to “to put one’s huaraches on (or shoes)”, e.g. kakaitsie neuketi “put on your huaraches” (Lit. “step on your huaraches”), “Kakai-tsie” says the girl pointing to the

Acquisition of referential and relational words in Huichol

111

barefoot rabbit, what it shows that the reference is to the general action of ‘put one’s huaraches on’ and not the object ‘huarache’. Moreover, when the girl says only kakai the situation shows that she is describing an action, as it is showed by the mother’s interpretation. Besides the syntactical context and the situation, the interpretation and the reinforcement of the adults were resources used to judge the employment of children’s words, specially in the case of tsitsi and kwita. These roots rarely appear as nouns in the adult input, they rather appear as a part of verbal bases, like tsitsi-miki ‘to want to suckle’, or tsitsi-tia ‘to breastfeed’; kwitama ‘with poop’ and kwita-miki ‘to want to poo’. Finally, I think that it is worth mentioning the case of the word wasie, not only because of its high frequency, but also because it represents the opportunity to know the origin and development of a word produced by children. At least in one case, we can observe that children tend to orientate towards individual actions or objects. The word washie (or waxie) is a creation of Maura. It arouse when her father was taking her hand to help her write her name while he repeated washie washie washie. From that moment on, she associated the situation of writing with the word washie, which is a Huichol traditional personal noun. The issue address is how the girl employs this word. Does she refer to the action of writing (or drawing), to the result of writing, or to the instrument used to write? The interpretation made by the adults ranged from the action of writing to the result “what it is written”, trying at the same time to be more cooperative with the girl and to offer their own interpretation9. The occurrences of this word in the girl do not present any morphosyntactic clue of its employment as a referential word. On the other hand, the situation seems to indicate that with such word the girl evokes the action in which it was originated. 5. Results I have tried to show the identification process of the relational words and referential words in Maura’s lexical production. Before presenting the production mentioned in detail, the following must be clear: kinship names are considered to be referential words in this counting; proper names nor onomatopoeias like wawau, miau and tas are taken into account. The general results are presented in tables I and II, where the lexical progressive production through six months of age is showed. In both tables,

112 Paula Gómez López words in boldface correspond to Spanish loans, and the words in cursive are exclusive of children lexicon. In table II words belonging to particle class appear in capital letters. All the occurrences of each word are presented, which allows the observation of abundant repetitions of some of them. The results show that the referential words do not exceed the relational words; on the contrary, there are more relational than referential lexemes. Taking the necessary precautions in order no to unconsciously assign the girl’s expressions to the classes of words for objects and the class of words for actions, and assigning the relational nouns (kinship and parts of the body) to the class of words for objects, I find a greater word production (types) referred to actions and states rather than objects. From 16 to 24 months of age, Maura used 34 words related to objects, and 38 related to events. On the other hand, if we eliminate the words that belong to the particles class from the words for events, we will get 33 verbs and 34 names, i.e. a fairly balanced situation. The relation among nouns and verbs is very far from being the one found in some cases of English: 80% were names and 20% were verbs (Shore 1995). 6. Conclusions It can be concluded that the results are similar to those obtained in the studies about Tzotzil (de Leon, 1998) and about other languages already mentioned: there is not any preponderance of the words for objects on the words for events. The girl rather shows a predilection for the words for events, judging by the lexicon exclusive of children, which includes only 1 referential word and 6 words referring to events. Moreover, the development of the word wasie exclusive of this girl does not show, by any means, that the referential use of the word is something primary; it rather shows that the girl associates, firstly, the general event or situation of writing or drawing with the word, and not with the result or instrument. It is worth adding that more than 30% of the words for objects are Spanish loans (in boldface, table I) and that in many occasions the recordings were in situations in which the girl was surrounded by objects: the game with illustrated books, toys and other different objects. These are not the normal situations for most of Huichol children in rural communities; they are for Maura, since she has mostly lived in an urban context, and her parents have provided her with an environment where books are everyday ob-

Acquisition of referential and relational words in Huichol

113

jects. The recording situation must be taken into account, since it can be a factor of the production of nouns and verbs, as the study by Gelman and Tardif (1998: 30) showed with English-speaking children and Mandarinspeaking children. These investigators found that there is a significant relation between the most frequent type of language and type of words in children’s production (nouns/verbs), but they have also found that the recording situations in which illustrated books and different objects are used stimulate the production of nouns and not that of verbs.

Notes 1. Gentner (1982) also found the same situation in other languages such as Japanese and Mandarin. She inferred that nouns are congnitively simpler to learn than verbs. 2. Tomasello and Merriman (1995: 2–7) point out that different reasons explain why investigators “are obsessed” with words for objects in children speaking field. They consider the belief that noun acquisition is the simplest part of lexicon acquisition to be the main reason. 3. Examples of studies about verb acquisition in Mexican and Central America languages are: Pfeiler (1998), Brown (1998), de León (1998) and Pye (2001). 4. Referring to the “cognitive simplicity” ascribed to nouns, de León and Rojas (2001: 23) explain that “In light of data about Mandarin Chinese, Corean, Tzotzil and Tzeltal acquisition, cognoscitivist predictions are refuted with data that show that verbs can be learnt as early as nouns in spite of their cognitive complexity” 5. In his study about English-speaking and Japanese children, Sakurai (1999) found a greater production of nouns than verbs in a similar relation to that found in the maternal input in both languages. This shows the importance of the study of the input. See also Smiley and Huttenlocher, 1995. 6. The formation of nouns from verbs is made through suffixation, as in abstract nouns(?utia-rika ‘writing’) and agentive nouns (of jobs: ?iitsií-kame ‘governor’). 7. Naturally, it should be taken into account that the reinforcement of adults are not always a manifestation of their interpretation of what children say, but their intention of correcting them. 8. I do not take into account some particles related to actions. These particles are used by the girl in the organization of the action, such as pa, patsi (they show beginning of action), ta and ri (they show change of action). 9. See V. Iturrioz (1998) about the interaction rules between the girl and the adults in the use of this word.

114 Paula Gómez López References Bloom, Paul 1993 The transition from infancy to language: Acquiring the power of expression. Cambridge, UK: Cambridge University Press. 2001 Roots of word learning. In: Melissa Bowerman & Stephen C. Levinson (eds.), Language Acquisition and Conceptual Development, 159–184. Cambridge: Cambridge University Press. Brown, Penelope 1998 La identificación de las raíces verbales en tzeltal (maya): ¿cómo la hacen los niños? Función 18: 121–146. Choi, Soonja 1997 Language-Specific Input and Early Semantic Development: Evidences from Children Learning Korean. In The Crosslinguistic Study of Language Acquisition, Vol. 5: Expanding the Contexts, Dan I. Slobin (ed.), 41–134. New Jersey /London: Erlbaum. de León Pasquel, Lourdes 1998 Raíces verbales tempranas en tzotzil: factores del input vs. Restricciones cognoscitivas. Estudios sobre la adquisición de algunas lenguas de México. Función 17–18: 147–174. de León Pasquel, Lourdes & Cecilia Rojas Nieto 2001 Sobre la adquisición del lenguaje: senderos en la conformación de un campo de estudio. In La adquisición de la lengua materna. Español, lenguas mayas, eusquera, Cecilia Rojas Nieto & Lourdes de León Pasquel (eds.), 17–50. México: UNAM/ CIESAS. Fisher, Cynthia 2000 Simple Structural Guides for Verb Learning: On Starting with Next to Nothing. In The Proceedings of the 30th Annual Child Language Research Forum, Eve V. Clark (ed.), 5–14. Stanford, CA: CSLI / University of Chicago Press. Gelman, Susan A. & Twila Tardif 1998 Acquisition of Nouns and Verbs in Mandarin and English. In: Eve V. Clark (ed.), The Proceedings of the 29th Annual Child Language Research Forum, 27–36. Stanford, CA: CSLI /University of Chicago Press. Gentner, Dedre 1982 Why nouns are learned before verbs: linguistic relativity versus natural partitioning. In Language development: Language, thought and culture, Stan Kuczaj (ed.), 301–334, Hillsdale, NJ: Erlbaum. Gentner, Dedre & Lera Boroditsky 2001 Individuation, relativity, and early word learning. In Language Acquisition and Conceptual Development, M. Bowerman & S. C. Levinson (eds.), 215–256. Cambridge: Cambridge University Press.

Acquisition of referential and relational words in Huichol

115

Gómez López, Paula 1998 Factores perceptuales y semánticos en la adquisición de la morfología en huichol. Función 18: 175–204. Gopnik, Alison & Soonja Choi 1995 Names, relational words, and cognitive development in English and Korean speakers: nouns are not always learned before verbs. In Beyond names for things: Young children’s acquisition of verbs, Michael Tomasello & William E. Merriman (eds.), 63–80. Hillsdale, NJ: Erlbaum. Iturrioz Leza, José Luis 1998 Acoplamiento estructural y adquisición del huichol como lengua materna. Función 17. Pfeiler, Barbara & Enrique Martín Briceño 1998 La adquisición de los verbos transitivos en el maya yucateco. Función 18: 97–120. Pye, Clifton 2001 La adquisición de la morfología verbal en español y maya k’iche’: Retos a la teoría de la ‘gramática universal. In La adquisición de la lengua materna. Español, lenguas mayas, eusquera, Cecilia Rojas Nieto y Lourdes de León Pasquel (eds.), 51–74. México: UNAM/ CIESAS. Rojas Nieto, Cecilia & Lourdes de León Pasquel (eds.) 2001 La adquisición de la lengua materna. Español, lenguas mayas, euskera. México: UNAM/CIESAS. Sakurai, Chikako 1999 A Cross-linguistic Study of Early Acquisition of Nouns and Verbs in English and Japanese. In Cognition and Function in Language, Barbara A. Fox, Dan Jurafsky & Laura A. Michaelis (eds.), 136–142. Stanford, CA: CSLI. (Center for the Study of Language and Information). Shore, Cecilia 1995 Individual Differences in Language Development. Thousand Oaks/ London /New Dehli: Sage Publications. Smiley, Paticia & Janellen Huttenlocher 1995 Conceptual development and the child’s early words for events, objects and persons. In Beyond names for things: Young children’s acquisition of verbs, Michael Tomasello & William E. Merriman (eds.), 21–62. Hillsdale, NJ: Erlbaum. Tomasello, Michael 1992 First Verbs: A case study of Early Grammatical Development. Cambridge: Cambridge University Press. Tomasello, Michael E. & William E. Merriman (eds.) 1995 Beyond names for things. Young Children’s Acquisition of Verbs. New Jersey /Hove, UK: Erlbaum

116 Paula Gómez López Table 1. Production of referential words Total = 34 lexemes mama ‘mom’ tuutu ‘flower’ tati ‘young goat’ kaaka ‘fruit’ kakai-ya ‘huarache’ teukari ‘grandfather’ wakana ‘hen’ patu ‘shoe’ kaxu ‘car’ ? iyaya ‘little daughter’ papa ‘dad’ turu ‘bull’ hawa ‘water’ tatatsi ‘uncle’ paapa ‘tortilla’ watsu ‘glass’ nunutsi ‘child’ kawayu ‘horse’ hixi ‘eye’ xupureru ‘hat’ kakawatitsi ‘peanut’ kwetsi ‘little brotherin-law’ xuriya ‘blood’ ? iwaru ‘little sisterin.law’ maxa ‘deer’ kaaru ‘banana’ kuka ‘chaquira’ yawi ‘key’ xikuri ‘handkerchief’ xari ‘saucepan’ kutsara ‘spoon’ wakaxi ‘cow’ munu ‘doll’ pixixii ‘little chicken’

(16 months 11* 13 14 1 1 2 1 1

15

16

11

17 1 2

18 1 8

19 1 7

20 6

24 months) 21 22 23 1 4 3 1

4

2 1

1 7 4 3 1 2 1

1

1 2 3 1

2

2 14

1 2

4 2 8 1

2

1

1 1 1

1

6 6

3

2

1 3

1

2 1 1 1 2

3 4 1 1 1 6 2 1 2 1 1 1

* Number of one-hour duration recordings from 16 to 24 months in the total of longitudinal data. Recording number 12 is lost.

Acquisition of referential and relational words in Huichol

117

Table 2. Production of relational words (16 months Total = 38 lexemes 11* 13 14 nunu ‘to sleep’ 2 4 mamu ‘to eat’ 5 KENA ‘give me’ 1 Kakai-tsie ‘to put shoes on’ ? uta ‘to keep’ 3 MANA ‘go away’ 4 KAAMI ‘it is ready’ 1 KI ‘(you) take’ tuni ‘to take’ kuku ‘to hurt’ mawe ‘There is/are not’ amunu ‘let’s go’ wasie ‘to write’ maima ‘to wash hands’ tsitati ‘to sit’ tsitsi ‘to suckle’ tseni ‘to bite’ mana ‘put’ ma ‘to put’ kutsu ‘to sleep’ hiina ‘take off’ (cloth) ka/tei ‘put’ tsina ‘to crush’ tsana ‘to break’ hauxina ‘to wash dishes’ wiya ‘to take’ taiya ‘to light’ hani ‘to take’ tara ‘to break’ kwita ‘to have poop’ upa ‘to pick up’ kawati ‘quiet’ ? AKUXI ‘(you) wait’ neuxei ‘(you) look’ name ‘to close’ wewi ‘to do/make’ wiwi ‘to throw’ yepi ‘to open’

15

4

16 2 3 1 1

17 1 1 1

18 3 4

19 4

2

1 1

20 1 1 1 5

4 4 1 1 1

24 months) 21 22 23 3 19 2 5 1 1 1 1

1

12 6 3

4

2 1 2 2

1 1 1 4

1 18 2 6 3

10

1 2

1

2 2 3

2

8 5

3 8

2

1 1 8

5 2 4 1

1

1 5 1 2 1

1

1

3 1 1

1

3 2 1 1 1

5 1 1

1 5 1 2 1 1 1

118 Paula Gómez López Abbreviations 1SG ADV AS CAUS DES DIM EPT GLB LOC ORN PL POSS REL REFL SBJ

first person singular adverbial assertor causative desiderative diminutive epenthesis global quality locative ornative plurality possessive relative clause reflexive subject

Culture-specific influences on semantic development: Learning the Tzeltal ‘benefactive’ construction 1 Penelope Brown

1. Introduction Can the specific language being learned – and its cultural context – have an early influence on children’s semantic development? Language acquisition theorists have generally suspected that the answer is no. The mainstream view has been that children start with concepts drawn from a universal repertoire of meanings, and only gradually modify them to language-specific shape in response to distributional properties of the input. Recent work, however, has provided evidence that children are actually remarkably sensitive to the language-specific semantic properties of certain types of relational words – spatial verbs and adpositions – from a very early age (from 14 months, children show a language-specific bias in comprehension; Bowerman & Choi 2001; Choi 2002). To date we have, however, relatively little evidence for children’s sensitivity to language-specific constructional meanings. The present paper aims to contribute to this discussion by examining how children acquire a particular construction – the ‘benefactive’ – in Tzeltal, a Mayan language spoken by approximately 200,000 people in southern Mexico. The so-called ‘benefactive’ is the basic Tzeltal three-argument ditransitive construction; it is the primary way to express three core arguments in this language. For example: (benefactive suffixes and glosses are in boldface type)2: (1)

ya k-ak’-be-t atzam ICP1ERG-give-BEN-2ABS salt ‘I give you (the) salt.’

This construction is standard for verbs of transfer and certain verbs of speaking (which may be construed as verbal transfer):

120 Penelope Brown (2)

ya ‘-pas-be-n waj ICP 2ERG-do-BEN-1ABS tortilla ‘You make tortillas for me.’

(3)

la

y-al-be-n y-a’yejal 3ERG-tell-BEN-1ABS 3ERG-story ‘He told me the story.’ CMP

The same construction is often used when the direct object of any transitive verb is a possessed noun – then the possessor is almost always promoted to the position of indirect object (‘possessor ascension’), as in: (4)

ya s-mulan-be-t a’-na ICP 3ERG-like-BEN-2ABS 2ERG-house ‘She likes you your house.’ (i.e., she admires it/covets it).

(5)

la

s-k’an-be-0 y-asarona 3ERG-want-BEN-3ABS 3ERG-hoe ‘She asked her for her hoe.’ [permanently] CMP

(6)

la

s-maj-be-0 s-k’ab 3ERG-hit-BEN-3ABS 3ERG-hand ‘He hit him his hand.’ CMP

In speech to small children, this construction is very often used in imperatives (e.g., ‘Tell-him your brother to come here’), warnings (e.g., ‘It will burn-you your hand! (fire)’, ‘Don’t touch-it the dog its-mouth!’), or offers (‘I’ll mix-you your corngruel, I’ll change-you your skirt’). It is therefore fairly frequent in input speech. Tzeltal has no inherently ditransitive verb roots – even ak’ ‘give/put’ is not a root ditransitive – and this benefactive construction is the only way to get a third argument into the core of the clause. Although massive argument ellipsis means that the three participants are rarely all overtly expressed as NPs, this construction ensures that the ‘recipient’ or ‘benefactee’ or ‘affectee’ participant is overtly marked on the verb.3 As we shall see, it has a much more general meaning than ‘benefactive’, so I call this Tzeltal construction BEN.4 There are two additional ways to get three arguments into a sentence, but both place the third argument in an adjunct. I will show that Tzeltal children acquire this BEN construction remarkably early; they start using it by the two-word stage and apparently have productive use by around age 2;6, well before they reliably cross-

Culture-specific influences on semantic development 121

reference other arguments – Agent, Possessor – with ergative marking (Brown 1998b). These early BEN suffixes appear in frozen forms at the one- and two-word stages, but rapidly appear with variation across person and across different verbs. Furthermore, even at an early stage children do not restrict this construction to expressing canonical transfer scenes, but use it much more generally to express situations where something implicitly good or implicitly bad is at issue. The children seem to be attuned to the language-specific semantics of the construction: they do not simply use it at first for transfer scenes which semantically require three participants (an Agent, a Patient or Theme, and a Goal/Recipient), but, from the beginning, scenes that semantically have two-participants are among those where this ‘affected’ participant is indexed by children. For example5: (7) Lus, age 2;0: _ -pojben alal ‘(He) steals me (my) doll.’ (8) Xan, age 2;2: _ yixlanbet laso ‘He played with (your) rope for/on you.’ (9) Xan, age 2;2: _ -lo’ben -tomut antun i ‘Antun ate me (my) egg.’ Children’s early mastery of the BEN construction is of interest to language acquisition theorists because, from a typological perspective, there is reason to expect children to have some trouble expressing three participant events. There is a considerable amount of variation both across languages, and within one language, in how these are linguistically coded; in particular they vary in terms of how participants involved in three-participant situations (the Agent, Theme, and Goal/Recipient) are encoded (Brown, Eisenbeiss & Narasimhan 2002; Narasimhan, Eisenbeiss & Brown 2007). Threeplace predicates have therefore been the focus of recent theoretical attention, as an obvious testing ground for current theories of syntax-semantics mapping (Margetts & Austin 2007). Because of this typological variability, three-place predicate constructions potentially provide a challenge for children acquiring language. Children have to determine (a) what linguistic devices can be used for encoding threeparticipant events in the language, (b) what are the respective semantic contributions of these devices, and (c) what are the construction- and languagespecific constraints on their combination. These can all cause difficulties for language learners. For example, German children produce characteristic kinds of case-marking errors when expressing three-participant events (Eisenbeiss & Matsuo 2003). Three-place predicates are therefore an important locus for examining how children acquire argument structure and how this process

122 Penelope Brown is influenced by the typology of the language they are learning as well as by culturally-specific semantic categories (Bowerman, Brown, Eisenbeiss, Narasimhan & Slobin 2002; Slobin, Bowerman, Brown, Eisenbeiss & Narasimhan in press). Research on language acquisition has assembled considerable evidence for how, across different languages, transitivity classes and argument structure are learned (cf. articles in Slobin (ed.) 1985, 1992, 1997; Pye 1985), yet acquisition of ditransitive and other three-place predicates has been much less thoroughly explored (exceptions include Billington 2002; Demuth 1998, 2003).6 The questions raised here focus on the expression of a third argument in the production of young Tzeltal children: When do Tzeltal children start using the BEN construction? Do they start with the prototypical benefactive semantics: ‘give’? When is the construction productive? What are their preferred verbs in the construction at different ages? What is the developmental trajectory? Do they have alternative solutions to expressing a third argument? The data also speaks to larger issues: How are transitivity classes and argument structure learned in a language with massive argument ellipsis? How do Tzeltal children learn what culturally appropriately gets expressed with the BEN construction – what is the role of particular cultural practices in socializing children to culture-specific ways of thinking in this case? The plan for this paper is as follows: in section 2, I outline a typological perspective on benefactives and present the ‘give’-schema as the putative prototype for these constructions crosslinguistically. In section 3, the Tzeltal BEN construction is described in relation to transitive and intransitive constructions, and its challenges to the learner are outlined. Section 4 presents the Tzeltal child data, showing the range of verbs three children use in the BEN construction at four age points (roughly ages 2;0, 2;6, 3;0, 3;6). In section 5, I discuss the pragmatic uses for this construction in both child speech and adult input speech. Section 6 presents my conclusions. 2. Benefactives crosslinguistically ‘Benefactives’ refers to those three-argument constructions which are derived from a transitive stem by verbal suffixes, or, in languages like English, by word order alternations. Like applicatives, benefactives are verbal affixes (or alternations) that increase valence, ‘those constructions in which beneficiaries are coded as arguments … rather than as adjuncts’ (Shibatani 1996: 159), as in (10), in contrast to (11):

Culture-specific influences on semantic development 123

(10) John bought Mary a book. (11) John bought a book for Mary. Many languages have two forms (alternations) for the benefactive: the double object form (as in (10)) and the prepositional phrase form (as in (11)). In English, the contrast is marked by word order and, in the second form, by a preposition marking the recipient of the action. Choice of form in this alternation (as in the causative, dative, or locative alternations) indicates the speaker’s current choice of perspective (see, for example, E. Clark 1991). In Tzeltal, there is no alternation, but simply the choice of the benefactive construction, marked by a specific set of affixes that index the person affected on the verb. As this three-argument construction also requires you to take a particular perspective, it could be expected to be more difficult for children to acquire than one- (intransitive) or two-argument (transitive) constructions that are associated with the majority of verbs. Shibatani (1996) argues that benefactives are typologically interesting for several reasons. First, they are interesting to syntacticians as an important example of one way languages can express relatively peripheral participants as central clausal (core) arguments (e.g., indirect objects, datives, ‘external possessors’); these are ‘voice alternations’ in the sense that you have to take a ‘perspective’ on the scene to formulate grammatical encoding. Second, benefactives are famous for displaying language-internal semantic specificity – verbs you can and cannot use the construction with; cf. Pinker (1989). There is generally a cline of acceptability for verbs used in a benefactive construction, with different cut-off points for different languages; compare these two sentences in English: (12) John threw Mary the ball. (13) ? John opened Mary the window. In contrast, in German the analogue to (13) is acceptable and the cut-off point is with the verb meaning ‘close’ (examples are from Shibatani 1996: 170): (14) Otto öffnet Karin die Tür. [lit.] ‘Otto opened Karin the door.’ (15) *Otto schliesst Karin die Tür. [lit.] ‘Otto closed Karin the door.’

124 Penelope Brown Many theorists think of ‘give’ as the prototype schema for three-argument verbs, and thus for benefactive scenes (Shibatani 1996); ‘give’ verbs have therefore received a lot of crosslinguistic attention (Newman 1996, 1997, 2002; Margetts 2004). Gleitman (1990) extends this perspective to language acquisition, using ‘transfer scenes’ expressed by verbs meaning things like ‘give’, ‘put’, as a core example of the kind of universal semantics/syntax mapping that enables children to ‘bootstrap’ verb meaning from syntax: Verbs that describe externally caused transfer or change of possessor of an object from place to place (or from person to person) fit naturally into sentences with three noun phrases, for example, John put the ball on the table. This is just the kind of transparent syntax/semantics relation that every known language seems to embody…. Restating this more positively, the component ‘transfer’ is inserted into a verb’s semantic entry in case it is observed to occur in three noun-phrase structures.

(Gleitman 1990: 30; itals added) Verbs of transfer (‘give’, ‘put’, ‘receive’, etc.) are often considered to be the canonical three-argument verbs for children learning language (e.g., Slobin 1985, Gleitman 1990). The claim, then, in these approaches to three-place predications is that to use a benefactive you need a situation construable as one of ‘giving’, for example in the sense of the ‘give’ schema of Shibatani (1996: 173–174): (16) ‘Give’ schema: Structure: [NP1 NP2 NP3 GIVE] NP1 = coded as subject NP2 = coded as primary object or dative IO NP3 = coded as secondary object or as a direct object Semantics: NP1 causes NP2 to have NP3; i.e. NP1= human agent, NP2= human goal, NP3= object theme NP2 exercises potential possessive control over NP3 NP1 creates the possessive situation on behalf of NP2 When this ‘give’ schema is applied to benefactives, two modifications are required. NP3 can be an event, not just an object (metonymically related): (17) John sang Mary a song. And it can be malefactive: NP1 did something to NP3 for either the benefit or the malefit (harm) of NP2:

Culture-specific influences on semantic development 125

(18) John hit me my hand. [see ex. 6] Tzeltal, however, presents an interesting challenge to the view that transfer events are the prototype for three-argument constructions, with ‘give’ the typical verb, since in Tzeltal, three-participant events are expressed with the BEN construction, and for this three-argument benefactive construction – although it can be used to express transfer (‘give’) events – the event types that are prototypical are centred more on activities that benefit or harm people than on the transfer of possessions per se. In constrast to Shibatani’s proposal that transfer events are prototypical here, I shall propose a competing hypothesis: there is no universal prototypical event type for three-argument constructions; rather, language-learners have to identify the event types relevant for mapping three-argument constructions in any one language. They must learn what the conventions are in that language and indeed, in their speech community, as such conventions may vary across different speech communities whose members speak ‘the same’ language. Let us turn to the details of the benefactive construction in Tzeltal. 3.

The Tzeltal BEN construction

3.1. Argument marking: one, two, or three arguments Tzeltal is a basic VOS language; arguments, when overtly expressed as NPs, generally follow the verb. There is obligatory aspect marking and person cross-referencing on the verb. Arguments are cross-referenced on verbs in an ergative pattern, with ergative prefixes for the subjects of transitive verbs, and absolutive suffixes for the objects of transitive and subjects of intransitive verbs, as shown in (19) and (20): (19) Person-marking on intransitive verbs (one argument): Aspect + Verb_Stem + Absolutive (SUBJ NP) e.g. ya x-way-on ICP ASP-sleep-1ABS ‘I am sleeping.’ (20) Basic person-marking on transitive verbs (two arguments): Aspect + Ergative + Verb_Stem + Absolutive (OBJ NP) (SUBJ NP) e.g. la s-maj-on CMP 3ERG -hit-1ABS ‘He hit me.’

126 Penelope Brown If there are three arguments, however, the indirect object is promoted to a position where it, rather than the object, engenders absolutive suffixes on the verb – i.e., the Recipient is treated like the Patient of a transitive verb (Dayley 1981)7; this is actually a crosslinguistically frequent pattern (Dryer 1986). So there are three core arguments in the Tzeltal BEN construction, although only two are cross-referenced on the verb; the Patient (direct object argument) is not demoted to an oblique case but is not marked on the verb and, if expressed, comes directly after the verb: (21) Person-marking on ditransitive verbs (three arguments): Aspect+Ergative+Verb_Stem+BEN+ABS (OBJ) (SUBJ) (RECIP) la smaj- -be-n8 (j-k’ab) CMP 3ERG- hitBEN-1ABS (1ERG-hand) ‘He hit me (my hand).’ As we saw in examples (1) - (6), this BEN construction is standard for verbs of transfer and certain verbs of speaking; furthermore, when the direct object of any transitive verb is a possessed noun then the possessor is almost always promoted to the position of indirect object (‘possessor ascension’).9 The referent of the benefactive object normally has to be the same as the possessor: (22) la

y-ajch’al-tes-be-n j-tzek 3ERG-mud-CAUS-BEN-1ABS 1ERG-skirt ‘He got my skirt muddy for me.’ CMP

(23) la

s-butz-be-n k-inam 3ERG-kiss-BEN-1ABS 1ERG-wife ‘He kissed me my wife.’ CMP

But there are exceptions to this generalization; for example the referent of the BEN argument is not the same as the possessor in (24) or (25): (24) ya k-ik’-be-t bel ta loktor te Xun-e ICP 1ERG-take-BEN-2ABS awaywards PREP doctor ART Xun-CL ‘I’ll take for you Xun [addressee’s son] to the doctor.’ (25) ya

j-chol-be-t s-k’op dios 1ERG-recite-BEN-2ABS 3ERG-word god ‘I recite God’s word for you.’ ICP

Culture-specific influences on semantic development 127

BEN with non-transfer verbs is thus not restricted to possessor raising. Crucially, no Tzeltal verb root, not even ak’ ‘give/put’, is obligatorily ditransitive in the sense that it always has to appear in BEN.10 With an expressed Recipient, verbs of transfer do require BEN. But, generally, use of a verb in a transitive or ditransitive construction is differentially preferred depending on the particular predication and also on cultural expectations. Some sentences are judged ungrammatical without BEN:11 (26) ya k-il-be-t a’-sit (*k-il) ICP 1ERG-see-BEN-2ABS 2ERG-eye ‘I see you your eyes’ [i.e., I look at you] but many verbs can appear in either a transitive or ditransitive construction equally happily: (27) ya

j-man a’-wakax OR ya j-man-be-t a’-wakax 1ERG-buy 2ERG-cow OR ICP 1ERG-buy-BEN-2ABS 2ERG-cow ‘I buy your cow’, or, ‘I buy (from) you your cow.’ (BEN is optional) ICP

(28) ya s-poj-be-0 s-mamalal ICP 3ERG-steal-BEN-3ABS 3ERG-husband ‘She steals (from) her her husband.’ (BEN optional but preferred) Using BEN can disambiguate the argument roles, an important feature in a language with free argument ellipsis: (29) ay antz ya s-mulan-be-n j-mamalal exist woman ICP 3ERG-like-BEN-1ABS 1ERG-husband ‘There’s a woman she likes me my husband.’ [i.e., she covets him] Without BEN, sentence (29) would be read as meaning ‘my husband’ is the liker of the woman. In addition, certain verbs can have two entirely distinct meanings when used in the transitive vs. the ditransitive (BEN) construction; for example, the verb na’ ‘to know’: in the ditransitive ya s-na’-be-n j-ba means ‘He knows me’, but in the transitive ya s-na’-on means ‘He thinks of me’. Is there a cline of acceptability for different verbs used in this construction? There is no evidence that this is the case for Tzeltal. A very large range of verbs – virtually any transitive verb – can be construed in terms of affecting someone or something in such a way as to generate reference to a third,

128 Penelope Brown Recipient/ Affectee argument. There is thus a very general meaning for the construction – not ‘transfer’, but ‘help/hurt Recipient/Affectee’, a meaning that is invoked whenever a verb is expressed in this construction (which therefore seems to be a single construction with very wide application). BEN conveys the sense that the action of doing something to someone’s bodypart or (potential) belongings has affective relevance to the possessor. Using the construction appropriately requires a culturally specific – and context specific – assessment of different ways actions on possessees can be relevant to, or affect their possessors. One occasionally hears speakers of English produce utterances with a similarly ‘malefactive’ flavor using the ‘on PRONOUN’ construction, though this construction is decidedly marked in English: (30) Overheard comment by bride (to guest at wedding) eating grapes: ‘I’m just afraid to eat them all on you’ (Kwedding: July 2002) (31) Remark of wife on phone to husband, about having told others about something he didn’t especially want her to: ‘There, I told them on you.’ (MD: Feb 2003) The meaning of this English ‘on PRONOUN’ construction seems to be generally ‘hurt, or be otherwise affectively relevant to Recipient/Affectee.’ Constructions with a meaning of this kind occur in many languages (O’Connor 1996). What is language- and culture-specific about Tzeltal BEN is the wide range of event types that the construction covers and the culture-specific knowledge necessary to construe a situation as warranting encoding with BEN. There are two alternative ways to express three participants in Tzeltal; neither of these cross-references the participant on the verb: (i)

an oblique prepositional phrase (this is not possible for expressing recipients):

(32) la

y-ak’-0 jun ta mexa 3ERG-give/put-3ABS book PREP table ‘He put the book on the table.’ CMP

(ii)

a possessed relational noun phrase with y-u’un ‘its relation’, or ‘because of me/you/he/she/it’:

Culture-specific influences on semantic development 129

(33) la

y-ak’-0 jilel jun y-u’un 3ERG-give/put-3ABS remaining_behind book 3ERG-RelN te Jxun-e ART Jxun-CL ‘He left behind a book for Xun.’ CMP

In contrast to BEN, neither of these constructions appears early in child speech. 3.2. The Tzeltal child’s task in learning BEN The Tzeltal-learning child has to learn to distinguish different semantic roles for arguments; plausibly, she has to learn the transitive construction first, cross-referencing A (agent) and O (direct object) arguments on the verb and expressing them (optionally) in an overt NP after the verb, and then modify this for the ditransitive, expressing A and R (recipient/benefactee) arguments cross-referenced on the verb and O (optionally) in an NP after it; absolutive suffixes on the verb now cross-reference R not O. Secondly, the child has to learn when to use this construction: very often (but not always) with some verbs (‘tell’, ‘give’, and any verb followed by a possessed noun as O), but sometimes with other verbs – namely, whenever taking a perspective that someone is ‘affected’ in an emotively relevant way by the action of the verb on the object. Learning when to take this perspective involves learning the contrast between the BEN perspective (agent does something to patient that affects someone/something else) and the transitive perspective (agent does something that affects patient). Depending on what the ‘something’ is that’s being done, different constructions are preferred. For example, you can say the sentence in (34), using BEN: (34) ya j-k’opon-be-tik dios ta s-tojol te chamel. ICP 1ERG-speak-BEN-1ERGPL god PREP 3ERG-concern ART sickness ‘We-inclusive speak to God for him about the sickness.’ But this would be more naturally expressed transitively, as in (35): (35) ya

j-chol-tik s-k’op dios a’w-a’y 1ERG-recite-1PL 3ERG-word god 2ERG-know/feel ‘We-inclusive speak to God you know.’ ICP

130 Penelope Brown Children have to learn the subtle understandings and values that condition which formulation is preferred in such cases. The learning task appears simple in some respects, difficult in others. Tzeltal inflectional morphology is in general highly regular and not particularly difficult for the learner (Brown 1998b). The morphology of BEN is straightforward, with the form of the benefactive suffixes transparently related to that of the absolutive, as can be seen in Table 1.12 Yet this morphology is cross-referencing a different kind of argument, a Recipient/Affectee rather than a Patient, which might be expected to confuse children. Table 1. Absolutive and benefactive suffixes in Tzeltal Absolutive

Benefactive + ABS

st

-on

-be-n

nd

2 sg.

-at

-be-t

3rd sg.

-0

-be-0

st

-(o)tik

-be-tik

st

-on jo’tik

-be-n jo’tik

nd

-ex

-be-x

rd

-ik

-be-ik

1 sg.

1 pl. incl. 1 pl. excl. 2 pl. 3 pl.

Furthermore, BEN raises a number of morphosyntactic complexities that the child has to grapple with in this context. These arise especially in connection with plural marking, which is in many contexts optional. Frequent argument ellipsis means that overt NPs often are not present to disambiguate whether arguments are singular or plural. One site of potential confusion is the interaction between ergative plural and BEN plural. Children have some trouble with this, as we shall see in section 4.3. Whether plural -ik has scope over the BEN argument or the ERG argument is sometimes in doubt – contrast (36) and (37) (cross-referencing affixes that belong to the same argument are in boldface): (36) ya s-pas-be-ik ICP 3ERG-do-BENpl ‘He does (it) for them’. (37) ya s-pas-be-ik ICP 3ERG-do-BEN-3ERGpl ‘They do it for him/her/it.’

Culture-specific influences on semantic development 131

There are also potential problems with morpheme order in the plural, and with the passive. The passive, marked with the suffix -(o)t, is possible in 3rd person only in BEN. Its form is identical with that of 2nd person singular absolutive of BEN, although as passive it is relatively rare in childdirected speech, so it is unlikely to present an early problem: (38) al-be-t tell-BEN- PASS ‘(It) was told to him.’ (39) k’an-oj-be-t winik-etik want-PERF-BEN- PASS man-PL ‘The men have asked him for it.’ (lit: ‘He has been asked for it (by) the men.’) There is thus considerable morphosyntactic complexity to master in connection with BEN. What about the semantics – can universal meanings like generic ‘give’ help the child here? 3.3. Problem with the ‘give’ schema as the core meaning of Tzeltal BEN The Tzeltal benefactive construction is THE three-argument construction (the only one). No alternative phrasing is possible to express three core arguments – Tzeltal has only one, general-purpose, preposition, which cannot be used to say things like ‘I bought the book FOR Mary’ using an oblique prepositional phrase. Therefore, as we have seen, although BEN can be used to express transfer events, it can also be used for events of many other kinds. BEN conveys the sense that the action of doing something to someone’s bodypart or (potential) belongings has affective relevance to the possessor. While this very general meaning is not specific to Tzeltal, using the construction appropriately requires a culturally-specific assessment of different ways actions on possessees can be relevant to, or affect, their possessors13. Focusing here on the semantics, we may ask: What kinds of situations take BEN descriptions in Tzeltal, and how do children learn this? And how do they learn the associated cultural knowledge?

132 Penelope Brown 4.

Child data

4.1. The data samples The database for the research reported here is based on recordings of spontaneous language production of Tzeltal children interacting with family members and/or the investigator in their homes; these were video- and/or audiotaped every 4 to 6 weeks over a period of several years.14 I report here on data from three Tzeltal children: two girls (Xan, Lus) and one boy (Mik), sampled for about 4-8 hours each, at four age points roughly six months apart (see Table 2). Table 2. Child samples for benefactive analysis LUS Sample

Age

XAN Age

MIK

MLU

Sample

MLU Sample

6 sessions 2;0

1.6

7 sessions 2;2

1.7

4 sessions 2;3

1.1

2 sessions 2;6

2;2

2 sessions 2;7

2.9

5 sessions 2;4 – 2;6

2.0

3 sessions 3;1

3.4

3 sessions 2;11 – 3.2 3;2

5 sessions 3;2

3.1

2 sessions 3;6

3.7

3 sessions 3;5

4 sessions 3;7

3.9

4.1

Age

MLU

In these samples I coded all child utterances for argument structure; I then extracted the forms with BEN marking for qualitative analysis of the semantics and pragmatics of the verbs used in this construction by the children. (I excluded from analysis utterances that were verbatim repeats of prior utterances, as well as any whose meaning was unclear in the context.) 4.2. What verbs do children use BEN with? Do children initiate use of BEN with verbs that have a transfer meaning? By ‘transfer’ here I mean physical movement of an object between persons or places, in line with Shibatani’s ‘give’ schema in (16) above. Of course one can imagine that metaphorical transfer is part of the meaning in virtually any example, but children this age are not likely to have much grasp of metaphor. Let’s look in detail at what verbs children first start using BEN

Culture-specific influences on semantic development 133

with in my samples. Note that, in most of these first BEN uses, the children omit aspectual markers and the ergative prefix (missing morphemes are indicated here by initial underline (for aspect) or hyphen (for ergative). Examples: LUS, 2;0: In her first samples, Lus uses BEN with 7 types of verbs, including the following: LUS, age 2;0 Transfer poj ‘steal, take away from’ [15 tokens] _ -poj-be-n alus ‘Alux steals it from me.’ _ -poj-be-n k-u’un ‘(He) steals mine from me .’ _ -poj-be -sijtz ‘(You) steal (her) shawl from her.’ ak’ ‘give, put’ [3 tokens] _ k-ak’-be-ix ‘I gave it to him.’ ak’-be-tik ‘Give it to us.’ _ k-ak’-be ek moch’ ‘I give her a basket too.’ tzak ‘take, grasp in hand’ [2 tokens] _ -tzak-be-n Nik ‘Nik takes it from me.’ tzak-be-n tal ‘Get it for me.’ LUS, 2;0: Non-transfer chol ‘change [clothing]’ chol-be ‘Exchange it [skirt] for her.’ [it should be: jcholesbe stzek ‘Change her her skirt.’] al ‘tell’ _ k-al-be me’tik lumine ‘I’ll tell it to Mrs. over there.’ kus ‘wipe’ _ -kus-be-t ‘(I) wipe it for you.’ At this age, there are not enough examples to tell whether BEN is productive; LUS uses it in all three persons but in most cases (except for poj ‘steal’ and ak’ ‘give/put’) in only one person for a given verb root.

134 Penelope Brown Examples: XAN, 2;2 XAN in her first sample uses BEN more prolifically than LUS, with 17 verb types. For example: XAN, age 2;2: Transfer poj ‘steal from’ [2 tokens] _ -poj-be-n -k’ab xutax ‘The scarecrow steals from me (my) hand.’ [it should be stzakben ‘he grabs it for me’] _ -poj-be-n alal ‘(He) steals from me the doll.’ ak’ ‘give, put’ [2 tokens] ya k-ak’-be-tik -we’ mut ‘We-incl. give it to the chicken(s) to eat.’ [it should be swe’el mut] _ k-ak’-be ‘I’ll ‘give’ it to him.’ [threat] ich’ ‘get’ _ k-ich’-be-t ‘I get it from you.’ XAN 2;2: Non-transfer puk’ ‘mix with hand’ [3 tokens] _ -puk’-be-n ‘(She) mixes (corngruel) for me.’ 16 verbs with one or two tokens each, e.g.: _ -koj-be-n j-ba-e ‘(We) knocked me (my head)’ (i.e., our heads knocked against each other) [Target: should be la jk’ojbe jba jo’tik] _ -ti’-be-n mut ‘(It) the chicken bit me (my-foot).’ _ -boj-be-t ‘(He) hit you (your-bodypart).’ _ -muk-be y-akan ‘(I) cover her her foot.’ [doll] ma me ‘-kas-be y-akan ‘Don’t break it it’s foot.’ la -k’ech’-be -nuk’ alal ‘(He) twisted off the doll’s neck for it.’ [i.e., for the doll] _ -tij-be-tik ‘We-inclusive push it for her.’ chuk-be-n ‘Tie it for me.’ _ -nup-be ‘(I) blow for it (the fire)’ [meaning: ‘I blow on the fire for it (the fire) so it will burn better.’] _ -lo’-be-n tomut ‘(He) is eating me (my) egg.’

Culture-specific influences on semantic development 135

_ y-ixlan-be-t laso antun i, … y-ixlan-be-t laso-tat antun i, laso ja’at i ‘Antun plays_with for_you rope, … Antun plays_with for_you rope father, you(r) rope.’ [telling her father, tattling on her brother Antun’s misdemeanor. Target: ya yixlanbet a’laso ja’at te antune.] Again, Xan uses BEN in all three persons and first person plural, but with the exception of ak’ ‘give/put’, not more than one person per verb type. Examples: MIK 2;3 There is less evidence in Mik’s data for this construction; he is in general less advanced than the other two children at this age. His first uses of BEN (in 12 samples examined beginning from age 2;0) do not appear till age 2;3, with 3 types. MIK: 2;3: Transfer pas ‘do/make’ [2 tokens] -pas-be ini. -pas ini. ‘Do this for it. Do this.’ [stack blocks] -pas-be-n i. ‘Do it for me.’ [open a plastic bag] ak’ ‘give, put’ [2 tokens] ak’-be-n la papa ‘Give it to me.’ ak’-be me ‘‘Give’ it to him.’ [threat] MIK 2;3: Non-transfer k’ej ‘put away’ [1 token] -k’ej-be Pontz ‘(We’ll) put it away for Pontz.’ [lollypop, for his brother Pontz] Thus, for two of his three verbs (ak’ ‘give/put’ and pas ‘do/make’), Mik contrasts two different persons with the same root. He also uses both of these verbs in transitive constructions. Summary The children use a wide range of verbs with BEN in their first samples, and they use BEN in different persons right from the beginning of the sampling period. For all three children, at this early stage (MLU 1.1 to 1.7), there is no evidence yet of morphological productivity – no verbs are used with all three (singular) person markers in the data sampled, and most appear with

136 Penelope Brown only one. But the range of verbs used with BEN rapidly widens with age, as can be seen in the summary presented in Table 3 below: Table 3. Cumulative summary of verb types used with BEN, all 3 children, (and tokens for those used more than once) LUS

XAN

MIK

age 2;0

types: 7 used more than once: poj ‘steal from’ (15) ak’ ‘give’ (3) tzak ‘grasp’ (2)

types: 17 used more than once: puk’ ‘mix by hand’ (3) chuk ‘tie’ (2) poj ‘steal from’ (2) ak ‘give’ (2) muk ‘cover’ (2) nup ‘blow on’ (2) ixlan ‘play with’ (2) ti’ ‘bite’ (2)

types: 2 more than once: pas ‘do/make’ (2) ak’ ‘give’ (2)

age 2;6

new types*: 6 pas ‘do/make’ (4) ich’ ‘get’ (4) kuch ‘carry’ (2)

new types*: 8 tilp’un ‘untie’ (4)

new types*: 9 chuk ‘tie’ (13) poj ‘steal from’ (3) jol ‘open’ (3)

age 3;0

new types: 10 man ‘buy’ (8) chuk ‘tie’ (3)

new types: 16 otzes ‘make-enter’ (6) ch’ol ‘pour’ (3) il ‘see’ (3) set ‘cut’ (3) ch’ay ‘drop’ (2) lok’es ‘make exit’ (2) jom ‘pierce’ (2)

new types: 15 tij ‘drive’ (5) man ‘buy’ (5)

age 3;6

new types: 12 tuch’ ‘break’ (4) le ‘look for’ (3) toch ‘peel’ (3)

new types: 11 pech’ ‘braid’ (3) toy ‘raise-high’ (2)

new types: 2 nit ‘pull’ (4) latz ‘pile’ (2)

* New types = for each sample, types new since the last sample, not produced in prior samples

If, as the criterion for a certain degree of productivity, we take production with several different verbs across all three persons, then the children appear to productively use BEN by their samples at age 28 to 32 months. BEN develops, in the sense that children use it initially with a few verbs

Culture-specific influences on semantic development 137

and a strong emphasis on first person; by age 2;4 to 2;8 they have many verbs and several persons per verb. They continue to add verbs throughout the data sampled (a total of 39 verb types with BEN in Lus’s samples, 59 for Xan, 27 for Mik.). 4.3. Errors and innovations Errors with BEN are relatively rare. Early on, children sometimes use a verb in a context that seems to need BEN without the BEN, so errors are errors of omission (of BEN, as well as of aspect, ergative and absolutive markers). There are a few person and number errors, as in: (40) ABS instead of BEN: Xan age 2;2: -ti’-at ‘bite you’ [instead of ya s-ti’-be-t ‘It bites (it) for you.’] (41) Person errors: Lus age 3;0: ma ba *la !k\ak’-be-n kaloj ek i antun i ‘Antun didn’t give me the car!’ [using lst-person ergative prefix (instead of 3rd) plus 1st-person BEN.] (This latter is not really a BEN error but an ERG error, involving a reanalysis of the root as being vowel-initial.) But I have found no obvious examples of semantic errors (inappropriate uses) in the data examined. 4.4. Input Much could be said about the input; here I report on a brief examination of one session of about one hour, with five adults and four older children (aged 4, 6, 8, and 11) present, all of whose speech I consider to be relevant input for the focal child (Tzeltal children receive a large proportion of their caregiving, and hence input speech, from older siblings and cousins). This sample consisted of 970 input utterances, of which 120 used the BEN construction, with 25 different verb types, representing all persons, all aspects, and many constructions including passives and reflexives with BEN. Clearly, there is no shortage of benefactive in speech to small children. The input verbs used three or more times with BEN during this one hour sample are given in Table 4.

138 Penelope Brown Table 4. Input verbs used with BEN in a one hour session Verb root

gloss

tokens

ak’ poj le al tzak pas man tij nit

‘give, put’ ‘steal, take away’ ‘search for’ ‘tell’ ‘grasp in hand’ ‘do, make’ ‘buy’ ‘move, push’ ‘pull’

55 15 6 6 5 5 4 3 3

It can be seen that ‘give/put’ is far and away the most frequent; on these grounds it might be considered to be the prototype for this construction. This makes it all the more interesting that children’s usage at age two does not favor ‘give’ in this construction. The input has examples of ‘give’ used both in the transfer meaning and in more grammaticalized complex verb constructions, with both inanimate and animate themes: (42) ak-’be-0 laj junuk ek ‘Give him one [puzzle piece] too.’ (43) ban ak’-be-0 s-me’ ‘Give her [baby] to her mother.’ BEN also occurs in the input with inanimate BEN arguments: (44) s-xat’-oj-be-0 y-akan ‘He has broken its wheel [of toy cart] for it.’ with explicit locative arguments: (45) y-ak’-oj-be-0 ta patna ‘He has put it [her lolly] for her at the back of the house.’ (46) chajp ya x-ba x-tuch-be-n ta s-be k-ik’ ‘It’s weird how it pinches me at my throat.’ [lit. ‘at its-path my breath’] [he has licked the bubble liquid] and with complex verbs:

Culture-specific influences on semantic development 139

(47) ak’-be-0 laj s-tik’ give/put-BE-3ABS QUOT 3ERG-insert ‘Let her insert (it).’ [puzzle piece] (48) ma me ‘w-ak’-be-0 s-ta-0 ajch’al NEG PT 2ERG-give/put-BEN-3ABS 3ERG-reach dirt ‘Don’t let it touch the dirt.’ [ puzzle piece] (49) la

y-ak’-be-n k-uch’ 3ERG-give/put-BEN-1ABS 1ERG-drink ‘She gave it to me to drink.’ [corn gruel] CMP

Complex verbs formed with roots other than ‘give/put’ also occur with BEN in the input: (50) pas-be-n k-il i do/make-BEN-1ABS 1ERG-see DEIC ‘Do it for me for me to see.’ An additional 16 other verbs were used once only, including verbs meaning things like ‘set-down-upright on top of something’, ‘know’, ‘clap (hands)’, ‘make-enter’, ‘copy (imitate)’, ‘mix in hand((corngruel)’, ‘burn’, ‘bite’, ‘order (something to be brought from far away)’, ‘break (hard solid thing)’, ‘tear (rope/cloth)’, ‘make-better’, ‘see’, ‘pour (liquid)’, ‘spill’, and ‘pay for’. Another feature of the input is that the same verb is sometimes used contrastively in both the BEN construction and in a transitive construction in the same context, for example: (51) a. CAL: cha’-lek-ub-tes-a xan tal. ‘Make it better again.’ [bubble liquid] b. CAL: lek-ub-tes-be-n tal i. ‘Make it better again for me.’ Thus there are contexts in the input where the special function of BEN – as taking the perspective of affective significance for an event participant – is highlighted, making it perhaps salient for the children. In short, even from this small sample we can see that the input features the BEN construction used with a wide variety of verbs in a range of constructions. It seems to provide ample opportunity for the child to identify a range of event types to which it applies, make the connections across events

140 Penelope Brown involving doing someone or something good (or harm), having an effect on their possessions (including transfers), and having an effect on their state of mind (telling and asking, for example). This in turn should enable them to delimit the domain for the three-argument construction in Tzeltal. 4.5. Significance of Tzeltal children’s BEN usage Several things are clear from the data. First, BEN is early – despite massive argument ellipsis (children never express all three arguments overtly at this age of 2;0-2;8) the children are using verbs in a three-argument construction. BEN shows some productivity for all three children by 2;4 to 2;8. Secondly, physical transfer is by no means the only meaning for verbs used with BEN, even at age 2;0. Thirdly, BEN develops: children use it initially with a few verbs and a strong emphasis on lst person; by age 2;6 they have many verbs and several persons and they continue to add verbs throughout the data sampled. Finally, it is noteworthy that BEN is accompanied by some of the most complex syntax the children show at each of the ages sampled. For example, Xan, age 2;2, can be seen struggling for the syntax to express all three arguments: (52) y-ixlan-be-t laso antun i, ... y-ixlan-be-t laso tat antun i, laso ja’at i. Antun plays_with for_you rope, … Antun plays_with for_you rope father, your rope. [telling her father, tattling on her brother Antun’s misdemeanor] BEN is a site for syntactic complexity also in the data at older ages; for example, Xan age 3;0 produced the following: (53) ya ba-n k-ak’-be-0 xan s-lo’-ik ICP (ASP)-go-1A 1ERG-give-BEN-1ABS again 3ERG-eat-3ERG-PL i matz mut i DEIC corngruel chicken DEIC ‘I’ll go give the chicken(s) corngruel again for them to eat.’ (54) la’ j-bul-be-tik y-ixim-ik i Come 1ERG-pull_up-BEN-1ERG-PL 3ERG-corn-3ERG-PL DEIC a’-me’jun i xi-e 2ERG-aunt DEIC said-CL ‘Come let’s pull out for her their corn your aunt,’ she said.

Culture-specific influences on semantic development 141

In these examples we have a hint that the semantic motivation to use BEN – talking about things in relation to an affectee – may be an important prod to children’s syntactic development. 5.

Discussion

We have seen that Tzeltal children are using the BEN construction with a variety of verbs even at the two-word stage. What are the children using the BEN construction for? 5.1. Illocutionary point of BEN Given its good/bad-affectee meaning, BEN is particularly useful, and often used, to do things like complain, accuse, tattle, plead, or request things to do with one’s belongings, or with services or help. These are speech acts small children are prone to, and are often subjected to by others. In speech to small children, the BEN construction is very often used in imperatives, warnings, threats, and in offers of food or assistence. It is therefore quite frequent in the input, although of course not nearly as frequent as transitive or intransitive constructions (Brown 1998b). It is also used in some cultural routines, for example, the ‘scarecrow’ routine illustrated in (55), where a brother (CAN, age 5;0) teases his 2;2-year-old sister XAN about the bad things the imaginary scare-monster ‘scarecrow’ did or will do to her, and she joins in the game: (55) [xan7feb95] CAN: la y-ak’-be-t nujkul. *XAN: *CAN: *XAN: *CAN: *XAN: *CAN: *XAN: …

‘He gave-you ‘leather’.’ [i.e., a beating] laj. ‘He did.’ la s-k’as-be-t a’-nuk’. ‘He broke-you your neck.’ ju’uk. jaej? ‘No. Huh?’ ju’uk laj. ‘No, it is said.’ [telling her what to say] la -k’ech’-be -nuk’ alal. ‘He twisted-off the doll’s neck for her.’ la -k’ok-be -nuk’ alal. ‘He broke-off her neck for her.’ jo’. ‘Yeah.’

142 Penelope Brown *CAN: la -poj-be-t ch’um ek. la -poj-be-t ch’um. *XAN: laj. *CAN: sok jelol ek. *XAN: sok. *CAN: lot ek’i. *XAN: -boj-be-n -k’ab xutax. *CAN: la -boj-be-t -k’ab xutax. *XAN: jo’ *CAN: aj. la -poj-be-t bel -tzek *XAN: laj. *CAN: aj. *XAN: laj. etc.

‘He stole-you squash too.’ ‘He stole-you squash.’ ‘He did.’ ‘With your namesake, too.’ ‘With.’ ‘It’s a lie, too.’ ‘He cuts-me my hand, the scarecrow.’ ‘The scarecrow cut-you your hand.’ ‘Yeah.’ ‘He stole-you (your) skirt.’ ‘He did.’ ‘Ah.’ ‘He did.’

These kinds of teasing routines are a characteristic feature of Tzeltal childrearing (Brown 2002), and provide ample scope for practice with the BEN construction. 5.2. What is the prototype? If we ask what are children’s favored meanings for verbs used in BEN in the sampled data, do we find ‘give’ the predominant meaning at any stage? Note that ak’ ‘give/put’ is by far the most frequent verb in the language likely to get used in this construction at this age (in terms of frequency it’s one of the top ten verbs in the input; Brown 1998b). On grounds of frequency alone, therefore, we would expect it to predominate in this construction, and indeed, it seems that ak’ with BEN achieves productivity (in the sense of occuring in all three persons) marginally earlier than other verbs. Yet ‘give’ as a prototype should, according to Shibatani (1996), have as part of its meaning that the recipient exercises an element of possessive control over the theme (see (16) above). That does not seem to be the case across the range of verbs used by children in this construction (nor indeed for the adults).15 Looking in each sample for the range of favored meanings expressed in with BEN, for each child (see Table 5 below), we find that, although the three children differ in the verbs they use, all of them use nontransfer verbs in the construction from the beginning of our samples around

Culture-specific influences on semantic development 143

age two. It’s hard to imagine, given the range of verbs represented in Table 5, that (à la Gleitman 1990) children insert ‘transfer’ into the semantic entry for each verb. Table 5. Range of verb types used in BEN construction by the 3 children (numbers after verb indicate tokens in data sampled; verbs are listed in decreasing order of frequency of use by the children) LUS

XAN

MIK

Age 2

puk’ ‘mix by hand’ (3) poj ‘steal from’ (15) pas ‘do/make’ (2) chuk ‘tie’ (2) ak’ ‘give’ (3) ak’ ‘give’ (2) ak’ ‘give’ (2) tzak ‘take in hand’ (2) k’ej ‘put away’ chol ‘exchange (clothes)’ muk ‘cover with cloth’ (2) nup ‘blow on’ (2) al ‘tell’ ixlan ‘play with’ (2) kus ‘wipe’ poj ‘steal from’ (2) ixlan ‘play with’ ti’ ‘bite’ (2) lo’ ‘eat’ boj ‘cut’ nit ‘pull on’ koj ‘knock against’ tij ‘push’ ich’ ‘receive’ pas ‘do/make’ k’as ‘break’ k’ech’ ‘twist off’

Age 2;6

pas ‘do/make’ (4) ich’ ‘get’ (4) kuch ‘carry’ (2) potz’ lok’el ‘peel off’ k’as ‘break’ k’an ‘want/ask for’

tilp’un ‘untie’ (4) k’ej ‘put away’ tzak ‘grasp’ butz ‘kiss’ kus ‘wipe’ tzot ‘twist’ al ‘tell’ k’an ‘want’

chuk ‘tie’ (13) poj ‘steal from’ (3) jol ‘pierce’ (3) tij ‘push’ tzak ‘grasp’ utziy ‘smell’ kus ‘wipe’ sut’ ‘tightly bind’ al ‘tell’

Age 3;0

man ‘buy’ (8) chuk ‘tie’ (3) tzis ‘sew’ ch’oj ‘throw’ mal ‘spill’ pak’ ‘stick on’ jip ‘throw’ t’otz ‘uproot’

otzes ‘insert’ (6) ch’ol ‘pour’ (3) il ‘see’ (3) set ‘cut’ (3) ch’ay ‘drop’ (2) lok’es ‘extract’ (2) jom ‘pierce’ (2) jut ‘pierce’ (2)

tij ‘push’ (5) man ‘buy’ (5) jul ‘pierce’ bon ‘paint’ il ‘see’ lok’es ‘extract’ ich’ ‘get’ mal ‘spill’

144 Penelope Brown Table 5, cont.

Age 3;6

uch’ ‘drink’ pech’ ‘braid’

jep ‘cut in half’ jol ‘open’ xi’ ‘fear’ k’iy ‘spread out to dry’ t’om ‘explode’ bul ‘pull up by roots’ k’am ‘roll up’ jitz’ lok’el ‘slide out’

lot’ ‘insert’ toy ‘raise’ jam ‘open’ lok’ ‘extract’ ch’ak ‘hire’ poch ‘peel’ jut ‘pierce’

tuch’ ‘hold up’ (4) le ‘look for’ (3) toch ‘peel’ (3) kotz’ ‘throw down’ (2) busk’in ‘spill’ xat’ ‘divide’ tz’ista ‘fart’ il ‘see’ tz’ibu ‘write on’ tzul ‘slide’ katz’ ‘chew’ jis ‘scrape’

pech’ ‘braid’ (3) toy ‘raise’ (2) lap ‘put on clothes’ xij ochel ‘insert’ jot’ ‘scratch’ tek’ ‘step on’ sew ‘cut’ balan ‘roll’ jux ‘file’ jip ‘throw’ kem ‘dig’

nit ‘pull’ (4) latz ‘stack’ (2)

6. Conclusion While children’s very first BEN forms may well be unanalyzed wholes (e.g., pojben ‘steal-from-me’, tzakben ‘grab-from-me’, chukben ‘tie-for-me’), and initial acquisition may proceed on a verb by verb basis á la Tomasello’s (1992) ‘verb islands’, Tzeltal children readily acquire wide use of the BEN construction very early, within a few months of the child’s first morpheme combinations at the two-word stage (age 24 to 30 months). There is no evidence that Tzeltal children have trouble expressing threeparticipant events; it seems likely that a dedicated construction with simple morphology makes the task easier. They do not utilize alternative solutions to expressing a third argument – for example with adjuncts – prior to tackling the benefactive construction. In general, transitivity and argument structure appear on the basis of data examined so far, to be relatively easy for Tzeltal children; they very rarely make transitivity errors (Brown 1998a), despite the fact that arguments are often ellipsed.

Culture-specific influences on semantic development 145

Children do not all show a preference for transfer meanings initially when using the BEN construction; (potential) transfer of possession is not the core of the constructional meaning. Rather, its meaning for them is something like: NP1 does something to NP3 to help/hurt NP2 (initially, usually ‘me’). This is suggestive evidence that the culture-specific use of the BEN construction leads to early culture-specific meanings for the construction – not ‘give’ or ‘transfer’, but a much more general meaning along the lines of ‘help/hurt Recipient/Affectee’. These are culture-specific in the sense that what ‘counts’ in the decision to use BEN is shaped by cultural knowledge and values. Appropriate use of the BEN construction requires cultural knowledge: what constitutes things you would want to happen as opposed to things that are not desirable. Many, but certainly not all of these, are universal. Thus, while we may presume that parents and children everywhere would agree it was undesirable to injure their bodyparts or steal things from each other, much more culture-specific knowledge is necessary to assess the pros and cons of other event types expressed using BEN by these Tzeltal children (‘play with you your rope, pull you your shirt, tie/untie my car (etc) for me…, tie-it-tightly-up for him (head, for headache), eat-me my banana, eat her (younger sister) it (banana), bite-him (cat) his fur), smell-him it, see you your-crotch!’ and ‘Come pull-out-ofthe-ground for your aunts their corn’). There is considerable cultural support available in caregiver’s speech and behavior for learning the Tzeltal BEN construction. Not only do the things naturally expressed by the BEN construction (offers, threats, requests) tend to be culturally important kinds of speech acts for children, but also, children are early drawn into caring for other children, so they are often asked to do things for others. They also routinely bait each other with what they – and others – will do to each other. Early access to this meaning is supported by specific cultural practices, including routine control threats (Brown 2002), and other cultural routines like the scarecrow one illustrated in (55) above. Furthermore, since there is no alternative phrasing possible (no possibility of saying something equivalent to ‘I bought the book FOR Mary’, utterances with BEN are reasonably frequent in the input. Perhaps if you have only one way of talking about a third participant, then you are more likely to generalize the construction rapidly. Is there any relation between how the Tzeltal children are construing events with the BEN construction and how children express three-participant events in other languages? There is evidence from Billington (2002)

146 Penelope Brown that English-learning children dissociate dative and benefactive constructions from a very early age, despite the fact that both can be expressed in the same two ways in English – either by a double object construction (give him the book, make him a cake) or a prepositional object construction (give a book to him, make a cake for him), indicating semantic priority for the benefactive.16 Demuth (1998, 2003) finds that 2-to-3-year-old children learning Sesotho productively use the applicative morpheme (of which one use is benefactive), and alternate with the same verb in applicative and nonapplicative uses, showing some productive knowledge at an age comparable to that of the Tzeltal children examined here. Of course, to explain exactly what is going on with the Tzeltal children’s use of BEN, further questions need to be addressed in future work: i. What is the precise relation between the children’s verbs and input verbs used in BEN: Is the children’s usage exactly mirroring the adult input in this construction? ii. Do children use the same verb root in both transitive and ditransitive constructions at the same time point? (i.e., do they have a basic twoplace or three-place meaning for each verb?) Do transitive uses precede ditransitive uses for each verb, or for any of the verbs? The present paper takes the first step, establishing that Tzeltal children use BEN with a wide range of verbs and with a culture-specific meaning at an early age, well before age 3. These facts, and indeed the very existence of the construction they are based on – a three-argument construction which adds an extra argument to two-argument verbs (assuming we can independently establish verb transitivity, and provided the construction is frequent in the input) – suggests that there may not actually be any universal prototype for three-argument constructions. If virtually all transitive verbs can be construed ditransitively, then the fact that a verb occurs in the ditransitive construction cannot tell you anything about that particular verb’s meaning. Children, then, have to learn how to take the relevant perspective, and what the local conventions are for adding a third participant when expressing an event.

Culture-specific influences on semantic development 147

Notes 1. This paper is based on a talk presented at the symposium on ‘Languagespecific influences on early semantic and cognitive development’ at the Vth conference of the International Association for the Study of Child Language, Madison, Wisconsin, in July 2002. An analysis of part of the same data for two children was reported in Brown (1998b). The paper’s current formulation also owes much to discussions and plans for a workshop in May 2003 at the Max Planck Institute for Psycholinguistics, which investigated the linguistic encoding of three-participant events from a crosslinguistic and developmental perspective (cf. Brown, Eisenbeiss & Narasimhan 2002; Narasimhan, Eisenbeiss & Brown 2007). I am grateful to Bhuvana Narasimhan and to an anonymous reviewer for extremely helpful comments on a first draft of the paper. 2. Tzeltal transcription conventions are based on a practical orthography; symbols correspond roughly to their English equivalents except that j = h, x = sh, and ’ indicates a glottal stop or glottalization of the preceding consonant. Abbreviations for glosses are as follows: 1,2,3 ERG indicates 1st, 2nd, 3rd person ergative prefixes (which mark both subjects of transitive verbs and noun possessors), 1,2,3 ABS indicates the corresponding absolutive suffixes, 1PLE is lst person plural exclusive, 1PLI = lst person plural inclusive, PL = 2nd or 3rd person plural, ASP = neutral aspect, = BEN benefactive suffix -be, CMP = completive aspect, ICP = incompletive aspect, PREP = preposition. 3. In none of the child data under examination were there any examples of all three arguments being overtly expressed, and this is rare in adult speech too (Brown, in press). Note that Tzeltal should present a learning problem analogous to that presented by Sesotho (Demuth 1998, 2003), since word order differs depending on the status of the Agent and Patient arguments on the animacy hierarcy (if A = P the order is VAP; if A is higher than P the order is VPA (Dayley 1981: 43, 1990; Robinson 2002). This issue is not pursued here, as it would require a different methodology, including eliciting utterances with explicitly expressed arguments from children. 4. Kaufman (1971) calls it ‘indirective’; Slocum (1948: 85) calls it ‘benefactiveindirective,’ Monod-Becquelin (1997) calls it ‘benefactive’. Dayley (1981: 44, 1990) calls it a ‘referential voice’. Berlin (1961ms: 59) says that ‘it may occur with all transitive bases to indicate a second object other than the object inherent in the transitive verb base.’ In work on the closely related Mayan language Tzotzil, Haviland (1988) calls it the benefactive or ditransitive suffix; Aissen (1974, 1987) calls it indirect object or ditransitive. 5. In the child examples, an initial hyphen – indicates missing ergative prefixes (cross-referencing Agent or Possessor); ergative and absolutive cross-referencing on the verb are obligatory in adult Tzeltal. An initial underline in front of the verb indicates a missing aspect marker, which is also obligatory.

148 Penelope Brown 6. One major exception to this is the English dative alternation, which has received more than its share of attention (cf. Osgood & Zehler 1981; Pinker 1989; Gropen, Pinker, Hollander, Goldberg & Wilson 1989; Goldberg 1995; Snyder & Stromswold 1997; Billington 2002). 7. In Tzotzil, a closely related Mayan language, BEN can apparently promote an instrument as well (Dayley 1981: 42): Ta-0-s-paj-b’e akuxa ti ka7e ASP-ABS3-ERG3-prick-INSTR needle the horse ‘He pricked the horse with a needle.’ Since Dayley could not find a comparable example for Tzeltal, nor have I found any examples of instrument advancement in my data, this possibility is not considered further here. 8. I write these BEN suffixes as two morphemes – with -be for BEN and then the ABS suffixes – a decision required by details of plural morpheme ordering, although they are probably not taken as two distinct morphemes by young children. 9. It is not however, contra Dayley (1981: 44), obligatory, as the examples show. Note that the relation between possessive and benefactive is also crosslinguistically frequent, and provides the basis for grammaticalization route (Margetts 2004; Payne & Barshi 1999) 10. In the ‘put’ sense you can definitely use ak’ ‘give, put’ without BEN (e.g., ak’a ta mexa ‘put it on the table’). The root ak’ has many light uses, and although it does occur in transitive sentences (i.e., without BEN), further linguistic analysis would be necessary to decide whether all the non-BEN uses of ak’ have a meaning distinct from plain ‘give’. Many other inherently three-argument verbs (with meanings like the English verbs show, feed, etc.), are complex verbs constructed with ak’: e.g., ak’be yil ‘make him see, show him’, or ak’be slo’ ‘give/make him to eat’). 11. Native-speaker judgements here are based on elicitation with one experienced male consultant, age 48. Crosslinguistic evidence suggests there may well be variability in such native-speaker assessments for three-place predications. 12. Mayanists differ as to whether they analyze the BEN suffix singular forms into two morphemes, since the ending is transparently related to that of the absolutive suffixes and BEN is historically derived from proto-Mayan *-be + absolutive suffixes. In Tzotzil and Tzeltal it is usually written as two suffixes, -be + -ABS, although (as mentioned above) synchronically the single forms are probably taken to be a single suffix by learners. 13. See O’Connor 1996 for an analogous argument for Northern Pomo possessorraising) 14. In connection with a large-scale comparative project on spatial language and its acquisition, I videotaped six-weekly longitudinal samples of natural interaction from children in five Tzeltal families, focussing on children from around the age of 1;6 to 4 or later. I also had the parents tape-record monthly samples

Culture-specific influences on semantic development 149 between Sept. ‘94 and July ‘98. This data was collected between the years of 1993 and 1998 in the community of Tenejapa, in Chiapas Mexico. Analyses of other parts of this child language database are reported in Brown 1997, 1998a, b, c, 2001, 2002, 2004, in press; and in Brown & Levinson 2000. The structure of the database can be seen by consulting the Max Planck Institute for Psycholingistics Browsable Corpus. [www.mpi.nl] Pioneering work on Tzeltal child language was carried out two decades earlier in the same community by Brian Stross (e.g., Stross 1970a/b, 1972, 1973). 15. Shibatani (1996: 191) actually suggests strict criteria for showing that his proposed schema is false, for a given language. One must show (1) that benefactive constructions are structurally different from the ‘give’ construction (which they are not, in Tzeltal), and (2) that benefactive constructions obtain more easily for situations remote from what is described by the ‘give’ situation than for ones that are closer to the schema. This too is probably not the case for Tzeltal. 16. Pye (p.c.) summarizes Billington’s (2002) study of the acquisition of 3-place predicates in English as follows: Billington extracted data from the CHILDES samples analyzed by Pinker & Stromswald, and found that both of these investigators had failed to separate the dative and benefactive constructions. Gropen et al. (1989) claim that children acquire both alternations of the dative at the same point, while Snyder & Stromswald (1997) claim that children acquire the double object form first. Billington separated the datives from benefactives and showed that children acquire the double object form of the dative first, but the prepositional form of the benefactive first. She also found a discontinuity between the children’s use of 2-place forms and 3-place forms of the same verbs. The children use 2-place forms of the verbs long before they use 3-place forms.

References Aissen, Judith L. 1983 Indirect object advancement in Tzotzil. In Studies in relational grammar, David M. Perlmutter (ed.), 272–302. Chicago: University of Chicago Press. 1987 Tzotzil clause structure. Dordrecht: Reidel. Austin, Peter K., Nick Evans, Anna Margetts & John Bowden 1999 Three-place predicates in the languages of the world. 2000 ARC Small Grant Application, Australian Research Council. Berlin, Brent 1961 The Tenejapan dialect of Tzeltal: A sketch of morphology. Unpublished Ms., Dec. 1961.

150 Penelope Brown Billington, Catherine E. 2002 To, for, six, eight: Patterns in the acquisition of dative and benefactive verbs in English. Unpublished PhD dissertation, University of Kansas. Bowerman, Melissa 2002 Language-specific semantic puzzles in first language acquisition. Talk presented at the 5th IASCL Conference, Madison, WI, July 2002. Bowerman, Melissa & Penelope Brown (eds.) in press Crosslinguistic perspectives on argument structure: Implications for learnability. Mahwah, NJ: Erlbaum. Bowerman, Melissa, Penelope Brown, Sonja Eisenbeiss, Bhuvana Narasimhan & Dan Slobin 2002 Putting things in places: Developmental consequences of linguistic typology. CLRF plenary session, April 12, 2002. In Proceedings of the 31st Annual Child Language Research Forum, Stanford, 2002, Eve V. Clark (ed.). Available from the CSLI Publications: http://cslipublications.stanford.edu/CLFR/2002/CLRF-2002-toc.html Bowerman, Melissa & Soonja Choi 2001 Shaping meanings for language: Universal and language-specific in the acquisition of spatial semantic categories. In Language acquisition and conceptual development, Melissa Bowerman & Stephen C. Levinson (eds.), 475–511. Cambridge: Cambridge University Press. Brown, Penelope 1997 Isolating the CVC root in Tzeltal Mayan: A study of children’s first verbs. In Proceedings of the 28th Annual Child Language Research Forum, Eve V. Clark (ed.), 41–52. Stanford, CA: CSLI/University of Chicago Press. [Republished in Spanish as: Brown, P. 1998. La identificación de las raíces verbales en tzeltal (maya): Cómo lo hacen los niños? Función 17–18: 121–146.] 1998a Early Tzeltal verbs: Argument structure and argument representation. In Proceedings of the 29th Annual Child Language Research Forum, Eve Clark (ed.), 129–140. Stanford, CA: CSLI. 1998b Children’s first verbs in Tzeltal: Evidence for an early verb category. In Elena Lieven (ed.), Special edition of Linguistics 36 (4):713–53. 1998c Conversational structure and language acquisition: The role of repetition in Tzeltal adult and child speech. Journal of Linguistic Anthropology 8(2): 197–221. 2001 Learning to talk about motion UP and DOWN in Tzeltal: Is there a language-specific bias for verb learning? In Language acquisition and conceptual development, Melissa Bowerman & Stephen C. Levinson (eds.), 512–543. Cambridge: Cambridge University Press. 2002 Everyone has to lie in Tzeltal. In Talking to adults, Shoshana BlumKulka & Catherine Snow (eds.) 241–275. Mahwah, NJ: Erlbaum.

Culture-specific influences on semantic development 151 2004

Position and motion in Tzeltal frog stories: The acquisition of narrative style. In Relating events in narrative, Vol. 2, Sven Strömqvist & Ludo Verhoeven (eds.), 37–57. Mahwah, NJ: Erlbaum. in press Verb specificity and argument realization in Tzeltal child language. In Crosslinguistic perspectives on argument structure: Implications for language acquisition, Melissa Bowerman & Penelope Brown (eds.). Mahwah, NJ: Erlbaum. Brown, Penelope, Lourdes de León, Barbara Pfeiler & Clifton Pye 2002 The acquisition of agreement in Maya. Symposium presented at 9th IASCL Conference, Madison, WI, 16–21 July, 2002. Brown, Penelope, Sonja Eisenbeiss & Bhuvana Narasimhan 2002 Proposal for Workshop The linguistic encoding of three-participant events: Crosslinguistic and developmental perspectives, May 2003, Max Planck Institute for Psycholinguistics, Nijmegen, NL. Brown, Penelope & Stephen C. Levinson 2000 Frames of spatial reference and their acquisition in Tenejapan Tzeltal. In Culture, thought and development, Larry P. Nucci, Geoffrey B. Saxe & Elliott Turiel (eds.), 167–197. Mahwah, NJ: Erlbaum. Choi, Soonja 2002 The relation between semantics and cognition: A crosslinguistic study of spatial categorization from infancy to adulthood. Talk presented in the symposium Language-specific influences on early semantic and cognitive development, 5th IASCL Conference, Madison, WI, July 2002. Clark, Eve V. 1991 Acquisition principles in lexical development. In Perspectives on language and thought, S. A. Gelman & J. B. Bynnes (eds.), 31–71. Cambridge: Cambridge University Press. Cook, Vivian J. 1976 A note on indirect objects. Journal of Child Language 3: 435–437. Dayley, Jon P. 1981 Voice and ergativity in Mayan languages. Mayan Linguistics 2 (2): 1–82 (Ms.). 1990 Voz y ergatividad en idiomas mayas. In Lecturas sobre la lingüística maya, Nora C. England & Stephen R. Elliott (eds.), 335–398. La Antigua, Guatemala: Centro de Investigaciones Regionales de Mesoamérica. Demuth, Katherine 1998 Argument structure and the acquisition of Sesotho applicatives. Linguistics, 36 (4): 781–806. 2003 Frequency effects on learning the syntax of Sesotho three-place predicates. Talk delivered at the workshop The linguistic encoding of

152 Penelope Brown three-participant events: Crosslinguistic and developmental perspectives, May 2003, Max Planck Institute for Psycholinguistics, Nijmegen, NL. Demuth, Katherine Ms. Syntactic and semantic productivity in the learning of three-place predicates. Unpublished Ms., Brown University. Dryer, Matthew 1986 Primary objects, secondary objects and the antidative. Language 62: 808–845. Eisenbeiss, Sonja & Ayumi Matsuo 2003 Encoding more than two participants in German and Japanese child language. Talk given at the workshop Linguistic Encoding of ThreeParticipant Events: Crosslinguistic and Developmental Perspectives. Max Planck Institute for Psycholinguistics, May 2003. Gleitman, Lila 1990 The structural sources of verb meanings. Language Acquisition 1 (1): 1–55. Goldberg, Adele 1992 The inherent semantics of argument structure: the case of the English ditransitive construction. Cognitive Linguistics 3 (1): 37–74. 1995 Construction grammar. Chicago: University of Chicago Press. Gropen, Jess, Steven Pinker, Michelle Hollander, Richard Goldberg & Ronald Wilson 1989 The learnability and acquisition of the dative alternation in English. Language 65: 203–257. Haviland, John, B. 1988 It’s my own invention: a comparative grammatical sketch of Colonial Tzotzil. In The Great Tzotzil Dictionary of Santo Domingo Zinacantán, Robert B. Laughlin (ed.), 79–121. Kaufman, Terrence 1971 Tzeltal phonology and morphology. University of California Publications Linguistics 61. Berkeley, CA: University of California Press. Margetts, Anna 2004 From implicature to construction: Emergence of a benefactive construction in Oceanic. Oceanic Linguistics 43 (2): 445–468. Margetts, Anna & Peter Austin 2007 Three-participant events in the languages of the world: Towards a cross-linguistic typology. Linguistics 45 (3): 393–451. Monod-Becquelin, Aurore 1997 Parlons Tzeltal. Paris: J. Harmitton Narasimhan, Bhuvana, Sonja Eisenbeiss & Penelope Brown 2007 ‘Two’s company, more is a crowd’: The linguistic encoding of multiparticipant events. In Encoding multi-participant events, B. Narasim-

Culture-specific influences on semantic development 153 han, S. Eisenbeiss & P. Brown (eds.), special issue of Linguistics 45 (3): 383–292. Newman, John 1996 Give: A cognitive linguistic study. Berlin /New York: Mouton de Gruyter. Newman, John 2002 Culture, cognition and the grammar of ‘give’ clauses. In Ethnosyntax: Explorations in grammar and culture, Nick J. Enfield (ed.), 74– 95. Oxford: Oxford University Press. Newman, John (ed.) 1997 The linguistics of giving. Amsterdam: John Benjamins. O’Connor, Catherine 1996 The situated interpretation of possessor-raising. In Grammatical constructions, Masayoshi Shibatani & Sandra Thompson (eds.), 125–156. Oxford: Clarendon. Osgood Charles & Annette Zehler 1981 Acquisition of bi-transitive sentences: Pre-linguistic determinants of language acquisition. Journal of Child Language 8: 367–384. Payne, Doris L. & Immanuel Barshi (eds.) 1999 External possession. Amsterdam: Benjamins. Pinker, Steven 1989 Learnability and cognition: The acquisition of verb argument structure. Cambridge, MA: MIT Press. Pye, Clifton 1985 The acquisition of transitivity in Quiche’ Mayan. Papers and Reports on Child Language Development 24: 115–122. 1990 The acquisition of ergative languages. Linguistics 28: 1291–1330. 1992 Acquisition of K’iche’ Maya. In The crosslinguistic study of language acquisition, Vol. 3, Dan Slobin (ed.), 221–308. Mahwah, NJ: Erlbaum. Robinson, Stuart 2002 Constituent order and voice in Tenejapan Tzeltal (Mayan). International Journal of American Linguistics 68: 51–81. Shibatani, Masayoshi 1996 Applicatives and benefactives: A cognitive account. In Grammatical constructions, Masoyoshi Shibatani & Sandra Thompson (eds.), 157– 194. Oxford: Clarendon. Slobin, Dan I. 1985 Crosslinguistic evidence for the language-making capacity. In The crosslinguistic study of language acquisition, vol.2: Theoretical Issues, Dan I. Slobin (ed.), 1157–1256. Hillsdale, NJ: Erlbaum. Slobin, Dan I. (ed.) 1985/ The crosslinguistic study of language acquisition, Vols. 1–5. Hills92/97 dale/Mahwah, NJ: Erlbaum.

154 Penelope Brown Slobin, Dan I., Melissa Bowerman, Penelope Brown, Sonja Eisenbeiss & Bhuvana Narasimhan in press Putting things in places: Developmental consequences of linguistic typology. In Event representations in language and cognition, Jürgen Bohnemeyer & Eric Pederson (eds.). Cambridge: Cambridge University Press. Slocum, Marianne 1948 Tzeltal (Mayan) noun and verb morphology. International Journal of American Linguistics 14: 77–86. Snyder, William & Karin Stromswold 1997 The structure and acquisition of English dative constructions. Linguistic Enquiry 28: 281–317. Stross, Brian 1970a Aspects of language acquisition by Tzeltal children. Unpublished PhD dissertation, University of California, Berkeley, CA. Stross, Brian 1970b Acquisition and componentiality. Papers from the 6th Regional Meeting of the Chicago Linguistic Society, 120–128 Chicago: Chicago Linguistic Society. 1972 A Tzeltal child’s acquisition of botanical terminology. In Meaning in Mayan languages, Munro Edmonson (ed.), 107–141. The Hague: Mouton. 1973 El contexto sociocultural en la adquisición de la lengua tzeltal. Estudios de Cultura Maya 9: 257–302. Tomasello, Michael 1992 First verbs. Cambridge: Cambridge University Press.

Bcuaa quiang – I stepped HEAD it! The acquisition of Zapotec bodypart locatives Kristine Jensen de López

1. Introduction1 Observations of children’s early language across different language groups have suggested that nouns are the very first word class children comprehend and produce (Gentner 1982). The primary function of nouns is often taken to be that of referring to a specific concrete object. However, in many languages (African, Mayan, Otomanguean languages and Thai) a subset of nouns is employed to refer to the designation of static spatial relations2 between two objects. In such languages this subset of nouns serves a purpose similar to that served by prepositions in English. Thus nouns are used to indicate where one object is in relation to the location of a second object. Thus, languages differ concerning whether they rely on prepositions for locating things and people or whether they rely on nouns for locating things and people. Children acquiring languages, which rely on prepositions for designating static spatial relationships produce in, on, up, down among their first 50 words, while between, back and front appear later. However, children do not always use such early words to refer to a static spatial relationship only (Brown 1973; Johnston & Slobin 1979; Tomasello 1987; Sinha et. al. 1994; Rohfling 2002). During the one-word stage of their language acquisition children may use the set of prepositions or locative words as verb particles, which then gradually become employed in the way adults use prepositions, namely to refer to spatial relationships (Tomasello 1987). Tomasello (1992) points out that words, which by adults are used to name places (prepositions in English), actually function as action requests for young children, and hence he suggests that such location words function as sentencestructuring verbs for children in their early language acquisition. In a dairy study of his English-acquiring daughter, Tomasello found that at the age of 16:26 months she produced ‘down’ as a request to be put down from her parent’s arms, at 17:25 months she produced ‘up-here’ as a request for help

156 Kristine Jensen de López to get up into her high chair. At the age of 18:13 months, the utterance ‘shoes-on’ was used in a locative situation while she was putting on someone’s shoes, and also ‘in’, was uttered as ‘put-it-in’ for location while commenting on a spoon in a cup. Finally at 19:16 months the child produced the locative utterance ‘under chair’, while she was putting a chair under the table, suggesting a locative use of a preposition. In Johnston & Slobin’s crosslinguistic study they explain the variation in the order of acquisition of spatial concepts to result from the linguistic complexity of the means available in the particular languages (Johnston & Slobin 1978). They point out that if a word is morphologically complex or if is homonymous, it may be acquired later than if this is not the case. Bowerman & Choi’s comparison of English- and Korean-acquiring children’s early language used for referring to dynamic spatial relationships showed two interesting differences across the two languages. First, due to the different language typologies English acquiring children employed prepositions, while Korean acquiring children employed verbs to identical spatial relationships, and second the two groups of children seemed to categorize the elicted set of dynamic relationships in different ways (Bowerman & Choi 2001). Hence the two language groups seem to have conceptualized the spatial relationships differently in accordance with the linguistic partitioning available in the particular language. The case of San Marcos Zapotec3 bodypart nouns which are employed to designate static spatial relationships presents yet another set of grammatical items with a different set of underlying conceptualisations, which varys from how a language that employs prepositions partitions the spatial array. This presents the Zapotec-acquiring children with a different type of task in acquiring their native language than is the case for children acquiring languages with prepositions. The comparison of children’s acquisition of locative items across typological different languages provides rich information concerning exactly which prelinguistic concepts children rely on for acquiring language, and rich information on the variety of pathways children have access to in learning their native language. Both are important questions raised by researchers investigating children’s acquisition of their first language. The present paper first describes the language task involved for children acquiring Zapotec, an Otomaguean language spoken in the south Mexican state of Oaxaca, and secondly presents a longitudinal case study of spontaneous production of Zapotec bodypart locatives. First I present the grammatical structure of the Zapotec bodypart system, which is the only lexem

Bcuaa quiang – I stepped HEAD it!

157

available for designating static spatial relationships – the language does not have prepositions. In continuation I describe the semantic system underlying locative bodyparts in Zapotec, and finally I present longitudinal data from one monolingual Zapotec acquiring boy’s spontaneous use of bodypart locative nouns. 2.

Background information on Zapotec

Zapotec forms part of the Otomanguean language family, which is spoken in the Southern Mexican state of Oaxaca. Zapotecan languages are primarily spoken in the central valleys near Oaxaca City, south from there to the Pacific coast, southeast to the isthmus of Tehuantepec, and northeast into the Sierra de Juarez. It has been estimated that there may exist as many as forty mutually unintelligible variants of Zapotec. It is not clear whether these varieties are ‘just’ dialectic varieties or whether, in some cases, they should more appropriately be described as different languages. Four groups of languages in the eastern Valley region have been identified by the Summer Institute of Linguistics; Eastern Tlacolula Zapotec (San Pedro Quiatoni Zapotec), San Pedro Güilá Zapotec, San Juan Guelavía Zapotec and Mitla Zapotec (Grimes et al. 1996). The specific Zapotec language, which has been analysed in this study is spoken by about 2,000 inhabitants living in a rural agricultural community located in the central valley, about 60 kilometres from Oaxaca City. The basic word order of San Marcos Tlapazola Zapotec is VSO, although SVO is also acceptable. 2.1. Bodypart locative nouns in Zapotec Zapotec languages differ from Indo-European languages in the grammatical forms they employ in referring to static spatial configurations. English, for example, employs prepositions whereas Zapotec relies, to a great extent, on bodypart nouns to designate a static spatial relationship. The same subset of nouns is also used to refer to human bodyparts. Zapotec bodypart locative nouns, hence, consist of nouns or noun-derived items, which are identical to the nouns used for reference to human body parts. This case of homonymous usage, which also involves polysemous usage, might posses a problem for the child in learning the distinction between the meaning of a bodypart noun used to refer to the location of an object as opposed to the

158 Kristine Jensen de López case when a bodypart noun is used to refer to a specific part of the human body4. There are however, some syntactic ‘clues’ distinguishing these two types of references, and which might be hints to the child in acquiring Zapotec. One difference lies in the fact that only the set of literal body parts become suffixed by a possessive pronominal enclitic; such as illustrated in the following three examples: (a) dets-a back-1SG my back

(b) lo-o face-2SG your face

(c) nii-b foot-3SG his foot

However, in some cases both possessed and non-possessed locative bodypart constructions can take on a definite clitic, in which case the utterance becomes ambiguous, as seen in example (1). (1)

nuu dets-a-ng. exist back.region-1SG=clitic ‘it’s on my back / it’s behind me.’

When referring to a static location of one object in relation to another object the employment of bodypart nouns is obligatory (similarly to prepositions in English). Zapotec linguists (see e.g. Butler 1988; Pickett & Black 1995; and Munro & Lopez 1999) classify bodypart nouns, when used with a locative meaning as prepositions or body-part prepositions. Cognitive linguists, on the other hand, have described the way bodypart nouns, when employed as locatives, are motivated by the metaphorical extension of the human body in its upright position, and argue that they are very different than prepositions (MacLaury 1989; Brugman, up.Ms.; Lakoff 1987). Brugman and Lakoff refer to Mixtec (a different Otomanguean language), which, in addition, relies on animal bodypart nouns for locating objects and people. Locative bodypart nouns occur as the first of two contiguous nouns in a noun + noun construction, as illustrated in example (2). (2)

quia yuu. head house. ‘on top of the house.’

Butler (1988), Pickett & Black (1995), Munro & López (1999) propose the existence of two classes of prepositions in Zapotec languages; those ex-

Bcuaa quiang – I stepped HEAD it!

159

pressed with bodypart words and those expressed with loan words derived from Spanish prepositions. One clear difference between these two classes of words is present in the syntactic properties followed by the prepositional phrase of either. Prepositional phrases constructed with bodypart nouns may express the prepositional object with either a possessive pronominal agreement clitic or with an overt noun following the preposition. This is also similar to the syntax followed by verbs in Zapotec. Prepositional phrases expressed with non-body-part spatial words (often Spanish loan words) on the other hand, express their object with a pronoun or an overt nominal. Hence, only bodypart derived spatial words take clitics suggesting that bodypart nouns used as locatives may form a different grammatical category than Zapotec nouns in that they syntactically resemble verb phrases. In addition to the above mentioned distinction, some types of spatial relationship in Zapotec are not expressed by use of prepositiona/bodypart phrases, but are intergrated as part of the verb, eg the applicative verbal suffix -ne, for expressing comitative ‘with’. 2.2. The core group of locative bodypart nouns San Marcos Tlapazola Zapotec relies on a core group of seven bodypart nouns, which are used for locative reference (Jensen de López 1998, 2002). These are illustrated in table 1. Table 1. core group of BODYPART locative grams Zapotec bodypart

Bodypart gloss

Locative gloss

Quia

Head

on, upper.region

Lo

Face

on, at, in front of.region

Ruu

Mouth

on.region, in.region

Láani

Stomach (internal organ)

in,region, under.region, through

Dets

Back

behind.region

Llaan

Butt

beside.region, under.region, on top of.region

Nii

Foot

under.region, lower.region

160 Kristine Jensen de López The fact that only a small core group of bodypart nouns are employed as locatives suggests that this group may form a closed class word group. This notion is consistent with descriptions of other Zapotec languages, however the specific inventory of bodypart nouns forming part of the subclass, as well as the semantics underlying each specific bodypart noun, may vary across languages. A glance at the semantic variation listed for each item under the colon titled locative gloss illutrates the degree of polysemous underlying the meanings of each locative item. 2.3. Word order and the bodypart locative phrase Bodypart locative phrases resemble prepositional phrase structures as illustrated in the examples (3) to (10). Similar to what Zlatev (2003) describes as ‘region nouns’ in his description of Thai, Zapotec bodypart locative nouns may be regarded as region nouns. This categorization is based on the notin that region nouns can be regarded as the heads of the noun phases they appear in (ibid: 16). The noun-like character of region nouns suggests that their meanings can be based on different values of regions. Examples (3) to (10) illustrate static spatial relationships in Zapotec. (3)

b-zu canast lo yiu. CPL-stand basket face.region soil. ‘Put the basket on the ground.’

(4)

b-seu ruu yuu. CPL-close mouth.region house. ‘Close the door.’

(5)

b-diat ra=y dets dxii. CPL-descend PL-PRO=clitic back.region then. ‘take them out off / down from the back region then’ (from the back of a toy animal).

(6)

b-zuub bidy quia yuu. CPL-sit chicken head.region house. ‘put the chicken on the house.’

(7)

b-teed avio-ng quia yuu. CPL-pass airplane=clitic head.region house. ‘pass the airplane over / across the house.’

Bcuaa quiang – I stepped HEAD it!

(8)

gu-lúu bidy láni.iu. CPL-put/in chicken stomach.region.house. ‘put the chicken inside the house.’

(9)

b-zuub bidy dets yallily. CPL-sit chicken back.region chair. ‘put the chicken behind the chair.’

161

(10) b-zuub bidy nii puant. CPL-sit chicken foot.region bridge. ‘put the chicken under the bridge.’ When referring to a static spatial relationship, bodypart nouns occur as the first of two contiguous nouns in a noun + noun construction. Hence, in this particular Zapotec language bodypart locatives do not take any overt genitive marking, although for other languages, for example in Ayoquesco Zapotec, it is suggested that the second noun bears a genitive relationship to the first (MacLaury 1989: 120). Specification of a region with respect to the landmark object may be expressed by reduplication of the bodypart noun, for example the upper region above a tree as in example (11). (11) quia quia yag. head head tree. ‘above the tree.’ As mentioned earlier, Tlapazola locative bodypart nouns can be inflected with a pronominal clitic referring to an implicit landmark object. This class of pronominal clitics are similarly employed with verbs. Bodypart nouns, when used for locative reference, are derivationally nominals and they also largely obey the grammatical rules employed for nouns. However to some extent, they also comply with the grammatical rules applying to verbs. Thus, similarly to what Zlatev (op.cite) reports for Thai and what Lucy (1992) reports for Yucatan Mayan, in these languages the categories of noun and verb appear to be modulus rather than selective. Thus there seems to be an interaction between closed-class (grammatical) and open-class (lexical) expression regarding spatial semantics (Zlatev ibid.). A preliminary analysis of the semantic range of San Marcos Tlapazola bodypart nouns is reported in the following section.

162 Kristine Jensen de López 3. Semantics of Zapotec bodypart locatives Cognitive linguists suggest that the usage of a linguistic system such as that of Zapotec involves the speaker using a conceptual system that is conventionally organised in a different way than is the case for speakers of an Indo-European language (Lakoff 1987). In other words, the semantics of the Zapotec spatial bodypart terms do not coincide with the underlying semantics of the English spatial system. This is illustrated by the fact that the bodypart locative láani (stomach) refers to the spatial configurations expressed by the English prepositions in, under and through. One principle difference between the English preposition in and the bodypart locative láani is that in involves containment and constraint of the object, often combined with the support of the object against gravity. The Zapotec locative bodypart noun láani, on the other hand, despite the possibility of expressing the notion of containment and constraint of the object, it does not involve the notion of support from gravity. For example an object located under a table or chair is referred to using láani as also is the case for an object placed inside a container object. 3.1. Bodypart nouns as a dative and a temporal marker Some bodypart nouns (especially lo) are used to refer to more abstract references. For example the term lo (face) may be employed in the sense of a dative marker. For example in the expression gunii lo laab ‘speak to her’. The following examples (12) illustrate the usage of lo (face) in reference to directed motion and to time. (12) a. ze-dli-ú lo bniin. FUT-go-2SG face.region child. ‘go look to the child.’ b. cuaa beecu lo Lipy. ∅-receive dog face.region Phillip. ‘take the dog away from Phillip.’ c. r.güia biny lo beecu. HAB-look person face.region dog. ‘the person is looking at the dog.’ d. cha-ib lo niaa. PROG-go-3SG face.region field. ‘he is going to the fields.’

Bcuaa quiang – I stepped HEAD it!

163

The bodypart noun used with a temporal meaning is illustrated in example 13. (13) lo iz 1906 biab guia.llub. face.region year 1906 CPL-fall rock.corn.grain. ‘In the year of 1906 it hailed.’ San Marcos Tlapazola Zapotec bodypart locatives employ an intrinsic as well as an extrinsic frame of reference (Lewinson, 1996). Zapotec of Ayoquesco, as reported by MacLaury on the other hand, relies on an extrinsic frame of reference, maintaining a vertical framework e.g. the bodypart quia (head) always refers to the highest part of the object, while nii (foot) refers to the lowest part, and dets (back) to the posterior region etc. (MacLaury 1989: 120). Zapotec spoken in Tlapaloza, although similar in many ways to Ayoqesco Zapotec, also differs in interesting ways. One difference is that the highest part of an object is far from always referred to using the bodypart noun quia, as is the case for Ayoquesco Zapotec. In Tlapazola Zapotec the location of an object lying on the surface of a table is always referred to as lo mes ‘face.region table’, despite the surface of the table being the highest part of the table, when standing in its canonical orientation. In addition, one may refer to the shorter side of an extended table as quia mes ‘the head of the table’. The location of a hot clay pot recently removed from the fire and placed on an inverted chair is referred to as being quia yallily ‘head.region of chair’, which suggests the speaker’s employment of an extrinsic frame of reference. What factors might motivate these semantic partitionings often violating the default hypotheses for describing the semantics of Zapotec bodypart nouns as suggested by MacLaury? One plausible explanation has been suggested by Brugman (op.cit.) in describing Mixtec, an Otomangeaun language which similarly relies on the metaphorical extension of body-part nouns as locatives. Brugman introduced the notion of associated space, in order to explain why, for example, the surface region of a table is referred to by violating the principle of quia used for reference to the highest region of an object. According to Brugman’s notion of associated space, since tables are mostly associated with humans interacting with them in a face-toface relationship, they are referred to as metaphorically possessing similar properties as human faces. This implicit, and cultural specific notion may be the core motivational factor explaining why Zapotec speakers ‘prefer’ lo face, as opposed to head quia in reference to particular spatial relationships

164 Kristine Jensen de López involving support from gravity, when the landmark object canonically is oriented facing a human. Cultural and social motivated schematizations appear to influence the semantic system of locative reference in San Marcos Tlapazola Zapotec. For example, the human in a horizontal orientation may account for the conceptualisation underlying the specific location of non-spherical objects, as in quia (head) referring to the top region of the extended mat that villagers sleep on. Similar a non-spherical object, such as a round table, does not have a region referred to as quia (head), while lo (face) is obligatorily employed to refer to the surface region of the table. Ends or protrusions of fruits are referred to as regions of quia (head) or llaan (butt) in accordance with the orientation in which they grow (although speakers do not always totally agree on this). The end of a banana which is attached to the banana palm is designated the llaan (butt) region of the banana. In summary, the employment of a specific frame of reference with a specific locative body part term depends mutually on pragmatics, the canonical functionality of the object of designation, geometry, schematisation, as well as on the social situatedness of the particular speaker employing and extending the system. The semantic and pragmatic structure present in the employment of bodypart locative nouns in Zapotec of Ayequesco suggests that the true structure of the Zapotec bodypart system can not be adequately accounted for in terms of a formal abstract system identical to the system underlying prepositions, such as in, on, over. Instead they claim that the system is based on a metaphorical extension of the human body in a canonical orientation. However, it is important to note that the system is not imposed in a rigid or formalized fashion, but involves less directly predictable extensions based on the social and interactional practices within the culture, thus reflecting the shared knowledge of the metaphors and cultural practices contextualizing the linguistic usage. 3.2. Associated space The parameter of associated space illustrates the notion of extended metaphor usage of bodypart nouns (Brugman, op.cit.). This usage is employed in contexts similar to those where speakers of Indo-European languages would employ adpositions or dative constructions. Here the semantic role of the bodypart noun is to identify a particular location of the subject in relation to the landmark object named by the nominal compound. One main difference between adpositions and bodypart nominals is that adpositions

Bcuaa quiang – I stepped HEAD it!

165

express the abstract locative relationship per se, whereas the bodypart nominals often rely crucially on the surrounding discourse context for a correct interpretation. For example the bodypart reading illustrated in example (14), is ambiguous because the nominal compound lo mes can be interpreted as either; ‘on the table’ or ‘in front of the table’, depending on the information surrounding sentence and discourse. (14) taz lo mes. cup face.region table. ‘the cup is on / in front of the table.’ 3.3. The case of quia (head) and lo (face) The bodypart locative nouns quia and lo, although they share common semantic features conflating the notion of support from gravity, also reflect interesting semantic distinctions regarding their reference to spatial relational configurations. The similarities between the terms quia and lo, are thus best illustrated when compared to the English preposition on, where the notion of support involving surface attachment is expressed. The semantic differences between the two Zapotec bodypart nouns, however, lie in the fact that lo often refers to horizontally extended surface objects (e.g. an extended mat), whereas quia, on the other hand, conflates the notion of the uppermost region of a three dimensional landmark object. Following this reasoning, a book on a bookcase is referred to as quia while located on the very top shelf of the bookcase, but referred to as lo.region while located on any of the lower shelves with the exception of the very bottom shelf, which thus is referred to as the nii-region (foot) of the shelf. The usage of quia to refer to three-dimensional landmark objects is suggested to be the least extended usage of this particular bodypart term, such as in the reference of ‘a chicken on the roof’. However, the metaphorical extension of lo used to express locations, which in English are referred to as ‘in front of a surface’ and which is the default of canonical social interactions, can outrule a less extended metaphorical usage of a bodypart word. Thus despite the fact that a table consists of a three-dimensional object and hence the trajectory object located on the table by default would be expected to be referred to as located quia.region of the table, Tlapazola Zapotec-speaker do not accept such usage. Instead the bodypart word lo is used to specify the location of an object on a table, which suggests that this particular metaphiorical extension is in violation of the rule posed by MacLaury.

166 Kristine Jensen de López In conclusion, lo is frequently used for reference to locative relations, which are based on functional, social and interactional properties. This is a property it does not necessarily share with quia. Brugman (ibid) suggests that this specific role played by the bodypart term for face has to do with the fact that in society faces are construed as important not because of their shape or relative location on the human body, but because they are salient identifying properties of people, since social interactions take place in faceto-face situations. The notion of social interaction, which is conflated in the locative meaning of lo makes it difficult to categorize lo in terms of basic abstract or geometrical properties that account for all uses of the noun. Examples 15a and 15b illustrate some of the polysemous usages of lo. In example (15a), lo expresses the surface of horizontally oriented flat objects, while in example (15b) lo is employed as a directive and determiner of social interaction. (15) a. beecu lo daa daa. dog face.region mat. ‘the dog is on the mat.’ b. ch-áa lo Nita. POT-go-1SG face.region Anita. ‘I am going to visit / see Anita.’ The usage of lo as goal of transfer or directedness of energy is expressed in predicates similar to the notions expressed in English by: ‘lie to’, ‘look at’ or ‘visit’. In general, lo seems to be ‘preferred’ over quia in reference to relations involving support of an object on a vertically extended surface. 3.4. The case of láani (stomach), nii (foot) and dets (back) The semantics of the bodypart term láani may involve the notion of canonical containment, however, similar to lo, láani may also conflate the source of activity path as example (16). (16) b-teed chiv láani coral. CPL-pass goat stomach.region corral. ‘Pass the goat through the corral.’ The configuration expressed in English as ‘under the table’, is expressed in Zapotec by either employing the bodypart noun láani (stomach) or the

Bcuaa quiang – I stepped HEAD it!

167

bodypart noun nii (foot). If the configuration specifically involves the upper region of the identified space as in ‘a string hanging from under the table’, this is always expressed as a láani configuration, whereas an object lying on the low ground area of the table (e.g. ‘under the table’) is expressed as a nii configuration. Again the distinction in interpreting the two uses can sometimes be made clear through the verb or the discourse context. Finally, the noun dets ‘back’ is used for reference to objects, which are displayed partly or fully out of sight. 3.5. Zapotec static spatial relations and the implications for language acquisition Although much more could be analysed and said regarding convergences between the semantics of basic spatial reference, the above analysis will be sufficient in order for the reader to follow the longitudinal case study of the acquisition of bodypart locative nouns presented in continuation. In the above sections I have illustrated the differences in the syntax and in the semantics of the basic lexical system of Zapotec static spatial reference compared to languages, which rely on prepositions for referring to static spatial relationships. In summary; Zapotec employs bodypart nouns, which do not encode the path of the trajector object, but only the specific region in relation to the landmark object. Bodypart region nouns are motivated by a metaphorical extension of human body parts onto the landmark object. These crosslinguistic differences suggest several competing and inconsistent hypotheses for children learning to refer to static spatial relations by use of bodypart region nouns. The embodiment hypothesis (e.g. Johnson (1987) or Lakoff (1987) might suggest that the child initially acquires the specific meanings of locative bodypart nouns, as derived from bodily experiences. Given that Zapotec-speaking children presumably know the names of their own body-parts, as is seen among young children acquiring prepositional languages, in the case that locative bodypart nouns should be acquired as separate lexical terms, this acquisition task may come into conflict with the child’s acquisition of nouns for human body parts. Thus one might hypotheses that Zapotec acquiring children, in the process of acquiring both the literal and the locative meaning of bodypart nouns, will start using bodypart nouns for reference to literal meanings, perhaps in reference to the child’s own body, and only later transfer this knowledge on the metaphorical extended meaning employed for referring to static spatial re-

168 Kristine Jensen de López lationships. Consequently this conflict at the lexical level may delay the child’s usage of bodypart terms as locative region nouns, at least until the metaphorical structure of these grams is fully in place. Oppositely, a dualsystem approach would suggest that the child acquire both systems in parallel to each other, and only gradually acquire the full subsystem of bodypart region nouns. The more general question concerning language acquisition is how young children go about acquiring such a complex, but at the same time semantically transparent system, and whether they follow a different development process involving different cognitive strategies than those of children acquiring prepositional languages, and hence in accordance with Slobin’s (and consistent with the view of Bowerman, ibid) notion of language acquisition in terms of typological bootstrapping (1996). 4. A longitudinal study of the early acquisition of Zapotec bodypart region nouns Until now it has not been analysed in detail when children acquiring bodypart nouns begin to produce these linguistic concepts in spontaneous speech. The data analysed in the present study consists of spontaneous speech from one monolingual Zapotec-acquiring boy age 15–33 months recorded during free interaction at his home in San Marcos Tlapazola. A total of 15 data points of each 90 minutes of video recording and/or audiorecordings were analyzed. On few occasions the child was video or audiorecorded by an older sister without my presence, although on most occasions I carried out the recording. All transcriptions were carried out in close collaboration with the child’s teenage sister who worked with me on a regular basis (see Jensen de López 2002). The final transcriptions were checked against context using the video-data5. The data were analyzed for the age at which bodypart region nouns (that is static spatial relational reference) were initially produced spontaneously by the child. The child’s first use of bodypart region nouns appeared when he was well into the two-word stage and perhaps consequently these early bodypart nouns, in some way, seemed to resemble predicate constructions. This may be viewed as an early emergence of bodypart locatives, when evaluated in relation to the fact that Tzeltal-acquiring children do not acquire bodypart locatives until a much later age (Brown, personal communication). Prior to producing bodypart locative nouns the child had produced verb predicates involving

Bcuaa quiang – I stepped HEAD it!

169

motion activity, hence bodypart nouns did not seem to play the role of sentence structuring verbs, as Tomasello suggests is the case for English children acquiring prepositions. A total of four bodypart nouns were employed as spatial locatives during the data period. These consisted of láani at the age of 24;12 months, quia at the age of 24;12 months, lo at the age of 27;23 months and dets at the age of 30 months. Interestingly, the very first bodypart region nouns identified in the speech of this child were constructions with clitics. This type of construction accounted for the child’s initial production across a whole of three of the produced bodypart nouns, namely láani, quia and lo. Hence, this suggests that the child did not hold clear grammatical categories concerning word classes, as clitics appeared with nouns as well as with verbs. 4.1. The child’s initial conceptualization of the meanings underlying bodypart region nouns The underlying conceptualizations motivating the child to employ bodypart locatives were assessed through detailed analysis of which referents were involved when the child employed a bodypart region noun as a locative. Each bodypart locative utterance was thus analyzed for its referent. In order to identify the exact referent I relied closely on the contextual situation as present in the video recording and in very few cases on the contextual cues noted down immediately after the audio recordings. All the child’s spatial relational utterances held clearly identifiable trajectory and landmark 6 objects with only few exceptions in the audio-recordings. Láani (and dets although dets appeared less frequently7) was the bodypart region noun, which seemed employed in the most consistent way by the children. However, láani as well as dets appeared overall less frequent in the data compared to the bodypart region nouns quia and lo. 4.1.1. Láani The child’s initial use of láani referred to containment relationships, which involved nonvisible trajectory objects. These utterances were in the context of the two situations; one in reference to a piece of bread in bowl of soup (24;12) and a second one in reference to a toy pistol which had fallen in between a pile of iron rods stacked on the ground (27;23). Thus both uses

170 Kristine Jensen de López referred to the notion of a trajectory object being constrained by the region of a landmark object, but also to objects, which had become nonvisible to the speaker. At a later age the child produced láani in utterances, which sometimes referred to the partly visibility of the trajectory object or still at other times referred to a total enclosure of the trajectory object. These contexts existed of; a small toy inside a container object, a puppy lying in a pile of corn ears and air inside a balloon. Again at a later stage láani was also used in reference to the notion of what perhaps was an ‘imaginative’ constraint of the trajectory object. The context of this particular usage consisted in the child playing with a non-cultural and novel object, namely my son’s Fisher Price garage, which the child had placed some toy bulls on the middle open-sided platform. The landmark object (the garage) did not resemble a container in anyway, since all four sides of the platform were open-sided. It did however involve the notion of constraint, in that if the bull moved beyond the boundaries of the platform it would fall to the ground. The fact that the platform was the middle section of three platforms suggested that the child might have employed the metaphor extension of a human in an up-right position to the landmark object (the garage) explaining his usage of láani to refer to this particular constraining spatial relationship8. Interestingly, the child’s initial uses of láani all involved the notion of constraint, although not all uses also conflated the notion of containment, as is the main properties underlying the preposition in in English. Cognitive Scientists take the notion of containment to be the most central and basic property of the preposition in. However, Zapotec bodypart region noun conveys the meaning of an inner.region of an object, independent of its orientation, rather than conflating the notion of containment or inside in relation to gravity, and this notion may explain why the Zapotec child mainly employed láani to mean constraint, rather than containment9. 4.1.2. Dets The child’s initial production of the bodypart locative dets was unclear in regard to whether he took the noun dets to refer to a static spatial relationship, to a literal bodypart or perhaps to convey both meanings at once. In any case the child’s initial usage of dets appeared within a full predicate construction referring to location, see examples (17a) and (17b). Note once more, that the locative bodypart appears with a clitic suffix, suggesting the meaning of its (the bottle’s) back.region.

Bcuaa quiang – I stepped HEAD it!

171

(17) a. Uttered while looking at a label stuck on a soda bottle – age 28;11 a xi cá ra bïny dets-ïng. EXPR that stuck PL person back.region=clitic. ‘Look, what someone has stuck on this.’ b. Uttered while standing behind his mother (who was squatting), and just about to kick a ball directly into her back – age 28;11 cua-a dets ma. put-1SG back.region mama. ‘I am going to throw the ball at mommy’s back.’ Although these two constructions appeared in the exact same recording session they suggest that the child relied on diverging conceptualizations concerning the meaning of dets. In example (17a) the child was playing with an empty soda bottle, when he suddenly expressed surprise seeing that a label was stuck on to it. The child’s locative production of dets suggested that he had not rely on an ego-centric conceptualization of the meaning of dets, but had projected the literal meaning of a body part noun onto the bottle, perhaps relying on an intrinsic frame of reference given the fact that the label was facing towards the child. In example (17b), however the child produced dets in a sense, which seemed to conflate the literal meaning of the noun ‘back’ with a locative meaning. In fact in this example it is totally ambiguous whether the child was referring to the mother’s back independent of the designation of the ball. However, since in example (17b) the child produced dets as a signal lexeme and not with a clitic pronominal, one might argue that the child might have relied on this grammatical distinction to separate the literal meaning of dets from the spatial relational meaning of dets. This, however, remains speculative until more data can be integrated for testing this hypothesis. Again at a later age dets was produced in reference to the location of a puppy behind a non-featured large straw basket, and did not involve a clitic pronominal. This utterance also reflected the child’s ability to impose a non-egocentric perspective in referring to the location of the puppy. At this age the child additionally demonstrated comprehension of the bodypart region noun dets when directed to him by his mother.

172 Kristine Jensen de López 4.1.3.

Lo and Quia

The child used the bodypart locatives lo and quia with a relatively high degree of polysemy, but at the same time these uses seemed associated with different underlying conceptual systems. 4.1.3.1. Quia The child’s initial usage of quia was not restricted to one single meaning based on a particular superimposed metaphorical extension. Although, some of the child’s utterances, which involved the bodypart region noun quia often referred to the placement of a trajectory object on the top region of a three-dimensional landmark object, they also often involved other types of profiling of static spatial relationships. The initial uses of quia, which were produced by the child are presented in the examples (18) and (19). (18) Uttered as the child is about to crawl up on top of a large object, presumable an inverted container object – 24;12 na-a na-a ch-epi-a quia=y. PRO-1SG PRO-1SG POT-ascend-1SG head.region=clitic. ‘I, I am going to climb up on top of that.’ (19) Uttered as the child had just stepped on an orange peel lying on the ground – 24;12 b-cua-a quia=ng. CPL-put-1SG head.region=clitic. ‘I stepped on it.’ At a later age the child used quia in reference to a source-path-goal activity profiling a location at the highest region of the landmark object, as illustrated in examples (20a) and (20b). (20) a. Uttered as a small toy doll had just fallen down from the roof of the driver’s cabinet of toy pick-up truck – age 28;11 B-ix mon ndee quia car. CPL-flip.over doll PRO head.region truck. ‘The man flipped over / down from this.’

Bcuaa quiang – I stepped HEAD it!

173

b. Uttered while the child stepped on a small toy sheep (he erroneously called it a goat – age 28;11 G-deed.c-a quia chiv. CPL-pass.exactly-1SG head.region goat. ‘I stepped on top of the goat.’ The child also used the bodypart region noun quia to express the path of an activity involving the notion of ‘over’. This was in relation to a trajectory object passing over the top region of a landmark object, which was in the context of a small toy car driving over a toy animal. Hence, the child used the region noun quia to express the path of the trajectory object rather than to express the abstract notion of support from gravity as is expressed by the English preposition ‘on’. Hence, the child seemed to be relying on a somewhat different conceptual system involving associations, which often conflated the notion of the uppermost region of a well-defined object with other associated space parameters such as the path of activity. 4.1.3.2. Lo Overall the child’s initial usage of the bodypart region noun lo was to conflate the perspective of social interaction combined with the notion of transfer of energy, similar to what is expressed by datives in English. From the child’s initial productions of this lo locative, it seemed differentiated from the more conservative meaning the child employed when producing the bodypart region noun quia. However, concerning the child’ underlying conceptualization of lo, he seemed to have developed two categories for which lo could be employed. On the one hand lo functioned to refer to abstract and socially interactive face.to.face interactions, which in a certain sense was similar to dative usage of prepositions in English. These were in contexts expressing the notion of a) a person lying to another person, b) a person looking at another person and c) a person taking something from another person. Examples (21a) to (21c) illustrate these particular employments of lo. (21) a. Uttered while joking with his older sister – 24;12 g-uaaci-a lo naa. CPL-LIE-1SG face.region PRO.1SG. ‘I am saying that I am lying to you.’

174 Kristine Jensen de López b. Uttered while looking at his sister – 28;11 rian lu-a Nita. put face-1SG Nita. ‘I am looking at Nita.’ c. Uttered while commenting on his own ongoing activity – 28;11 xi rian lu-a beecu dxindx. that put face-1SG dog small. ‘And I am looking at a dog.’ d. Uttered as the child was just about to throw a ball at his sister – 28;11 ey i-cu-a=ng lo. EPR POT-put-1SG=CLITIC face. ‘Yes, I am going to throw it at you.’ The second category of conceptualizations underlying the child’s employment of the bodypart noun lo expressed the notion of support of a trajectory object in relation to ‘a close-to-ground’ or a lower region landmark object. Examples of this particular kind of meaning are illustrated in example (22). (22) Uttered as the child was looking into the back region of a toy pick-up truck – age 28;11 chúu lo=g? be face.region=clitic? ‘What’s inside it?’ The main contexts in which the child used the bodypart region noun lo in reference to a spatial relationship between two objects were as follows; for trajectory objects lying on a straw mat spread on the ground, for the child lying on a mat, for a bag placed on a self, for a container object placed on top of the driver’s cabinet of a toy truck and for a sliver stuck into the child’s hand. Although many of these uses conflate the notion of support from gravity, most uses were for close-to-ground relationships or for relationships, which were at eye level of the speaker. The child’s use of lo also conflated the notion of source-path-goal, and again in relation to a lower region activity. Examples of these are illustrated in examples (23a) and (23b). (23) a. Uttered as the child was about to step on some toys – age 28;11 ni-a=ng lo nde. IRR-1SG=CLITIC face.region PRO. ‘I was going to pass over / across of this.’

Bcuaa quiang – I stepped HEAD it!

175

b. Uttered as the child requested the mother to take a doll out from the back of a toy pick-up truck – age 28;11 gu-le=g lo car=ig. CPL-take.out=clitic face.region truck=clitic. ‘Take it out of this truck.’ Overall the child produced bodypart nouns as datives as well as regional locatives at an early age and these often conflated several features within the specific meaning. In conclusion specifically the bodypart nouns quia and lo were used with different syntactic constructions and with polysemous meanings by the child. These two types of bodypart nouns were also the most frequently produced bodypart nouns throughout the data sample. These also seem to be the most frequent bodypart locative nouns in naturalistic speech. 4.2. In order to test the hypothesis regarding whether the child’s acquisition of abstract locative meanings of bodypart nouns was influenced by the his acquisition of the literal meanings of bodypart nouns, I investigated the correlation between the child’s spontaneous production of each locative bodypart noun and his spontaneous production of each literal bodypart noun. The results are reported in the following section. First I present data regarding the child’s spontaneous usage of literal body parts throughout the data sample. 4.2.1. The child’s reference to literal body parts During the specific period of data collection the child used a total of four body part terms in reference to real body parts; namely the terms nii ‘foot’, lo ‘face’, naa ‘hand’ and dets ‘back’. The child used nii productively in suffixed constructions when referring to his own foot, to the foot of his sister and also to the leg of a table. The child also expressed comprehension of the literal meaning of lo, which was used in reference to his face. Dets and naa were both used in reference to his mother’s back and hand. Table 2 illustrates the correlations between the child’s usage of bodypart terms in reference to real body parts and spatial relations. A √ marks positive cases

176 Kristine Jensen de López of correlation between a regional and a literal use of a bodypart, while X marks negative cases, that is when the particular bodypart noun was used with one type of meaning, but not with the other type of meaning. The notion of own versus other refers to whether the utterance was in reference to child or to someone else. Table 2. Correlations between the child’s production of literal referring and locative region referring bodypart nouns BODYPART word

Láani

Lo

Quia

Dets

Nii

Regional meaning

√

√

√

√

X

Literal meaning

X

√ (own)

X

√ (other)

√ (own)

As illustrated in the table above, the child’s initial acquisition of body part nouns did not reflect an overall one-to-one correlation between the child’s use of the literal body part nouns with the source domain meanings and the child’s use of bodypart region nouns in reference to spatial relationships. Only three out of the total of four produced bodypart nouns appeared in both types of meanings (lo, dets and nii) while quia and láani, despite being among the high frequent bodypart locative nouns produced by the child were not produced with both meanings. This has implications for the embodiment hypothesis, in that spontaneous production of bodypart locative nouns can develop independent of the child’s production of bodypart nouns with literal meanings.10 5. Discussion The bodypart nouns quia (head) and lo (face) where the most frequently appearing bodypart locatives at this early age. At first glance the meaning of these may seem to share a great deal of resemblance with the English preposition on in for example the notion of support from gravity. However, Zapotec bodypart locatives seem to involve a semantic distinction, which prepositions do not rely on. In Zapotec one must distinguish between whether a support relationship appears at a high level region of a landmark object, and for which quia is most often employed, as opposed to whether a support relationship appears at a close-to-ground level region of the landmark object, and for which lo is employed. Moreover, the use of lo often

Bcuaa quiang – I stepped HEAD it!

177

conflates the notion of support from gravity with the notion of interpersonal interaction at eye-level. The results from the analysis suggest that at a very early age this Zapotec acquiring boy was aware of these language-specific semantic distinctions in relation to referring to spatial relationships between two objects. The meanings underlying the child’s earliest uses of bodypart nouns seemed related to social interaction and transfer of activity rather than to the mapping of the source domain meanings of literal body parts in a oneto-one fashion. None of the child’s uses of the bodypart locatives were identified as erroneous by native adult Zapotec speakers, and thus seemed to mirror adult usage of the bodypart system. A final observation from the above analysis was that despite the fact that Zapotec bodypart locatives appear in a sentential position (or in a preposition phrase), which is similar to the structure for many Indo-European languages, the child’s early bodypart locatives did not occur in holophrastic constructions as is the well-known case for children acquiring prepositions in Indo-European languages. Bodypart locatives, on the other hand, initially appeared in clitic constructions. Additionally, the concept of behind dets was acquired earlier than reported for children acquiring behind in English, and furthermore the appearance of dets was similar to the age at which the notion of support and constraint appeared. This challenges a pure cognitive approach to the acquisition of spatial language where cognitive complexity is the key prediction for acquisition rather than linguistic complexity. A cognitive prediction might view a difference in the lexical ‘overlap’ as a problem, which needs to be worked out by the child in a one-to-one fashion before he is able to deal with both lexicons in parallel. However this study does not find support for such prediction. Overall the acquisition of San Marcos Tlapazola Zapotec bodypart locative terms, as appearing in the early spontaneous production of this particular child, suggests an organization and conceptualization, which differs in structure from what has been attested for children acquiring a prepositional language. Although Zapotec bodypart locative nouns emerge at least just as early as prepositions do they initially follow a different acquisition process in the speech of the child. The fact that bodypart locatives initially appear as predicates makes their appearance more sudden than the process observed from children acquiring verb particles/prepositions, and for which it has been suggested prepositions serve as ‘sentence-structuring’ verbs (Tomasello 1987). It may be the case that, while children acquiring prepositions proceed from verb particles to predicates, children acquiring locative

178 Kristine Jensen de López nouns, such as is the case for Zapotec, use bodypart locative nouns as full predicates from the very beginning. This suggests two different routes accounting for children’s acquisition of language used to refer to spatial relations, and may be explained in terms of Slobin’s notion of typological bootstrapping. The notion implies that crosslinguistic differences in children’s language acquisition may be due to linguistic complexity rather than a cognitive complexity (Slobin 1997). (A view compatible with that of Bowerman & Choi 2001.) 6. Conclusion This preliminary study suggests that the process and emerging order for children’s acquisition of spatial relational terms is dependent on the linguistic complexity within the language being acquired. Whereas prepositions appear early in the one word utterances of children, and frequently involve the child’s own body as either the trajectory or the landmark object, bodypart terms do not emerge until well into the multi-word stage of language acquisition, and are not initially used in reference to the child’s own body. Semantically, prepositions/particles incorporate both path and region information and they are both portmanteau items. San Marcos Tlapazola Zapotec bodypart nouns are less relational (in part because they do not encode path) than prepositions/particles, but are also the only means of locating a specific region in the spatial relational array. Hence, bodypart locative nouns do not serve as sentence-structuring verbs as they might for children acquiring English. At the language-specific conceptual level these differences are reflected in the children’s varied conceptualizations of the notion of support. At the level of reference, Danish and English children are more likely to use their own bodies as trajectory or landmark objects than this Zapotec child did. However, it is unlikely that these differences are only due to the difference language structures as the particular way a language is used within culture practices also plays a crucial role. In terms of embodiment and metaphor theory it does not seem to be the case that the embodied semantics of Zapotec bodypart terms plays a facilitating role in early acquisition (Sinha & Jensen de López 2000). There is some indication, though, that the Zapotec child progresses faster towards constructing these terms into a system than children acquiring prepositions do. Whether this is due to the transparent and embodied semantics or it is due to the landmark-focused, non-portmanteau semantics of Zapotec bodypart terms, is a topic for future research.

Bcuaa quiang – I stepped HEAD it!

179

Further support for the notion that acquiring Zapotec bodypart nouns may provide Zapotec children with a conceptual system, which differs from the habitual system developed by children acquiring different languages structures is attested in my cross-cultural investigation of language comprehension and action imitation of spatial relational concepts. The study showed that Zapotec and Danish children perform in systematic different ways when acquiring spatial comprehension and cognition (Jensen de López 2002a, 2002b, 2006). Further research on the acquisition of Zapotec bodypart locative nouns would benefit from analyzing data from Zapotec child-directed speech in order to compare the effect of input with the analysis presented in this paper and in order to permit a more detailed appreciation of what influences children’s acquisition of bodypart locative nouns.

Notes 1. I use the following abbreviations in this paper: 1, first person; 2, second person; 3, third person; SG, singular; PL, plural; PRO, pronoun; CPL, completive aspect; FUT, future aspect; HAB, habitual aspect; POT, potential aspect; PROG, progressive aspect; EXPR, expressive. 2. An example of a spatial relationship is of the kind ‘the cat is on the mat’, where the preposition on is the locative word expressing where the cat is in relation to the mat. 3. I refer to this specific dialect of Zapotec using the general term Zapotec rather than San Marcos Zapotec. 4. To my knowledge, this is the first study investigating the acquisition of bodypart nouns at a very early age, however studies of children acquiring Mayan languages which also make use of bodypart locative nouns show that children are above 5-years of age before they show mastery within this category of words (Penny Brown, personal communication). 5. Data collection was carried out during an extended period of fieldwork starting in August 1995. My rutine visits several days a week to the child’s home created a natural relaxing atmosphere for video-recording the spontaneous daily interactions of the child. 6. The concept of Trajector refers to the entity whose location is of relevance and is similar, while the concept of Landmark refers to the entity in relation to which the location of the Trajector is determined (Langacker 1987). For the example in i, the cat is the trajector and the mat landmark. 7. As dets only occurred once and furthermore this was late into the period, the notion of consistent usage is only preliminary.

180 Kristine Jensen de López 8. Note that the notion of support from gravity was not the central concern of the child in his conceptualization of this particulat spatial relationship with novel objects. Elicitation of novel objects may be seen as the most precise way to tap into the underlying conceptualization of spatial objects, as it demands the speaker to generate the spatial item without relying on previous reference to the specific trajectory-landmark relationship. 9. The distinction between containment involving the physical properties of constraint merged with the properties of containment, as opposed to constrainment involving the inner (stomach).region of a particular landmark object causes a problem for semantic theories. It also challenges the claim made by mainstream cognitive developmentalists, who argue that lexical concepts emerge as mappings from a set of universal prelinguistic concepts, and lies at the heart of the Mandler-Bowerman debate.

References Bowerman, Melissa & Soonja Choi 2001 Shaping meanings for language: universal and language-specific in the acquisition of spatial language categories. In Language acquisition and conceptual development, Melissa Bowerman & Stephen C. Levinson (eds.), 475–511. Cambridge: Cambridge University Press. Brown, Roger 1973 A first language. Cambridge MA: Harvard University Press. Brugman, Claudia Ms. Metaphor in the Elaboration of grammatical categories in Mixtec. Unpublished Ms. Bulter, Inez 1988 Gramatica Zapoteca de Yatzachi el Bajo. Serie gramática de lengua indigena de México. Summer Institute of Linguistics. Clark, Eve 1973 Non-linguistic strategies and the acquisition of word meanings. Cognition 2: 161–182. Gentner, Dedre 1982 Why Nouns are Learned Before Verbs: Linguistic Relativity versus Natural Partitioning. In Language development, Vol. 2: Language, thought, and culture, Stan. A. Kuczaj (ed.), 301–334. Hillsdale, NJ: Erlbaum. Jensen de López, Kristine 1998 Learning to organize space by use of prepositions, body parts, verbs of motion and disposition. Paper presented to the Conceptual, Discourse and Language Conference, Atlanta, USA, October 10 –12.

Bcuaa quiang – I stepped HEAD it! 2002a

181

Baskets and body-parts: a cross-cultural and cross-linguistic investigation of children’s development of spatial cognition and language. Danish and Zapotec children’s spatial cognition and language. PhD dissertation, Institute of Psychology, University of Aarhus. 2002b Language-specific patterns in Danish and Zapotec children’s comprehension of spatial grams. The 31st Stanford Child Language Forum, Eve Clark (ed.), 50–59. Stanford University Press. http://cspublications.stanford.edu/CLRF/2002/CLRF-2002-titel.html. 2006 Culture, language and canonicality: Differences in the use of containers between Zapotec (Mexican indigenous) and Danish children. In Doing things with things, Alan Costall & Ole Dreier (eds.), 87–109. London: Ashgate. Johnson, Mark 1987 The body in the Mind: the bodily Basis of Meaning. University of Chicago Press. Johnston, Judith R. & Slobin, Dan I. 1979 The development of locative expressions in English, Italian, SerboCroatian and Turkish. Journal of Child Language 6: 529–545. Lakoff, George 1987 Women, Fire, and Dangerous things: what categories reveal about the Mind. The University of Chicago Press. Langacker, Ronald W. 1987 Foundations of cognitive grammar, Vol. 1: theoretical prerequisites. Standford, CA: Standford University Press. Levinson, Stephen C. 1996 Relativity in spatial conception and description. In Rethinking linguistic relativity, John J. Gumperz & Stephen C. Levinson (eds.), 177– 202. Cambridge University Press. Lucy, John A. 1992 Grammatical categories and cognition; a case study of the linguistic relativity hypothesis. Cambridge: Cambridge University Press. MacLaury, Robert E. 1989 Zapotec Body-Part Locatives: Prototypes and Metaphoric Extensions. American Linguistics 55 (2): 119–153. Munro, Pamela & Fellipe H. Lopez 1999 San Lucas Quiaviní Zapotec Dictionary, Vol. 1. University of California, Los Angeles: Chicano Studies Research Center Publications. Pickett, Velma & Black, Cheryl 1995 Gramática Popular del Zapotec del Istmo, versión preliminar. Summer Institute of Linguistics. Rohlfing, Kathrine J. 2002 UNDERstanding; How infants acquire the meaning of UNDER and other spatial relational terms. Bielefeld University. PhD dissertation.

182 Kristine Jensen de López Sinha, Chris G., Lis A. Thorseng, Mariko Hayashi & Kim Plunkett 1994 Comparative spatial semantics and language acquisition: evidence from Danish, English and Japanese. Journal of Semantics 11: 253– 287. Sinha, Chris G. & Kristine Jensen de López 2000 Language, culture and the embodiment of spatial cognition. Cognitive Linguistics 11 (1/2): 17–41. Slobin, Dan I. 1997 The Universal, the Typology, and the Particular in Acquisition. In The crosslinguistic study of language acquisition, Vol. 5: expanding the contexts, Dan I. Slobin (ed.), 1–39. Hillsdale, NJ: Erlbaum. Tomasello, Michael 1987 Learning to use prepositions: a case study. Journal of Child Language 12: 79–98. 1992 First Verbs: A case study of early grammatical development. Cambridge: Cambridge University Press. Zlatev, Jordan 2003 Holistic spatial semantics of Thai. In Cognitive Linguistics and NonIndo-European Languages, Gary B. Palmer & Eugene H. Casad (eds.). Berlin /New York: Mouton de Gruyter. Cognitive Linguistics 18.

“Lo oye, lo repite y lo piensa.” The contribution of prompting to the socialization 1 and language acquisition in Yukatek Maya toddlers Barbara Pfeiler

1. Introduction In some European languages certain interaction practises between adult caregivers and children, such as dialogic conversation, question-andanswer sessions and naming games, are said to function as socialization and language teaching tools that facilitate acquisition2. However, in cultures and languages like Kaluli (Schieffelin 1985; Ochs & Schieffelin 1995), Samoan (Ochs 1988), Basotho (Demuth 1986) and the K’iche’ (Pye & Rekart 1990), Tzeltal (Brown 1998) and Tzotzil (de León 1999, 2000) Mayan languages, infants between 6 and 12 months of age are not considered preferred interlocutors, that is, they are not treated as active listener/speakers. Face-to-face interaction among Inuit children is mainly with siblings, cousins, peer friends or adolescents with whom a non-simplified conversation can be begun accompanying physical play without toys (Crago 1997). A similar situation occurs among Yukatek Maya children in the state of Yucatan, Mexico, who demand less attention from their mothers and have more social interaction with siblings or other family caregivers (Daltabuit Godás 1992: 199; Gaskins 2000). Infant babbling in these societies is not represented as a linguistic aspect, and conversation with infants only begins after they have learned how to say “mama” or “dada” and the need is felt to teach them language. Discursive practises can vary by culture and according to the behaviours parents exhibit toward their children and expect from them. Prompting is a language socialization practise in which adults formulate what the child should say within a framework of “say…” (Demuth 1986; Schieffelin 1990; Sidnell 1997; Watson-Gegeo & Gegeo 1986). Promptings can include the activity of making a child speak, as well as the correction of words, grammatical forms or just pronunciation. A number of prompting strategies have been reported in different cultures, such as Kaluli (Schieffelin 1979),

184 Barbara Pfeiler Chicano (Eisenberg 1982), African-American English speakers (Ward 1971) and Mayan speakers like the K’iche’ (Pye 1986) and Tzeltal (Brown 1998). Kaluli mothers say it is necessary to correct children’s expressions in areas equivalent to phonology, morphology, lexicon and pragmatic and semantic meaning (Schieffelin 1985). Other strategies are repetitions, by which caregivers instruct children what is proper to say in a given situation (Schieffelin 1990; Ochs 1988; Watson-Gegeo & Gegeo 1986), and expansions, such as adult responses to children’s unclear or confused expressions in which they verbally reformulate the child’s expression in the correct grammatical form (Ochs & Schieffelin 1995: 80). Expansions are thought to facilitate child grammatical competence since they match children’s forms with adult forms. 2. Interaction between Mayan caregivers and children Research on childhood in Mayan societies has shown that dyadic interaction between caregivers and infants is generally minimal or absent (Vogt 1969: 165; Wagley 1949: 29–30 cited in Pye 1992: 241; Brown 1998; de León 1999). K’iche’ mothers in highland Guatemala seem most concerned about attracting the child’s attention or controlling its activities. They use the verb ch’a’ ‘(s)he says’ to linguistically intervene between their child and a person addressing the child to isolate her/him from direct contact with strangers. They do not elicit vocal-verbal responses and these routines apparently serve only a single function: to delimit specific conversational routines and provide immature speakers with correct verbal responses. For Tzeltal-speaking children in the Chiapas (Mexico) highlands, the dialogic repetition they are exposed to during socialization can be important for learning the structural properties of both the language and the collaborative verbal interaction style preferred by adults (Brown 1998). In the interaction between Tzotzil caregivers and their small children, also from Chiapas, de León (1999) reports direct elicitation with declaratives that teach behaviour beginning when children become mobile. Adults address children using expressions intended to control and focus behaviour, and are thus concentrated on what to do and not what to say. Expressions directed towards small children include rhetorical questions, a simplified register with its own vocabulary, deliberate annoyance and dyadic addressing intended to control the child’s behaviour or provoke affection. Among the Yukatek Maya in the lowlands of Yucatan, Mexico, Gaskins states that in socialization of children adults rarely verbally address chil-

“Lo oye, lo repite y lo piensa.”

185

dren that are just taking their first steps, but that children that can stand unassisted receive more language from their parents: “More speech is directed at upright babies who now understand and will obey some directives. They are also questioned about what they want or where they are going” (1996: 352). The earliest language interactions between adults and children are restricted to themes of danger, and adults do not tend to instruct children in what or how to play. Language is thought to be acquired on its own and children learn by “observing”, for instance, before the age of 2 years children spend 40% of their time looking at other people and things (Gaskins 1996: 352, 2000). This continues between 2 and 3 years of age when, even though their involvement in the social world increases daily, children remain focused on observation of others’ activities. It is during this period that the ability to understand events and activities grows. Children are not actively taught the language, “because parents believe that development just happens, that children move to a new stage by themselves, they do not feel the need either to take the direct initiative to teach children who fail to meet some developmental ideal or to feel anxiety about their responsibility for it.” (Gaskins 1996: 355). Parents thus do not feel the need to actively engage in their children’s language development since they consider it to be just another event, object or activity in their surroundings. In Yukatek Maya acquisition, however, data does exist showing use of prompting by caregivers with children that initially seems comparable to the socialization instruments observed in European cultures, such as the use of repetitions, question-and-answer routines and the naming game. These prompts are directed at what the child should say, as well as at what it should do. Based on this data, I did a survey of parents’ beliefs about language development, and the role of caregivers. The quantitative study3 was done in the town of Yalcobá4, which is also the site of research on Yukatek Maya acquisition. An interview was applied to 91 people (29 fathers and 62 mothers) between the ages of 15 and 75 years. The women were all homemakers and the men worked in construction, business or as farmers5. Themes on how children acquire language were selected to understand caregiver beliefs about language development, prompting and corrections. Of the 91 interviewees, 15 stated that the origin of Maya is innate whereas Spanish is learned. Most of them (75/91) saw it as important that children be in contact with adult speech (“to speak in front of children so they can hear and repeat”), and that this be reinforced by speaking slowly and clearly or through discursive practises such as repetition, correction or object naming. To determine if a child understands parents’ emissions it is

186 Barbara Pfeiler customary to ask the child to find an object. Slightly fewer interviewees (63/91) admitted to modifying speech addressed to children by speaking more loudly, simplifying and/or through repetition. By far the most homogenous response (76/91) was the need to correct children’s speech. 3.

The form and function of Yukatek Maya promptings

To understand explicit use of prompts and, among these, the forms of correction in directed speech, data were selected from three longitudinal studies of differently-aged children: Sandi, a girl (1;09 to 3;04); Armando, a boy (1;01 to 2;08); and David, a boy (2;01 to 2;02, and 2;05). The data for these children pooled for this analysis consisted of 58 recordings of caregivers and children. The smallest sample of the three was for David due to different recording circumstances. Participants in the recorded sessions include: the toddlers Sandi (SAN), “Chito” (Armando, ARM) and David (DAV); Sandi and David’s mother, Filomena (FIL); Armando’s mother, Lorena (LOR); Armando’s father, Pancho (PAN); and the children’s grandmother Narcisa (NAR). The recordings average 60 minutes in length and the recorded situations are spontaneous speech, play among children, and, in the case of David, interactions between him, his mother and his older sister Sandi. 3.1. Prompting form in caregiver speech The most frequent prompts were calling, questioning and naming routines with utterances that teach behaviour and a series of discursive practises. Promptings to report direct speech were found in dyadic (between two people) and triadic interactions (between three people).6 Triadic prompts, understood as those in which the speaker teaches the child how to respond to another person with a particular verbal construction, are mainly used in question routines, message transmission, and, occasionally, to make the child speak or do something for the other person in a general conversation or for politeness. In modern Yukatek Maya there are a number of verbal strategies used as elicitation techniques, such as the verbs ‘to say’ and ‘to speak’ and the particle kech. The most frequent and least specific verb is a’l-ik ‘to say it’, although t’aan ‘to speak’ (the root for ‘tongue’), is also frequently used to

“Lo oye, lo repite y lo piensa.”

187

refer to the initial emission in a conversation, both as a transitive verb (t’an-ik) and an intransitive (t’aan) verb. Both verbs can either precede or follow sections of reported speech, which can be an exact quote or indirect discourse (Lucy 1993). The verb a’l ‘to say’ was found in two forms, a’l-ti’ ‘say to him/it’ and a’l-eh ‘say it’. A’l-ti’ was found in triadic interactions to report direct speech or indirect speech indicating order, for example: (1)

Sandi (2;05.00) SAN: mor: eng:

He’

le=ten=o’ he’l=o’. DEICT DEF=PRO=DIST2 DEICT-DIST2 ‘Mine is there.’

FIL: mor: eng:

A’l-0=ti’ le=miis ka’h=u=haant(-eh)=o’. say-IMP=LOC DEF=cat PTL=he=eat(-SUBJ)=DIST2 ‘Tell the cat to eat it’!

SAN: mor: eng:

He’l=a’ miis. DEICT=DIST1 cat ‘Here it is, cat!’

A’l-eh was used in triadic interactions to report direct speech indicating elicitation + explanation, to prompt the child to say something or to make her/him call to someone, for example: (2)

Armando (1;10.06) LOR: mor: eng:

Granada kech=ti’ a’l-eh. granada PTL=LOC say-IMP ‘Say to him pomegranate, say it!’

NAR: mor: eng:

Ba’ax k-a=haant-ik. what INC-2ERG=eat-INC ‘What do you eat?’

ARM: Ba’l=a’ granada. mor: thing=DIST1 pomegranate eng: ‘This thing pomegranate.’ LOR: eng:

Granada. ‘Pomegranate.’

ARM: Granada. eng: ‘Pomegranate.’

188 Barbara Pfeiler This example shows the use of different strategies for prompting Armando to speak with kech=ti’ and a’l-eh, as well as with questions and finally repetition. T’aan and t’aan-eh were documented with the meaning ‘to call’ someone by their name, for example: (3)

Armando (1;08.08) LOR: mor: eng:

T’aan-eh. Novia kech=ti’. call-IMP girlfriend PTL=LOC ‘Call her! Say girlfriend to her.’

(…) ARM: Novia. eng: ‘Girlfriend.’ Lucy (1993: 91) states that the particle kech means ‘say (it) like this’ (i.e. the quotative function)7, and is always inflected for person. It indicates a verbatim speech quote or the re-enactment of a gesture, and occurs frequently in glosses. It always follows the cited discourse and is inflected according to the person to whom it is attributed.8 This form is extremely common in daily Maya conversation in which speakers routinely cite the addressee or other co-present individuals, imputing humorous expressions to them as a way of playing, having fun or notifying. Speakers also commonly use it to represent an interlocutor’s expression with a minor change in form, thus implying a totally different interpretation and usually casting a ridiculous light on the interlocutor. In comments, kech provides a generally semantic, efficient way of changing places in the citation and thus taking control of this portion of the reported speech functions. In the acquisition data, the adults make children repeat using the kech(-ti’) ‘say it like this (to her/him)’ particle. This was documented in triadic interactions to report direct speech, showing the child what to say in a wide array of contexts (see Table 1), such as giving an order, showing, teaching how to interact to attain a desire, treatment of animals, caring for infants, defending himself, overcoming fear, explaining, naming, correcting, asking and responding, and denying or forbidding something. In example (4) the mother asks her daughter, Sandi, to repeat the sentence using two voices of kech=ti’, the first time to have a third participant, through Sandi, carry out an action, and the second time to add information about the action to be carried out.

“Lo oye, lo repite y lo piensa.”

(4)

189

Sandi (2;02.09) FIL: T’ok-0=ten flores kech=ti’. mor: break off-IMP=PRO flowers PTL=LOC eng: ‘Say to him “break off the flowers for me!”’. SAN: T’ok Armando . mor: break off Armando eng: ‘Break (them) off, Armando!’ FIL: Ka’=p’éel kech=ti’. mor: two=CL.INANI PTL=LOC eng: ‘Say “two” to him.’ SAN: Ka’=p’éel Chito . mor: two=CL.INANI Chito . eng: ‘Two Chito.’

In example (5) Sandi’s mother, Filomena, speaks to her daughter to show her the improper behaviour of the third participant. The girl mentions another of the third participant’s behaviours and then the mother emphasises using the kech=ti’ command in a discursive structure similar to example (4). (5)

Sandi (2;02.09) FIL: Táan= u=puch’-ik-ech=e’ Armando=o’ . mor: PROG= A.3=crush-INC-B.2.SG-TOP Armando=DIST2 eng: ‘Armando is crushing you.’ SAN: Mak-0 a=chi’ nene’. mor: shut-IMP 2.A.SG=mouth baby eng: ‘Be quiet baby!’ FIL: Ma’ a=puch’-ik-en kech=ti’ . mor: NEG 2.A.SG=hit-INC-B.1.SG PTL=LOC eng: ‘Say “do not hit me” to him!’ SAN: Puch’-ik-en nene’. mor: hit-INC-B.1.SG nene eng: ‘Hit me baby!’ FIL: Yah-0 kech=ti’. mor: painful- B.3.SG PTL=LOC eng: ‘Say “it hurts” to him.’ SAN: Yah-0 nene’. mor: painful- B.3.SG baby eng: ‘It hurts baby.’

190 Barbara Pfeiler Examples (4) and (5) are both sequences with two turn takings, which is consistent with an adult command emission and its repetition by the child. Kech was also used in dyadic interactions to report indirect speech in elicitation or naming contexts. In the following example Armando’s grandmother (NAR) tries to motivate him to speak by asking him a question and showing him what to answer. (6)

Armando (1;08.13) NAR: mor: eng:

Kux túun a=mama. and so 2.A.SG=Mom ‘What about your Mom?’

NAR: mor: eng:

Táan uy=ichkil kech. PROG 3.A.=bath PTL ‘Say “she is bathing”.’

ARM: K-ichki . mor: ?-bath eng: ‘She bathes.’ Some examples were recorded in which direct speech reported with the a’l=ti’ indication was followed by direct speech reported with the kech=ti’ indication in contexts like giving an order + kech=ti’, naming + kech=ti’, telling someone what to do + kech=ti’ + explanation, and calling + giving an order. In example (7) use of the verb ‘to say’ (a’l=ti’) plus the particle ka’(ah) suggests that it is used in the sense of ‘speak to him so he does …’, with the aim of getting the attention of the third participant. (7)

Sandi (1;05.17) SAN: mor: eng: FIL: mor:

eng:

Wa’l-en Chito wa’l-en, Chito. stand up-IMP Chito stand up-IMP Chito. ‘Stand up Chito, stand up, Chito!’ A’l=ti’ ka’h wa’l-ak-0. Wa’l-en say=LOC(B.3.SG) PTL stand up-SUBJV-B.3.SG stand up-IMP ka’h ts’íib-nak-o’on kech=ti’. PTL write-SUBJV-B.1.PL PTL=LOC ‘Tell him to stand up. Stand up, so we can write, say it to him!’

“Lo oye, lo repite y lo piensa.”

SAN: mor: eng:

191

Wa’l-*eh Chito, wa’l-*eh Chito wa’l-*eh. stand up-IMP Chito stand up-IMP Chito stand up-IMP ‘Stand up! Chito, stand up! Chito, stand up!’

The possibility of the two prompts being used in the same act of speech offers the child more material with which to build her responses. Forms that require the child to use or change the pronoun form can be difficult for children since the change has to be correct.9 In addition to the ‘to say’ verbs, Yukatek Maya has two particles that are central to indexing reported speech. One is bin, usually glossed as ‘they say’ or ‘it is said’10, which is invariable and apparently does not inflect for person. Bin is used to report every day knowledge such as descriptive reports of current expressions like he’ bin u táale’ ‘it is said he will surely come’ (Hanks 1993). This particle specifies declarative sentences enunciated by a third person as not directly witnessed by the speaker, a common form in literary genders like stories (Pfeiler 1992), traditional histories and ritual speech (Hanks 1984). In the acquisition data employed here, use of evidentials is linked to children’s confused expressions11. Ki is the other particle used to indicate verbatim repetition in a spoken quote or in the re-enactment of a gesture. This evidential always follows the quoted discourse and is inflected for the addressee. Yukatek Maya adults use it as part of a routine of quoting the addressee or other co-present individuals to impute humorous expressions to them, have fun or even notify. They can also use it to represent the expression of an interlocutor with a minor form change that implies a totally different interpretation and usually makes the interlocutor seem ridiculous. Ki is less frequent than that of bin, but it has the same function of quoting the child’s confused expressions: (8)

Armando (1;01.29) FIL: mor: eng:

Ma’. Vieja ki. NEG old EV ‘No. Old lady, he says.’

ARM:

???

FIL: mor: eng:

Púuts’-0 bin. escape B.3.SG EV ‘She is escaping, he says.’

Example (8) shows how the caregiver – in this case the child’s aunt – clarifies the child’s unintelligible expression by guessing what the child wanted

192 Barbara Pfeiler to say based on context. This practise varied in the two mothers’ directed speech: Sandi’s mother (FIL) repeated the children’s unclear expressions while Armando’s mother used it to make fun of her son’s speech. Examples of corrections and repetitions12 were also documented. In example (9) Sandi refers to a cat nursing its kittens and emits a verbal root attributed to the animals’ actions. Filomena (FIL), Sandi’s mother, corrects the form by adding the aspect and person morphemes and finally presents her daughter with a verbal phrase with plural inflection linked to the situation that is then copied by the child. (9)

Sandi (2;02.14) SAN: eng:

Chan miis. ‘Small cat.’

SAN: mor: eng:

Chu’ch=a’. nurse=DIST1 ‘Nurse.’

FIL: mor: eng:

Tun

chu’ch-o’b. nurse-3.PL ‘They are nursing.’ PROG.(3.A.SG)

(…) FIL: mor: eng: sit:

Lóoch-a’n-o’b. hug-PART-B.3.PL ‘(They are) hugging.’ referring to the kittens.

SAN: mor: eng:

Ha. Lóoch-a’n-o’b. AFF hug-PART-B.3.PL ‘Yes (they are) hugging.’

FIL: mor: eng:

Bey in= lóoch-ik-ech=e’. ADV 1.A.SG=hug-INC-B.2.SG-TOP ‘Like I hug you.’

In example (10), the participants are eating fruit and Sandi indicates the fruit using a numerical classifier corresponding to animate beings. Filomena corrects her by replacing it with the numeral classifier for inanimate beings. The girl repeats the incorrect expression and is again corrected until she finally imitates the correct (adult) expression.

“Lo oye, lo repite y lo piensa.”

193

(10) Sandi (2;07.09) SAN: mor: eng:

He’

hun=túul-i’ . DEICT one=CL.ANIM-PARTIT ‘Here is one (animate).’

FIL: mor: eng:

Hun=p’éel ma’ hun=túul-i’ . one=CL.INANIM NEG one=CL.ANIM-PARTIT ‘One (inanimate), not one (animate).’

SAN: eng:

Hun=túul. ‘One (animate).’

FIL: eng:

Hun=p’éel . ‘One (inanimate).’

SAN: eng:

Hun=p’éel . ‘One (inanimate) yes.’

FIL: mor: eng:

Ma’ túun aw=ohel-i’ . NEG so 2.A.SG=know-NEG ‘Well, you do know it.’

The above examples show that children are exposed to a wide variety of explicit prompts, although these are used only sporadically in the data for these three children. Given this, and that interactions and communicative strategy use in many cultures are similar in form but not communicative function (Demuth 1986), an analysis was done of the explicit prompts according to their function in communicative situations, their frequency and the speaker. The imperative form of the verb ‘to say’13 a’l(=ti’) was mainly used to indicate activities while t’aan(-eh) generally served to call someone. All the caregivers used both prompts. The particle kech(=ti’), served to teach the social uses of the assertive voice (see Table 1), including giving orders, explaining, questioning, demanding or reporting and was mainly used by the mothers with the children.

194 Barbara Pfeiler Table 1. Prompt form, function and frequency in speech directed at three children acquiring Yukatek Maya. Prompt form

Function

…, kech(=ti’) ‘…, say it like this (to him)’

Single-word expressions: to name or indicate (objects, proper names) Expressions of two or more words: to elicit, correct, negate, explain How to interact with people and animals How to give orders How to forbid How to prevent danger How to care for children How to defend oneself How to greet

Frequency 35 16 16 9 4 3 2 2 2 Total: 89

A’l=ti’… Indicate activities ‘Say to him... /tell him’ Give an order

11 2 Total: 13

A’l(-eh) ‘Say it…’.

To name (objects, proper names) To call, question

7 3 Total: 10

A’l=ti’…, kech=ti’ # ‘Say to him…, say it like this to him’.

To call … give an order Give an order … elicit

T’aan… ‘Call…’.

To call

T’aan…, kech=ti’ # ‘Call…, say it like this to him’.

To call … name

2 5 Total: 7 18 Total: 18 4 Total: 4

In the speech addressed to the three children, most of the uses of kech(=ti’) cause the child to repeat words related to activities. This is done with the imperative mode and verbal phrases, and, to a lesser degree, by asking the child to repeat words naming objects, proper names, adjectives, numbers and quantifiers. This discursive practise reinforces data reported for early verb acquisition (Pfeiler 2003).

195

“Lo oye, lo repite y lo piensa.”

In the data for all three toddlers the particle kech=ti’ is the most frequent explicit prompt to report direct speech and is used more by the children’s mothers than by other caregivers. It is a prompt that mothers use in a variety of situations. This same prompt is used by children with their younger siblings during play14, but also to demonstrate behaviour like greetings. For example, at the age of 2;08 Sandi (Armando’s older sister) was already using the particle kech in the kech=ten ‘say to me’ form in a communication directed at Armando (1;01). It should be mentioned that the kech=ti’ prompt generally occurs in speech directed by caregivers to children and not vice versa, although Sandi, at 3 years of age, did direct this prompt to her mother when playing with dolls. In Sandi’s data, the kech=ti’ particle is used most frequently up to when she is 3 years old, after which these prompts are used in ways similar to in adult conversations. In Armando’s data, the most complete data set, kech=ti’ appears beginning in the first recording session, at 1;01, when the boy tries, unsuccessfully, to imitate the word. From an early age this prompt has the function of making a child participate in a social group, and appears to be effective since Armando reacted in all but one out of forty recorded instances in his data set. Armando began imitating prompts formulated with kech=ti’ more clearly beginning at 1;08, and more consistently after 2;01, though confused and partial responses did continue. From 2;01 onward, use of prompts is infrequent and even absent in speech addressed to Armando. Explicit prompts are used in both direct and indirect speech, and are most frequent as triadics, but do occur in dyadic addressing. Table 2. Frequency of dyadic, triadic, direct and indirect forms. Adult>child

Dyadic

Triadic

Total

Direct

Indirect

Total

3

49

52

47

5

52

11

49

60

52

8

60

David

–

3

3

3

–

3

Total

14

101

115

102

13

115

Sandi Armando

Triadic forms15 were by far the most frequent in speech addressed by adult caregivers to the three children. This common use of the explicit prompt addressed to third parties responds to the presence of different adults and children16 in the recordings, but more importantly it allows the child to be

196 Barbara Pfeiler treated as a person addressed within a group of more than two participants and to familiarize him/her with the triadic conversational routine. Input expressions were mostly direct, and when the children emitted sentences with the prompts they were in direct speech and only dyadic. The dyadic prompts were expressed with the particle kech ‘say’ and a’l(-eh) ‘say (it)’, and less frequently with t’aan(-eh) ‘say it/talk to him/call him’. The triadic prompts, in contrast, usually employed kech=ti’ ‘say it like this to him’ and a’l(=ti’) ‘to say (to him)’; the verb t’aan ‘to call’ as a triadic form was used with people that are seen in the distance or can no longer be seen17.

Concluding remarks This quantitative description shows how caregivers used different language techniques to support toddler’s speech. The three forms used in prompting occurred in dyadic and triadic interactions in direct and indirect speech, meaning the children heard both simple grammatical forms and complex forms like subordinate phrases. In many cases, direct triadic instructions were used to report information, and their function was to demonstrate the language and teach how to interact in specific situations; this characterizes the role of language in early childhood socialization. Analysis of the toddler’s reactions and responses to the recorded prompts showed that from 1 to 2 years of age, the child’s ability to understand what s/he was asked to do and then repeat one-word sentences grew continually. This is just how Mayan parents describe acquisition of Maya, “the child hears it and repeats it” (i.e. the language). After 2 years of age the data contain less frequent use of explicit prompts since the children no longer needed to be motivated to speak and conversations between adult(s) and child(ren) became increasingly complex in both linguistic form and content. So the strategy does not change with age but the number of social circumstances that the child is fully competent to participate grows. Another aspect that may contribute to acquisition is the use of evidentials in the input. The evidentials bin and ki were mainly used by mothers to reformulate a confused infantile message and provide the intended meaning or a grammatically correct form. Other strategies used to prompt toddler speech in daily conversation and in games included questioning, correction, repeating and eliciting routines. The least frequent practise was to correct a toddler’s grammatically incorrect emission18, be it examples of inalienable

“Lo oye, lo repite y lo piensa.”

197

possession, numeral classification system, verbal clause (agreement and aspect auxiliaries) or simply pronunciation. The existence of distinct communicative practises that are closely linked to a specific culture may explain the acquisition of different linguistic forms in the world’s languages (Schieffelin & Ochs 1986). The most commonly used linguistic forms in Yukatek Maya linked to the studied explicit prompts are the imperative forms (in dyadic addressing)19 and the subjunctive mood (in triadic addressing). In previous studies Pfeiler (2003) and Pfeiler & Martín Briceño (1997) documented early and frequent use of imperative forms, as well as subjunctives, though less frequently, and concluded that in acquisition of Yukatek Maya the discursive practise of prompting contributes in a certain degree to producing early development of verbs. In conclusion, the present study demonstrates that prompting routines provide toddlers with a model of speech as such, but that in most cases it has a social purpose, rather than explicit language instruction. In the earlier recordings, the particle kech(-ti’) is linked more to language (i.e. repetition, that it is said this way) and a’l-ti’ to activities implying a conjugation change. Although the studied explicit prompts are not frequent in the context of the overall language, kech(-ti’) contributes most to earlier socialization due to its relatively frequent use and its combined function of teaching behaviour and language. This is especially the case in its triadic form because it represents indirect instruction in which the toddler, instead of being merely an addressee, is introduced as an active and skillful interlocutor in certain situations.

Notes 1. This research has been supported by a grant from the National Council for Science and Technology (CONACYT # 4639 H). I am grateful to Suzanne Gaskins for her comments on this paper. 2. Within communicative practises children learn to attain communicative results by referring to the effectiveness of their role in dialogue with adults. Documented communicative situations include formats such as directing attention towards objects, give-and-take exchanges, appearance routines, disappearance and reappearance, indicating figures in books, question-and-answer sessions, naming games and expansions (Bruner 1978). 3. This research was done as field practice for students in their last year of the studies of Linguistics and Anthropology, Universidad Autónoma de Yucatán, in 2003.

198 Barbara Pfeiler 4. Most children in Yalcobá acquire the Maya language as their mother tongue and Spanish as a second language. Mothers speak Maya to children though fathers speak to them in Spanish since men generally leave the community for work and thus have a better understanding of Spanish. The children’s grandparents speak Maya and their siblings speak Maya and Spanish. Children generally spend the first years of life under the care of the mother and other family members such as grandparents, single aunts (or uncles) and older siblings. As they grow they begin to relate with cousins and friends, meaning they are in surroundings in which speech comes from people of multiple ages and roles. 5. Daltabuit Godás (1992) reports that Yalcobá’s inhabitants have a traditional lifestyle. Women see to childcare, housework and “backyard” income-generating activities like weaving hammocks and sewing huipiles. Single and married young men have replaced farm work with construction work in Cancun, in the state of Quintana Roo, or as labourers on private ranches, meaning they are absent during the work week and return home at weekends. 6. Among the Tzotzil Maya this technique is only used to teach older children and transmit messages to third parties. 7. In Tzotzil the direct report xi ‘(s)he says’ is also used in speech with toddlers with the construction X xi (‘one must say “X”) (Haviland 1998: 188). 8. This form appears to be grammatically intermediate between verb and particle. Although it inflects for all persons and takes a dative object, it cannot take a time, aspect or mood marker, cannot be used as a verb alone in a relative clause, and cannot be questioned, negated or modified in a way that reflects the class of expression it reports (Lucy 1993). 9. Demuth (1986) states that at 30 months of age children make correct transpositions. 10. The particle la ‘they say’ is extremely frequent in adult Tzotzil. The reported xi is also used to quote the ‘interpreted’ expressions of prelinguistic babies (de León 2000: 139–140). 11. Bin is also used to gloss from another language into Yukatek in informal translation. As a special case of this it is often used to comment on non-verbal expressions by infants (reaches, shrugging shoulders, turning their backs, etc.) if one person wants to interpret or comment on this behaviour to another. And it is used when someone’s speech has not been heard. (Gaskins, personal communication, 2006). 12. Making children repeat begins early on, and by the time they have the ability to speak, the habit of repeating is well established. 13. Among Chicano parents the term dile ‘say to him’ is used specifically to teach forms of politeness, to recite the names of people or objects or correct a child’s expression (Eisenberg 1982). 14. Demuth (1986) reports a similar situation among Basotho children. 15. Triadic prompts are also the most commonly used prompts with Kaluli, Mexican-American, Sesotho, and Kwara’ae-speaking children (Demuth 1986).

“Lo oye, lo repite y lo piensa.”

199

16. Children are continually immersed in multiparty interactions, for example with members of the extended family or members of two nuclear families living on the same lot, sharing the same cooking hearth and/or the same clothes washing basin. 17. In Basotho this form is used to distract a child’s attention or change the course of an interaction. 18. Kwara’ae adults translate infantile expressions they consider as language (Watson-Gegeo & Gegeo 1986: 27), while the Inuit, Kaluli and Samoa do not use this strategy to expand confused infantile expressions. Samoan caregivers, for example, tend to respond to children’s unclear messages in a way that forces them to work harder to understand others. Instead of reformulating the message so the child could pretend to signify, or proposing that he confirm or discount it, they choose to either ignore the expression, indicate confusion or laugh. (Ochs & Schieffelin 1995: 80–83). 19. A higher frequency of imperatives in directed speech has been reported for Tzotzil (de León 1999), Ch’ol (de León, this vol.) and Yukatek Maya (Pfeiler 2003).

References Brown, Penelope 1998 Conversational structure and language acquisition: The role of repetition in Tzeltal adult and child speech. Journal of Linguistic Anthropology 8 (2): 197–221. Bruner, Jerome S. 1978 The role of Dialogue in Language Acquisition. In The Child’s Conception of Language, Anne Sinclair Robert J. Jarvella & Willem Johannes Maria Levelt (eds.), 241–256. Springer Series in language and communication. Berlin /Heidelberg /New York: Springer-Verlag. Crago, Martha B., Shanley E. M. Allen & Wendy P. Hough-Eyamie 1997 Exploring Innateness through Cultural and Linguistic Variation. In The inheritance and innateness of grammars, Myrna Gopnik (ed.), 70–90. Vancouver Studies of Cognitive Science 6. New York: Oxford University Press. Daltabuit Godás, Magalí 1992 Mujeres Mayas. Trabajo, nutrición y fecundidad. México: Instituto de Investigaciones Filológicas, Universidad Nacional Autónoma de México. de León, Lourdes 1999 Verb roots and caregiver speech in Tzotzil Mayan acquisition. In Language, Cognition, and Function, L. Michaelis & B. Fox (eds.), Stanford Center for Language and Information, Stanford University.

200 Barbara Pfeiler 2000

The emergent participant: Interactive patterns of socialization of Tzotzil (Mayan) children. Journal of Linguistic Anthropology 8 (2): 131–161. Demuth, Katherine 1986 Prompting routines in the language socialization of Basotho children. In Language Socialization Across Cultures, Bambi B. Schieffelin & Elinor Ochs (eds.), 51–79. Cambridge: Cambridge University Press. Eisenberg, Ann R. 1982 Language acquisition in cultural perspective: talk in three Mexicano homes. PhD dissertation, University of California. Gaskins, Suzanne 1996 How Mayan Parental Theories Come into Play. In Parent’s Cultural Belief Systems. Their Origins, Expressions, and Consequences, Sara Harkness & Charles M. Super (eds.), 345–363. New York / London: The Guilford Press. 2000 Children’s daily activities in a Mayan village: A culturally grounded description. Cross-Cultural Research 34 (4): 375–389. Hanks, William F. 1984 Sanctification, structure and experience in a Yucatec Maya ritual event. Journal of American Folklore 97: 131–166. 1993 Metalanguage and pragmatics of deixis. In Reflexive Language: Reported Speech and Metapragmatics, John A. Lucy (ed.), 127–159. Cambridge: Cambridge University Press. Haviland, John B. 1998 Early Pointing Gestures in Zinacantán. Journal of Linguistic Anthropology 8(2): 162–196. Lucy, John A. 1993 Metapragmatic presentations: reporting speech with quotatives in Yucatec Maya. In Reflexive Language: Reported Speech and Metapragmatics, John Lucy (ed.), 91–125. Cambridge: Cambridge University Press. Ochs, Elinor 1988 Culture and language development. Language acquisition and language socialization in a Samoan village. Studies in the Social and Cultural Foundations of Language 6. Cambridge: Cambridge University Press. Ochs, Elinor & Bambi Schieffelin 1995 The Impact of Language Socialization on Grammatical Development. In The Handbook of Child Language, Paul Fletcher & Brian MacWhinney (eds.), 74–94. Oxford: Blackwell.

“Lo oye, lo repite y lo piensa.”

201

Pfeiler, Barbara 1992 Manuel J. Andrade & Hilaria Máas Collí. Cuentos Mayas Yucatecos, Tomo I (Review). Boletín de la Escuela de Ciencias Antropológicas de la Universidad de Yucatán 103(16): 78–84. 2003 Early acquisition of the verbal complex in Yucatec Maya. In First Verbs: on the way to mini-paradigms, Dagmar Bittner, Wolfgang U. Dressler & Marianne Kilani-Schoch (eds), 379–399. Berlin / New York: Mouton de Gruyter. Pfeiler, Barbara & Enrique Martín Briceño 1997 Early verb inflection in Yucatec Maya. Papers and Studies in Contrastive Linguistics 33: 117–125. Pye, Clifton 1986 Quiché Mayan speech to children. Journal of Child Language 13: 85–100. 1992 The Acquisition of K’iche’ Maya. In The Crosslinguistic Study of Language Acquisition, Vol. 3, Dan I. Slobin (ed.), 221–308. Hillsdale, NJ: Erlbaum. Pye, Clifton & Deborah Rekart 1990 La Adquisición del K’iche. In Lecturas sobre la Lingüística Maya, Nora C. England & Stephen R. Elliot (eds.), 115–126. La Antigua, Guatemala: Centro de Investigaciones Regionales de Mesoamérica. Schieffelin, Bambi B. 1979 How Kaluli children learn what to say, what to do, and how to feel: an ethnographic study of the development of communicative competence. PhD dissertation, Columbia University. 1985 The acquisition of Kaluli. In The Crosslinguistic Study of Language Acquisition, Vol. 1, Dan I. Slobin (ed.), 525–593. Hillsdale, NJ: Erlbaum. 1990 The Give and Take of Everyday life. Language Socialization of Kaluli Children. Studies in the Social and Cultural Fondations of Language 9. Cambridge: Cambridge University Press. Sidnell, Jack 1997 Organizing Social and Spatial Location: Elicitation in Indo-guayanese Village Talk. Journal of Linguistic Anthropology 1 (2): 143–165. Vogt, Evon Z. 1969 Zinacantán: A Maya Community in the Highlands of Chiapas, Cambridge, MA: Harvard University Press, Ward, Martha Coonfield 1971 Them Children: A Study in Language Learning. New York: Holt, Rinehart & Winston.

202 Barbara Pfeiler Watson-Gegeo, Karen A. & David W. Gegeo 1986 Calling-Out in Kwara’e children’s Language socialization. In Language Socialization across Cultures, Bambi B. Schieffelin & Elinor Ochs (eds.), 17–50. Cambridge: Cambridge University Press.

List of contributors

Barbara Pfeiler Unidad Académica de Ciencias Sociales y Humanidades

Paula Gómez López Departamento de Estudios de Lenguas Indígenas

Universidad Nacional Autónoma de México Calle 43 s/n x 44 y 46, Col. Industrial Mérida, Yucatan 97150, Mexico

Universidad de Guadalajara Juan de Mena 279 Guadalajara, Jalisco 44170 México

[email protected]

[email protected]

Penelope Brown

Kristine Jensen de López

Max Planck Institute for Psycholinguistics PB 310 NL6500AH Nijmegen The Netherlands

University of Aalborg Kroghstræde 3 DK-9220 Aalborg Denmark [email protected]

[email protected] Carlos Carrillo Carreón Unidad Académica de Ciencias Sociales y Humanidades Universidad Nacional Autónoma de Mexico Calle 43 s/n x 44 y 46, Col. Industrial Mérida, Yucatan 97150, México

Pedro Mateo Linguistics Department University of Kansas Lawrence, Kansas 66045 USA [email protected]

[email protected] Lourdes de León Center for Higher Studies in Social Anthropology Juárez 87, Tlalpan México, D.F. 01040, México [email protected]

Clifton Pye Linguistics Department The University of Kansas Lawrence, Kansas 66045 USA [email protected]

Index

absolutive, 16–17, 21–23, 44, 47, 49, 50– 53, 56–57, 59, 62–63, 66, 89, 125–126, 129–130, 137, 147–148 acquisition, 1–6, 8–13, 42–44, 46, 69, 71, 73–75, 77–81, 85–86, 94–95, 101, 103–104, 107–108, 115, 122, 144, 148–156, 167–168, 175–180, 183, 185, 188, 191, 196–197 early ~, 6, 46, 69, 86, 168, 178 ergative language ~, 3–4 (see also Ergative) inflectional ~, 42 language ~, 1–6, 8, 16, 42, 46, 69, 78, 85, 103, 119, 121–122, 124, 155, 167–168, 178, 183 later ~, 86 lexical ~, 88, 94 lexicon ~, 103, 113 Maya ~, 185 noun ~, 113 semantic ~, 6 suffix ~, 46 Tzeltal ~, 113 Tzotzil ~, 85 verb~, 2, 5, 103, 113, 194 vocabulary~, 104 affectee, 120, 128, 130, 141, 145 affix, 25, 41 cross referencing ~, 130 agent, 3–4, 6, 69–71, 76–79, 89, 121, 124, 129, 147 Agent case marking, 4 agglutinative morphology, 16 Agr elements, 56 agreement, 16, 18, 25, 42, 47–49, 51, 53– 58, 60–66, 197 absolutive ~, 17, 51, 53

accusative ~, 58–59, 61–62 AgrO, 57 AgrS, 57 ergative ~, 17, 48, 51–54, 58–66, 68 Mayan~, 48 nominative ~, 57, 60, 62 object ~, 57, 58, 63, 64, 69 pronominal ~, 61, 159 subject ~, 30, 56, 58, 63 verb ~, 53 applicative, 21, 25, 26, 36, 44, 146, 159 argument, 6, 48–49, 61, 69–70, 76, 77, 85, 89, 90, 93, 95, 106, 119–132, 138, 140, 144, 146–148 affectee, 128 BEN, 126, 130, 138 core ~, 119, 126, 131 ERG, 130 locative ~, 138 nominal ~, 5, 85 patient ~, 147 verb ~ aspect, 4, 7, 16–18, 20, 23, 25, 37, 42–43, 46, 51, 54, 57, 59, 61–65, 71, 75, 79, 89, 107, 125–126, 133, 137, 147, 192, 196, 198 completive ~, 17, 18, 20, 22, 23, 44, 52, 147, 179 future ~, 179 habitual ~, 179 imperfective ~, 59, 79 incompletive ~, 44, 52, 65, 147 neutral ~, 147 perfective ~, 59, 71, 79 potential ~, 179 progressive ~, 52, 179 Ayoquesco Zapotec, 161, 163

206 Index belief, 113 caregiver ~, 185 benefactive, 6, 119, 120, 122–126, 130, 132, 137, 144, 146–148 benefective, Tzeltal, 131 bias, 86, 94, 119 body part, 6, 157, 158, 163–164, 167, 171, 175–176 literal ~, 158, 175–177 case, 4, 49, 54–63, 65–67, 121, 126 accusative ~, 49, 61 ergative ~, 61 inherent ~, 54, 59, 63 nominative ~, 49 oblique ~, 126 overt ergative ~, 63 case-marking, 121 CDS, 86, 88, 90, 91, 92, 93, 94 Child Development, 7 clause type, 4, 54, 58, 64–65 Complementizer projection, 58 complex, verbal ~, 17, 41, 42, 46 complexity, 42 cognitive ~, 113, 177–178 linguistic ~, 156, 177–178 morphosyntactic ~, 131 semantic ~, 42 syntactic ~, 140 concordance, 4, 70 constraint, language-specific ~, 121 construction, ditransitive ~, 6, 127, 146 three-argument ditransitive ~, 119 three-place ~, 121 containment, 162, 166, 169 context, 1, 3, 5, 17–18, 46, 75, 86, 89, 92–93, 108, 110–112, 128, 130, 132, 137, 139, 168–169, 173, 192, 197 communicative ~, 94

cultural ~, 5, 85, 92, 95, 119 discourse ~, 90, 165, 167 ~ learning, 5, 93 left edge ~, 17 nominal ~, 110 person ~, 22 predicative ~, 110 right edge ~, 20 sentence edge ~, 23 situational ~, 106, 108 sociocultural ~, 1 syntactic ~, 23 correction, grammatical ~, 7 declarative, 93, 191 derivation, 15, 22, 37, 40–41, 42, 46, 104 verb ~, 35 development, 2, 4–7, 23, 41–42, 69, 72, 75–80, 85–87, 92, 106, 108, 111–112, 122, 141, 147, 168, 180, 185, 197 child ~, 7 cognitive ~, 147 early ~, 4, 5, 106, 197 inflectional ~, 46 language ~, 42, 46, 108, 185 lexical ~, 85–87, 92 morphological ~, 5, 69, 77, 79, 85 phonological ~, 41 semantic ~, 6, 119 syntactic ~, 141 verb ~, 23 directive, 93, 166 discourse, 4, 48, 65–66, 86, 90, 93, 165, 167, 187–188, 191 D-language, 60, 61 edge, left word ~, 18–19, 27, 28 right word ~, 21–23, 32–34, 37 elicitation, 2, 148, 184, 186–187, 190 elision, 5, 90 ellipsis, argument, 120, 122, 127, 130, 140

Index ergative, 3–4, 16–18, 26, 44, 47–66, 69, 70, 79, 88–89, 121, 125–126, 130, 133, 137, 147 ergativity, 2–4, 47–50, 54, 57, 60, 62, 64, 66 Mayan, 50, 54 mixed ~, 2, 4 morphological ~, 49–50 split ~, 47, 50–57, 59, 62, 64–66 syntactic ~, 49 error, 41, 66, 137 evidential, 7, 191 expansion, 7 experiencer, 4, 69–71, 76–79 extension metaphorical ~, 158, 163–165, 167, 172

207

indirective, 147 input, 2–3, 7, 15, 23–24, 27–42, 72–74, 77–79, 85–87, 90–92, 94, 104, 109, 111, 113, 119–120, 122, 137–139, 141, 142, 145–146, 179, 196 interaction, dyadic ~, 7, 184, 190 triadic ~, 186, 187, 188, 196 intransitive, active ~, 69, 71, 75–77, 79 inactive ~, 69, 71, 76–77, 79 intransitivity, split ~, 4, 80 Ixil, 52–53 Jakaltek, 52, 63–64

Huichol, 1, 5, 86, 103, 104–107, 109, 111, 112

landmark, 161, 164–165, 167, 169–170, 172–174, 176, 178–179 dimensional ~, 165 language, Eastern Mayan, 48 ergative ~, 3, 4, 47, 49–50, 55–57, 61– 62, 65–66, 79, 88 head marking ~, 89 head-marking ~, 90 mixed ergative ~, 3–4 Tzeltalan, 95 language development, 42, 46, 108, 185 languages, mesoamerican ~, 1, 5–6, 8 lexicon, early ~, 5, 85 locative, spatial ~, 6, 169 locative noun, bodypart ~, 157, 160, 164–65, 167, 168, 175, 176, 177, 178, 179

I-languages, 58–61 imperative, negative ~, 17, 18, 93 positive ~, 3, 17, 19, 23, 30–32, 93

marking, case ~, 59–60, 66, 121 Mayan, Tzotzil, 85–87

factor, discursive ~, 5, 88, 95 left edge ~, 23, 27, 31 right edge ~, 23, 32, 40, 44 form, double object ~, 123 frozen ~, 6, 121 framework, 7, 48, 54, 163, 183 frequency, child ~, 28–29, 31, 33, 35, 37–38, 40 input ~, 28–29, 31, 34–35, 37–38, 40, 42 game, naming ~, 183, 185, 197 generalization, 65, 126 goal, 2, 15, 121, 124, 166

208 Index Mayan language family, Eastern ~, 26, 33, 50–51, 54, 64–65 meaning, constructional ~, 6, 119, 145 method, comparative ~, 2, 47 Mochó, 4, 50–51, 54, 57, 61, 63 mood, 16, 46, 197, 198 Mopan, 51–52, 57, 61, 64 morphology, ergative ~, 63 name, proper ~, 88, 111, 194 Natural Morphology, 71, 81 noun, locative ~, 176–177, 179 region ~, 160, 167–169, 176 object, landmark ~, 161, 164–165, 167, 170, 172–174, 176, 178 trajectory ~, 165, 169, 172–174 order, head-initial syntactic ~, 48 overgeneralization, 42 paradigm, Mayan aspect ~, 17 parameter setting, 3, 12 patient, 59, 69, 71, 76–78, 89, 121, 126, 129–130, 147 pattern, ergative-absolutive crossreferencing ~, 48 PM, 18, 20, 36 Pokomam, 52 positional, 71 possessor, 6, 120–121, 124, 126–128, 131, 147 practice, discursive ~, 1, 6 practises, discursive ~, 183, 185–186

prefix, 15, 17–18, 21, 44 absolutive ~, 17, 51 aspect ~, 23, 63 aspectual ~, 16–18, 44 completive aspect ~, 18 ergative ~, 50–51, 125, 133, 137, 147 incompletive ~, 17 inflectional ~, 3, 40 verb ~, 105 promp, dyadic ~, 196 prompt, 93–94, 187, 194–196 explicit ~, 193, 195–197 triadic ~, 186, 196, 198 prompting, 7, 92, 183, 185–186, 188, 196–197 recipient, 120–121, 123, 126–130, 142, 145 relation, spatial ~, 6, 155–156, 159, 164–165, 167–171, 174, 175–179 relationship, spatial ~, 155–156, 159, 164, 168, 170, 174, 176–177, 179 static spatial ~, 155–157, 160–161, 167, 170, 172 support ~, 176 repetition, 7, 184–185, 188, 190–191, 197 request, 93, 141, 155 root, 3, 15, 16, 18, 21–23, 25–26, 30, 32– 35, 40–41, 46, 65, 71, 75, 108, 120, 135, 137, 148, 186 bare ~, 3, 15, 25–26, 33, 37 canonical ~, 16 CVC ~, 3 ditransitive ~, 120 edge ~, 29, 35 nominal ~, 105 verb ~, 2, 15, 17–21, 23, 25–41, 77, 120, 127, 133, 138, 146 verbal ~, 3, 76, 105, 192 Root Infinitive Hypothesis, 16

Index San Marcos Tlapazola Zapotec, 157, 159, 163, 164, 177–178 Semantic bootstrapping, 3 semantics, embodied ~, 178 Set A, 43, 66, 69–79, 83, 89 Set B, 66, 69–79, 83, 89 socialization, language ~, 2, 92, 95, 183 speech, child directed ~, 18, 24, 29, 34 Mayan Input ~, 24 speech act, 93–94, 141, 145 split, ergative ~, 47, 48, 50–54, 58–59, 61– 62, 64–65 stage, early ~, 104, 121, 135 multi-word ~, 178 one-word ~, 155 two-word ~, 6, 24, 120, 141, 144, 168 status, 3, 16, 18, 20–21, 23, 25–26, 30, 34, 35–38, 40, 41–44, 46, 54, 64, 69, 71, 75, 83, 108, 147 structure, argument ~, 90, 95, 121–122, 132, 144 intransitive verb ~, 56 transitive verb ~, 55 verb root ~, 39 subgroup, Kanjobalan, 50, 52, 54, 58 Mamean, 4, 53–54 Yukatek, 4 subject, 3–4, 16–18, 21, 23, 25, 30, 47, 49–53, 56–64, 69–71, 79, 88–90, 124– 125, 147, 164

209

system, split intransitive ~, 4, 69–70, 79 template, 16 inflectional ~, 25, 33 verb ~, 16, 37 theme, 121, 124, 142 theory, minimalist ~, 55 Tlapazola Zapotec, 163, 165 transitivity, 16, 18–19, 21, 65, 122, 144 verb, 16, 21, 47, 146 transparency, morphological ~, 4, 69, 71, 78–79 unergative, 62 usage-based theory, 46, 82 verb, light ~, 95 spatial ~, 95, 119 verbs, of transfer ~, 119, 124, 126, 127 Wastekan, 50 Wixarika, 103 word, referential ~, 103, 106, 108, 111–112, 116 relational ~, 5–6, 103, 106–112, 117, 119 Yukatek, 1, 4, 7, 15–31, 33–41, 43–44, 57, 69–71, 78–79, 86, 94–95, 183–186, 191, 194, 197–198 Yukatek Maya, 4, 7, 183–186, 191, 194, 197 Yukatek Mayan, 4 Yukatekan, 26, 33, 50–51, 54, 64–65