The Linguistics of Endangered Languages

THE LINGUISTICS OF ENDANGERED LANGUAGES

The Linguistics of Endangered Languages Contributions to Morphology and Morphosyntax

W. Leo Wetzels (ed.)

LOT Utrecht 2009 Published by LOT phone: +31 30 253 6006 Janskerkhof 13 fax: +31 30 253 6406 3512 JK Utrecht e-mail: [email protected] The Netherlands http://www.lotschool.nl/

Cover illustration: Ingaricó girl learning to write. Manalai village, Roraima, Brazil. Photo by Odileiz Cruz, 1999.

ISBN 978-90-78328-98-8 NUR 616

W. Leo Wetzels (ed.)

Contents

Kees Hengeveld (University of Amsterdam) Preface ...... 1

1. Leo Wetzels (Paris III, Sorbonne Nouvelle/CNRS, LPP and Vrije Universiteit Amsterdam) Introduction ...... 3

Part I : South America

2. Jesús Mario Girón (Vrije Universiteit Amsterdam) Morfología y función de las construcciones nominalizadas en wãnsöjöt (puinave) ...... 15

3. Dany Mahecha Rubio(Vrije Universiteit Amsterdam) El nombre en Nkak ...... 63

4. Sérgio Meira (Leiden University Centre for Linguistics) and Spike Gildea (University of Oregon) Property concepts in the Cariban family: Adjectives, adverbs, and/or nouns? ...... 95

5. Eithne B. Carlin (Leiden University Centre for Linguistics) Truth and knowledge markers in Wayana (Cariban), Suriname ...... 135

6. Francesc Queixalós (CELIA/CNRS Paris) La posture du corps dans la classification et la localisation: l’exemple du sikuani ...... 151 7. Willem Adelaar (Leiden University Centre for Linguistics) Inverse markers in Andean languages: A comparative view ...... 171

Part II : Africa

8. Mulugeta Seyoum (Ethiopian Languages Research Centre and Leiden University Centre for Linguistics) Negation in Dime ...... 189

9. Azeb Amha (Leiden University Centre for Linguistics) The morphosyntax of negation in Zargulla ...... 199

10. Anne-Christie Hellenthal (Leiden University Centre for Linguistics) Possession in Sheko ...... 221

11. Kofi Dorvlo (Language Centre, University of Ghana) Noun class system and agreement patterns in Logba (Ikpana) ...... 243

12. Mercy Lamptey Bobuafor (Leiden University Centre of Linguistics) Noun classes in Tafi: A preliminary analysis ...... 267

Part III : Middle East and East-Timor

13. Masayoshi Shibatani (Rice University) and Khaled Awadh Bin Makhashen (Hadhramout University and Universiti Sains, Malaysia) Nominalization in Soqotri, a South Arabian language of Yemen ...... 311

14. Aone van Engelenhoven (Leiden University Centre of Linguistics) On derivational processes in Fataluku, a non-Austronesian language in East-Timor ...... 333

Index ...... 363 Preface

Kees Hengeveld (Amsterdam Center for Language and Communica- tion, University of Amsterdam)

At the recommendation of an advisory committee consisting of Tjeerd de Graaf, Silvia Kouwenberg, Maarten Mous, Pieter Muysken and Leo Wetzels, the Netherlands Organisation for Scientific Research (NWO) started a research programme in 2002 that aimed at encouraging research on endangered languages and at disseminating the results of that research among the peoples concerned, the general public, and the academic community. The Endangered Languages Programme (ELP) implemented the first aim by funding seven projects on languages and language groups from Africa, Asia, and South America. The board of the programme, consisting of Kees Hengeveld (chair from 2005), Rosalyn Howard, Ul- rike Mosel, Stephane Robert, and Rieks Smeets (chair until 2005), and guided and assisted first by Frank Zuijdam and later by Marc Linssen from NWO, selected these projects from a large number of proposals of high quality, which made clear that there is a great interest and an enormous potential for language documentation among Dutch linguists and the colleagues with which they collaborate internationally. This was reason for NWO to ask Maarten Mous, participating in the ELP, to take the initiative for a European programme on endangered languages. Such a programme was indeed approved by the European Science Foundation, and has now started under the name EUROBABEL. Within each of the ELP projects attention was paid to the documentation needs of the peoples concerned, such as the production of teaching materials, dictionaries, and audiovisual materials registering daily life and culture. The general public was informed about language endangerment through cultural events for the interested laymen, and through the development of a multimedia course on language endangerment for use in secondary schools. This course was developed by Gotze Kalsbeek and Cecilia Odé and is freely available in Brazilian Portuguese, Dutch, English, and Russian at the website www.endangeredlanguages. nl. The academic community was kept informed during the programme through a series of conferences, organized by NWO in collaboration with 2 KEES HENGEVELD

Leo Wetzels at VU University Amsterdam in 2004, Kees Hengeveld at the University of Amsterdam in 2007, and Maarten Mous at the Univer- sity of Leiden in 2010. This book too is meant to inform the academic community about the results of the programme. It testifies to the enormous richness of human languages in general, and to the need to pre- serve and document this richness in particular. It samples highlights of the grammars of the languages studied, which are described in full in the language descriptions coming out of the individual projects, which are listed as they appear at the programme’s page at NWO’s website www. nwo.nl. The board of the ELP would like to thank W. Leo Wetzels for taking it upon him to act as the editor of this volume; Katherine Demuth, Pattie Epps, Francesc Queixalós, Anne Schwarz, Miriam van Staden, Hein Steinhauer, Mauro Tosco, and Johan van der Auwera for acting as reviewers; and Jeroen van de Weijer for correcting the English contributions and for preparing the manuscript for publication. Introduction

W. Leo Wetzels (Paris III, Sorbonne Nouvelle/CNRS, LPP and Vrije Universiteit Amsterdam)

The articles in this volume originate from two sources. The manuscripts by Sérgio Meira & Spike Gildea, Francesc Queixalós, Masayoshi Shi- batani & Khaled Awadh Bin Makhashen, and Willem Adelaar were selected by the editor. All others were written by researchers engaged in the Endangered Languages Programme financed by the Netherlands Or- ganization for Scientific Research. The papers are organised according to the geographical distribution of the languages discussed, resulting in the following division, respectively: South America (Colombia, Suriname, Brazil and the Andes), Africa (Ethiopia, Ghana), and Asia (Yemen, East- Timor).

South America The language called Wãnsöjöt Yedókjet, more often referred to as Pui- nave, is spoken by some 5000 individuals. These people are divided into two communities, one of which lives in the region of the Inírida River in Colombia, and the other, a smaller group, resides on the shores of the Venezuelan Orinoco. Wãnsöjöt has sometimes been classified as belonging to the Makú-Puinave linguistic family, together with the Colombian languages Nikak, Kakua, and the eastern (mainly Brazilian) Maku languages. However, there is a lack of strong evidence for a genetic relationship between Wãnsöjöt and any of these languages. Therefore, it must be considered a linguistic isolate for the time being. Although the Wãnsöjöt represent a relatively numerous group, the number of speakers who master the traditional language is rapidly decreasing, mainly because over half of the population now live in urban areas and many children and adolescents opt for Spanish. In the article “Morfología y fun- ción de las construcciones nominalizadas en wãnsöjöt (puinave)”, Jesús Mario Girón discusses the question of nominalization in Wãnsöjöt, which has nominalized structures that function either as predicates or as participants. Deverbal nominalizations create agentive or objective nouns. In addition, action nominals and locative nominalizations may be 4 W. LEO WETZELS derived from verbs, which serve as oblique arguments. Nominalization is not limited to the lexicon; syntactic constituents may also be nominalized. Extended incomplete clauses and clauses that suffix the resultative -pn or the nominal past -hin function as subordinate and coordinate clauses. Furthermore, the conditional construction is formed by the addition of the locative case marker -u to the initial clause (protasis) of the conditional, whereas the past nominal hin, suffixed to one of two clauses in a sequence, creates adversative complex sentences. The Wãnsöjöt data show a close relation between nominalization and predicates of relative clauses and other complement constituents. The Nkak and the Kakua once formed a cultural unit in the Vau- pés region, Colombia, from which part of the population split off and migrated to the North. The 200 or so members of the Kakua people, also known as Bará-Maku, still live in southeast Colombia between the Que- rari and Papuri rivers, both tributaries of the Vaupés. It is the northern group that is known as the Nkak, a nomadic people who were only officially contacted in 1988. Their number is estimated at roughly 550 individuals, all of whom live in the interfluvial region of the middle Guavi- are and the upper Inírida. Soon after official contact, the Nkak suffered a series of drastic changes in their way of life, most notably the reduction of about 40% of their population, their entrance onto the market as un- skilled workers hired by the farmers who surround their land, and, since 2002, the massive displacement of various groups owing to the presence of armed forces in their territory. For these reasons, the survival of the Nkak language and culture is not at all certain. In her paper “El nombre en nkak”, Dany Mahecha Rubio discusses the phonological, morphological, and syntactic properties of nouns, nominalizations and nominal classification in Nkak and compares nominalization and nominal classification with similar constructions in other languages of the region (Wãnsöjöt, Yujup, Hup, Nadeb, and Dâw). In particular, she presents the few derivational processes the language has by which nouns are derived from verbs or nouns, and offers detailed observations on the structure of simple and complex nouns (both of which lack number and gender), the various types of compounds which may have nouns and verbs as their structuring elements, determiners, and case marking. Lastly, Mahecha Rubio provides arguments to distinguish two types of nominal classification. INTRODUCTION 5

The forty or so Cariban languages are spoken over an area that extends from the mouth of the Amazon River to the Colombian Andes and from Maracaibo, Venezuela to Central Brazil. Current estimates for speakers that are fluent in one of the Cariban languages range from 40,000 to 100,000. Given these comparatively impressive populations, the Cariban languages, together with Tupi and Arawak, represent one of the larger linguistic families of lowland South America. While most of the Tupi languages are spoken to the south of the Amazon River, the Ca- riban languages are mainly found to the north of it. The paper “Property Concepts in the Cariban Family: Adjectives, Adverbs, and/or Nouns?” by Sérgio Meira and Spike Gildea contributes to the typological issue of establishing what the universal lexical categories are, giving particular focus to the status of adjectives as a universal lexical category, as recently asserted by Dixon. Most modern descriptions of the Cariban languages have claimed that this language group does not have a separate category of adjectives, but that property concepts are divided between nouns and adverbs. Dixon reanalyzes data from reference grammars of three Cariban languages, claiming that the entire adverb category is better labeled “adjective”, and that the subset of property concept nouns is better considered a separate syntactic category of “Adjective2”. Meira and Gildea first demonstrate that Dixon’s specific analysis is unconvincing, but then that a closer look at the Cariban data yields a syntactic distinction between two subsets of the adverb class, one of which contains exclusively adjectival concepts and could, therefore, be considered a class of adjectives. They continue to suggest, however, that debating whether to call this a ‘subclass of adverbs’ or an ‘independent lexical category of adjectives’ is not an empirical, but rather a theoretical question (and therefore unanswerable on objective criteria alone). They instead investigate the diachronic origins of the unusual distribution of adjective concepts in the pan-Cariban lexical categories, arguing that this leads to a better understanding of the actual situation of “adjectives” and “adverbs” than simple binary classifications. In her paper “Truth and Knowledge Markers in Wayana (Cariban), Suriname”, Eithne Carlin addresses the functionality of a set of grammatical markers in the Cariban language Wayana, spoken by some 1200 people spread out over an area that overlaps with borders between Suri- name, French Guiana, and Brazil. The data presented are taken from the Surinamese Wayana, who live in the village of Pïlëoimë, along the Ta- panahoni River. The Wayana harbour a negative attitude towards their 6 W. LEO WETZELS traditional language, directly attributable to the marginalized position of the indigenous communities vis-à-vis the national context. The Wayana in Suriname are rapidly learning Sranantongo, the lingua franca of the country, which is, however, still inadequate for improving their social standing especially in the state capital of Paramaribo, since Sranantongo, like Wayana itself, is a stigmatized language in formal settings. The grammar of Wayana contains a set of markers that includes a facsimile (similative) marker, several assertive or emphatic markers, a marker with the meaning ‘through and through’ or ‘truly’, a frustrative marker, nominal past suffixes, and evidential marking. Carlin concentrates on the nonverbal categories that express epistemological ideas of realities and truths. There is a pervasive regularity in Wayana and other Cariban languages of encoding states of being or not being, intimately connected with the Wayana world view. As Carlin shows, all the relevant markers, with the exception of the frustrative, express some temporal aspect of permanence vs. non-permanence, revealing an underlying temporal dimension ranging from simultaneity, progressivity, temporariness, to non- fluctuating permanence. Moreover, when nominal categories are marked by the frustrative, insights into cultural norms may be gleamed, but when it appears on verbs, it falls into the ‘knowledge’ domain. Until fifty years ago, the nomadic and semi-nomadic Sikuani lived in small groups in the tropical savannas of eastern Colombia, between the Orinoco, Meta, Manacacias, and Vichada rivers. Beginning in the 1950s, some groups migrated toward the jungle areas of the Eastern Middle Orinoco in Venezuela and the Guaviare river in Colombia. Cur- rently, the approximately 30,000 Sikuani populate both sides of the Middle Orinoco, with 70% of the population being in Colombia and the remainder in Venezuela. The Sikuani language is the largest of the small Guahibo family, which also contains the languages Cuiba (2,500 speakers), Guayabero (1,000 speakers), and Hitnü (250 speakers). It has been observed crosslinguistically that languages use terms related to the human body in a way that often transcends the naming of physical entities, properties, states or events. Sometimes they go as far as grammati- calizing the expressions in which these terms are embedded. The Sikuani language offers a nice illustration of this fact, as is shown by Francesc Queixalós in his paper “La posture du corps dans la classification et la localisation: l’exemple du sikuani”. There are four verbs describing the positions of the body that, on the one hand, construct a genuine categorization of the lexicon of nouns akin to more common nominal classifica- INTRODUCTION 7 tion systems, whereas, on the other hand, these verbs function to construct predications of localization and identification. Moreover, in their occurrences as auxiliaries, they go beyond the notion of space, expressing aspect and even modality meanings. In his paper “Inverse markers in Andean languages: A comparative view”, Willem Adelaar compares the use of inverse markers in the verbal morphology of three unrelated Andean languages: Quechua, Puquina and Mapuche. The languages of the Quechuan linguistic family are spoken by some 8 million people in Peru and also in south-western and central Bolivia, the Andean region of Colombia and Ecuador, north-western Ar- gentina, and northern Chile. Taken together, the Quechuan languages are the most widely spoken indigenous languages of the Americas. Puquina, a linguistic isolate, is now extinct. It was spoken in South Peru, on the border with Chili (Arequipa, Moquegua, Tacna). The Mapuche live in Central and Southern Chile and Southern Argentina. Of the more than one million Mapuche, who make up about 4% of the Chilean population, only some 200,000 may be considered fluent speakers. In Argentina, the Mapuche population is estimated at 60,000. As Adelaar argues, the data from Quechua, Puquina and Mapuche show a comparable development in their personal reference marking systems, in which inverse markers play a crucial role. Inverse markers allow languages like these, which display a limited set of personal reference endings (e.g. with subject markers only, or with an incomplete set of endings encoding both an actor and a patient in a transitive relation), to expand their inventory without having recourse to object markers specified for grammatical person. Instead, the absence of fully specified object markers is compensated by conveying the role of patient to what is normally a subject or agent marker. Inverse markers are used to indicate such a change of roles.

Africa Dime, Zargulla, and Sheko, spoken in Ethiopia, belong to the Omotic linguistic family, a primary branch of Afro-Asiatic, which also includes Berber, Chadic, Cushitic, Egyptian and Semitic. Like many other Omotic languages, Dime (with approximately 5,400 speakers) and Zargulla (with around 7,800 speakers) are endangered. Besides the low number of speakers, there is a strong socio-political pressure put on the minority ethnic groups of south-western Ethiopia in general, which contributes to the endangerment of their languages and cultural traditions. The Zargulla live to the west of Lake Chamo. Their lifestyle is similar to that of other 8 W. LEO WETZELS groups in the area, with the economy based on a balance between agri- culture and herding. The Dime are sedentary and horticulturalists. Two mutually intelligible dialects have been identified within the Dime society, Us’a and Gerfa. The Us’a dialect is the main source of data for Mulugeta Seyoum’s paper “Negation in Dime”. In Dime, negation is expressed through the use of the negative morpheme -káy. This morpheme also occurs following a copular verb. The negative nominal clause is headed by the obligatory negative copular verb yi- and, again, the negative marker -káy. In negative constructions, the aspectual and tense distinctions used in the affirmative are neutralized. Within the Omotic linguistic family, the neutralization of tense and aspect in the negative appears to be a special feature of Dime. On the other hand, it is a general feature of all the Omotic languages to use a special variant -koy of the negative morpheme for the negative imperative and optative. In her paper “The morphosyntax of negation in Zargulla”, Azeb Amha discusses the morphemes of the language that mark clausal negation; five of which are formally related and they are used in main clauses. The distribution of the five negative markers is determined by the predicate type (i.e. auxiliary, simple lexical verb or a complex predicate), tense-aspect and mood. Negation is marked by affixes attached to the verb in the non-past tense of declarative and interrogative clauses and in dependent clauses. In past declarative and interrogative clauses and in the imperative/optative mood however, it is expressed by independent negative verbs. There is partial similarity between the independent negative verbs and negative affixes, suggesting a (historical) link between the two. Amha attempts to account for the formal correspondences among these morphemes by identifying the direction of the historical derivation. Yet another Omotic language reported on in this volume is Sheko, spoken by approximately 40,000 people, who live in the forested hills between Mizan Teferi and Tepi and on the Guraferda plateau in South- west Ethiopia. Together with languages such as Dizi and Nayi, Sheko belongs to the so-called Dizoid group of languages, a subgroup of the North Omotic branch. In her paper “Possession in Sheko”, Anne-Christie Hellenthal discusses the three constructions that the language possesses to express a possessor-possessum relation. Each of these constructions is of interest from a typological or historical-comparative point of view. Firstly, predicative possession in Sheko is similar to existential predication, in accordance with existing typological knowledge. However, the existential construction in Sheko permits only the possessum as a sub- INTRODUCTION 9 ject, contrary to what is found in other languages, for which predicative possession divides into two types, one of which has the possessor as a topic and the other the possessum. Secondly, attributive possession, in particular possessive noun phrases headed by baab ’father’ or bé ‘mother’, show a peculiarity in gender marking, using bé irrespective of gender in definite phrases, even though masculine is the default gender in Sheko. Hellenthal also discusses possessor ascension in relation to body part nouns, taking into consideration inalienable possession in the related language Dizi. The author shows that the two constructions involved have different semantics, i.e. possessor ascension emphasizes the part (the possessum), while possessive noun phrases emphasize the whole (the possessor). Sheko and the other Dizoid languages remain atypical in marking inalienable possession by a case marker employing possessor ascension. Logba and Tafi belong to the southernmost cluster of the Ghana- Togo Mountain (GTM) languages, a subgroup of the Niger Congo family. Logba and Tafi are two of fourteen different languages spoken in the hilly regions of the Ghana-Togo frontier, in the Volta Region of Ghana. Logba, spoken by roughly 7500 people, and Tafi, with approximately 4,400 speakers, are threatened by Ewe, the surrounding regional language to which both adults and children are shifting and (Ghanaian) Eng- lish, the official language of Ghana. In his paper “Noun Class System and Agreement Patterns in Logba (Ikpana)”, Kofi Dorvlo studies the intricate noun class system of the language, where ‘noun class’ is understood as a set of nouns that can be characterized by ‘a single set of morphological concords’. In noun class languages of the Niger Congo family in general, nouns have a particular prefix in the singular, while a different prefix is used in the plural. These languages usually also show a system of morphological concord between a noun and the verb. The GTM languages are reported by most researchers to have noun class systems and Logba is no exception as it shares the general noun class features of the Niger-Congo languages. Dorvlo argues that Logba noun classes correspond to semantic groupings, which contain subgroups with a somewhat looser semantic definition. There is a limited agreement within the head noun of the NP and a group of morphemes comprising a subset of the cardinal numerals (when used as modifiers), the demonstrative, and the interrogative. Moreover, there is general agreement between the subject NP and the verb, which is con- 10 W. LEO WETZELS trolled by the argument in subject position: the verbal agreement prefix is selected in function of the noun class that functions as the subject. Just like Logba, Tafi has an active noun class system in which each noun belongs to a particular class identified by a prefix, as is shown by Mercy Lamptey Bobuafor in her paper “Noun Classes in Tafi: A Pre- liminary Analysis”. As in Logba, the head noun determines the type of affix that is used for agreement with the verb. The author argues that ten different noun classes must be distinguished for Tafi, based on the distribution of the class prefixes, the subject-verb agreement markers that are used to cross-reference the subject arguments on the verbs, and the shape of their pronominal forms. As in Logba, some modifiers show concord with the head noun while other word classes, like the adjectives and the ordinals, do not show such agreement. Some modifiers can be substantiv- ized by adding class prefixes and the pronominal forms of the various noun classes to them and thus they function as nouns. The way borrowed nouns are integrated into the noun class system was also examined. Loans may also be allocated to a given class based on phonological and/or semantic criteria, or are assigned to the default word class.

Asia Soqotri is a pre-Islamic Modern South Arabian language. With some 50,000 speakers, it is the native language of the Soqotrans, spoken on the islands of Soqotra, Abd-el-Kuri, and Samhah off the southern coast of Yemen. The language is regarded as endangered, since Arabic has become the official language on the island and because the youth over- whelmingly exhibit fluency in Arabic and an imperfect ability in their traditional language. The paper “Nominalization in Soqotri, a South Ara- bian language of Yemen” by Masayoshi Shibatani and Khaled Awadh Bin Makhashen argues for the nominalization analysis for several of the so-called headless NP constructions including the ones known as headless relative clauses. Contrary to the popular deletion analysis for expressions analogous to John’s (as in This book is John’s), (We must help) the poor, and (I am reading) what John recommended to me, arguments are advanced for the analysis of treating these as grammatical nominalizations, which in their basic function play a referring function. Soqotri, along with a number of other languages such as Chinese, Japanese and Lahu, makes use of the same nominalization morphology for all these constructions. An important distinction, however, is drawn between event nominalizations, on the one hand, and argument nominalization INTRODUCTION 11 and what is called genitive nominalization. Soqotri presents a clear case where different types of argument nominalization function as a modifier element of the relative clause construction, suggesting that relative clauses are nominalization-based rather than involving full subordinate sentences or clauses as in the standard analysis. Fataluku is a non-Austronesian language spoken in the Lautém district in the eastern part of the Republic of East Timor. It genetically belongs to the non-Austronesian Timor-Alor-Pantar branch of the Trans- New-Guinea phylum. The Fataluku speech community is estimated to be comprised of around 30,000 people. Although the constitution of East Timor acknowledges that all the languages spoken on the territory of the republic must be protected, the promotion of Tetum as the language of communication in education and public life and its status as the co- official language (alongside Portuguese) has caused a shift from Fataluku to Tetum. It is for this reason combined with the redistribution of power (and prestige) in the newly created state, that the transmission of Fata- luku to the younger generation is seriously declining. In his contribution to this volume “On Derivational Processes in Fataluku, a Non-Austro- nesian Language in East-Timor”, Aone van Engelenhoven discusses the (very few) productive derivational morphological processes of the language. The main nominalizing device of the language is a nominalizing morpheme /-(n)u/, which may be suffixed to the same roots as the verbalizing morpheme /-(n)e/. These mechanisms, which are mutually exclusive, give rise to formations like tupur-u ‘woman’ versus tupur-e ‘be a woman, feminine’. Formations of this type raise the question as to whether the roots to which these suffixes attach are pre-categorial. One of the few productive morphological processes of Fataluku takes state verbs as its input and creates outputs that function as modifiers (adverbs) of to the following verb. The same process of deverbal adverbialization creates serial verb constructions in which the derived adverb fills the object slot of the bivalent verb. The noun referring to the instrument, product, or theme participant of the scene evoked by the verb is encoded as the object of ‘take’ in the directly preceding clause.

Within the remit of the general theoretical debate, the contribution of theory-directed data is crucially important to improve data-oriented theory formation. For any kind of theoretical linguistics, the discovery of new linguistic patterns is a matter of great importance, since the empirical validity of theoretical models is rightly put to the test continually by 12 W. LEO WETZELS confrontation with the balance as it is actually found to exist between both the striking similarities between languages and also the differences between them. If considered from this perspective, the usefulness of the study of languages that are at the edge of disappearance is self-evident. It is the shared hope of all the contributors to this book that the topics discussed here serve to emphasize the importance of data from endangered or dying languages to further develop the field of linguistics. Part I : South America

Morfología y función de las construcciones nominalizadas en wãnsöjöt (puinave)

Jesús Mario Girón (Vrije Universiteit Amsterdam)

1. Introducción* Se entiende corrientemente que ‘nominalización’ es “volver nombre un verbo o un adjetivo” generalmente con la adición de algún elemento morfológico, dando por resultado el nombre de la acción (‘action nouns’, infinitivos, gerundios) o nombres para los argumentos del verbo (agente, paciente, localidad, manera, propósito, etc.).1 En esta presentación trata- remos sobre las nominalizaciones en wãnsöjöt2 en un sentido amplio,

* Este estudio se hizo gracias al apoyo de la Organización Holandesa para la Investigación Científica (NWO) y hace parte del Programa de Investigación de Lenguas en Peligro (Endangered Languages Research Program) y del proyecto A Grammatical Description of the Puinave and Nikak Languages (Occidental Maku, Colombia) en el cual participé con la beca 256-00-520. Agradezco al profesor Leo Wetzels y al profesor Francesc Queixalos por sus críticas; cualquier error es responsabilidad mía. JMG. 1 Comrie & Thompson (1985: 358) explícitamente indican que los ‘action nominals’ son frases nominales que contienen un nombre derivado de un verbo, pero que también tienen “uno o más reflejos de una proposición o un predicado” (argumentos internos, determinaciones adjetivales o adverbiales), siendo una instancia media entre la frase nominal y la oración. 2 El wãnsöjöt yedokjet o puinave es hablado por unas 5000 personas en la cuenca del Río Inírida. Esta lengua ha sido clasificada como perteneciente a la familia lingüística Makú-Puinave, pero las evidencias sólo indican una lejana relación; por lo que debe considerársele como aislada. La notación de los ejemplos es parcialmente fonética y parcialmente fonológica: el fonema /h/ se escribe j; los alófonos aproximantes de las vocales /i/ y /u/ en ataque se escriben y y w; el fonema glotal // se representa por (’). No se representa la sonorización de explosivas /p,t,k/ en coda seguidas de vocal en límite morfémico con el fin de reflejar la representación fonológica de las raíces léxicas; las explosivas en coda seguidas de los morfemas AGT y AL no siguen dicha sonorización. La explosiva alveolar (d) en posición intervocálica y después de explosiva dorsal 16 JESÚS MARIO GIRÓN mostrando aquellas que se hacen sobre recursos morfológicos y crean nuevas palabras, y otras nominalizaciones funcionales donde constituyentes de tipo frasal y clausal, cuya cabeza léxica es un verbo, funcionan como nominales.3 En wãnsöjöt se hace un extenso uso de construcciones nominalizadas para designar la acción como acontecimiento o estado y para designar participantes de la oración. Las construcciones nominalizadas en sentido estricto se hacen principalmente con recursos derivativos. Las nominalizaciones pueden funcionar como predicado de cláusulas nominales, o como argumento nuclear o periférico (adjuntos) de cláusulas finitas. Las construcciones nominalizadas pueden tener la forma de deri- vación verbal, de frase verbal con argumentos, o de cláusula subordinada en función de complemento (cláusulas adverbiales o cláusulas coordina- das de oraciones complejas). Así, encontramos nominalizaciones formadas por la secuencia V+morfema derivativo: sin-yu //cernir-NMZ// ‘cernidor’, o frases verbales: {MP-V+AGT}: ka-sek-e //3PL-robar-AGT.SG// ‘el que los roba’, {MP+V+caso}: ja-map-a //3SG-pescar-AL// ‘a/para su pesca (de él)’, o cláusulas relativas con sus propios argumentos, como se muestra en el siguiente ejemplo (la cláusula relativa se indica entre corchetes; la co-referencia de argumentos se indica mediante subíndices):

(1) tiwai-ati kaj-duk yem-otj [{i-tk-di’k-ot}j-kom a-towai-atk]C.REL. niño-ERG 3PL-ver trampa-PL ATR-COMP-dejar-PL-MIR 1SG-primo-ERG El niño vio las trampas (las) que -me di cuenta- dejó mi primo.

Antes de describir las nominalizaciones se hace una corta presentación del módulo verbal en cláusulas finitas en wãnsöjöt, con base en el cual se pronuncia vibrante. Los tonos reflejan la forma en superficie; las sílabas no marcadas se han de interpretar como tonos bajos [L]. Para una descripción de la fonología segmental ver el capítulo correspondiente en Girón (2008), y para la descripción del tono ver Girón & Wetzels (2007). Al final del artículo se provee una lista con las abreviaturas usadas en los ejemplos. 3 Noonan (2005) señala diez tipos de nominalizaciones en lenguas Sino- tibetanas, más allá del sentido estricto de la conversión de un verbal en un nominal. Ver también http://www.uwm.edu/~noonan/VersatileNom.pdf. Las lenguas amazónicas y en general las lengua nativas de América hacen extenso uso de diversos tipos de nominalización, en opinión de M. Koptjevskaja-Tamm (2006). CONSTRUCCIONES NOMINALIZADAS EN WÃNSÖJÖT 17 contrastaremos las construcciones nominalizadas. Seguidamente presentamos las nominalizaciones más frecuentes, diferenciándolas por los tipos semánticos resultantes (nominalización de participantes y de acción nominalizada simple o extendida con participantes internos) y por el tipo de mecanismos morfológicos (formantes nominalizadores derivacionales y flexivos que corresponden aproximadamente a las nominalizaciones de participante y de acción nominalizada).

2. Generalidades sobre la estructura morfosintáctica y en particular de la palabra verbal finita en wãnsöjöt 2.1 Rasgos generales de la lengua En wãnsöjöt las raíces léxicas pueden ser libres (generalmente aquellas que corresponden a nombres, pronombres, demostrativos y partículas adverbiales) o ligadas (nombres dependientes, verbos); ambas clases tienen marcaje tonal. No hay una clase léxica de adjetivos, sino que esta función se expresa con diversas morfologías (adjetivos verbales (verbos estativos), derivaciones, composición verbo-nominal). Los nombres flex- ionan en número (SG=Ø, PL= -ot), y el orden de determinación entre dos nominales en una relación genitiva, posesiva o atributiva es determinante-determinado (wen-jupuk //cerro-coronilla// ‘cima del cerro’). Hay un marcaje de caso gramatical -at (ERG) sobre el nominal en función de argumento agente,4 homófono con la marca de argumento dativo, el cual

4 Para las funciones gramaticales utilizamos la nomenclatura U = argumento único de intransitivas, A = agente o argumento 1 de transitivas; P = paciente, objeto o argumento 2 de transitivas. Se usa esta notación de tipo semántico porque en puinave no se puede postular que la organización sintáctica esté regida por la relación Sujeto-Predicado, sino por una elección semántico- pragmática de argumentos más o menos potentes en foco. Sólo en predicaciones con nominalizaciones de agente o paciente podemos hablar de una relación Su- jeto-Predicado (ver §3.1) como una relación sintáctica de Sujeto en cláusulas intransitivas bifurcadas (dos constituyentes) como función establecida por concordancia y orden de constituyentes. En tales casos tendremos las construcciones nominalizadas agentiva con S(A) y atributiva con S(U~P). Fuera de estos casos puntuales, las convenciones S,O,V sólo se usan para designar los órdenes de constituyentes suplidos por los argumentos agente, paciente y predicado verbal o nominalizado. 18 JESÚS MARIO GIRÓN se indica en las glosas como OBL. En wãnsöjöt hay una tendencia a la polisíntesis mediante aglutinación, lo que se manifiesta en la prefijación de marcas de actancia, en la posibilidad de incorporar el nominal del objeto, en la sufijación de marcas de tiempo, de modalidad de oración y de caso sobre las raíces léxicas, y en la afijación de determinaciones léxicas adverbiales de manera y aspecto sobre predicados verbales. Los marcadores de persona (MP) funcionan como posesivos con raíces y frases nominales; los radicales verbales pueden prefijar dos MP correspondientes a los argumentos paciente y agente respectivamente en un orden rígido (MP(P)-MP(A)-V). Hay morfemas que alternan con los marcadores de persona definida en la zona preverbal, actuando como índices o nexos con argumentos no definidos. Estos son: el morfema atributivo i- (ATR), la marca de “objeto no definido” ebu- (OND) y la marca de complemento o recesivo -t- (COMP/RCS). Normalmente las oraciones presentan al menos dos constituyentes: el predicado (verbal o nominal) y el argumento base de predicación o sujeto en el orden VS, siendo el orden SV una estrategia de focalización. Pero es posible tener predicaciones activas con sólo la palabra verbal como constituyente único, y en tal caso son obligatorias al menos un MP y una marca de modalidad ilocutiva, de tiempo o de aspecto. Estas palabra-cláusulas son formas verbales finitas. Cuando en la predicación verbal finita están presentes los constituyentes nominales de los argumentos nucleares en la cláusula, los constituyentes pueden presentar cualquier orden, excepto el orden *OSV, es decir el orden interno de la palabra verbal finita. Por su parte, las cláusulas cuyo predicado es alguna palabra verbal nominalizada (nominalizada ya sea por la marca de agentivo o por el atributivo) tienen restricciones para tener el predicado en posición inicial (*VSO, *VOS). En cláusulas transitivas es marcado el agente con la marca de ergativo, y en cláusulas atributivas (predicado nominalizado por morfema atributivo) está marcado de la misma forma el argumento único de verbos activos (ergatividad escindida).

2.2. Sobre el módulo verbal La secuencia MP-V constituye un módulo verbal con importantes impli- caciones en la derivación morfológica y en la sintaxis. Este par de morfemas son los constituyentes mínimos de una cláusula. Por ello, en wãnsöjöt la mera adición de una marca discursiva a este módulo hace predicación: ka-yoi-at //3PL-reir-CPLT// ‘ellos se rieron’. Por ello se CONSTRUCCIONES NOMINALIZADAS EN WÃNSÖJÖT 19 puede proponer, que en wãnsöjöt hay una regla de formación de palabra que demanda que una raíz, categorizada como ‘verbal’, requiera obliga- toriamente ser prefijada por una marca de persona, o marca de argumento. Esta condición conlleva a que existe un nodo ‘frase verbal’ del que dependen, al menos un argumento y un verbo como nodos terminales. En cláusulas finitas intransitivas este puesto de argumento es suplido por el argumento único {MP(U)-V}y en cláusulas transitivas finitas por el argumento agente {MP(A)-V}, como se indica en la figura 1. El argumento 2 (objeto o paciente), aunque implementado como un marcador de persona prefijado al módulo antes mencionado, parece estar bajo el nodo del que también depende el módulo antes mencionado. Es decir, este segundo argumento es ‘hermano’ del módulo {MP(A)-V} en su conjunto, tal como se indica la figura 1 a), y no de cada uno de sus constituyentes. Si fuese de está última forma, la frase verbal transitiva compacta sería un módulo de estructura plana, como se indica en b) de la figura 1.

Fig.1. Representación de los argumentos del verbo en wãnsöjöt

(2) a. FV b. *FV

FN(P) FV MP(P) MP(A/U) V

FN(A/U) V

MP MP

ka- a- wp 3PL 1SG flechar Yo los flecho

En este sentido, la palabra verbal de una oración finita como en (3a) tiene la misma estructura que toda la oración en (3b). Esta hipótesis permite analizar algunas de las derivaciones y construcciones nominalizadas que se mostrarán en este artículo. 5

5 Son varias las secuencias de palabras y estructuras de palabras del wãnsöjöt que corroboran esta hipótesis, tales como la inserción del atributivo en la zona 20 JESÚS MARIO GIRÓN

(3) a. ka-a-sikom-pan-dei bi-kable-ot 3PL-1SG-SIML-cortar con machete-dejar 1PL-cable-PL Estoy dejando cortados nuestros cables (TH. relato 6) b. d’ a-bt’k fuego 1SG-encender Prendo/í fuego

Una palabra verbal transitiva finita puede presentar como único marcador el marcador del objeto a condición de que se haya sacado el agente del verbo para expresarlo léxicamente, generalmente precediendo la palabra verbal, y marcado tal constituyente con el caso ergativo -at, como se ve arriba en la cláusula matriz en el ejemplo (1).

3. Variedad de construcciones nominalizadas en wãnsöjöt Presentaremos las construcciones nominales en tres grupos, según las características semánticas de las construcciones resultantes: nominalizaciones de participantes, de instrumento y lugar, y de acción nominalizada. A nivel morfológico las dos primeras nominalizaciones tienen la forma de palabra o frase verbal, mientras que la acción nominalizada puede ser tanto una palabra derivada, una frase nominal o una cláusula. En puinave no es relevante distinguir entre ‛palabra’ y ‛frase verbal’ ac- tiva en razón de la estructura del módulo o frase verbal presentada arriba (MP-MP-V). Pero es posible formar palabras derivadas sobre una base léxica verbal sin expresión argumental, las cuales pueden funcionar como un nominal ({V-nominalizador}FN), y que luego del proceso de formación léxica, pueden tomar una marca de argumento en función de poseedor o caso genitivo: FN/MP-{V-nominalizador}FN. Esto sucede en derivaciones con el nominalizador -yu (NMZ) o el agentivo (AGT): ma- {map-yu} 2SG-{pescar-NMZ} “tu anzuelo”. También hay nominalizaciones que se constituyen directamente a partir del módulo verbal (MP- V), como sucede con el formante jet- que forma el nombre de instrumentos (ver 3.2.3), o el formante -dipn que forma el nombre del lugar donde se realiza una acción (ver 3.2.1) de actancia (ver §3.1.2.) y la estructura de los verbos separables, en la cual cada morfema constituyente del verbo toma un MP: (MP-V-MP-V): ja-bík-ka-sãk- ma ja-jaá //3SG-pensar1-3PL-pensar2-RPR 3SG-sobre// “dizque ellos se quedaron pensando sobre él” (CM cuento Jṍmdom “Abuelo de perezozo”). CONSTRUCCIONES NOMINALIZADAS EN WÃNSÖJÖT 21

En cuanto a la función de las nominalizaciones en la cláusula, unas construcciones funcionan exclusivamente como argumentos, ya sea como argumentos nucleares del verbo o adjuntos, tal como sucede con las palabras o frases formadas con el nominalizador -yu o con el alativo- a, mientras que otras construcciones pueden funcionar tanto como predicados o como complementos. Ejemplos de este último caso son las nominalizaciones de frases verbales atributivas que inician por el morfema i- ATR, la derivación agentiva, o la palabra verbal no finita formada con el eventivo -dik (verbo existencial en función de auxiliar). En los siguientes cuadros se presentan las nominalizaciones según los tres tipos semánticos de nominalización mencionados arriba, indicando la categoría funcional formada:

semántica construcción y formante nomi- etiqueta categoría nalizador (en negrilla) del for- funcional mante resultante agentivo {MP/ATR-V}-v-(PL) AGT nominal atributivo, agen- i-{(MP)-V}-(RES)-(PL) (N) ATR predicado, tivo (en v. esta- comple- tivos e intransi- mento, ad- tivos activos), jetivo verbal participante {MP-NEGV-V}-i NZNEGV adjetivo negativo habituativo, {V}-noba-{M/F} HABP adjetivo cualificativo Cuadro 1. Nominalizaciones de participantes y de atributos semántica construcción y formante nomi- etiqueta categoría nalizador (en negrilla) del for- funcional mante resultante locativo {MP-V}-dipn NZLOC nombre derivado contenedor {(MP)-V}-ot TC nombre derivado instrumento jet-{MP-V} INS nombre derivado instrumento, {(MP)-V}-yu NMZ Nombre nombre abstracto derivado de acción Cuadro 2. Nominalizaciones de instrumento, lugar y acción 22 JESÚS MARIO GIRÓN

semántica construcción y formante nomi- etiqueta del categoría nalizador (en negrilla) formante funcional resultante resultativo, {(FN/MP)-V-(AGT)}-pn-(PL) RES nombre de- perfectivo, rivado, pre- acción como dicado hecho atributivo pasado nomi- (MP-V}-jin PSNM predicado, nal, caducidad, cláusula adversativo apódosis en oraciones disyuntivas proposición, {( FN/ATR/MP)-V}-dik EV predicado, acción como comple- hecho mento cláusula relativa Determinantes nominalizantes: casos, partículas adverbiales y otras localizaciones evento (sede, {( FN/MP)-V}-a AL oblicuo, meta) dativo subordinada existencial no-finita ámbito, evento, {( FN/ATR/ MP)-V}-u ADH oblicuo, acción como cláusula sede prótasis en condicional adverbial fi- (MP-V}-jee FIN oblicuo nalidad/razón adverbial de (FN/MP-V}-tee INES oblicuo localización Cuadro 3. Acción nominalizada (estados, eventos, hechos)

3.1 Nominalizaciones de participante Se presentan aquí las nominalizaciones que tienen por dominio una frase verbal derivada, y cuyo resultado semántico es una nominalización de argumento. En unos casos las nominalizaciones pueden argumentos internos expresados por índices pronominales ligados, o por frases nominales adjuntas. Mostramos primero las nominalizaciones de agentivo y de atributivo que corresponden a nominalizaciones de los argumentos agente y objeto/paciente del verbo respectivamente; luego se presentan dos derivaciones de participante de poca productividad. CONSTRUCCIONES NOMINALIZADAS EN WÃNSÖJÖT 23

3.1.1 Nominalización de agentivo La derivación de agentivo (AGT) 6 sufijada a la raíz verbal crea la des- ignación del participante que efectua7 la acción -{V-AGT}, pero este con- structo no puede ser expresado sin adicionar un marcador de objeto (argumento paciente), con lo cual se forma una frase verbal nominalizada agentiva (4a,b) {MP(P)-{V-AGT}}. El paciente también puede estar expresado por una incorporación nominal (6a,b), o de manera indefinida mediante los morfemas i- (ATR) (4d,e) o ebu- (OND) (4c,e).8 Otra vez aquí la estructura sintáctica interna es similar a las vistas para -yu y -dipn.

(4) a. ’m preguntar, pedir ja-’m- el que le pide /pregunta 3SG-pedir-AGT ja-’m--t los que piden/preguntan 3SG-pedir-AGT-PL

6 El agentivo tiene dos alomorfos: a) copia vocálica del último núcleo de la raíz y b) nasal labial. El primero de ellos produce cambios morfofonológicos en radicales con coda dorsal (k→k/h/__) (4c,d). Otro grupo de verbos terminados en /k/ presentan un segundo alomorfo, el cual es la sustitución de la coda dorsal por una nasal labial (4b). El plural del primer alomorfo se hace sufijando -t a la copia vocálica (4a); el plural del alomorfo nasal labial se hace sustituyendo la coda dorsal del radical por -t (4b). La melodía tonal de los agentivos AGT no es previsible, aunque hay tendencias (ver Girón 2008). 7 Una caracterización aproximada del rol del argumento de esta construcción es el de “actor” propuesto por la RG (Van Valin & LaPolla 1997: 141 y sgts.), el cual incluiría agentes de transitivas y argumentos únicos de verbos intransitivos activos (A y U(A)) 8 En las lenguas Caribe hay una estructura similar para la nominalización de agentivo (sujeto) como en tiriyo (Meira 1999: 167-72), donde dicha nominali- zación tiene un prefijo de argumento objeto (poseedor?). En dicha lengua la referencia al agente puede distinguir entre agentes reales o potenciales -ne/to(- no) respectivamente. En verbos intransitivos también hay dos paradigmas para agentes y objetos reales y potenciales. En wãnsöjöt hay distinción en cuanto a definición o indefinición sólo para los objetos, mediante el paradigma de MP definidos y las marcas i- y ebu- de indefinición. 24 JESÚS MARIO GIRÓN

b. puk envolver ja-pu(k)-m el que lo envuelve 3SG-envolver-AGT.SG ja-pu(k)-t los que lo envuelven 3SG-envolver-AGT.PL c. ’ok chuzar ebu-’o(k)j-o el que chuza algo OND-chuzar-AGT d. tak pisar i-tak-a el que pisa ATR-pisar-AGT e. -bep trabajar ebu-bep-e~ i-bep-e el que trabaja algo OND-trabajar-AGT ATR-trabajar-AGT

En la estructura de esta derivación el agentivo (AGT) y el verbo están bajo un nodo nominal (figura 2). Y a diferencia del módulo verbal de la palabra finita (figura 1), el participante que está bajo el mismo nodo nominal que el verbo nominalizado es el paciente, mientras que el marcador de agente sale de la frase verbal, donde sólo queda representado de manera indiferenciada por toda la construcción regida por el nominalizador. Dado que el agente aún no está especificado, esta construcción exige la expresión léxica de dicho argumento en un constituyente que debe preceder a la nominalización en un orden rígido AV. De ello resulta que dicha frase nominal es co-referencial con el predicado nominalizado {P-V-AGT} y está focalizada en inicio de cláusula (o inmediatamente antes de la nominalización) por una marca de modalidad de oración (-da de asertivo (6a), -te de interrogativo (6c), -ya de futuro, etc.). Esto lo inter- pretamos como una construcción clivada donde hay una relación Sujeto- predicado, y parece que por esta razón en esta construcción el agente nunca tiene la marca -at de ergativo. La estructura de la nominalización se representa en la siguiente figura: CONSTRUCCIONES NOMINALIZADAS EN WÃNSÖJÖT 25

(5) Figura 2. Representación de la predicación con palabra verbal agentiva Cláusula

FN FN

V NMZ (AGT)

jede(A)-da a(P)- ’m -(A) DEM-ASR 1SG preguntar-AGT Aquel/la (es) quien me preguntó.

Las frases verbales nominalizadas con el agentivo funcionan como predicado de una cláusula donde se focaliza el agente. En esta construcción la expresión léxica del paciente (OD) puede ir opcionalmente, ya sea en posición final (6c) y sin marca alguna, o topicalizado en inicio de oración como en (6d). El objeto indirecto o participante beneficiario usualmente va después del verbo y marcado con la marca -at de oblicuo (OBL) (6a):

(6) a. j boni-da dapyuij-b-mi denk-at DEM hombre-ASR anillo-dar-AGT mujer-OBL Este hombre fue el que dio el anillo a la mujer (lit: este hombre fue anillo-dador a la mujer) b. bi-ya-jam ja-juyakj-si- ’jok-oi WH(quién)-FUT-INQU 3SG-cabeza-PROS bajar algo-AGT ¿Quién se encargará de bajar su cabeza (de cierto personaje que en una narrativa fue trozado por las hormigas para bajarlo desde la cima de un cerro)? (MMr1) c. juan-te ma-pou-o (mam) Juan-INT 2SG-golpear-AGT PR2SG ¿Juan fue el que te golpeó? d. naa yamsi ji, jede-da ja-map-a DEM pez ‘bocón’j TOP DEMi-ASR 3SGj -pescar-AGTi (En cuanto a) este ‘pez bocón’ aquel lo pescó (aquel fue su pescador) 26 JESÚS MARIO GIRÓN

Rara vez la frase agentiva aparece en función de argumento oblicuo, o como predicado de cláusula subordinada. Las cláusulas con esta nomi- nalización como predicado no tienen mucha frecuencia, pero son necesarias para introducir información de manera concisa cuando se quiere identificar el agente. La palabra agentiva puede cumplir con los roles semánticos de agente de transitivas (ver ejemplos arriba), o de causa como en (7), cuando se adiciona el causativo a la base verbal.

(7) bii jaj-jui mo-pin-yui , o’oi-da jaj-kik-’-mi PRIND 3SG-oir PNE-relatar-NMZ PRNE-ASR 3SG-llorar-CAUS-AGT Ella escuchó una noticia, eso la hizo llorar

La cláusula con predicado agentivo puede ser usada como cláusula complemento de objeto (8a), o como relativa (8b). En el primer caso la cláusula agentiva entera es el complemento de la matriz; en el segundo caso el sujeto (agente) de la relativa es el objeto de la cláusula matriz:

(8) a. dumat jaj-pisi-kai-wak-at epin-oti-at [bjadik-u Ahora 3SG-saber-3PL-saber-CPLT persona-PL-ERG quién-ADH 9 ja-’i’-m jose-jit ]j C.COMPL. de O. 3SG-matar-AGT José-PSNM Ahora sabe la gente quién fue el que mató al finado José b. osm-da jaj-bii-jeu dumat jaj-bii-’i’k-dik Por eso-ASR3SG-1PL-buscar ahora 3SG-1PL-matar-EV [j jak-’i’-mj-pn ]j C.REL. de O. (THr8) DEM3PL-matar-AGT-RES Por eso lo buscamos ahora para matarlo a ese que lo mató

3.1.2 Nominalización atributiva o de objeto El morfema i- (ATR) hace que la secuencia de radicales verbales situados a su derecha, y sus marcas asociadas, se interpreten como atributo del argumento único de cláusulas intransitivas o del argumento paciente expresado por fuera de la secuencia nominalizada. El morfema atributivo no designa una persona gramatical sino un nexo entre una propiedad y un argumento, de tal manera que dicho argumento queda indeterminado en

9 El alomorfo de base del pasado nominal es -jin (ver §3.2.2). CONSTRUCCIONES NOMINALIZADAS EN WÃNSÖJÖT 27 cuanto a la categoría Persona. Si el radical al que afija el atributivo es un verbo intransitivo estativo, se crea un adjetivo verbal (9a,b) y el argumento único asociado a esta palabra tiene un rol de sede o tema del atributo10. Si el radical es un verbo intransitivo activo la palabra resultante expresa un agentivo (9c,d) y el argumento único sobre el que predica la palabra atributiva cumple un rol de tema o agente de verbo iner- gativo:

(9) a. i-pik lo negro ATR-negro b. i-mon lo hondo, que es hondo ATR-hondo c. i-bok el que camina ATR-caminar d. i-jap-ot los que reman ATR-remar-PL

Si el radical es un verbo transitivo, el argumento representado por la con- strucción atributiva representa el argumento paciente u objeto y puede cumplir con un rol semántico de sede, meta o paciente de la acción des- ignada por el verbo. Por definición, la palabra o construcción atributiva es co-referencial con la expresión léxica del argumento sobre el cual predica. Esta característica hace que la construcción atributiva sea usada preferencialmente para expresar cláusulas relativas11 de argumento único de intransitivas (10a), o relativas del argumento objeto de transitivas (10b):

10 En wãnsöjöt la función adjetiva no corresponde a una clase léxica definida. La mayoría de “adjetivos” se alinean con los verbos (estativos) que presentan la construcción atributiva y usualmente toman un participante cuyo rol semántico es análogo al macro-rol “undergoer” propuesto por van Valin (ibid). 11 La relación entre relativización y nominalización está extensamente docu- mentada en varias regiones del mundo. Delancey (1986) y Noonan (2005) indican el sincretismo de estas funciones en familias lingüísticas del tronco Indo- Altaico. 28 JESÚS MARIO GIRÓN

O V (10) a. jduk-tai [[i-dik-di]i wam-tee]C.REL. jai-tj-wok-at animal-masa ATR-haber-PAS olla-INES 3SG-COMP-comida-CPLT La carne que estaba en la olla se la comieron

V O b. a-kedk a-dek-dik [i-a-nak-pn]C.COMPL. de O 1SG-poder 1SG-leer-EV ATR-1SG-escribir-RES Yo puedo leer lo que yo he escrito

En estas cláusulas relativas, nominalizadas por atributivo, puede haber facultativamente un argumento agente, indicado dentro de la palabra verbal por un marcador de persona, como en el ejemplo anterior (10b), o por la marca de complemento -t-, o por medio de la expresión léxica marcada con -at (ERG), o por ambos medios (11). Esta estructura se representa en la figura 3, análoga a la figura 1, pero donde la frase nominal del argumento 2 u objeto se suprime para ser sustituida por el atributivo:

(11) tiwai-ati kaj-duk yem-otj [[i-tk-di’k-ot]j-kom niño-ERG 3PL-ver trampa-PL ATR-COMP-dejar-PL-MIR a-towai-atk]C.REL. kaj-ti-jai-mapai oyem 1SG-primo-ERG 3PL-COMP-3SG-tocar CONJ El niño vio las trampas [que dejó mi primo] y las tocó CONSTRUCCIONES NOMINALIZADAS EN WÃNSÖJÖT 29

(12) Figura 3. Representación de la palabra verbal atributiva

FN(P) FN (FN(A))

FN(P) FV FN ERG

(FN(A)) V ATR MP/COMP

ja’i-da [i- -ja-/-t-j -min’k]i juan -atj canoa-ASR ATR-MP/COMP-fabricar Juan-ERG Juan hace la canoa (lit: canoa es lo que hace Juan)

Como en la nominalización de agentivo vista en §3.1.1, la palabra formada por el ATR flexiona en número (SG y PL, ver (11) y (9d) como los nombres, y requiere la especificación del argumento único o el argumento paciente sobre el cual predica. Esto se hace mediante una frase nominal aparte, la cual generalmente precede a la palabra atributiva, aunque dicho orden U/PV que focaliza el objeto, no es obligatorio en este caso. A diferencia de la nominalización de agente, esta construcción es altamente productiva. Con esta construcción se expresa el predicado de cláusulas intransitivas (13a,b), o predicados dirigidos al paciente de transitivas (13c,d), o información secundaria en cláusulas relativas, frases adverbiales y otros constituyentes adjuntos, como se muestra en (14) (en negrilla constituyentes de interés en esta demostración).

(13) a. jede-da bon i-yuk dem DEM-ASR hombre ATR-venir ayer Ese hombre es el que vino ayer b. i-won-pak-da o’o ATR-emborrachar-DUR-ASR PRNE Él mantiene borracho(lit: el que se emborracha asiduamente es él) 30 JESÚS MARIO GIRÓN

c. madio-ati ij-kak pun-yon-ot j, bi-yewai oyem Mario-ERG ATR-cocinar palo-hoja-PL 1PL-comer CONJ Mario lo que cocino fue hojas de palo, y nosotros comimos (MMr2) d. i-tm-da-wi dui i-’a-duk ATR-uno(Num)-ASR-REST Luis ATR-1SG-ver Al único que ví fue a Luis (lit: solamente fue Luis el que yo ví)

(14) a. a-jawai-wa ja-yewai i-bak-jet, 1SG-hermana-hijo 3SG-comer ATR-mucho-COM ja-bu-patjei-sik-da-tep ikom 3SG-hambre-SPRL-hambre-ASR-ENF SIML/CONCS Mi sobrino comió mucho, pero todavía tiene hambre b. madia ja-nok-suk bogota, osm bn María 3SG-ir-INM Bogotá por eso antes i-t-bep padatá ja-jeu ATR-COMP-trabajar dinero 3SG-buscar María quiere ir a Bogotá, por eso antes trabajó para buscar plata

Los marcadores de persona a la izquierda del atributivo están menos in- tegrados a la palabra-predicado. Un marcador en dicha posición tiene la misma función que la expresión pronominal focalizada del argumento único u objeto de la palabra atributiva, como se muestra en (15a). Sin embargo, una marca de argumento agente de un verbo transitivo puede aparecer en tal posición, cuando se tiene la expresión del objeto del verbo focalizado en primera posición (15b):

(15) a. ai -je-maj-u-tei-deno nan 1SG-esperar-2SG-IMP-esperar-rato aquí ai-[i-wok-to-nok ]k oikom 1SG-ATR-tomar-FREC-ir SIML Espéreme aquí mientras voy a tomar (lit: Espéreme aquí, yo el que va a tomar (un número indeterminado de) veces, mientras) b. kok-te ma-[i-jeu ] ají-INT 2SG-ATR-buscar/querer ¿Ají es lo que tu buscas? CONSTRUCCIONES NOMINALIZADAS EN WÃNSÖJÖT 31

El carácter frasal y/o clausal de la construcción atributiva está frecuentemente documentado en el corpus. En (16) se aprecian cláusulas relativas que contienen información episódica, espacial, de tiempo y modalidad relacionada con el argumento de la cláusula matriz representado en la relativa:

(16) a. “-at”, kai-mok-ma [nat i-jaj-nam-nok-pn-ot]i C.REL. si-CPLT 3PL-decir-RPR DEM ATR-3SG-botar-IPEL-RES-PL “Sí” ellos dijeron, éstos los que él había dejado tirados (THn.7) b. osm-ma yi-otj kai-’i’-bk jank , Por eso-RPR pez-PL 3PL-matar-dar D3SG [[i-tk-ja-k’nok-ot]j-a ja-k’in-dik-a]C.REL. ATR-COMP-3SG-llevar-PL-AL 3SG-mamá-haber-AL Dizque por eso ellos mataron pescado para él, para que él los (pescados) llevara donde su mamá

El morfema i- (ATR) puede ser usado hasta dos veces en la misma palabra. Esta doble aparición se presenta como una relativa dentro de otra relativa y con ella se obtiene la nominalización atributiva del objeto de un verbo transitivo:

(17) ja-mi-ma-u-bk [i-[i-jeu]] 3SG-valor-2SG-IMP-dar (comprar) ATR-ATR-necesitar Compre lo que se necesita

La relativa puede ir topicalizada en inicio de oración, lo que no contra- dice la aseveración sobre la restricción de oraciones con verbo inicial en las cláusulas transitivas con predicado nominalizado, ya que la relativa no es el predicado principal de la oración. En el ejemplo siguiente la cláusula matriz es un predicado finito en posición final de oración; le an- tecede el objeto en la posición normal, mientras que la relativa en inicio de oración es co-referencial con el argumento poseedor del objeto de la matriz wok-ti ‘comidita’: 32 JESÚS MARIO GIRÓN

OI O (18) [i-yu-sog-ot-ya] i ji kai-wok-ti ATR-venir-estar vertical-PL-FUT TOP 3PL-comida-DIM (S)-V bi-si-pi-bk-ya (CDr1) 1PL-PROS-también-dar-FUT Nosotros también vamos a dar su comidita a los que van a venir (lit: Los que van a venir y estar parados, su comidita nosotros también daremos)

El predicado atributivo puede quedar incluido dentro de otras nominalizaciones producidas por el agentivo, por las marcas de caso -a (AL) y -u (ADH), -pn (RES) y por el pasado nominal -jin (PSNM); lo que indica su posición interior en el proceso de formación de palabra.

(19) a. b-at [i-ma -bep]-u mam-at Quién-OBL ATR-2SG-trabajar-ADH PR2SG-ERG ¿Para quién estas trabajando? (lit: en lo de quién es lo que tu trabajas) b. oyat ja-ji-yai-pai-no ja-’u entonces 3SG-ANF-hacer oscilar-golpear-IPEL 3SG-esposa [i-tai-’ou-teno-wn]-jin ATR-PAT-dormir-DIMV-estar colgado-PSNM Entonces él la hizo sacudir golpeando (la hamaca donde) su esposa patéticamente estaba dormidita colgando.

Las dos nominalizaciones frasales vistas hasta aquí, la agentiva y la atributiva, tienen la particularidad de ser predicados y a la vez nombres de- verbales que funcionan como expresión de argumentos. Las construcciones de agentivo y atributivo son co-referenciales con un constituyente nominal externo a la palabra que ellas forman. Esto permite que estas nominalizaciones puedan ser consideradas como “construcciones de categorías mixtas”.12 Al contrastar estas nominalizaciones notamos que la

12 Las construcciones de categorías mixtas son descritas por Bresnan (1997: 1 pdf) como “construcciones en las cuales una sola palabra encabeza una frase, la cual es un híbrido sintáctico de dos tipos de categorías diferentes”. CONSTRUCCIONES NOMINALIZADAS EN WÃNSÖJÖT 33 palabra atributiva conserva la sintaxis de la frase verbal finita, mientras que en la nominalización agentiva la frase verbal nominalizada tiene por argumento al objeto o paciente, además de presentar un rígido orden 13 S(A)V con la expresión léxica del argumento agente. Incluimos entre las nominalizaciones de participante las siguientes dos derivaciones que crean palabras cualificativas que pueden funcionar como predicado atributivo de un nombre o como sustituto de él. Una de las construcciones se hace sobre la negación verbal, mientras que la otra es un proceso derivacional léxico.

3.1.3 Nominalización de la negación verbal mediante el sufijo -i El morfema verbo-final -i es una nominalización de evento y/o participante de la negación verbal (NZNEGV). La negación verbal finita tiene la estructura MP-san-V.14 El carácter nominalizado de la sufijación de -i puede no ser muy evidente desde que aplica a palabras verbales finitas como en (20). Pero si se sustituye el marcador de persona en inicio de la frase verbal por los morfemas i- (ATR) o ebu- (OND) (21a,b), o se suprime el marcador de persona (21c), se crean cualificativos que pueden flexionar en número (PL) (21d). La aparición de esta construcción es escasa en el corpus disponible, y aún el tipo de formas obtenidas por cues- tionario son pocas:

13 Aplica a nuestro análisis de estas nominalizaciones en wãnsöjöt una generali- zación de Haspelmath (1995: 56) citada por Bresnan & Mugane (2006: 23 pdf) a propósito de las nominalizaciones en Gikuyu y su relación con las construcciones de categorías mixtas, según la cual: “(a) In words derived by inflectional word-class-changing morphology, the internal syntax of the base tends to be preserved. (b) In words derived by derivational word-class-changing morphology, the internal syntax of the base tends to be altered and assimilated to the internal syntax of primitive members of the derived word-class.” 14 En la negación los verbos terminados en explosiva dorsal cambian ésta por la explosiva [coronal, +anterior] /t/; pero la nominalización de la negación verbal se sufija al radical verbal provocando las mismas modificaciones morfofono- lógicas en las codas dorsales que hace la derivación de agente (AGT): [k], [h], [] y [Ø]. 34 JESÚS MARIO GIRÓN

(20) a. wau-b-da bi-’ep-’ -’, osm bn lluvia-inconveniente -ASR 1PL-mal-CAUS-AGT por eso antes bi-san-ked(k)-i (CG&GMd1) 1PL-NEGV-poder-NZNEGV La lluvia fue que nos daño, por eso no pudimos terminar (lit: la lluvia…, por eso antes no fuimos terminadores) b. osm-da bi-tim-san-min-i, bi-ti-ot por eso-ASR 1PL-conuco-NEGV-crecer-NZNEGV 1PL-hijo-PL pi, también Por eso nuestro conuco no crece, nuestros hijos tampoco

(21) a. ebu-san-du(k)j-i OND-NEGV-mirar-NZNEGV ciego (el que no ve) b. i-san-yat-i ATR-NEGV-hablar-NZNEGV mudo, que no habla c. ye-san-wai-i comer-NEGV-comer-NZNEGV ayuno d. ot-da i-san-to(k)j-i-t PRNE.PL-ASR ATR-NEGV-nadar-NZNEGV-PL Ellos no son los que nadan

3.1.4 Nominalización de habituativo mediante el formante -noba El morfema bisílabo -noba- sufijado al verbo es usado en un puñado de entradas para designar características habituales (cualificativos) de una persona, al parecer con connotación peyorativa. La palabra resultante funciona como un cualificativo que puede adicionar los morfemas para masculino -di y femenino -da (22c) y flexiona en plural como los nombres, como se observa en (22d). Las marcas de masculino (M) y femenino (F) sólo se presentan en un puñado de entradas de palabras cualificativas referidas a humanos: CONSTRUCCIONES NOMINALIZADAS EN WÃNSÖJÖT 35

(22) a. yat hablar yat-noba-di locuaz, chismoso hablar-HABP-M b. yowan’ep decir obscenidades yowan’ep-noba-di hombre zafio decir obscenidades-HABP-M c. ptok pelear ptok-noba-da mujer buscapleitos pelear-HABP-F d. tesau mentir tesau-noba-di-ot mentirosos mentira-HABP-M-PL

3.2. Nominalizaciones de lugar e instrumento Reseñamos aquí las construcciones donde se crea un nombre derivado con base en una raíz verbal y un formante, la mayoría de éstos, sufijados y cuyo dominio es la palabra. Dependiendo de la estructura argumental y semántica del radical verbal las derivaciones pueden formarse sobre la raíz, o sobre un módulo {MP-V}. Esto lleva a que, en muchos casos, haya referencia a un argumento de aparición obligatoria, con diversas roles semánticos en la nominalización. Dicho argumento podrá ser ‘poseedor’, ‘experimentante’, ‘paciente’, ‘instrumento’, ‘sede’, ‘agente’, etc. En algunos casos hay referencia cruzada de dicho argumento con constituyentes externos a la nominalización, correferencia establecida por índices presentes en los otros constituyentes.

3.2.1 Nominalización de lugar mediante el formante -dipn (NZLOC) La forma -dipn puede tener dos funciones. En un caso se trata de una única forma que sufija a verbos, formando una nominalización locativa o nombre del lugar donde se desarrolla la actividad descrita por la base verbal. En este contexto, el módulo V-dipn puede prefijar un marcador de persona definida en función de posesivo, o la marca ebu- (OND) en función de complemento (23d). En tal sentido la estructura de esta nomi- nalización sigue el modelo de la descrita para el nominalizador -yu, es decir, la estructura es: {Det(FN){V-NZLOC}FN}. Usualmente el módulo {V- dipn} es marcado por un caso locativo (-a AL, -at ABL, -’u ADH) (24). 36 JESÚS MARIO GIRÓN

Esta derivación es productiva, habiendo algunas formas completamente lexicalizadas, pero la construcción puede expresar designaciones nuevas, a necesidad del hablante.

(23) a. bat cruzar mo-bat-dipn puente, sitio de cruce PNE-cruzar-NZLOC b. jui aprender mo-jui-dipn escuela PNE-aprender-NZLOC c. kak cocinar mo-ka-dipn cocina PNE-cocinar-NZLOC d. pa-kt sufrir ebu-[pa-i-kt]-dipn infierno OND-sufrir-NZLOC e. sk lavar mo-sk-dipn lavadero PNE-lavar-NZLOC

(24) a. a-’ep-da dumat, a-nok-at a-[’ou-dipn]N]FN-a 1SG-mal-ASR ahora 1s-ir-CPLT 1SG-dormir-NZLOC-AL ya-st 2PL-ANT Ahora estoy enferma, me voy a mi dormitorio (distanciándome) de ustedes b. ja-kan ja-’ok-ma, bn ja-yk-ma 3SG-hamaca 3SG-sacar-RPR antes 3SG-ir-RPR

ja-[di-dipn]N]FN-at 3SG-estar-NZLOC-ABL Dizque su hamaca sacó, ya salió de su estancia (su sitio)

c. yeu-ot ja-du-dik-ma jm ja-[map-dipn]N]FN-’u nutria-PL 3SG-ver-EV-RPR allí 3SG-pescar-NZLOC-ADH Dizque (siempre) encontraba nutrias allí en su pescadero (MMn2)

La segunda función de la forma -dipn siempre está asociada a alguna negación adverbial (yau “nada”, bidut “caritivo”). Esta construcción expresa una acción no realizada, lo que puede interpretarse como un perfectivo negativo o incompletivo (ICMPL). En esta construcción no hay referencia al evento como una localización; luego no se trata aquí de una nominalización de argumento, ni de una derivación que tenga por CONSTRUCCIONES NOMINALIZADAS EN WÃNSÖJÖT 37 dominio la palabra, sino de una nominalización de la acción de extensión clausal (25).

(25) yau ja-bi-du-dipn, bn ja-t-kom-’nok pitat-a nada 3SG-1PL-ver-ICMPL antes 3SG-RCS-MIR-llevar hospital-AL No la/lo vimos, ya se lo/la habían llevado al hospital.

En ambas funciones podríamos suponer que la forma -dipn es una forma compuesta por la secuencia -di (PAS), o -di(k) ‘haber’, más -pn (RES). En la segunda función no se encuentra el marcador mo- (PNE). En § 3.3.1 se tratará sobre la nominalización de resultativo -pn y se re- tomará la discusión sobre la forma dipn asociada a la negación. La siguiente nominalización semeja una forma compuesta por el carácter léxico del morfema nominalizador. Se trata de una nominali- zación de instrumento poco productiva donde el formante nominalizador afija directamente al radical verbal {V-formante}FN.

3.2.2 Nominalización por composición nominal por el formante -ot “contenedor” Sólo hay unos cuantos términos formados por la adición de un verbo o un módulo verbal y el morfema léxico -ot que significa ‘casa’, ‘estuche’ o ‘contenedor’ y que funciona como un lexema clasificador (término de clase). Las entradas que se conocen son formas lexicalizadas:

(26) a. [wokV-otTC]N b. [mo-tek]FV-otTC]N beber-contenedor15 PNE-sentar-contenedor canoa para hacer la chicha asiento

3.2.3 Nominalización de instrumento con el formante jet En wãnsöjöt hay dos formas jet con diferente distribución pero con significado análogo. En posición final y átona expresa el caso comitativo con funciones de sociativo e instrumental; antepuesto al módulo verbal forma la estructura {jet-{MP-V}FV}FN y crea la designación ‘instrumento

15 No se trata aquí de la marca -ot de plural, con la cual es homófona, porque para que los verbos afijen el PL deben estar previamente nominalizados por el ATR i-, u otro nominalizador; ej: i-wok-ot ‘los que beben’. 38 JESÚS MARIO GIRÓN para X’ o ‘cosa para X’, donde X refiere a la acción expresada en el verbo, pero sólo se encuentra con algunos radicales verbales. En esta posición presenta un tono alto (prosódicamente autónomo) que no condi- ciona la melodía del módulo verbal que le sigue. Esta construcción no es productiva y las ocurrencias encontradas parecen completamente lexicalizadas, incluyendo el marcador de persona no definida (PNE) que hay entre el formante y el verbo. Esta nominalización y el atributivo (ver §3.1.2) son los únicos formantes nominalizadores en posición prefijada al radical verbal, haciendo una estructura del tipo {nominalizador-FV}FN; ambos refieren al objeto directo, en este caso como instrumento, en el otro como paciente o experimentante:

(27) a. kak cocinar jet-mo-kak fogón INS-PNE-cocinar b. ’ok chuzar jet-mo-’ok cosa para chuzar INS-PNE-chuzar c. butuk’i jugar jet-mo-butuk’i juguete INS-PNE-jugar

3.2.4 Nominalización -yu (NMZ) de nombre de acción e instrumento El morfema -yu sufija exclusivamente a radicales verbales (verbos activos y estativos) y produce nombres de entidades concretas (nombre de instrumentos), o abstractas (nombre de acontecimientos y estados). De- pendiendo del tipo y rasgos semánticos del verbo la derivación requiere la expresión de morfemas de persona prefijados al verbo derivado MP- {V-yu}. Proponemos que la derivación tiene una estructura del tipo

{MP{V-yu}FN}FN, y no {MP-V}FV}yu}FN.; lo que se representa en la figura 4. Nótese en (29a,b) que las nominalizaciones que designan objetos materiales alienables no requieren prefijación de marcas de argumento, y son al parecer las formas más lexicalizadas. El marcador de persona cumple con el rol de poseedor o sede del estado de cosas expresado en la nominalización. Se podría dudar que la estructura propuesta es V-yu por la posibilidad de aparición del atributivo prefijando esta secuencia (29c), ya que la distribución propia del atributivo es no prefijar nombres. Pero tal restricción se obvia si suponemos que primero se genera la estructura ATR-V: i-nak ‘el que escribe’, sobre la cual se sufija el nominalizador: {{i-nak}-yu}. Esto supone que la derivación atributiva es más interna CONSTRUCCIONES NOMINALIZADAS EN WÃNSÖJÖT 39 que la nominalización con -yu. El morfema -yu tiene especificación tonal H, pero sólo se evidencia frente a raíces con tono [L] (29b,d):16

Fig.4. Representación de la nominalización con -yu

(28) a. FN b. *FN

Det(FN) FN FV NMZ

MP V NMZ FN V

a- ki(k) -yu a- ki(k) -yu 1SG llorar mi llanto

(29) a. -sin cernir sin-yu cernidor b. -kk barrer k-yu’ escoba c. -nak escribir i-na-yu escritura ATR-escribir-NMZ d. pnpat rato ja-pnpat-yu edad (de él/la) 3SG-duración-NMZ

Esta nominalización con algunos verbos intransitivos designa los nombres de la acción como estado o hecho, como ocurre claramente en los ejemplos en (30):

(30) a. ’ep estar enfermo mo-’ep-yu enfermedad b. tek sentarse mo-te(k)-yu sentada de alguien c. ma-yoi tu ríes ma-yo-yu’ tu risa

16 De las cuatro melodías tonales de raíces (alto, bajo, ascendente y descendente) el tono alto del morfema no se manifiesta ante las raíces con tono alto o descendente; frente a raíces con tono ascendente, es el componente alto del ascendente el que se manifiesta por escisión del tono de la raíz monosilábica. 40 JESÚS MARIO GIRÓN

d. ja-y(k)-yu yamat 3SG-salir-NMZ sol la salida del sol e. ja-d(k)-yu pun 3SG-caer-NMZ árbol la caída del árbol f. om-da ja-pip jede bon ja-’om-yu-at ahí-ASR 3SG-temblar DEM hombre 3SG-temer-NMZ-ABL Ahí tembló ese hombre de (debido a su) temor

La base de derivación puede ser una base compleja, ya sea formada por dos raíces, o verbos separables, en cuyo caso sólo aparece el marcador de persona del poseedor:

(31) a. ji-jei estar alegre mo-ji-jei-yu alegría, fiesta PNE-corazón-bueno-NMZ b. jui-suk sumar, reunir ebu-jui-’i-su-yu reunión,integración OND-reunir-ATR-reunir-NMZ c. ji-sok apoyar vertical ja-ji-so-yu anclaje 3SG-CONX-vertical-NMZ

3.3. Nominalizaciones de acción: estados, eventos, hechos Presentamos a continuación las nominalizaciones de acción que refieren al nombre de la acción misma como estado, evento, hecho o proposición. En estas construcciones no se crean palabras nuevas, sino que se produce un crecimiento de la secuencia morfémica con base en el módulo verbal (MP-V) y la adición de marcas que no son libres. Por ello se crean secuencias frasales, pedazos de cláusula o cláusulas que funcionan como un único constituyente en un contexto sintáctico mayor. Es discutible si el proceso de nominalización que crean estos formantes es su principal función; pero las secuencias que ellos cierran pueden funcionar como expresión de argumentos y adjuntos, y es por esto que hablamos aquí de nominalización en un sentido amplio, más funcional que morfológico. Todas las construcciones descritas a continuación involucran algún participante interno, muchas veces expresado por los marcadores de persona o una marca de indefinición de argumento; también pueden incluir marcas aspectuales, modales y determinantes léxicos. Mostraremos aquí nominalizaciones de frases verbales en función de acción nominalizada, CONSTRUCCIONES NOMINALIZADAS EN WÃNSÖJÖT 41 mediante la partícula aspectual -pn, de aspecto nominal (el formante -jin de pasado nominal) y la palabra no finita MP-V-dik que llamamos eventivo (EV). En el caso de la construcción eventiva (-dik), el constituyente puede funcionar como predicado principal. Posteriormente se muestran las determinaciones subordinadas de los casos locativos -a de AL y -u de ADH (§3.4), y de partículas adverbiales y locativas (jee de finalidad o propósito, -teé de inclusión) que hacen que frases verbales complejas y cláusulas funcionen como constituyentes nominalizados (§3.5). Presentamos en esta sección tres nominalizaciones que tienen dis- tinto carácter pero que comparten el hecho de constituir frases verbales cuyos formantes nominalizadores parecen gramaticalizaciones de radicales léxicos. Se trata de los formantes -pn y -jin que funcionan como morfemas derivativos en la formación léxica a partir de radicales nominales o verbales, y del formante -dik que es la misma forma del verbo cópula “estar, haber”. Los formantes -pn y -jin pueden generar adjetivos derivados sobre morfemas verbales, y ambos formantes también funcionan como determinaciones aspectuales sobre frases verbales o incluso sobre cláusulas. Las secuencias integradas por -pn y -jin generalmente funcionan como acciones nominalizadas en virtud de la determinación aspecto-temporal que hacen sobre las acciones, presentándolas como eventos cerrados. Por ello, la función de estas marcas también puede interpretarse simplemente como determinaciones gramaticales de aspecto sobre palabras verbales finitas. Las dos construcciones se diferencian entre ellas porque la forma -jin tiene un uso a nivel de la cláusula en oraciones complejas que expresan sentido de adversativo (ver §3.3.2).

3.3.1 Nominalización de resultado de la acción mediante la marca pn (RES) El morfema -pn de “resultativo” (RES) sufijado a raíces verbales expresa el resultado de la actividad y de ahí su función aspectual. La acción nominalizada por pn se usa como nombre o como cualificativo, y flexiona en número (SG= Ø, PL= -ot). La secuencia con cabeza verbal creada por pn es análoga al contexto N-pn, como tim ‘conuco’ → tim-pn ‘rastrojo, conuco en barbecho’, o sim ‘pie’ → sim-pn ‘pisada, huella’. En algunas construcciones usadas como nombre hay una completa lexi- calización de pn con el radical verbal (32a), habiéndose perdido en algunos términos, como en este ejemplo, la marca inicial de dependencia de la raíz verbal: 42 JESÚS MARIO GIRÓN

(32) a. ’um-pn pilar-RES harina, (masa) pilada b. i-st-pn ATR-moquear/ahumear-RES lo que es(tá) moqueado, ahumado 17 c. i-mo-map-pn ATR-PNE-pescar-RES lo que uno ha pescado d. ja-t-yau-dei-pn mo-yan ja-ubat-ot-at 3SG-COMP-destruir-dejar-RES PNE-pueblo 3SG-enemigo-PL-ERG la destrucción del pueblo por los enemigos

En (33) la marca pn en solitario expresa un evento, con o sin duración, en un pasado cerrado (perfectivo). Dado que pn señala una acción con- cluida y fijada, es incompatible con la marca -ju de imperfectivo (IPFC). Pero la localización temporal de la palabra verbal que sufija pn está determinada por el contexto, o por las marcas de tiempo -ni, y -di (34a,b). Cuando en la palabra determinada por pn está también presente el morfema atributivo, el efecto nominalizador de pn es posterior y de otro carácter que el producido por i- (ATR) (34), el cual produce una nomi- nalización desde la perspectiva de un argumento (el objeto), mientras que pn nominaliza la acción al caracterizarla como un estado.

(33) a. busik-pn-da tiwai tener hambre-RES-ASR niño Hambreado estaba el niño b. om-ma i-minyuk ja-’u butuk’ i-ot ahí-RPR ATR-llegar 3SG-esposa jugar-PL du-no-m-pn ver-ir-AGT-RES Ahí dizque llegó la mujer de él, quien había ido a ver a los ju- gadores

17 Ahumeado, ahumado o moqueado, es una forma muy corriente de conservar alimentos, especialmente carne, en la región amazónica. CONSTRUCCIONES NOMINALIZADAS EN WÃNSÖJÖT 43

(34) a. ’am-da i-ja-ji-pn-ni PR1SG-ASR ATR-3SG-silbar-RES-PSVG A mí es a quien él silbó b. i-kut-bok-pn-di-da mam ATR-huir-caminar-RES-PAS-ASR PR2SG El que terminó por huir fuiste tú

La palabra verbal que afija pn puede funcionar como cláusula relativa, aunque no es este morfema el que causa la relativización.En el ejemplo siguiente la relativización se produce por la yuxtaposición recursiva de una cláusula relativa de otra no principal, secuencia que a su vez es subordinada de la matriz en posición inicial:

(35) osm-da jaj-bii-jeu dumat, [jaj-bii-’ i’k-dik por eso-ASR 3SG-1PL-buscar ahora 3SG-1PL-matar-EV [jj jak-’i’-mj-pn]C.REL] C.SUBORDINADA

éste 3SG-matar-AGTi-RES Por eso loj buscamosi ahora, (para) matarloj (a) esej que lok mató

A su vez, una palabra verbal cerrada por pn puede recibir otros determinantes del nombre, como -jin (PSNM) (36b). Una nominalización previa puede sufijar el resultativo en función de aspecto perfecto, como se muestra en (34) y (36b) con el atributivo, y en (35) y (36a) con el agentivo:

(36) a. oi-da jai-mok ja-pin-dik juani, naai así-ASR 3SG-decir 3SG-contar-EV Juan DEM [den-si-’m-i ]-pn mujer-?-pedir-AGT-RES Así decía contando Juan, éste mismo que pidió a la muchacha b. on-da [[i-’i]-pn]-jin bi-duk-a allí-ASR ATR-morir-RES-PSNM 1PL-ver-AL Allí vimos al que ya se había muerto

Vimos en §3.2.1 que hay dos formas -dipn, una de ellas es la nominali- zación locativa vista en dicha sección, y otra es la forma asociada a una negación adverbial, de la cual mostramos algunos ejemplos en (37) (en negrilla los morfemas en discusión). 44 JESÚS MARIO GIRÓN

(37) a. yau-ma ja-ye-si-wai-dipn i-jei-jet nada-RPR 3SG-comer-RPT-comer-ICMPL ATR-bien-COM Dizque nada él volvió a comer bien b. bidut-ma i-i-du-dipn i-bu-yu-at caritivo-RPR ATR-ATR-ver-ICMPL ATR-oscuro-NMZ-ABL Nada (era lo que) se veía en la oscuridad

En esta construcción la frase verbal con -dipn funciona como un complemento del morfema de la negación, especificando tanto el participante como el evento que se caracteriza como no real.

3.3.2 Determinación de frases verbales por el morfema -jin de pasado nominal La función básica del morfema -jin es indicar la función de caducidad en nominales (38). Sin ser propiamente una marca de nominalización, este morfema también se encuentra sufijado a verbos,18 o a frases verbales indicando que una actividad, un acontecimiento o un estado ya no son vigentes (39), o no funcionan en la situación presente. Cuando la secuencia determinada por -jin funciona como complemento, -jin aplica sobre la frase verbal y sus dependientes y adjuntos.

(38) a. bi-dou-jin otro-fecha-PSNM fecha vencida, momento pasado o retrasado b. om-ma ja-’in-jin ja-y (THr3) ahí-RPR 3SG-mamá-PSNM 3SG-soñar Ahí él soñó con (el fantasma de) su mamá

18 Algunos nombres de enfermedades, tanto mágicas como atribuidas a causas naturales, se forman por la sufijación de jin a raíces nominales y verbales. CONSTRUCCIONES NOMINALIZADAS EN WÃNSÖJÖT 45

(39) jn ja’-jaa-ma ka-ma-no-jin ot allá piedra-sobre-RPR 3PL-bañar-ir-PSNM PRNE.PL yu-yu-ma i-t-sak-ot (THr1) salir-NMZ-RPR ATR-COMP(A)-morder-PL Allá dizque estaban sobre la piedra (donde) ellos habían ido a bañarse, a su salida (de ellos) dizque fueron mordidos.

El morfema -jin en combinación con el morfema -pn (RES) produce cualificativos que designan un estado o característica que se torna per- manente:

(40) a. i-jut-pn-jin ATR-amarillo-RES-PSNM El que es (quedó) pálido b. jamai-pn-jin hacer pereza-RES-PSNM perezoso c. ja-ot i’ep-pn-jin tee te-jam 3SG-estuche ATR-mal-RES-PSNM INES-HYP-INQU ja-ma-i-tep-b ’an 3SG-2SG-ATR-ENF-dar D1SG ¡Como así que me va entregar eso con el forro dañado! (lit: su estuche el que ya quedó malo, dentro de él, pregunto, lo entre- ga usted (con énfasis) a mí!)

El morfema -jin (PSNM) tiene por dominio la cláusula en oraciones complejas donde las cláusulas constituyentes están en una relación adversa- tiva, y es la cláusula marcada con -jin la que expresa la frustración por un evento no realizado en contraste con otro que sirve de referencia. El formante jin puede estar en ambas cláusulas si ambas tienen carácter de evento no realizado o que causa frustración. La cláusula marcada con pasado nominal puede recibir determinaciones adicionales de tiempo y modalidad (41b), pero tales determinaciones son exteriores a la frase verbal cerrada por -jin. Esto muestra que la categoría tiempo gramatical es exterior a esta determinación de aspecto nominal: 46 JESÚS MARIO GIRÓN

(41) a. “i-mon” a-san-mok-sut “mon” ATR-hondo 1SG-NEGV-decir-DES chontaduro a-mok-suk-jin 1SG-decir-DES-PSNM/AVRS Yo no quería decir ‘hondo’ sino ‘chontaduro’ b. a-san-yuk-sut]-jin]-ni-da nan, ’am, 1SG-NEGV-venir-INM-PSNM/AVRS-PSVG -ASR aquí PR1SG a-map-a-da a-nok-suk-jin 1SG-pescar-AL-ASR 1SG-ir-INM-PSNM/AVRS Yo no quería venir aquí, sino que quería ir a pescar

3.3.3 Nominalización del contenido proposicional mediante -dik (EV “eventivo”) La construcción MP-V-dik (PEV) es altamente productiva y cumple con diversas funciones que tienen en común presentar la acción con su o sus argumentos, es decir los elementos mínimos de una cláusula contenidos en el módulo verbal, como un solo constituyente sintáctico. Dado que el verbo -dik es un verbo existencial, podría suponerse que su función sería de complemento verbal, produciendo algo así como una forma existencial o factual, con alcance de cláusula, la cual podría glosarse como “hay el evento de A(B) haciendo X”. En cualquier caso, se trata aquí de un proceso de gramaticalización de dicho radical. La construcción no admite la negación verbal, sino que debe ser negada como un todo por la nega- ción ’i, la cual es usada para nominales. En principio podemos decir que se trata de una forma no finita, análoga en sus usos al infinitivo, 19 ya que con ella puede designarse el nombre de la acción incluyendo sus argumentos. Pero a diferencia de los infinitivos de las lenguas indoeuropeas, esta construcción no puede ser usada para designar el nombre de la acción fuera de contexto y sin argumentos. La construcción con -dik como ‘eventivo’ puede ser usada como palabra verbal determinada en una construcción auxiliar, o como complemento adverbial,20 o como complemento de objeto (42). Puede ser el

19 Seguimos aquí la caracterización de Ylikosky (2003) sobre no finitos. 20 Los radicales que hasta el momento podemos caracterizar como formantes que hacen determinaciones adverbiales en wãnsöjöt son los siguientes: -dik “haber/EV”, -bk “dar/contributivo(¿)”, -nok “ir/incoativo/impeletivo”, -dei “dejar en determinada posición o estado”, -’k “hacer/CAUS”, -to(k) CONSTRUCCIONES NOMINALIZADAS EN WÃNSÖJÖT 47 predicado principal de una cláusula, pero siempre que dicha cláusula esté en relación de coordinación, sucesión o implicación con otras cláusulas. Esta construcción presenta la acción como un acontecimiento com- pleto, no necesariamente concluido ni pasado, siendo su significación temporal dependiente del contexto, tanto discursivo como lingüístico. No obstante, cuando es usada sin marcas adicionales generalmente indica una acción pasada y cerrada. No se puede decir que esta construcción exprese la acción como algo estático, y en ello es similar a la predicación finita. También se parece a la construcción finita (PF) en que no admite el plural, y que puede afijar marcas de aspecto. Esta construcción no finita guarda concordancia con otras palabras verbales y cláusulas mediante el marcaje de sus argumentos y generalmente está en posición final de oración, o de cláusula cuando está en coordinación con otras palabras verbales o cláusulas en oraciones complejas.

(42) a. ji-da jaj-jeu-ei [a-kut-dik]j C.COMPL de O. PR3SG-ASR 3SG-querer-AGT 1SG-huir-EV Él/la quiere que yo huya (él/la (es quien) ésto quiere ‛mi huida sea’) b. sai-’au-oti-ma-di jaj-not jui-ui-t [noche-hormiga sp]-PL-RPR-PAS 3SG-ruido oir-AGT-PL [naa ja-’ak-dik]j C.COMPL de O. (MMn1) DEM3SG-gritar-EV Las hormigas nocturnas dizque fueron los que escucharon el ruido de éste que gritaba c. Dio-at ya-pin-di bidat [bjadik-u Dios-OBL hablar-relatar-PAS D1PL cómo-ADH ebu-ja-min’k-dik bn] C. COMPL. de O. OND-3SG-hacer-EV antes Dios cuenta a nosotros cómo todo fue creado antes

Los elementos de la cláusula cuyo predicado es una construcción con -dik se comportan de igual manera a los elementos de las cláusulas finitas. El ejemplo en (43) consta de dos oraciones complejas con cláusulas subordinadas. La primera de ellas se compone inicialmente de una palabra-cláusula finita con dos complementos: el primero es una cláusula

“vez/frecuentativo”. 48 JESÚS MARIO GIRÓN que funciona como objeto, cuyo predicado es una palabra verbal eventiva con un argumento agente interno en primera posición marcado con el ergativo; el MP de la palabra verbal eventiva corresponde al objeto, no al agente, como se mostró en (1). La segunda subordinada de la primera oración es una cláusula adverbial de finalidad. En la segunda oración compleja del ejemplo la palabra verbal eventiva forma parte de una con- strucción auxiliar (PF-PEV).

(43) bij -ja i-san-pi-pin’ [sp-at k 1PL-3SG-NEGV-también-relatar serpiente-ERG jam(O)-sak-dik] C.COMPL. de O. [ja-bi-wn-jee 3SG-morder-EV 3SG-1PL-remedio-FIN pisi-bi-wak-jet.] C.ADV. -san ja-dik-’u om-da saber-1PL-saber-COM ?-NEGV-3SG-haber-ADH ahí-ASR ja-bi-ket’ ja-bi-’nok-dik pitat-a (THr1) 3SG-1PL-poder 3SG-1PL-llevar-EV hospital-AL Tampoco él nos avisó (que) la serpiente la mordió, para nosotros remediarla con lo que sabemos. Si no sucede así, ahí la podemos llevar al hospital.

Mostramos a continuación oraciones donde una cláusula con palabra verbal eventiva -dik (PEV) es el predicado principal.21 No obstante que las cláusulas de las palabras eventivas son independientes, tienen algún grado de coordinación con las adyacentes y dependen de ellas para la referencia o para el sentido en el discurso:

(44) a. o-ma i-pak, ja-pak-ma . o-ma-wi PRNE-RPR ATR-vomitar 3SG-vomitar-RPR PRNE-RPR-REST ja-’ep-dik (THr2) 3SG-mal/enfermarse-EV Ese dizque vomitó, dizque vomitó, dizque ahí mismo se en- fermó

21 Es frecuente en wãnsöjöt encontrar secuencias de radicales verbales, lo que da pie para hablar de serialización en esta lengua. Pero el caso de la secuencia donde el último radical verbal es el verbo -dik es claramente un caso de gramaticalización. CONSTRUCCIONES NOMINALIZADAS EN WÃNSÖJÖT 49

b. … ja-mok-ma kadat. “jou” ka-mok-ma 3SG-decir-RPR D3PL INTERJ 3PL-decir-RPR o-ma-wi ka-nok-dik PRNE-RPR-REST 3PL-ir-EV (Algo)…dijo a ellos. “Bueno” dijeron ellos. Inmediatamente se fueron. c. i-pi wyu-ma ja-di ja-min-dau ATR-tres día-RPR 3SG-haber 3SG-llegar bajando el río ketyem o-wi ja-’i-dik después PRNE-REST 3SG-morir-EV Tres días estuvo después que llegó (bajando el río). Ahí mismo murió

Cláusula relativa con palabra verbal eventiva:

(45) kaunok wyu wuk, o-kom-tep ka-ibi nat cuatro día pasar PRNE-MIR-ENF 3PL-tío (paterno) DEM.PL i-’i-ot, [de’-pik ja-tam-dik] C.REL. ATR-morir-PL cerdo de monte-piel 3SG-nombrar-EV Cuatro días pasaron, me dí cuenta que fue el tío de estos muertos, “cuero-cafuche” se llamaba.

Hemos visto que frecuentemente la palabra verbal no finita con -dik no es la única palabra verbal de la oración. Las oraciones en (46) son enten- didas como construcciones auxiliares modales:

(46) a. -san-ja-dik-’u om-da [ja-bi-ket’ ?-NEGV-3SG-haber-ADH ahí-ASR 3SG-1PL-poder ja-bi-’nok-dik] pitat-a (THr1) 3SG-1PL-llevar-EV hospital-AL Si no sucede así, ahí la podemos llevar al hospital. b. bi-sk-da dumat bidut [ja-san-ket 1PL-celo-ASR ahora PR1PL 3SG-NEGV-poder [kok, sayu, kak] bi-wok-dik] ají sal cocinado 1PL-comer-EV Nosotros ayunamos ahora, nosotros no podemos comer ají, sal, ajicero 50 JESÚS MARIO GIRÓN

Sin embargo en (47) las dos palabras verbales, finita y no finita (eventiva), están siendo usadas en su pleno sentido léxico, siendo tal construc- ción una forma de expresar verbos en relación de determinación léxica (en negrilla la construcción en discusión):

(47) “ma-ye-ku-wai”, ja-mok-ma ja-’u jan. 2SG-comer-EXHOR-comer 3SG-decir-RPR 3SG-esposa D3SG “’i” ja-mok-ma. ja-’om-ma ja-yewai-dik (THr2) NEG 3SG-decir-RPR 3SG-temer-RPR 3SG-comer-EV “Coma” dijo su mujer a él. “No”, dijo. Él dizque temía comer

Con las palabras verbales no finitas con -dik se expresan imperativos negativos nominalizados (48a) y adjuntos en oraciones complejas (48b):

(48) a. ’i-da ya-jon-dik NEG-ASR 2PL-subir-EV No suban b. ja-mo-’k-jee mo-wnot-ot mo-bik-mo-sak-ji 3SG-PNE-hacer-FIN PNE-rendal-PL PNE-pensar-PNE-pensar-TOP [ja-jupt-u ka-mo-jip-dik] 3SG-adelante-ADH 3PL-PNE-amarrar-EV Para uno hacer rendales22 piensa primero sobre los (anzuelos) que uno amarra

3.4 Determinación de frases verbales y cláusulas por casos y partículas adverbiales En la identificación de construcciones nominalizadas encontramos sen- cuencias de tipo frasal y clausal que funcionan como acciones nominalizadas en posición subordinada y en función de constituyentes oblicuos, o como adjuntos de cláusulas matrices o predicados principales. Los formantes responsables de estas construcciones subordinadas son de dos tipos: casos locativos que conectan frases verbales como argumentos periféricos y partículas adverbiales que hacen similar función con secu-

22 Rendales o “guarales” es la designación en castellano local de un sistema de pesca que consiste en una sarta de anzuelos que penden de un único hilo de pesca. CONSTRUCCIONES NOMINALIZADAS EN WÃNSÖJÖT 51 encias frasales o clausales. Presentamos inicialmente las construcciones con caso y luego las determinaciones adverbiales (3.5).

3.4.1 Nominalización locativa mediante el caso alativo -a El módulo {MP-V}-AL23 es una construcción subordinada de alcance frasal mediante la cual se hace referencia a la actividad como una meta, una sede o ámbito, o un estado a alcanzar. Hacia este estado de cosas o ámbito se dirige un participante expresado mediante un marcador de persona o una frase nominal prefijada a la raíz verbal. Según la estructura semántica y argumental del radical verbal, el argumento tiene un rol de poseedor, experimentante o agente de la nominalización locativa (49a, b). Es decir, la estructura de la nominalización es análoga al modelo de la figura 4b. Es decir, en lugar del agrupamiento {MP-{V-AL}FN}, ten- dríamos aquí una frase verbal {MP-V} determinada en su conjunto por la marca de alativo: {{MP-V}-AL}FN. En este caso el módulo verbal es to- mado como un nominal por la partícula casual que está estableciendo la subordinación del grupo bajo una cláusula matriz. El hecho que la subordinación se haga directamente por una partí- cula causal sobre una frase verbal o una cláusula aparentemente finita lleva a considerar la dos hipótesis: a) que los radicales que consideramos verbales en wãnsöjöt tienen la propiedad de comportarse como verbos o nombres en este contexto, y que la marca casual de alativo tiene una dis- tribución especial que le permite afijar directamente a verbos, o b) que habría que considerar un nominalizador [Ø] que aplica al módulo verbal para ser factible el marcaje con caso. Dejamos planteada esta discusión sobre categorización léxica y dependencia en wãnsöjöt. La construcción es muy productiva y funciona en la oración como un argumento oblicuo. Cuando el verbo de la construcción es el verbo existencial -dik ‘haber’, la construcción designa lo que hemos llamado en wãnsöjöt ‘dativo existencial’ (49c). En (49d) hay dos marcadores de persona, pero el primero de ellos al parecer funciona como una referencia cruzada con el argumento representado en la primera palabra. En tal caso, la nominalización ma-’ok-a sería co-referencial con el argumento representado por el índice ja- en ambos constituyentes.

23 A diferencia de las sonorización que se produce en juntura de morfemas, el alativo (al igual que el agentivo) hacen que la explosiva final de la raíz verbal /p, t, k/ se conserve sorda. 52 JESÚS MARIO GIRÓN

(49) a. bi wyu-dou itm bon-t ti-nok-ma [jai-map-a]k PRIND dia-fecha uno hombre-joven-ir-RPR 3SG-pescar-AL Un día un joven dizque fue a su pesquería (actividad de pesca) b. biduti-ni bi-biksak bi-nok-dik [bii-bep-a]k PR1PL-PSVG 1PL-pensar 1PL-ir-EV 1PL-trabajar-AL Nosotros pensábamos ir a nuestro trabajo c. bi-padataj ’i-sikom, bii-si-no-ya [jaj-dik-a]k 1PL-dinero NEG-PROG 1PL-PROS-ir-FUT 3SG-haber-AL Dado que no tenemos plata, la conseguiremos (lit: nuestro dinero no hay, iremos hacia donde él está) d. jaj-si-yau-wn jaj -[mai-’ok-a]j 3SG-PROS-lejos-POT 3SG-2SG-sacar-AL Ha de ser lejos (el sitio) donde usted saca (oro) (diálogo entre LM&OG sobre actividad de minería del oro)

En la construcción con el alativo se puede sustituir el marcador de persona inicial por la correspondiente frase nominal. Tal construcción es frecuentemente encontrada en el corpus disponible:

(50) a. ja-t-pek-’yuk [ja-j-ot]FN-dik-a 3SG-COMP-INTS-traer 3SG-pariente-PL-haber-AL paujil-te-ja (THr2) Paujil-INES-TRNSL Finalmente lo trajeron donde está su familia en el Paujil b. jaj-mai-du(k)j-i duu den-pekj, dunk 3SG-2SG-ver-INTV esa mujer-adulta allá [[mai-di-dau]-nok-a’]k (GG&PUd2) 2SG-haber-ir río abajo-IPEL24-AL ¿Usted vio a esa mujer (adulta) allá río abajo (donde) usted fue?

Al ser preguntada una información oblicua (el ‘para qué’ o la intención de la acción), el pronombre interrogativo se marca con el alativo:

24 Es discutible si las secuencias de raíces verbales como las del ejemplo constituyen una serialización verbal, o si muchas de las series verbales son secuencias de radicales en función de determinaciones adverbiales. CONSTRUCCIONES NOMINALIZADAS EN WÃNSÖJÖT 53

(51) bidik-a ma-juibek-su bi-jet Para qué 2SG-reunir-DES 1PL-COM ¿Para qué se quiere reunir usted con nosotros?

Como partícula casual, el alativo sufija a construcciones previamente nominalizadas, como se muestra en el ejemplo con el formante de lugar - dipn. Pero aquí el caso alativo no es una nominalización sino el uso corriente de su función como caso que indica localización de una entidad con referencia a un sitio; la misma función que tiene en djk-a //puerto- AL// ‘al puerto’:

(52) ja-nok-at-ni [[ja-’ou-dipn]-a] (THr5) 3SG-ir-CPLT-PSVG 3SG-dormir-sitio-AL Ya se fue ya a su dormitorio.

3.4.2 Nominalización de acción mediante el locativo -u El locativo -u (ADH)25 nominaliza la acción indicando la actividad como un ámbito, una sede. Con este locativo se han lexicalizado algunas entradas cuya cabeza puede ser una raíz nominal (53a,b), o verbal (53c,d).

(53) a. yeN-’u hierba-ADH sabana b. [wen-tat]N-u cerro-talón/base-ADH Pata de cerro (n. propio) c. bkV -u blanco-ADH sabana (lit: en lo blanco,) d. pinV-’u relatar-ADH noticia

A diferencia de la localización expresada por el alativo -a (ver §3.4.1), no hay aquí direccionalidad sino contacto o inmersión (54):

25 La función de -u como caso locativo o espacial indica localización sin movimiento y con contacto de una entidad en un ámbito. Se combina con otras marcas de caso espacial para expresar otros rasgos de la localización espacial (ver Girón 2008). La glotal del morfema se introduce epentéticamente cuando la base a la que se afija el morfema termina en vocal o nasal. Otra regla post- léxica indica que hay sonorización de oclusivas ante morfemas que inician por vocal (p/t/k -o/u → [bo/bu, o/u, go/gu]). Reglas léxicas para alativo y agentivo previenen dicha sonorización. 54 JESÚS MARIO GIRÓN

(54) a. mo-bepV-u PNE-trabajar-ADH en lo que usted trabaja b. b-at i-ma -bep-u mam-at Quién-OBL ATR-2SG-trabajar-ADH PR2SG-ERG ¿Para quién estas trabajando? (lit: ¿Para quién (es) en lo que estás trabajando? c. a-st-u bima-u-bk juan-at 1SG-moqueado 26-ADHalgo 2SG-IMP-dar Juan-OBL Déle un poco de mi moqueado a Juan. (lit: En cuanto a mi moqueado algo déle a Juan)

La construcción verbal cerrada por esta marca casual funciona como una cláusula subordinada (55), y en tanto cláusula subordinada funciona como paráfrasis de uno de los argumentos.

(55) a. yam-ji ja-pisi-ya-san-wat-di [’an PR2PL-TOP 3SG-saber-2PL-NEGV-saber.NEG-PAS casabe ya-min’k-u] C.COMPL de O. 2PL-hacer-ADH En cuanto a ustedes no lo sabían [acerca de ustedes hacer casabe]. b. [ka-pin’u om-da] C.ADV. ka-pisi-ka-wa-ye 3PL-relatar-ADH ahí-ASR 3PL-conocer-3PL-conocer-REC En su conversación ahí, ellos se conocieron entre sí c. bidut-da i-’ak-juk-ot [ya-jupt-’u yam]C.ADV. PR1PL-ASR ATR-gritar-primero-PL 2PL-delante-ADH PR2PL Nosotros (fuimos) los que gritamos adelante que vosotros

Pero el uso con alcance de cláusula más productivo de la partícula -u es el marcaje que hace este formante sobre cláusulas que integran oraciones

26 Ya mencionamos que “moqueado” o “ahumado” es una forma de conservar alimentos en Amazonas. Es de presumir que como rasgo cultural regional, en las lenguas dominantes se usa la designación de manera análoga a como se hace en algunas lenguas nativas: “ahumado” puede referir al resultado o a la entidad. CONSTRUCCIONES NOMINALIZADAS EN WÃNSÖJÖT 55 complejas condicionales (56), donde se sufija a la cláusula que hace de prótasis del condicional:

(56) a. [ja-yewai-’u]-ma-ikom ja-t-tep-pak-ma (THr2) 3SG-comer-ADH-RPR-SIML 3SG-COMP-ENF-vomitar-RPR Dizque si él comía, ahí mismo él lo vomitaba b. [a-mapyu-ot ja-dik-’u ] a-kuk-map jam’o 1SG-anzuelo-PL 3PL27-haber-ADH 1SG-FRUS-pescar ahí Si yo tuviera anzuelos pescaría ahí (lit: mis anzuelos habiéndolos, yo intentaría pescar ahí) c. [wau-dk-u bdem] san-yut jam’o lluvia-caer-ADH mañana NEGV-venir.NEG ahí Si mañana llueve yo no vengo (lit: en el evento de llover mañana yo no vengo)

3.5 Determinaciones adverbiales y espaciales con efecto nominalizante Presentamos por último dos casos de construcciones adverbiales y espaciales formadas con base en partículas adverbiales y casos locativos combinados, las cuales son ejemplo de acciones nominalizadas en fun- ción de argumentos adjuntos de cláusulas principales. Ya hemos visto que morfemas de caso como el alativo -a o el adhesivo -u son altamente productivos para crear acciones nominalizadas subordinadas que funcionan como argumentos nucleares y oblicuos. Otros casos de tipo espacial (topológicos) también determinan frases verbales como en (57), constituyendo expresiones que funcionan como adjuntos (determinaciones adverbiales) o cláusulas subordinadas:

27 El marcador de persona del verbo existencial no concuerda en número con el objeto en posición inicial; lo que sugiere que dicho marcador hace referencia a la situación existencial como objeto del verbo -dik. 56 JESÚS MARIO GIRÓN

(57) ja-yuiot-tk-tee-da ja-t-kom-tep-’nok 3SG-vestido-mojado-INES-ASR 3SG-COMP-MIR-ENF-llevar pitat-a hospital-AL En sus ropas mojadas (me di cuenta que) se la llevaron al hospital

Concluimos esta descripción de las nominalizaciones en wãnsöjöt con el análisis de las cláusulas nominalizadas por la partícula adverbial jee que indica propósito o ‘finalidad’. Frases adverbiales de finalidad con argumentos internos pueden ser el complemento oblicuo de cláusulas cuyo predicado es una nominalización agentiva o de objeto. En (59) el agente de la subordinada (j) es el argumento objeto de la matriz (construcción agentiva). Nótese que la nominalización de finalidad o propósito puede tener un objeto interno, expresado tanto por un marcador de persona en el módulo verbal (segundo marcador de persona a la izquierda del radical), como por un constituyente nominalizado final (k):

(58) ka-duk [ka-jui-jee] C.ADV. 3PL-entrar 3PL-aprender-FIN Ellos entraron a estudiar

(59) ji-da yaj-kui-u i [kak-yaj-min’k-jee [nat PRDEM-SR 2PL-ordenar-AGT 3PL-2PL-hacer-FIN DEM mo-bep-ot]k]C.COMPL. OI PNE-trabajar-PL Él les ordenó a ustedes (para) que ustedes hagan estos trabajos

Muchas ocurrencias de jee se pueden analizar como si esta partícula tuviera dos objetos: un antecedente (algo previo que está, ocurre o se hace), y un objeto terminal que informa sobre el propósito o meta. El morfema va sufijado a la cláusula que designa el objeto terminal. El objeto o in- formación antecedente generalmente se pone en una cláusula cuyo predicado es una palabra verbal no finita (-dik) (60). Sin embargo es posible tener el orden inverso (61a), pero se conserva el nexo lógico entre ambos constituyentes: CONSTRUCCIONES NOMINALIZADAS EN WÃNSÖJÖT 57

(60) a. o-wi bi-upt-dik ja-bi-duk-jee a-jawai PRNE-REST 1PL-volver-EV 3SG-1PL-ver-FIN 1SG-hermana Inmediatamente regresamos para ver a mi hermana b. o-ma-wi ka-sion-ot ka-’ok-dik ka-nok-jee PRNE-RPR-REST 3PL-catumare-PL 3PL-sacar-EV 3PL-ir-FIN ya-bat-ta lejos-cruzar-AL Dizque inmediatamente sacaron (cogieron) sus catumares para ir al otro lado c. 120.000 ja-mi-ka-b-dik ipuk ja-dapde 3SG-valor-3PL-dar-EV brujo 3SG-muñeca woi-nok-jee partir-EV-FIN 120.000 costó para que el brujo le partiera la muñeca (de su mano) a él

4. Las nominalizaciones en el discurso wãnsöjöt Las palabras verbales y cláusulas nominalizadas alternan con la forma finita de múltiples maneras, ya sea como expresión de los argumentos del verbo o como complementos verbales, generalmente portando informa- ción subsidiaria o de menor jerarquía, con la cual se especifica la infor- mación sintáctica principal codificada en las palabras verbales finitas. Las nominalizaciones son usadas como cláusulas complemento y relativas, o como constituyentes periféricos del discurso, mediante los recursos morfológicos vistos (atributivo i-, finalidad jee, evento -dik, etc.), como se ve en el siguiente texto (en negrilla las nominalizaciones):

(61) dinkwista-ji o-da [i-jui ja-jaa nat lingüista-TOP PRNE-ASR ATR-aprender 3SG-sobre DEM moyeduk-ot ka-juno(k)-jet i-yad-ot] C.REL. [ja-t-ket’k-jee palabra-PL 3SG-clase-COM ATR-hablar-PL 3SG-COMP-poder-FIN ja-t-jui-dik bt epin-od-at] C.ADV. 3SG-COMP-aprender-EV PRIND.PL gente-PL-OBL Lingüista es el que aprende sobre estas clases de palabras para que puedan aprender otras personas

En el trozo de discurso en (62) tenemos una cláusula imperativa en inicio, seguida de una oración condicional con dos cláusulas que expresan 58 JESÚS MARIO GIRÓN una citación, y por último una cláusula finita cuyo objeto es la cláusula condicional que le precede. Obsérvese que el argumento objeto de la primera cláusula del condicional está expresado por el índice jam en el verbo -di’k y por una nominalización en posición postverbal. La segunda cláusula del condicional tiene un argumento único (n) expresado también por un marcador de persona en el verbo finito y por una nomi- nalización frasal postverbal. Aunque la segunda cláusula del condicional es intransitiva, vemos que la nominalización que expresa el argumento único puede tener un argumento objeto interno en razón de la valencia del verbo de la nominalización; pero dicho argumento interno a la nomi- nalización es completamente independiente de la estructura argumental del verbo de la cláusula matriz.

O- A-V-condicional O (62) yau-pi-kit. [jam-yaj-di’k-’u yaj-ki-yum ], nada-SBSC-llorar.NEG 3SG-2PL-dejar-ADH 2PL-llorar-NMZ U-V U om-da jan-si-’i naa jap-’i’-m-pnn” ahí-ASR 3SG-PROS-morir DEM 3SG-matar-AGT-RES jai-mok-ma kat’atj 3SG-decir-RPR D3PL “Nada de llorar. Si dejan su lloraderam, ahí sí morirá éste que lo mató” les dijo a ellos (el chamáni a los parientesj de una personap que se supone fue muerto por el participante único de la segunda cláusula del condicional)

5. Conclusión Hemos visto la formación de palabras verbales y cláusulas nominalizadas por diversos mecanismos morfológicos y cumpliendo diversas funciones en el discurso. Vimos las nominalizaciones de argumento, circunscritas en la mayoría de los casos a la formación de designaciones de nombres de participante, instrumento, lugar. Entre estas nominalizaciones sobresalen las nominalizaciones agentiva y atributiva como formas frasales, las cuales disponen el predicado desde la perspectiva del agente o del paciente, y la nominalización de acción nominal e instrumento imple- mentada por el morfema -yu. Se ha mostrado también cómo una cláusula deviene en complemento de objeto o adverbial mediante la construcción CONSTRUCCIONES NOMINALIZADAS EN WÃNSÖJÖT 59 verbal no finita (PEV), o mediante morfemas de caso y partículas adverbiales que nominalizan la construcción. Este extenso uso de nominalizaciones en wãnsöjöt es complementario de la construcción verbal finita, pero es indicativo de una tendencia discursiva a crear expresiones referenciales y predicativas perifrásticas. De otro lado se reafirma la especifi- cación categorial de las raíces léxicas, en tanto las raíces que consideramos verbales deben ser modificadas por las marcas nominalizadoras para constituir secuencias nominalizadas en la oración y funcionar como argumentos de palabras verbales finitas. Por otro lado se reconoce tam- bién la posibilidad de que constituyentes nominalizados sean predicados principales o subordinados en el discurso wãnsöjöt, lo que indica la alta frecuencia de predicaciones intransitivas en la urdimbre del discurso wãnsöjöt.

Convenciones 1 = 1ª persona, 2 = 2ª persona, 3 = tercera persona, A = agente, primer argumento de transitivos, ABL = ablativo, ADH = adhesivo, ADV = adver- bio/adverbial, AGT = agentivo, AL = alativo, ANF = anafórico, ANT = an- tagónico, ASR = aseverativo, ATR = atributivo, AVRS = adversativo, C = cláusula, CAUS = causativo, COM = comitativo, COMP = complemento, CONCS = concesivo, CONJ = conjunción, CONX = conexión, CPLT = com- pletivo, D = pronombre dativo, DEM = demostrativo, DES = desiderativo, DIMV = diminutivo verbal, DUR = durativo, ENF = enfático, ERG = ergativo, EV = eventivo, EXHOR = exhortativo, F = femenino, FIN = finalativo, FN = frase o sintagma nominal, FREC = frecuentativo, FUT = futuro, FV = frase o sintagma verbal, HABP = habituativo peyorativo, HYP = hipotético, ICMPL = incompletivo, IMP = imperativo, IPEL = impeletivo, INES = inesivo, INM = inminente, INQU = inquisitivo, INS = instrumental, INT = interrogativo, INTRJ = interjección, IPFC = imperfectivo, M = masculino, MIR = mirativo, MP = marcador de persona, N = nombre, NEG = negación (adverbial), NEGV = negación verbal, NMZ = nominalizador, NZLOC = nominalizador locativo, NZNEGV = nominalizador de la negación verbal, O = objeto, OBL = oblicuo, OD = objeto directo, OI = objeto indirecto, OND = objeto relacional no definido, P = paciente, 2° argumento de transitivos, PAS = pasado, PAT = paté- tico, PEV = construcción/palabra eventiva, PF = construcción/palabra finita, PL = plural, PNE = marcador de persona definida no especificada, PR = pronombre, PRIND = pronombre indefinido, PRNE = pronombre no específico, PROS = prospectivo, PSNM = pasado nominal, PSVG = pasado vigente, RCS = recesivo, RES = resultativo, REST = restrictivo, RPR = reportativo, RPT = repeti- tivo, S = sujeto, SBSC = subsecuente, SG = singular, SIML = simultáneo, 60 JESÚS MARIO GIRÓN

SPRL = superlativo, TC = término de clase, TOP = tópico, TRNSL = transla- tivo, U = argumento único de intransitivas, V =verbo, v = vocal armónica de agentivo, WH = pronombre interrogativo. Las iniciales entre paréntesis en los ejemplos corresponden a los colaboradores y el texto o relato de donde es extraido el ejemplo.

Referencias Bresnan, Joan (1997). Mixed Categories as Head Sharing Constructions. En Miriam Butt y Tracy Holloway King (eds.), Proceedings of the LFG97 Conference, Uni- versity of California, San Diego. On-line, Stanford University: http://www.csli. stanford.edu/publications/LFG2/lfg97.html. 17 pages (postscript). Bresnan, Joan y John Mugane (2006). Agentive Nominalizations in Gikuyu and the Theory of Mixed Categories. En Miriam Butt, Mary Dalrymple y Tracy Hollo- way King (eds.), Intelligent Linguistic Architectures: Variations on themes by Ronald M. Kaplan, 201-234. CSLI Publications, Stanford, California. 39 pp. http://www.stanford.edu/~bresnan/kikuyu.pdf. Comrie, Bernard (1976). The Syntax of Action Nominals: a cross-language study. Lin- gua 40: 177-201. Comrie, Bernard &Sandra A. Thompson (1985). Lexical Nominalization. En Timothy Shopen (ed.), Language Typology and Syntactic Description, Vol. 1, 349-398. Cambridge: Cambridge University Press. DeLancey, Scott (1986). Relativization as Nominalization in Tibetan and Newari. Paper presented at the 19th International Conference on Sino-Tibetan Languages and Linguistics. Girón, J.M. (2006). Contraste de Oclusivas y Nasalidad en wãnsöhöt (puinave). Amerindia, 81-96. Vol. 29-30. Paris: CNRS. Girón, J.M. (2008). Una Gramática del Wãnsöjöt (Puinave). Doctoral dissertation, Vrije Universiteit Amsterdam. Utrecht: LOT. Girón, J.M. & W. Leo Wetzels (2007). Tone in Wãnsöhöt (puinave). En Leo Wetzels (ed.), Language Endangerment and Endangered Languages. Serie Indigenous languages of Latin America n. 5, 129-156. Leiden: CNWS Publications. Haspelmath, Martin (1995). Word-class-changing inflection and morphological theory. En Geert Booij & Jaap van Marle (eds.), Yearbook of Morphology, 43-66. Dordrecht: Kluwer. Koptjevskaja-Tamm, M. (2006). Nominalization. En Keith Brown (ed.), Encyclopedia of Languages and Linguistics, Vol. 8, 2nd ed., 652-659. Amsterdam: Elsevier. Noonan, Michael (2005). Nominalizations in Bodic Languages. Milwaukee, University of Wisconsin. http://www.uwm.edu/~noonan/Nom.paper.pdf. Meira, Sérgio (1999). A grammar of Tiriyo. Ph.D. dissertation (UMI microfilm 9928570). Houston: Rice University. Van Valin Jr.,Robert D. & Randy J. LaPolla (1997). Syntax-Structure, Meaning and Function. Cambridge: Cambridge University Press. CONSTRUCCIONES NOMINALIZADAS EN WÃNSÖJÖT 61

Ylikosky, Jussi (2003). Defining Non-finites: Action Nominals, Converbs and Infinitives. SKY Journal of Linguistics 16: 185-237. http://cc.oulu.fi/~jylikosk/ defining.pdf.

El nombre en Nkak1

Dany Mahecha Rubio (Vrije Universiteit Amsterdam)2 Vrije Universiteit Amsterdam

0. Introducción Este artículo describe características fonológicas, morfológicas y sintác- ticas de los nominales en la lengua Nɨkak. Para empezar presenta un panorama del contexto cultural y lingüístico de los Nɨkak, incluyendo los principales rasgos tipológicos de la lengua. En seguida presenta el nombre, algunos nominalizadores, las estructuras de los nombres simples y complejos, las formas de composición semántica, las formas de determi- nación, las clases nominales, y las marcas de caso. En segunda instancia, discute los argumentos que muestran la coexistencia de dos sistemas de clasificación nominal, el de términos de clase y un conjunto de nominales que operan como clasificadores. Así como las similitudes de las características de la clasificación nominal y las marcas de caso entre el Nɨkak y otras lenguas de la familia lingüística Makú-Puinave. Final- mente, recapitula los hallazgos relevantes y discute los argumentos de considerar al Nɨkak como una lengua de esta familia.

1. Los Nɨkak Los Nɨkak son un pueblo de tradición nómada de la Amazonia colombiana. Su territorio ancestral se encuentra localizado en el departamento

1 La información presentada en este artículo hace parte del proyecto de investi- gación “La gramática del Nɨkak” dirigido por Leo Wetzels en la Vrije Universi- teit de Ámsterdam, y financiado por The Netherlands Organization for Scienti- fic Research NWO con la beca No 256-00-521. Agradezco los comentarios hechos a versiones previas de este manuscrito hechas por Leo Wetzels, Elsa Gomez-Imbert y Pattie Epps. 2 Antropóloga, Master en Estudios Amazónicos y estudiante del Doctorado en letras de la Vrije University of Amsterdam. Actualmente se desempeña como docente de la Universidad Nacional de Colombia Sede Amazonia. 64 DANY MAHECHA RUBIO del Guaviare (véase mapa) y la población actual se estima en 600 personas. Desde 1988, año en que los Nɨkak fueron contactados oficialmente, han enfrentado un sinnúmero de cambios de todo orden en su modo de vida. Entre los cambios más notorios es pertinente mencionar: una re- ducción de la población cercana al 40% en los primeros años de contacto; su ingreso al mercado como mano de obra en las fincas de los colonos que rodean su territorio y con la venta de productos artesanales; y los efectos de la presencia de actores armados en su resguardo3, que han generado el desplazamiento masivo de varios grupos locales hacia San José del Guaviare, la capital del Departamento, desde el año 2002. La mayor parte de los intentos de retorno al territorio han fracasado por los problemas de seguridad que persisten en el área y los conflictos sus- citados entre distintos grupos locales en un intento de reorganizarse terri- torialmente. En junio de 2006 se encontraban cerca de 230 Nɨkak (el 38% de la población) de distintos grupos locales en calidad de despla- zados en áreas aledañas al perímetro urbano de San José del Guaviare. En este contexto, es claro que la supervivencia de los Nɨkak, su cultura y como parte de ella el idioma, enfrenta un verdadero desafío. Por tanto es urgente que se tomen las medidas necesarias para garantizar la protec- ción de los derechos de este pueblo. El Nɨkak es inteligible con el idioma de los Cacua o Bara ubicados en el interfluvio de los ríos Papurí-Querarí en el Vaupés.4 El idioma Nɨkak ha sido clasificado como genéticamente afiliado a la familia lingüística Makú-Puinave (cf. Mason, 1950; Ortiz, 1965; Rivet y Taste-

3 Las autodenominadas Fuerzas Revolucionarias de Colombia (FARC) han hecho presencia en el área que bordea el territorio Nɨkak desde 1984. A me- diados de 1996 se inició una disputa por el río Guaviare entre las FARC y un grupo de las Autodefensas Unidas de Colombia (AUC). Desde ese momento toda la población colona e indígena del área ha vivido las consecuencias del conflicto armado, entre los que sobresalen el desplazamiento masivo hacia cen- tros urbanos y la vinculación de los jóvenes a los bandos del conflicto (Cabrera et al 1994, 1999; Mahecha et al 1998; Mahecha 2005). 4 En 1988 cuando un grupo de Nɨkak fue trasladado al Caño Wacará en el Vaupés, se empleó un traductor Cacua para comunicarse con ellos. En 1993, mientras se realizaba un documental sobre este pueblo, en él que la autora de este texto participó, también se contrató un traductor Cacua para colaborar con la traducción del Nɨkak al español y viceversa. EL NOMBRE EN NKAK 65 vin, 1920). Recientes investigaciones lingüísticas confirman la pertenencia a esta familia de las lenguas Hup (Epps, 2005, 2008), Yujup (Ospina, 2002), Dâw (S. Martins, 2004), y Nadeb (V. Martins, 2005).

Mapa 1

En contraste, los hallazgos de algunas de estas indagaciones señalan a las lenguas Cacua, Nɨkak y Wánsöhöt (puinave) como las más distantes de ese conjunto (Girón, 2006: 16-17); y cuestionan la pertenencia de la in- clusión en la familia Makú-Puinave (Epps, 2005: 6-7, 2008), sin que se cuente con información conclusiva hasta el momento.5 Los trabajos de campo6 en los que se basa este artículo se efectu- aron en los asentamientos de Villa Leonor en un sector del Resguardo del Refugio y en el de Agua Bonita. La mayoría de las personas que se encuentran allí se autoidentifican como Wajari muno ‘gente del Guaviare’, y Meu muno ‘gente de la coronilla’ y tienen dialectos propios7. No ob-

5 Martins e Wetzels (en preparación) están comparando datos fonológicos y morfológicos de las lenguas Nɨkak y Wánsöhöt (puinave) con el de las lenguas Hup, el Yujup, el Dâw, el Nadeb, denominadas por Epps (2005, 2008) como Nadahup y por V. Martins (2006) como Makú Occidental, en aras de determi- nar el parentesco de estos dos conjuntos. 6 La duración de los trabajos de campo abarcó cerca de nueve meses, en cuatros periodos entre el 2005 y el 2008. 7 De acuerdo a investigaciones previas en la lengua Nɨkak las diferencias dia- 66 DANY MAHECHA RUBIO stante la mayor parte del material lingüístico, empleado en este texto, fue colectado con los Meu muno, quienes tradicionalmente ocupaban el sector nororiental del territorio (véase en el mapa los grupos identificados con la letra B). Según este análisis fonológico, el idioma Nɨkak tiene seis vocales y 16 consonantes. La nasalidad es una propiedad suprasegmental que afecta al morfema. El rasgo [+nasal] en la trascripción fonológica es no- tado con el símbolo ~ precediendo los morfemas nasales. Esta lengua presenta un alto número de palabras monosílabas. El idioma Nɨkak presenta un patrón prosódico tono acentual. Los contrastes tonales se encuentran en los lexemas nominales monosilábi- cos, en los que se distinguen dos tonos fonológicos, uno ascendente [ ̌ ] y uno alto [ ´ ]. Las palabras monomorfémicas bisilábicas y trisilábicas llevan el acento en la primera sílaba. Las sílabas no acentuadas se consideran tonos bajos que se realizan por defecto. En palabras estructuradas la correlación entre tono y acento varía dependiendo del tipo de lexemas que la constituyen y de la relación semántica entre los mismos. Aún está por definir las características de los patrones prosódicos en relación a las funciones gramaticales de los distintos tipos de palabra en la lengua. En la transcripción fonológica de las palabras monomorfémicas bisilábicas y trisilábicas no se marcará el acento, en cuanto este aspecto es predecible. En las palabras estructuradas se marcará el tono y el acento, por cuanto este aspecto está en proceso de análisis y definición. El acento se mar- cará con el símbolo [  ]. El Nɨkak presenta en la caracterización de los nominales y adjetivos rasgos de una lengua aislante, mientras que en la morfología verbal los de una lengua aglutinante8. Las principales clases léxicas son los lectales son notorias en el léxico y es probable que también existan diferencias en el nivel fonológico. Cabrera et al. (1994) señalaron la existencia de tres grandes dialectos, y datos de esta investigación permitieron identificar cinco variedades dialectales: la de los Waɟari muno ‘el grupo del río Guaviare’; los Meu muno ‘grupo de la coronilla’; Takajud muno, ‘grupo del centro’; Mua beʔ muno ‘gente de la Laguna grande’ y los Mipa muno ‘gente del río Inírida’. La propuesta de gramática nɨkak elaborada por Hess et al (2005) fue hecha con los materiales colectados con los grupos locales del nororiente los Wayari muno y del suroriente los Mua beʔ muno. 8 Como anota Lyons (1975: 198) no es conveniente definir los tipos de lenguas con una tipología nítida, en cuanto no existen tipos puros y resulta más apropi- EL NOMBRE EN NKAK 67 nombres, los adjetivos y los verbos. La morfología nominal presenta es- cazas formas derivadas, no tiene distinciones gramaticales de género ni número y la composición nominal es un proceso muy productivo en la formación de palabras y en la incorporación de neologismos. Hasta el momento se han identificado cinco casos que son: acusativo, dativo, locativo, genitivo y comitativo, de estos los cuatro primeros se sufijan a los nombres. En contraste la morfología verbal es más compleja, enuncia las categorías de actancia, tiempo, aspecto y modo, y no presenta distinciones gramaticales de generó y número. El orden de los constituyentes prevalente es S V O aunque la marca de objeto indirecto -~da puede sufijar al sujeto y se pueden presentar otros órdenes dependiendo del contexto pragmático.

2. El nombre El nombre como categoría gramatical tiene la función de designar una relativa estabilidad (cf. Givón, 2001). Según Frawley (1992) esta carac- terística del nombre es reconocida desde tres aproximaciones distintas. De acuerdo a Hopper y Thompson (1984), esta estabilidad del nombre es informacional, individualizada y destacada por los participantes del discurso. Mientras que para Givón (1984) la estabilidad del nombre es temporal y es asociada a un fenómeno espacialmente fijo. Y finalmente para Langacker (1987) se trata de la estabilidad cognitiva, una región densa e interconectada en un espacio conceptual (cf. Frawley, 1992: 69-70). De tal manera que “los nombres pueden no siempre ser personas, lugares, o cosas, pero personas, lugares y cosas casi siempre pueden convertirse en nombres. Entidades, relativamente estables y discursos atemporales, on- tológicos, y fenómenos conceptuales, que motivan clases de forma” (Frawley, 1992: 68).9 En el idioma Nɨkak el nombre es una clase de palabra abierta, que presenta diversas formas de composición productivas. Esta clase está in- tegrada por un conjunto de entidades animadas (humanas y no humanas) e inanimadas que hacen parte de la cosmovisión. También por objetos,

ado reconocer que de acuerdo a las clases de palabra pueden presentar rasgos que los acercan más a una clasificación tipológica que a otra. 9 La traducción es de la autora. 68 DANY MAHECHA RUBIO sustancias, fenómenos meteorológicos, tipos de paisaje, antropónimos y topónimos. Los nombres pueden tener una estructura morfológica simple o compleja. Los simples son formados por una raíz léxica y pueden ser monosílabos y bisílabos. También existe una pequeña proporción de trisílabos. La mayoría de estos son nombres de especies animales (sobre todo de peces) y vegetales entre las que se encuentran varias plantas cul- tivadas. Algunas de estas palabras son de origen Arawak, por ejemplo ~kedaɹiʔ ‘espejo’ y es probable que otras que se refieren a especies culti- vadas también lo sean como cɨpɨɹiʔ, ‘piña (Ananas comosus)’, ~ukuɹiʔ ‘cierto tubérculo’, daɹuaʔ ‘cierto calabazo (Ligenaria siceraria)’. Los nombres que tienen estructura morfológica compleja pueden ser derivados, compuestos y los originados por procesos de nominali- zación. En lo que sigue presento las características de estos nombres.

2.1 Nombres derivados Los nombres derivados en Nɨkak están constituidos por una raíz léxica y un morfema gramatical derivacional. Estos constituyen unidades fonoló- gicas y sintácticas que denotan antropónimos, nombres de época10, y aumentativos y diminutivos.

2.1.1 Los Antropónimos. Los nombres propios pueden ser lexemas que no se pueden segmentar y cuyo significado denota exclusivamente nombres propios como se observa en 1-3. También existen nombres propios que se derivan de animales, vegetales y partes del cuerpo. Un número importante de estos nombres propios femeninos se sufijan con el morfema -~ba, véase 4-5, mientras que los masculinos se sufijan con el morfema -daʔ, véase 6-8. Al parecer la función de estos sufijos es constituir nombres propios de personas. Aquí vale resaltar que el uso del sufijo - ~ba es más recurrente que el -daʔ, sin que se haya encontrado una expli- cación para esta diferencia. Contemporáneamente los Nɨkak tratan de evitar el uso de los nombres propios en la lengua nativa, y prefieren adoptar nombres en español, para relacionarse con personas foráneas. Sin embargo, los adul-

10 Se toma el nombre de ‘época’ propuesto por Ospina (2002: 191) en su de- scripción del Yuhup para designar un morfema que tiene la misma función. EL NOMBRE EN NKAK 69 tos que aún usan los nombres en Nɨkak los emplean con estos sufijos. Ospina (2002: 191) señala que el Yujup tiene el prefijo ʔɨh- para marcar los nombres propios. Y que este prefijo parece cumplir la función de denotar la cualidad humana del referente evitando la ambigüedad con los lexemas que semánticamente se refieren a especies vegetales o animales. Queda por indagar si en el Nɨkak estos sufijos tienen esa función. La lengua Dâw también tiene un prefijo que cumple la función de formar nombres propios (Martins y Martins, 1999, citados en Ospina, 2002:191).

(1) ~bauɹɤ ‘héroe mítico que dio origen a la humanidad actual’ (2) hauɟaɤ ‘nombre propio masculino’ (3) ~ɨbu ‘nombre propio femenino’ (4) ~kada-~ba ‘nombre propio femenino’ maíz-MNPR (5) teip-~ba ‘nombre propio femenino’ mano-MNPR (6) pahi-daʔ ‘nombre propio masculino’ pez sp cucha-MNPR (7) hiw-daʔ ‘nombre propio masculino’ felino-MNPR (8) puɟu-daʔ ‘nombre propio masculino’ algodón silvestre-MNPR

Según los Nɨkak, cuando una persona muere, todos sus objetos personales deben ser destruidos y su nombre no debe volver a mencion- arse y menos cerca de los parientes que se encuentran vivos. Los Nɨkak argumentan que esto es una muestra de respeto a los parientes y una forma de prevenir que uno de los espíritus del muerto regrese a este nivel del mundo a llevarse a sus seres queridos vivos. La única forma de hablar de las personas fallecidas es mencionar la relación de parentesco y agre- gar el sufijo -bitui ‘fallecido’. Es por esto que cuando muere una persona, los Nɨkak de este grupo local dejan de usar las palabras que constituían el nombre en su propio dialecto y comienzan a usar un nombre que define propiedades similares, o adoptan el léxico de otra variedad dialectal. Por ejemplo, una persona que se llamaba kíb ‘ojo’ falleció, luego sus parientes no podían usar esta palabra más y en su lugar adoptaron la palabra ~ed-hát ‘ver-NR’. Esto explica el que algunos nombres de animales y vegetales, cuando son usados como nombres propios, puedan tener hasta 70 DANY MAHECHA RUBIO cuatro sinónimos, y que un grupo de hablantes de un dialecto pueda adoptar los nombres de ciertas especies de otro grupo en determinado momento (véase ejemplos 9 a, b, c, d).

(9) anaconda se denomina: a. ~bai por los Meu muno ‘sebucán o matafrío’ b. túi por los Waɟari muno ‘deslizar’ c. ~bɨh bu-ɹiʔ por los Takajudn munɤ ‘dueño del caño’ caño dueño-LOC d. tui-~di-de por los Muabeʔ muno ‘literalmente la deslizadora’ deslizar-ACT.E-FREC

2.1.2 Nombres de época. Los nombres de época se pueden constituir de dos maneras, una es la adición a una raíz lexical del sufijo -ɟubɤ que denota ‘época’ o en el español regional ‘tiempo de’. La raíz lexical generalmente denota una especie vegetal o animal, véase los ejemplos 10-11.

(10) ~wada-ɟubɤ ‘época de ~wada’ frutal sp.-EP (11) púi-ɟubɤ ‘época de rana’ cierto batracio-EP

La otra forma de constituir los nombres de época es la adición a una raíz verbal nominalizada del sufijo -ɟubɤ, como en 12-13.

(12) bɨg-hát-ɟubɤ ‘tiempo de lluvia o invierno’ Lluvia, invierno-NR-EP (13) hihi-hát-ɟubɤ ‘tiempo de hacer calor o verano’ hacer calor-NR-EP

Estos nombres pueden estar determinados por un pronombre como en el ejemplo que sigue:

(14) Carlos a húi-~hát -ɟubɤ Waɟaɹi ~budɤ hiw-ka bé AguaBonita-~da NPR 3SGM venir-NR EP Guaviare gente haber-NEG-PASREC NPR-OD ‘En el tiempo en que Carlos vino la gente del Guaviare no estaba en Agua Bonita’ EL NOMBRE EN NKAK 71

La lengua Yuhup presenta el morfema -tóh, el cual tiene carac- terísticas morfológicas muy similares al de -ɟubɤ en Nɨkak de acuerdo a la descripción de Ospina (2002: 192), quien además describe las otras estructuras sintácticas en que se emplea este morfema y que están por explorarse en Nɨkak como el de marca hipotética. Finalmente, se debe anotar que en algunas lenguas de la familia lingüística Tucano oriental como el Cubeo y el Macuna se presentan formas nominales que operan como los nombres de época que presentamos en esta sección11.

2.1.3 Aumentativos y diminutivos. El morfema aumentativo -beʔ se sufija a nombres de entidades animadas e inanimadas e indica un au- mento en la magnitud del tamaño, véase ejemplos 15-17.

(15) Luis-beʔ ‘Luis el grandote’ Luis-AUM (16) ~díh-beʔ ‘tucán grandote’ (Ramphastos tucanus)ʼ Tucán-AUM (17) ít-beʔ ‘hueco grandote’ hueco-AUM

El morfema diminutivo -bit se sufija a nombres de entidades animadas e inanimadas para indicar una disminución del tamaño, como en los ejemplos 18-20.

(18) ~wɨʔ́ -bit ‘hijito’ hijo-DIM (19) hǔ-bit ‘mascotica’ seno-DIM (20) kaɟu-bit ‘pollito’ pollo-DIM

11 En las lenguas Cubeo y Macuna los nombres de época también son sufijos. En acuna se denomina -rodo, por ejemplo ote-rodo ‘tiempo de frutas’ (cf. Smothermon & Smothermon (1993). En Cubeo el sufijo que corresponde a nombre de época es -nɨmɨ. como en mijinùmù ‘tiempo mítico, cuando inicialmente tomaron el yajé’ (Morse et al 1999: 410). 72 DANY MAHECHA RUBIO

Los sufijos -beʔ aumentativo y -bit son átonos, por lo que las palabras constituidas con estos morfemas presentan una correlación de tono y acento en la primera silaba siguiendo el patrón prosódico de los bisílabos y trisílabos monomorfémicos en Nɨkak.

2.2 Nominalizadores La lengua Nɨkak presenta formas de nominalización que se constituyen a partir de raíces verbales y predicados verbales sufijados por nominalizadores, los cuales se describen a continuación. El morfema nominalizador -hát se sufija a los lexemas verbales para crear nombres de partes del cuerpo, los objetos o entidades que están involucrados en la acción, como en 21-23 o conceptos abstractos como en 24. En el proceso de derivación la raíz verbal mantiene el tono.

(21) ~cɨi–hát ‘pie’ pisar, patear-NR (22) hui-hát ‘oído’ escuchar, percibir-NR (23) ta-hát -~dá ‘ropa’ atravesar-NR-plano (24) kab-hát ‘enfermedad’ enfermar-NR

El frecuentativo de se sufija a los predicados verbales para constituir nombres que denotan tipos de actividades o nombres propios que describen los hábitos o características de los animales como en 25-26.

(25) 'kúʔ-~di-'de ‘enfermero’ curar-ACT.E-FREC (26) ak-~di-'de ‘gallineta sp’ rascar-ACT.E-FREC (literalmente que siempre se rasca)’

El morfema -péʔ se sufija a los lexemas verbales para nominalizar el evento del que se está hablando, como en los siguientes ejemplos:

(27) i wab-péʔ ‘lo que ellos comen’ 3PL comer algo blando -NR EL NOMBRE EN NKAK 73

(28) ~bi diu-péʔ ced-~wɨd kɤpe ‘Lo que ella dio a luz dos 3SF parir-NR estar juntos-COL nuevo nuevos12 (bebés)’

2.3 Nombres compuestos En Nɨkak los nombres compuestos se constituyen a partir de la yuxta- posición de dos o más morfemas de la misma clase gramatical o de dis- tinta. La mayoría de los componentes de los compuestos son raíces léxi- cas que se pueden usar en las formas libres, como las de 29-30 pero otros incluyen morfemas que se sufijan a los verbos como en 31.

(29) tɨ́a-pukuʔ ‘ceniza’ leña montón (30) ~dei-butu ‘roedor (Dasyprocta sp.) o chaqueto pequeño’ roedor sp. pequeño (31) tak-hát-~daʔ ‘carrete de hilo’ coser-NR cilíndrico

Los rasgos que permiten identificar los compuestos son: 1) la cohe- sión como unidades fonológicas en las que los compuestos tienden a mantener las formas segméntales, pero pueden perder su tono lexical ya que el conjunto presenta un sólo acento de intensidad13; y 2) la cohesión como unidades semánticas en cuanto los compuestos se refieren a lexemas nuevos. Sin embargo, en la lengua Nɨkak se distinguen dos tipos de compuestos de acuerdo al grado de cohesión semántica, ya que algunos denotan significados más lexicalizados, menos o no composicionales, y otros totalmente composicionales.14 En los compuestos más lexicalizados

12 Entre los Nɨkak a los recién nacidos no se les asigna nombre y se les denomina literalmente kɤpe ‘nuevos’. 13 El rasgo más prominente para definir un compuesto son las características prosódicas en la correlación tono acento entre los componentes. En cuanto en los nombres compuestos una palabra acentuada prevalece sobre los demás componentes (cf. Bloomfield 1933: 228, citado en Katamba, 1993: 294). 14 Las descripciones de las lenguas Yujup (Ospina, 2002), Hup (Epps, 2005) y Wánsöhöt (Girón 2008) señalan características fonológicas similares en la con- strucción de los compuestos. Específicamente el Hup también presenta diferencias en el patrón prosódico de los compuestos más lexicalizados y el de los 74 DANY MAHECHA RUBIO generalmente el acento coocurre con el tono en N1 (o en la primera sílaba de este) y se elide el tono de N2 como en los ejemplos 32-34. En los compuestos menos lexicalizados se mantienen los tonos, pero coocu- rren en N2 con un acento de mayor intensidad como en 35-37.

(32) ~deba-taʔ ‘arco iris’ mariposa -tubular (33) áb-pu ‘delfín’ subir-soplar (34) íg~di ~daʔ ‘linterna’ brillante-tronco (35) íbɟi tɨ́t ‘cordón de amarrar la hamaca’ hamaca cuerda (36) nɨ́k páʔ ‘paladar’ lengua tapar (37) kíb ~céd ‘cejas’ ojo pelo

Sintácticamente los nombres compuestos operan como argumentos o predicados no verbales en la misma forma que los nombres simples y los sintagmas nominales. En concordancia con esto pueden ser modificados por un determinante calificativo o nominal, o pueden modificar un nombre. La composición nominal es además uno de los procesos más productivos en esta lengua, y ello se refleja en la incorporación de un extenso vocabulario acerca de las herramientas y objetos industriales que los Nɨkak adoptaron luego del contacto15, como se ilustra en los ejemplos 38 a, b y c. En estos nombres se establecen asociaciones que involucran conceptos de forma a-b y de función c.

menos lexicalizados (cf. Epps, 2005: 186-188, 2008). 15 Payne señala que la propiedad semántica dominante de los compuestos es que el significado del compuesto constituye una unidad semántica que origina un nuevo concepto, y el significado de este puede ser más preciso o diferente del de los componentes (cf. 1997: 93). EL NOMBRE EN NKAK 75

(38) a. kib paʔ-hát daʔ ‘botón’ ojo cerrar-NR redondo b. wǎm tɨt ‘manija’ olla cuerda c. we-hat-dɤʔ ‘armario’ guardar-NM recipiente

En seguida se presentan los tipos de composiciones nominales de acuerdo a las clases lexicales que pueden combinarse y a las relaciones semán- ticas entre constituyentes.

2.3.1 Nombre + Nombre. Los compuestos constituidos por la secuencia de dos nombres denotan tres tipos de relaciones de acuerdo a su función semántica: a. La relación de poseedor-poseído en la que el núcleo del compuesto es siempre N2 como en 39-40.

(39) ~debep ~tɤ̌i ‘yuca de los nemep16’ nemep-yuca (40) buɟup-~budɨ ‘chontaduro de colibrí’ colibrí-chontaduro b. La relación unidad-parte. En estos compuestos, que se refieren a las partes del cuerpo humano, animal y vegetal, el primer constituyente denota un concepto genérico que es precisado por el segundo. Sintáctica- mente presentan una relación de determinante determinado, y denotan una relación semántica de pertenencia a una entidad completa, tales como mano → dedo, flor → pétalo, cabeza → cornamenta, véase 41-46.

16 De acuerdo a los Nɨkak los nemep son otro de los espíritus que una persona tiene. Cuando la persona fallece estos se quedan en el bosque, tienen una apariencia peluda y son dañinos para los seres humanos y no humanos vivos, pues se alimentan de la sangre. 76 DANY MAHECHA RUBIO

Nombres de partes del cuerpo humano (41) téip tíb ‘dedo de la mano’ mano dedo (42) teip ~tɤʔ ‘muñeca’ mano cabeza17

Nombres de partes de plantas (43) cá wawa ‘pétalo’ flor hoja (44) cá tíu ‘estambres’ flor estar acanalado

Nombres de partes del cuerpo de los animales (45) ~céɨʔ peu ‘cresta’ cabeza cartílago (46) ~ceɨʔ-hawi-ɹɤ-hát ‘cornamenta’ cabeza-sembrar-brotar-NR c. Una relación semántica que expresa localización u origen. En estos compuestos, que tienen la cabeza sintáctica a la derecha, N1 ejerce la función de determinante y encierra la noción de locativo, y el segundo constituyente N2 es el determinado. En los ejemplos 47-48 el resultado de la composición denota hábitos característicos de una especie de abejas y de un tipo de gente, para el primero en 47, es posarse alrededor de los ojos de los humanos y para los segundos en 48 es vivir en la selva.

(47) kib-~de-~wɨd → kib ~de-~wɨd ‘abeja Ducke’ (Paratrigona impunctata) ojo-larva-COL

(48) ɟéʔ-~budɤ → ɟéʔ ~budɤ ‘guerrilleros’ selva, monte- gente

2.3.2 Nombre + Verbo estativo descriptivo. Los compuestos son constituidos por un nombre más un verbo estativo descriptivo como en los

17 ~tɤʔ ‘cabeza’ en el dialecto de los Wayari mɨno. EL NOMBRE EN NKAK 77 ejemplos 49-50. Este orden de los constituyentes es el mismo de los sintagmas nominales, en los que estos verbos pueden actuar como calificativos, pero se distingue de ellos por las reglas fonológicas que caracteri- zan la prosodia de estos compuestos como una unidad, por la ausencia de las marcas de actancia y porque el contenido semántico corresponde a una entidad específica. Estos compuestos presentan una relación de determinado-determinante.

(49) híw buru → híw -buruʔ ‘jaguar negro (Panthera onca)’ felino negro

(50) ~bɨdɨ ɟɤɹe → ~bɨdɨ-ɟɤɹe ‘chontaduro amarillo (Bactris gasipaes)’ chontaduro amarillo

En los sintagmas nominales también se ha evidenciado el uso de dos lexemas calificativos kɤpe ‘nuevo’ y bǎ ‘viejo’ que pueden preceder al nombre como 51-52, pero no se han encontrado palabras compuestas que presenten este orden de los constituyentes, es decir que existan compuestos del tipo lexema calificativo + nombre.

(51) ~ bi bǎ ɟéd-ɟi ‘el huerto viejo de ella’ 3SGF viejo huerto ¿? (52) ~kad kɤpe wawa ‘ese cogollo de árbol’ VIS.S.M.D nuevo hoja

2.3.3 Nombre + Verbo activo. Los compuestos constituidos por la secuencia nombre + verbo activo de acuerdo a su función semántica denotan composiciones locativas y descriptivas. Las composiciones locativas tienen las mismas características sin- tácticas de la secuencia nombre + nombre. En los que el primer nombre ejerce la función de determinante y encierra la noción de locativo y el segundo constituyente ejerce la función de determinado como en 53. El producto de la composición, en este ejemplo, es el nombre de una especie de abeja que le gusta colgarse de los cabellos de los humanos. 78 DANY MAHECHA RUBIO

(53) ~céɨʔ-hig-~wɨd → ~céɨʔ hig-~wɨd ‘abeja Della Torre’ (Scaptotrigona pectoralis) cabeza-amarrar-COL Literalmente ‘las que se agarran en la cabeza’

En las composiciones descriptivas de la secuencia nombre + verbo activo la semántica de los componentes se refiere a la función y la apariencia del objeto o entidad que designan como en 54-55.

(54) ~ceɨʔ- paʔ ‘cachucha’ cabeza-tapar (55) kib paʔ-hát daʔ ‘botón’ ojo cerrar-NR redondo

2.3.4 Verbo activo + Nombre. Los composiciones de la secuencia verbo activo+ nombre son descriptivas y expresan la función y/o aspecto del objeto o entidad que designan, véase ejemplos 56-58. En estas composiciones la relación entre N1 y N2 es determinante determinado.

(56) cɨi-tiw-hát daʔ → ‘balón de fútbol’ pisar-jugar-NR redondo

(57) ~hɨɟ-hát ~bakida → ‘grabadora’ (literalmente maquina de escuchar) escuchar-NR maquina

(58) hí-hat tab → ‘pluma’ saltar-NR hoja

Los nombres compuestos constituidos por la yuxtaposición de una palabra en Nɨkak y una palabra en español son escasos. La palabra ~bakida ‘maquina’ es el único ejemplo registrado que tengo de un lexema foráneo que ejerce la función de N2 como una especie de clasificador en este tipo de compuestos, y es muy productivo. Así se encuentra que las palabras que designan todos los aparatos electrónicos se componen de un lexema verbal que denota la función principal del aparato o la forma como funciona y la palabra ‘~bakida’, véase 59-60. EL NOMBRE EN NKAK 79

(59) kiriʔ-hát ~bakida ‘cámara fotográfica’ Hacer un ruido como un click-NR máquina

(60) ~ɟeke-hát ~bakida ‘computador’ escribir-NR máquina

Los ejemplos 56-60 muestran procesos de derivación en el que N1 es un lexema verbal que se nominaliza y al yuxtaponerse a un nominal que tiene propiedades de clasificador nominal da origen a referentes que indican la función de un objeto18.

2.3.5 Verbo activo + Verbo activo. Los compuestos constituidos por la secuencia verbo activo + verbo activo son descriptivos y la semántica de los compuestos se refiere a la forma, función y características de la entidad que describe, como se observa en 61-62. En estos ejemplos la nomi- nalización surge al yuxtaponer los lexemas verbales sin marcas específi- cas que la denoten, aunque cada uno de estos puede recibirlas en otros contextos. Este tipo de compuestos es el que presenta menos ocurrencias.

(61) téw-pede ‘cerbatana’ coger-apuntar (62) áb-pu ‘delfín’ subir-soplar

2.3.6 Clasificación nominal. Los compuestos menos lexicalizados en Nɨkak presentan una estructura en la que el primer constituyente N1 es un nombre específico y el segundo se refiere a una parte del primero (literalmente). El conjunto de lexemas que se presenta en N2 son determinados por el primer componente y la extensión semántica que representa el concepto N2 define un estado inherente al resultado de su procesamiento y/o a la forma del objeto o entidad19. En seguida se presentan

18 Aikhenvald (2000: 220, citada en Epps, 2005: 121) señala que estas funciones derivacionales son comunes para los clasificadores nominales en otras lenguas del Noroeste Amazónico. 19 Ospina (2002, 2004-2005) y Epps (2005, 2007) describen ejemplos de com- posición nominal en las lenguas Yujup y Hup que evidencian estas característi- cas. 80 DANY MAHECHA RUBIO ejemplos de los compuestos con esta estructura de acuerdo al concepto que denota el segundo componente: partes de la planta, clases de materiales o sustancias y tipos de objetos de acuerdo a la forma.

Partes de la planta Entre los morfemas que se emplean para designar partes de la planta en Nɨkak se tienen: ~daʔ ‘tronco’, tekeɹe ‘fruto de corteza dura’, wawa ‘hoja’, y cú ‘corteza’, véase los siguientes ejemplos de nombres compuestos en los que estos morfemas van ligados. En 63a el formema ~daʔ ‘tronco’ o ‘cilíndrico’ precisa que se trata del tronco de una especie vegetal específica. En los ejemplos 63b y c se evidencia la extensión semántica del morfema para clasificar objetos ci- líndricos.

(63) a. ɟáb ~daʔ ‘tronco del milpesos (Oenocarpus bacaba)’ Milpesos tronco b. hé ~daʔ ‘cerro, montaña’ piedra tronco c. ~deke  hat ~daʔ ‘lápiz, esfero’ escribir-NR-tronco

En 64a el morfema tekeɹe ‘fruto de corteza dura’ es especificado por una especie vegetal y en 64b se emplea como extensión semántica para describir un producto granulado como la fariña.

(64) a. popeɹe tekeɹe ‘semilla de milpesillo (Oenocarpus mapora)’ milpesillos fruto b. ~dě-ɟi tekeɹe ‘fariña’ cazabe-¿? fruto

En 65a el morfema wawa ‘hoja’ es especificado por una especie vegetal y en 65b es una extensión semántica para designar objetos lamiformes.

(65) a. ~huda wawa ‘hoja de platanillo (Phenakospermun guyanense)’ platanillo hoja b. ~ ɟeke-hat wawa ‘hoja de cuaderno’ escribir- NR hoja EL NOMBRE EN NKAK 81

En 66a el morfema cǔ especificado por una especie vegetal, y en 66b y c corresponde a una extensión sémantica que denota la forma laminar de un objeto y de una parte del cuerpo.

(66) a. tegebɤ. cǔ ‘corteza de tegebɤ’ frutal sp. corteza b. ~cɨɟ-hát cǔ ‘chancleta’ pisar-NR corteza c. '~húɟap cǔ ‘pabellón del oído’ oído corteza

Clases de materiales o sustancias Hasta el momento se han identificado cuatro morfemas que denotan sustancias o materiales en Nɨkak, estos son ~baʔ ‘caldo, jugo’, ɟéb ‘hueso’, céd ‘cabello, pelo’, cúi ‘carne, masa’. En 67a y b los compuestos designan el resultado del procesamiento del primer compuesto, en c una sus- tancia producida por el cuerpo y en d, e y f denota parte de los cuerpos de animales.

(67) a. ɟab ~baʔ ‘jugo de milpeso (Oenocarpus bataua)’ milpeso jugo, caldo b. kahiwa cui ‘masa de yuca dulce’ yuca dulce carne, masa c. kíb ~baʔ ‘lagrima’ ojo jugo, caldo d. ~ babo  ɟéb ‘hueso de venado’ venado hueso e. patcu ~céd ‘pelo de choruco (Lagothrix Lagotricha)’ Choruco pelo f. hiwi cui ‘carne de danta (Tapirus terrestres)’ danta carne, masa

Formas y/o clasificaciones de las entidades y objetos Los morfemas ~dá ‘plano’, dɤʔ recipiente, daʔ ‘redondo’, tɨ́t ‘filiforme’, ~taʔ ‘tubular’ designan tipos de objetos y entidades que presentan estas formas. En los ejemplos 68 a, c, d, e, f y g el morfema sirve para afinar la descripción al relacionar la entidad con una forma específica. En 68b es- pecífica que se trata del fruto de la planta y en h de una especie de frutal 82 DANY MAHECHA RUBIO comestible, cuya morfología es similar a la de una liana con muchas variedades.

(68) a. ~ɟeke-hat ~dá ‘libro’ escribir-NR plano b.  kuraʔ daʔ ‘fruto de la palma real’ palma sp. redondo c. cɨi-tiu-hát daʔ ‘balón de futbol’ pisar-jugar-NR redondo d. we-hát dɤʔ ‘armario’ guardar-NR recipiente e. wě dɤʔ ‘pilón’ arbol sp. recipiente f. ~wǎb tɨ́t ‘manija de la olla’ olla filiforme g. ~euʔ20 ~táʔ ‘cuello’ cabeza tubular h. ciwaʔ ~táʔ ‘inga sp’ inga sp tubular

Los constituyentes de los compuestos que se han descrito en esta sección se relacionan semánticamente como ‘términos de clase’, enten- diendo que estos son: (…) “morfemas clasificadores de claro origen lexical y presentan variados grados de productividad en el lexicón de una lengua. Uno de los dominios semánticos más comunes de los términos de clase es el de el mundo de las plantas, donde las lenguas especifican la diferencia entre árboles y frutas por un proceso de composición (…)” (Grinevald, 2000: 59). Aunque algunos aspectos semánticos permiten diferenciar dos tipos de compuestos. En unos casos la función de los morfemas es precisar los nombres de las especies vegetales o animales de las que se está enunciando una parte o un atributo. En estos casos los significados de las palabras constituidas son composicionales como ‘tronco de milpeso’, ‘hoja de platanillo’, ‘corteza de determinado árbol’, etc. Mientras que en los contextos en que la función de los morfemas es

20 En el dialecto de los Muabeʔ muno ~euʔ significa cabeza, sin embargo los Meu muno siempre designan cuello como ~euʔ~taʔ. EL NOMBRE EN NKAK 83 operar como extensiones semánticas para referirse a cualidades, carac- terísticas o atributos similares al del nominal determinado, los conceptos denotan significados menos composicionales como ‘cerro’, ‘lápiz’, ‘fariña’, ‘armario’, ‘balón’ entre otros. Desde esta perspectiva, los morfemas que ocupan la posición N2 presentan características de clasificadores nominales que definen la forma del objeto o entidad y son al mismo tiempo un mecanismo y un recurso muy productivo para incorporar neologismos.21 En Nɨkak todos los morfemas que tienen características de clasificadores nominales se encuentran en formas libres y funcionan claramente como nominales, es decir pueden constituir núcleos de sintagmas nominales, ser determinados por un lexema calificativo, ser modificados por un pronombre posesivo, ser referidos anafóricamente o substituidos por un pronombre22 como sucede en los ejemplos 69a, b, c y d.

(69) a. wa ed-~da-bé a ~daʔ-~da ~daba taka-ɹiʔ 1SG ver-ACT.A- PASREC 3SGM tronco-OD camino medio-LOC ‘Yo vi un tronco en la mitad del camino’ b. ~kúdˀcǔ a 'bara INVIS.S.F.D tela 3SGM blanca ‘Esa tela es blanca’ c. wa dɤʔ a tab~de e ·S3SGM llenar-DUR ‘mi cajón está lleno’ d. ded di ɹáʔ wa tɤɹɤ dɤʔ ~bi ɟawap wa ~ed-da-béʔ ced-wɨd23 dɤʔ Dónde IT1SG timbo recipiente otro día 1SG ver- ACT.A- PASREC estar juntos COL recipiente ‘¿Dónde está mi timbo? Ayer yo vi dos timbos’

21 Esta es otra de las características ampliamente documentadas en el Yujup (Ospina 2002, 2004 -2005) y el Hup (Epps 2005, 2007). 22 La función anáforica de los clasificadores nominales ha sido reportada como un rasgo tipológico de las lenguas amazónicas (Derbyshire y Payne 1990: 243). 23 En la lengua Nɨkak esta es la manera denominar una cantidad que contempo- ráneamente se expresa en español como el número dos o tres. 84 DANY MAHECHA RUBIO

De hecho se puede observar nítidamente como están cambiando el significado semántico de algunos morfemas que apuntan a conformar un conjunto de clasificadores nominales como se ve en (70a, b, c, y d).24

(70) a. ~daʔ ‘tronco’ → cilíndrico b. tɨ́t ‘hilo’ → filiforme c. cǔ ‘corteza’ → lamiforme d. dɤʔ ‘recipiente’→ ahuecado

Recapitulando la composición nominal en Nɨkak presenta las características de un sistema de términos de clase y el desarrollo paralelo de un sistema embrionario de clasificación nominal. Este sistema com- parte rasgos similares al de los Yujup descrito por Ospina (2002) y al de los Hup por Epps (2008) como se verá más adelante con la comparación y discusión de estos aspectos en otras lenguas de la familia Makú- Puinave.

2.4 Flexión Nominal En Nɨkak los nombres operan como sujeto o núcleo de sintagma nominal (71), objeto (72), o como constituyentes de frases genitivas y posposi- cionales (73-74).

(71) hiwbeʔ a cɨ́~da ~babɤ-~da Jaguar onza 3SGM encontrar- ACT.A venado- OD ‘El jaguar encuentra un venado’

(72)  ~kád a héb-ɨi-~dá ~hiwbeʔ -~da VIS.S.M.D 3SGM comer-desear- ACT.A Jaguar onza-OD ‘Ese desea comerse el jaguar’

(73) a ɟád 'kúʔ-~di- 'de- ~íʔ 3SGM esposa curar- ACT.E- FREC- POS ‘La esposa del enfermero’

24 Gomez-Imbert (2007: 419- 420) describe entre los marcadores de clases que denotan formas en Tatuyo morfemas que tienen el mismo significado semántico que en nɨkak. El Tatuyo es una lengua clasificada como Tucano oriental. EL NOMBRE EN NKAK 85

(74) ~kád a téu-~da-bé teɹu-~iʔ patcu-~da a ~céɨʔ- ɹiʔ VIS.S.M.D 3SGM coger- ACT.A PASREC dardo- POS choruco-OD 3SGM cabeza-LOC ‘Ese capturó al choruco con su dardo en la cabeza’

El Nɨkak presenta cinco marcas de casos gramaticales que afectan los nominales, estos son -~dá, acusativo, -~íʔ genitivo, -ɹéʔ dativo, -ɹíʔ locativo y -~hid comitativo. La marca de -~dá acusativo es obligatorio para entidades anima- das25 y opcional para las inanimadas.26 Este morfema generalmente sigue el patrón tonal del nominal al que se sufija, excepto cuando esta sufijado a palabras de tono alto con dos o más sílabas. En estos casos el tono es bajo.

(75) William dupi a wɨ-tá-ɟu-~da-bé ~pad-ɟat-~dá NPR todos- 3SGM dar-repartir-MA?ACT.A-PASREC comida-NR-OD ‘William le dio a todos comida

En Nɨkak las palabras que se refieren a las partes del cuerpo y a los términos de parentesco se consideran posesiones inalienables como en 76a-b. Cuando se trata de dos nombres que se encuentran en el mismo sintagma nominal, el orden de los constituyentes poseedor-poseído expresa de por sí una relación de posesión que no requiere marcas adicionales. Sin embargo, en la expresión de la posesión alienable el morfema - ~íʔ genitivo marca al poseedor opcionalmente para enfatizar esta relación como en 76c.

(76) a. wa ~íd ‘mi madre’ 1SG madre b. wa ~ceɨʔ-~da ‘mi cabeza’ 1SG cabeza-OD

25 Los pronombres personales libres y los demostrativos deícticos también llevan este sufijo constituyendo palabras fonológicas más reducidas. Por motivos de espacio no se desarrollará este aspecto en este artículo. 26 Zuñiga (2007) muestra como la marca opcional de OD para entidades inanimadas es un rasgo común a varias lenguas de la región (cf. 219-220). 86 DANY MAHECHA RUBIO

c. ~do-beʔ- ~íʔ a ̍baru ced- ~wɨd NPR-grande-POS 3SGM perro estar juntos -COL ‘los dos perros de Manuel’

El morfema -ɹéʔ dativo, tiene como función denotar estatus de destinata- rios o receptores y se sufija a los nombres y pronombres como en 77a y b.

(77) a. ~díd ~wawa bɨrɨp -ɹéʔ VIS.S.M.P hoja canasto- DAT ‘esta hoja es para el canasto’ b. ~wiʔ-ɹéʔ 1S.POS-DAT ‘para mi’

El morfema -ɹíʔ se sufija a los nombres y a los sintagmas nominales que denotan locación como en los siguientes ejemplos.

(78) a. wa ď -hɨpɨ Caño Jabón-ɹíʔ a ~ɟá-hat 1SG hermano mayor NPR-LOC 3SGM estar-NR ‘Mi hermano mayor está en Caño Jabón’ b. a ~beu-ɹíʔ ‘en la coronilla’ 3SGM coronilla-LOC c. ~kud-ɹíʔ ‘allá’ INVIS.S.M.D-LOC

El morfema ɹíʔ también se sufija a la forma de interrogación para locación.

(79) dedn díʔ ɹaʔ ~ba ɟed-ɟi ‘Dónde está su huerto?’ Dónde-LOC INT 2SG huerto-¿?

El morfema ~hid se sufija a los nombres ejerciendo la función de comitativo.

(80) ~dída 'hɨi- ~di 'dau-ka-~di heʔ'ɟabutu -~hid VIS.S.M.P 3SGM venir-ACT.E CANT-NEG-ACT.ERESTRmilpesillo-COM ‘El viene solo con un racimo de milpesillo’ EL NOMBRE EN NKAK 87

El morfema ~wɨd se sufija a los nombres para denotar una noción de plural, véase 81a y b. Los nombres de algunos insectos, como las abejas que viven en enjambres, llevan este sufijo como uno de los compuestos que constituyen sus nombres.

(81) a. ded-ɹáʔ wa ̍ipopɨ-~wɨd Dónde- INT1SG tíos-paternos-COL ‘¿Dónde están mis tios?’ b. wɤɟɤ-~wɨd ‘abeja sp’ ¿? COL

3. Comparación y discusión con otras lenguas de la familia lingüística Makú-Puinave Los rasgos de la composición nominal en Nɨkak muestran similitudes con las descripciones de las lenguas Hup (Epps 2005, 2007, 2008), Yu- jup (Ospina 2002; 2004-2005) y Dâw (S. Martins 2005). Entre los rasgos que comparten los sistemas de clasificación nominal de estas lenguas están: presentan términos de clase, tienen una fuerte inspiración en las raíces lexicales que describen los vegetales, y son un potente mecanismo para incorporar neologismos. En particular las lenguas Hup, Yuhup y Nɨkak están desarrollando incipientes sistemas de clasificación nominal. Mientras que las lenguas Nɨkak y Dâw presentan composiciones nominales calificativas, locativas, y descriptivas cuyas estructuras, en cuanto a las clases sintácticas de las palabras que constituyen los componentes de los compuestos y la función de uno de estos como cabeza sintáctica, son prácticamente idénticas (véase S. Martins 2005: 146-148). En contraste la descripción elaborada del Wãnsohot por Girón (2007: 183) muestra que en esta lengua los términos de clase son un grupo heterogéneo de lexemas dependientes. También se encuentran morfemas en Yujup y Hup cuyos rasgos fonológicos, morfológicos y semánticos podrían evidenciar alguna rela- ción con el Nɨkak Por ejemplo las lenguas Yujup y Nɨkak tienen el mismo morfema tɨ́t ‘hilo, cuerda’ que designa objetos filiformes. En Hup el morfema tat designa frutas comestibles y frutas redondeadas (Epps, 2005: 209), y en Nɨkak la función de denotar frutas y objetos circulares y redondos la ejerce el morfema -daʔ. En Yujup el morfema -do denota 88 DANY MAHECHA RUBIO abertura (Ospina, 2002: 215) y en Nɨkak -doʔ designa los objetos ahue- cados que ejercen función de recipientes. No obstante, las similitudes entre los sistemas de clasificación nominal entre las lenguas Hup, Yuhup y Nɨkak podrían ser también resultado de la influencia del contacto con lenguas Tukano oriental. Epps (2007, 2008) sugiere que esta relación puede ser parte de lo que está mo- tivando el desarrollo de clasificadores en el Hup. Además, las descripciones de las lenguas Tukano oriental han señalado como una de sus características la existencia de un amplio repertorio de clasificadores que se sufijan (cf. Gomez-Imbert, 2007: 420). Específicamente en el caso de los Nɨkak se ha establecido que al menos una parte los grupos ancestrales que ocuparon el actual territorio eran parientes de los Cacua, y que antes de migrar al norte tenían relaciones con los vecinos territoriales entre los cuales estaban los Cubeo27 (cf. Mahecha 2007). Por otra parte, la descripción de los casos gramaticales en Nɨkak muestra que la marca del caso acusativo, presente en otras lenguas de la familia Makú-puinave, opera de manera similar al del Hup y Dâw. En el Hup ǎn, en Nɨkak ~daʔ, y en Cacua díʔ la marca de OD es obligatorio para entidades animadas y opcional para las inanimadas, así como para pronombres y demostrativos que sean cabezas de los sintagmas nominales (cf. Epps, 2005: 224-228), mientras en el Dâw es de carácter opcional (S. Martins, 2004: 158). En el Yuhup ~díh marca al argumento paciente (…) “cuando los dos constituyentes representan argumentos de igual valor en la jerarquía de agentividad y puede haber ambigüedad. Y cuando el constituyente que representa el argumento agente presenta un valor de menor jerarquía, en una escala de + animado a –menos animado, que el argumento P paciente” (Ospina, 2002: 139). En el Wánsöhöt el caso gramatical –at también se afija a argumentos único, agente y dativo (cf. Girón, 2007: 195-196). Sin embargo, Zúñiga (2007: 225) señala, de acuerdo a información preliminar sobre estas lenguas, que el desarrollo independiente de las marcas de acusativo en las lenguas de la familia

27 Aún cuando no se ha establecido un número preciso de cognados entre las lenguas Nɨkak y Cubeo se han encontrado bastantes palabras que designan especies vegetales y animales con rasgos fonológicos similares. EL NOMBRE EN NKAK 89

Makú-Puinave se dio en una fase previa y sugiere una influencia más reciente del Tucano en la lengua Hup. Las características de la posesión son similares en Hup, Yujup, Dâw y Nɨkak Éstas distinguen entre la posesión inalienable y la alienable, el orden de los constituyentes es poseedor -poseído y se marca al poseedor (cf. Epps, 2005: 292; Ospina, 2002: 242-243; S. Martins, 2005: 158). Aquí vale señalar la existencia de sufijos posesivos en Hup, Dâw y Nɨkak, mientras que parecen ser una innovación reciente en el Yuhup (Ospina, 2002: 255). La similitud fonológica en las marcas del genitivo Hup -~dǐh, Yujup-~dәh, Dâw -~ɛɟ, Cacua ~íʔ, y Nɨkak ~íʔ, también puede ser indicador evidencia de un origen común, o un contacto previo entre estas lenguas. Respecto a los demás casos que tiene el Nɨkak, ɹéʔ, dativo, ɹíʔ locativo y ~hid comitativo, se debe mencionar que solo el Hup presenta marcas de caso de locativo e instrumental (Epss, 2005: 224). Sin embargo una comparación sobre la forma como operan estos morfemas en ambas lenguas requiere robustecer el análisis de la morfología y sintaxis del Nɨkak el cual se encuentra en curso.

4. Consideraciones finales La aproximación a las características de los nominales en Nɨkak evidencia una escasa derivación nominal; la ausencia de marcas gramaticales de género y número; procesos de composición nominal que evidencian un sistema de términos de clase y el desarrollo elemental de un sistema de clasificadores de objetos y entidades de acuerdo a su forma; y la existencia de marca de caso acusativo para el OD. Estos rasgos tipológicos de la morfología de la lengua Nɨkak son también parte de las características de las lenguas Hup, Dâw, Cacua y Yujup consideradas como afiliadas a la familia lingüística Makú-Puinave. Y a estas semejanzas se suma la existencia de una estructura sintáctica y semántica similar para denotar la relación poseedor-poseído. Así como algunas similitudes fonológicas entre los sufijos que designan estas relaciones gramaticales, las cuales ameritan un seguimiento riguroso para probar o descartar un origen común. De acuerdo a los datos presentados el Nɨkak presenta estructuras morfológicas y sintácticas en común con el Hup, el Yuhup y el Dâw, pese a que son ininteligibles y se encuentran geográficamente distan- 90 DANY MAHECHA RUBIO ciadas contemporáneamente. Parte de las cuales pueden provenir de un contacto común previo a las migraciones Tucano oriental en el área, como sugiere Zuñiga (2007) para las marcas de OD. Algunas, como el desarrollo de los sistemas de clasificación nominal pueden ser una influencia más reciente de las lenguas Tucano oriental como sugiere Epps (2005, 2007, 2008). Y otras podrían ser explicadas como la dispersión de un rasgo en el área del Noroeste amazónico, por ejemplo las similitudes con las formas de composición nominal de las lenguas Tucano oriental. Por ahora se puede señalar que los datos presentados sobre el Nɨkak en este artículo muestran semejanzas importantes con las lenguas Hup, Yuhup y Dâw que ameritan futuras investigaciones comparativas en torno a la morfología y la sintaxis de estas lenguas. Así como sobre la historia de la ocupación del área del Noroeste amazónico por parte de estos pueblos, las cuales podrían dar luces sobre el desarrollo de los procesos de gramaticalización de los sistemas de composición y clasifica- ción nominal, las marcas de caso y el uso de posposiciones en estas lenguas. Esta información es pertinente en aras de tener mayores elementos para definir la pertenencia o no de este idioma a la familia lingüística Makú Puinave.

CONVENCIONES 1SG Primera persona singular 2SG Segunda persona singular 3SGF Tercera persona singular femenina 3SGM Tercera persona singular masculina 1S.POS Posesivo primera persona singular ACT.E Actante estativo y paciente ACT.A Actante agente AUM Aumentativo CANT Cantidad COL Colectivo COM Comitativo DAT Dativo DIM Diminutivo DUR Durativo EP Época FREC Frecuentativo EL NOMBRE EN NKAK 91

INT Interrogación INVIS.S.F.D Invisible singular femenino distante INVIS.S.M.D Invisible singular masculino distante INVIS.PL Invisible plural LOC Locativo MA? Marca de aspecto sin definir MNPR Marca de nombre propio NEG Negación NPR Nombre propio NR Nominalizador OD Objeto directo PASREC Pasado reciente POS Posesivo VIS.S.M.P Visible singular masculino próximo VIS.S.M.D Visible singular masculino distante RESTR Restrictivo

Bibliografía Cabrera, Gabriel, Carlos Franky y Dany Mahecha (1994). Aportes a la etnografía de los nɨkak y su lengua - aspectos sobre fonología segmental. Tesis de grado en Antropología. Universidad Nacional de Colombia (Bogotá). Cabrera, Gabriel, Carlos Franky y Dany Mahecha (1999). Los nɨkak. Nómadas de la Amazonia colombiana. Bogotá: Unibiblos. Cathcart, Marilyn (1979). Fonología del Cacua. En: Sistemas fonológicos colombianos 4. Lomalinda: Townsend. Pp. 9-45. Derbyshire, Desmond y Doris Payne (1990). Noun Classification Systems of Amazo- nian Languages. En: D. Payne (ed.) Amazonian Linguistics Studies in Lowland South American Languages. Austin. University of Texas. Pp: 243-271. Epps, Patience (2005). A Grammar of Hup. Ph.D. thesis. University of Virginia. Epps, Patience (2007). Birth of a noun classification system: the case of Hup. En: W. Leo Wetzels (ed.) Language Endangerment and endangered languages. Lin- guistic and anthropological studies with special emphasis on the languages and cultures of the Andean-Amazonian border area. Leiden: CNWS. Pp: 107-128. Epps, Patience (2008). A Grammar of Hup. Berlin, New York: Mouton de Gruyter. Franky Carlos y Dany Mahecha (2007). Desplazamiento y perspectivas de procesos escolares para los nukak. Ponencia presentada en el “Primer Foro de Comu- nidades Indígenas del Guaviare Y desplazamiento”. Secretaria de Educación Departamental, San José del Guaviare, diciembre 17 y 18. Ms. Frawley, William (1992). Linguistics Semantics. Hillsdale: N. J. Erlabum 92 DANY MAHECHA RUBIO

Girón, Jesús Mario (2006). Léxico, historia cultural y clasificación genética del wánsöhöt (puinave). Ponencia presentada en el 52º Congreso Internacional de Americanistas, Simposio LING13, Sevilla. Girón, Jesús Mario (2007). Una gramática del wánsöhöt (puinave). Ph.D. Thesis. Vrije Universiteit (Amsterdam). Utrecht: LOT Dissertation Series 185. Givon, Talmy (2001). Syntax. Amsterdam / Philadelphia: J. Benjamins. 2 v. Gomez-Imbert, Elsa (2007). Tatuyo and Other Tukanoan Class Systems. En: L. Wet- zels (ed.) Language Endangerment and endangered languages. Linguistic and anthropological studies with special emphasis on the languages and cultures of the Andean-Amazonian border area. Leiden: CNWS. Pp: 401-428. Grinevald, Colette (2000). A morphosyntactic typology of classifiers. En: G. Senft (ed.) Systems of nominal classification. Cambridge: Cambridge University Press. Pp: 50-92. Hess Richard, Kennet Conduff y Jan Ellen Conduff (2005). Gramática Pedagógica Provisional del idioma Nukák. Bogotá: Iglesia Nuevos Horizontes. Katamba, Francis (1993). Morphology. Basingstoke: Macmillan. Lyons, John (1975). Introducción a la lingüística teórica. Barcelona: Teide. Mahecha, Dany, Carmen Fajardo, Carlos Franky y Gabriel Cabrera (1998). Los nɨukak: un mundo nómada que se extingue. Bogotá: Fundación Gaia Amazonas. Docu- mento de trabajo No. 6. 21 p. Mahecha, Dany (2005). Ms. Informe de campo sobre la situación de los nɨkak ubicados en el Resguardo del Refugio (Julio 14-Septiembre 29 de 2005). Documento de trabajo. Mahecha, Dany (2007). Los nɨkak: experiencias y aprendizajes del contacto con otras gentes. En: L. Wetzels (ed.) Language Endangerment and endangered languages. Linguistic and anthropological studies with special emphasis on the languages and cultures of the Andean-Amazonian border area. Leiden: CNWS. Pp: 91-106. Martins, Silvana Andrade (2004). Fonologia e Gramática Dâw. Ph.D. Thesis. Vrije Universiteit (Amsterdam). Utrecht: LOT Dissertation Series 98. Martins, Valteir (2005). Reconstrução Fonológica do Protomaku Oriental. Ph.D. Thesis. Vrije Universiteit (Amsterdam). Utrecht: LOT Dissertation Series 104. Mason, A. (1950). The languages of South American Indians. En: Handbook of Southamerican Indians, ED Mason. VI. Washington D.C. Morse, Nancy L., Jay K. Salser and Neva Salser (ed.) (1999). Diccionario ilustrado bilingüe: cubeo-español español-cubeo. Bogotá: Editorial Alberto Lleras Camargo. Ortiz, Sergio Elias (1965). Familias lingüísticas de Colombia. En: Prehistoria, Historia Extensa de Colombia, Tomo 3, Vol I. Bogotá: Lerner. pp. 28-165 Ospina, Ana María (2002). Les structures élémentaires du yuhup makú, langue de l'Amazonie colombienne: Morphologie et sintaxe. Thése de Doctorat. Université Paris 7-Denis Diderot (Paris). Ospina, Ana María (2004-2005). Clasificación nominal en yuhup. En: Amerindia No 29/30. Paris. Pp: 179-194. EL NOMBRE EN NKAK 93

Payne, Thomas (1997). Describing Morphosyntax. Cambridge: Cambridge University Press. Rivet, Paul y Constant Tastevi (1920). Affinités du Makú et du Puinave. En: Journal de la Societé des Américanistes 12: 69-82. Seifart, Frank (2005). The structure and use of shape-based noun classes in Miraña (North West Amazon). Ph.D. Thesis. Radboud Universiteit Nijmegen. Wage- ningen: MPI series in psycholinguistics 32. Smothermon, Jeffrey R. and Josephine H. Smothermon (ed.) (1993). Masa ye, gawa yerãca ãmara tuti (Macuna-español diccionario de 850 palabras). Bogotá: Editorial Alberto Lleras Camargo. Zúñiga, Fernando (2007). The discourse-syntax interface in Northwestern Amazonia. Differential object marking in Makú and some Tucanoan languages. En: L. Wetzels (ed.) Language Endangerment and endangered languages. Linguistic and anthropological studies with special emphasis on the languages and cultures of the Andean-Amazonian border area. Leiden: CNWS. Pp: 209-227.

Property concepts in the Cariban family: Adjectives, adverbs, and/or nouns?

Sérgio Meira (Leiden University Centre for Linguistics) and Spike Gildea (University of Oregon)

1. Introduction1 It has been a staple of typology since Dixon (1977, 1982) that the adjective class is not universal, with “property concepts” (semantic adjectives) sometimes found in a lexical class of adjectives, but sometimes in other word classes, especially verbs and nouns. Much of the descriptive typological discussion since has focused on the question of whether a “miss- ing” adjective class is a subcategory of nouns or a subcategory of verbs. The theoretical discussion, too, has focused on the ways in which adjectives are midway between nouns and verbs, e.g. Givón’s (2001) sugges-

1 Data sources for this work: Hixkaryana: Derbyshire (1965, 1979, 1985); Ma- kushi: Abbott (1991), Amodio & Pira (1996), Raposo (1997); Tiriyó: Meira (1999), field notes, Carlin (2003); Akawaio: Gildea (2005), Fox (2003). Abbre- viations used in this work: 1 = first person; 1+2 = first person dual inclusive; 1+3 = first person plural exclusive; 2 = second person; 3 = third person; 3ANA = third person anaphoric; 3R = third-person reflexive possessive (coreferential with subject); A = subject of transitive verb; ADJ = adjective; ADV = adverb; AGT = agent; AN = animate; ATTR = attributivizer (essive marker); AZR = adverbializer; C.NZR = circumstance nominalizer; COL = collective (number); COP = copula; DETR = detransitivizer; DIR = directional; EMPH = emphatic; ERG = ergative; FRUST = frustrative; HAVE = ‘having’ (predicative possession) marker; HRSY = hearsay; IMMED = immediate; IMPER = imperative; INSTR = instrumental; INTNS = intensity marker; LK = linker or relator prefix; LOC = locative; NEG = negation; NEW = new information marker; NZR = nominalizer; O = object of transitive verb; O.NZR = object nominalizer; POS = possessed form, possession marker; POT = potential adverbializer (‘good for V-ing’); PRES = present; PRPS = purpose; PST = past; PTC = particle; QNT = quantity; RECP = reciprocal; REDUP = reduplication; REIT = reiterative; S = subject of intransitive verb. 96 SÉRGIO MEIRA & SPIKE GILDEA tion that adjectives semantically fall between the time stability of nouns and the time instability of verbs; cf. also Croft’s (2002.87ff) more in- depth discussion of properties as midway between objects and actions also in terms of relationality, stativity, and gradability. However, beginning with Derbyshire (1979, 1985), most modern descriptions of Cariban languages have argued that there is no category “adjective,” but rather that property concepts are divided between the lexical categories of “noun” and “adverb” (e.g. Koehn & Koehn 1986 for Apalaí, Abbott 1991 for Makushi, Hawkins 1998 for Waiwai, Meira 1999 and Carlin 2004 for Tiriyó (aka Trio), Tavares 2005 for Wayana). One purpose of this paper is to provide a clear statement of the data and argumentation for this analysis. In his introductory article to a more recent book on this topic, Dixon (2006) reverses course, asserting that a structural word class “adjective” actually should be identifiable in every language. Of relevance to the Cariban family is his claims in §8 that what has been called the “adverb” class in Hixkaryana and Tiriyó (and by extension, other northern Cariban languages) is better labeled an adjective class, and in §9 that Abbott’s Makushi analysis misses two classes of adjectives, one which Abbott calls adverbs and the other descriptive nouns. A second purpose of this paper is to demonstrate that Dixon’s arguments for this position are unconvincing, but that nonetheless, a more careful look at the Cari- ban data yields a clear syntactic distinction between two subsets of the adverb class, one of which contains exclusively adjectival concepts. This finding leads us to consider more closely the theoretical criteria by which we might decide whether to call this latter category a subcategory of adverbs or an independent lexical category of adjectives. We begin with a brief synopsis of open word classes in northern Cariban languages (section 2), after which we offer a somewhat detailed discussion of the syntactic constructions via which property concepts are attributed to or predicated of nouns (section 3). Following this first pass at the morphosyntactic facts, we next turn to the details of the argumentation for identifying a category of adjective hiding within either the previously identified category of nouns or of adverbs (section 4). We conclude (section 5) with a call for further research on the typologically interesting question of word classes and property concepts in other Cariban languages, and in under-documented languages more generally. PROPERTY CONCEPTS IN CARIBAN 97

2. Morphosyntactic properties of word classes in Northern Cariban languages The Cariban languages so far described have presented morphologically and syntactically defined categories of verb, noun, postposition, and a host of particles and ideophones. Noun and verb are large open classes, with large numbers of underived roots and extremely productive derivational morphology. Alongside these classes is one more lexical category containing semantic adverbs and adjectives; this is a relatively small basic lexical category that becomes an open class through productive derivational morphology. Postpositions, particles, and ideophones, on the other hand, are relatively large closed classes that are not, or only mar- ginally, augmented by productive derivational morphology. In section 2.1, we lay out the fundamental inflectional morphology and syntactic behavior that distinguishes between the three open word classes; in section 2.2, we summarize the derivational morphology that enables stems of one class to become stems in the other two.

2.1 The morphosyntax of the main word classes 2.1.1 Verbs The category of verbs is identifiable in all Cariban languages by its morphological properties: there usually is a number of affixes that are characteristic only of verbs. The number of affixes may vary from language to language, but it includes at least imperative markers (usually including, besides a static, also a dynamic or ‘go do it’ imperative, plus a few unique person-marking prefixes) and class-changing affixes (adverbializ- ers: the supine or ‘purpose of motion’ form; participant nominalizers refering to A, O, S, and to a general circumstance/instrument). Gildea (1998) reconstructs (among others) the aforementioned nominalizers, which can be consistently used to identify (via their reflexes) the category of verbal root. Gildea further identifies seven different clause types across the Cariban family; one of these has a unique set of person-marking prefixes and tense-aspect-mood-number suffixes, but the other six clause types share their inflectional morphology with nouns and adverbials (postpositions). Although Gildea did not discuss imperatives, we assert that the imperative clause type is cognate in all Cariban languages described to date, and therefore it can always be used to distinguish the category of verbs from other lexical categories. Because the 98 SÉRGIO MEIRA & SPIKE GILDEA category of verbs is not at issue in the adjectives debate, we leave this as an assertion to be demonstrated in future work. Given that the unity has been called into question for both the noun and adjective classes in Northern Cariban, we offer somewhat more detail for each of these.

2.1.2 Nouns Nouns have specific morphological properties, such as markers of possession (both of possessed state and of the person of the possessor; see Table 1), number (traditionally called ‘collective’) as well as a certain number of meaning-changing elements (suffixes or particles, depending on the language) marking features such as past (‘ex-N’), diminutive (‘small N’), etc. (see Table 2). There are also class-changing affixes that convert nouns into verbs or adverbs, many of which are exclusive to nouns and can thus identify them (not illustrated here). The possessive prefixes are mostly shared with other word classes (they also occur on postpositions and certain verb forms); the meaning-changing elements are mostly exclusive to nouns, though this varies from language to language for specific elements; for each language, the ones exclusive to nouns can be used to define the category.

TIRIYÓ HIXKARYANA MAKUSHI maja ‘knife’ kanawa ‘canoe’ ewɨʔ ‘house’ 1 ji-maja(-ɾɨ) ro-kanawa-ɾɨ uj-ewɨʔ 2 ə-maja(-ɾɨ) a-kanawa-ɾɨ aj-ewɨʔ 3 i-maja(-ɾɨ) ɨ-kanawa-ɾɨ it-ewɨʔ 1+2 kɨ-maja(-ɾɨ) kɨ-kanawa-ɾɨ uj-ewɨʔ-kon 3R tɨ-maja(-ɾɨ) t-kanawa-ɾɨ t-ewɨʔ Table 1. Examples of possessive morphology PROPERTY CONCEPTS IN CARIBAN 99

S = suffix; P = particle; N = exclusive to nouns; — = non-existent. ELEMENT TIRIYÓ HIXKARYANA MAKUSHI -kon (S, N) komo (P) -kon (S, N) number i-maja -kon wewe komo pena on-kon (collective) ː ɾ ‘their knive(s)’ ‘trees’ ‘ancient ones’ - p (S, N) -mpə (S, N) tho (P/S, N)2 ɾɨʔ ɨ past / deva- u-je- p maja-mpə ho jkomo tho ɾɨʔ ɨ lued ɾ ‘my former ‘old, ex-knife’ ‘(dead) old man’ tooth’ -pisi(kə) (S, N) t ko (P) ʃ m kk (P, N) diminutive oto-pisi kana t ko ɨɾɨ ɨ ʃ (no examples) ‘a little animal’ ‘small fish’ Table 2. Examples of nouns with some meaning-changing elements (including number)

Nouns also have specific syntactic features. They can function as subjects and objects of transitive and intransitive verbs. They can also occur as arguments of postpositions (including the adverbializing particle and/or postposition pe/me, part of the copular construction described in section 3.1.2). All nouns can occur in the possessor slot of a possessive phrase, and most can also be the possessum, as illustrated in the second row of Table 3 (note the linking element j- which occurs in certain languages, like Makushi and Hixkaryana, but not in others, like Tiriyó).

TIRIYÓ HIXKARYANA MAKUSHI Postposi- sikoɾo pona ɾo-mɨn j-aka waikin pɨkɨɾɨ tional school to 1-house.POS LK-to deer after phrases ‘to the school’ ‘to my house’ ‘following the deer’ j pahko i-pata bɨɾ ekomo j-oknɨ i-san-tonon j-ewɨʔ Possessive 1.father 3-village boy LK-pet 3-mother-COLLLK-house phrases ‘my father’s village’ ‘the boy’s pet’ ‘their mother’s house’ Table 3. Examples of nouns (NPs) as objects of postpositions and in possessive phrases

2 Derbyshire distinguishes two tho’s in Hixkaryana: a suffix and a particle (1985: 245). Both are exclusive to nouns. 100 SÉRGIO MEIRA & SPIKE GILDEA

Nouns identified in Cariban languages by using the above properties are mostly semantically consistent with the expected time-stable referents. It is, however, not very difficult to encounter meanings typically translatable into Indo-European languages with adjectives.

TIRIYÓ HIXKARYANA MAKUSHI aene ‘alive (one)’ aweʃenɨ ‘wrong (one)’ aimutun ‘white (one)’ akɨpɨɾɨ ‘hard (one)’ eɲhoɾu ‘goodness, good’ anneʔ ‘stingy (one)’ iwape(tɨ) ‘deep (place)’ eɲʃemnɨ ‘not alive (one)’ aʔkiʔku ‘sweet (one)’ j mono ‘big (one)’ hoɾ e ‘big (one)’ inon ‘big (one)’ tɨpɨi ‘thick (one)’ (ɨ)khana ‘deep (place)’ moɾɨ ‘good (one)’ Table 4. Examples of nouns with property (“adjectival”) meanings

This group of ‘property nouns’ has not yet been studied in detail in any Cariban language. As far as the available data goes, there does not seem to be any important morphosyntactic difference between them and other semantic groups of nouns: besides having typically nominal roles such as subject and object, they can, as is shown in Table 5, also bear possessive morphology, co-occur with meaning-changing elements, and be arguments of postpositions.

TIRIYÓ HIXKARYANA MAKUSHI i:-mono koso j-amusu-nu possessive 3-big:POS deer.sp LK-heavy-POS morph. ‘its size’ ‘the deer’s weight’ mono-pisi

meaning- ‘a little big’ j hoɾ e-tho inon-kon ‘big ones’ changing mono-mpə ‘no longer big’ moɾɨ-kon ‘good ones’ elements ‘no longer big’ mono-ton ‘big ones’ aʔneʔ jaʔ mono pə ekeh hona with post- hot.one into big.one on sick.one to(ward) positions ‘into the hot one (wa- ‘on the big one (tree)’ ‘to(ward) the sick (one)’ ter)’ Table 5. Typically nominal behavior of ‘property nouns’

Note that these property nouns also need the essive pe/me particle when they occur as copular complements (1a; see section 3.1.2) in those lan- PROPERTY CONCEPTS IN CARIBAN 101 guages where this particle is obligatory (e.g. Makushi), and also when they occur as manner modifiers of verbal predicates (1b).

(1) a. aʔneʔ pe u-puʔpai man MAK hot.one ATTR 1-head 3.COP ‘My head is hot.’ b. a-pon ekaʔmaʔ-kɨ kaʔneʔ pe 2-clothes put.on-IMPER fast.one ATTR ‘Put on your clothes fast.’

2.1.3 Adverbs Unlike nouns and verbs, adverbs in Cariban languages do not present inflectional morphology: no person-, number- or tense-aspect-marking affixes are attested. Their only morphological possibility is nominalization (described in section 2.2 below). Syntactically, adverbs typically have the same distribution as postpositional phrases, serving as complements of the copula or as modifiers of verbal predicates.

(2) a. kuɾe tɨ-ɾə-e i-:ja TIR good PST-make-PST 3-AGT ‘He made it (=a blanket) well.’ b. təɾemine wɨtoto nɨ-tən with.song person 3S-go.PST ‘The person went/walked singing.’

j (3) a. kaɾ he ɾmahaʃa n-te-he HIX fast/strong CONTRAST 3S-go-PRES ‘This one goes/walks very fast.’ b. asako ɾo nɨ-nɨh-tʃownɨ two totally 3S-sleep-PST ‘He slept twice (= two nights).’

(4) j-aɾɨ-ʔpɨ-i-ja aminke tuna kata pɨʔ, MAK 3O-carry-PST-3A-ERG far water DIR about pɨɾanna j-aɾakkita pɨʔ sea LK-middle about ‘He (=frog) carried him (=man) far into the water, to the middle of the ocean.’ 102 SÉRGIO MEIRA & SPIKE GILDEA

When looking at the members of the adverb class in Cariban languages, one is struck by discovering many meanings typically translatable into Indo-European languages with adjectives: size (Tiriyó pija ‘small’), shape (Hixkaryana tamnoɲe ‘round’), other basic physical properties (Makushi saʔme ‘hard’), color (Tiriyó sikinme ‘black’), speed (Hixkary- ana kɨɾhɨɾaɾo ‘slow’), and even human propensities and feelings j (Hixkaryana tukhoɾ e ‘gentle, polite’).3 The same class also includes more typically adverbial meanings: time (Tiriyó kokoro ‘tomorrow’), location (Hixkaryana tano ‘here’), direction (Makushi miarɨ ‘hither’), manner (Hixkaryana huɾuhuɾhe ‘floating’, Makushi amaʔpe ‘stealth- ily’), quantity (Makushi tamɨʔnawɨɾɨ ‘all’, Tiriyó tapɨime ‘many’, Hixkaryana asako ‘two’). All of these share the morphological property that they can be nominalized, a property we turn to in the next section.

2.2 Category-changing processes: adverbs from nouns and nouns from adverbs All members of the three open classes of words in northern Cariban languages can shift categories to each of the others via productive derivational morphology. Verbs can directly become nouns or adverbs, nouns can directly become verbs or adverbs, and adverbs can directly become nouns, whereupon they can then take advantage of nominal verbalizing morphology to become verbs. Once again, we leave aside illustration of the derivational processes involving verbs, limiting our exposition to the processes that derive nouns from adverbs and adverbs from nouns.

3 Given the derivational relation found between adjectives and (usually manner) adverbs in most European languages (e.g. English happy  happily, etc.; in German, an undeclined adjective like gut ‘good’ can also be an adverb, meaning ‘well’), a fact duly pointed out in traditional grammars, the connection between adverbs and adjectives in Cariban languages is perhaps not so surprising. Most theoretical work on word classes, however, does not seem to consider it important: adjectives are mostly treated as intermediate between nouns and verbs, both in the functionalist-typological literature – e.g. Givón’s (2001) time- stability continuum – and in generative/formalist approaches – e.g. with syntactic features like +V, +N (Haegeman 1994, or, in a more nuanced way, Baker 2003), and adverbs as a heterogeneous ‘default’ category. The Cariban case described in this paper shows, we hope, that the relation between adjectives and adverbs deserves more attention. PROPERTY CONCEPTS IN CARIBAN 103

First, we discuss how nouns become adverbs. There are basically two processes, one based on a prefix t- plus a range of similar and probably historically related suffixes (-ke, -ne, -ɾe, -je, -e...) forming synchronic circumfixes, and the other on a suffix and/or particle (the essive pe/me).4 Some reseachers see semantic differences between these morphemes (see Carlin 2004: 470ff); what is clear, however, is that they define morphological subclasses of adverbs.

TIRIYÓ HIXKARYANA MAKUSHI tɨ-maja-ke ‘having a knife’ t-ot-ke ‘having meat food’ (maja ‘knife’) (otɨ ‘meat food’) tɨ-pana-ke ‘having an ear’ t-amta-ke ‘wide’ (pana ‘ear’) (amta ‘width’) it-ewɨʔ-ke‘having a house’ tɨ-pana-e ‘able to hear’ tɨ-ɾwo-ɲe‘talking, able to talk’ (ewɨʔ ‘house’) tɨ-pɨ-je ‘married, wifed’ (ɾwo ‘talk, language’) noɾa pe ‘dirty’ j (pɨ ‘wife’) t-ahoʃe-ɾ e ‘strong’ (noɾa ‘dirt’) tɨ-katɨ-ne ‘fat, fatty’ (ahoʃe ‘strength’) moɾɨ pe ‘good’ (i-katɨ ‘fat’) tɨ-hɾo-je ‘by, on foot’ (moɾɨ ‘good one’) tɨ-pəmu-ɾe ‘blossoming’ (hɾo ‘foot’) sɨɾɨɾɨ pe ‘now, today’ (ipəmu ‘flower’) ekeh me ‘sick, ill’ (sɨɾɨɾɨ ‘this one’) wəɾi me ‘feminine, female’ (ekehɨ ‘ill, dead one’) (wəɾi ‘woman’) toto me‘human, like a human’ tɨpɨi me ‘thick’ (toto ‘human being’) (tɨpɨi ‘thickness’) hawana me ‘as visitors, visiting’ kumu me ‘brownish’ (hawana ‘visitor’) (kumu ‘palm sp.’) Table 6. A few adverbialized nouns

The syntactic origin of these processes is evident: pe/me is still a particle or postposition (depending on the specific language, or even the specific noun in a specific language) that can adverbialize full noun phrases in all

4 Further adverbializing processes can be found in negation: negative suffixes (depending on the language, -hɾa, -mɾa, -ːɾa, -pɾa, -mna, -nna, etc.) also create negative adverbs; these can also be nominalized. Since their specificities are not relevant for the topic at hand, derived negative adverbs will not be further discussed in this paper. 104 SÉRGIO MEIRA & SPIKE GILDEA

Cariban languages so far described. The t- -ke circumfix is clearly relat- able to the third-person reflexive possessive prefix t- and the instrumental postposition ke (‘with the subject’s own N’); the other suffixes would have come from interactions between stem-final consonants and a single adverbializing suffix, possibly -je (also attested as a perlative marker on postpositions). The synchronic differences between these sources and their constructions are, however, significant enough to warrant a different treatment. For instance, the source elements t- ‘3R’ and ke ‘INSTR’ can still co-occur with their source meaning (‘with/using the subject’s own N’), contrasting with the meaning of t- -ke (‘having N’), as seen in the Tiriyó and Hixkaryana examples below. Note that the t- ‘3R’ prefix occurs on a possessed stem in the source construction, as the suffix -ɾɨ ‘POS’ makes clear in Hixkaryana; and even in Tiriyó we can still observe a reflex of this prefix in the form of vowel length (tɨ-maja-ke and tɨ-majaː=ke form a minimal pair).

(5) a. tɨ-maja-ke nai TIR AZR-knife-HAVE 3.COP ‘S/He has a knife.’ b. tɨ-majaː=ke n-ahkəː-jan 3R-knife.POS=INSTR 3A-cut-PRES ‘S/He is cutting it with his/her own knife.’

(6) a. t-amo-ke HIX AZR-hand-HAVE ‘having a hand’ j b. t-amo-ɾɨ ke ɾma n-ekaɾ me-konɨ heno 3R-hand-POS INSTR PTC 3A-tell-PST PTC ‘He said it with his own hands (without speaking).’

For pe/me, one observes a continuum ranging from cases with predictable meaning (pe/me = ‘as’, ‘like’; Tiriyó taɾəno me ‘as, like a Tiriyó’, from taɾəno ‘Tiriyó (person)’) via cases with more specific meanings becoming frequent (see Tiriyó kumu me ‘brownish’ from Table 6 above, a color, not simply ‘as, like a certain species of palm tree’, though the latter meaning still remains possible) to cases in which there is only one specific meaning, the source word often being no longer synchronically PROPERTY CONCEPTS IN CARIBAN 105 available (Tiriyó wapəme ‘bluish’, sikinme ‘black’; *wapə and *sikin are not attested as nouns). Just as all nouns can be adverbialized, all adverbs can also become nouns. This is usually done with several different suffixes that define morphological subclasses (e.g. Tiriyó -no, -to, -mɨ; Hixkaryana -no, -mɨ; Makushi -n, -nan). Interestingly, the subclass defined by the suffix -mɨ contains only adverbs derived with the prefix t- and its various possible co-suffixes (-ke, -je, -ne, -nje, -ɾe, -se, -e, -so, etc.).

TIRIYÓ HIXKARYANA MAKUSHI kuɾe ‘good’  kuɾa-no ohʃe ‘good’  ohʃa-no kuɾeʔne ‘big’  kuɾeʔna-n j j pija ‘small’  pija-n kaɾ he ‘strong’  kaɾ he-no tiwin ‘one’  tiwin-nan əːkənə ‘two’  əːkənə-n asako ‘two’  asako-no pəeɾa ‘stupid’  pəeɾa-to omeɾoɾo ‘all’  omeɾoɾo-no teːɾeʔmase ‘visible’  əːseːnə ‘ill’  əːseːnə-to oɾoke ‘yesterday’oɾoke-no teːɾeʔmase -n ‘one that can əiɾe ‘wild’  əiɾa-to jake ‘many’  jake-no be seen’ 5 j j taːmiːɾe ‘red’  taːmiːɾe-n tutʃuɾ e ‘red’  tutʃuɾ e-mɨ təːnakəe ‘liar’  təːnakəe-n tɨhje ‘married’  tɨhje-mɨ tɨɾetɨke ‘horned’  tɨɾetɨke-n tonoso ‘edible’  tonoso-mɨ Table 7. A few nominalized forms

The meaning of the resulting nominalizations is ambiguous between that of an entity having that property (usually when not possessed: 7a), or the property itself (when possessed: 7b-c).

5 In Tiriyó, and perhaps also Makushi, the nominalizing suffixes -mɨ and -no tend to reduce to -n word-finally. If CCV-initial suffixes or clitics follow the word, the difference between these two nominalizers is maintained. With the nominal past suffix -mpə, for instance, pijan ‘small one’ and taːmiːɾen ‘red one’ become pijano-mpə ‘the one which was small’ and taːmiːɾemɨ-mpə ‘the one which was red’. Note also that -mɨ, unlike -no, does not cause a stem-final e to change to a. 106 SÉRGIO MEIRA & SPIKE GILDEA

TIR (7) a. iɾə mao tɨw-əe-se kawə-no-ton, ma, soni, that at.time PST-come-PST high-NZR-COL NEW vulture.sp watəikə, akaɾaman... vulture.sp vulture.sp ‘At that moment came all the high ones (= the ones who live up high), the soni vulture, the watәikә vulture, the akaraman vulture...’ b. eːkaːɾə nai, kanawaimə i-kawə-no nono pəe? how 3.COP airplane 3-high-NZR ground from ‘How high from the ground is (that) airplaine flying?’ (Lit. How is that airplane’s height from the ground?) c. kananama-n, i-siɾiɾima-no yellow-NZR 3-blue-NZR ‘It is yellowish blue.’ (Lit. Its blue is yellow.)

HIX j (8) a. kaɾ he-no kaʃe mak tɨ n-eh-ʃakonɨ ha strong-NZR because PTC HRSY 3S-COP-PST INTNS ‘It was because (he was a) strong (man).’ b. ɨ-matkɨ-ɾɨ kaw n-a-ha, un metɾu me n-a-ha, 3-tail-POS long 3S-COP-PRES one meter ATTR3S-COP-PRES ɨ-matkɨ-ɾɨ kawo-no-nɨ 3-tail-POS long-NZR-POS ‘Its tail is long, it is one meter, its tail’s length.’

Given that each word class can transition to the other via derivational morphology, the logical possibility arises that a single root could make the transition more than once, e.g., that a noun could be adverbialized, then renominalized (see 9-11 below), and perhaps then even re-adverbialized (not yet attested).

TIR (9) ma, iɾə mao tɨw-əe-se ikɨː-jamo ma-n ton NEW this TEMP PST-come-PST 3:y.br-COL ATTR-NZR COL ‘Well, then came those who are her younger brothers.’ PROPERTY CONCEPTS IN CARIBAN 107

HIX (10) onokna komo ɨ-mʃek-rɨ me-no-hnɨ jak mokɾo ha creature COL 3-child-POS ATTR-NZR-NEG PTC that.one PTC ‘That one (the causer of our problems) is not just the offspring of animals.’

(11) ajawɨ pa-n mɨːkɨɾɨ MAK madness ATTR-NZR that.one ‘That one is a madman.’

Finally, we identify the property of reduplication in Tiriyó (and possibly also in the neighboring languages Apalaí and Wayana, but apparently not in any other member of the Cariban family) that distinguishes adverbs and verbs (and their nominalizations) as distinct from underived nouns. The reduplicated adverbials add the meaning ‘all around’, ‘all over the place’, ‘to all’ (13a-b); the reduplicated nominalized forms usually have the meaning of ‘many entities of the same kind (scattered all around)’ (12, 13c).

TIR6 (12) kuɾe kawə tɨkoɾoːje ‘good’ ‘high, tall’ ‘white’    kuɾa-no kawə-no tɨkoɾoːje-n ‘good one’ ‘tall one’ ‘white one’    kuɾa-kuɾa-no kawə-kawə-no tɨko-tɨkoɾoːje-n ‘many good things’ ‘many tall people/things’ ‘many white things’

6 The change of final vowel caused by nominalization in kuɾe  kuɾa-no is also present in the reduplicant, which shows that reduplication logically follows nominalization. In general, reduplication is a widespread post-lexical process in Tiriyó, affecting e.g. fully inflected verbs, with the reduplicant including both inflectional prefixes and part of the verb stem (e.g. w-ekaɾama ‘I gave it’, weka-w-ekaɾama ‘I gave it (many times, or to many people)’; for details, see Meira 2000). 108 SÉRGIO MEIRA & SPIKE GILDEA

(13) a. aɾe-aɾehtə nai TIR REDUP-3.on.top 3.COP ‘(the roots) are all linked to each other’ b. epo-epo-ne ɾəken REDUP-enough-COL only ‘It will be enough for everybody.’ (lit. on top of each other) c. təinken pa nai əɾinə, tɨpa-tɨpanake-n once again 3.COP clay.pot REDUP-having.ears-NZR ‘Again there is a clay pot, one with several ‘ears’ (= handles).’

In sum, the lexical categories of noun and adverb are robust: nouns present inflectional morphology whereas adverbs do not, the syntactic distribution of the two classes do not overlap, and rich derivational morphology allows for free passage of stems between the two categories.

3. The grammar of property predication and modification Property concepts modify participants in two different ways: as predicates (‘the man is big’) and as attributive modifiers (‘the big man’), typically inside the NP headed by the modified noun. Given that property concepts are divided between the lexical categories of noun and adverb, one might guess that each word class plays a somewhat different role in nominal modification. This is the case: typically, predicate modification is carried out by means of adverbs and attributive modification by nouns. Since the semantic makeup of lexical categories in Cariban languages has not so far been studied in detail, we cannot fully determine if there are family-wide patterns in which certain semantic subcategories of property concepts align with the syntactic categories of noun or adverb (but see section 5 for some tendencies). Note, however, that which lexical category a given root falls into is relatively unimportant: as we have already seen, all adverb roots can readily become noun stems and all noun roots can readily become either adverb stems or arguments of postpositional phrases which are syntactically indistinguishable from adverbs. In section 3.1, we explore the grammar of nonverbal predicates, and in section 3.2, the grammar of attributive modification. In section 3.3, we turn our attention to another typical use of adjectives: compara- PROPERTY CONCEPTS IN CARIBAN 109 tive constructions (‘the man is bigger than the breadbox’).

3.1 Nonverbal predicates Nonverbal predicates in most Cariban languages have not been thor- oughly described in terms of either grammar or semantics. Most grammars include examples of different nonverbal predicate constructions without further details about the semantic difference between these constructions or the classes of predicates that might (or might not) be compatible with each construction. Payne (1997) identified six major functions of nonverbal predication: equative, proper inclusion, attributive, locative, existential, and possession. All six have been illustrated from the Akawaio language (Gildea 2005), coded via two different constructions: nonverbal predicates without a copula do the first three functions plus possession, whereas those with a copula do all six functions. Both constructions may be used to predicate property concepts of a subject, the former requiring a nominal predicate (14a) and the latter an adverbial predicate (14b). In other words, the Akawaio copular construction has only adverbial complements, whereas the non-copular construction has only nominal complements.

(14) a. juwaŋ kɨɾə-ɾə b. juwaŋ be maŋ hunger 3AN-EMPH hunger ATTR 3.COP.IMMED ‘He’s hungry (always).’ ‘He’s hungry (now; a fact).’

The semantic difference between these examples (as indicated in the glosses) is consistent with Pustet’s (2003) finding that the absence of the copula correlates with stability (essence, permanence), whereas its presence suggests instability (temporariness, contingence). Note that the property concept ‘hungry’ is a noun in Akawaio, so in order to occur in a copular predicate, it must be marked with an adverbializing morpheme, in this case, the attributive or essive marker be (14b). In contrast, the Akawaio color term aimuʔne ‘white’ is an adverb and shows the opposite pattern: it occurs in its basic form in the copular predicate (15a), but must be nominalized to serve as the noncopular predicate (15b). A similar pattern is found with Akawaio derived adverbials like tuzubaraige ‘having a cutlass’: the adverbial form occurs in the copular predicate (15d) and a re-nominalized form in the noncopular predicate (15c). 110 SÉRGIO MEIRA & SPIKE GILDEA

(15) a. aimuʔne Ø-eʔ-tai white 1S-COP-PST ‘I was white...’ b. taːne seɾəbe taːne juweːi Ø-eʒi but now but red 1S-COP.PRES ‘...but now, I am red.’ c. tu-zubaɾa-iɡe-naŋ kɨɾə-ɾə AZR-cutlass-HAVE-NZR 3AN-EMPH ‘He owns a cutlass.’ (Lit. ‘He is a cutlassed one’; it makes him who he is) d. tu-zubaɾa-iɡe Ø-eʔ-aik AZR-cutlass-HAVE 1S-COP-PRES ‘I have a cutlass.’

In sections 3.1.1 to 3.1.3, we examine examples of these constructions from the three Cariban languages mentioned by Dixon: Tiriyó, Hixkary- ana, and Makushi. We first look at the non-copular construction with nominal complements (3.1.1) and then at the copular construction with adverbial complements (3.1.2). In the final subsection, we discuss the “mixed” constructions found in Tiriyó and Hixkaryana (but not in Akawaio or in Makushi), in which nouns occur in copular predicates and adverbs in noncopular predicates (3.1.3).

3.1.1 The non-copular construction In all three languages, we find a non-copular construction parallel to that seen in Akawaio, with nouns or nominalizations serving as the predicate. Text examples were not difficult to find in all three languages, usually showing the semantics of stability expected for this construction, as can be seen below (the non-copular constructions are underlined):

(16) a. atɨtoːme iɾə apo n-ka-n ji-pɨ, TIR why this like 3S-say-PRES 1-wife kuɾa-no ji-pɨ i-jomi, tɨː-ka-e good-NZR 1-wife 3-language PST-say-PST ‘ “Why is my wife talking like this? (Usually) her language is good (= has no accent),” (he) said.’ PROPERTY CONCEPTS IN CARIBAN 111

b. owa, mono, tɨː-ka-e, mono jaɾawaɾe NEG big.one PST-say-PST big.one Yaraware ‘No, he is big,’ (he) said, ‘Yaraware is big.’

(17) a. mojoɾo-no mokjamo ha, woɾɨskomo heno ha, HIX elsewhere-NZR those.AN INTNS woman QNT INTNS j Ø-ke-konɨ hatɨ, ʃaɾ emna ha 3S-say-PST HRSY otter INTNS ‘“They are the ones far away, the women,” said the otter.’ b. ɨto-no-tho uɾo there-NZR-PST 1 ‘I am the one who was, used to be there.’

MAK (18) a. miaɾɨ toʔ wanɨ-ʔpɨ, it-un saʔne enkaɾuʔna-n thither 3COL COP-PST 3-father PITY blind-NZR ‘They were there, (and) his father was blind.’ b. mɨːkɨɾɨ teseuɾɨno tusawa that.AN third.one chief ‘That one was the third chief.’ (Part of a list of all past chiefs of a certain village.)

3.1.2 The copular construction In all three languages we find a copular construction that takes as its predicate an adverbial complement, whether a simple or derived adverb or another type of adverbial, such as a postpositional phrase (including here also nominals marked with the attributivizer or essive morpheme, be in Akawaio, pe in Makushi, me in Hixkaryana and Tiriyó). And again, it is a simple matter to find examples like the following in texts.

(19) a. moːɾaimə nai, mono me TIR armadillo 3.COP big.one ATTR ‘The armadillo is big.’ b. moːɾaimə maɾə ɾe kaːɾi me t-ee-se armadillo also FRUST force ATTR PST-COP-PST ‘The armadillo is also strong, but in vain.’ c. ma, kuɾe nai seɾə, uɾu-tə nai, wət-uɾu-to NEW good 3.COP this advise-POT 3.COP DETR-advise-NZR 112 SÉRGIO MEIRA & SPIKE GILDEA

apo ɾo pa nai like EMPH REIT 3.COP ‘Well, this is good, this is good for advising, it is just like advising (=good education).’ (An old man talking about a recently published book of traditional stories in Tiriyó.)

HIX (20) a. ohʃe w-eh-ʃaha b. t-ono-so n-a-ha kjokjo good 1S-COP-PRES AZR-eat-AZR 3S-COP-PRES parrot ‘I am well.’ ‘Parrot can be eaten.’ (Lit. ‘Parrot is edible.’) c. toto me n-eh-ʃakonɨ amɲehɾa haka, kuɾumu person ATTR 3S-COP-PST long.ago then buzzard ‘The buzzard used to be a man at that time, long ago.’

(21) a. tiwin wei toʔ wanɨ-ʔpɨ emiʔne MAK one day 3COL COP-PST hungry ‘One day they were hungry.’ b. innapeɾɨ kaʔneʔ pe nai really fast.one ATTR 2.COP ‘It is really true, you are fast.’ c. kusan pe i-puʔpai siʔpo wanɨ-ʔpɨ length ATTR 3-head hair COP-PST ‘His head hair was very long.’

3.1.3 “Mixed” constructions Having illustrated the constructions common to all four languages, let us now look at “mixed” patterns which are not found in all these languages and are less frequent even in the languages in which they are found. At this time, we cannot speculate about the meaning differences associated with these mixed, and possibly innovative, constructions (see section 4.3). Cases of a predicate adverbial occurring in the non-copular construction have been found in Tiriyó, where they are actually not infre- quent: PROPERTY CONCEPTS IN CARIBAN 113

(22) a. pahko kuɾe, tɨː-ka-e TIR 1:father good PST-say-PST ‘ “My father is well,” (he) said.’ b. ma, anja i-moitɨ əːseːnə, wəɾi nəɾə, winihpə NEW 1+3 3-relative ill woman 3ANA Winihpë eka, mɨnome 3.name pregnant ‘Well, our relative is sick, she is a woman, her name is Winihpë, she is pregnant.’

Less frequent, but still attested (in Tiriyó, Hixkaryana, and Makushi), are cases of copular constructions with non-adverbial complements, e.g. a nominal without the adverbializer:

(23) tɨ-ːna-ke, kuɾa-no n-ai, i-ːnan me, TIR ADV-flute-PROP beautiful-NZR 3.COP 3-flute ATTR iɾə-npə pəe tɨwəɾən this-PST from other ‘There were flutes (in the show), it was beautiful, like flutes, and then there was another (type of flute).’

(24) a. ohʃa-no haɾha mokjamo n-eh-tʃownɨ ha HIX good-NZR back.again those.AN 3S-COP-PST INTNS ‘Those people became good people again.’ b. moɾo-no mokɾo n-ah-ko ɾo-hetʃe, Ø-ke-konɨ hatɨ there-NZR that.AN 3S-COP-PST 1-wife 3S-say-PST HRSY ‘ “The one who is over there, that one has become my wife,” (he) said.’

(25) toʔ saːkɨɾɨɾo-no a-wanɨ-ʔpɨ ʒeɾonimu MAK 3COL four-NZR 3S-COP-PST Jeronimo ‘The fourth one (chief) was Jeronimo.’

This quick overview of attributive predicates in three Cariban languages shows that the two most frequent constructions present a clear semantic difference (stability/instability), coupled with distributional differences between nouns and adverbs (the copular construction requires adverbial complements, the non-copular construction nominal ones). This clarity 114 SÉRGIO MEIRA & SPIKE GILDEA is, however, called into question by the existence of “mixed” cases, in which nominals and adverbials each occur in the construction characteristic of the other, with unclear semantic consequences (but see section 4.3). We now turn to attributive modification, which uses only nouns.

3.2 Nominal modification In the area of attributive nominal modification (the big man), the property concept must be a noun, which occurs in a construction that is not clearly grammaticalized like the noun phrases in more familiar languages: in some languages (e.g., Tiriyó), property nouns may precede the modified (26a), follow the modified (26c-d), or even be non-contiguous (26b);7 in other languages (Hixkaryana), a pause always seems to occur between modified and modifier (27a-b). In yet others, ordering constraints seem to be emerging (note that, in the Makushi examples 28a-b, modifier nouns precede the modified noun, though, as far as we know, there are no further phonological or morphosyntactic properties of a phrasal constituent). We follow Payne (1993) in interpreting this flexibil- ity, when present, as evidence for a more ‘appositional’ strategy, with juxtaposed nominals (including possible property nominals) prag- matically assumed to refer to the same real-world entity without necessarily being joined in a single syntactic constituent.

(26) a. oːni po nai, kuɾa-no epeɾu, əmɨja-n epeɾu TIR that LOC 3.COP good-NZR fruit soft-NZR fruit maɾə, tɨː-ka-e too PST-say-PST ‘ “Over there (there) are good fruits, soft fruits too,” (he) said.’ b. kuɾe iɾə j-ekeima-to ə-:ja, kuɾa-no good this 1-do.evil-C.NZR 2-AGT good-NZR w-ekeima ə-emi 1A-do.evil.PST 2-daughter ‘It is OK that you want to do evil to me, (for) I have done evil to your good daughter.’

7 We assume the observed order variation reflects some pragmatic distinction. PROPERTY CONCEPTS IN CARIBAN 115

c. konopo mono n-eː-jan rain big.one 3S-come-PRES ‘Big (= a lot of) rain is coming.’ d. seɾə po nai pɨː mono, tɨː-ka-e this LOC 3.COP mount big.one PST-say-PST ‘ “Here there’s a big mountain,” (he) said.’

j (27) a. Ø-to-tʃowɨ bɨɾ ekomo komo, asako-n komo HIX 3S-go-PST boy COL two-NZR COL ‘Two boys went.’ Also: ‘Two of the boys went.’ (Lit. The boys went, the two.) b. hɨː... ka-je hatɨ, wajamo, wosɨ all.right say-PST HRSY turtle woman ‘ “All right...” said the turtle, the woman/female (turtle).’

(28) a. kaiwan kuɾeʔna-n moɾɨ paːka MAK fat.one big-NZR good.one cow ‘A good cow is big and fat.’ b. ʒezus-ja uj-aɾɨ-toʔpe-nɨkon kaʔ pona, moɾɨ pata ja, Jesus-ERG 1O-take-PRPS-COL sky DIR good place DIR moɾɨ tɨ-n-konaka-ʔpɨ ja good 3R-O.NZR-make-PST DIR ‘Jesus will take us all to heaven, to the good place, to the good (place) that he made.’

In contrast to the case of predication, where both Hixkaryana and Tiriyó presented multiple exceptions to the restriction of nouns in noncopular predicates and adverbs in copular predicates, we have encountered only one exceptional case of an adverb modifying a noun attributively in any of the Cariban languages we have worked on: Abbott (1990: 89) illustrates the claim that sometimes in Makushi, numbers can directly modify nouns with the example in (29).

MAK (29) t-ekɨn-kon jaɾɨ-ʔpɨ-i-ja asakɨʔne maikan-jamɨ 3R-pet-COLL take-PAST-3-ERG two foxes-PL ‘He took his own animals, two foxes.’ 116 SÉRGIO MEIRA & SPIKE GILDEA

To summarize, the grammar of nominal modification does not fit the expected prototype of a dependent modifier internal to a noun phrase headed by the modified noun. The lack of clear evidence for a NP constituent is common in the family, especially evidence for a syntactic connection to mirror the semantic connection between the modifying and modified nouns. The next section shows a similar lack of evidence for an entrenched construction.

3.3 Comparative constructions Comparative constructions, when available, are an important tool for identifying and defining an adjectival class. In the case of Cariban languages, there usually are no grammaticalized comparative constructions, but simply specific morphemes (normally postpositions) with meanings such as ‘more than’, ‘stronger/bigger than’, ‘superior to’, ‘too much for’, etc. These postpositions often still retain a locative meaning in other contexts (e.g. Hixkaryana oho, also ‘above’). The examples below illustrate the use of such morphemes in a more typically comparative context (with a property as the term of comparison: 29a, 30a, 31a-b), as well as in other contexts (occurring by themselves: 29b, 30b-c; or with an inflected or nominalized verb as the term of comparison: 29c, 31c). Note that examples with a term of comparison are much less frequent than examples without them – simply ‘I am more than you’, with the pertinent property either inferable from context or irrelevant. Even when a term of comparison is present, the pauses (marked as commas) between it and the ‘comparative’ postposition stress the looseness of their syntactic bond. This supports the claim that there are no really grammaticalized comparative constructions (just as there is no really grammaticalized construction for nominal modification; see previous section): the ‘comparative postpositional phrases’ are perhaps better seen as simple adjuncts, similar to other postpositional phrases (and maybe only a metaphorical step removed from locative postpositional phrases). Comparative sentences expressing equality (as good as) are even less frequent than their superior- ity/inferiority counterparts, but they also seem to support this claim: the examples found in the corpora use an adverbial or particle meaning ‘equally’ or ‘the same’ (also found elsewhere with the same meaning) without any construction-specific features (29d). PROPERTY CONCEPTS IN CARIBAN 117

TIR (29) a. kuɾe nai məe, tɨː-ka-e, aipɨ me, [anja i-wae], well 3.COP this.AN PST-say-PST speed ATTR 1+3 3-more kɨ-wae-ne, tɨː-ka-e 1+2 -more-COL PST-say-PST ‘ “This one is good,”, (they) said, “he is faster than us,” (they) said. b. ji-wae manae, iwa, ji-wae manae, tɨː-ka-e, 1-more 2.COP iguana 1-more 2.COP PST-say-PST tɨw-əːsina-e PST-cry-PST ‘ “You’re more than me, iguana, you’re more than me,” (= stronger, more powerful) (he = jaguar) said, (he) cried.’ c. menjaːɾə m-əːs-apəkəma-e ji-wae, tɨː-ka-e now 2S-DETR-suffer-PRES 1-more PST-say-PST ‘ “Now you are suffering more than me,” (he) said.’ (= You had made me suffer before, I am now taking my revenge.) d. pai, pənjeke, əis-apo ɾo kuɾe, k-otɨ me tapir peccary RECP-like EMPH good, 1+2-meat/game ATTR ‘Tapir is as good meat/game as peccary.’ (Lit. ‘Tapir, peccary, like each other they are good, as our meat/game.’)

HIX j (30) a. [kajkusu j-oho] n-a-ha, ɨ-hoɾ me-no-nɨ, ɾokmo dog LK-more 3S-COP-PRES 3-big-NZR-POS wolf ‘He is bigger than a dog, the wolf.’ b. oj-oho n-a-ha ha, 2-more-NZR that.IN INTNS ‘That (is) too much for you.’ c. [ɾo-muɾu j-osnaka] n-a-ha, o-muɾu 1-son LK-less 3S-COP-PRES 2-son ‘Your son is smaller than mine.’ (also: less important than mine) j j d. kaɾ he [o-to-nɨ-ɾ j-oho], kaɾ he ɨ-te-he fast 2-go-NZR-POS LK-more fast 1S-go-PRES ‘I will run faster than you.’ (Lit. ‘Fast, more than your going, I will go fast.’) 118 SÉRGIO MEIRA & SPIKE GILDEA

(31) a. kusan pe mɨːkɨɾɨ wanɨ, [tɨ-ɾui j-entai] MAK tall.one ATTR that.AN COP 3-o.br. LK-more ‘He is taller than his older brother.’ (Lit. ‘He is tall, more than his older brother.’) b. meɾuntɨ paːpa [tamɨʔnawɨɾo-n-kon j-entai-non] strength god all-NZR-COL LK-more-NLZR ‘God is stronger than everyone.’ c. uj-eʔma-kɨ [ tiaɾon-kon j-eʔma-Ø-ja j-entai ] 1O-pay-IMPER other-COL LK-pay-3-ERG LK-more ‘Pay me more than you paid the others.’

This concludes our presentation of the basic grammar of words coding property concepts, as seen through the eyes of the authors of the grammars of Hixkaryana, Makushi, and Tiriyó. We now turn to the question of whether a more perspicacious analysis of these patterns might not reveal an adjective category hiding in one or both of the categories of nouns and verbs.

4. Should we separate a class of adjectives from adverbs and/or nouns? In the very first modern description of a Cariban language (Hoff 1968, on the Carib language of Suriname, or Kari’nja), the label “adjective” was used for the class analogous to what we have been calling adverbs in this paper. Hoff (to appear) further argues for the label ‘verbal adjectives’ to describe derived forms that the analyses above would consider a mix of nominalizations and derived adverbs; Courtz (2008), working on the same language as Hoff, also prefers to describe adjectives. Coming from a different perspective, Dixon (2006) considers all of what we have called “adverbs” to be better labeled “adjectives,” and in Makushi, he further considers the seven property nouns listed in Abbott (1991: 88) to constitute a small category that he calls adjective2. In this section, we first examine the reasoning behind the initial proposal to call this category “adverbs” (from Derbyshire 1979, 1985), and we consider how well this reasoning might extend to the cognate category in the other northern Cariban languages (4.1). We then consider Dixon’s (2006) critique of this analysis, seeking to test the reliability and PROPERTY CONCEPTS IN CARIBAN 119 validity of his arguments for the alternative analysis (4.2). One crucial element in question will be the role of semantic evidence for category membership. Following this, we construct a more fine-grained analysis of the semantic, syntactic, and morphological sub-categories of the adverb class, showing that there is indeed a syntactic subclass of adverbs that contains only property concept meanings, and which might therefore be considered as a candidate for a distinct adjective category (4.3). We find no support for the hypothesis that a subset of nouns should constitute a distinct adjective category in any of the languages in question.

4.1 The adverb analysis Derbyshire (1979) was the first to propose that there was no need for a category of adjectives in Hixkaryana; he recognized the existence of property concept nouns and adverbs, an analysis which was subsequently adopted in most of the descriptions that followed (Koehn & Koehn 1986 for Apalaí, Abbott 1990 for Makushi, Hawkins 1998 for Waiwai, Meira 1999 and Carlin 2004 for Tiriyó/Trio, and Tavares 2005 for Wayana). Derbyshire first demonstrated that each category had a number of morphsyntactic properties that united its membership in a single structural category. The noun category was sufficiently clear semantically as to require no further justification. However, the adverb category was truly heterogeneous semantically, containing adverbial and adjectival meanings. He then relied on two criteria to decide on the label adverb rather than adjective. First, he estimated that most words in this category (especially most monomorphemic words) had clearly adverbial, not adjectival meanings: “all but a few members of this large class pertain to semantic types usually associated with adverbs” (1985:13). Second, he argued that the syntactic properties of the members of this category were closer to those of adverbs than to those of adjectives: “their syntactic properties correlate with (modifying or sentence) adverbials” (1985:14). These properties were basically the ones described in sections 2 and 3 above. In considering the theoretical validity of these arguments, we begin with the unquestioned premise in descriptive linguistics that language- internal categories must be determined based on language-internal patterns. Without question, Derbyshire has followed this criterion in diag- nosing his two categories. The use of the argument of “semantic majority” for deciding to la- 120 SÉRGIO MEIRA & SPIKE GILDEA bel a class, however, is criticizable, for several reasons: (a) the meanings in question may sometimes be difficult to distinguish (as is the case between “adverbial meanings” and “adjectival meanings;” e.g., hard, fast, etc.); (b) derived and underived members of the category may give different results (the majority of meanings of underived terms is adverbial, but the majority of the meanings of derived terms may not be, due to productive class-changing processes that could, e.g., derive new “adjectival” meanings from any given noun or verb); (c) there are different types of “majority” (should one count the number of “adjectival” vs. “adverbial” meanings in a given standard wordlist, or look at the occurrence of tokens of these meanings in a representative corpus of texts?). These same objections could be raised against the analyses proposed by Hoff (1968, to appear) and Courtz (2008), in which the label “adjective” is used without any argumentation whatsoever, either against the adverb analysis or in favor of a competing adjective analysis. The analysis thus appears to be based entirely on semantics. Although we do consider semantics to be relevant to the task of naming any relatively homogeneous category identified through morphosyntactic tests, given the co-existence of the “adjectival” meanings with all the most frequent and most typical “adverbial” meanings (e.g., manner (well), place (here), time (now), etc.), we do not find it compelling in this case. In contrast, we find the syntactic argument substantially more compelling: the category shares syntactic distributional properties with postpositional phrases, including (i) the ability to occur as the predicate of a copular clause, (ii) the ability to modify a verbal predicate, and (iii) the need to be nominalized in order to attributively modify nouns.

4.2 The proposed categories of adjective1 and adjective2 We turn now to Dixon’s proposal, which basically states (2006: 28-30) that the entire class of words here termed adverbs would be more felicitously analyzed as forming an adjective class with some members having adverbial meanings. His morphosyntactic arguments are (a) that “Euro- centrism” led Derbyshire and Meira to believe that “words which cannot function as modifier within an NP (except in the nominalized form) may appear un-adjective-like”, and (b) that the label adverb “is scarcely appropriate; an adverb cannot normally occur as copula complement.” The first argument is actually a claim about the motives of the ana- lysts, and one with which it is difficult to agree, given the amount of care PROPERTY CONCEPTS IN CARIBAN 121 and detail given to morphosyntactic arguments in their publications. Derbyshire, the first to use the “adverb” label, did not seem concerned by the lack of modifying uses for words of this category (a fact which he did not even explicitly mention), but rather by the syntactic roles typical of adverbials and postpositional phrases. The second argument fares little better under even casual inspection, as Dixon himself observes further down the page: “It is perhaps not surprising that the Carib adjective class, which functions only as copula complement and as adverb, should include words of place and time which are typically coded as adverbs in other languages.” And indeed, a quick review of the adverbs listed by Derbyshire (1985) reveals words that readily occur as complements of copulas in many well-known languages: e.g., English: I am late; the game is today, she isn’t here; or French: nous sommes ici, il n’est pas là, c’est trop). Left unmentioned are important patterns in Cariban languages that might argue against an adjectival analysis. For instance, adjectives do not typically occur modifying verbal predicates, whereas the Cariban class of adverbs typically does. In addition, adjectives do not usually pattern morphosyntactically with adpositional phrases. In the languages in question, however, adpositional phrases share with adverbs all the morphosyntactic properties mentioned in sections 2 and 3; both can be seen as members of a larger class of adverbials. In sum, the arguments against Dixon’s category “adjective1” appear more substantial than the arguments against the category of adverb. Turning to the small category “adjective2” in Makushi, this receives no argumentation at all, but is simply asserted based on the semantics of the seven-member illustrative list of “descriptive nouns” from Abbott (1991: 88). As seen in sections 2-3, all Cariban languages described to date treat a substantial subset of property concepts as lexical nouns (there are many more than seven in Makushi as well). There do not seem to be differences in the morphosyntactic properties (as far as this has already been researched) that would distinguish descriptive nouns as a special subclass (see Table 5 in section 2.1.2 above); and, as far as a comparative construction can be assumed to exist, it does not seem to differentiate them from other nouns. In sum, at this point, Dixon joins Hoff in offering only semantic criteria to separate this “adjective” category from other nouns. If at all, they pattern together with the nominalized adverbs (e.g. the properties of Tiriyó mono ‘big one’, a syn- 122 SÉRGIO MEIRA & SPIKE GILDEA chronically underived descriptive noun, are the same as the properties of kuɾa-no ‘good one’, from kuɾe ‘good’). If future research identifies morphosyntactic grounds for setting up a class of “adjectives2” for these descriptive nouns in a Cariban language, a parallel analysis will likely hold for the cognates in the other languages; but for the time being there still seems to be no reason for that.

4.3 Towards an adjectival subclass of adverbs From the discussion above, we conclude that renaming the entire adverb category “adjective” hides more than it reveals. However, an argument might be made for the identification of subclasses of adverbs, and then one might debate whether or not any subclasses are distinct enough to deserve the status of independent word classes, adverb and adjective. In this section, we turn first to a finer-grained examination of the syntactic distribution of semantic subclasses of adverbs, next we attempt to correlate the syntactic subclasses with morphological properties, and then we end by discussing the implications of these semantico-syntactic subclasses. We begin with the observation that claims about the syntactic behavior of word classes in Cariban are typically somewhat coarse-grained, with a few examples being presented and their behavior then asserted to hold true over the entire category. But adverb classes are notoriously heterogeneous; most researchers, e.g., Schachter & Shopen (2007: 19-20), see them as a default category for words that do not fit in other, more or- derly, classes. In Cariban languages, the adverb class would appear to be even “messier” semantically, as it includes words with the aforementioned adjectival meanings. In order to examine any possible patterns, we separate the adverbs into the following subclasses: typical adverb meanings (including time, place, and manner), and typical adjective meanings (including dimension/size, physical properties, color/ pattern, quantity/order, age, speed, and human propensities). Having made such divisions in our lists of adverbs, we scoured our corpora for examples of each semantic subclass presenting as many as possible of the syntactic behaviors discussed in section 3 as typical of the entire class. As seen in Table 8, members of every subclass were found in nominalized form as attributive modifiers of other nouns; similarly, members of every subclass were found as complements of the copula (a YES means that at least one member of the category in question was PROPERTY CONCEPTS IN CARIBAN 123 found in at least one example of the construction in question). However, for four of the meaning subclasses – all falling within the area of what we would call “adjectival meanings” – we were unable to encounter any examples of a member modifying a verbal predicate as verbal adjuncts.8 This distributional property immediately suggests a division of the larger adverb category into two syntactic subclasses, one which remains heterogeneous (a mix of adverb and adjective meanings), the other of which contains purely adjective meanings. A search through the morphological subcategories of adverbs (mentioned in section 2.1 above) reveals that the t-adverbs are mostly found in the four subclasses that do not modify verbal predicates (though there are exceptions, like təɾemine in ex. 2b, in section 2.1.3 above), so we cannot reinforce the division with excep- tionnless morphological properties. Against this analysis is the caution that must always be exercised when arguing from the small corpora we are able to assimilate on these languages: absence of evidence cannot be taken as evidence of absence. In fact, we would not be surprised to find members of these other categories modifying verbal predicates when semantically or idiomatically appropriate, similar to English smile thinly/ widely, speak sharply/softly, talk much, behave maturely, etc. So we can now weigh the evidence: two positive morphosyntactic properties continue to unify the category, whereas one negative property divides it. If one’s goal is to seek out differences that allow a category of “adjective” to be identified, then the one negative property is well-

8 Interestingly, the same distinction was found among certain postpositional phrases (yet another feature that joins postpositions and adverbs as adverbials): certain postpositions apparently occur only as copular complements and never as verbal adjuncts. These prepositions would tend to fall in the “mental state” or “human propensity” (“experiencer”) semantic area: e.g., Tiriyó se ‘wanting, desirous of’, pɨːnə ‘caring, protective toward’, ino ‘afraid of’, waːɾə ‘knowing’, eiɾe ‘angry at’, je:nə ‘afflicted with (disease)’, etc. In fact, one could say that the postpositional class in the Cariban languages in question is as “strange” or unexpected as its adverbial class, since it includes typically adjectival/verbal meanings such as the above. Meira (2004) treated these postpositions in detail and suggested that they are derived from more complex constructions, in a way that parallels the history of adverbs as developed at the end of this section. 124 SÉRGIO MEIRA & SPIKE GILDEA

“QUASI- COMPLEMENT IN ADJUNCT/ MODIFICATION” SEMANTIC SUBCLASS THE COPULAR MODIFIER OF A WHEN NOMINAL- CONSTRUCTION VERB IZED (N N) Time (now, later, long YES YES YES ago...) Place and Direction YES YES YES (here, thither, hence...) Manner YES YES YES (well,...) Speed YES YES YES (fast, slow...) Human Propensity YES YES YES (sad, angry, sleepy...) Quantity and Order YES YES YES (much, few, two...) Dimension/Size YES YES NO (big, small, long...) Physical Property YES YES NO (hard, sharp, thin...) Color and Pattern YES YES NO (red, blue, striped...) Age YES YES NO (new, old, ...) Table 8. Syntactic distribution of various semantic subclasses of Cariban adverbs situated to help meet that goal – the most clearly “adverb-like” trait is modification of verbal predicates, and the group of “adverbs” that lack this trait all translate felicitously as adjectives. If one’s goal is to seek out empirical validity for a category – that is, to privilege categories that are identified by more than one property – then the two positive properties provide the necessary criteria: ability to nominalize via one of the two nominalizing suffixes and ability to serve as the complement of a copula. One is therefore left with an age-old problem in linguistics: when is a property sufficient to identify an independent word class, as opposed to a subclass of a larger class? In this case, two properties versus one might PROPERTY CONCEPTS IN CARIBAN 125 be sufficient for us to propose a single lexical category (adverbs) with a small subclass (adjectival adverbs). Or we might even prefer to dismiss the negative property as reflecting semantically-based variation in behavior: we could propose that, given an appropriate verbal predicate, there is no grammatical reason why any of these ‘adjectival’ adverbs could not be used to modify a verbal predicate – if a plausible story or metaphor could be found that makes sense of the meaning, as in the English examples above (speak softly, etc.). One could indeed debate this issue, and never conclusively resolve it, just as one is hit by various waves of polemics concerning the existence of a noun-verb distinction in certain languages of the US Pacific coast (especially Nootka). It is not clear to us that the labeling issue is important for the languages themselves. In fact, after immersing our- selves in this problem, what strikes us as important is not the label game, or whether we have subclasses versus separate classes, but rather the question of why this interesting system of lexical items and constructions takes the form that it does. Let us explore some “why” questions. We begin with the question of why the Cariban category of adverbs should include so many “adjectival” (property/quality) meanings. Clear- ly, the answer must be historical, since patterns of lexicalization are not amenable to synchronic analysis – speakers do not choose the part of speech to use with a given concept, they inherit the form-meaning pairing and their identifying properties from their ancestors. A necessary preliminary to a historical explanation would be a reconstruction of how this state of affairs came into being. In order to generate hypotheses about the historical evolution of word classes, we need to understand better the comparative distribution of property concepts into the noun and adverb classes, and also to look for evidence of older morphological complexity in each class. A quick look at the three languages examined in this paper shows that the respective categories differ in size: in Tiriyó and Hixkaryana, most of the “adjectival” meanings occur as adverb roots, which can then be nominalized, whereas in Makushi, many more occur as noun roots, which can then be derived into adverbs. And in fact, a number of apparently monomorphemic adverbs in the other two languages correspond to Makushi Noun + pe constructions: e.g., Tiriyó kuɾe, Hixkaryana ohʃe, Makushi moɾɨ pe ‘good’; moɾɨ ‘good one’ being a noun that corresponds semantically to the Tiriyó and Hixkaryana nominalizations kuɾa-no and 126 SÉRGIO MEIRA & SPIKE GILDEA ohʃa-no. Examining the apparently monomorphemic property concepts in all three languages, there is another asymmetry: a high number of synchronically monomorphemic adverbs in Tiriyó and Hixkaryana contain what look like former derivational morphemes. For example, many end in me, like saːsaːme ‘happy, satisfied’; in the absence of a corresponding noun root *saːsaː ‘happy one’, this adverb must be considered monomorphemic, but it does not take a leap of faith to imagine that it was once derived. Similarly, several other adverbs have an identifiable – though synchronically no longer productive – derivational element, like the -a(ka) ending in amɨma(ka) ‘heavy,’ atuma(ka) ‘warm, hot’, ku- tuma(ka) ‘painful, bitter’, etc. This situation suggests a preliminary hypothesis: some Cariban languages apparently developed a considerable number of adverb roots from earlier property nouns. Many of these nouns have been lost in some languages, such that the now-basic adverbs must be nominalized in order to modify other nouns. This process is perhaps more advanced in Hixkaryana and Tiriyó than in Makushi, but its effects can be seen in all three languages. The older property concept nouns were used frequently in adverbial constructions, either with adverbializing morphology or as arguments of postpositions, and over time the original nominal roots fell out of use.9 This then raises the question of why property concepts should occur so frequently in adverbial constructions, to which the obvious answer is that attributive predicates in copular constructions are primarily (or exclusively, in Makushi and Akawaio) adverbials. This, then, raises its own question: why do some northern Cariban languages allow only predicate adverbs to serve as complements of the copula, and why are predicate adverbials more frequent even in those languages that allow nominal complements with a copula? This appears to be a typologically unusual configuration (Dixon 2006 even used it as an argument against applying the label adverb to the category), and so again one might ask

9 Makushi, with a lower number of synchronically underived adverbs, may be closer to the earlier state of affairs, which, in Proto- and/or Pre-Proto-Cariban times, one might speculatively reconstruct as having no synchronically underived adverbs, but only property or quality nouns, postpositions, and adverbializing constructions. PROPERTY CONCEPTS IN CARIBAN 127 how this situation came to be. Some of our data point towards an interesting hypothesis. As is typologically common, the predicate locative construction in Cariban contains an intransitive locative verb, reconstructed as *eti ‘dwell’ (the reflex of the nominalized form *w-eti-topo is still attested as ‘dwelling place’ in several modern languages, e.g. Kari’nya weitopo ‘dwelling place’, Hoff 1968.141). The reason adverbs function as complements of the copula would be that, etymologically, copular complements were not true complements, but adverbial modifiers of the locative verb: ‘he dwells over there’ > ‘he is over there’. With the further evolution of *eti towards being a copula, the locative construction extended into other nonverbal predicate functions: ‘he dwells happily’ > ‘he is happyADV’;

‘He dwells as a hunter’ > ‘he is [a hunter]ADV’; and ‘he dwells as my father’ > ‘he is [my father]ADV’; etc. In addition to expanding its functional domain, in at least Hixkaryana and Tiriyó, modern reflexes of *eti have moved closer to being a true copula in that they can now take nominal complements (although they are still less frequent, and the semantic distinction contributed by this new construction remains unclear). To sum up our historical hypotheses, we posit that property concepts were formerly a subset of nouns, with adverbs being limited to more traditional concepts like place, time, and manner. When the innovative copular locative construction began to be used for attributive predication, the nominal property concepts had to become derived adverbs in order to occur in these predicates. All property concepts that could be predicated occurred in this construction, and therefore even those that did not modify other sorts of verbal predicates required (and began to occur in) an adverbial form. Copular property predicates have become more frequent than the non-copular predicate, and so the higher-frequency adverbial form of the property concepts began to be seen as more basic, which in some cases has led to attrition of the original nominal roots. This scenario makes sense of the synchronic Cariban facts, and allows us now to return to the question of categorization. Under this scenario, the adjective analysis is historically meaningless: synchronically, adjectives are at best a nascent category. If the hypothesis proposed here is correct, it is more insightful to seek meaningful unity in the historical process whereby property nouns became adverbials in order to function as copular complements than to discuss whether or not one is dealing with one class with a smaller subclass, or with two classes. 128 SÉRGIO MEIRA & SPIKE GILDEA

5. Implications and questions At the end of our paper, we have reached little by way of final conclusions. Rather than reiterate our analysis, we prefer to consider some implications of our hypotheses and to explore descriptive questions for future fieldwork with these (and other Cariban) languages. First, a typological implication. Given that the semantic denota- tions of adverbs and adjectives can co-exist so comfortably in a single word class, we are moved to ask whether the two perhaps share more properties functionally than is usually assumed. We might see the semantic fields of adjectives and adverbs as all being property concepts of one kind or another, and therefore all as plausibly modifying NPs, modifying verbal predicates, or serving as nonverbal predicates. The question, then, is how the grammar of individual languages will code these functions. One could imagine that all three would be done with the same word class, as it is for the English time, place and (some) manner adverbs, e.g., this man here, he put it here, and he’s here (cf. the first three rows of columns 2-5 in Table 10). Elsewhere in English, we have the well-known dichotomy between adjectives in the first two functions and adverbs in the third (the further isolation of quantifiers from adjectives is also shown). We could contrast English with a language where all three would be done with different word classes, e.g., where property concepts are verbs (rather than copular complements) for predication, deverbal adjectives/nouns for nominal modification, and deverbal adverbs when needed to modify another verbal predicate. From this perspective, an obvious logical possibility is the Cariban case, where a single word class modifies verbal predicates and also serves as the complement of the copula, in opposition to the nominal word class that can ‘modify’ nouns (the final three columns of Table 10). We wonder how many permutations of such patterns might be observed if the adjectival and adverbial concepts of more languages were to be sorted into such tables. PROPERTY CONCEPTS IN CARIBAN 129

English Cariban

SEMANTIC NOUN COPULAR VERB NOUN COPULAR VERB SUBCLASS MODIF COMPL. MODIF MODIF COMPL. MODIF Time N/ADV- ADV/N- (now, later, N ADV COP ADV ADV ADV NZR AZR long ago...) Place and Direction N/ADV- ADV/N- N ADV COP ADV ADV ADV (here, thither, NZR AZR hence...) Manner N/ADV- ADV/N- N ADV COP ADV ADV ADV (well,...) NZR AZR Quantity and Order QUANT COP QUANT- N/ADV- ADV/N- ADV (much, few, N QUANT AZR NZR AZR two ...)

Speed N/ADV- ADV/N- ADJ N COP ADJ ADJ-AZR ADV (fast, slow,...) NZR AZR Human Pro- pensity N/ADV- ADV/N- ADJ N COP ADJ ADJ-AZR ADV (sad, angry, NZR AZR sleepy, ...) Dimension /Size N/ADV- ADV/N- ADJ N COP ADJ (ADJ-AZR) ?ADV (big, small, NZR AZR long, ...) Physical Property N/ADV- ADV/N- ADJ N COP ADJ (ADJ-AZR) ?ADV (hard, sharp, NZR AZR thin, ...) Color and Pattern N/ADV- ADV/N- ADJ N COP ADJ (ADJ-AZR) ?ADV (red, blue, NZR AZR striped, ...) Age N/ADV- ADV/N- ADJ N COP ADJ (ADJ-AZR) ?ADV (new, old,...) NZR AZR Table 10. Mapping functions into modifying structures: adverbs and adjectives in English and Cariban 130 SÉRGIO MEIRA & SPIKE GILDEA

A second (theoretical) implication concerns the theory of part of speech systems. Word classes are traditionally identified with the help of morphosyntactic properties. As historical syntax teaches us, morphosyntactic properties – constructions, morphemes, position constraints, etc. – are the result of diachronic evolution, with the specific diachronic paths being important to explain the specific details of each given morphosyntactic property. This implies that word classes themselves also have a diachronic dimension, which can also be relevant, or even crucial, for understanding its synchronic situation. If the hypothesis put forth here is correct, the Cariban class of adverbs owes its very existence to the lexicalization of adverbial constructions based on (property) nouns – a phenomenon reminiscent of how a class of auxiliaries comes into existence (English auxiliaries like be, have, inasmuch as one wants to see them as forming a class, exist because of the reanalysis of constructions in which they occurred with their etymological functions – copula, possessive predicate – but which evolved further into progressive and perfect constructions: is a-going > is going, has a book written > has written a book). We wonder if famous word class problems like the verb-noun distinction in Nootka and other languages in the North-Western United States and Canada would not become more tractable with a similar diachronic perspective that would consider the historical development of the properties proposed to identify nouns and verbs in these languages, and therefore also the historical development of the (emerging) classes themselves.10 A third implication is more inward-looking, at the pre-history of South America. To the extent that our hypothesis survives a more con-

10 One might imagine, for instance, that even if there are languages without a noun-verb distinction, these languages should be diachronically unstable: the typological prototypes of ‘nouns’ and ‘verbs’ (see Croft 2001: 63) that would tend to cause certain meanings (‘cat’,‘person’, ....; ‘go’, ‘break’, ‘build’, ...) to align with certain syntactic behaviors (being subjects and objects; being predicates) would lead over time to the birth of syntactic categories that one might felicitously name nouns and verbs. It is probably the case that, even in the absence of a clear syntactic distinction, there would already be a statistical corre- lation: words with ‘nominal’ meanings are probably more often used as subjects and objects, the reverse being probably true for words with ‘verbal’ meanings, even if both kinds of words could in principle perform all these functions. PROPERTY CONCEPTS IN CARIBAN 131 centrated analysis of more extensive lexical data, we may be able to reconstruct a stage in pre-Proto-Carib in which property concepts are lexical nouns rather than adverbs or adjectives. In two nearby language families, Tupían and Jê, recent years have seen multiple papers on the status of attributive predicates and NP-internal modifiers (cf. Queixalós 2001 for Tupí-Guaranían; Meira 2006 for Sataré-Mawé, with notes on the Tupían family; Oliveira (2003) for a review of the literature in Jê). These papers argue over whether the property predicates are headed by descriptive/stative verbs or by property-concept nouns in nonverbal predicates. While the synchronic debate is far from over, it is worth pointing out that both Tupían and Jê could end up with property concepts reconstructed exclusively to nouns, which could provide another tenuous step in the direction of relating the three into a superfamily, TuKaJê (Rodrigues 1996, Drude & Meira to appear). We conclude this paper with the observation that there are few full grammars of Cariban languages, and even the best of these do not examine the subclasses of nouns and adverbs in much detail. We propose that such an examination might yield interesting discoveries in future descriptive work on Cariban languages, and that certain questions might lead in the direction of those interesting discoveries. First, in checking through a list of property concepts, (i) What proportion are nouns and what proportion adverbs, and which concepts are which? (ii) What morphology is used to move each root to the other class? And (iii) For apparently monomorphemic roots, can a “deeper”, perhaps archaic, root be identified inside synchronically unproductive derivational morphology? Sec- ond, in checking through the constructions involving property concepts, (i) How many possible types of nonverbal predicates are there? We predict every language will have NP NP and COP ADV constructions, but we do not know how widespread the NP ADV and NP COP NP constructions might be. (ii) How does nominal modification work, and in particular, (a) can adverbs modify nouns directly, and (b) is there evidence for order or contiguity restrictions? (iii) What is the grammar that accomplishes the comparative function, and in particular, can nouns, verbs, and adverbs participate equally, regardless of semantic value? Finally, (iv), are there restrictions on whether individual property concept adverbs can modify verbal predicates, and if so, is there any evidence for semantic coherence among those that cannot? We look forward to joining the fieldworkers who will take the op- 132 SÉRGIO MEIRA & SPIKE GILDEA portunity to ask such questions in the years to come.

References Abbott, Miriam (1991). Macushi. In Desmond C. Derbyshire & Geoffrey K. Pullum (eds.), Handbook of Amazonian languages, vol. 3, 23-160. Berlin: Mouton de Gruyter. Amodio, Emanuele & Vicente Pira (1996). Língua Makuxi / Makusi Maimu. Roraima: Diocese de Roraima, Missionários de Scarboro (Canada), and the British and German Embassies in Brazil. Baker, Mark C. (2003). Lexical categories: verbs, nouns and adjectives. Cambridge Studies in Linguistics, vol. 102. Cambridge: Cambridge University Press. Carlin, Eithne B. (2004). A grammar of Trio. Frankfurt am Main: Peter Lang Euro- päischer Verlag der Wissenschaften. Croft, William (2002). Radical construction grammar: syntactic theory in typological perspective. Oxford: Oxford University Press. Derbyshire, Desmond C. (1965). Textos Hixkaryâna. Belém (Brazil): Museu Paraense Emílio Goeldi, Publicações Avulsas no. 3. Derbyshire, Desmond C. (1979). Hixkaryana. Lingua Descriptive Studies, vol. 1. Am- sterdam: North-Holland. Derbyshire, Desmond C. (1985). Hixkaryana and Linguistic Typology. Dallas: Summer Institute of Linguistics and the University of Texas at Arlington. Dixon, R. M. W. (1977). Where have all the adjectives gone? Studies in Language 1:19-80. Dixon, R. M. W. (1982). Where have all the adjectives gone? The Hague: Mouton de Gruyter. Dixon, R. M. W. & Alexandra Y. Aikhenvald (eds.) (2006). Adjective classes: a cross- linguistic typology. Oxford: Oxford University Press. Drude, Sebastian & Sérgio Meira (to appear). A preliminary reconstruction of Proto- Mawetí-Guaraní. Fox, Desrey (2003). Zauro’nödok Agawayo yau: Variants of Akawaio Spoken at Wara- madong. Houston: Rice University Ph. D. dissertation. Gildea, Spike (1998). On reconstructing grammar: comparative Cariban morphosyntax. Anthropological Linguistics Series. New York, Oxford: Oxford Univer- sity Press. Gildea, Spike (2005). Nonverbal predication in Akawaio. Working Conference on the Grammar of Cariban Languages, Centre d’Etude des Langues Indigènes d’Amé- rique, Paris, 5-9 December. Givón, T. (1984/2001). Syntax: An Introduction. 2 vols. Amsterdam: John Benjamins. Haegeman, Liliane (1994). Introduction to government and biding theory. (2nd ed.) Oxford: Basil Blackwell. Hawkins, Robert E. (1998). Wai Wai. In Desmond C. Derbyshire & Geoffrey K. Pullum (eds.), Handbook of Amazonian languages, vol. 4, 25-224. Berlin: Mouton de Gruyter. Hoff, Berend J. (1968). The Carib language. The Hague: Martinus Nijhoff. PROPERTY CONCEPTS IN CARIBAN 133

Koehn, E. & S. Koehn (1986). Apalaí. In Desmond C. Derbyshire & Geoffrey K. Pullum (eds.), Handbook of Amazonian languages, vol. 1, 33-127. Berlin: Mouton de Gruyter. Meira, Sérgio (1999). A grammar of Tiriyó. Houston: Rice University Ph. D. dissertation. Meira, Sérgio (2000). Reduplication in Tiriyó (Cariban). Languages of the World, vol. 17. München: LINCOM Europa. Meira, Sérgio (2006). Stative verbs vs. nouns in Sateré-Mawé and the Tupian family. In Grażyna J. Rowicka and Eithne B. Carlin (eds.), What’s in a verb? Studies in the verbal morphology of the languages of the Americas, 189-214. LOT Occasional Series. Utrecht, The Netherlands: LOT Oliveira, Christiane C. (2003). Lexical categories and the status of Descriptives in Apinajé. International Journal of American Linguistics 69, 243-274. Payne, Doris L. (1993). Nonconfigurality and discontinuous expressions in Panare. In David A. Peterson (ed.), Proceedings of the Nineteenth Annual Meeting of the Berkeley Linguistic Society, February 12-15, 1993: special session on syntactic issues in native American languages, 121-138. Berkeley, CA: Berkeley Linguis- tics Society. Payne, Thomas E. (1997). Describing morphosyntax: a guide for field linguists. Cam- bridge: Cambridge University Press. Pustet, Regina (2003). Copulas: universals in the categorization of the lexicon. Oxford: Oxford University Press. Queixalós, F. (ed.) (2001). Des noms et des verbes en tupi-guarani: état de la question. LINCOM Studies in Native American Linguistics, vol. 37. Munich: LINCOM Europa. Raposo, C. (1997). Makuusiyamî’ya teserukon ko’mannîpî [The Makushi keep their tradition]. Unpublished manuscript. Rodrigues, Aryon D. (1996). Grammatical affinities among Tupi, Carib, and Macro-Jê. Unpublished manuscript. Brasília: University of Brasília. Schachter, Paul & Timothy Shopen (2007). Parts-of-speech systems. In Timothy Shopen (ed.), Language Typology and Syntactic Description (2nd ed.), Vol. 1, 1- 60. Cambridge: Cambridge University Press. Tavares, Petronila (2005). A grammar of Wayâna. Houston: Rice University Ph. D. dissertation.

Truth and knowledge markers in Wayana (Cariban), Suriname

Eithne B. Carlin (Leiden University)

1. Introduction ‘Truth and knowledge markers’ is the term used to group together a seemingly disparate set of grammatical markers that is pervasive in the Cariban, and also in some Arawakan languages of the Guianas in South America. The markers in question express epistemological ideas of realities and truths. In the existing grammatical descriptions, at best, a formal description of the morphemes in question as emphatic, similative, and frustrative markers may be found scattered through various parts of the morphology, where they are dealt with in a structural manner; however, what is needed, besides a structural analysis, is a systematic look at and detailed analysis of their semantic, pragmatic and indeed philosophical import, both as separate morphological items and also collectively as a system of truth and reality markers. In this contribution, Wayana data are presented to exemplify at least parts of this system.1 In section 2, a short typological overview of the language is given. In section 3, the elements of the putative truth and knowledge-marking system are presented. Sec- tion 4 examines the notions of truth and knowledge and that of entrenchment as an explanation of the system. Section 5 presents some conclusions.

1 Wayana is a Cariban language spoken by some 1200 people spread out over 3 neighbouring countries, Suriname, French Guiana, and Brazil. The data presented in this paper are from the Surinamese Wayana who live in the village of Pïlëoimë (a.k.a. Apetina) along the Tapanahoni River. I would like to thank the storytellers Kulepeman and Same for their myths and stories and also for their help in getting me to understand them. 136 EITHNE B. CARLIN

2. Typological overview of Wayana Wayana is a polysynthetic language that uses mainly suffixes (see (1)); the only prefixes found are person markers and derivational diathesis markers; to date, two infixes have been found. It is a head-marking language, as shown in (2):

(1) wë-të-pïnï-tpë ei-topo-npë shoot-able.to-PRIV.NOM-PST 3POSS.be-TMP.NOM-PST ‘Story of the person who couldn’t shoot’

(2) a. Ronnie -pakolo-n b. i-pakolo-n Ronnie 3POSS-house-POSS 3POSS-house-POSS ‘Ronnie’s house’ ‘his house’

The open word classes are: nouns, verbs, adverbs; and the closed classes are: personal and demonstrative pronouns, postpositions, interjections, ideophones, interrogatives, and particles. Tense is marked on both verbs and nominals as an obligatory category, using different markers according to word class; see examples of the nominal past in (3a,b):

(3) a. ï-kamisa-tpï 1POSS-cloth-PST ‘my old clothes’ b. ï-n-ipanakma-tpï-lëken w-ekalë-ja-i 1POSS-3O-hear.NOM.PST-only 1/3-give-PRES-CERT ‘I’m just telling what I heard (I am telling my former hearing thing)’

The verb types are intransitive, transitive, and (derived) reflexives. The clause types are: verbal, non-verbal, clauses with the verb ‘to be’, and quotative. The order of constituents depends on the construction and discourse type, and also on pragmatic considerations; important information is generally fronted. With transitive verbs the constituent order is mostly OV(A), as in (3b) and VS (4a) or SV (4b). Wayana has an evidential system based on witnessed versus non-witnessed events in the past, the form of the latter doubling as a reportative with the verb ‘to say’. In the present tense, a system which is based on certainty versus non-certainty is used. TRUTH AND KNOWLEDGE MARKERS IN WAYANA 137

(4) a. moloinë t-ële-ta-i mule then COREF-liver-VRBLZ-NF child ‘then the child rested’ b. masike mule-psik t-ële-ta-i so child-dim COREF-liver-VRBLZ-NF ‘so the little child rested’

3. Truth and knowledge markers One of the first things one notices when studying the Cariban languages of the Guianas is the great amount of morpheme marking on the main word classes that is often not only difficult to translate into European languages, even in paraphrased translation, but also even to comprehend, mainly because our languages lack the distinctive categories that are culturally entrenched and obligatorily expressed grammatically in these languages. In missionary works of the early seventeenth century, many of the markers presented here were said to be “ornate particles” without which “the sentence is perfectly fine” (Hardman 1986: 113). The grammatical marking in question is found in Wayana (in fact in almost all Cariban languages) as either enclitics, suffixes, infixes, or particles and includes:  a facsimile (similative) marker -me; (suffix on nominals)  several assertive or emphatic markers (clitics; infix on adverbs)  a marker that has the meaning ‘truly’or ‘through and through’ (clitic)  a frustrative marker, -lep (clitic)  nominal past tense markers (suffixes)  evidential marking (witnessed vs. non-witnessed; reportative vs. everything else) (affixes)

With the exception of the evidential marking, which is not considered further in this paper, all the above elements are marked on nominals and/or adverbs and postpositions. In fact, with a few exceptions, the focus of this paper is restricted to the kind of marking given above that is found on the nominal class, where we will see indeed that Wayana (and the Cariban languages in general) allows for a very high degree of specification. Consider the examples (5) through (9), with the facsimile -me, emphatic -le, ‘truly’ -lë, frustrative -lep, and nominal past -npë/-npï and 138 EITHNE B. CARLIN

-tpï/tpë, where the function of the relevant grammatical markers on the nominals is given in small caps on the right, after each sentence:

(5) a. FACSIMILE Kulum, wëlïi-me Kulum kun-eha-k, eagle.sp woman-FACS eagle.sp 3.PST-be-R.PST waluhma-me young.woman-FACS ‘Eagle was a woman, a young woman (manifestly but not inherently)’ (instrinsically an eagle but in the outer casing of a woman) b. FACSIMILE tuwalë manai mëlë Tïliyo-me ë-w-esi-ke know you.are DP.INAN.MED Trio-FACS 2-1TR-be-INST ‘you know that because you’re a Trio’ (not biologically a Trio)

(6) INTENSIFIER + EMPHATIC Wajana-h_le man inëlë Wayana-INTENS_EMPH he.is 3PRO.ANIM.ANA ‘he’s a real Wayana’ (at least he behaves totally like a Wayana)

(7) Wajana-lë inëlë TRULY Wayana_TRULY 3PRO.ANIM.ANA ‘he’s a real (pure-blooded) Wayana’ (as opposed to the offspring of a mixed union,e.g., a combination Wayana/Wayãpí)

(8) a. ï-pawana_lep manai FRUSTRATIVE 1POSS-friend_FRUST you.are ‘you are my friend’ (you are my friend but I don’t have many ad- vantages from that)2

2 The term pawana ‘friend’ also has the meaning ‘trading partner’ and -pawana eitop ‘being someone’s friend/trading partner’ is something of an institution among the Cariban groups; it entails having obligations towards the partner in question, and is based on a high degree of reciprocity (see Carlin 2004: 22-23). TRUTH AND KNOWLEDGE MARKERS IN WAYANA 139

b. FRUSTRATIVE paila tëkalëi pïlëo malë i-të-top-kom, bow he.gave arrow also 3POSS-go-TMP.NOM-PL tëhem wë-top_lep meat shoot-TMP.NOM_FRUST ‘He gave him a bow and arrows for their journey, a means for shooting game animals’ (but this man didn’t use them to shoot meat: the bow and arrows were, in vain, instruments for shooting because the man couldn’t shoot)

(9) moloinë t-ëne-i i-pakolo-tpï NOMINAL PAST then COREF-see-NF 3POSS-house-PST ‘then he saw his house (former house)’

In the examples above, the extra morphology marked on the nominals has to do with specifying whether or not the referent of a noun X is inherently so, really so, truly (through and through) so, in reality so but with at least one flaw, or formerly so. In examples (5a,b) neither referent marked with the facsimile marker -me is inherently what they are said to be, although for all intents and purposes they now are manifestly what is denoted by the noun ‘woman’ in (5a) and ‘Trio’ in (5b). Examples (6), with the emphatic -le, here preceded by the intensifying infix -h-, and (7), with the ‘truly’ or ‘through and through’ marker -lë, in many contexts, can be used as synonymous forms. However, the difference lies in the fact that the referent in (6) counts as being a Wayana, though he may be of mixed blood, which does not actually make him less Wayana if he lives as a Wayana, although his DNA may prove otherwise; on the other hand, in (7) the speaker is stating that the referent is a full-blooded, non- mixed Wayana, that is, there is a certain degree of inherency and permanence in (7) that is not included in the meaning of (6). In example (8), with the frustrative marker, the speaker is saying that that the addressee is indeed his friend but that the friend is not fulfilling all the obligations that friendship brings, such as helping his friend when in need, giving him things (reciprocally) etc.; likewise, the bow and arrows in (8b) are inherently bows and arrows but they are not being used for the purpose that bow and arrows have, namely shooting game animals, in order to provide the family with meat. Example (9), with the nominal past marker -tpï, shows us one of the obligatory categories in Wayana, namely that of marking a nominal as being ‘former’, when the referent of that noun no 140 EITHNE B. CARLIN longer exists or functions as such, or when its possessor has died, or when the possessive relationship has ended. In fact all the above categories are obligatorily marked when the pragmatics of the situation require this, and where the marker has undergone grammaticalization, as is the case, for example, with the facsimile in its depictive usage. In the free translation of the examples above, I have added in pa- renthesis the less readily translatable content of the utterance. However, a Wayana translating the above sentences into a European language (one of the national languages of Suriname or French Guiana, namely Dutch or French) does not usually add the parenthetical information, rather they offer a translation of the utterances without a translation of the facsimile, ‘truly’, frustrative, and nominal past. The examples (10-13) show the nouns without the extra morphology, where one sees that the translations are identical to the translations in (5-9) above, the parenthetical information left aside. This is the reason why early works could refer to these markers as “ornate particles”, because in translation they seemingly added nothing to the content of the utterance. However, it is clear from the above parenthetical translations that there is quite a difference in meaning between the forms of the nouns with and without the additional morphological markers.

(10) Palasisi ë-w-esi-ke (see 5b) white.person 2-1TR-be-INST ‘because you are a white person’

(11) Wajana inëlë (see 6,7) Wayana 3PRO.ANIM.ANA ‘he is a Wayana’

(12) ï-pawana ëmë (see 8a) 1POSS-friend 2PRO ‘you are my friend’

(13) helë man ï-pakolo-n (see 9) DP.INAN.PROX it.is 1POSS-house-POSS ‘this is my house’

But how is this difference, and thus the meaning of these markers to be characterized? Indeed a number of questions arise here: for example, TRUTH AND KNOWLEDGE MARKERS IN WAYANA 141 why do speakers of Wayana (and the other Cariban languages) feel the necessity for such fine-grained specifications? Why is it that precisely these morphemes have become grammaticalized? Is the world of the Wayana so confusing as to whether something really exists in the here and now, or are their nominal concepts so loosely defined that for example a noun only has limited classificatory value, or do they know something about the nature of concepts that speakers of standard average European languages do not? And what is then the meaning of the linguistic symbol? It was these questions that led us to examine the meaning and function of all the ‘extra’ information that is provided in the extra morphology. This is not to say that what is found in the examples given cannot be expressed in standard average European languages, on the contrary, they can be, in various, generally periphrastic ways, as evidenced by the parenthetical translations. However, an explanation is here being sought for the ways in which such markers become grammatically obligatory in a language such as Wayana. Before we attempt to answer these questions, I give below a short overview of the markers themselves and show when and how they are used.

Facsimile The facsimile (a.k.a. essive-translative or similative) -me is used to express that the denotee of the noun is not inherently but rather manifestly the denotee, as shown in (5a) above taken from a mythological text, and in (5b) in everyday usage where the person in question was not a Trio Amerindian but a white person. The facsimile marker is used to express ‘change of state’ as in (14a,b), where the Wayana had turned into birds and a woman into a spider monkey respectively. When used with non- transformational verbs it expresses a non-permanent or transient state (15):

(14) a. mëkpalë-me_tot t-ëtï-he tree-dwelling.animal-FACS_PL COREF-become-NF tolopït-(t)o(m)-me bird-PL-FACS ‘they had (manifestly) become tree-dwelling animals, (manifestly) birds’ 142 EITHNE B. CARLIN

b. alimi-me t-anukta-i wëlïi spider.monkey-FACS COREF-transform-NF woman ‘the woman changed into a spider monkey

(15) mule-me-hnë ïw-aptao child-FACS-PERSIST 1-when ‘When I was (still) a child’

The marker -me is used syntactically as a marker of secondary predication or as a depictive to express a physical or psychological state of one of the participants (16).

(16) tolopït-me_tot t-ëhalë-i birds-FACS_PL COREF-disperse-NF ‘They dispersed as birds (no longer as Wayana)’

As in all the Cariban languages, -me has been grammaticalized to form adverbs, for example tïnme ‘quiet’, talanme ‘ maybe’. Furthermore, some discourse cohesion markers which have become lexicalized are likewise formed with -me, for example, malonme ‘then’, mëlëme ‘then (being in that state)’. In addition, -me combined with the nominalizer -to(po) has a purpose reading as in apëih-toh-me ‘in order to grab it’.

Emphatic markers Wayana has several means for expressing intensity, emphasis, assertion, and speaker’s strong intentions. It is not always clear what the difference is between the different markers. ‘Intensity’ is expressed by the infix -h-, which occurs after the first vowel of an adverb or postposition. The longer the -h- is pronounced (17a,b), or when it is pronounced with a strong burst of air followed by an extended glottal closure, the greater the intensity. The intensity infix is often, but not obligatorily used in combination with one of the emphatic markers (18).

Intensity: (17) a. upak ‘long ago’ uhhpak ‘really long ago’ b. hemalë ‘now, today’ hehmalë ‘right now’ TRUTH AND KNOWLEDGE MARKERS IN WAYANA 143

(18) wewe pole-h_le tïï-kë tree in.alignment-INTENS_EMPH do-IMP ‘put it exactly in alignment with the tree!’

The emphatic clitic -le (often preceded by the intensifier infix) is found on nominals (19), finite verbs, adverbs, and postpositions. Both the infix -h- and the emphatic -le are found in free variation with the emphatic marker -nma found on adverbs and postpositions (20). The clitic -le is also found with imperatives to strengthen the command.

(19) ipokanu-h_le ‘he is a really good person’ ipoke_nma man, i-se_nma wai good_EMPH it.is 3-DESID_EMPH I.am ‘It’s really good, I really want it’

(20) kole-nma ‘very many’ Or: kohhle ‘very many’

Another marker similar to -le is -lë which has the meaning ‘real’, ‘thorough(ly)’,‘true’ or ‘truly’ as in (21a,b) whereby (b), which is example (7) repeated here, seems to be synonymous with Wayana-h-le ‘a real Wayana’. In Wayana, the difference between the ‘truly’ marker -lë and the emphatic marker -(h)le is that of permanence or constancy, as opposed to ‘really X’ at a given moment in time.

(21) a. ipokan_lë inëlë ‘he is a truly good person (always has been, always will be)’ b. Wayana_lë inëlë ‘he’s a real (pure-blooded) Wayana (as opposed to the offspring of a mixed union, e.g. a combination Wayana/Wayãpí)’

There are other derived markers that also express notions of thorough- ness or totality, such as: -phele ‘up to the limit’: ijume-hpele man ‘he is totally mature (as mature as is possible for a person)’ -pkë_lë_le ‘really absolutely’: ipoke-pkë_lë_le ‘ really absolutely brilliant’ (e.g. news)’ but since they are generally found in combination with, and sometimes derived from the basic emphatic markers they are not discussed separately here. 144 EITHNE B. CARLIN

Frustrative A feature quite commonly expressed in the languages of Amazonia is the one termed ‘frustrative’. In Wayana, the frustrative is expressed by means of the clitic _lep: it is used in clauses to express that an action is unsuccessful or in vain, as shown in (22a), where the protagonist left to go hunting but didn’t shoot any game. The frustrative marker can also be marked on nouns to express that the referent of the noun is lacking in at least one semantic feature of the noun, or that the object expressed by the noun is not used for its inherent purpose as shown in (22b), a repetition of (8b), where the bow and arrows the man had been given were not used for the purpose of shooting and so were ‘in vain’ a means of shooting game.

(22) a. anumalë tïtëi inëlë, koko-psik tomorrow COREF-go-NF DP.ANIM.ANA night-DIM tï-të-i_lep COREF-go-NF_FRUST ‘The next day he left, he left early in the morning (but he didn’t shoot any game)’ b. paila tëkalëi pïlëo malë i-të-top-kom, bow he.gave arrow also 3POSS-go-TMP.NOM-PL tëhem wë-top_lep meat shoot-TMP.NOM_FRUST ‘He gave him a bow and arrows for their journey, a means for shooting game animals’ (but this man didn’t use them to shoot meat: they were an instrument for shooting but were never shot)

The frustrative can be seen as a non-fulfillment of an expectation or belief, and as such it contrasts with negation marking which negates an affirmative statement. In addition, a combination of the emphatic markers and the frustrative is frequently found in Wayana, resulting in very precise specifications such as (16): TRUTH AND KNOWLEDGE MARKERS IN WAYANA 145

(16) lome heinë i-waliktao_lë_lep3 but this.side 3-behind.LOC_TRULY_FRUST ‘but not exactly behind them on this side’ (practically right behind them but not quite)

Nominal past The final marking that is relevant for the purposes of this paper is obligatory past marking on nouns by means of the suffixes -npë/-npï and -tpë/tpï, used e.g., when the referent of that noun no longer exists or functions as such (cf. (9) above), or when its possessor has died, or when the possessive relationship has ended, see (17a,b).

(17)a. kumakaimë-tpë ‘a former (felled) kankantri (large tree sp.)’ b. j-etatï-npï ‘my former hammock’ (e.g. that is no longer usable; or after I have died)

While I have just given a general picture of what type of markings there are, what I want to focus on now is the ubiquity in these languages of the markers throughout the grammar, to express notions relating to intrinsic or non-intrinsic or transient values and the function thereof.

4. The meaning of the concept: real or not? In the anthropological literature on Amazonian cultures, one often comes across references to how difficult it is to make sense of the content of myths since protagonists are now human, now spirit, now animal, in the words of one of the leading anthropologists Joanna Overing (1990: 602), one is faced with “chaos, obscurity, ambiguity, and confusion”. As another leading anthropologist, Peter Rivière (1994: 256) pointed out in his article WYSINWYG in Amazonia:4 “One of the lessons I learnt when trying to obtain exegesis on myths was the futility of trying to find out whether a particular character was a human, animal, or spirit”. In general, however, this confusion arises from the translations of the mythical

3 This example was collected by Karen Hough while carrying out the Nijmegen Space Games battery test in 2007; see Hough (2008). 4 WYSINWYG stands for ‘What you see is not what you get’. 146 EITHNE B. CARLIN texts. As example (5a) shows, Eagle was manifestly a woman, a ‘facsimile woman’ so to speak, but was she now an eagle or a woman? Both in fact, which may seem irrational or contradictory to the western mind, since in our view an eagle cannot be simultaneously an eagle and a human woman. In order for (5a) to be a true statement, we need a frame of reference within which this is possible. The western mind seeks to explain the double-identity of Eagle here by means of metaphor because our languages lack such a frame of reference that allows two things to be the same while being intrinsically different.5. But what is then the frame of reference that does allow this? In fact it was in Overing’s seminal article from 1990 that she found a means to dissipate the seeming confusion she was facing while working on myths, namely through the work of the American philosopher Nelson Goodman, in particular his 1978 book Ways of Worldmaking in which he asks in what sense there could be a “multiplicity of worlds” or “versions of the world”. Indeed Goodman’s work is in many ways an elaboration of the German philosopher Ernst Cassirer’s insights into symbolic forms. For Cassirer (1946: 8) “myth, art, language and science appear as symbols; not in the sense of mere figures which refer to some given reality by means of suggestion and al- legorical renderings, but in the sense of forces each of which produces and posits a world of its own”. If this is the case then, Goodman argues, there must be a multiverse rather than a universe, several worlds as opposed to one, or at least several versions of the world. As Overing (1990:603) points out: “The scientist, artist, myth-teller or historian, and shaman-curer are ‘doing much the same thing’in their constructions of

5 Whereas compounds such as ‘spiderman’ exist in the western world of films, whereby the rightmost element is the head, and as such combines the characteristics of two creatures, man and spider, basically a man but with the attributes and behavioural elements of a spider, there is a major constitutional difference between this type of compound and that expressed by the Wayana noun plus -me concept. The notion N-me highlights the sameness in appearance rather than in behaviour. If spiderman were to have only the appearance of a spider while being human then the two notions could be considered comparable, however, this is not the case. Perhaps a more suitable analogy would be that found in films about aliens who take on a human form and attributes whereby their alien essence is hidden from the naked eye, that is, on the outside they are human but in essence they are not human; they merely appear to be so. TRUTH AND KNOWLEDGE MARKERS IN WAYANA 147 versions of worlds. However, while the thought processes for construct- ing worlds are in many ways similar, the facts of which these worlds are made are very different indeed.” It is now a generally recognized fact that the ‘world’, or rather version of the world, of Cariban (and also many other Amerindian) peoples differs in some respects from our western world in that it has as one of its canons that the anima or spirit can manifest itself in different forms, so it can take the form of now an anaconda, now a jaguar, and now a human, that is, the ‘outer casing’ can change; the world is a transformational one. Within the Cariban world the ambiguity and confusion that we find from our point of view does not exist because a transformation from one state to another is grammatically marked. This realization leads us to the conclusion that this world can only be described within the frame of reference for that particular world or version of the world, that is, how the symbol refers is dependent on the system of symbolization within which the symbol is found. Our version of the world does not allow for transformations of the Amazonian type, and our grammars do not need to account for them, which is why we end up with irrational or contradictory statements. We tend not to focus on states of being or changing states. In other words, we are confined to certain ways of describing what is being described. In describing, or referring to such an instance of transformation, the Cariban peoples have a linguistic system that makes compatible the two world versions of example ‘Eagle being at the same time an eagle and a young human woman’. As evidenced by the parenthetical translations in the examples above, our languages tend to resort to long expla- nations in order to reconcile the eagleness and the humanness of the referent ‘eagle’ because we have not entrenched such a mode of reference in our languages. However, the fact that we ultimately can translate such sentences adequately enough to capture the meaning shows us that there is in fact no difference in principle between the predicates we use and those we could use, but there is a pragmatic difference in habit, or, in Goodman’s terms, of “entrenchment” of ways in which we refer to the world. And the entrenchment “provides the required distinction” (Good- man 1984: 38). The distinction relevant here is a thread that runs through most aspects of the Cariban world, namely that rather Wittgensteinian notion of ‘states of being’ and this is duly a category that has to be expressed grammatically. So in which ways do the morphemes presented above act as markers of states of being (again we are still keeping within 148 EITHNE B. CARLIN the nominal domain)? And how do they represent either truth or knowledge? In short we can characterize the markers as follows: (a) Facsimile -me: charts transformations of spirit between states; ‘being’ simultaneously (b) Emphatic _le: gives a qualitative value of a current state, generally temporary or at a given point in time (c) ‘Truly’ _lë: gives an instrinsic value, such as biological, inherent; has a non-fluctuating permanence (d) Frustrative _lep: in the nominal domain: at least one semantic feature of that denoted by noun is not fulfilled, i.e., ‘flawed’, ‘not functioning as X should’: this marker gives us insight into cultural norms and culturally pertinent modes of behaviour (cultural entrenchment); in the verbal domain: ‘knowledge’ of outcome of state of affairs, i.e. ‘in vain’ (e) Nominal past –npë/-npï and tpë/tpï: transformations between present, former and future states; progression along a temporal line What the above characterizations show is that there is an underlying temporal dimension with all the markers with the exception of the frustrative _lep. In fact, it would appear now that the frustrative, at least when marked on nominals, has nothing whatsoever to do with either truth or knowledge, and everything to do with cultural norms and other expectations. For example, the noun eluwa ‘man’ marked with the frustrative refers to a man who does not fulfill the conditions of his man- hood, namely he is not a good hunter, he does not provide his family with meat. In the western world eluwa_lep ‘man_FRUST’ would likely have very different connotations. The emphatic and the ‘truly’ clitics in (b) and (c) would in one sense seem to be opposites of each other, expressing a non-permanent versus a permanent quality, that is, ‘he’s really good (right now)’ as opposed to ‘he is good through and through’. The facsimile -me allows the expression of two simultaneous states, namely the instrinsic and the adopted state, as we saw with the eagle being both eagle and woman above. There is no temporal progression involved here, both are co-existent ongoing states, that is, the actual state of being is foregrounded rather than the beginning or end of the state. This is in contrast with the nominal past markers which do show a temporal progression, namely a change of state that is complete, such as an ex-wife, a former house, or (my) future grandchildren ïpalïnpïtom, which is literally ‘my grandchildren (descendants) after I have passed on’. TRUTH AND KNOWLEDGE MARKERS IN WAYANA 149

5. Conclusions The aim of this contribution was to look for the regularities and common semantic denominator in the contrasting markers given, namely the facsimile, the emphatic, the ‘truly’ marker, the frustrative and the nominal past. The specificity with which nominals are marked allows the speaker to give as much information as is required for the listener to get as full a picture as possible, that is, the linguistic system itself portrays an entrenched specificity that Standard Average European languages do not have. All the markers, with the exception of the frustrative, have a temporal aspect in them, namely one of permanence vs. non-permanence. There is a pervasive regularity in Wayana and other Cariban languages of encoding states of being or not being because this is what is habitually projected by the speakers and the system which has been built up is a useful and efficient one within their terms of reference, that is, version of the world. A sentence such as (5a) ‘Eagle was a woman, a young woman’ may be a contradiction of the known truth in our version of the world that an eagle is a bird and a woman is a human being, but perhaps right versions of worlds do not necessarily coincide with what is true but with what is right and fitting in that version of the world. Goodman’s proposal that the term ‘truth’ should be replaced by “rightness of symbolic function”, and ‘knowledge’ by “understanding”, notwithstanding the philosophical implications that these terms are not co-extensive, may bring us to a greater understanding of the system we are trying to un- ravel. The relationship between “rightness of symbolic function” (truth), entrenchment (projection), and “understanding” (knowledge) is both a decisive and determining one, because it allows us to apply a scientific rigour to the constructed system and provides us with strict rules by which we can determine what constitutes a possible world version versus a spurious one. The existing system is the result of the entrenchment of the importance of instrinsic versus non-intrinsic values of nominal de- notees, extending far beyond the domain of mythology as is evidenced by the examples above that pertain to normal daily life. External evidence for the claim that such markers form a system together comes from at least one language contact situation where the Arawakan language Mawayana (spoken in Suriname) has borrowed some of the Cariban functional categories described here, namely a facsimile category, nominal past marking, and a reportative/evidential category, without actually borrowing the morphology itself, that is, 150 EITHNE B. CARLIN

Mawayana pressed into service old functionally defunct markers to do the job of marking fascimile etc., or it created new markers to express these categories (see Carlin 2006). While I have only dealt with part of the system in this paper on the Wayana markers, namely some of the nominal marking, a full study would have to include the role of evidentiality and reportative marking, as well as the role of temporality through the system.

Abbreviations: A: agent; ANA: anaphoric; ANIM: animate; CERT: certainty; COREF: coreferential; DESID: desiderative; DIM: diminutive; DP: demonstrative; FACS: facsimile; FRUST: frustrative; IMP: imperative; INAN: inanimate; INST: instrumental; MED: medial; NF: non-finite; NOM: nominalizer; O: object; PERSIST: persistive; PL: plural; POSS: possessive; PRES: present; PRIV: privative; PRO: pronoun; PROX: proximal; PST: past; R.PST: re- mote past; S: subject; TMP.NOM: time, manner, place nominalizer; TR: transitive; VRBLZER: verbalizer

References Carlin, Eithne B. (2004). A Grammar of Trio, a Cariban Language of Suriname. Duisburg Papers on Research in Language and Culture 55. Frankfurt am Main: Peter Lang. Carlin, Eithne B. (2006). Feeling the need: The borrowing of Cariban functional categories into Mawayana (Arawak). In Alexandra Y. Aikhenvald & R.M.W. Dixon (eds.), Grammars in Contact: A Cross-Linguistic Perspective. Explora- tions in Linguistic Typology, Vol. 4, 313-332. Oxford: Oxford University Press. Cassirer, Ernst (1946). Language and Myth. New York: Dover. Goodman, Nelson (1978). Ways of Worldmaking. Indianapolis: Hackett. Goodman, Nelson (1984). Of Mind and other Matters. Cambridge, Mass.: Harvard University Press. Hardman, Martha James (1986). Data-source marking in the Jaqi languages. In Wallace Chafe & Johanna Nichols (eds.), Evidentiality: The Linguistic Coding of Episte- mology, 113-136. Norwood, New Jersey: Ablex Publishing Corporation. Hough, Karen (2008). The Expression and Perception of Space in Wayana. Leiden: Sidestone Press. Overing, Joanna (1990). The shaman as a maker of worlds: Nelson Goodman in the Amazon. Man, New series, Vol.25/4: 602-619. La posture du corps dans la classification et la localisation: l’exemple du sikuani

Francesc Queixalós1 (CELIA/CNRS Paris)

J’étudie dans cet article, qui reprend en les reformulant les présentations de Queixalós 1992 et 1998, l’usage que la langue sikuani2 fait d’une certaine lecture des configurations spatiales que le corps humain peut prendre. Ces configurations s’expriment dans un paradigme de quatre verbes. La première partie de l’article traite de la façon dont les êtres sont discriminés en fonction de leur posture spatiale, la seconde aborde l’exploitation des verbes de posture aux fins de localisation dans l’espace.

1. Comment la posture classifie les êtres Il existe quatre verbes de posture, qu’en première approximation je glose comme suit:

(1) eka ‘être assis’ nuka ‘être debout’ boka ‘être étendu’ ruka ‘être suspendu’

Une première ligne de clivage s’établit entre le dernier et les trois autres: assis, debout, étendu comportent l’idée de contact avec le sol, et se distinguent entre eux par la configuration géométrique, alors que suspendu, ruka, n’indique que l’absence de contact avec le sol, et ne dit rien sur la configuration géométrique. Notons en passant que ce dernier verbe fait l’originalité du paradigme sikuani par rapport à des exemples plus connus de forte thématisation de la posture, tels qu’on les observe en

1 CELIA (CNRS, IRD, Inalco, Université de Paris 7). 2 Bassin du moyen Orénoque, Colombie et Vénézuela. 152 FRANCESC QUEIXALÓS

Amérique du Nord, en Afrique, ou dans les langues germaniques conti- nentales (Lemmens 2002a, 2002b3). Quelques idées préliminaires sur le contenu de ces verbes, obtenues en plaçant l’informateur en situation expérimentale devant des objets non issus du contexte culturel quotidien, tels les cubes en bois de l’école, nous aideront à cerner leur sémantisme. Un objet oblong, cylindre ou parallélipipède rectangle, est dit être debout, nuka, s’il est dressé, et étendu, boka, s’il est couché. Le critère se trouve bien sûr dans le rapport de la base – plutôt: la plus grande longueur mesurable dans la base – à la hauteur. Mais étendu s’emploie aussi pour désigner la posture d’un objet plus ou moins isotropique, c’est-à- dire ayant les mêmes propriétés physiques dans toutes les directions. Pour un locuteur sikuani, la sphère, mais aussi le cube, ont cette caracté- ristique. La posture assis, eka, est centrée sur deux idées: avant tout celle de base, ou face à vocation de contact stable avec le support; puis celle de la proportion entre base et hauteur, lesquelles, contrairement à debout et étendu, doivent être de dimensions comparables. Un cylindre dont la hauteur est en gros équivalente au diamètre de la base est assis s’il repose sur la base, et étendu s’il repose sur le flanc. Si nous en venons aux objets réels de la vie quotidienne, on voit que la plupart d’entre eux sont non isotropiques: ils sont configurés ou rendus asymétriques par les filtres culturels qu’à la perception des objets imposent les façons, socialement déterminées, d’interagir avec eux. Une calebasse en tant que fruit d’une plante est un objet isotropique: quelle que soit sa position, on dira qu’elle est étendue, boka. Qu’on y fasse un trou et l’évide pour en faire un récipient: elle a acquis une base, à l’opposé de l’ouverture. Si elle repose dessus, elle est assise eka, et dans tout autre posture, étendue. Dans la combinaison d’un verbe de posture intransitif et un sujet, il se fait jour un certain nombre de restrictions concernant l’acceptabilité sémantique de l’association. Autrement dit, le lexique des noms est sous- catégorisé selon une évidente prototypicité de la combinatoire entre une entité et une posture corporelle. Dans la confrontation d’un nom au paradigme de (1) se révèlent trois catégories: celle où l’association à un verbe

3 Je dois à Colette Grinevald d’avoir pu prendre connaissance des travaux de M. Lemmens. LA POSTURE DU CORPS EN SIKUANI 153 donné est exclusive – tel nom ne peut être sujet que de tel verbe, tout autre association aboutit à un énoncé étrange –, celle où l’association est canonique – tel nom est préférentiellement sujet de tel verbe, tout autre association est marquée sémantiquement, fortement informationnelle –, et celle où elle est libre ou indéterminée – tel nom s’associe avec autant de facilité à un verbe qu’à un autre.

1.1 Noms à posture exclusive Cette catégorie, dont chaque élément n’est combinable qu’à un seul verbe, comporte presque exclusivement des noms non animés.

1.1.1 Les assis. L’idée de stabilité est sans doute une composante séman- tique importante pour eka ‘assis’. On y associe: le monde ou territoire, la maison, le nid, le village, les flaques et le marigot, les poudres fines contenues ou répandues (voir plus bas), le rocher, la colline, le cratère constituant la partie visible du nid des fourmis manioc, et la termitière de savane, qui a pourtant l’aspect d’une colonne pointue; également les plantes sans tige, comme l’ananas – et son fruit –, ainsi que l’agave; le fœtus dans le ventre est assis (et je le suppose classé comme non animé, puisque le tout nouveau-né l’est encore, grammaticalement, jusqu’aux premiers cris et soins).

1.1.2 Les debout. Le mur, le ravin surplombant une rivière sont vus sous l’aspect de nuka, ‘debout’.

1.1.3 Les étendus. L’idée de contact large avec le sol prédomine ici, avec les noms pour la surface du sol – considérée comme une extension qui supporte le monde habité –, les pierres, les rochers en forme de dalle, la surface dégagée devant la maison, la plantation, la forêt, la savane, la plage, les plantes rampantes. Les araignées terrestres telle la mygale s’associent à boka, ‘étendu’, ce qui amène à préciser que ce n’est pas la surface de contact proprement dit qui caractérise cette posture, mais la surface délimitée par les points de contact. Rocher et pierre sont désignés par le même vocable, iboto, et la distinction entre une petite pierre et un gros rocher se fait sur la base du verbe de posture, respectivement étendu et assis, eka. Le locuteur sikuani explicite que la différence entre une pierre et un rocher tient dans la pos- sibilité / impossibilité de bouger l’objet. La stabilité comme trait consti- tutif de la posture assis se confirme. 154 FRANCESC QUEIXALÓS

1.1.4 Les suspendus. Cette posture s’attribue aux nuages, à la brume, au ciel diurne, à la Voie Lactée. Les aoûtats, désignés par un nom animé, ont pour posture exclusive ruka, ‘suspendu’. Des notions plus immaté- rielles appartiennent à cette catégorie, comme le temps qui reste avant une échéance, le contenu des histoires, et, inféré-je indirectement: les mœurs ou tradition, puisqu’instituer une tradition se dit suspendre les mœurs; le sens ou message, puisque ‘traduire X dans la langue Y’ se dit: ‘suspendre X dans la langue Y’, et envoyer un message se dit ‘suspendre la parole’ (pour les verbes transitifs de posture, voir ci dessous section 6).

1.2 Noms à posture canonique Ici nous voyons des noms qui en situation non marquée sémantiquement et pragmatiquement apparaissent préférentiellement avec tel verbe, tout en se réservant la possibilité d’une association avec un autre (d’autres) verbe(s). De cette catégorie font autant partie des animés que des non animés.

1.2.1 Les plutôt assis. On peut faire adopter à un objet à peu près n’importe quelle posture, en imagination du moins. Dans la vie quotidienne, les objets, de par leur configuration et leur usage, ont le plus souvent une posture canonique, typique, naturelle, normale. Les contenus, je l’ai dit, sont assis, eka, ainsi que les poudres, même répandues. Mais les poudres à grains visuellement saillants, comme la farine de manioc, sont assises comme contenu et étendues, boka si on les répand. Plausiblement, l’idée de compacité fait partie du sé- mantisme de la posture assis. Le tabouret traditionnel, bas sur pattes, et les récipients de rapport base / hauteur proche de 1, tels que les marmites, tasses, flacons renflés, apparaissent comme assis dans leur position normale, et étendus s’ils sont renversés. L’assiette devrait être, d’après sa configuration géométri- que, étendue, la base se présentant comme bien supérieure à la hauteur. Mais le filtre culturel lui impose d’être assise, puisque pour un récipient être étendu signifie être renversé. Une cavité souterraine, grotte, terrier, est assise, mais sera perçue comme suspendue, ruka, si on y voit le lieu d’un déplacement, quelque chose comme un tunnel. La notion de posture canonique, par la variabilité qu’elle renferme, a une affinité certaine avec le domaine de l’animé: en gros, un non animé LA POSTURE DU CORPS EN SIKUANI 155 prend plus d’une posture par l’effet d’une force extérieure; chez un ani- mé les postures sont davantage fonction de son comportement propre. La posture secondaire par excellence d’un animé, quelle que soit sa posture canonique, est celle d’étendu, boka, à prendre presque invaria- blement comme équivalente de mort. Parmi les quelques animés à posture canonique assis figurent les crapauds et les grenouilles terrestres. On les dit avoir les extrémités courtes, quoique visibles.

1.2.2 Les plutôt debout. Les plantes à tronc ou à tige sont canonique- ment debout, nuka. Les objets nouveaux s’incorporent aux catégories existantes. La bouteille, la table, le camion se tiennent debout. La bouteille, boteya-bü, montre une collision entre son verbe de posture canonique, qui implique une forme allongée, et son classificateur -bü, qui ren- voie à la classe dont le prototype est la sphère (un autre membre atypique de cette classe d’objets renflés est le tubercule de manioc). La table, de par ses dimensions, devrait être assise. Mais la longueur de ses pattes, et donc leur visibilité, la classent comme debout. Le même critère prévaut – les roues tenant lieu de pattes – pour le camion. La plupart des objets debout (et aussi assis) dans leur posture normale se tiennent étendus lorsqu’ils prennent une autre position qui garde contact avec un support. En quelque sorte, une marmite renversée, un arbre couché, sont ‘dé-configurés’ par rapport à une norme perceptive. De là l’extension de boka, ‘étendu’ à l’assiette renversée – confirmant que la notion d’étendu tient compte non de la surface de contact avec le support mais de la surface délimitée par les points de ce contact – et la nécessité de lui assigner la posture assise dans sa position normale. L’observation des postures critiques – celles qui font apparaître un seuil de rupture d’une posture à l’autre – nous en apprend plus sur le contenu des verbes. L’inclinaison du pilon appuyé contre un support vertical montre qu’entre debout et étendu le seuil est – très approximative- ment, bien sûr – de l’ordre des 40. Un arbre ou un palmier qui reste ac- croché à une liane au moment de sa chute est suspendu. Ce qui donne à entendre que la position suspendue peut être compatible avec l’idée de contact avec le sol si, toutefois, le point de contact n’est pas proprement la base de l’objet, c’est-à-dire la surface sur laquelle repose l’objet. Les animaux aux extrémités tenues pour longues sont debout, en quoi ils diffèrent des batraciens mais se rapprochent de la table. On y trouve les mammifères quadrupèdes – aussi bien grands: jaguar, tapir, 156 FRANCESC QUEIXALÓS cerf, sanglier, que petits: agouti, rats, souris. Au repos allongés sur le sol, ces animaux sont étendus, boka. Parmi eux, certains peuvent adopter la posture assise eka, quand ils reposent sur leur séant, comme le jaguar, le chien, l’agouti. Cette posture secondaire est impossible pour le tapir, le cerf, le sanglier, le fourmilier, le cabiai. De cette classe, tous les animaux à tanière – renard, lapin, agouti – sont aussi dits être assis lorsqu’ils s’y trouvent terrés et dissimulés à la vue. Enfin le jaguar et le puma perchés sur l’arbre sont vus comme suspendus, ruka. Presque à l’autre extrémité de l’échelle zoologique, les mouches – posées – et les grillons s’adjugent aussi la posture canonique debout, à cause, dit-on, de leurs pattes distinctement visibles. Mais la mouche est suspendue quand sa position défie la gravité.

1.2.3 Les plutôt étendus. La natte pour s’asseoir par terre, le grand plat en terre pour cuire le manioc, la galette de manioc séchant sur la pente du toit sont étendus, boka. Une association pour ainsi dire a priori existe entre le métal et cette posture, avant même la considération de la forme et la posture de l’objet métallique. Etendu est aussi la posture d’une masse, comme le manioc râpé, qui s’étale sur son support. On peut voir dans l’idée d’étalement, d’extension, de contact large et intime avec le support, l’explication de l’association de ce verbe aux notions vues plus haut de surface du sol, de savane, de plage – moyennant peut-être l’hypothèse que ces objets, qui constituent notre sol, reposent eux- mêmes sur un support. Il en va de même du chemin. Ainsi que des fleu- ves et rivières, en contraste avec ce que nous avons vu plus haut du marigot, bras de rivière immobilisé par un brusque changement du cours, qui est assis. La stabilité est bel et bien une idée constitutive de ce dernier verbe. Le chemin et le cours d’eau ont en commun non seulement d’être étendus, mais aussi d’être suspendus en posture secondaire. A l’instar de la cavité souterraine perçue comme un tunnel, nul changement de posture évidemment ici, mais un changement de perspective de la part du locuteur. La posture étendu montre le chemin ou la rivière comme une ligne orientée vers quelque part (on ne dit rien sur son orientation précise). Je tenterai ci-dessous une élucidation de cet emploi. Le degré de visibilité relative des extrémités sur lesquelles repose l’animal fait que les différents tatous sont classés dans la posture étendu, posture qui caractérise, par excellence, le groupe des serpents, lézards – gros et petits – et tortues. Dans la taxinomie zoologique ce groupe porte LA POSTURE DU CORPS EN SIKUANI 157 la dénomination pe-bokae-wi, ‘les étendus’. Sous la surface du sol, pour ceux qui vivent dans les terriers – en particulier les tatous –, ils conti- nuent d’être étendus. Du serpent lové on dira qu’il est assis, eka, et du lézard qui redresse la tête qu’il est debout, nuka. La posture étendu est aussi non marquée pour la raie. Dire d’une raie qu’elle est suspendue, ruka, revient à dire qu’elle nage. Il peut sembler insolite que le cafard, dont les pattes sont tout de même très apparentes, s’associe à la posture étendu. Mais les pattes du cafard sont étalées, et leurs points de contact sur le support délimitent, à l’instar de ceux des pattes de la mygale, une surface importante. Enfin les vers et chenilles sont aussi étendus, bien sûr, mais sur la branche leur posture est variable en fonction du support: si le rapport entre le diamètre du corps de l’animal et le diamètre de la branche est nettement inférieur à 1 ils sont étendus, mais s’il est proche de 1 ils sont suspendus, ruka. Une chenille qui sera dite suspendue sur une certaine branche est dite étendue sur un support plan de largeur identique au dia- mètre de la branche. Plutôt que des dimensions relatives des corps en présence, c’est de la gravité qu’il est question ici: à l’équilibre stable s’associe étendu – comme pour assis –, à l’équilibre précaire s’associe suspendu.

1.2.4 Les plutôt suspendus. Tout ce qui est dépourvu de contact avec un support apparaît comme suspendu, ruka. Cela comporte les objets suspendus dans un fluide, air ou eau: avion, pirogue, tronc flottant, etc., et les objets accrochés au-dessus du sol comme toile d’araignée, feuillage, plantes parasites, pont, ainsi que les ustensiles et meubles qu’on accroche aux poutres, murs et toit de la maison: hamacs, arc et flèches, lignes de pêche, sacs. Les plantes grimpantes, telles les lianes, sont également suspendues. A cette classe appartiennent les parties du corps. Y compris le pied, alors que la chaussure, même enfilée, est étendue, boka. A terre, la plupart de ces objets prennent la posture étendu. (Les longilignes peuvent aussi y être debout, nuka – arc adossé, par exemple.) Notons que les racines et les tubercules, dans la terre, sont suspendus. Si le sens abstrait de ce verbe est bien ‘immergé dans’ par opposition à ‘posé sur’, il faut substituer au fluide mentionné plus haut la notion de matière ou milieu, comprenant l’air, l’eau et la terre. De façon tout à fait régulière, les tubercules extraits et posés par terre apparaissent comme étendus. 158 FRANCESC QUEIXALÓS

Il reste les astres et les étoiles, suspendus bien entendu. Mais avec la particularité qu’aussi bien le soleil que la lune et les étoiles sont dits assis, eka, lorsqu’ils se trouvent près de l’horizon. Distinction importante, puisqu’elle intervient dans l’énoncé de l’heure. Revenons au chemin, la rivière et le souterrain ‘suspendus’. Jus- qu’ici nous avons considéré la notion de ‘suspendu’ comme uniquement inscrite sur la dimension de verticalité: elle ne caractérise pas une forme géométrique propre à l’objet considéré, mais seulement la position relative de l’objet dans l’espace, où il se trouve immergé, et non posé. La gravité fait que tout corps non suspendu est borné par en dessous, par la base de contact avec le support. Il est plausible que le non bornage vertical impliqué par l’idée de suspendu puisse être reporté sur la dimension horizontale, où il signifierait l’absence de contact, de délimitation, de confins, dans une direction donnée. La rivière, le chemin et le tunnel en- visagés comme les supports d’un parcours, d’un trajet, seraient donc vus ouverts, ou non bornés, à un bout. On doit peut-être aller plus loin. Si nous effectuons un deuxième report, cette fois de l’espace sur la durée, nous voyons que le temps qui reste avant une échéance, dont on dit aussi, avons-nous vu, qu’il est suspendu, ressemble au chemin et la rivière en ce qu’il est également le support d’un trajet, puisqu’on le parcourt en direction de l’échéance.4 Les habitants des branches, des airs et de l’eau sont en général suspendus. Sauf exception, que je mentionnerai, en dire qu’il sont étendus, boka, équivaut à dire qu’ils sont morts. Les autres postures secondaires sont: les mammifères tels les singes sont, sur la branche, assis, eka, ou debout, nuka (le singe abattu et accroché à une branche reste dans la posture suspendu); la localisation normale des grenouilles arboricoles est sous la branche ou la feuille, où on les dit suspendues; si elles passent au- dessus de la branche, alors elles sont assises; les oiseaux posés se voient attribuer une certaine diversité de postures: les échassiers et autres oiseaux à pattes longues sont debout, mais les oiseaux à pattes courtes sont assis, tels le perroquet; la visibilité des pattes intervient ici, comme le confirme le fait que certains oiseaux, comme le vautour, sont debout sur le sol et assis sur la branche; on explique que perchés ils adoptent une position plus fléchie; dans leur nid ils sont étendus, sauf le geai qui y est

4 A la différence près qu’en toute logique l’échéance devrait représenter une borne. LA POSTURE DU CORPS EN SIKUANI 159 suspendu parce que son nid est une poche suspendue à un fil; les pois- sons inversent, par rapport à la raie, la prééminence des postures; de suspendus comme posture canonique ils passent à étendus dans deux cir- constances hormis la mort: quand ils se tiennent au-dessus de leur nid ou qu’ils nagent près de la berge; ils sont sans doute vus comme touchant le fond de leur ventre. Beaucoup de petits animaux – scorpion, sauterelles, fourmis – apparaissent comme suspendus, ainsi que les parasites du corps. Ces der- niers, hors du corps, sont étendus, même vivants. La posture suspendue des sauterelles va assez de soi: ou l’insecte vole, ou il est sur les feuilles. Plus difficile à justifier semble la posture des autres petits animaux. Ces référents sont perceptuellement peu individualisés, distincts. Il peut y avoir un facteur d’indétermination configurationnelle ici, qui résulterait en une attribution de posture ‘par défaut’: ni assis, ni debout, ni couché. Je terminerai par les larves vivant dans les troncs d’arbre et dont beaucoup sont comestibles. Dans le bois elles sont suspendues, que le tronc soit debout ou à terre. On pense à l’idée d’immersion dans un milieu, évoquée à propos des tubercules. Directement au contact du sol elles sont étendues. Par rapport aux vers et chenilles, elles inversent la pré- éminence des postures. Sur la branche, suspendu et étendu sont possi- bles, sous les mêmes conditions d’équilibre que pour les vers et chenilles.

1.3 Noms versatiles Leurs référents sont les moins déterminés en termes de posture. On le constate à la grande liberté de ces noms pour s’instituer en sujet d’un verbe de posture. Il s’agit, on s’en doute, des noms comportant le trait humain. Le corps humain n’a pas de posture canonique, ou, plus juste- ment, les quatre postures sont une schématisation des postures canoni- ques du corps humain, plaquées sur le reste des êtres qui peuplent le monde. Elles sont, en ce sens, anthropocentriques. On aura remarqué que les mammifères supérieurs sont proches de l’humain sous ce rapport – et, ajouterai-je, les esprits sont identiques à lui. Tout au plus peut-on noter que pour les humains: 1) assis et suspendu sont des attitudes très fré- quemment adoptées – et verbalisées – dans la vie quotidienne, la pre- mière parce qu’elle s’applique à la position accroupi sur les talons, la seconde en raison de l’usage du hamac; 2) la position debout n’est pas per- çue comme particulièrement distinctive des humains; 3) en revanche, toujours pour les humains, être étendu sur le sol est senti comme em- 160 FRANCESC QUEIXALÓS preint d’une certaine anomalie5, dont la mort, nous l’avons vu pour l’ensemble des animés, constitue l’avatar extrême; d’ailleurs, l’emploi de ‘étendu’ comme auxiliaire dans une construction à sujet humain à glissé vers la modalité commisérative:

(2) Aitahibi tsa-bokae. Tsik! yakaba-boka. (Il)EstIvre gérondivant-EtreEtendu onomatopée vomir-(il) EstEtendu ‘Il est ivre, le pauvre! Il vomit, le pauvre!’

2. Bilan sémantique Les verbes de posture indiquent une façon d’être statiquement dans l’espace, à base anthropocentrique.

2.1 Synopsis La première distinction s’établit entre le fait de se trouver ou non en contact avec un support. Le non contact est désigné par un verbe, ruka, ‘suspendu’, le contact par trois verbes. Ces trois verbes se distinguent entre eux par la forme géométrique du corps considéré. Deux d’entre eux indiquent des configurations bien caractérisées par le rapport de grandeur entre la base et la hauteur. Il s’agit de eka, ‘assis’, à rapport proche de 1, et nuka, ‘debout’, à rapport nettement inférieur à 1. Le troisième verbe, boka, couvre à la fois une configuration donnée par un rapport base / hauteur nettement supérieur à 1, et une configuration plus ou moins isotropique.

5 Pour une observation équivalente sur le néerlandais, voir Gibbs et al. (1994) LA POSTURE DU CORPS EN SIKUANI 161

contact – +

base/hauteur

=1  1

< 1 > 1

ruka eka nuka boka ‘suspendu’ ‘assis’ ‘debout’ ‘étendu’

La configuration géométrique des corps est ‘filtrée’ par une perception culturellement déterminée, qui fait que les corps ou objets de la vie réelle ont, en fin de compte, une façon d’être dans l’espace doublement condi- tionnée: par leur géométrie telle qu’elle frappe nos sens et par la position ou l’attitude que, sur des bases socioculturelles, nous leur assignons comme normale. De ce fait les verbes de posture produisent des ramifica- tions sémantiques dérivées, souvent plus abstraites. Je vais reprendre les verbes un par un.

2.2 Suspendu Par l’effet de la gravité, il y a un contact par en dessous qui devrait se faire et ne se fait pas. Le contact ne se fait pas par en dessous parce qu’il se fait partout, quand le corps est porté par la matière même du milieu dans lequel il se trouve immergé. Nous avons rencontré quatre espèces de milieu: l’air, l’eau, la terre, le bois (rien ne dit qu’elles soient les seu- les), portant l’oiseau, le poisson, le tubercule, la larve respectivement. Ou bien le contact ne se fait pas en dessous parce qu’il se fait au-dessus, pour tout ce qui est accroché. On notera la transmissibilité de la posture: quand je dis du grand-père qu’il est suspendu, c’est le hamac qui se trouve de fait suspendu, grand-père repose simplement sur le hamac; mais, naturellement, la posture du grand-père est ce que je veux caracté- riser, pas celle du hamac. Reste la possibilité que le contact par en dessous se fait bel et bien, mais il est trop précaire pour garantir à lui seul la stabilité. D’où tout le jeu entre suspendu et étendu à propos du ver sur la branche, mais aussi l’arbre arrêté dans sa chute par une liane, ou, à nou- 162 FRANCESC QUEIXALÓS veau, grand-père dans son hamac, qui avec les pieds touchant le sol ne laisse pas d’être suspendu. Enfin, le manque de contact fait de ruka une posture ‘ouverte’ à un bout qui, transposée sur l’horizontale, donne l’idée d’échappée dans une direction, et transposée sur la durée donne l’idée de délai.

2.3 Assis Ce sont les idées de compacité, immobilité, stabilité, permanence qui prévalent en eka. Dans ces acceptions abstraites trouvent leur origine des emplois qui, au premier abord, semblent éloignés de la stricte géométrie. Par exemple la position assise de la termitière de savane, pourtant à forme de colonne, s’explique par l’immobilité et la stabilité; le critère de la longueur-visibilité des pattes par la compacité.

2.4 Debout En nuka se combinent les notions de dressé et de longiligne. On se de- mande pourquoi la termitière est assise alors que l’arbre est debout. Il pourrait y avoir une différence subtile entre les statismes de l’arbre et de la termitière. L’arbre est debout en posture canonique – il peut tomber et continuer d’être arbre –; son statisme renferme du changement en puis- sance. La termitière est assise en posture exclusive, car tombée elle de- vient fragments de termitière6; la stabilité prime.

2.5 Etendu Le contact large et / ou intime avec un support caractérise cette posture. L’intimité expliquerait la position non seulement de certains types de ter- rain (savane, plage), mais aussi celle du fer, découlant des effets de la gravité. D’autre part, si l’hypothèse de l’isotropie physique est juste, il en découle que: 1) la mort, associée à la posture étendu même pour une mouche ou une araignée, revient à une déconfiguration de l’être considé- ré; si l’on veut, un cadavre est un objet dépourvu des propriétés qui fai- saient du vivant ce qu’il était aux yeux de l’observateur; 2) la catégorie des étendus est un fourre-tout, la classe-poubelle où entrera de prime abord la posture d’un objet géométriquement et socialement insolite ou inclassable – les microbes, par exemple, dont la connaissance est une

6 Importants culturellement, car utilisés en guise de pierres pour construire les foyers à même le sol. LA POSTURE DU CORPS EN SIKUANI 163 nouveauté –; ce trait, propre à boka en effet, donne à penser que la classification nominale par postures, qui double la classification en genres- classes (cf. Queixalós 1998), est, elle aussi, exhaustive.

3. Postures composites Les verbes de posture, comme d’autres verbes à teneur spatiale, sont à même de fonctionner comme auxiliaires. Dans ce rôle, les verbes de posture peuvent conserver leur contenu spatial, mais la plupart de leurs occurrences les montrent endurant certaines dérives sémantiques qui les rendent aptes à exprimer l’aspect ou la modalité. Nous avons vu en (2) un exemple de ces emplois subduits. Cependant, les fortes restrictions sur la combinatoire d’un membre de la classes des verbes de posture avec un membre de la même classe font penser que les quelques exemples disponibles relèvent davantage de la composition que de l’auxiliarisation. Nous les examinons maintenant. ‘Debout-assis’, nuka-eka, est la posture exclusive de la montagne, à comparer à celle de la colline, ‘assis’. Elle s’applique également à une entité se trouvant en position debout sans être franchement dressée, par exemple pour évoquer l’attitude d’un homme mal en point, voûté, dissi- mulé.

(3) nexatha pina nuka-eka baharaponüyo Alors Citatif EtreDebout-EtreAssis Celui-ci ‘alors cet individu se tenait là, dissimulé, dit-on’

‘Etendu-debout’, boka-nuka, se dit lorsqu’on est par terre avec le haut du corps dressé – l’informateur prit, pour me faire comprendre cela, la position de celui qui fait des flexions des bras au sol, mais dans la phase de bras tendus –, ou bien debout appuyé-incliné contre un support vertical, ou bien le haut du corps couché sur un support horizontal tel une table, les pieds touchant le sol. C’est, en fin de compte, la même posture que la troisième posture composite attestée, debout suspendu, nuka-ruka, à ceci près que la partie haute est retenue, dans l’une par un support en dessous, dans l’autre par une attache au-dessus, comme il apparaît dans l’illustration qui suit. (On notera que la seconde posture peut aussi, nous l’avons vu, se dire simplement ruka.) 164 FRANCESC QUEIXALÓS

boka-nuka nuka-ruka ‘étendu debout’ ‘debout suspendu’

4. Dynamisation Par adjonction du verbe pona, ‘aller’, dans la position post-verbale typique des auxiliaires, il est possible de conjuguer l’idée de déplacement avec les quatre postures de base. Ruka-pona, ‘aller-suspendu’, peut rester proche de sa littéralité, s’appliquant par exemple au déplacement d’un poisson, mais également renvoyer à l’idée d’un déplacement interrompu par une pause, comme celui de quelqu’un qui, étant de passage, rend une courte visite et repart. C’est, évidemment, l’action, et non le référent de l’actant, qui, dans la deuxième acception, se trouve ‘suspendue’ pendant un laps de temps court. Eka-pona, ‘aller-assis’, indique une façon de marcher lente, carac- téristique des personnes âgées ou amoindries. En revanche nuka-pona, ‘aller debout’, fait allusion une démarche à l’allure vive et alerte. Enfin boka-pona, ‘aller-étendu’, signifie, assez littéralement, se dé- placer en rampant, à la façon du serpent. Mais il donne également lieu à l’expression d’un paraprocès (Boons 1987: 31): une notion de déplace- ment se projette sur une configuration parfaitement statique pour donner l’idée d’étendue (la route va de tel endroit à tel endroit). On dit ‘aller- étendu’ d’une chaîne de montagnes, d’une bande de forêt, d’une étendue de terre, pour parler de la portion d’espace qu’elles couvrent.

5. Comment la posture localise les êtres Les langues ne classent pas pour classer, mais pour construire des quanti- tés, pour effectuer des mesures, pour exprimer des utilisations, pour LA POSTURE DU CORPS EN SIKUANI 165 pointer sur un référent ou le suivre à la trace, et – ce que nous avons ici – pour localiser7. En effet, il n’existe dans cette langue aucun type formel distinct pour la prédication locative, laquelle se ramène à une prédication verbale personnelle intransitive. Les quatre verbes de posture sont utili- sés à cette fin. La localisation d’une entité passe moins par son repérage dans un espace où elle s’inscrit, le site, que par la représentation d’un contour, d’une figure, devant venir frapper les sens. En d’autres mots, on ne pré- cise pas où chercher, mais quoi chercher (Levinson 1991: 11). Ainsi, à la question

(4) ika Phauna? ‘où est Phauna?’ on peut répondre

(5) hota raha nuka Ici Assertif (Il)EstDebout ‘il est ici’ en explicitant les coordonnées spatiales, mais il est au moins aussi natu- rel de répondre

(6) baha nuka-he bo Accompli (Il)EstDebout-Miratif Exclamatif ‘il est debout’

(étant entendu que les morphèmes glosés en italiques sont interchangea- bles entre (6) et (7), où l’absence de coordonnés spatiales est, le cas échéant mais pas de façon obligatoire, palliée par un mouvement du regard ou une protrusion des lèvres dans une direction particulière). L’exemple (4) montre une construction interrogative idiomatique. La façon régulière – mais moins fréquente – de poser une question concernant la localisation apparaît en (7), avec l’adverbe spatial hota in- troduit par la particule interrogative de, et la préfiguration d’une posture possible ou plausible exprimée par un des quatre verbes qui nous oc- cupent.

7 C’est la nature grammaticalement parasitique de la classification nominale. 166 FRANCESC QUEIXALÓS

(7) de hota Phauna nuka? Interrogatif Ici Phauna (Il)EstDebout ‘où est Phauna?’

(Le choix de la posture debout indique non seulement que Phauna est supposé se tenir debout, mais aussi qu’on ne s’attend pas à ce qu’il soit en train de se déplacer.) Autres exemples:

(8) de hota xura eka? Interrogatif Ici Perroquet (Il)EstAssis ‘où est le perroquet?’ présuppose que l’oiseau est posé;

(9) de hota hera boka? Interrogatif Ici Pirogue (Il)EstÉtendu ‘où est la pirogue?’ présuppose que la pirogue est amarrée;

(10) de hota bitsabi ruka? Interrogatif Ici Arc (Il)EstSuspendu ‘où est l’arc?’ présuppose que l’arc est accroché sous le toit de la maison. Les exemples suivants ne présupposent rien, puisqu’ils portent sur des référents à posture exclusive:

(11) de hota pübübana eka? Interrogatif Ici NidDeFourmisSp. (Il)EstAssis ‘où est le nid de fourmis sp.?’

(12) de hota Hapa unu boka? Interrogatif Ici Hapa Forêt (Il)EstÉtendu ‘où est la forêt de Hapa?’

Dans l’extrait suivant, un garçon perché sur un arbre cueille des fruits pour sa mère qui l’attend en bas et qu’il ne voit pas. Craignant l’abandon LA POSTURE DU CORPS EN SIKUANI 167

– mais l’ignorant encore alors qu’il vient en effet de se produire – l’enfant se rassure:

(13) abüxü taena eka! Encore MaMère (Il)EstAssis ‘ma mère est encore là’

(quelques événements mineurs surviennent, puis:)

(14) abüxü taena nuka! Encore MaMère (Il)EstDebout ‘ma mère est encore là’

Les postures choisies le sont strictement en fonction de leur plausibilité – contre ‘suspendu’ et ‘étendu’ – puisque, sa mère étant déjà partie, rien ne peut indiquer au garçon l’attitude corporelle exacte de cette dernière. Et le choix, j’insiste, est obligatoire, la prédication de localisation passant nécessairement par l’emploi d’un verbe de posture. Les adverbes ou cir- constants spatiaux sont, eux, facultatifs. On va ainsi de la posture à la localisation. Il est vrai que dans bien des cas, le choix d’un verbe en regard d’un sujet donné fournit plus d’indications sur le site qu’on ne le croirait au premier abord (ceci ne vaut, bien sûr, que pour les noms à posture non exclusive, puisqu’il y a choix). Je me contenterai de rappeler deux exemples: dire d’un vautour qu’il est debout c’est le localiser sur le sol, et en dire qu’il est assis c’est le localiser sur une branche; si dans la forêt je lève la tête en disant qu’une chenille est suspendue, mon interlocuteur cherchera un rameau frêle, mais si je dis qu’elle est étendue il cherchera une grosse branche. En parlant de la figure on décrit, ainsi, le site. La localisation peut se voir comme une existence spatiale. La meil- leure lecture d’une prédication locative avec sujet interprétable comme indéfini est l’existentiel:

(15) halerito ruka LarveSp. (Il)EstSuspendu ‘il y a une larve sp. (dans l’arbre)’

On va même plus loin: de la posture localisée à l’identité. Une question comme (16) – où le verbe est nominalisé – peut très bien s’interpréter 168 FRANCESC QUEIXALÓS comme portant sur la localisation de Phauna, mais l’énoncé concret pro- féré visait à l’identification de Phauna dans un groupe d’hommes debout, immobiles entre deux danses:

(16) de hota Phauna piha-nukae? Interrogatif Ici Phauna Possessif3º-ÊtreDebout ‘lequel est Phauna?’ plus littéralement: ‘où est la station debout de Phauna?’. En sikuani, l’existence spatiale s’exprime plus par une ‘façon de se tenir ou d’être là’ que par un ‘lieu où l’on est ou se tient’. L’emploi de (16) pour identifier un référent s’expliquerait par un tour de pensée, à l’allure assez phénoménologique, qui réduirait la ‘façon d’être’ à la ‘fa- çon d’être là’.

6. Conclusion Les postures sont, fondamentalement, des formes géométriques obtenues par projection d’une figure sur un plan vertical, d’où la distinction entre la posture suspendu et les trois autres. Elles semblent couvrir l’ensemble des êtres, grâce, en particulier, au caractère de fourre-tout que la notion de déconfiguration confère à la posture étendu. Ces projections de la figure sur le plan vertical – sa silhouette, si l’on veut – constituent la pre- mière donnée prise en considération pour la localisation d’une entité, sa projection sur un plan horizontal – ses coordonnées spatiales, ou site – venant secondairement compléter l’information. Nous sommes loin d’avoir épuisé les propriétés originales que cette classe de verbes présente aussi bien dans le lexique que dans la gram- maire. J’énumère rapidement ces propriétés (cf. Queixalós 1998): 1) le paradigme de quatre verbes examiné ici est doublé d’un paradigme de quatre verbes transitifs, eta, nuta, buata, ruta, contreparties causatives des premiers; 2) chacun des deux paradigmes, intransitifs et transitifs, est doublé d’un paradigme de verbes distensifs (Queixalós 2002), qui véhicule des notions quantitatives concernant le procès (aspect) ou les actants (nombre); 3) les verbes de posture – dans leurs versions intransitive, transitive, ten- sive et distensive – jouent aussi le rôle d’auxiliaires verbaux, comme ils le font en d’autres langues telles le néerlandais; l’information qu’ils ap- LA POSTURE DU CORPS EN SIKUANI 169 portent est littérale, c’est-à-dire spatiale, mais aussi aspectuelle ou, plus originalement, modale; 4) à l’intérieur de l’acception spatiale, l’opposition entre auxiliaires intransitifs et auxiliaires transitifs donne lieu à une réorientation posturale que j’appelle aiguillage diathétique: sur la base d’un procès à deux participants directs, les auxiliaires intransitifs décrivent la posture du référent du sujet, et les auxiliaires transitifs celle du référent de l’objet; qu’on en juge

(17) ne-taeya-eka-me Objet1º-Regardant-EtreAssis-Sujet2º ‘tu me regardes (toi assis)’

(18) ne-taeya-eta-me Objet1º-Regardant-Asseoir-Sujet2º ‘tu me regardes (moi assis)’

Ce mécanisme, assez inhabituel dans les langues, en dit long sur la pré- gnance de la posture corporelle dans le codage de l’expérience au sein de la société sikuani.

Références Boons, Jean-Paul (1987). La notion sémantique de déplacement dans une classification syntaxique des verbes locatifs. Langue française, L’expression du mouvement 76, 5-40. Gibbs, Raymond W., Dinara A. Beitel, Michael Harrington & Paul E. Sanders (1994). Taking a stand on the meanings of stand: bodily experience as motivation for polysemy. Journal of Semantics 11, 231-251. Lemmens, Maarten (2002a). Tracing referent location in oral picture descriptions. In Andrew Wilson, Paul Rayson & Tony McEnery (eds.), A Rainbow of Corpora – Corpus Linguistics and the Languages of the World, 73-85. Munich: Lincom Europa. Lemmens, Maarten (2002b). The semantic network of Dutch posture verbs. In John Newman (ed.) The Linguistics of Sitting, Standing, and Lying, 103-139. Amster- dam: John Benjamins. Levinson, Stephen C. (1991). Relativity in spatial conception and description. Ms., International Symposium: Rethinking linguistic relativity, Ocho-Rios (Jamaica), 27 p. 170 FRANCESC QUEIXALÓS

Queixalós, Francesc (1992). Auxiliares de postura corporal en sikuani. II Congreso del CCELA, Lenguas aborígenes de Colombia. Memorias Bogota CCELA, U. de los Andes, n°2, 185-197. Queixalós, Francesc (1998). Nom, verbe et prédicat en sikuani. Louvain: Peeters. Queixalós, Francesc (2002). Sur la distensivité. Bulletin de la Société de Linguistiques de Paris: La pluralité 55-71. Louvain: Peeters. Inverse markers in Andean languages: A comparative view

Willem F.H. Adelaar (Leiden University)

1. Introduction The purpose of the present contribution is to compare the use of inverse markers in the verbal morphology of three unrelated Andean languages: Quechua, Puquina and Mapuche. It will be argued that inverse markers tend to develop as a result of typological convergence among languages with a predominantly suffixing morphology. Inverse markers allow languages with a limited set of personal reference endings (e.g. with subject markers only, or with an incomplete set of endings encoding both an actor and a patient in a transitive relation) to expand their inventory without having recourse to object markers specified for grammatical person. In- stead, the absence or insufficiency of fully specified object markers can be compensated by assigning the role of patient to what is normally a subject or agent marker. Inverse markers are used to indicate such a switch of roles.

2. Function of the inverse marker In the three languages under scrutiny, the inverse marker is used in tran- sitional1 endings encoding two different speech act participants, a subject or agent and an object or patient. At least in one of these languages (Quechua), the encoded object need not coincide with the direct object of a transitive verb and may function as a recipient or beneficiary, so that the use of inverse markers is not confined to transitive verbs in the strict sense. From a strictly morphological point of view, however, there are never more than two categories involved. Therefore, we will refer to the

1 We will conveniently use the term ‘transition(al)’ in relation to subject-object combinations that are encoded in the verb, rather than ‘transitive’, which may cause confusion in this context. Spanish colonial grammarians used to refer to these combinations as ‘transitions’. 172 WILLEM F.H. ADELAAR two participants that can be encoded in a verb form as the ‘subject’ (S) and the ‘object’ (O). The function of the inverse marker is to assign the role of object to a personal reference marker that normally specifies a subject. Character- istically, the use of inverse markers is subject to a hierarchy of grammatical persons, which means that they can only occur when a subject is lower in hierarchy than its object. Of the languages considered here, an explicit grammatical person hierarchy has only been proposed for Mapuche (Salas 1992, Arnold 1996), which will be presented below. It is likely that a similar hierarchy operates in the two other languages, although for Puquina the data are too limited to draw any final conclusions.

3. Inverse markers in Quechua Quechua is a language family, rather than a language.2 Personal reference marking may vary considerably throughout the present-day linguistic varieties (whether languages or dialects) that together make up the Quechua family. In many of these modern varieties, the original system of personal reference marking that must have existed in Proto-Quechua has suffered considerable alterations. In order to appreciate the function of inverse markers in Quechua as it originally was, we have to set out from that proto-language or from any conservative variety that has remained close to the proto-language in this respect. An uncontroversial reconstruction of personal reference marking in Proto-Quechua verbs cannot be achieved because of the difficulty to reconstruct certain endings, namely, the first person subject ending and the combined ending for a first person subject acting upon a second person object. Both endings are highly variable throughout the dialects. How- ever, the general structure and the distinctions that characterize the personal reference system of the Proto-Quechua verb are straightforward. The Central Peruvian Quechua dialects of the Quechua I branch (following the classification in Torero 1964) have retained the essence of the

2 Quechuan languages are found distributed over different South American countries, including Argentina, Bolivia, Ecuador and Peru. For a general overview of these languages see Cerrón-Palomino (1987), Adelaar with Muysken (2004). INVERSE MARKERS IN ANDEAN LANGUAGES 173 original Quechua system of personal reference coding. For the present discussion, we will conveniently draw our examples from the very conservative dialect of Pacaraos (originally spoken on the upper Chancay river in the department of Lima, but now moribund), because it has not suffered many significant mutations in its system of personal reference coding.3 The original Quechua personal reference system is based on a four- fold distinction between first person (‘me’), second person (‘you’), third person (‘neither me nor you’) and inclusive person (‘both me and you’). When personal reference markers encode the possessor of a noun, explicit number (plural) can only be expressed periphrastically. The plural- ity of a verbal participant can also be indicated periphrastically, or it can be expressed by means of specific suffixes, which operate independently from the personal reference markers.4 Each grammatical person can refer to singular or plural entities, except for the inclusive, which always refers to a group of two or more persons. (The inclusive is often called the ‘fourth person’ in the literature on Quechua and Aymaran, but we will refrain from using this term here in order to avoid confusion.) Verbs with transitional endings contain an indication of both a subject and an object. There are no transitional endings encoding a third person object. To put it differently, the personal reference marker for a third person object is null. Personal reference markers referring to an object are always first person, second person or inclusive person. As far as a hierarchy can be established, it appears to be first person > second person / inclusive > third person. No hierarchy can be established between second person and inclusive person (=1st+2nd). Verbal subject markers (and nominal possessive markers) consist of a set of suffixes. However, the verbal set may differ depending on the tense and mood categories with which it is combined. For Pacaraos Quechua, the set of verbal subject markers used in the unmarked (present) and future tenses is shown in Table 1:5

3 For a detailed sketch of Pacaraos Quechua, see Adelaar (1987). 4 In Southern Bolivian and Argentinean Quechua, person and number markers have become fused to such an extent that they can no longer be separated (cf. Adelaar 1995). 5 In other Quechua dialects we find -ni, -nki, -n, -nčik (Ayacucho) or -V́:, -nki, -n, -nči (northern Junín) in the unmarked tense. 174 WILLEM F.H. ADELAAR

present future 1st person -V́y -šaq 2nd person -nki -nki 3rd person -n -nqa inclusive person -nsi -šun Table 1: Verbal subject markers in Pacaraos Quechua

In the past tense, which is indicated by a specific tense marker -rqa-, Pacaraos Quechua has a zero ending for 3rd person. For 2nd person it uses -yki, an ending that is also found on nouns and in nominalized verbs. The transitional endings that encode both a subject and an object are heterogeneous in structure. The ending that encodes a first person subject acting upon a second person object varies considerably depending on the dialects. Most central Peruvian dialects (Quechua I) use a portmanteau suffix -q (or -k), the southern Peruvian and Bolivian dialects (Quechua IIC) use another portmanteau suffix -yki, and Pacaraos Quechua uses a combination of suffixes -mu-V́y.6 There is a special portmanteau ending for the future tense, -šqayki, which has reflexes in most Quechua dialects. Note that the relation of a first person subject acting upon a second person object is not counter-hierarchical and hence does not involve inverse marking. What we do see, however, is a great amount of insecurity in the formation of this ending. In parts of the verbal paradigm of several Quechua dialects, no ending is available for the transitional relationship of a first person subject acting upon a second person object, and a plain first person subject marker is used instead. A first person object in Pacaraos Quechua is indicated by means of the affix -ma(:)-,7 whereas the accompanying subject markers are the expected ones: 2nd S> 1st O -ma-nki, 3rd S> 1st O -ma-n. Note that the subject marker may vary according to tense and mood, and that the two affixes of the transition can be separated by tense markers, adverbializing affixes or nominalizing affixes, as in (1):

6 The original function of -mu- is to indicate a motion towards the speaker or an action performed in a designated location other than that of the speaker. 7 In Quechua II dialects the first person object marker is -wa-, which may also have been the Proto-Quechua form (-ma(:)- < *-mu-wa-). INVERSE MARKERS IN ANDEAN LANGUAGES 175

(1) tapu-ma:-na-n-rayku8 ask-1ST PERSON OBJECT-NOMINALIZER-3RD PERSON SUBJECT-MOTIVE ‘for the sake of his questioning me.’

Although the two transitional relations in question are both counter- hierarchical, there is no sign of inverse marking. In the transition of a third person subject acting upon an inclusive object, -ma(:)- occurs with the inclusive subject marker, but the latter encodes the object instead of the subject. (The ending is -ma-nsi in the unmarked tense of the indicative mood.) This combination is counter- hierarchical, so -ma(:)- acquires the status of an inverse marker, while losing its canonical function as a first person object marker. For the counter-hierarchical transition of a third person subject acting upon a second person object, a special inverse marker is used: -šu-.9 It is combined with the second person subject marker, which then refers to the object and not to the subject. (The ending is -šu-nki in the unmarked and future tenses of the indicative mood, as well as in the potential mood; it is -šu-…-yki with the past tense of the indicative mood and with most nominalizers.) As in the previous cases, the elements of this combination can be separated by other suffixes, as in (2), or be distributed over different phonological word forms, as in (3):

(2) mika-na:-šu-nqa-yki eat-CAUSE.DESIRE-INVERSE-NOMINALIZER-2ND PERSON SUBJECT ‘that you get hungry’

(3) nyi-šu-q ka-nki say-INVERSE-NOMINALIZER be-2ND PERSON SUBJECT ‘He used to say to you.’

To sum up these facts in a different perspective: (a) Quechua has a first person object marker -ma(:)- or -wa- (depending on the dialect). (b) When a first person object interpretation is excluded (because the subject marker overlaps semantically with first person), -ma(:)-/-wa- is

8 The affix -ma(:)- has a long vowel in open syllables. 9 In many Quechua II dialects the inverse marker is -su-. 176 WILLEM F.H. ADELAAR interpreted as an inverse marker, the person of the subject marker is reassigned to the object, and the true subject is third person. (c) When -ma(:)-/-wa- is not available as an inverse marker (because the combination is already in use for a case of straightforward first person object marking), the special inverse marker -šu-/-su- is selected, the person of the subject marker is again reassigned to the object, and the true subject is third person. Traditionally, the Quechua personal reference system has never been treated in terms of a hierarchy and inverse markers. There seem to be good reasons why -šu- has not been recognized as an inverse marker. As a matter of fact, it has developed into a true second person object marker in a number of dialects, viz. in the northern Peruvian dialect of Incahuasi and Cañaris (department of Lambayeque), in the province of Santiago del Estero (Argentina) and, to a lesser extent, in Cochabamba (Bolivia) and in Cajamarca (northern Peru) (cf. Adelaar 1995). Also, the combination *-šu-n, which would be the expected ending for a third person subject acting upon a second person object if -šu- were indeed the object marker for second person, is blocked, because the ending -šun already has another function. It refers to an inclusive future or imperative. However, the inverse marker interpretation has the advantage of showing that the early Quechua speakers apparently constructed their personal reference marking system by using fewer suffixes than the number of categories they intended to express. We will see that similar developments can be assumed for Puquina and Mapuche. Consequently, the inverse marker mechanism appears to play a frequent, if not universal role in the development of complex pronominal reference systems in suffixing languages.

4. The inverse marker in Puquina The extinct Puquina language has only been known through a mostly translated religious text from the early seventeenth century (Oré 1607). Therefore, in the present state of our knowledge, it is not possible to give a full account of the personal reference marking system of Puquina. Only some basic facts can be recovered.10

10 For more information on Puquina, see Torero (2002) , Adelaar (2004), and INVERSE MARKERS IN ANDEAN LANGUAGES 177

The Puquina language shares a number of lexical elements and structural features with the Arawak language family, which is widely distributed throughout the tropical lowlands of South America. The most conspicuous points of resemblance with the Arawak languages can be found in the inventory of personal pronouns and in a set of pre-clitic personal reference markers that indicate nominal possession. On the other hand, personal reference marking on verbs is achieved by means of suffixes and suffix combinations, as in Quechua. This hybrid character of the personal reference system suggests that the Puquina language may have had non-Andean origins but that it adopted some of the features of neighboring languages, such as Aymara and Quechua, by indirect diffusion (cf. Aikhenvald 2002). There are indications that Puquina was syntactically ergative, but this fact does not seem to have had any particular relevance for the morphological make-up of the verb. Oré’s Puquina data, here presented in their original orthography, demonstrate the existence of an inverse marker -s-, which could be combined with the subject markers -k(i) < qu(i), c, gu(i), gue> for first person and -p(i) for second person. Second person with inverse is illustrated in (4), as opposed to (5), where there is no inverse marker:

(4) pampa-cha-gue-s-p-anch level-MAKE-FUTURE-INVERSE-2ND PERSON-DECLARATIVE ‘He will forgive you.’

(5) quichu-gue-p-anch grieve-FUTURE-2ND PERSON-DECLARATIVE ‘You will grieve.’

The following examples illustrate the use of first person with inverse (6) and without inverse (7, 8).

(6) ore-gue-s-c-anch tell-FUTURE-INVERSE-1ST PERSON-DECLARATIVE ‘He will tell me.’

Adelaar and Van de Kerke (in press). 178 WILLEM F.H. ADELAAR

(7) ni-ch baptiza-gue-nch I-AGENT baptize-1ST PERSON-DECLARATIVE ‘I baptize (you)’

(8) yti-n-qui-nch receive-PLURAL-1ST PERSON-DECLARATIVE ‘We receive (it).’

There is no indication of a second person object in baptiza-gue-nch in example (7). Possibly, the presence of a full pronoun with the agent- marking suffix -ch11 leads to such an interpretation. If we assume that the third person in Puquina is lower in hierarchy than first person and second person, all combinations using the inverse marker -s- are counter- hierarchical. The combination of a second person subject acting upon a first person object is attested in the imperative form ore-suma ‘tell me!’. Whether or not the ending -suma contains the inverse marker as well, cannot be told for sure, considering the incompleteness of the material on which this analysis is based. Unlike in Quechua and Mapuche, there are no attested examples in which an inverse marker and a personal reference marker are separated from each other by other affixes.

5. Inverse markers in Mapuche12 The rather elaborate personal reference system of the Mapuche language encodes the parameters of person and number. The number system is based on a three-fold distinction of singular, dual and plural. The personal reference system of the verbs comprises four categories: first person (speaker), second person (addressee), and two third persons (a non- speech act participant previously in focus, and a non-speech act participant that is not previously in focus). The difference between the two third persons only comes to light in transitional endings involving both a

11 In other contexts -ch is an ablative case marker. 12 The Mapuche language, also known as Mapudungun or Araucanian, was the ancient language of Chile. It is nowadays spoken in parts of southern Chile and southern Argentina. For an overview of Mapuche language facts, see Salas (1992) or Adelaar with Muysken (2004); for a detailed grammatical study, see Smeets (2008). INVERSE MARKERS IN ANDEAN LANGUAGES 179 third person subject and a third person object. The amount of contextual salience of the entities referred to as third person determines the choice between two competing endings. Following a practice borrowed from the Algonquian language studies, Arnold (1996) assigns the denominations ‘proximate’ and ‘obviative’ to the two third person categories of the Mapuche verb. She characterizes Mapuche as a language with an opposition between inverse and direct voice, taking into account the morphological make-up of the Mapuche verb in terms of person and number marking, as well as the syntactic behavior of its arguments. As in other languages with an alleged inverse system, Mapuche features a grammatical person hierarchy, which has been defined as 1st person > 2nd person > 3rd person [proximate] > 3th person [obviative] (Salas 1992, Arnold 1996). A drawback of the Mapuche hierarchy is that at least one transitional ending (first person singular subject acting upon second person singular object) contains an inverse marker without being counter-hierarchical, and that such unexpected use of the inverse marker was even more extensive at an earlier stage of the language, as it is today in a particular dialect of the language (Huilliche, see below). The Mapuche verb has two affixes that can be interpreted as inverse markers: -e- and -mu-.13 The personal reference markers in the unmarked (indicative) mood are shown in Table 2:

1st person 2nd person 3rd person singular -(ï)n -(ï)ymi -(ï)y dual -yu -(ï)ymu -(ï)y ((e)ŋu) plural -yiny -(ï)ymïn -(ï)y ((e)ŋïn) Table 2: Indicative personal reference markers in Mapuche

With a third person subject marker, the indication of dual and plural by means of the elements (e)ŋu and (e)ŋïn, respectively, is optional.14 In addition, there is one exclusive object marker -fi-, which indicates an (obviative) third person object, as in (9) and (10):

13Smeets (2008: 163-5) discusses a number of disadvantages of the inverse approach in relation to these affixes. 14 The forms -ŋu, -ŋïn are clitics. The initial e- appears in non-clitical use. 180 WILLEM F.H. ADELAAR

(9) l̯aŋïm-fi-n kill-3RD PERSON OBJECT-1ST PERSON SINGULAR INDICATIVE ‘I killed him.’

(10) l̯aŋïm-fi-y15 kill-3RD PERSON OBJECT-3RD PERSON INDICATIVE ‘He (proximate) killed him (obviative).’

There are special singular subject markers for the hortative-imperative mood (1st person -či, 2nd person -ŋe, 3rd person -pe). The non-singular subject markers of this mood are identical to those of the unmarked in the first person, but slightly different in the second person. The conditional mood, which is used for syntactically subordinate verbs, is characterized by the presence of a marker -l- and has special endings for first person singular (-i) and third person (-e). When co-occurring with inverse or object markers, the mood marker -l- stands between them and the subject markers. The indication of tense is cumulative and can be combined almost freely with person, mood (except imperative) and nominalization. Counter-hierarchical transitional endings involving a third person subject are formed by inserting the inverse marker -e- before a (subject) person marker, to which is then reassigned the role of object. The person markers themselves are followed by an element -mew or -mu. In other contexts mew is a postposition denoting oblique case. This element is reduced to -ew after the singular person markers of the unmarked tense (with elimination of -i of the second person marker), as in (11-13):

(11) ramtu-e-n-ew ask-INVERSE-1ST PERSON SINGULAR INDICATIVE-POSTPOSITION ‘He asked me.’

(12) leli-e-ym-ew watch-INVERSE-2ND PERSON SINGULAR INDICATIVE-POSTPOSITION ‘He watched you.’

15 rd In this combination, the 3 person marker -y is usually not pronounced; l̯ is an interdental lateral. INVERSE MARKERS IN ANDEAN LANGUAGES 181

(13) l̯aŋïm-e-y-ew kill-INVERSE-3RD PERSON INDICATIVE-POSTPOSITION ‘He (obviative) killed him (proximate).’

Inverse transitions can also be expressed in complement clauses and relative clauses, where a nominalized verb in -e-t-ew encodes transitions with a third person subject (obviative when the object is also third person). All the other transitions, including the inverse relation of a second person acting upon a first, are reflected in nominalized verbs ending in -fi-el. (Note that -fi- need not refer to a third person object in this case.) The person identity of the object with nominalizations in -e-t-ew, as well as subject and object with nominalizations in -fi-el, is indicated non- morphologically by an interplay of pronouns and word-order. In order to denote the counter-hierarchical transition of a second person singular subject acting upon a first person singular object, the inverse marker -e- is inserted before the first person singular subject marker, as in (14):

(14) leli-e-n watch-INVERSE-1ST PERSON SINGULAR INDICATIVE ‘You watched me.’

The same form is used for the hortative-imperative mood (‘watch me!’), even though the first person singular subject marker of the hortative- imperative paradigm is not -n but -či, as in (15):

(15) leli-či watch-1ST PERSON SINGULAR HORTATIVE ‘Let me watch!’

When a second person subject acts upon a first person object and either one of them, or both, are non-singular, the inverse marker -mu- is used, instead of -e-. In that case, it is possible to distinguish between a singular, a dual or a plural first person object (-mu-n, -mu-yu, -mu-yiny), but the number of the second person subject remains unspecified. As in Quechua, the combination of a first person subject acting upon a second person object reflects a certain amount of insecurity and inconsistency. When both the subject and the object are singular, the first 182 WILLEM F.H. ADELAAR person dual ending -yu is used in combination with the inverse marker -e-, even though the transition is not counter-hierarchical, for instance, in (16):

(16) leli-e-yu watch-INVERSE-1ST PERSON DUAL ‘I watched you (singular).’

It appears that the function of the inverse marker here is to separate a pair of individuals expressed by the subject marker and to assign the role of object to one of the two (the addressee). Thus, even though -e- is no longer an inverse marker, its manipulative function is still obvious. When either the subject or the object in this transition is non-singular, the subject marker for first person plural -yiny is used in combination with the reflexive-reciprocal marker -(ï)w-. More precise number distinctions can not be expressed, for instance, in (17):

(17) leli-w-yiny watch-RECIPROCAL-1ST PERSON PLURAL ‘We watched you.’ ‘I watched you (two or more).’ ‘We (two or more) watched each other.’

It appears that the ending -(ï)w-yiny originated by semantic extension from a combination that originally had the more restricted function of a first person plural reciprocal form.16 The main source for early 17th century Araucanian (Valdivia 1606), as well as modern dialect data from the Huilliche variety of southern Chile17 (Augusta 1990, Salas 1992), indicate an even less restricted use of the inverse marker -e- in the transition of a first person subject acting upon a second person object. These sources report the use of -e- in combination with a second person subject marker, as in (18):

16 It can be argued on distributional grounds that the suffix -(ï)w- is no longer an instance of the reflexive-reciprocal marker when it is used in a transitional combination (Smeets 2008: 293). 17 The status of Huilliche as a separate language cannot entirely be determined due to lack of data. INVERSE MARKERS IN ANDEAN LANGUAGES 183

(18) elu-e-ymi give.to-INVERSE-2ND PERSON SINGULAR INDICATIVE ‘I/we give it to you (singular).’

We may conclude that in such cases, -e- is no longer an inverse marker functioning under the dominance of a person hierarchy, but rather a kind of universal role reverser.

6. Concluding remarks The discussed data from Quechua, Puquina and Mapuche show a re- markable amount of shared tendencies in the development of their personal reference marking systems. In all three systems, inverse markers appear to play a role. As the case of -ma(:)-/-wa- in Quechua shows, the function of an inverse marker can be a derived function assigned to affixes that originally had a different meaning. The main reason for inverse markers to emerge seems to be a shortage of suitable object markers (no second person object marker in Quechua, neither a first nor a second person object marker in Mapuche). Inverse markers have the advantage of reducing the number of affixes needed for the composition of complex personal reference systems in verbal morphology. The examples also show that the existence of a person hierarchy is not an indispensable prerequisite for the use of inverse markers. Inverse markers tend to function counter-hierarchically, but they can also follow the hierarchy, as occurs in the first person subject to second person object transition in Mapuche. In the latter case, the function of the inverse marker may become that of a universal role reverser. On the other hand, the relevancy of a hierarchy in direct-inverse systems can be deduced precisely from the difficulties surrounding the genesis of markers denoting a first person subject acting upon a second person object in all three languages. As long as there is no explicit second person object marker, a hierarchically restricted inverse marker is of no help for distinguishing between second and third person objects (or null objects), forcing the language user to develop ad hoc strategies. As for the person hierarchy itself, given its predictable character (1st > 2nd > 3rd, etc.), one may wonder if it should not be defined in terms of distance from ego, rather than as a culturally or linguistically determined choice. 184 WILLEM F.H. ADELAAR

A final question to answer is: Why did three unrelated languages develop such similar strategies in order to set up a complex personal reference system, which they probably did not have at an earlier stage of their development? It seems that interaction with the geographically proximate Aymaran languages may have played a role. The Aymaran languages (Aymara and Jaqaru) have closely-knit, complex systems of verbal personal reference marking, which encode both a subject and an object without having recourse to inverse markers or other affixes used in a derived function. The Aymaran personal reference systems are probably the oldest and the most opaque systems to be found in the An- dean region, and they may have served as a model for the neighboring languages. It is significant that these neighboring languages (especially Puquina) were on their way to develop a complex suffix morphology, which was already present in the Aymaran languages. Diffusion from the Aymaran languages to Quechua and Puquina comes as no surprise, but in the case of Mapuche it is problematic, considering the geographical distance between central Chile and southern Peru. Mapuche has assimilated a number of lexical loans from Aymara, but there are no historical records pointing at a particularly close relationship between the two languages in the past. As we have shown, linguistic evidence suggests that some sort of interaction between Mapuche and the other three languages may have existed at a certain point of time, even though the resemblances are structural rather than formal.

References Adelaar, Willem F.H. (1987). Morfología del quechua de Pacaraos. Lima: Universidad Nacional Mayor de San Marcos, Facultad de Letras y Ciencias Humanas. Adelaar, Willem F.H. (1995). Raíces lingüísticas del quichua de Santiago del Estero. In Ana Fernández Garay & Pedro Viegas Barros (eds.), Actas II Jornadas de Lingüística Aborigen (Buenos Aires 15-18 Nov. 1994), 25-50. Buenos Aires: Universidad de Buenos Aires, Facultad de Filosofía y Letras, Instituto de Lingüística. Adelaar, Willem F.H. (2004). Puquina and Callahuaya. In Willem F.H. Adelaar, with Pieter C. Muysken, The Languages of the Andes, 350-62. Cambridge: Cambridge University Press. Adelaar, Willem F.H., with the collaboration of Pieter C. Muysken (2004). The Lan- guages of the Andes. Cambridge: Cambridge University Press. Adelaar, Willem F.H. & Simon C. van de Kerke (in press). La lengua puquina. To appear in Mily Crevels & Pieter C. Muysken (eds.), Las lenguas de Bolivia. INVERSE MARKERS IN ANDEAN LANGUAGES 185

Aikhenvald, Alexandra Y. (2002). Language Contact in Amazonia. Oxford, New York: Oxford University Press. Arnold, Jennifer (1996). The inverse system in Mapudungun and other languages. Revista de Lingüística Teórica y Aplicada 34, 9-48. Concepción (Chile). Augusta, Félix José de (1990). Gramática mapuche bilingüe (re-edition of 1903). Santiago de Chile: Seneca. Cerrón-Palomino, Rodolfo M. (1987). Lingüística quechua. Cuzco: Centro Bartolomé de Las Casas. Oré, Luis Jerónimo de (1607). Rituale seu Manuale Peruanum. Naples. Salas, Adalberto (1992). El mapuche o araucano. Fonología, gramática y antología de cuentos. Madrid: Editorial MAPFRE. Smeets, Ineke (2008). A Grammar of Mapuche. Berlin, New York: Mouton de Gruyter. Torero Fernández de Córdova, Alfredo A. (1964). Los dialectos quechuas. Anales Cien- tíficos de la Universidad Agraria 2,4: 446-78. Lima: La Molina. Torero Fernández de Córdova, Alfredo A. (2002). Idiomas de los Andes: Lingüística e historia. Lima: Instituto Francés de Estudios andinos, Editorial Horizonte. Valdivia, Luis de (1606). Arte y gramatica general de la lengua que corre en todo el Reyno de Chile, con un vocabulario, y confessionario. Lima: Francisco del Canto (reprint 1887, Leipzig: Teubner.)

Part II : Africa

Negation in Dime

Mulugeta Seyoum (Ethiopian Languages Research Centre and Leiden University Centre for Linguistics)1

1. Introduction Every language has mechanisms to express negation. Variant forms of negatives in different languages may be expressed through the position of negative markers, their form, interpretation, and the way in which they are used. Formally, negative morphemes can appear as verbal affixes, free morphemes, or even verb forms (Payne 1985). For instance, Hamar negative imperatives are formed by way of the verb gärä ‘leave, stop’ following the negated lexical verb as in, kumä ‘eat!’ vs. kuman gärä ‘don’t eat!’ (Lydall 1976: 427), while in some other languages such as in Maale there are portmanteau morphemes indicating both negation and aspect (Amha 2001: 223). In some languages, like Lithuanian, the negative marker precedes the verb, whereas in others, like Japanese, it follows the verb (Collberg & Hakansson 1999: 30). Occasionally, languages allow for double negation, while others use single negation exclusively. In Turkish (Asli & Kerslake 2005), the negative marker is inflected according to person, while the verb always remains in a neutral form. In agglutinative2 languages the negative marker is part of the verbal inflection. Since Dime, like most Ethiopian languages, is an agglutinative

1 Fieldwork on Dime was made possible by the Ethiopian Languages Research Centre (ELRC) and by the Endangered languages program of the Dutch Science Foundation (NWO) for recent and previous fieldwork trips, respectively. My informants Tadesse Gelbo and Shiftaye Yisan are native speakers of Dime, aged 23 and 30, respectively. I am grateful to them for their interest and devo- tion to provide me with valuable data. 2 In agglutinative languages the morphemes are joined together relatively “loosely” (e.g. Hungarian). In such kinds of languages, it is usuallyeasyto determine where the boundaries between morphemes are. For instance, in Swahili each bound morpheme carries (ordinarily) only one meaning. Dime morphology also fits this general pattern. 190 MULUGETA SEYOUM language, the verb is inflected for negation. This paper deals with negation in verbal and nominal clauses and in polar and non-polar interrogatives.

2. Negation in Dime3 2.1 Negation in verbal clauses In Dime, negation is always marked by -kay, while the affirmative is unmarked. Compare the following affirmative and negative constructions, illustrated in (a) and (b) below, respectively:

(1) a. nú adéén ‘he comes’ b. nú ad-kay ‘he does not come’

The verb morphology of Dime does not distinguish between perfective and imperfective aspect in negative constructions. This seems to be a specific feature of Dime. The verb only shows the negation marker -káy. From the point of view of word-order typology, Dime belongs to the class of SOV languages, and, more specifically regarding the position of negation, to the sub-type SOV-NEG. Our point of reference here is a survey by Dryer (1988), who studies the position of the sentential negation marker in SOV languages. Most of the 345 SOV languages he considered showed the location of NEG as either SOV-NEG or SO-NEG-V. The Dime negative morpheme is expressed by either -ká, -káy or -k’áy. The variant -ká is a reduced form of -káy and it occurs in non- final clause position affixed to a copula or a main verb. The morpheme -k’áy, whose onset consonant is glottalized, occurs following ejective consonants and the velar nasal (ŋ). This means that the occurrence of each allomorph is conditioned by their environment. In negatives, the aspectual distinctions are neutralized, as shown below: when compared with the non-negative correspondents, it is clear that the aspectual differences do exist in the affirmative.

3 Dime is an endangered Omotic language, spoken in the southern region of Ethiopia by a population of 5,400 individuals (1994 census). Dime is the least studied language among the South Omotic groups. NEGATION IN DIME 191

(2) a. ná íní ád-káy 3SF.SUBJ today come-NEG ‘She does not come today’ b. ná gáim ád-káy 3SF.SUBJ tomorrow come-NEG ‘She will not come tomorrow’ c. ná náái ád-káy 3SF.SUBJ yesterday come-NEG ‘She didn’t come yesterday’ d. wótu gáim wunt’-i-k’áy 1PL.SUBJ tomorrow work-NEG ‘We will not work tomorrow’

The negative forms of the verb wunt’ in (2d) and the paradigm of tí ‘go’ in (3) illustrate that the initial consonant, k, of the negative marker changes to k’ after ejectives and ŋ.

(3) até tí-k’áy ‘I do/will/did not go’ wotu tí-k’áy ‘We do/will/did not go’ yáay tí-k’áy ‘You do/will/did not go.’ yesé tí-k’áy ‘You do/will/did not go’ nú tí-k’áy ‘He does/will/did not go’ ná tí-k’áy ‘She does/will/did not go’ kété tí-k’áy ‘They do/will/ did not go’

After verbs ending in consonants other than ejectives and ŋ, the negative suffix is -káy, as, for example, kété gaaz-káy ‘they will not curse’. The neutralization of the aspectual differences in negative construction is not a general feature of the Omotic languages. For example, Aari has a perfective and imperfective negative marker k- and -y-, respectively (Ben- der 1991: 97). In Dime, refusal is expressed through a slightly different negative construction. As already mentioned and demonstrated in example (3), tense-aspect is generally not expressed in negative verb forms. In the expression of refusal, however, the existential verb déén/déét and the morpheme -tub, which marks future tense, follow the negative marker -k’á(y)/-ká(y), as illustrated in (4). 192 MULUGETA SEYOUM

(4) wótú gáim wunt’-k’a-déét-tub 1PL.SUBJ tomorrow work-NEG-exist-FUT ‘We shall not work tomorrow’ (lit. We are expected to work tomorrow, but we refuse to work)

The structure of the verb in example (4) is complex as it involves two verbs: wunt’- ‘work’ and déét ‘exist’. The final verb dééttub is observed in nominal clauses. All types of negative constructions, such as verbal or non-verbal, independent or a subordinate are characterized by having the suffix -ka or -kay. The negative morpheme -káy occurs following a copula verb, but it should be noted that it is located in the final position of a sentence as shown below:

(5) nú sóó ád-ká dáh-im 3SM.SUBJ here come-NEG stay-ACC ‘He has not come yet.’ (6) nú nááe ád-káy 3SM.SUBJ yesterday come-NEG ‘He did not come yesterday.’ (7) nú kní yi-ká-déé 3SM.SUBJ dog COP-NEG-PF ‘It was not a dog.’ (8) is-ko kní yi-káy 1SG.OBJ-GEN dog COP-NEG ‘I have no dog.’

In negative imperatives or prohibitions, the final vowel of the basic verb is dropped and the special negative marker -kóy, rather than the negative declarative marker -káy, is attached to the verb root, as in the following examples:

(9) Commands Prohibitions yízí ‘Run!’ yíz-kóy ‘Do not run!’ géhé ‘Push!’ géh-kóy ‘Do not push!’ óló ‘Hurry!’ óló-kóy ‘Do not hurry!’ dáhin ‘Wait!’ dáhin-kóy ‘Do not wait!’ wúy ‘Stop!’ wúy-kóy ‘Do not stop!’ NEGATION IN DIME 193

The use of a special negative marker for imperative and optative form is a feature of many Omotic languages. For instance, Haro (Woldemariam 2003: 201) and Maale (Amha 2001: 229) use a special negative marker for imperative and optative forms. The verbal inflection of negative verbs is summarized in the following table:

Affirmative Negative Aspect marker Person (no person marker) 1SG/PL 2/3SG/PL Imperative No person No person -koy marker marker IPF -déé -t -n -ká/káy PF -i -t -n -ká/káy FUT -tub No person no person -ká-déét-tub marker marker Table 1: Verbal suffixes

2.2 Negative Nominal Clauses The negative nominal clause is headed by the negative copula yi- and the negative marker -káy. Equative, attributive, existential, locative as well as possessive negative nominal clauses use yi-káy. In examples (10a-c) the present negative nominal clause is illustrated:

(10) a. nú kní yi-káy 3SM.SUBJ dog COP-NEG ‘It is not a dog.’ b. kní yi-káy dog COP-NEG ‘There is no dog.’ c. is-ko kní yi-káy 1SG.OBJ-GEN dog COP-NEG ‘I have no dog.’

As mentioned earlier, in verbal constructions the negative marker -káy, is added to the main verb. This is illustrated in (11). 194 MULUGETA SEYOUM

(11) kn-ís ád-káy dog-DEF come-NEG ‘The dog doesn’t come.’

The past negative nominal clause is constructed with the sequence yi-ká- déé (COP-NEG-IPF) as shown (12-13) for equative, locative, and possessive nominal clauses. If the negative nominal clause expresses some future action, the morpheme -tub is suffixed, as is illustrated in (14).

(12) nú kní yi-ká-déé 3SM.SUBJ dog COP-NEG-PF ‘It was not a dog/he had no dog.’ (13) kní yi-ká-déé dog COP-NEG-PF ‘There was no dog.’ (14) kní yi-ká-déé-tub dog COP-NEG-PF-FUT ‘There will be no dog.’

In connection with the past negative form illustrated above, two important facts should be noted: first, the morpheme -déé, which was identified as the imperfective aspect marker in main verbs, is used as the perfective aspect marker in negative nominal clauses, as in (12-14). Sec- ondly, when it precedes the perfective marker in negative nominal clauses, and generally in medial position, the negative marker is realised as -ká instead of -káy.Importantly, the negative marker in this context should not be confused with the perfective aspect marker -ká in affirmative past nominal clauses, i.e., déén-ká. The copula is obligatory in negative nominal clauses and in tensed nominal clauses in contrast to non-tensed ones. Polar negative interrogatives, to which we turn in the next section, are not marked for aspect and person. Both the perfective and the imperfective negative forms have the same verbal structure. Moreover, there is no variation in the verbal form due to person, number and gender of the subject. That means no marking at all concerning the number, person and gender of the subject. NEGATION IN DIME 195

2.3 Polar negative interrogatives In polar negative interrogatives, the marker -áá, which is used only with second person subjects in affirmative interrogatives, is attached to all negative interrogative verbs, irrespective of the person of the subject and the aspect of the verb.

(15) até ád-k’áy-áá ‘Didn’t I come?’ wətú ád- k’áy-áá ‘Didn’t we come?’ nu ád- k’áy-áá ‘Didn’t he come?’ ná ád- k’áy-áá ‘Didn’t she come?’ kété ád- k’áy-áá ‘Didn’t they come?’ yá ád- k’áy-áá ‘Didn’t you (SG) come? yesé ád- k’áy-áá ‘Didn’t you (PL) come?’

The following are sentential examples:

(16) ná wunt’-is-im bos-káy-áá 3SF.SUBJ work-DEF-ACC finish-NEG-Q ‘Didn’t she finish the work?’ (17) ná wunt’-is-im bos-káy-áá 3SF.SUBJ work-DEF-ACC finish-NEG-Q ‘Doesn’t she finish the work?’ (18) yá bay-im íts-káy-áá 2SG.SUBJ food-ACC eat-NEG-Q ‘Aren’t you eating the food? (19) kété ná-ó bosin-ká tíŋ-k’áy-áá 3PL.SUBJ water-LOC place-CNJ go-NEG-Q ‘Aren’t they going to the river at all?’ (20) kn-ís gím-á até yin-kó kiyó dog-DEF speak-CNV1 1SG.SUBJ 2SG.OBJ-GEN there k’ót-a dót yá gáa-k’áy-áá et’-á úis-i-n arrive-CNV1 if 2SG.SUBJ eat-NEG-Q say-CNV1 ask-PF-3 ‘The dog asked (the hyena) by saying “If I come down to you, will you not eat me?”

In the remaining part of this section, we discus tag/confirmation questions. This is a type of yes/no question that consists of a declarative clause followed by a “tag” that requests confirmation or rejection of the 196 MULUGETA SEYOUM declarative clause (cf. Payne 1997). The examples in (21-25) question confirmation of a negative statement.

(21) yá kín-im yéf-káy, (yá)yéf-áá 2SG.SUBJ 3MS.OBJ-ACC see-NEG, see-PF:Q ‘You did not see him, did you? (22) šiftaye t’úlim šál-káy, (nú) šál-de́é šiftaye swim can-NEG, 3MS.SUBJ can-IPF:Q ‘šiftaye can not swim, can he?’ (23) šiftaye t’úl-im šál-káy, (nú) šál-í šiftaye swim-ACC can-NEG, 3MS.SUBJ can-PF:Q ‘šiftaye could not swim, could he?’ (24) até kn-im gís’-káy, (até) gís’-í 1SG.SUBJ 3PL.OBJ-ACC beat-NEG, 1SG.SUBJ beat-PF:Q ‘I did not beat them, did I?’ (25) yá t’úlim šál-káy, (yá) šál-áá? 2SG.SUBJ swim can-NEG, 2SG.SUBJ can-Q:2 ‘You can not swim, can you?’

The structure of the verb in the “tag” question is the same as that in regular interrogative clauses. Confirmation questions after affirmative statements are expressed by a copy of the verb followed by the negation marker -káy. The suffix -áá is added to the verb following the negative marker for all persons. Examples:

(26) p’et’ros yín-im madd-i-n, (nú) mad-káy-áá? Peter 2SG.OBJ-ACC help-PF-3 (3MS) help-NEG-Q ‘Peter helped you, didn’t he?’ (27) mante sakiyó déén-ká, (ná) yi-káy-áá? mante there exist-PF (3FS) COP-NEG-Q ‘Mante was there, wasn’t she?’ (28) até dáh-í-t, (ati) da-káy-áá? 1SG.SUBJ late-PF-1, I be late-NEG-Q ‘I’m late, am I not? (29) wó-n k’iy, šál-káy-áá? 1PL.OBJ-DAT go, can-NEG-Q ‘Let’s go, can’t we? NEGATION IN DIME 197

2.4 Negative interrogatives clauses with content question words Negative interrogatives clauses with content question words are suffixed with the negative marker -k’áy. In this clause type, aspect or tense distinctions are not marked on the verb:

(30) ná amóid dime-n tí-k’áy 3SF.SUBJ when dime-DAT go-NEG ‘When is it that she does not go to Dime?’ (31) áyi ád-k’áy who come-NEG ‘Who did not come?’ (32) ameh-id níts-af ád-k’áy how-many-PL child-PL come-NEG ‘How many of the children are not coming?’

To sum up, from the typological point of view Dime negative constructions display special features. It has no aspectual or tense distinction in negative constructions. In other Omotic languages aspect or tense distinctions are made in negative constructions. In some languages like Maale there are portmanteau morphemes which indicates both negation and aspect, while in other languages the negation marker is affixed to verbs either preceding or following the aspect or tense marker as in Nayi (Ephrem 2007: 49). Dime has a single negative marker for different temporal and aspectual values, including the declarative and interrogative. In both perfective and imperfective negative polar and non-polar interrogatives, the negative marker -k’áy/-kay is suffixed to the verb. In negative tag/confirmation questions, the interrogative is marked by -áá for all persons, preceded by the negative marker -kay. The negative imperative is often marked by a special negative morpheme. The final vowel of the basic verb is deleted and the special negative marker -kóy is suffixed to the verb. Using special negative markers for imperative and jussive is a common phenomenon in Omotic languages. Thus, Dime shares this feature with other Omotic languages (cf. 9 above). 198 MULUGETA SEYOUM

Abbreviations used in this paper 1SG first person singular ABL ablative 2SG second person singular INST instrumental 3SM third person masculine singular PL plural 3SF third person feminine singular SUBJ subject 1PL first person plural OBJ object 2PL second person plural NOM nominative 3PL third person plural ABS absolutive IPF imperfective PF perfective FUT future COP copula DAT dative DEF definite ACC accusative RELT relative GEN genitive Q question M male F female

References Asli, Göksel & Celia Kerslake (2005). Turkish: A Comprehensive Grammar. London: Routledge. Amha, Azeb (2001). The Maale Language. CNWS Publications Vol. 99. Research School of Asian, African, and Amerindian Studies, Universiteit Leiden. The Netherlands. Bender, M. Lionel (1991). Comparative Aroid Syntax and Morphosyntax. Afrika und Übersee 74: 87-110. Collberg, Sheila Dooly & Gisela Hakansson (1999). Negative imperatives and the parametric typology of Negation. Working papers 47: 25-37. Department of Linguistics. Lund University. Dryer, Matthew (1988). Universals of negative position. In Michael Hammond, Edith Moravcsik & Jessica Wirth (eds.), Studies in Systematic Typology, 93-124. Amsterdam/Philadelphia: John Benjamins. Ephrem, Dejene (2007). Verb Inflection in Nayi. MA Thesis. Addis Ababa University. Getachew, Ayen (2006). Inflectional Morphology of Kwama/North Mao/. M.A Thesis, Addis Ababa University. Hirut, Woldemariam (2003). The Grammar of Haro with Comparative Notes on the Ometo Linguistic Group. PhD thesis, Addis Ababa University. Horn, Larry (1989). A Natural History of Negation. Chicago: University of Chicago Press. Lydall, Jean (1976). Hamer. In M. Lionel Bender (ed.), The Non-Semitic Languages of Ethiopia, 393-438. London: Oxford University Press. Mulugeta, Seyoum (2008). The Grammar of Dime. PhD Thesis. LOT Publication. Leiden University. Payne, John R. (1985). Negation. In Timothy Shopen (ed.), Language typology and syntactic description 1, 197-242. Cambridge: Cambridge University Press. The morphosyntax of negation in Zargulla

Azeb Amha (Leiden University Centre for Linguistics)

1. Introduction1 Zargulla is the name by which the language examined in the present study is known among linguists and in official documents in Ethiopia. The 1994 national census, for instance, reports that there are 7800 mother tongue speakers of Zargulla. The speakers refer to themselves as Gamo and to their language as Gamotsoto. They use ‘Zargulla’ to refer to the area where they live and in reference to their ritual chief (senior sacrificer) whom they call Zargulla kaati (chief/king). Confusing to the outsider, there is another dominant linguistic group (700,000 people in the 1994 census) who also refer to themselves as Gamo and to their language as Gamotsoto and are known officially by this name. The Zargulla people, however, identify this group as Zeege or ɗaac’e and they refer to the language of these people as Zeegetso or ɗaac’etso. The Zargulla and the ‘official’ Gamo (i.e., Zeegetso or ɗaac’etso) languages are not mutually intelligible. However, some of the members of the two language groups live in the same villages and are bilingual in Zargulla and Gamo. The two groups are agriculturalists in a similar ecological setting but, according to the British social anthropologist Dena Freeman, who did extensive field work in the area, they have different social structures, oral traditions and rituals (cf. Freeman 2006).

1 I would like to thank the Dutch National Science Foundation (NWO), for its financial support under its Endangered Languages Program, which made research on Zargulla possible. I am grateful to Sasha Aikhenvald, Felix Ameka, Maarten Mous and the editor of the volume, Leo Wetzels, for helpful comments. An earlier version of this paper was presented at the 5th International Conference on Cushitic-Omotic Languages, 16-18 April 2008, LLACAN/Inal- co Paris. I thank the participants of the conference for comments and stimulat- ing discussion. All remaining errors and shortcomings are of course my responsibility. 200 AZEB AMHA

The Zargulla language belongs to the East Ometo branch of Omo- tic, whereas Gamo belongs to the North Ometo branch (cf. Fleming 1976). The current administrative division in Ethiopia (introduced in 1992) is mainly based on ethno-linguistic identity. Due to their shared ethnic identity name (Gamo), the Zargulla and North Ometo Gamo are considered the same. The two are part of the same administrative unit, and primary education is given to Zargulla children using the (North Ometo) Gamo language, which has become one of the few written local languages in the area. This will enhance bilingualism in Gamo and may endanger the continuation of Zargulla as a distinct language. Omotic is part of the larger Afroasiatic language family, whose other members are Berber, Chadic, Cushitic, Egyptian and Semitic. Lan- guages that are closely related to Zargulla and are also classified as East Ometo languages are Koorete, Haro, and Zayse and a number of other little-known languages or dialects, such as Garbantsa, Mäle and Ganta. Zargulla and Zayse are the closest within the branch and they are claimed to be dialects of the same language. The two indeed are very similar but there are some lexical as well as morphological differences between them. Zargulla is a tone-accent language with a limited number of monosyllabic words. That is, although high and low tone are readily distinguishable and form some minimal pair contrasts, the distribution of the high tone suggests a lexical rule whereby every disyllabic or multi- syllabic word must have at least one high tone, with obligatory contrast in accentual prominence among the syllables. There are no multi-syllabic words with low tone on all syllables.2 SOV word order in clauses is frequent but variations are also attested. The case roles of subject, object as well as genitive, dative and ablative are morphologically marked on the noun (cf. Yimam 1994). Depending on verbal focus and mood, the subject may also be co-indexed on the verb through agreement suffixes. Ob- ject is not marked on the verb. Verbs are also morphologically marked for tense-aspect, mood and in certain tenses for negation. All nominal and verbal bound morphemes are suffixes; there are no prefixes in the language.

2 In the present study, high tone-accent is marked by the grave accent (´ ), low tone is not marked. NEGATION IN ZARGULLA 201

In the present contribution we deal with clausal negation, which involves at least seven different morphemes used in different tenses and moods in main and dependent clauses.3 Two of the seven negation markers examined in the present contribution are only used in dependent clauses. The remaining five, which exhibit formal correspondences with each other, are used in the main clause. Negation is marked by bound morphemes attached to the verb in the non-past tenses of declarative and interrogative moods and in dependent clauses. In the past tense of declarative and interrogative clauses and in the imperative/optative mood, however, it is expressed by negative verbs. There is partial similarity between the independent negative verbs and negative affixes, suggesting a (historical) link between the two. The purpose of the present work is to address analytical problems emerging from the formal similarity between the bound negative markers and independent negative verbs. Specifi- cally, the question arises whether the bound negative markers reflect morphologization of a syntactic (periphrastic) negative verb phrase and whether the non-inflecting negative verbs in the declarative emerged through simplification of inflectional paradigms. Based on comparison of negation marking in declarative, interrogative and imperative/optative moods in Zargulla and taking note of the situation in related languages and general typological tendencies we suggest an answer to this question. As background for the discussion on negation, we will present in section 2 a brief summary of predicate types in Zargulla and point out how each of these is marked for negation. In section 3, we elaborate on bound and independent morphemes that mark negation and illustrate their use in context. This section is intended to show that the distribution of the negative markers is determined by the predicate type, tense and mood. Sec- tion 4 presents a summary of the interaction of mood/modality and negation. In section 5 we briefly discuss negation in dependent clauses and conclude the paper.

2. Types of verbal predicates in main clauses in Zargulla On morphological grounds, we distinguish three main verbal predicate types in Zargulla. The largest group comprises verb roots which inflect

3 Term-negation, negative quantifiers and adverbs will not be discussed. 202 AZEB AMHA for (present, past, future and progressive) tenses or negation. We will refer to verbs from this class as ‘simple lexical verbs’ in contrast to ‘compound or complex predicates’ and ‘existential-predicative (auxiliary) verbs’ which form the remaining two verb types. The structure of the three predicate types and how negation is marked on them is briefly discussed in sections 2.1-2.3.

2.1 Existential and predicative verbs yese and yene The verbs yese and yene can head a main clause or they may serve as auxiliary verbs, occurring in combination with other lexical verbs. These two are affirmative verbs characterized by the fact that they do not inflect for subject agreement. Moreover, tense and negation are expressed by replacing yese and yene with their corresponding suppletive forms. The verb yese has the lexical meaning ‘exist’ as in (1a) and with a dative-subject it expresses possession (1b).

(1) a. košáll-aa ʔuúnnó-y yésa košalle-LOC malaria-NOM EXIST:PRES ‘There is malaria in Koshalle (i.e. name of a village)’ b. ʔésú-s lagó gaiddeé-tte yésa 3PL-DAT a.lot ox-FOC exist:PRES ‘They have a lot of oxen’ yene is a general predicative verb, comparable to the “BE-verbs” in Eng- lish (2).

(2) a. k’ottó-y ʔánn-aa yéne axe-NOM where-LOC BE:PRES ‘Where is the axe?’ b. yés’s’-í díkk-áwo taí Ɂoótso sing-SS:CNV NEG.VB-3PL:OPT 1SG.SBJ work gallá-tte yéne body-FOC. BE:PRES ‘Let them not sing! I am working’

The past tense form of both yese and yene is yéšše, as illustrated in (3a) and (3b), respectively. While affirmative simple and complex verbs dis- NEGATION IN ZARGULLA 203 tinguish present, past, progressive and future tenses, existential verbs only make a distinction between present and past tense.

(3) a. nú gáde ga mačó bínnatteyéšše 1SG:EXCL:POS land LOC a.lot heat EXIS:PAST ‘There was too much heat in our area’ b. ʔísóy s’oolínté-tte yéšše 3F.SG:SBJ star-FOC BE:PAST ‘It was a star’ (lit. ‘She was a star’)

Moreover, the verbs yese and yene have a common declarative negative form baáʔa ‘there is no X’, ‘there was no X' or ‘(He, she, etc.) is/was not X’, as illustrated in the examples in (4). Tense is not distinguished in negative existential or predicative declarative clauses; subject-agreement and other verbal categories are also not marked (4a,b). Such morphological reduction in the negative is attested in many languages (cf. Miestamo 2005). However, as we will show in section 3.2., a verbal base b-, formally similar to baáʔa, is inflected for subject-agreement in negative interrogative clauses in Zargulla.

(4) a. ɗimáll-a ʔuúnno-y baáʔa ɗimalle-LOC malaria-NOM exist.NEG ‘There is/was no malaria in Dimalle’ b. ʔésí ʔayyéla mala dič'ó baáʔa 3M.SG.SBJ ayyele:ABS SIML tall BE:NEG ‘He is/was not so tall as Ayyele’

2.2 Simple lexical verbs Simple lexical verbs may only be marked for tense, e.g. ɗáy-ínne in (5a) or they could be fully inflected taking tense and subject-agreement as in (5b) when the verb is in focus. Fully inflecting main verbs are derived from complex verbs comprising two verbs (V1 and V2) but the V2 is often not expressed and morphological material associated with it are pronounced as part of V1 (cf. Amha 2007a). For this reason focused verbs such as yew- in (5b) are treated on a par with simple verbs in this work. 204 AZEB AMHA

(5) a. budó ʔekk-í ʔol-í ʔudúla gá-tte fire:ACC take-SS:CNV give.up-SS:CNV mortar interior-FOC ɗáy-ínne throw-PAST ‘He took fire and put it in a mortar’ b. ʔaadé yew-aá-tt-us-éne true come-PROG-FOC-3PL-FUT ‘They are (really) coming’

Like the affirmative declarative past tense marker -ínne in (5a) above, the non-past negative declarative marker, -aáʔa, is directly attached to the verb root, as in example (6).

(6) hínnó-y hamm-aáʔa that:F-NOM go-NEG:NON-PAST:DCL ‘That one (F) does not go/ is not going’

In contrast to the affirmative (cf. 5b), the negative imperfective verb in (6) does not take other affixes such as subject-agreement or focus. In terms of tense-aspect, a situation expressed by a verb root + -aáʔa is viewed as non-complete: depending on context, it can be interpreted as a present tense form, a progressive, or a habitual action. Thus only past and non-past distinction is made in the negative declarative. Past negative declarative clauses are formed using a complex predicate which is discussed in the next section (2.3).

2.3 Complex predicates Complex predicates in Zargulla comprise two verbal elements. The V1 is either a converb (7a-b), a derived intensive or progressive verb stem or an ideophonic verb. V2 may be one of the existential-predicative verbs (as in 7c) or one of a restricted set of simple lexical verbs including directional verbs such as yew- ‘come’, ham- ‘go’ or aspectual verbs such as kes- ‘go out’, ʔol- ‘give up’ which, respectively, express the ‘start’ or ‘completion’ of the state of affairs expressed by V1, illustrated in (7a,b). In example (7) the complex predicates are underlined. NEGATION IN ZARGULLA 205

(7) a. kiítá-y ham-úm boč’oc’-átt-us message-NOM go-DS:CNV present-FOC-3PL:SBJ ʔekk-í yeénne take-SS:CNV come:PAST ‘The message having gone (to them), (the hyenas) brought a present’ b. ʔudúla ga ɗay-í ʔol-í budó mortar:ABS inside throw-SS:CNV give.up-SS:CNV fire ga waatsé-tte huúʔʔ-ínne inside water-FOC pour-PAST ‘After hiding (some of the fire) in a mortar (he) spilled water on the hearth (i.e., to extinguish the rest of the fire so that his friend would not have fire)’ c. gožž-éne-tte-s híyy-áa yéne get_drunk-FUT-FOC-3M.SG say-PROG BE:PRES ‘He is a little drunk’

As mentioned earlier, the past negative declarative is also expressed by a complex predicate comprising the converb form of the negated lexical verb and the negative existential verb baáʔa. Note for example, the structural similarity between the underlined affirmative complex predicate in (7a) and the past tense negative verb in (8).

(8) ʔésí núná haar-í baáʔa 3M.SG:NOM 1PL:EXCL:OBJ rule-SS:CNV NEG:PAST:DCL ‘He did not rule/administer us’

To summarize the present section, the three predicate types in Zargulla declarative clauses are associated with different ways of marking negation. Existential-predicative verbs have a corresponding negative existential verb: baáʔa ‘there is/was not, it is/was not.’ In contrast, the negative of simple or complex predicates may be formed by affixing a bound negative morpheme to the verb or by combining a converb form of the verb and the negative verb baáʔa. Next to baáʔa, there are two other negative verbs: b- and dikk-/dokk- which combine with a simple or complex predicate. The choice of using a bound negative marker or any of the three negative verbs baáʔa, b- or dikk-/dokk- depends on the tense 206 AZEB AMHA and mood/modality value of the clause, as we will show in sections 3 and 4.

3. Negation in main clauses 3.1 Negation in declarative clauses In past affirmative clauses, negation is indicated by the verb baáʔa which is preceded by the converb form of the negated lexical verb. Hereafter, negative verbs are highlighted using boldface letters.

(9) háy taa mišš-í baáʔa this:M-NOM 1SG:OBJ satiate-SS:CNV NEG:PF:DCL yaá-tte-inne gaámmé-y that_say-FOC-PAST lion-NOM ‘“This (food) did not satisfy me” said the lion’

As an existential verb baáʔa does not distinguish tense-aspect; it may be used to express a denial of an event in the past or present. But in combination with a lexical verb it denotes only the past negative. As such, it contrasts with -aáʔa which is directly affixed to the verb root and designates a non-past negative declarative clause (10). The latter is clearly a bound morpheme. Other than derivational affixes such as the causative or the passive (10b), no other inflectional material can intervene between the verb root and -aáʔa, and no other inflectional or derivational affix can follow it.

(10) a. dokkó maahé-y míy-aáʔa cabbage:ABS leopard-NOM eat-NEG:NON-PAST:DCL ‘A leopard does not eat cabbage’ b. ʔúsúní téʔ-útt-aáʔa 3PL:NOM stop-PASS-NEG:NON-PAST:DCL ‘They are not prevented/ They will not be prevented’

The imperfective negative declarative marker -aáʔa is formally similar to the negative existential verb baáʔa. This formal similarity leads to rais- ing the following question: does the verb baáʔa consist of more than one morpheme or is the imperfective negative marker -aáʔa a phonologically reduced part of a single morpheme baáʔa? This issue will be discussed in NEGATION IN ZARGULLA 207 section 4, where we propose the first analysis based on evidence from other parts of the grammar. As we show in the next section, there are other formally corresponding independent negative verbs and bound negation markers in Zargulla.

3.2 Negation in interrogative clauses In the present section we address polar interrogatives only. The negative declarative and negative interrogative differ mainly by the fact that the negative verb in the interrogative indicates person, number and gender distinctions of the subject whereas the negative declarative does not inflect for subject as we showed in sections 2.1-2.3 and 3.1. In negative interrogative clauses past, progressive, and future or present tenses are distinguished. The head in the past and progressive negative interrogative clauses is a complex predicate and subject inflect- tion is marked on the negative verb which occurs as V2. The examples in (11a-c) illustrate subject-agreement distinctions among third person masculine singular, first person plural, and first person singular. The full paradigm of interrogative negative verbs is given in (16).

(11) a. ʔú keelá-y kong-í b-áye 3PL:POS terrace-NOM fill-SS:CNV NEG:VB:PAST:Q-3M.SG gáda land ‘Didn’t their (Woraze people’s) terrace fill the land?’ (i.e. ‘There are many terraces made by them’) b. took’ám-eɗe híyy-útt-a b-áwo took’ame-PL say-PASS-INT NEG:VB:PAST:Q-3PL ‘Aren’t they called took’ame?’ (took’ame = name of a group/descent line) c. harge-í táná hátte hátte ʔaíkk-í disease-NOM 1SG:OBJ now now hold-SS:CNV wóɗ-ó ʔaík-um dákk-í b-ánna kill-INT hold-DS:CNV send-SS:CNV NEG:VB:PAST:Q-1SG ‘Didn’t I divorce her when disease kept catching me frequently and started to kill me?’ (i.e., ‘ I divorced her when I got sick frequently’) 208 AZEB AMHA

The progressive takes the same negative verb (i.e. b-) and similar subject agreement markers as the past tense form but in the former the lexical verb must end in a long vowel -áa as illustrated in (12).

(12) tá ʔoots-áa b-ánna 1SG work-PROG NEG:VB:PAST:Q-1SG ‘Am I not busy working?’

The suffix -áa is also used in the declarative negative to indicate the progressive aspect (13).

(13) tá ʔoots-áa baáʔa 1SG work-PROG NEG:PF:DCL ‘I am not working’

In the future and present tense form, the interrogative negative is marked by -íkk- (14). This morpheme is directly attached to the lexical verb and it is followed by the same subject-agreement markers that are used in the past and progressive negative interrogative forms discussed earlier in the present section.

(14) híkko taí tá ʔaccó ʔul-í INTJ 1SG:SBJ 1SG:POS meat.ACC return-SS:CNV ʔep-íkk-ánna take-NEG:NON-PAST:Q-1SG ‘Well in that case, won’t I/don’t I take my meat back?’

It is possible to leave out the negation markers (b- or -íkk-) from an utterance as a consequence of which the interrogative subject-agreement markers appear directly affixed to the verb. In such cases the subject agreement markers function as the only morphological indicators of negation and interrogation. For example, the sentence in (15) cannot be a declarative clause; it can, however, be ambiguous with the affirmative optative, which takes the same subject-agreement markers as negative interrogative and imperative/optative verbs. It is possible that intonation plays a role in distinguishing between the negative interrogative and imperative/optative interpretation but this has not yet been fully investi- gated. NEGATION IN ZARGULLA 209

(15) gomé maák’k’-iya hí ʔute curse/taboo become-3F.SG that reason ‘Doesn’t it become a taboo because of that?’ Or: ‘Let it become a taboo because of that.’ [Not: ‘It becomes a taboo because of that’]

In (16a) and (16b) we illustrate the full paradigm of past and future/present negative interrogative verbs. Note the alternation of the verb into hang- ~ ham- ‘go’ depending on tense-aspect. Such stem alternation, corresponding to tense-aspect and mood, is observed only for a few verbs (cf. Amha 2007a) and it will not be further discussed here because it is not directly relevant for our analyses.

(16) a. Negative Interrogative, Non-Fut: 1SG. hang-í b-ánna ‘Didn’t I go?’ 2SG. hang-í b-ay(e) ‘Didn’t you go?’ 3F.SG. hang-í b-íya ‘Didn’t she go?’ 3M.SG. hang-í b-áye ‘Didn’t he go?’ 1PL.EXCL. hang-í b-únno ‘Didn’t we go?’ 1PL.INCL. hang-í b-ínno ‘Didn’t we go?’ 2PL. hang-í b-íte ‘Didn’you (pl.) go?’ 3PL. hang-í b-áwo ‘Didn’t they go?’ b. Negative Interrogative, Future/ present: 1SG. ham-íkk-ánna ‘Will I not/Don’t I go?’ 2SG. ham-íkk-ay(e) ‘Won’t you/Don’t you go?’ 3F.SG. ham-íkk-íya ‘Won’t she go/Doesn’t she go?’ 3M.SG. ham-íkk-áye ‘Won’t he go/Doesn’t he go?’ 1PL.EXCL. ham-íkk-únno ‘Won’t we go/Don’t we go?’ 1PL.INCL. ham-íkk-ínno ‘Won’t we go/Don’t we go?’ 2PL. ham-íkk-íte ‘Won’t you go/Don’t you go?’ 3PL. ham-íkk-áwo ‘Won’t they go/ Don’t they go?’

The examples in (11-15) and the paradigm in (16) show that there is no dedicated question marker in negative interrogative clauses. The same holds for positive/affirmative interrogative clauses which also have no special question morpheme. Moreover, intonation patterns in declarative and interrogative clauses do not differ significantly. The main difference between affirmative interrogative clauses and their declarative counter- 210 AZEB AMHA parts is the absence of the focus marker -tte-, which is generally present either on the verb or on one of its complements in affirmative declarative clauses, as in (17a) and (17b), respectively. (The focus marker is formally identical to the masculine copula marker -tte, see example (18)). When none of these two constituents is focused, e.g. when the speaker wishes to highlight the subject, a special construction based on the relative clause form of the verb is used (cf. Amha 2007b). In example (17a), the verb is in focus and it is marked by -tte-. The corresponding interrogative sentence in (17c) is structurally identical to that in (17a), except that the morpheme -tte- is absent.

(17) a. ʔayyele-í šanká hám-á-tte-s-ínne Ayele-NOM low_land go-ThV-FOC-3MS-PAST ‘Ayele went to the low land’ b. ʔayyele-í šanká-tte-s hám-ínne Ayele-NOM low_land-FOC-3MS go-PAST ‘Ayele went to the low land’ c. ʔayyele-í šanká hám-á-s-ínne Ayele-NOM low_land go-ThV-FOC-3MS-PAST ‘Did Ayele go to the low land?’

All of the negative interrogative forms in the paradigm in (16a) may alternatively be replaced by the invariable hang-í baás-wa, which is a nominalized relative clause followed by the interrogative copula morpheme -wa (18).

(18) hang-í b-aás-wa go:PF-SS:CNV NEG:VB:PAST-REL-COP:Q ‘Didn’t I/you/he/she etc. go?’ [lit. ‘Am I/ aren’t you, etc. not one who is gone?’]

The use of the morpheme -wa as a copula marker in non-verbal predict- ative clauses is illustrated in the question and answer pairs in (19), where -wa in the question form contrasts with the affirmative copula markers -tta/-tte in the response to the question. The morpheme -tte also functions as a focus marker (cf. Amha 2008 for a discussion of the copula construction in Zargulla). NEGATION IN ZARGULLA 211

(19) haí ʔoidé-wa hóo ʔoidé-tta / -tte this chair-COP:Q yes, chair-COP:AFF ‘Is this a chair?’ ‘Yes, it is a chair’

That the whole negative interrogative paradigm can be replaced by an invariable nominalized form, such as that illustrated in (18), suggests that the language is developing towards reducing inflectional paradigms. In declarative and interrogative affirmative clauses, full verbal inflection involving subject-agreement is used only when focus is involved. Other- wise, the verb occurs as an invariable form with morphemes marking tense-aspect or negation (cf. verbal types discussed in section 2) or in the relative clause form. Moreover, a fully inflected declarative or interrogative affirmative verb can also be optionally replaced by a nominalized relative clause parallel to that in (18).

3.3 Negation in imperative and optative clauses The negative imperative is formed by a complex predicate consisting of a converb and a negative imperative verb: dókk- or díkk-.

(20) a. ʔep-í dókk-o take-SS:CNV NEG.IMP:VB-2SG ‘(You M/F) don’t take!’ (Compare: ʔepp-á ‘take (2SG)!’) b. ʔep-í díkk-íte take-SS:CNV NEG:IMP:VB-2PL. ‘Don’t take (2PL)!’ (Compare: ʔepp-á-ite ‘take (2PL)!’)

A formally related negative verb is used to express negative wishes (negative optative) for first and third person, as illustrated in (21).

(21) a. ham-í díkk-ánna go-SS:CNV NEG:IMP:VB-1SG. ‘Let me not go!’ b. ham-í díkk-ínno go-SS:CNV NEG:IMP:VB-1PL:INCL ‘Let’s not go!’

As the examples in (20-21) illustrate, the negative imperative/optative verb is affixed with the subject-agreement markers which are also used 212 AZEB AMHA in the negative interrogative (cf. section 3.2). These subject-agreement markers are used with affirmative imperative/optative verbs too. The difference between affirmative and negative optative paradigms is that in the negative, the agreement markers are attached to the negative verb, whereas in the affirmative these are affixed to the lexical verb. Negative and affirmative imperative/optative constructions are illustrated in (22a) and (22b), respectively, using the verbs ham- ‘go’ and ʔepp- ‘take’. The second person negative and affirmative forms (highlighted in boldface) show slight differences in subject co-indexation.

(22) a. Negative Imp/Optative: ham-í díkk-ánna ‘Let me not go’ ham-í dókk-o ‘Don’t go!’ ham-í díkk-iya ‘Let her not go’ ham-í díkk-aye ‘Let him not go’ ham-í díkk-únno ‘Let us not go (EXCL)’ ham-í díkk-ínno ‘Let us not go (INCL)’ ham-í díkk-íte ‘Don’t go!’ ham-í díkk-uússo ‘Let them not go’ ʔepp-ó ‘Let him/her/them take!’ b. Affirmative Imp/Optative: ʔepp-aná ‘Let me take’ ʔepp-á ‘take! (2SG)’ ʔepp-iíšša / ʔepp-íya ‘Let her take’ ʔepp-eésse / ʔepp-áye ‘Let him take’ ʔepp-uúnno ‘Let’s take (EXCL)’ ʔepp-iínno ‘Let’s take (INCL)’ ʔepp-á-ite ‘take! (2PL)’ ʔepp-uússo /ʔeppá-wo ‘Let them take’

Some of the subject agreement markers in the examples in (22), namely the first person plural inclusive and exclusive morphemes -uúnno and -iínno, and one of the three third person plural markers, -uússo, partly resemble their respective independent subject pronouns núní ‘we (EXCL)’, níní ‘we (INCL), ʔúsúní ‘they’. As will be discussed in the next section, a more obvious formal correspondence is observed between independent subject pronouns and verbal agreement markers in focused affirmative declarative clauses. NEGATION IN ZARGULLA 213

4. The interaction of modality and negation In the previous section, we demonstrated the use of five negative markers, i.e., verbal lexemes and affixes that are partially similar. The distribution of the morphemes is related to modal distinctions. Declarative main verbs have sets of negation markers that are different from the ones used in imperative and optative clauses. As for the interrogative, half of its negative markers are formally similar to the ones used in the declarative, whereas the other half correspond to the ones used in the imperative/optative (see the broken lines in the boundaries of the column for negative interrogative). The highlighted segments in Table 1 show the formal similarities among bound negation markers and independent negative verbs and their realization in different tenses and moods.

Negative Negative interrogative Negative declarative Imperative/ Opt Person Non-past Past: Past: Non-past: lexical V + lexical V + lexical V + lexical V + lexical V + 1SG díkk-ánna -íkk-ánna b-ánna 2SG dókk-o -íkk-ay b-ay 3F.SG díkk-iya -íkk-íya b-íya baáʔa -aaʔa 3M.SG díkk-aye -íkk-áye b-áye 1PL:EXCL díkk-únno -íkk-únno b-únno INCL díkk-ínno -íkk-ínno b-ínno 2PL díkk-íte -íkk-íte b-íte 3PL díkk-uússo -íkk-áwo b-áwo Table 1: Negation markers in main verbs in Zargulla

It seems that the five negative markers, díkk-/dókk-, -íkk-, b-, baáʔa and -aaʔa, are derived from two underlying lexical sources. However, the derivational relation among the morphemes in the different moods is not directly apparent. It seems reasonable to assume that the non-past negative interrogative suffix -íkk- is derived from the inflecting negative verb díkk-/dókk- by deletion of the initial consonantal segment. Similarly, the past negative interrogative marker b-+INFL and the past negative declarative verb baáʔa are formally related. The question is what is the direction of the derivation in this case? The first possible analysis is that the invariable declarative negative verb baáʔa is the basic lexical source from which its inflecting counterpart b- in the interrogative evolved. This claim appears to be reasonable since in its use in existential-predicative 214 AZEB AMHA clauses baáʔa is comparable to affirmative verbal lexemes such as yese, yene and yešše, which are also not inflecting for subject as we showed in section 2.1. Alternatively, baáʔa could be derived from the inflecting negative verb b- which we observed in negative interrogative clauses. In this second analysis, baáʔa could be regarded as a complex form comprising b- and the non-past declarative negative marker -aáʔa. This analysis is parallel to what we claimed for the relationship between the negative imperative/optative verb dokk-/dikk- and the non-past interrogative negative suffix -ikk. Language-internal and external considerations suggest that the second analysis is more plausible. That is, baáʔa is derivative of the inflecting negative interrogative verb b-, which is structurally and functionally similar to the negative imperative/optative verb dókk-/díkk. The verbs b- and dókk-/díkk take identical subject-agreement markers and they have the same function of expressing negation. The two differ only in their distribution: b- is used in assertive/propositional utterances (i.e. in de- claratives and interrogatives) whereas dVkk- is used in directives (in imperatives and optatives). The question then is what is the motivation for the emergence of a non-inflecting negative verb (i.e. baáʔa) in the declarative? An answer to this question must consider both a cross-linguistic typological tendency and the language-specific characteristics of Zar- gulla declarative clauses. Language-internal support for our hypothesis comes from the pattern of subject agreement marking in Zargulla. As we showed in section 2.2, the affirmative declarative does either not inflect for subject-agreement at all or, when it is inflected in the case of focused verbs, these agreement forms can be shown to be recently grammaticalized from preverbal independent pronouns. The subject-agreement markers in focused declarative verbs are the highlighted morphemes in the paradigm in (23), which represents an example of the so-called ‘extended paradigm’ representing focused declarative verbs. NEGATION IN ZARGULLA 215

(23) Focused affirmative declarative verb: yewe ‘to come’ táy/tání yeétt-á-tte-t-ínne 1SG:NOM come-EMPH-FOC-1SG-PAST ‘I came’ néní yeétt-á-tte-n-ínne ‘you (sg.) came’ ʔísí yeétt-á-tt-iš-ínne ‘she came’ ʔésí yeétt-á-tte-s-ínne ‘he came’ núní yeétt-á-tt-un-ínne ‘we (EXCL) came’ níní yeétt-á-tt-in-ínne ‘we (INCL) came’ wútúní yeétt-á-tt-it-ínne ‘you (pl.) came’ ʔúsúní yeétt-á-tt-us-ínne ‘they came’

Earlier works on Zargulla and on the related language Zayse have shown that the extended paradigm is a recent innovation, derived from a complex verb form (Amha 2007a) or from a syntactic clause (cf. Hayward 1990, 1991). The subject-agreement markers in this construction are derived from preverbal short pronouns. A piece of evidence for this analysis is the similarity between the independent subject pronouns táy/tání ‘I’, néní ‘you’, etc. and their corresponding verbal subject-agreement markers -t-, -n-, etc. in (23). Moreover, non-cliticized preverbal subject pronouns are still used in the related east Ometo languages Haro (cf. Woldemariam 2003) and in Koorete (Hayward 1982, Mendisu 2008). In contrast to the agreement morphemes in (23), which are used only in focused construction, the subject agreement forms in the negative interrogative and imperative/optative forms (cf. Table 1) are obligatory and they seem to be well entrenched (older) verbal categories. Considering the presence of verbal subject-agreement marking in affirmative and negative declarative clauses in most branches of Omotic, including the North and West Ometo branches, it is probable that the affirmative and negative declarative have lost earlier verbal inflectional material whereas the imperative/optative and negative interrogative kept the agreement markers. Functional (un)markedness and frequency of the declarative could have motivated the simplification in this mood (cf. Givon 1995, Bybee & Hopper 2001, among others, on the important role of markedness and frequency in innovation or recycling of morphological forms). Within the declarative, the affirmative tends to be more affected by innovation than the negative (cf. Poplack (2001) on the retention of older inflectional forms in the negative forms in Canadian French). From the point of view of this general tendency in language, and considering lan- 216 AZEB AMHA guage internal and comparative Omotic, the inflecting negative interrogative and imperative/optative forms of Zargulla seem to represent a more archaic form and the invariable negative declarative baáʔa is a later derivation. The fact that all of the inflecting negative interrogative forms may optionally be replaced by an invariable nominalized relative clause followed by the interrogative copula marker -wa (cf. example 18) suggests that simplification of paradigms is affecting negative verbs as well. One question for the proposed analysis remains to be answered: what is the source of the morpheme -aáʔa in non-past negative verbs and in the negative existential verb baáʔa? This question cannot be fully answered at this point. However, what seems to be a cognate morpheme, -áʔʔa, is found in the south Ometo language Maale. The morpheme -áʔʔa is affixed to verbal roots to express ‘emphatic denial’ or ‘refusal’ as in (24a). In Maale, -áʔʔa may also co-occur with other negative markers: -ib(á)- (perfective negative), -úw(a)- (imperfective negative) or -ind- úw(a)- (future negative) as illustrated in (24b-c), taken from Amha (2001: 224-234).

(24) a. táání pétte baazzi t-áʔʔo maɗ-áʔʔa 1SG:NOM one thing BE-CNV2 do.work-EMPH:NEG ‘I shall not do anything’ b. ʔízá máɗɗ-ó kurs-ib-áʔʔa d-á 3F.SG:NOM work-ABS finish-PF:NEG-EMPH:NEG BE-IPF:Q ‘Didn’t she finish the work?’ c. ʔíntsí wór-ó ʔááɗ-úw-áʔʔa d-á 2PL.NOM river-ABS go-IPF:NEG-EMPH:NEG BE-IPF:Q ‘Aren’t you (PL or SG polite) going to the river at all?’

In the examples in (24b,c), -áʔʔa is not a negative marker but one that identifies the utterance as a (negative) assertion which is then questioned by the interrogative copula verb d- which has scope over the entire prop- osition.

5. Negation in dependent clauses In the present section we briefly discuss two bound negative morphemes which are exclusively used with dependent verbs. The purpose of this section is to show that negation in dependent clauses is only marked by NEGATION IN ZARGULLA 217 bound morphemes affixed to the verb, whereas in main clauses it is marked by negative verbs, and in some tenses, by affixes, as discussed in sections 3 and 4. Further work needs to be done for a more complete account of negation in dependent clauses. In purposive adverbial clauses, negation is marked by suffixing -aattes to the verb of the dependent clause. Such a dependent clause is then obligatorily followed by the complementizer malá. The dependent clause often precedes the matrix clause, although this order can be switched without causing ungrammaticality.

(25) tání neé-s ʔátt-aattes malá ʔaččó 1SG:NOM 2SG-DAT remain-NEG:PURP COMP meat:ACC ʔals-í-tte-t muúnne finish-SS:CNV-FOC-1SG eat:PAST ‘I ate all of the meat so that it will not be left for you’

An affirmative purposive clause that corresponds to the example (25) is given in (26). In such clauses, the verb of the dependent clause is marked by -ádes and, just like the negative purposive clause, it is followed by the complementizer malá.

(26) tání neés ʔátt-ádes malá ʔaččó 1SG:NOM 2SG-DAT remain-AFF:PURP COMP meat:ACC ʔeerés-ette-t muúnne little-FOC-1SG eat:PAST ‘I ate a little bit of the meat so that there will be some left for you’

Dependent clauses which designate events that are simultaneous to the one expressed in the matrix clause are marked for negation by -áčče. The matrix clause itself may be negative (27a) or affirmative as in (27b).

(27) a. nu gawó ʔáde laáʔ-áčče 1PL:EXCL:POS stomach father kiss-NEG:SIML ʔaats-aáʔa let_pass-NEG:IPF:DCL ‘We don’t let our chief pass by without kissing (him)’ 218 AZEB AMHA

b. yee laátso ne ɓók’k’-áčče yés-um VOC chief 2SG open_mouth-NEG:SIML exist-DS:CNV šaató-z-í waákk-éne child-DEF:M-NOM what:BE-FUT ‘Now, chief! If you do not say something, what will happen to the child?’

A negative simultaneous clause marked by -áčče can take the focus marker -tte as well, as in (28).

(28) taa laáʔ-áčče-tte-n ʔol-ínne 1SG:OBJ kiss-NEG:SIML-FOC-2SG give_up-PAST hí-seési baáʔa say-REL:NMZ EXIST:NEG ‘There is no one who says “you did not kiss me’”

Another dependent verb that is directly marked for focus is the affirmative anterior/simultaneous converb.

6. Summary and conclusion The paper addresses negation marking in Zargulla, an Omotic language spoken in Ethiopia. Zargulla has several (independent and bound) morphemes that mark clausal negation. The distribution of negative markers is determined by the predicate type (i.e. auxiliary, simple lexical verb or a complex predicate), tense-aspect, mood and on the status of the clause as a dependent or main clause. Negation is marked by affixes attached to the verb in the non-past tense of declarative and interrogative clauses and in dependent clauses. In past declarative and interrogative clauses and in the imperative/optative mood however, it is expressed by independent negative verbs. There is partial similarity between the independent negative verbs and negative affixes in main clauses, suggesting a (historical) link between the two. We attempted to account for the formal correspondences among these morphemes by identifying the direction of the historical derivation. To this end, we considered language internal evidence as well as material from related languages and views on the interaction of frequency in language use and markedness. The negative verbs in the imperative/optative and the past tense interrogative are dókk-/díkk- and NEGATION IN ZARGULLA 219 b- respectively. We claim that the non-past interrogative negative -íkk- and the negative declarative verb baáʔa, respectively, originate from these two inflecting verbs.

Abbreviations used ABS Absolutive/Accusative case M Masculine AFF Affirmative NEG Negative CNV Converb NMZ Nominalizer COMP Complementizer NOM Nominative COP Copula OBJ Object DAT Dative case PASS Passive DCL Declarative clause PAST Past tense DEF Definite PF Perfective DS Different subject PL Plural EMPH Emphatic POS Possessive EXCL Exclusive PRES Present tense F Feminine gender PROG Progressive FOC Focus PURP Purposive FUT Future tense Q Question/interrogative IMP Imperative mood REL Relative clause INCL Inclusive SG Singular INFL Inflection SIML Simultaneous INT Intensive SS Same subject INTJ Interjection VB Verb IPF Imperfective aspect VOC Vocative LOC Locative

References Amha, Azeb (2001). The Maale Language. Leiden: Research School CNWS, Leiden University (CNWS Publications, vol. 99). Amha, Azeb (2007a). Are -a- and -o- in the indicative verb paradigms of Zargulla nominalizers? In Azeb Amha, Graziano Sava & Maarten Mous (eds.), Omotic and Cushitic Languages Studies: papers from the Fourth Cushitic Omotic Con- ference, Leiden, 10-12 April 2003. Pp. 1-22. Cologne: Köppe. Amha, Azeb (2007b). Questioning forms in Zargulla. In Rainer Voigt (ed.), “From Beyond the Mediterranean”: Akten des 7. internationalen Semitohamitistenkon- gresses, Berlin 13-15 Septemebr 2004, 197-210. Aachen: Shaker Verlag. Amha, Azeb (2008). Gender distinction and affirmative copula clauses in Zargulla. In, John D. Bengtson (ed.), In Hot Pursuit of Language in Prehistory: essays in the four fields of anthropology in honour of Harold Crane Fleming, 39-48. Amster- dam/Philadelphia: John Benjamins. 220 AZEB AMHA

Bybee, Joan and Paul Hopper (2001). Introduction to frequency and the emergence of linguistic structure. In Joan Bybee and Paul Hopper (eds.), Frequency and the Emergence of Linguistic Structure, 1-24. Amsterdam/Philadelphia: John Ben- jamins. Fleming, Harold (1976). Omotic Overview. In Marvin L. Bender (ed.), The Non-Semi- tic Languages of Ethiopia, 299-323. East Lansing: African Studies Center, Mich- igan State University. Freeman, Dena (2006). Who are the Gamo? And who are the D’ache? In Siegbert Uhlig (ed.), Proceedings of the XVth International Conference of Ethiopian Studies, Hamburg, July 20-25, 2003, 85-91. Wiesbaden: Harrasowitz Verlag. Givón, T. (1995). Functionalism and Grammar. Amsterdam/Philadelphia: John Ben- jamins. Hayward, Richard J. (1982). Notes on the Koyra language. Afrika und Übersee 65(2): 211-268. Hayward, Richard J. (1990). Notes on the Zayse Language. In Richard J. Hayward (ed.), Omotic Language Studies, 210-355. London: School of Oriental and Afric- an Studies. Hayward, Richard J. (1991). Concerning a vocalic alternation in North Omotic verb paradigms. Bulletin of the School of Oriental and African Studies 54(3): 535-553. Mendisu, Binyam Sisay (2008). Aspects of Koorete verb morphology. PhD thesis, Fac- ulty of Humanities, Department of Linguistics and Scandinavian Studies, Univer- sity of Oslo. Poplack, Shana (2001). Variability, frequency, and productivity in the irrealis domain of French. In Joan Bybee and Paul Hopper (eds.), Frequency and the Emergence of Linguistic Structure, 405-428. Amsterdam/Philadelphia: John Benjamins. Woldemariam, Hirut (2003). The grammar of Haro with comparative notes on the Ometo linguistic group. PhD thesis, Addis Ababa University. Yimam, Baye (1994). Some aspects of Zargulla morphology. In Bahru Zewde, Richard Pankhurst and Taddese Beyene (eds.), Proceedings of the Eleventh International Conference of Ethiopian Studies, Vol. I, 419-428. Addis Ababa: The Institute of Ethiopian Studies. Possession in Sheko

Anne-Christie Hellenthal (Leiden University Centre for Linguistics)

1. Introduction 1.1 The notion and expression of possession Possession is a notion which seems simple at first, but is semantically complex at closer inspection. At its simplest, possession is the relation between a possessor and a possessed item, the possessum. The central type of relation seems to be one of ownership (permanent possession), but part-whole (inalienable possession), abstract possession, and other relations also fall under the domain of possession (Herslund & Baron 2001: 11).

(1) a. He has a gun (ownership) b. He has long hair, a cube has six sides (inalienable possession) c. He has a problem (abstract possession)

The linguistic constructions used for prototypical possessive relations are often also used for non-possessive notions, such as location or goal. (See Heine 1997: 47, for a list of sources of possessive constructions.) Sheko (Omotic, Ethiopia) has three different constructions which can express a possessor-possessum relation. They go along “the three major dimensions which constitute the principal linguistic realisations of possession: predicative possession, attributive possession and (...) external possession or possessor ascension,” (Herslund & Baron 2001: 4). Each of them is of interest from a typological or historical-comparative point of view.

1.2 Overview Section 2 discusses attributive possession, e.g. possessive noun phrases. The head noun (the possessum) is marked tonally. Then a subgroup of possessive noun phrases is discussed of which the head is formed by baab ‘father’ or bé ‘mother’. The grammaticalization of these morphemes is illustrated and their role for the expression of gender is inves- tigated. Curiously, when a possessive noun phrase headed by baab is 222 ANNE-CHRISTIE HELLENTHAL made definite, bé has to be used. This could be evidence that in the past, the default gender was feminine, whereas currently the default gender is masculine. Grammatical gender is a recurring topic in Omotic studies (e.g. Fleming 1976: 36, Bender 1990, 2007: 736-42). Section 3 compares possessive clauses to existential and locative clauses, and shows that predicative possession in Sheko looks like existential predication. The three types of predicate nominals are similar, which is expected from a typological point of view. According to Herslund & Baron (2001: 9), predicative possession can be divided into two types, of which one has the possessor as a topic and subject, and the other the possessum. Sheko happens to use only one of these types: only the possessum can be the subject in predicative possession. Significantly, the possessum cannot be marked for definiteness. The asymmetry is caused by the properties of the predicative construction, which typically introduces a new participant. Finally, section 4 concerns possessor ascension. It compares the ascension construction and the attributive construction in relation to body part nouns, taking into consideration the discussion about inalienable possession in the related language Dizi (Alan 1976; Claudi & Serzisko 1985). It is shown that possessor ascension is the neutral way of presenting body part nouns and that spatial terms related to body part nouns behave in the same way. Essential in the analysis is the semantic difference of the two constructions involved. In opposition to a possessive noun phrase, which more or less focuses on the possessor, i.e. the whole, possessor ascension basically centralizes the possessed, i.e. the part. Section 5 presents the conclusions of the discussion contained in this paper.

1.3 The Sheko language Sheko (called .ókú nōōgù by the speakers) is an Omotic language of the Majoid branch, also called the Dizoid branch after its better-known language Dizi. It is spoken by approximately 40,000 people (estimate provided by the language consultants). The Sheko people are subsistence farmers, producing coffee and honey for extra income. They live in the forested hills between Mizan Teferi and Tepi, and on the Guraferda plateau in Southwest Ethiopia. The main research area was Boyta, a village close to Sheko town. Some characteristics of the Sheko language are the following: the language displays four level tones, written here v (lowest), v (mid-low), POSSESSION IN SHEKO 223 v (mid-high), and v (highest). Tone plays an important role in person and aspect/mood marking. Nouns can be marked for gender-definiteness in the singular by --s (DEF-M), - (F-DEF). Most inanimate nouns are masculine.

(2) zegu ‘ox’ zeg-n-s ‘the ox’ otì ‘cow’ oyt- ‘the cow’

For plural referents, gender marking consists of a vowel -u for masculine and -i for feminine. These occur only with a handful of words which end in a consonant.

(3) tosa ‘myth’ M tosa-s ‘myths’ ááb ‘fruit’ M ááb-ùs ‘fruits’ túún ‘spring’ F túún-ìs ‘springs’

Gender is also marked in demonstratives, adjectives, relative clauses, and in 3rd pers. pronominal elements. NPs are marked for case, with nominative unmarked. Sheko employs different verbal morphology for final (main) verbs and for converbs, i.e. non-final verbs in a clause chain. Fur- thermore, the final main verbs carry a modality marker which gives information on the type of utterance, i.e. realis declarative, irrealis declarative, negative, imperative/jussive and optative. Sheko is verb-final, and dependent clauses precede the main clause, but next to suffixes the language uses prefixes as well, and some modifiers follow the head noun.

2. Attributive possession 2.1 The characteristics of attributive possession In attributive possession, the possessor always precedes the possessum. Both can be expressed in an NP (4-5), or the possessor can be expressed anaphorically by a possessor prefix, as is shown in table 1 below. The relation between the possessor and the possessum is not necessarily one of ownership, and one could also call this construction an associative or genitive construction. 224 ANNE-CHRISTIE HELLENTHAL

(4) yi-s endriyàs èmà t-k DEM-M Endrias clothes COP-REAL.STI ‘These are Endrias’s clothes.’ (5) ōtì bāācī án-n-kì-b tengì bàtà íì-sok’-u-kì-b-i-s cow skin put-NEG2-be-REL tree.sp on 3pl-sleep-u-be-REL-DEM-M ‘those who didn’t use a cow hide, what they were sleeping on was tengi.’

The head of the NP is marked tonally, as is every head which is preceded by a modifier. The following table shows the tonal changes in disyllabic nouns for all six tonal melodies. The numbers in brackets indicate the tone level of the two syllables of the noun. Tone 4 is replaced by tone 2 and all other tones are replaced by tone 1. Note that four of the six contrastive melodies are neutralized by this replacement.

noun in isolation pre-modified noun kábí (44) ‘axe’ há-kābī (22) ‘his axe’ ət’ì (41) ‘maize’ há-t’ì (21) ‘his maize’ ema (33) ‘clothing’ há-èmà (11) ‘his clothing’ budà (31) ‘pumpkin’ há-bùdà (11) ‘his pumpkin’ kāfà (21) ‘bird’ há-kàfà (11) ‘his bird’ t’ètu (13) ‘pebble’ há-t’ètù (11) ‘his pebble’ Table 1. Tonal changes of head noun

It is of course possible to put more than two NPs together, in which case each modified head undergoes the change in tone. In example (6) the nouns baab ‘father’ and náánú ‘elder brother’ are modified.

(6) dād--s bààb nāānū-rā ō.k-ít- child-DEF-M father elder_brother-ACC call.IMP-PL-STI ‘Call the elder brother of the boys’ father.’

A possessive noun phrase is distinguishable from a compound, even though in compounds the same tonal process takes place. The difference is that in compounds the first noun cannot be made definite or plural or be otherwise modified. This is shown in example (7). (Example (8) proves that íírú ‘rain’ can be made definite when it is not in a compound.) POSSESSION IN SHEKO 225

(7) íírú bèngì-k’à gáydú anga kì-á-k rain year-LOC problem very be-3ms-REAL.STI ‘In the rainy season there are many problems.’ *íír-ń-s bèngì rain-DEF-M year intended: ‘the rainy season’ (in any context) (8) íír-ń-s m-bàtà nea há-k’yār-ū-k rain-DEF-M 1SG.POSS-on firmly.ELAT 3ms-beat-u-REAL ‘The rain drenched me.’ (lit. the rain beat firmly on me)

2.2 Attributive possession with baab ‘father’ / bé ‘mother’

2.2.1 Function and distribution Possessive noun phrases may use baab ‘father’ and bé ‘mother’ as the second element. baab ‘father’ has an allomorph bab and bé ‘mother’ has an allomorph báy ~ béy. The baab / bé construction can be used to express or assert ‘ownership’ (9). It is also used to denote a characterizing property of an entity or group of entities (10).

(9) íncù bààb dàtà -téé-t -ás-a óóc’-á-m wood father near 1sg-go-SS 1sg-3MS-ACC ask-POT-IRR.STI ‘I’ll go to the owner of the wood and ask him.’ (10) yi-nì úń bē tē-k DEM-F flower mother COP-REAL.STI ‘This one has flowers.’ (lit. this is a mother of flower)

Besides in possessive constructions, baab ‘father’ and bé ‘mother’ are employed in a range of other constructions. In fact, one can consider baab and bé nominalizers. They are used to form agent nouns (11) and nominalize time expressions (12, 13) and adverbs (13). These morphemes are also used in Irrealis complement clauses (14) and after case- marked noun phrases (15). Without baab, the word yistà ‘at that’ would be interpreted as referring to time (16) instead of location. In the examples below, baab and bé are written as suffixes, even if the distinction between suffix and separable noun is difficult to draw. 226 ANNE-CHRISTIE HELLENTHAL

(11) wūnk’ù-bààb há-kòb-k stealing-father 3ms-take-REAL.STI ‘A thief took it’ (12) gonà-bēy-rā ē-bàr-kìy-à yesterday-mother-ACC forget-throw_away-be-2sg.Q ‘Did you forget yesterdays’ ? ’ (13) yi-s-k adik’à únà-bàb-kn gərì-tà DEM-M-DAT after long_ago-father-DAT head-LOC2 anga-bàb yāāb kay--s íì-āmān-k very-father man god-DEF-M 3pl-believe-REAL.STI ‘After that, many more people than before believed in God.’ (14) nātā íì-bààb nōōgù-rā í-yáz-m-bàb-rā 1SG 3PL.POSS-father word-ACC 3pl-be.able-IRR-father-ACC anga m-bààs-kì-k very 1sg-want-be-REAL.STI ‘I want very much that they can (speak) their father’s language.’ (15) yi-s-tà-bààb í-ŋ ááb-ara há-ūm-t-á DEM-M-LOC2-father wood:DEF:M-DAT fruit-ACC 3ms-eat-SS-3ms kù-ù-k be.ill-u-REAL.STI ‘He ate berries of that tree and became ill.’ (16) yi-s-tà í-ŋ ááb-ara há-ūm-t-á DEM-M-LOC2 wood:DEF:M-DAT fruit-ACC 3ms-eat-SS-3ms kù-ù-k be.ill-u-REAL.STI ‘Then he ate berries of the tree and became ill.’

Ideophones (17) and forms derived from adjectival verbs (18) cannot feed this nominalization. The forms derived from adjectival verbs are already nominalized by the definiteness-gender marking and function as adjectives.

(17) *óó-bààb IDEO[look_intently]-father POSSESSION IN SHEKO 227

(18) kyanu ts’aa-n-s / ts’aa-n-s-əb kyànù dog become.black-DEF-M become.black-DEF-M-REL dog ‘black dog’ *ts’aans-bààb

Only -bé is acceptable in negative sentences. This might be due to the semantics of the feminine gender, which can be used as a diminutive, whereas masculine is neutral with regard to size. The use of the feminine/diminutive form emphasizes that even the least of what might be expected did not take place.

(19) yír-bē-rā ats-ar-á-kì-k what-mother-ACC give-NEG-3ms-be-REAL.STI ‘He didn’t give anything whatsoever.’

Furthermore, many plant and animal names are compounds with -bé ~ -báy as the second element. In most cases, the first half does not occur without -bé, therefore the names are written as a whole.

(20) kúmbē ‘ant, sp. (tiny)’ c’ínc’úbē ‘mosquito’ háánhanubē ‘bird of prey, sp.’ sántàbē ‘bird, sp.’ wōp’mbē ‘chameleon’ úngúbēy ‘plant with blue flowers, sp.’ írkùbē ‘yam sp. with sharp taste’ óngúbāy ‘yam sp.’ ibē ‘water taro’

Sheko is not the only language showing the grammaticalization of the terms for ‘father’ and ‘mother’. For instance, the cognate forms of Sheko baab / bé show a similar behavior in the Omotic languages Dime and Bench. In Dime, agentive nominals are derived by -bab (with a H tone). The word for ‘father’ in Dime is bábe (Mulugeta 2008: 58). In Bench, the geographical neighbour language of Sheko, -bày occurs in names of plants and animals, bāb and bày are used with the semantics of ‘owner’, and nouns and adjectives can be bases for a nominalization process that suffixes -u-bāb, u.bày (Rapold 2006: 213ff). 228 ANNE-CHRISTIE HELLENTHAL

2.2.2 On gender in constructions with baab ‘father’ / be ‘mother’ As a nominalizer, baab ’father’ and bé ‘mother’ have lost their connota- tion of ‘ownership’ and only convey that the referent of the nominalization is masculine or feminine. The present section returns to the possessive noun phrases with baab ’father’ and bé ‘mother’ in order to show a fascinating grammatical property of possessive noun phrases, namely that masculine definite NPs take the feminine bé instead of baab, while still adding a masculine marker after the definiteness marker. Gender is evidently not straightforward in the case of definite possessive noun phrases. The construction with baab ’father’ and bé ‘mother’ can be plural- ized. In the plural, nothing special happens. Just like other nouns, the masculine takes the suffix -ùs and the feminine the suffix -ìs.

(21) a. eki bààb-ùs money father-M.PL ‘rich men’ b. eki bē-ìs money mother-F.PL ‘rich women’

The ‘mismatch’ between masculine and feminine gender appears when constructions with baab ‘father’ and bé ‘mother’ are made definite. (The category of definiteness is confined to the singular.) When baab ‘father’ is made definite, the construction becomes ungrammatical; instead, bé ‘mother’ plus the masculine definiteness marking must be used. Thus, in table 2 below the odd one out is the masculine definite form.

indefinite definite M eki bààb eki bēy--s money father money mother-DEF-M ‘rich man’ ‘the rich man’ F eki bē eki bēy- money mother money mother:F-DEF ‘rich woman’ ‘the rich woman’ Table 2. Gender in definite possessive noun phrases POSSESSION IN SHEKO 229

Below, two sentential examples are given. Sentence (22) illustrates definiteness with feminine gender, sentence (23) with masculine gender. Evidently, the meaning of -bé ‘mother’ is sufficiently bleached in this context to allow masculine gender agreement.

(22) íncù hà-nì tiirà bēy- t-k wood PROX-F shadow mother.F-DEF COP-REAL.STI ‘This tree here (f) is shadow-giving.’ (23) āāptù bē--s kom-s-rā intoxication mother-DEF-M chief:DEF-M-ACC há-gāsk-ū-k 3ms-insult-u-REAL.STI ‘The drunkard (m) insulted the chief.’

The construction given above represents the only instance in the Sheko language where feminine gender morphology (-bé) is used with both feminine and masculine nouns. Could it be a proof for a historical shift from feminine to masculine as the default gender? Another possible piece of evidence for such a shift is reported for Bench, the geographical neighbour of Sheko. In contemporary Bench, the default gender is masculine, while the plural demonstratives are more similar to the feminine than to the masculine gender (Rapold 2006: 389). While it may be hasty to draw conclusions from these isolated facts, they should be considered elements in the discussion about gender in (Proto-)Omotic. The gender situation for Proto-Omotic is unclear, because present-day Omotic languages vary in their function of gender. For some languages, gender is reported to be non-grammatical, i.e. only some words display inherent gender, based on biological distinctions (Fleming 1976: 36). Other languages have feminine gender as the default gender, e.g. Maale (Amha 2001: 45), whereas yet others, like Sheko, have masculine gender as the default gender.

3. Predicate possession Many languages use the same kind of constructions for existential, locative and possessive predicate nominals, because all express a stative situation in which something is situated with respect to a location (Payne 1997: 127). In other words, location, being a fundamental and concrete 230 ANNE-CHRISTIE HELLENTHAL notion, is a basic ingredient of possession (Herslund & Baron 2001: 22). In the case of possessive clauses the location, i.e. the possessor, is often animate.

3.1 The possessive clause in Sheko In Sheko, the possessive clause best compares with an existential clause. Existentials make use of the verb ki (L) ‘be present, exist, live’. An example of an existential clause is given in example (24).

(24) gyanu kì-á-k coffee be-3ms-REAL.STI ‘There is coffee.’

In a possessive predicate, the possessor NP is in the dative case and the possessum NP occurs as the subject of the predicate 'be present, exist'.

(25) gyanu íì-k kì-á-k coffee 3PL-DAT be-3ms-REAL.STI ‘They have coffee.’ (lit. coffee exists to them) (26) bāà yí-nà.à-k kì-tà, ... work 3FS.POSS-husband-DAT be-COND ‘If her husband has work,...’ (lit. if there is work to her husband)

In Sheko, the dative case is used on NPs expressing Benefactive and Re- cipient roles (27). According to Belle & Van Langendonck (1996, introduction) the dative marks the target or ‘pole’ to which an action is oriented. The target can be a goal in a case of transfer or an interested party which experiences the outcome of an action to his advantage or detri- ment.

(27) ii--s yi-z-k’à kì-b yāb-m-s-k house-DEF-M DEM-M-LOC be-REL man-DEF-M-DAT m-ba-a-m 1pl-work-POT-IRR.STI ‘We will work for the man who lives in that house.’ POSSESSION IN SHEKO 231

A locative clause is also existential in Sheko and contains a noun phrase marked by one of the case markers -k’à (for ‘containment’) or -tà (for a more general location).

(28) goydù kar-k’à há-kìy- -sēē-k guereza_monkey forest-LOC 3ms-be-DS 1sg-see-REAL.STI ‘I saw a guereza monkey in the forest.’ (lit. there was a guereza in the forest; I saw it)

In locative expressions using a body part noun both dative and locative case are present (29). Note that the locative expression as a whole must be marked by locative case.1 The locative expressions are discussed in section 4 on body part nouns/ possessor ascension.

(29) ēēz--s bāc--s-k bow-k’à kì-í-k cat-DEF-M bed-DEF-M-DAT belly-LOC be-3fs-REAL.STI ‘The cat is under the bed.’

3.2 Definiteness marking and the function of predicative possession According to Herslund & Baron (2001: 9), predicative possession branches into HAVE-like possession, i.e. a construction in which the possessor is the topic and grammatical subject, and BELONG-like possession, i.e. a construction in which the possessum is the topic and grammatical subject. This distinction is valid cross-linguistically (Heine 1997: 33). Sheko has only a BELONG-like construction.

1 It is ungrammatical to use only a dative case to mark a place in Sheko, as is demonstrated in the sentence below. This is unlike the cognate in Dizi, where a NP marked by -kn can be interpreted as denoting a place (Claudi & Serzisko 1985: 149). *bac-n-s-kn kì-á-kə bed-DEF-M-DAT be-3ms-REAL.STI intended: ‘It/he is at the bed’ 232 ANNE-CHRISTIE HELLENTHAL

In the predicative possession construction, only the possessum can be the grammatical subject. It is possible to change the order of possessor and possessum, but this does not affect the grammatical marking (cf. example (25), repeated here as (30) for convenience, and (31) below).

(30) gyanu íì-k kì-á-k coffee 3PL-DAT be-3ms-REAL.STI ‘They have coffee.’ (lit. coffee is to them) (31) m-bāād--s-k gyanu p’úc’á kì-á-k 1SG.POSS-younger_sibling-DEF-M-DAT coffee a_lot be-3ms-REAL ‘My brother has a lot of coffee.’

Interestingly, the subject cannot be marked for definiteness, as shown in example (32). Asymmetry in definiteness between possessor and possessum is to be expected, although an indefinite possessum in a BELONG- construction may not be usual (cf. Heine 1997: 30).

(32) ?*gyan-n-s m-bāād--s-k kì-á-k coffee-DEF-M 1SG.POSS-younger_sibling-DEF-M-DAT be-3ms-REAL intended: ‘The coffee is to my brother’, i.e. ‘the coffee belongs to my brother.’

The systematic absence on the subject of definiteness marking is related to the type of construction and its function. Possessive predication typically asserts possession, whereas in attributive possession the possession is typically presupposed (Heine 1997: 26). Thus, in possessive predication the possessum is presented as a newly introduced referent, just like existential predication also characteristically introduces a new referent and is not normally used to provide given or known information. If one wants to present the possessor as the grammatical subject, it is possible to use a copula-construction, with which it is possible to have either the possessor (33) or the possessum (34) as the topic. In this construction, the subject is easily marked as definite, hence it can refer to known, topical referents.

(33) m-bāād--s gyanu bààb t-k 1SG.POSS-younger_sibling-DEF-M coffee father COP-REAL.STI ‘My brother is a coffee-owner/ owns coffee’ POSSESSION IN SHEKO 233

(34) gyan-n-s yi-s kéta mèngistì-k-bààb t-k2 coffee-DEF-M DEM-M all government-DAT-father COP-REAL.STI ‘All this coffee belongs to the government’ (lit. is father of ‘to the government’)

Note that, strictly speaking, the above construction is not a case of predicative possession, but of equation. It is only the copula complement, i.e gyanu bààb or mèngistìkbààb, which involves an NP expressing a possessive relation. This type of NP is discussed in section 2.2.

4. Body part nouns/ possessor ascension In the past, there has been some controversy about inalienable possession in Dizi, the closest relative of the Sheko language. While Allan (1976) stated that body parts are inalienably possessed, this is contested by Claudi & Serzisko (1985), who claim that the Dizi possessive constructions involving body part nouns represent the phenomenon of possessor promotion in which the possessor occurs with locative case. Recently, the idea of possessor promotion or possessor ascension has come under attack itself. More precisely, the underlying assumption that the alienable (‘normal’) construction and the inalienable (‘promoted’) construction have the same meaning appears not to hold (Chappel & McGregor 1996: 7). The inalienable construction expresses that the person, i.e. the whole, is affected, whereas the alienable construction does not take the whole into account but focusses on the part. Since the semantics are different, both constructions are equal and a speaker can describe a situation with regard to the whole or the part by choosing one or the other. In other words, discourse features play a role in the choice between a ‘normal’ possessive and a ‘promoted’ possessive construction. The term possessor ascension is used here as a label for the construction described in section 4.1 below.

2 Without -baab one gets a benefactive/recipient reading. gyan-n-s yi-s kéta mèngistì-kn tə-kə coffee-DEF-M DEM-M all government-DAT COP-REAL.STI ‘All this coffee is for the government’ (e.g. to be given as a form of taxes) 234 ANNE-CHRISTIE HELLENTHAL

The following section shows that possessor ascension is the neutral way of presenting body part nouns in Sheko. It also illustrates the different semantics of possessor ascension and attributive possession constructions with examples.

4.1 Possessor ascension in Sheko Possessor ascension occurs not only with persons and their body parts, but also with inanimate possessors and their parts. These parts are often expressed by the same body part noun, as in the examples (35-36). Body part nouns in Sheko occur almost never in a possessive noun phrase, but nearly always in a construction with dative case marking on the possessor (35-36). Thus, a construction with possessed body part nouns can be compared with a predicative clause denoting a possessive relation (37), also illustrated in section 3 above.

(35) baka--s-k a.u ān-á-k stool-DEF-M-DAT leg be_broken-3ms-REAL.STI ‘The stools’ leg is broken.’ (36) endriyàs-k a.u ān-á-k Endrias-DAT leg be_broken-3ms-REAL.STI ‘Endrias’ leg is broken.’ (37) endriyàs-k mèkinì kì-á-k Endrias-DAT car be-3ms-REAL.STI ‘Endrias has a car.’ (lit. to Endriyas there is a car)

Word order may be changed in predicative possessive clauses, but not in possessor ascension. Compare example (36) with (38).

(38) *a.u endriyàs-k ān-á-k leg Endrias-DAT be_broken-3ms-REAL.STI intended ‘Endrias’ leg is broken.’ possible with benefactive interpretation ?‘a leg has been broken for Endrias’

Let us first observe that, unlike in other languages, in Sheko and Dizi only body part nouns are expressed with the inalienable possessive construction, whereas in most other languages, according to Payne (1997: 105) even in all other languages, also kinship terms are included. The POSSESSION IN SHEKO 235 two examples below show that Sheko encodes both ordinary (alienable) things and kinship terms by possessive noun phrases.

(39) endriyàs byàk’ ān-á-k Endrias spear be_broken-3ms-REAL.STI ‘Endrias’ spear is broken.’ (40) endriyàs dàdù sg-ítî Endrias child see-2pl.Q ‘Have you seen Endrias’ child?’

However, Sheko is not alone in treating kin and body parts differently. In their typological survey, Chappel & McGregor (1996) conclude that languages differ regarding the categories they treat as inalienable. Kin, body parts, bodily fluids and spatial relations terms may all be viewed as inalienable, but languages may consider only a subset as inalienable. Like Sheko, many Australian languages treat body parts but not kin as inalienable (Dixon 1980: 293). In other languages, such as Ewe, kin and spatial terms but not body parts are inalienable (Ameka 1996: 827f). Secondly, a construction employed with inalienable possession is usually morphologically less marked/complex than a construction employed with alienable possession. However, the reverse holds in the Dizoid languages, as was already noted by Claudi & Serzisko (1985: 134). In Sheko, the construction with body part nouns makes use of a dative case marker on the possessor noun phrase, whereas in constructions with other nouns the two noun phrases are juxtaposed without in- tervening phonological material. The ‘markedness’ of the inalienable construction in Dizi was one of the reasons why Claudi and Serzisko analysed it as possessor promotion. However, Claudi and Serziskos’ analysis of possessor promotion to a locative case is equally not in line with what one usually finds in languages: possessors are commonly “promoted to” a direct object or an indirect object, not to a locative (1985: 141). The Dizi case marker in question is -kŋ. Without going into much detail, the latest description of Dizi gives -kŋ as a genitive, not a locative, and -is as a dative case. For an overview of Dizi case marking, see Beachy (2005). While a genitive or dative case are frequently attested in possessive constructions in the languages of the world, the marking of inalienable possession by a case marker in the Dizoid languages remains 236 ANNE-CHRISTIE HELLENTHAL atypical. Therefore, future research should look for an explanation of how this came to be.

4.2 Semantic differences between possessor ascension and attributive possession As was said above, body parts often occur in the possessor ascension construction. Two more Sheko examples are given below. Sentence (41) is about traditional Sheko marriage customs and sentence (42) is taken from a fable.

(41) yí-nà.wà-k ááb-a séé-r-í-k’yá-m 3FS.POSS-husband-DAT eye-ACC see-NEG-3fs-leave-IRR ‘She didn’t see her husbands’ face.’ (42) ás-k éd-k’à yí-bàr- twèètwèè ás-k 3MS-DAT mouth-LOC 3fs-throw-DS IDEO 3MS-DAT foori-k’à há-gé-b-àà-s-tà throat-LOC 3ms-say-REL-PROX-M-LOC2 ‘She threw [the hot pebble] in his mouth, and while it said ‘tweet- wee’ in his throat (while his throat got burned)...’

There are basically two contexts in which body parts occur in attributive possessive noun phrases. The first context is where the body part is alienable, i.e. there is no part-whole relation between the possessor and the possessum, but a different one, e.g. a relation of ownership. Thus, the bone in (43) is not part of the body of the speaker, but it is an animal bone which the subject had given to the addressee to eat. Another example is (16), repeated here as (44) for convenience. The sentence tells about a tanned cow hide, not about the skin of a living cow.

(43) -ūūs--s-ā āts- yí-gē- 1SG.POSS-bone-DEF-M-ACC give.IMP-STI 3fs-say-DS ‘ “Give my bone,” she said...’ (44) ōtì bāācī án-n-kì-b tengì bàtà cow skin put-NEG2-be-REL tree.sp on íì-sok’-u-kì-b-is 3pl-sleep-u-be-REL-DEM.M ‘those who didn’t use a cow hide, what they were sleeping on was tengi.’ POSSESSION IN SHEKO 237

The second context in which a possessive noun phrase is used, places emphasis on the possessor. Example (45) below makes this very clear: only clause (b) can follow (a) as an explanation, (c) cannot. It is of course possible to use a possessor ascension construction, but then again the semantics change (46).

(45) a. wosa hà-z -kū.ū-kā -ts’àf-ù-k letter PROX-M 1SG.POSS-hand-INSTR 1sg-write-u-REAL.STI ‘I wrote this letter by my (own) hand.’ b. ts’àhafì--s nā-ŋ ts’af-ar-á-kì-k clerk-DEF-M 1SG-DAT write-NEG-3ms-be-REAL.STI ‘The clerk didn’t write it for me.’ c. *-kòmpùtèrì-kā ts’af-en-kì-k 1SG.POSS-computer-INSTR write-NEG:1sg-be-REAL.STI ‘I didn’t write it on the computer.’ (46) wosa hà-z nā- ŋ kú.ú-ka -ts’àf-ù-k letter PROX-M 1SG-DAT hand-INSTR 1sg-write-u-REAL.STI ‘I wrote this letter by (my) hand.’ (not on the computer)

Another example is given in (47).

(47) í-gāyd--s yáát--s-əb há-fòòt-àb-rā 3PL.POSS-problem-DEF-M be_big-DEF-M-REL 3ms-become-REL-ACC -āāb-kā -sēē-k 1SG.POSS-eye-INSTR 1sg-see-REAL.STI ‘I saw with my own eyes that their problem is enormous.’

Notice that in the examples (45a) and (47) the possessor is the same as the subject/agent of the verb. It may not surprise that a reflexive in Sheko makes use of a possessor prefix and the noun for ‘head’ (48). Here too, the sentence is not about the part (the head) but very much about the whole, i.e. the possessor, who is at the same time the subject/agent. Compare (48) with the idiomatic (49), which is used as a warning for un- ruly children.

(48) hā-grì kóót- 2SG.POSS-head watch.IMP-STI ‘Watch (it) yourself’/ ‘Look after it yourself’ 238 ANNE-CHRISTIE HELLENTHAL

(49) yē-k gərì kóót 2SG-DAT head watch.IMP ‘Watch your head’ (i.e. ‘Beware’)

Thus, it appears that possessor ascension is an unmarked way to talk about possessed body parts, whereas a possessive noun phrase with body parts puts emphasis on the possessor or indicates that the body part is alienable.

4.3 Spatial terms In Sheko, as in many other languages, most spatial terms (locational nouns) are related to body parts. A table is given below.

gərì ‘head’ gərì-k’à ‘on top of’ bow ‘belly’ bow-k’à ‘in, under’ íú ‘side’ í-tà ‘at the side of, near’ adi ‘footprint’ adi-k’à ‘after, behind’ sān ‘forehead’ saantà ‘in front of, before’ Table 3. Locational body part nouns

Spatial terms occur with possessor ascension.3 Two examples are given in (50) and (51).

(50) kyaan-s ás-k gərì-rā yááná-k bow-k’à dog:DEF-M 3MS-DAT head-ACC pot-DAT belly-LOC tóórá há-wùsk-ù-t downward 3ms-enter-u-SS ‘The dog entered his head down in the pot and he...’ (51) téré--s-ā taamu-k í-tà tóót- coffee_pot-DEF-M-ACC fire-DAT side-LOC2 erect.IMP-STI ‘Put the coffee pot next to the fire’

3 Since spatial terms are often derived from body parts, it is plausible that a language treats both as inalienable, but not necessarily so: Ewe distinguishes the two, treating spatial terms as ‘inalienable’ and body parts as ‘alienable’ (see Ameka 1996: 810ff for an explanation). POSSESSION IN SHEKO 239

Likewise, inherent parts of a location may be treated as a body part.

(52) hàà-z kyaan-s kàcawó-k’à ei-k PROX-M dog:DEF-M still up.there-LOC stone-DAT kop’arà-k’à há-bààs-kì-k open.space.in.forest.or.stone-LOC 3ms-want-be-REAL ‘Here the dog is still searching over there at the rock’s crevices.’

Locational body parts are not only used in the spatial frame, but also in the time frame, as shown in example (53).

(53) í-ts’yāāts’-ū-t-íì c’òr--àb-k adi-k’à 3pl-tie-u-SS-3pl finish-CAUS-REL-DAT footprint-LOC p’eet’à búúts-ú-t thatch mow-u-SS ‘after they finish tying they cut the thatch and…’ (lit. in the foot- prints to their finishing)

5. Conclusion Attributive possession in Sheko takes the form of possessive noun phrases, in which the head, i.e. the possessum, is marked tonally for being modified. Possessive noun phrases headed by baab ‘father’ or bé ‘mother’ are discussed in particular. This construction can be utilized to assert ownership, a central relationship between possessor and possessum. baab ‘father’ and bé ‘mother’ are grammaticalized to nominalizers and thus occur also in other constructions. Possessive noun phrases with baab show a peculiarity in gender marking. When baab ‘father’ is made definite, the construction becomes ungrammatical; instead, bé ‘mother’ plus the masculine definiteness marking must be used. The use of bé irrespective of gender is peculiar since the language has masculine as its default gender. The gender situation for Proto-Omotic is unclear, because present-day Omotic languages vary in their use of grammatical gender (e.g. Bender 2007: 736-742). Some languages have feminine as their default gender, others, like Sheko, masculine. The behavior of bé might point to a historical shift in default gender. Predicative possession is similar to existential predication, which corroborates typological findings. The possessum is the subject of the 240 ANNE-CHRISTIE HELLENTHAL predicate whereas the possessor is marked by the dative. According to Herslund & Baron (2001: 9), predicative possession can be divided into two types, of which one has the possessor as a topic, and the other the possessum. Sheko happens to use only one of these types: only the possessum can be the subject in predicative possession. Moreover, it is not a topic, which is proven by the absence of definiteness marking. The asymmetry is caused by the function of predicative possession, which typically asserts possession and thus presents the possessum as a new referent, similar to the existential construction, which also typically introduces a new participant. The possessor ascension construction in Sheko makes use of a dative case marker on the possessor noun phrase, whereas in constructions with other nouns the two noun phrases are juxtaposed without interven- ing phonological material. Besides, possession ascension is grammatically different from the predicative possession construction, in that the order possessor-possessum may not be reversed. Body parts occur frequently in the possessor ascension construction. A similar situation occurs in the related language Dizi, which gave rise to an discussion in which Alan (1976) claimed that Dizi shows inalienable possession, whereas Claudi & Serzisko (1985) argued Dizi employs possessor promotion. However, the underlying assumption in the argument appears not to hold, i.e. the alienable (‘normal’) construction and the inalienable (ascension) construction do not have the same meaning (Chappel & McGregor 1996: 7). In opposition to a possessive noun phrase, which more or less focuses on the possessor, i.e. the whole, possessor ascension basically centralizes the possessed, i.e. the part. This has been demonstrated with body part nouns: possessor ascension is the means to present (inalienable) possessed body parts, whereas a possessive noun phrase with body parts puts emphasis on the possessor or indicates that the body part is alienable. Spatial terms involving body part nouns behave in the same way. The marking of inalienable possession by a case marker in the Dizoid languages remains atypical, since a construction employed with inalienable possession is usually morphologically less marked/complex than one employed with alienable possession.

Acknowledgements Many thanks to my language consultants and friends in Sheko, especially to Ayna Bejih, P’et’ros Kiatus, Defera Chonu and T’erata Alemu; and to POSSESSION IN SHEKO 241

Azeb Amha, who read a very early and a later version of this paper and made useful comments on both, as well as to Maarten Mous and an anonymous reviewer. My research on Sheko is funded by and part of the project ‘Morphosyntax of two modal categories in Omotic languages’ under the Endangered Languages Programme of the Dutch Scientific Or- ganisation (NWO).

Abbreviations ACC accusative -əra LOC2 general locative -ta CAUS causative -s M masculine COND conditional -nta NEG negation marker ara COP copula tə NEG2 negation marker n DAT dative kn PASS passive -t’ DEF definiteness -n PL plural DEM demonstrative POSS possessor DS different subject -n POT potential -a converb F feminine PROX proximal demon- ha strative H High tone Q question marking IDEO ideophone REAL realis declarative -k ELAT elative REL relative IMP imperative SG singular INSTR instrumental -ka SS same subject con- -tə verb IRR irrealis declara- -m STI indirect stance tive L Low tone TPCLZ topicalizer LOC locative -k’a

Pronouns and possessor prefixes are written in SMALL CAPITALS, verbal subject clitics in normal small letters.

References Allan, Edward J. (1976). Inalienable possession in four Ethiopian languages. Afrika und Übersee 59 (4): 300-307. Ameka, Felix K. (1996). Body parts in Ewe grammar. In Hilary Chappell & William McGregor (eds.), The grammar of inalienability, 783-840. Berlin/New York: Mouton de Gruyter. Amha, Azeb (2001). The Maale Language. (CNWS Publications vol. 99.) Leiden: Research School CNWS. Beachy, Marvin (2005). An overview of Central Dizin phonology and morphology. MA thesis, University of Arlington, Texas. 242 ANNE-CHRISTIE HELLENTHAL

Belle, William van & Willy van Langendonck (eds.) (1996). The Dative, vol 1: descriptive studies. Amsterdam/Philadelphia: John Benjamins. Bender, Marvin L. (1990). Gender in Omotic. Journal of Afroasiatic Linguistics 2(2): 203-221. Bender, Marvin L. (2007). Topics in Omotic morphology. In Alan S. Kaye (ed.), Mor- phologies of Asia and Africa, 729-751. Winona Lake IN: Eisenbrauns. Chappell, Hilary & William McGregor (1996). Prolegomena to a theory of inalienability. In Hillary Chappell & William McGregor (eds.), The grammar of inalienability: a typological perspective on body part terms and the part-whole relation, 3-30. Berlin/New York: Mouton de Gruyter. Claudi, Ulrike & Franz Serzisko (1985). Possession in Dizi: inalienable or not? Journal of African Languages and Linguistics 7: 131-154. Dixon, Robert M. W. (1980). The languages of Australia. Cambridge: Cambridge University Press. Fleming, Harold C. (1976). Non-Semitic languages (section 2.1, Cushitic and Omotic). In Marvin L. Bender, J. Donald Bowen, Robert L. Cooper & Harold Fleming (eds.), Language in Ethiopia, 34-53. London:Oxford University Press. Heine, Bernd (1997). Possession: cognitive sources, forces and grammaticalization. Cambridge: Cambridge University Press. Herslund, Michael & Irène Baron (2001). Introduction: dimensions of possession. In Irène Baron, Michael Herslund & Finn Sørensen (eds.), Dimensions of Posses- sion, 1-25. Amsterdam/Philadelphia: John Benjamins. Mulugeta, Seyoum (2008). A Grammar of Dime. Utrecht: LOT (Netherlands Graduate School of Linguistics). Payne, Thomas E. (1997). Describing morphosyntax. A guide for field linguists. Cam- bridge: Cambridge University Press. Rapold, Christian J. (2006). Towards a grammar of Benchnon. PhD dissertation, Leiden University. Noun class system and agreement patterns in Logba (Ikpana)

Kofi Dorvlo (Language Centre, University of Ghana)

0. Introduction The Ghana Togo Mountain (GTM) languages are well known for their noun class system and agreement patterns. However, an insight into the grammar of Logba shows that this language behaves differently from its presumed genetic relatives in the Na sub group. This paper investigates the semantics of Logba noun classification and the associated agreement. First, I present the markers of the noun classes and examine the semantics of each class, noting the semantic grouping that can be set up within each class. The agreement patterns within the NP on the one hand and NP subject and verb on the other are then described. I show that Logba has a singular-plural pairing for nouns, except for those that refer to liquid and mass nouns. In the NP, demonstratives, interrogatives, cardinal numbers from one to six show concord with the noun head. However, adjective and intensifier do not show any agreement relation with the head noun. The selection of the vowel prefix, however, depends on the class of the noun and the [ATR] value of the vowel in the verb stem.

1. Socio-linguistic and historical context of Logba Logba is one of the fourteen languages on the hills of the Ghana-Togo frontier. These languages have been referred to as the Togorestsprachen (Struck 1912) Togo Remnant languages or the Central Togo Languages (Dakubu and Ford 1988). They are now commonly referred to as Ghana- Togo Mountain Languages (henceforth GTM) (Ring 1995). There are still differences in opinion on the classification of GTM languages. Westermann and Bryan (1952) consider these languages as an isolated group because they have vocabulary items which show a relationship to Kwa and a noun class system that is similar to the Bantu languages. Greenberg (1963a) classifies them among the Kwa sub-group B of the Niger-Congo family. Based on a comprehensive linguistic comparison, 244 KOFI DORVLO

Heine (1968) sub-classified them into KA and NA1. Stewart (1989) sub- mits that these two groups belong to two branches of Kwa: The KA group belongs to the Left Bank branch together with Gbe including Ewe. The NA group to which Logba belongs is in the Nyo branch which includes Akan and Ga-Adangbe. Williamson and Blench (2000) suggest that the KA and the NA sub groups branch out from Proto Kwa. Blench (2001:5) points out the difficulty in establishing the GTM languages as a group in relation to Kwa, and suggests that these languages may be given a more comprehensive analysis.

1.1 Nominal class system According to Schuh (1995: 128), the term, ‘noun class’ has been used in at least two senses in African languages. In one use, it refers to ‘a single set of morphological concords’. In another sense, it refers to ‘a paired set of morphological concords’ where the member of the pair refers to singular and the other member is its plural equivalent. I use noun class in the first sense. In this paper, I refer to the classes by the nominal prefixes which I write in bold in lower case. In a noun class language of the Niger-Congo family generally, nouns have a particular prefix in the singular, while for the plural a different prefix is used. There are some nouns, especially mass nouns which, as a result of their meaning, do not have a number differentiation. Also, there is a system of morphological concord between a nominal and the verb. The GTM languages are reported by most researchers to have noun class systems. Logba shares the general features that noun class languages of the Niger-Congo family are known to have. It however differs from its presumed genetic relatives. For example, both Logba and Likpe are classified as Na-Togo languages but Logba has no nominal prefix consonants, besides the nasal prefix. However, the data presented in Ameka (2002) suggests that Likpe has both CV- and V- nominal prefixes (see Ameka 2002).

1 According to Blench (2001) the KA and NA division by Heine is based on the word for ‘flesh’ in these languages. NOUN CLASSES AND AGREEMENT IN LOGBA 245

1.2 Typological features of the language. Logba has seven vowel phonemes. The two sets of mid vowels are distinguished by the feature [ATR], as shown in Table 1 below:

Front Central Back [+ATR] [–ATR] [+ATR] [–ATR] Close i u

Mid e  o 

Open a Table 1: Logba vowel system

All the affixes have two forms, one [+ATR] and the other [–ATR]. Selection of either of the forms depends on which harmony set the vowels in the lexical stem belongs. If the stem is [+ATR], one of the following vowels /o, u, i, e/ will be selected. If, on the other hand, it is [–ATR], the vowels selected will be one of these: /, /. The following words illustrate this:

[+ATR] [–ATR] (1) a. è-bí ‘beans of cocoa’ b. -kp ‘year’ è-gbè ‘stone’ -dz ‘women’ ó-dó ‘feather’ -k ‘custom’ ù-kpó ‘mountain’ -g ‘hunger’

The stems of words impose a restriction on the vowels in the affixes making them undergo a change. Words with [–ATR] mid-vowels in the stem trigger [–ATR] affixes while those with [+ATR] trigger [+ATR] affixes. In (2a), the verb stem lé ‘climb’ has the prefix /o/ and in (2b) the verb stem z ‘sell’ has the prefix //.

(2) a. olé y á b. z idz á o-lé -y-á -z i-dz-á 3SG-climb CM-tree-DET 3SG-sell CM-yam-DET ‘he climbed the tree’ ‘she sold the yam’ 246 KOFI DORVLO

The two close vowels /i/ and /u/ and the central vowel /a/ have no harmonic partners. /i/ and /u/ are [+ATR] because each triggers [+ATR] prefix. The verb stem tsí ‘sit’ in (3a) has /i/ and nú ‘hear’ in (3b) has /u/. Each of them triggers /o/ as the 3SG subject prefix.

(3) a. Tsami ŋkpsikp otsí onukpanago etsi tsami ŋkpsikp o-tsío-nukpa.nago etsi linguist every 3SG-sitCM-chief.big under ‘every linguist is under the paramount chief’ b. onu ikú é o-nú i-kú-é 3SG-hear CM-song-DET ‘he hears the song’

A verb stem with /a/ is [–ATR] because it takes an [–ATR] prefix. The verb stem wá ‘tell’ in (4) has /a/ and takes // as 3SG subject prefix.

(4) wám -wá-m 3SG-tell-1SGOBJ ‘she told me’

Logba has two tones H and L. Tone is realised on vowels and syllabic nasals. Monosyllabic words can be Low tone or High tone. This is exemplified in (5)

(5) bà ‘kill’ mì ‘take’ n-tá ‘hand’ bú ‘ask’ ŋú ‘see’ ŋ-gb ‘rashes’

Noun roots can also have Low or High tone. The noun prefix is either a Low toned vowel or a syllabic nasal. This is shown in (6)

(6) ì-sò ‘faeces’ ì-và ‘thing’ n-wù ‘dresses’ m-và ‘medicines’ ù-kú ‘bone’ ò-tú ‘gun’

All possible tonal patterns (LL, LH, HH, HL) are attested in disyllabic roots in (7). NOUN CLASSES AND AGREEMENT IN LOGBA 247

(7) LL àsè ‘thank’ fìfì ‘break’ LH tòl ‘send’ gànú ‘greet’ HH í ‘love’ húhú ‘wave hand’ HL dónù ‘shrink’ dzúbà ‘return’

Tone combinations in disyllabic noun roots are shown in (8)

(8) L-LL à-bùbà ‘termite’ à-dzàgò ‘millet’ L-HH à-búkpá ‘shoulder’ è-bítsí ‘child’ L-LH à-dzàyí ‘firewood’ ù-zùgbó ‘head’ L-HL à-fásà ‘landlord’ ò-zúmè ‘tomorrow’

In Logba, tone has both lexical and grammatical functions. Tone is used to indicate the difference between the Habitual aspect and the Past progressive aspect. While the former requires a low pitch, the latter is realised with a high pitch. (9a) and (9b) illustrate Habitual aspect and Past Progressive aspects respectively.

(9) a. tkp ìdz b. tkp ìdz -t-kp ì-dz -t-kp ì-dz 3SG-HAB-eat CM-yam 3SG-PSTPROG-eat CM-yam ‘He/She eats yam’ ‘He/She was eating yam’

From here on, only High tone will be marked. A syllable unmarked for tone should be interpreted as carrying Low tone. The examples in this paper are presented in four lines. The first line is an orthographic representation showing the Logba word divisions. The second line provides a morphemic representation written in bold with morpheme boundaries indicated by hyphens (-). The interlinear English gloss is in the third line and a free English translation is provided in single quotes in the fourth line. The basic constituent order of the clause is subject followed by the verb and in a transitive clause the verb is followed by a direct object. In a double object construction, the Goal comes before the Theme. In locative constructions, the first post-verbal object is the Theme followed by a second object which is the Locative. The adjunct occurs after the core arguments in the clause. The linear order of constituents in the clause is shown in (10). 248 KOFI DORVLO

(10) SUBJ VERB OBJ 1 OBJ 2 ADJUNCT

This basic constituent order can be different for reasons of topicali- sation and focalisation. In topic constructions, a noun phrase may be pre- posed to the clause. Also, in focus constructions the constituent that is focused is fronted to the left periphery of the clause. The verb usually appears with a vowel or a nasal prefix. The vocalic or nasal pronominal prefix therefore signals the agreement between the subject and the verb. In (11a) and (11b) the /o-/ and // refer to sá ‘man’ and Selorm respectively. In (12a) nú ‘water’ a liquid noun triggers /n/ on the verb t ‘pour’ whilst in (12b) in ‘meat’ a mass noun marks /i-/ on the verb tsi ‘be.in’.

(11) a. sá á ozi b. Selorm n bladzo -sa-á o-zi Selorm -n bladzo CM-man-DEF SM.SG-be.good Selorm SM.SG-buy plantain ‘The man is good’ ‘Selorm bought plantain’

(12) a. Nú nt uzi é yó N-ú n-t u-zi-é yó CM-water SM-pour CM-door-DET skin ‘Water poured on the door’ b. In itsi fútsú é nu i-n i-tsi fútsú-é nu CM-meat SM-be.in soup-DET in ‘Meat is in the soup’

Apart from the vowel pronominal prefix, the verb is able to host aspectual markers. All these occur as prefixes to the verb root. For example, in (13) the following markers: /o/ pronominal prefix, lo progressive aspect marker and mo negative marker are attached to the verb u ‘be’.

(13) Agb t olómou anú odu Agb t o-ló-mo-u a-nú o-du CM-dog say 3SG-PRSPROG-NEG-be CM-mouth CM-sickness ‘The dog says it is not attacked with “mouth-sickness” ’ NOUN CLASSES AND AGREEMENT IN LOGBA 249

This feature of the verb differentiates it from other categories: Nouns have prefixes to mark the class they belong to, but cannot be a host of aspect and negative markers. The rest of the paper is organised as follows: Section 2 is devoted to the prefix classes. In section 3, the semantic basis of noun classes is discussed. Section 4 deals with the various agreement patterns displayed in the language and section 5 summarizes the results of the discussion contained in this paper.

2. The prefix classes Singular nouns in Logba have a low tone vowel prefix. However, nouns referring to liquids and pourable substances show a syllabic nasal prefix. The largest number of nouns takes an a-prefix: These nouns have the nasal prefix as their plural. Nouns that have u- in the singular have e-/- in the plural; and those that have e-/- in the singular have a nasal prefix in the plural. The nouns that have o-/-prefix in the singular have i-prefix in the plural. A group of nouns with the i-prefix are mass nouns. There is another group of nouns which do not have a prefix. These zero prefix nouns are borrowed words. When used in the plural, the stem of the noun is maintained. The prefix of the singular is replaced except the plural prefix of mass and liquid nouns, as shown in Table 2 below:

PREFIX SINGULAR PREFIX PLURAL SEMANTICS a- a-biá ‘chair’ N- m-biá ‘chairs’ animals, in- a-gbé ‘dog’ ŋ-gbé ‘dogs’ sects, artifacts u- u-gusa ‘brother’ e-/- e- gusa ‘broth- kinship terms, u-bme ‘town’ ers’ social group e-bme ‘towns’ terms e-/- e-fieyi ‘cala- N- n-fieyi ‘cala- natural ele- bash’ bashes’ ments, items for e-kele ‘grass’ n-kele ‘grasses’ ritual and religious practices. o-/- -s ‘horse’ i- i-s ‘horses’ God, big ani- o-dró ‘elephant’ i-dró ‘elephants’ mals 250 KOFI DORVLO

NON-COUNT NOUNS N- n-da ‘liquor’ - - liquids, pourable substances i- i-n ‘meat’ - - mass nouns, abstract substances Table 2: Singular and plural prefixes

Some plural nouns with a nasal prefix add an additional plural suffix -w:

(14) a-gut ŋ-gut-w ‘bat’ a-dzimi n-dzimi-w ‘mudfish’ e-féshí n-féshí-w ‘sheep’

Some other nouns have [-w] suffixed to the noun with the prefix typical of the singular. These nouns are loans either from Ewe or from another language in the area. The plural morpheme in Ewe is [-wó]. It is probable that the Logba form [-w] is based on this morpheme. These nouns show the same prefix as their singular counterparts:

(15) a-kpn a-kpn-w ‘biscuit’ u-kpl u-kpl-w ‘table’ a-kó a-kó-w ‘parrot’

There are other prefixless nouns that take the plural suffix -w. These nouns can be traced to Ewe. Some of the words, for example a-bladzo, have prefixes in Ewe, the language from which these nouns are probably borrowed.

(16) mang mang-w ‘mango’ bladzó bladzó-w ‘plantain’ fesre fesre-w ‘window’ seƒóƒó seƒóƒów ‘flowers’

As was observed in a fieldwork session, there is a simplification of the singular-plural prefix system going on among younger speakers. Stu- dents between 16 to 21 years of age attending the Jim Borton Memorial NOUN CLASSES AND AGREEMENT IN LOGBA 251

Secondary School in Adzakoe used the -w suffix as the plural marker for all the nouns, without any class-marking prefix. When I presented these elicited data to older speakers in Klikpo, Adiveme and Alakpeti, they frowned on these forms, describing them as ungrammatical and a careless adulteration of the language. One can foresee that in a not too distant future, the class system will have ceased to function in the plural nouns. This does not mean that the -w suffix is inappropriate. There are many nouns for which this plural suffix is the accepted form. In nouns referring to peoples there can be a singular suffix parallel to the plural suffix. A good candidate to exemplify this point is the noun A-kpana- nyi, which means ‘a person who hails from Akpana (Logba)’. There is a template for prefixing and suffixing on the noun in the language. Also there is an internal shift in the language in favour of the suffix, -w.I propose that the noun prefix is a language internal grammatical feature and the -w suffix is a product of the contact with Ewe. Westermann (1903) points out that those nouns that have been borrowed from surrounding languages and are still perceived as foreign words have kept their plural suffix. Bertho (1952: 1051) notes that this is a borrowing from Ewe. It is probable that the process of simplification of the singular-plural prefix system will continue until a large number of the prefixes will be ‘bleached out’ of the language. The -w suffix has become a default plural in the language. When nouns with the -w suffix are followed by an agreeing modifier, the modifier as well as the verb take the a- prefix which is used for plural nouns in the language. This is attested in example (17) below:

(17) peyaw am atsoenú peya-w a-m a-tsoenú pear-PLU AM-DEM SM.PLU-good and dry ‘Those pears are good and dry (not watery).’

There is another set of nouns that appears without any plural affix. These are non-count or mass nouns. Instead of the expected i-prefix that is used for mass and individuated nouns, they have the a-, u-, o-/- prefixes. These nouns also include undifferentiated nouns like e-gbe ‘stone’, e-tsi ‘land’ and nouns like e-vi ‘sun’ and a-bobí ‘moon’. It is probable that the worldview of the people makes them conceive these nouns as having no identifiable plural. A discussion with the native speakers 252 KOFI DORVLO brings this to light. They argue that it is not possible to have a plural of these entities in real life so the plural of these nouns is not acceptable. Some examples are below:

(18) a. a- b. u- a-bobí ‘moon’ ú-sú ‘urine’ a-bu ‘valley’ u-súsfolí ‘bladder’ a-débí ‘kidney’ u-mnta ‘salt’ A-drúva ‘Thursday’ u-múshi ‘smoke’ a-nú ‘mouth’

c. o-/- -dntí ‘waist’ -dzá ‘fire’ o-kúnu ‘anus’ o-lómí ‘testis’

Thirdly there are a few nouns that have suppletive or compounded plurals. The stem of the plural a-há ‘persons’ is similar to the Ewe word ha ‘group’. The plural forms for ‘man’ and ‘woman’ are compounds using the word ina- ‘person’ as the first element of the compound followed by the word for ‘man’ or ‘woman’. This is illustrated in (19):

(19) i-na ‘person’ a-há ‘persons’ -sá ‘man’ i-ná-sá2 ‘men (literally: person men)’ u-dz ‘woman’ i-ná-dz3 ‘women (lit.: person women)’

3. Semantic basis of noun classes It has been demonstrated for some languages that their noun classes are semantically based. (see Breedveld 1995 for Fulfulde, and Aikhenvald 2000, for example for a general typology). However, the patterns displayed in languages are not universal, there are language specific differences. A loose semantic definition is typical of each noun class. In

2 There are some speakers who have the plural as asáw. 3 There are some speakers who have the plural as dzw. NOUN CLASSES AND AGREEMENT IN LOGBA 253

Logba, each class has nouns which are not easily accounted for by a single semantic feature. a- class The a- class hosts a large number of nouns. Three semantic sub-groups emerge. These are: a. animals. b. body parts c. artefacts. The semantic subgroup of animals includes nouns referring to insects. The o- / - prefix is used to refer to them.

(20) a. Animals a-n.d ‘cat’ a-gb ‘dog’ a-n.k ‘chicken’ a-kl ‘goat’ a-gú ‘antelope’ a-lá ‘scorpion’ a-kpakpla ‘frog’ a-gbíglm ‘spider’ b. Insects a-zuz ‘housefly’ a-ny ‘louse’ akpakpla ‘frog’ is borrowed from Ewe and is integrated in this class. This is based on its form and meaning in Ewe and Logba. Nouns referring to visible parts of the body form a second semantic group within this class. These terms can be applied to body parts of animals as well.

(21) a-tr ‘breast’ a-ŋaŋa ‘rib’ a-fuí ‘thigh’ a-gbashi ‘arm’ a-kukli ‘finger nails’

Nouns referring to the semantic group of artefacts are human made, from clay, wood or cotton. These are used in the daily activities of the people.

(22) a-l ‘clay-bowl’ a-kntí ‘basket’ a-kpó ‘farm-bag’ a-s ‘pot’ a-biá ‘chair’ a-fúta ‘cloth’ u- class The u-class contains at least four semantic clusters: kinship terms (23a), social organization terms (23b), human category terms (23c), and important socio-cultural possessions (24). 254 KOFI DORVLO

(23) a. Kinship terms u-gusa ‘brother’ u-tí ‘father’ u-gu ‘husband’ u-má ‘mother’ b. Social group terms u-sá ‘clan’ u-nánsa ‘chief’ (status) u-bme ‘town’ c. Human category terms u-dz ‘woman’ u-bí ‘child

Socio-culturally salient entities and entities that belong to the same semantic or associative sphere are part of this class. The word u-tsa ‘house’ can be considered as both belonging to the social organisation terms and the socio-culturally salient ones. Among the latter are words for ‘barn’ and ‘grinding stone’, which are found in the home or in the farm:

(24) u-tsá ‘house’ u-loégbé ‘grinding stone’ u-bo ‘farm’ u-mútsí ‘barn’ e- class The e-class is a small class comprising natural elements and items relating to ritual and religious practices. Dixon (1982) and Lakoff (1987) point out that the class system of a language may be based on the myth and belief system of the culture. When some of these myths are lost, it is difficult to understand why a given group of items are in a single class. I attended a ritual carried out to placate the gods for an attempted suicide in one of the Logba villages and observed that some of the items in this group were used or referred to during the ceremony, confirming their functional unity to the native speakers of Logba. The nouns that belong to this class are given in (25) below:

(25) e-ví ‘sun’ e-gbe ‘stone’ -kp ‘year’ e-kelé ‘grass’ e-tsí ‘ground’ e-fiéyí ‘calabash’ e-feshi ‘sheep’ e-te ‘tooth’ NOUN CLASSES AND AGREEMENT IN LOGBA 255 o- class The nouns in this class refer to God, man, important people, big animals and soft and attached human body parts. This class can be referred to as the class of important entities. It includes the words related to male gender, including specific body parts, such as testis. The nouns belonging to the different subgroups include:

(26) a. God and important people -kpaya ‘God’ o-núkpá ‘king’ -sá ‘man’ b. Big animals -sámínángo ‘leopard’ o-gbómí ‘monkey’ o-dró ‘elephant’ -s ‘horse’ o-ló ‘crocodile’ c. Soft and attached body parts o-lómí ‘testis’ -t ‘cheek’ o-tsóe ‘ear’ o-núkpá ‘king’ is perhaps loaned from Ga. The word o-ló for ‘crocodile’ is similar to Ewe e-ló, but the prefix is different. The word for ‘horse’ could be borrowed from Ewe also. In fact, in some Ewe dialects the word e-s ‘horse’ has o- prefix. For example, in the Peki dialect, it is o-s.

N- class This class is dominated by nouns referring to non-individuated entities especially liquids. Some of its members are:

(27) n-da ‘liquor’ n-ú ‘water’ n-fú ‘oil’ i- class Non-count nouns which refer to either abstract entities like peace or masses made up of particles such as rice or sand are in this class. 256 KOFI DORVLO

(28) i-be ‘time’ i-yóyú ‘peace’ i-n-fieyi ‘sand’ i-yánu ‘air’ i-tsí ‘soil’ i-múnyí ‘hair’ i-kágo ‘rule’ i-hanágo ‘indiscipline’ i-m-bí ‘rice’

Agreement patterns in relation to the noun class system are discussed in the next section.

4. Agreement Agreement is a formal relationship between elements whereby a form of one word requires a corresponding form of another’ (Crystal 2004). I will adopt the terminology proposed by Corbett (2006), who proposes to refer to the element that determines the agreement as the controller and the element whose form is determined as the target. Corbett refers to the syntactic environment in which agreement occurs as the domain of agreement and further notes that singular, dual and plural are agreement features. The singular classes have identical agreement patterns but different noun prefix and different plural pairings. However, members of a small group of singular nouns identified as artifacts have a different agreement pattern. The prefixes of the nouns can either be a vowel or a nasal. These are cross referenced on the verb and the demonstrative to signal agreement. These are shown in Table 3 below:

Prefix Verb Agreement Demonstrative a-SG o-/- o-/- N- PLU N- N- u- SG o-/- o-/- e-/- PLU e-/ a- e-/- SG o-/- o-/- N- PLU a- N- o-/- SG o-/- o-/- i- PLU i- i- a- SG a- a- Table 3: Noun prefixes and agreement patterns NOUN CLASSES AND AGREEMENT IN LOGBA 257

4.1 Agreement within NP In the subsections that follow, I will describe agreement phenomena within the NP. A noun word in Logba consists of a noun class prefix, a noun stem and possibly a plural suffix. In a simple NP the adjective occurs after the head noun. The adjective is followed by a quantifier and a determiner. An intensifier follows a determiner at the NP boundary. The quantifier slot can be filled by the cardinal numbers i-kp ‘one’ n-ny ‘two’, and so on. The determiner slot is filled by elements which include demonstratives i-m ‘this’ m-m ‘these’. The head noun determines the agreement of the elements in the NP.

4.1.1 Possessive NP Possession is expressed by the juxtaposition of the possessor and the possessed. A determiner obligatorily occurs on the possessed entity. A class marker of the possessed noun is maintained except for kinship terms. In (29) the possessed noun umá ‘mother’ is a kinship term which is used without the class marker /u-/. However, otú ‘gun’ in (30) is used as a possessed noun with the class marker /o-/ because it is not a kinship term.

(29) Kofi má á (30) Ivanuvo otú é Kofi má á i-vanuvo o-tú é Kofi mother DET CM-hunter CM-gun DET ‘Kofi’s mother’ ‘The hunter’s gun’

In an NP in which possession is expressed, the adjective follows the possessed noun. The quantifier and the demonstrative occur after the adjective in this order. This is shown in (31).

(31) Ivanuvo otú kŋkl glaŋkp am alé akponu i-vanuvo o-tú kŋkl glaŋkp a-m a-lé CM-hunter CM-gun old seven AM-DEM SM.PLU-be akpo-nu farm.bag-containing region ‘Those seven old hunter’s guns are in a farm bag’ 258 KOFI DORVLO

4.1.2 Head noun and adjective Logba has one underived adjective and processes by which adjectives can be derived from other categories like verbs and nouns. Within the NP, the adjective immediately follows the head noun. Unlike the determiner, which has number agreement with the head noun, the adjective does not take any prefix to mark agreement. In sentence (32), kŋkl ‘old’ is an adjective and appears after m-fúta ‘clothes’ the head of the NP. Both n-ny ‘two’ and m-m DEM have N-prefix to indicate agreement with m-fúta ‘clothes’, the head noun.

(32) Man mfúta kŋkl nny mm pétée Ug Ma-n m-fúta kŋkl n-ny m-m pétée Ug 1SG-buy CM-cloth old AM-two AM-DEM all Accra ‘I bought all those two old clothes in Accra’

4.1.3 Head noun and numeral When used as modifiers, the cardinal numbers ‘one’ to ‘six’ (which have the i-prefix when counting), exhibit prefix variation, in order to show agreement with the head noun. However, the agreement is shown only when the head noun is individuated. With the singular prefix classes, nouns belonging to u-, e-/-, o-/- take the o-/- agreement on the number. This is shown in (33).

(33) Ebitsi -kp E-bitsi -kp CM-child AM-one ‘one child’

Countable nouns belonging to the N-class take the N- agreement marking on the number as shown in (34).

(34) Ngb nny N-gb n-ny CM-dog AM-two ‘two dogs’

The a-prefix class of artifacts takes a- agreement marker. In (35) afúta ‘cloth’ is an artifact with an a-prefix and triggers an a-prefix agreement NOUN CLASSES AND AGREEMENT IN LOGBA 259 marker on the numeral but adzi ‘day’ triggers N-prefix agreement marker on the numeral in (36).

(35) Afúta drui a-kp (36) Adzi ŋ-kp4 A-fúta druia-kp A-dzi ŋ-kp CM-cloth red AM-one CM-day AM-one ‘one red cloth’ ‘one day’

The e-/- plural nouns trigger the a-agreement marking on the numeral.

(37) Ebitw any E-bit-w a-ny CM-child-PLU AM-two ‘two children’

Other cardinal numerals are not known to have exhibited agreement relation.

4.1.4 Head noun and Demonstrative The choice of prefix for the demonstrative is determined by the number (SG/PLU) and the semantic class to which the noun head belongs. There is an agreement relation between the noun and the demonstrative. The noun head is the controller and the target is the demonstrative. In the singular, o-/- is used as a prefix to the demonstrative, a- for plural, i- for mass nouns and other i-prefix nouns. N- is the agreement marker for liquid nouns and any other N-prefix nouns. The sentences below demonstrate this. (Agreement markers are in bold face)

(38) SG. COUNT NOUN Ebiti -m á ‘This child is fat’ PLU. COUNT NOUN Ebitw a-m aá ‘These children are fat’ SG. -ANIMATE Afuta a-m azi ‘This cloth is good’ PLU. -ANIMATE Mfuta m-m nzi ‘These clothes are good’ LIQUID Na m-m mbo intse ‘This drink is strong’ MASS Ia i-m ibo iu ‘This metal is heavy’

4 This is lexicalised in the language and it is used in the introduction in story telling. 260 KOFI DORVLO

Two forms of demonstratives are distinguished in Logba based on how near or far away the noun is from the speaker. These are:

(39) +near m ‘this’ –near m ‘that’

4.1.5 Head noun and interrogative The agreement relation between the noun and the question word b ‘how much’ is identical to that of the demonstrative, but restricted to the plural classes. The plural agreement marker a- is therefore used. For mass nouns and other i-prefix nouns, i-agreement marker is used. N-agreement marker is used for both liquid nouns and N-prefix nouns. Other question words which combine with singular nouns use o-/- as a prefix to the question word. This is shown in (40) below:

(40) SG. COUNT NOUN sa -má ‘Which man?’ PLU. COUNT NOUN Aha a-má ‘Which persons?’ SG. -ANIMATE Afuta a-má ‘Which cloth?’ PLU. -ANIMATE Mfuta m-má ‘Which clothes?’ LIQUID Nu m-má ‘Which water?’ MASS Ia ib i-lá ‘What is the time?’ (Lit: Iron how many beat?)

4.1.6 Head noun and intensifier The intensifier follows the determiner and any item that comes after it cannot be part of the NP. It has no agreement relation with the head noun. The sentence in (41) below is an extract from libation prayer. i-m AM-DEM has i-prefix to show agreement with i-va ‘thing’, the head noun, but there is no prefix on the intensifier petee ‘all’ to mark agreement.

5 (41) Iva gbali im pétée ta iz im loo! iva gbali i-m pétée ta i-z i-m loo! CM-thing bad AM-DEM all let 3SG-go AM-DEM ADR ‘Let all these bad things go away here’ (libation)

5 This is taken from the libation prayer in one of the performances in Logba by the elders in Alakpeti. NOUN CLASSES AND AGREEMENT IN LOGBA 261

I will now move on to another agreement domain, the clause, in which agreement takes place between the subject NP and the verb.

4.2 Agreement between the subject NP and the verb It is a grammatical requirement in Logba to cross reference the subject on the verb with a concord marker. Arguments in object position in a clause are not cross referenced on the verb. The nouns which have the o- /- verbal concord are by far the largest group. They comprise nouns with the following prefixes. o-/-, u-, e-/-, and a-6. These are all singular nouns. Nouns that trigger the i-verbal concord are those that take the i-prefix. Nouns whose stems belong to the o-/- prefix take the o-/- singular. Nouns that are cross referenced by the N- prefix take the N-noun prefix. Nouns with the e-/- plural noun prefix trigger the same prefix as concord. The verbal subject marker is glossed as SM.SG when it refers to a singular noun and SM.PLU when it refers to a plural noun. The subject marker is not marked for singular or plural when it refers to non count nouns like liquid and mass nouns. Prefixes on the nouns are glossed as CM for both singular prefix and plural prefix. The additional suffix -w attached to some nouns is glossed as PLU. Other noun class agreement markers such as those on demonstratives and quantifiers are glossed with AM (agreement marker). The lexical subject is not obligatory in the language. If the lexical noun is removed from example (43) leaving -z dzátsum ‘it went to the kitchen’ the sentence will be grammatical. It is however obligatory to have a subject marker on the verb when a lexical subject is used as in (42-44) below: The object is not marked on the verb and there are no special markers on lexical nouns in subject and object position. The noun in both subject and object position can be pronominalized as there are distinct forms of pronouns for them.

6 A small group of a-prefix nouns, which I describe as artifacts because they are wood, clay, cotton and metal objects, take the a- prefix as agreement marker in the singular. In the plural, they fall in the class of nasal prefix nouns and take N- as agreement marker. 262 KOFI DORVLO

The Subject agreement classes of all the Noun classes are further illustrated with example sentences in Table 4 below:

Noun Verbal Prefix Concord Example sentence a- ó- 42. A-gbi- ó-ŋú n-wo u-tsa. CM-dog-DET SM.SG-see CM-bee CM-home ‘The dog saw a bee hive.’ - 43. A-nd-á -z -dzátsúm. CM-cat-DET SM.SG-go CM-kitchen ‘The cat went to the kitchen.’ N- Ń- 44. N-k-á ń-dó. CM-fowl-DET SM.PLU-go.out ‘The fowls went out.’ a- á- 45. A-fútá-a á-k a-gli-é yó. CM-cloth-DET SM.SG-hang CM-wall-DET skin ‘The cloth hangs on the wall.’ u- ó- 46. U-dzi- ó-gl belet. CM-girl-DET SM.SG-tie belt ‘The girl has a belt on her waist.’ - 47. U-nansá-á -z Tota. CM-chief-DET SM.SG-go Tota ‘The chief went to Tota.’ e- é- 48. E-nansá é-bá Klikpo. CM-chief SM.PLU-come Klikpo ‘Chiefs came to Klikpo.’ e- ó- 49. E-feshi-é o-bo u-tsá nu. CM-sheep-DET SM.SG-stay CM-house in ‘The sheep is in the room.’ - 50. E-bitsí-é -fnyí kuatsia. CM-child-DET SM.SG-peel banana. ‘The child peels banana.’ e- á- 51. E-kele-w á-lé a-fá-á nu. CM-grass-PLU SM.PLU-be AM-house-DET in ‘Grasses are in the house.’ o- ó- 52. O-ló ó-rí-é. CM-crocodile SM.SG-hold-3SGOBJ ‘crocodile caught him.’ NOUN CLASSES AND AGREEMENT IN LOGBA 263

- 53. -gblawò-é -la Kofi. CM-teacher-DET SM.SG-beat Kofi ‘The teacher beat Kofi.’ i- í- 54. I-w i-bo a-fá-á nu. CM-mortar SM.PLU-stay CM-house-DET in ‘Mortars are in the house.’ N- Ń- 55. N-ú ń-t u-zí-e yó. CM-water SM-pour CM-door-DET skin ‘Water pours on the door.’ i- i- 56. I-n í-tsi futsu-é nu. CM-meat SM-be.in soup-DET in ‘Meat is in the soup.’ Table 4: Subject agreement classes with example sentences

4.3 Agreement systems A combination of noun phrase internal agreement and external verb agreement results in nine different agreement classes: five singular and four plural classes. Two of the agreement classes also contain nouns that have no number distinction. These are mass nouns with a noun prefix i- and liquid nouns with a nasal prefix. There are two singular noun classes with a noun prefix a-. These are distinguished in verb agreement only. There are two plural noun classes with a noun prefix N-. These are also distinguished in verb agreement only. The singular classes have identical agreement patterns but different noun prefix and different plural pairings.

5. Conclusion The paper argues that the noun classes in Logba are in semantic grouping but each group has sub-groups which cannot be accounted for by a single semantic feature. In the a-class are sub-groups like animals, visible body parts and artefacts. The u-class has kinship terms, social group terms and human category terms. The o-class has God and important people, big animals and soft and attached body parts. The N-class contains nouns referring to liquid and other pourable substances. The i-class is for non- count mass nouns and the e-class is for items used for ritual and religious practices. There is agreement relation within the NP. The cardinal numerals one to six which have i-prefix when counting show agreement relation 264 KOFI DORVLO

with the head noun. However, the adjective and the intensifier are the constituents in the NP which do not show any agreement relation with the head noun. There is also agreement between the subject NP and the verb. The argument in subject position controls the agreement and the verb is the target. The verbal concord is in the form of a vowel prefix or nasal prefix on the verb. The selection of the vowel prefix however depends on the class of the noun and the [ATR] value of the vowel in the stem. The noun class system in Logba exhibits a partial disintegration when compared to what is reported on in other GTM languages. The data presented in Ameka (2002) for Likpe a Na Togo GTM language shows that it has CV, V and N-prefixes. Schuh (1995) also reports CV prefix nouns and a more coherent class system for Avatime, a Ka Togo GTM language which is a close geographical neighbour of Logba. In his findings, Heine (1968), shows that Logba has lost all nominal prefix consonants except the nasal. However, in an earlier work exclusively on the grammar of Logba, Westermann (1903)7 notes that the feature of nominal prefixes in GTM languages is present in Logba in a broken down form. This gives the impression that Logba had a coherent system which has broken down as a result of possibly a contact with other languages in the area.

Abbreviations [+ATR] Advanced tongue root DEF Definite [–ATR] Unadvanced tongue root H High tone 1SG First person singular HAB Habitual 3SG Third person singular L Low tone AM Agreement marker NEG Negative CM Class marker OBJ Object PRSPROG Present progressive PSTPROG Past progressive SM Subject marker SM.PLU Subject marker, Plural SM.SG Subject marker, Singular

7 The original version in German was translated by Juliet Huber, Leiden Uni- versity Centre for Linguistics, Netherlands. NOUN CLASSES AND AGREEMENT IN LOGBA 265

References Aikhenvald, Alexandra Y. (2000). Classifiers. A typology of noun categorization devices. Oxford: Oxford University Press. Ameka, Felix K. (2002). The progressive aspect in Likpe: Its implications for aspect and word order in Kwa. In Felix K. Ameka & E. Kweku Osam (eds.), New directions in Ghanaian Linguistics, 85-112. Department of Linguistics, Legon. Bertho, R. P. (1952). Les dialectes du Moyen-Togo. Bulletin de l’IFAN, B, 14: 1046- 1107. Blench, Roger (2001). What have we learnt since Heine? Paper presented at the 32nd Annual Conference on African Linguistics and subsequently revised. Berkeley, 23-25th March 2001. Breedveld, J. O. (1995). Form and meaning in Fulfulde. A morphological study of Maasinakoore. PhD thesis Research School CNWS, Leiden. Corbett, Greville G. (2006). Agreement. Cambridge: Cambridge University Press. Crystal, David (2004). A Dictionary of Linguistics and Phonetics Oxford: Blackwell Publishing. Dixon, R. M. W. (1982). Where have all the adjectives gone? And other essays on syntax and semantics. Berlin: Mouton. Egblewogbe, Eustace Y. (1992). The remnant concept and the distribution of Guan. Afrika und Übersee 75: 43-57. Greenberg, Joseph H. (1963). History and present status of the Kwa problem. In M. Houis, G. Manessy & S. Sauvageot (eds.), Actes du second colloque international de linguistique negro-africaine, 215-217. Dakar: Université de Dakar and West African Languages Survey. Heine, Bernd (1968). Die Verbreitung und Gliederung der Togorestsprachen. Berlin: Dietrich Reimer. Kropp Dakubu, M. E. & K. C. Ford (1988). The Central Togo Languages. In M. E. Kropp Dakubu (ed.), The Languages of Ghana, 119-153. London: Kegan Paul International. Lakoff, George (1987). Women, fire and dangerous things: what categories reveal about the mind. Chicago: University of Chicago Press. Ring, AndrewJ. (1981).Ewe as a second language: a sociolinguistic survey of Ghana’s Central Volta Region, Legon: Institute of African Studies. Schuh, Russell G. (1995). Avatime Noun Classes and Concord. Studies in African Lin- guistics 24: 123-149. Stewart, J. M. (1989). Kwa. In In J. Bendor Samuel (ed.), The Niger-Congo languages, 217-246. Lanham: University Press of America. Struck, R. (1912). Einige Sudan-Wortstamme. Zeitschrift fur Kolonialsprachen 2: 233- 253, 309-323. Welmers, William. E. (1973). African Language Structures. Berkeley: University of California Press. Westermann, D. (1903). Die Logbasprache in Togo. Kurzer Abrib der Grammatik und Texte. Zeitschrift fur Afrikanische und Ozeanische Sprachen 7: 23-39. Westermann, D. & M. A. Bryan (1952). Languages of West Africa. Handbook of African Languages 2. Oxford University Press for International African Institute. 266 KOFI DORVLO

Williamson K. & R. M. Blench (2000). Niger-Congo. In Bernd Heine & Derek Nurse (eds.), African Languages: An Introduction, 11-42 Cambridge: Cambridge Uni- versity Press. Noun classes in Tafi: A preliminary analysis

Mercy Lamptey Bobuafor (Leiden University Centre for Linguistics)

1. Introduction Tafi is a Ghana-Togo-Mountain (GTM) language (Ring 1995) spoken in Ghana. The GTM languages are spoken in Ghana, Togo and Benin. Ear- lier works on these languages referred to them, in German, as “Togo- estsprachen” (Struck 1912) or, in English, as “Togo Remnant Lan- guages”. They have also been referred to as “Central-Togo languages” (Kropp Dakubu & Ford 1988). The genetic affiliation of these languages within Niger-Congo has been much contested. Westermann and Bryan (1952: 96) considered them as constituting an isolate group and observe that they have some similarity to the Kwa languages in terms of vocabulary while their noun class system is reminiscent of Bantu languages. Greenberg (1996) puts them in Kwa. Heine (1968a) also classifies them as Kwa and sub-divides them into NA- and KA- groups. These two groups, now called NA-Togo and KA-Togo, are presumed to branch out from Proto-Kwa (Williamson and Blench 2000; Blench 2001). Tafi belongs to the KA-Togo group. The Tafi language is spoken by about 4,400 people in the Volta Region in the South-eastern part of Ghana (Gordon, 2005), in four communities: Agɔme, Madɔ, Abuiƒé1 and Atome. The Tafi people call themselves Bagbɔ2 and their language Tɩgbɔ. South of the Tafi area live the Nyagbo people. To the east of Tafi are the Avatime and Logba people. Even though Logba and Tafi are geographical neighbours, Logba belongs to the NA-Togo group, so it is only distantly related to Tafi, unlike Nyagbo and Avatime. The Nyagbo people have sometimes been thought of as constituting one group with the Tafi, and their languages show a certain degree of mutual intelligibility. Tafi is one of the least studied languages, not only among the GTM languages but also in Ghana (Ring 2000). It is threatened by Ewe (the

1 The IPA symbol for orthographic ‘ƒ’ is /ɸ/. 2 Tafi is the name given to them by the Ewes. 268 MERCEY LAMPTEY BOBUAFOR lingua franca of the region), to which both adults and children are shifting. Tafi is also threatened by English, the official language of Ghana. The only available study on the language was done by Funke (1910). Tafi also receives some attention in Heine (1968a), Ford (1973) and Da- kubu & Ford (1988). Tafi, like all the other GTM languages, has an active noun class system in which each noun belongs to a particular class identified by prefixes. Thus, it marks singular/plural alternations with prefixes and often requires concord of other elements in the sentence with the governing noun. In this way, the head noun determines what affix is used for agreement. The aim of this paper is to describe the nominal class and concord system of Tafi drawing attention to its distinctive properties and placing the findings in an areal and typological context. The database for this study includes spontaneous spoken texts recorded in the field by the author, a Frog Story (FS) narrative description elicited using the wordless picture book Frog, where are you? Mayer (1969) and responses to various picture stimuli designed for the investigation of topological relations and the semantics of positional verbs. Data elicited with these picture stimuli are marked as TRPS (Topological Relations Picture Series, Bow- erman and Pederson (1993)); and PSPV (Picture Series for Positional Verbs, Ameka et al. (1999)).

2. Typological Features of Tafi Tafi has a total of 31 consonant phonemes including /bh/ and /ɸ/. /bh/ is an aspirated voiced labial stop. It has no voiceless counterpart in the language and occurs in some few words including kábhā ‘top’ and bhulí ‘small’. /ɸ/ is a voiceless labial fricative which has no voiced counterpart in the language. It appears that this sound was introduced into the language as a result of borrowing from Ewe. It only occurs in two words which are aɸu ‘sea’ and Abuiɸé3 ‘the name of one of the Tafi communi-

3 Abuiɸé is the Ewe appellation for this community. The Tɩgbɔ name is Ofú. Abuiɸé means Abui’s hometown, aɸé ‘home’ has been borrowed from Ewe and attached to the name Abui. It appears that the first person who settled in that village is called Abui. NOUN CLASSES IN TAFI 269 ties’. Tafi has double articulated labial velar stops /kp/ and /gb/ as well as alveolar and palatal affricates /ts/, /dz/ and /tʃ/, /dʒ/4 respectively. Tafi has a nine vowel system. All these vowels except /o/ have nasalized counterparts. The vowels participate in ATR vowel harmony. Thus, they are divided into two sets, [+ATR] and [-ATR]. As shown in Table 1 below, the vowels / i, e, o, u / are [+ATR] and / ɩ, ɛ, ɔ, ʊ / are [-ATR]. The /a/ vowel may be said to be neutral because it can occur with vowels from both sets.

Front Central Back [+ATR] [-ATR] [+ATR] [-ATR]

High i ɩ u ʊ Mid e ɛ o ɔ Low a Table 1: Tafi Vowel Chart

The near minimal pairs in (1) illustrate the contrastive nature of these vowels:

(1) Vowels Examples i / ɩ yí ‘kill’ yɩ́ ‘3SG OBJ’ ɩ / e ɩ-nyɩ́5 ‘firewood’ e-nyí ‘names’ u / i tsú ‘dig’ tsí ‘die’ ʊ / a bʊ-yā ‘farm’ ba-yā ‘farms’ a / u tá ‘shoot’ tú ‘pound’ ɔ / i sɔ́ ‘hoe’ sí ‘run’ a / ɔ ká ‘fold’ kɔ́ ‘give, offer’

ATR vowel harmony in Tafi is root-controlled, in the sense that the vowel of prefixes harmonizes with the vowel of the first syllable of the root. Thus, depending on the [ATR] value of the initial vowel in the root, the prefix may have varied forms, as shown in the examples below:

4 The orthographic representation of the IPA symbols /tʃ/ and /dʒ/ are ‘tsy’ and ‘dzy’ respectively. 5 The IPA symbol for ‘ny’ is /ɲ/. ‘ny’ is the orthographic representation. 270 MERCEY LAMPTEY BOBUAFOR

(2) a. á-ba-ŋa fufuō b. é-be-tú fufuō 3SG-FUT-eat fufu 3SG-FUT-pound fufu ‘S/he will eat fufu.’ ‘S/he will pound fufu.’ c. ɔ́-bɔ-vɩ sukúū d. ó-bo-plǔ a-wi=ń 2SG-FUT-go school 2SG-FUT-wash CM-dress ‘You will go to school.’ ‘You will wash the dress.’

As shown in the examples above, the vowels in the verbal root in (2a) and (2c) trigger a process of [-ATR] assimilation that targets the vowels of the subject pronoun and future prefixes which surface as [-ATR], while the vowels in the verbs in (2b) and (2d) impose their positive ATR value upon the vowels of the prefixes. Moreover, the future marker is underlyingly ba but in (2c) and (2d), the vowel in the future prefix is rounded. This results from a process of progressive assimilation triggered by the 2SG subject pronoun, which is phonologically [+round]. Tafi is a register-type tone language. In connected speech, it displays both level and contour tones. Tafi has three level tonemes: High (H), Mid (M) and Low (L) plus two complex tones, Falling (F) and Ris- ing (R) tones which are phonetically realised on the peak of one syllable.6 The contour tones can be lexical or they can be generated in context as in (4b) where the falling tone could be analysed as a sequence of H and L. Tones are used for lexical contrast as in (3), and to express grammatical functions, as shown in (4). Tones are marked with the following diacritics:

H ́ acute accent M ̄ macron L ̀ grave accent used on syllabic nasals (vowels with low tones are left unmarked) Rising ̌ hacek Falling ̂ circumflex

6 The tone system of Tafi is not clear yet. It presents several complications which require further investigation. NOUN CLASSES IN TAFI 271

(3) tú ‘pound’ tū ‘beat (a person)’ kilé ‘tooth’ kilě ‘lizard’ kɩmɔ́ ‘breast’ kɩmɔ̄ ‘rubber, gum’

(4) a. béé-tú fufuō b. bée-tú fufuō 3PL:PRSPROG-pound fufu 3PL:PSTPROG-pound fufu ‘They are pounding fufu.’ ‘They were pounding fufu.’

The difference between (4a) and (4b), representing the present and past progressive aspects, respectively, is exclusively tonal. In Tafi, each syllable is a tone bearing unit. Possible syllable shapes are given below and the syllable boundaries are marked by a dot (.).

(5) V á.kā ‘father’ o.sí ‘tree’ CV sʊ́.bha ‘rain’ bʊ.pá ‘house’ CLV ká.vlɔ̄ ‘towel’ o.dre.ke.kpú ‘corner’ CGV bʊ.mwɩ̄ ‘salt’ ɔ.bhɩā ‘friend’ N ká.m̀.pie.sí ‘armpit’ kó.ŋ́ ‘very much’

Tafi is a language whose basic word order is Subject-Verb-Object (SVO), as illustrated by the sentences in (4). In double-object constructions, the Goal precedes the Theme. In locative constructions, the Theme occurs before the Locative. Moreover, adjuncts may occur after the core arguments in the clause. The rest of the paper is organised as follows: in section 3, I discuss Tafi noun classes, drawing attention to the prefixes that identify them, the subject-verb agreement markers and the way loanwords are accommodated in the class system. Section 4 discusses agreement patterns within the Noun Phrase, between the head noun and adjectives, numerals, ordinals, the indefiniteness marker, demonstratives, interrogatives, and other nominal modifiers. Section 5 deals with pronouns. Sec- tion 6 looks at agreement within possessive pronominal constructions. Section 7 summarises the discussion. 272 MERCEY LAMPTEY BOBUAFOR

3. Tafi noun classes Schuh (1995: 130) observes that the term noun class is used in two different ways in the description of African languages, namely, it is used to refer to “a single set of morphological concords which may show up as affixes on noun stems, affixes on modifiers, and pronominal referents to nouns” and “a paired set of [morphological] concords where one member of the pair has a singular referent and the other member is a plural corresponding to that singular.” In this paper, I use the term following the first usage, because one can account better for the noun class system if the individual classes are treated separately since this will avoid the complication of many-to-one relation between singular and plural pairings. Moreover, using the singular to plural pairing leaves the tɩ- class which contains non-count nouns an orphan. Consequently, I distinguish noun classes on the basis of the prefixes, subject-verb agreement, the pronominal forms of the classes and agreement within the noun phrase. During the discussion, it will be observed that the noun class system in Tafi is gradually decaying.

3.1 Noun class prefixes In this section, I examine the individual sets of noun classes. First, I look at the prefixes the nouns occur with. Subsequently, I discuss the morphological concords each class governs. The various classes identified are 1 2 1 2 a -, ba(a)-, o-, i-, ki-, a -, ka-, bʊ -, bʊ -, and tɩ-. Five of these classes 1 2 are made up of singular classes (a -, o-, ki-, ka-, and bʊ - classes), four 2 1 are plural (ba(a)-, i-, a -, and bʊ - classes) and one (tɩ-) class contains 2 non-count nouns. The ba(a)- and bʊ - classes contain the plural forms of nouns in more than one class. Most often, in less careful or spontaneous speech, consonantal onsets are dropped, an observation already made by Funke (1910). In sentence initial position or in isolation, the full forms of 1 2 the CV noun class prefixes, ba(a)-, ki-, ka-, bʊ -, and bʊ - are used, but elsewhere their initial consonant may be dropped except for the prefix tɩ- whose full form is always used in any position. This is one of the signs that the noun class system in Tafi is in the process of decaying as it is in Nyagbo (see Essegbey, to appear). Also, historically, Logba is reported to have CV prefixes which have eroded to become only V prefixes (Heine 1968, Dorvlo, 2008 and this volume). NOUN CLASSES IN TAFI 273

Table 2 below gives the noun class prefixes and agreement markers that verbs and elements within the noun phrase take.

Class Class SubjM Subj Free Poss. Dem. Num. In- Int. Pfx rk Pron Pron Pron. Pfx Pfx def. Qtf. (CM) (SM) mrk. Pfx. 1 a - a/e/Ø- a/e- a/e- yɩ́ a- a- te- te- - ba(a)- ba/be- ba/be- ba/be- (b)alɩ́ a- -alɩ́ a- tɩa/tie- te- tɩa- o- ɔ/o- ɔ/o- ɔ/o- ɔlɩ́ ɔ- ɔ- to- to- - lɔ/lo- i- ɩ/i- ɩ/i- ɩ/i- ɩlɩ́ ɩ- ɩ- tɩ/ti- ti- tɩɩ lɩ/li- ki- kɩ/ki- kɩ/ki- kɩ/ki- (k)ɩlɩ́ ɩ- ɩ- tri- ti- - 2 a - a/e- a/e- la/le- alɩ́ a- a- ta/te- te- taa- ka- ka/ke- ka/ke- ka/ke- (k)alɩ́ ɩa- a- tie- tie- - 1 bʊ - bʊ/bu- bʊ/bu- bʊ/bu- (b)ʊlɩ́ ʊ- ɔ-/ ʊ- tʊ/tutu- tʊʊ- 2 bʊ - bʊ/bu- bʊ/bu- bʊ/bu- (b)ʊlɩ́ ʊ- ɔ-/ ʊ- tru- tu- -

tɩ- tɩ/ti- kɩ/ki- kɩ-/ki- (k)ɩlɩ́ ɩ- ɩ- - ti- tɩɩ- Table 2: Noun class prefixes, agreement markers and pronouns

As shown in the table, the same noun class prefixes are used as agreement forms for subject cross reference (SM) and subject pronouns except for the tɩ- prefix. Thus, although they are mostly identical in form their functions are distinct and I will like to keep them distinct. Also, the free pronoun is the form that occurs in object position as well as in possessive constructions.

3.1.1 The a1- class. With the exception of some borrowed nouns, nouns in the a1- class are characterised by the use of one of the prefixes a- or e-, depending on the [ATR] specification of the vowel in the initial syllable of the nominal root. The following are some examples:

(6) á-kā ‘father’ é-nī ‘mother’ a-dzɩ̄ ‘woman’ e-kusí ‘chief’ a-ga ‘animal’ é-fū ‘guest, stranger’ 274 MERCEY LAMPTEY BOBUAFOR

Depending on the ATR value of the first vowel in the verb, the subject- verb agreement marker that is used to cross-reference nouns in this class on verbs are a- or e-, as in (7a) and (7b). The subject pronominal form is also illustrated in (7c) and (7d). Consider the following examples taken from a frog story narration:

(7) a. á-nyɩ́nʊ́vɔɛ̄ =̄ ń7 á-mɔ o-zi to-lí nɩ́ o-sí CM-boy=DEF SM-see CM-hole AM-INDEF LOC CM-tree gbɩgblǎ nɩ́-m̄. big DEF-inside ‘The boy saw a hole in the big tree.’ (FS) b. á-nyɩ́nʊ́vɔɛ̄ =̄ ń é-wū o-sí=ń CM-boy=DEF SM-climb CM-tree=DEF ‘The boy climbed up the tree.’ (FS) c. é-féké ka-kudzɔgɛ=̌ ń 3SG-lift CM-dog=DEF ‘He (= the boy) lifted the dog.’ (FS) d. á-nyɩ́nā o-sí to-lí nɩ́ ki-trekpú=ń ɔbha 3SG-hold CM-tree AM-INDEF LOC CM-anthill=DEF near ‘He (= the boy) held a tree beside the anthill.’ (FS)

In (7a) and (7b), the subject NP has been cross-referenced on the verb. The alternation between a- and e- of the SM is explained by ATR assimilation. In (7c) and (7d), the subjects of the sentences are pronouns. Some nouns in this class are prefix-less. These nouns are in this class, because they share with the prefixed nouns in this class the agreement markers, as in (8a) and (8b). Examples of such nouns are Kpáyā

7 The full form of the definiteness marker in Tafi is nɩ́ ‘the’. Most often, its vowel is elided and the reduced form is cliticised on to the word immediately preceding it. The tone carried by the deleted vowel, however, remains and it moves leftwards on to the alveolar nasal /n/. However, when it is followed by certain postpositions such as kábhā ‘on, top’, kedē ‘behind’, back’ and kesí ‘under ‘beneath’, etc., whose initial consonants are dropped, the vowel of the definiteness marker remains even though after their initial consonant is dropped, the resultant word that follows nɩ́ begins with a vowel, as in example (9a). Also, in utterance-final position, nɩ́ and ń are in free variation. NOUN CLASSES IN TAFI 275

‘God’; sã́hwɩ̌̃ ‘spider’; gbɔkɔɛ́ ‘toad’; and túŋ́gbá ‘antelope’. These nouns can be said to take a Ø-prefix.

(8) a. sã́hwɩ̌̃=ń é-tsirí bu-pí nɩ́-m̄. Spider =DEF SM-be.on CM-ceiling DEF-inside ‘The spider is on the ceiling.’ (TRPS. 7) b. á-nyɩ́nʊ́vɔɛ̄ ̄ te-li a-bá-zā kɩlɩ́ CM-boy AM-INDEF SM-come-be:PST_LOC CONJ gbɔkɔɛ́ tsyí á-zā. toad too SM-be:PST_LOC ‘A boy happened to be there and a toad was there too.’ (FS)

As shown in the examples in (8), the agreement marker for the noun sã́hwɩ̌̃ ‘spider’ in (8a) is e-, while á-nyɩ́nʊ́vɔɛ̄ ̄ ‘boy’ and gbɔkɔɛ́ ‘toad’ in (8b) ) take the same agreement marker a-. The difference in vowel quality is due to the different ATR specification of the verb-initial vowels, as expected. Also, the pronominal form of these nouns is the same as that of the other nouns in this class, as illustrated by the pronominal form for sã́hwɩ̌̃ ‘spider’ in (8c).

c. é-tsirí bu-pí nɩ́-m̄ 3SG-be.on CM-ceiling DEF-inside ‘It is on the ceiling.’

There are also some nouns in this class that do not exhibit any singular/plural distinction. They are in this class because their pronominal and subject-verb agreement forms are the same as those of the other nouns in this class. The following sentences contain some examples:

(9) a. sʊ́bhā=ń e-púlú a-wʊlakpá nɩ́-ábhā Rain=DEF SM-wash CMPL-leaf DEF-top ‘The rain washes the leaves.’ b. dzyosǔ á-lɩ̄ yɩ́ ɔ-hwɩ=ń í-shú Blood SM-be.LOC 3SG CM-arm=DEF CM-body ‘He has blood on his arm.’

Most borrowed nouns, mostly from Ewe, also belong to this class. Some loanwords retain the prefixes of the source language while others have 276 MERCEY LAMPTEY BOBUAFOR their prefixes changed or they do not occur with prefixes at all.8 Exam- ples given here are all words from Ewe origin:

(10) a-fɔmízī ‘rabbit’ < Ewe afɔmízi ‘rabbit’ a-wī ‘dress’ < Ewe awu ‘dress’ e-nyī ‘elephant’ < Ewe enyi ‘cow’ i-dzyi ‘heart’ < Ewe (e)dzyi ‘heart’ dɔkū ‘turkey’ < Ewe dɔku ‘turkey’ tédzyí ‘donkey’ < Ewe tedzyi ‘donkey’

The SM or pronominal form of the loanwords is either a- or e-, depending on the ATR value of the first syllable of the verb as is the case of the nouns in the a1- class. The a1- and ba(a)- classes (see section 3.1.2) seem to be the default classes so if there is a case of agreement mismatch, people resort to these classes. All these are indications that the noun class system is gradually decaying. It must however, be noted that the adult speakers do not take lightly to this and are always complaining that some speakers especially, the younger ones are ‘destroying the language’. With regard to the semantics of the nouns in this class, it contains most animate nouns, almost all human nouns including kinship terms, some wild and domestic animals, fluids, meteorological phenomena as well as most borrowed words. The semantics of the borrowed words have no limitations.

3.1.2 The ba(a)- class. The ba(a)- class comprises plural forms of the 1 2 a - class and the bʊ - class nouns and a sub-set of the ki- and ka- classes, which will be discussed in sections 3.1.9, 3.1.5 and 3.1.7 respectively. Nouns in this class have the plural prefixes ba(a)- or be(e)-. As usual, the ba(a)~be(e) alternation is governed by the ATR value of the initial root vowel. Ba-/be- are used to cross-reference the plural subject on the verb and they also occur as their pronominal forms. Consider the following examples:

8 Loanwords will be discussed in detail in section 3.2. NOUN CLASSES IN TAFI 277

(11) a. ba-wī=ń á-sɔmɩ́ kí-húí nɩ́-ábhā CMPL-dress=DEF SM-hang CM-drying_line DEF-top ‘The clothes are hanging on the drying line.’ (TRPS. 36) b. bá-sɔmɩ́ kí-húí nɩ́-ábhā 3PL-hangCM-drying_line DEF-top ‘They are hanging on the drying line.’

(12) a. bee-kusí=ń é-tsoku ’ʊ-pá nɩ́-m̄ CMPL-chief=DEF SM-enter CM-house DEF-inside ‘The chiefs entered the house.’ b. bé-tsoku ’ʊ-pá nɩ́-m̄ 3PL-enter CM-house DEF-inside ‘They entered the house.’ c. be-tū a-lɩ́ ɔ-ma nɩ́-m̄ CMPL-mountain SM-be.LOC CM-town DEF-inside ‘There are mountains in the town.’

The subjects of the example sentences in (11a), (12a), and (12c) are plural nouns, however, instead of the expected agreement markers ba- and be-, the ones used in these sentences are a/e-. The examples in (11a) and (12a) were produced by my consultants during elicitation sessions. Sen- tence (11a) was given during a picture-stimuli elicitation session. As it turns out, the consonantal onsets of the subject-verb agreement markers are deleted in spontaneous speech. The same has happened to the nominal complement ‘house’ in (12a) and (12b) whose citation form is bʊ-pá. In (11b) and (12b), I have the pronominal form for the nominals in this class occurring as subjects. The singular forms of ba-wī ‘clothes’ in (11a) and bee-kusí ‘chiefs’ in (12a) belong to the a1- class, while that of be-tú ‘mountains’ 2 in (12c) which is bu-tú ‘mountain’ belongs to the bʊ - class. So far, I have come across only two words, be-ní ‘rivers’ and be-dzigā ‘traders’, whose plural prefix belongs to the ba(a)- class, while their singular forms take the ke- prefix, typical of the ka- class (see section 3.1.7 below). Some nouns can take more than one plural prefix depending on how a speaker wants to construe them. For instance, some speakers use the ba(a)- prefix to mark agreement on the plural forms of kɩ-drɔgá̌ ‘bedbug’ and kɩ-zɔ̄ ‘housefly’ yielding ba-drɔgá̌ and ba-zɔ̄ respectively. 278 MERCEY LAMPTEY BOBUAFOR

The unmarked plural forms of these nouns are a-drɔgá̌ and a-zɔ̄ whose prefix is identical to the one used by the nouns of the a2- class (see section 3.1.6). Thus, if a speaker construes kɩ-drɔgá̌ ‘bedbug’ and kɩ-zɔ̄ ‘housefly’ as a distinct set of animates, for example, insects, then the a2- prefix will be selected for their plural forms. However, if they are generally construed as animate nouns then the ba(a)- prefix will be selected to form their plurals.

3.1.3 The o- class. Nouns belonging to the o- class take an o/ɔ- prefix in the singular, distributed in accordance with the ATR harmony require- ments. Examples of these nouns are:

(13) a. o-nugbū ‘mouth’ b. ɔ-tɔmɩ́ ‘beard’ o-lí ‘neck’ ɔ-ma ‘town’ o-sí ‘tree’ ɔ-shɛ̌̃ ‘branch’ o-yĩtsí ‘hawk’ ɔ-ɖá ‘metal, iron’ o-ní ‘soup’ ɔ-bhɩa ‘friend’

These nouns are cross-referenced on the verbs of which they are the subject as o/ɔ- and their subject pronominal form (l)o/(l)ɔ-. The realisation of the onset /l/ depends upon the individual speaker’s choice. Below are some examples:

(14) a. o-seyuyū=ń ɔ́-sɔmɩ́ o-sí=ń í-shú CM-fruit=DEF SM-hang CM-tree=DEF CM-body ‘The fruit is hanging on the tree.’(TRPS. 27) b. (l)ɔ-sɔmɩ́ o-sí=ń í-shú 3SG-hang CM-tree=DEF CM-body ‘It is hanging on the tree.’

(15) a. o-hui=ń ó-tsiri ’e-sukpogunū nɩ-abhā CM-rope=DEF SM-be.on CM-tree_stump DEF-top ‘The rope is on the tree stump.’(TRPS. 42) b. (l)o-tsiri ’e-sukpogunū nɩ-abhā 3SG-be.on CM-tree_stump DEF-top ‘It is on the tree stump.’ NOUN CLASSES IN TAFI 279

The subject-verb agreement marker of the a1- class also tends to be used for the o- and other classes by some speakers, especially young adults and children, as illustrated by the following sentence, which was ob- tained during a picture-stimuli elicitation session:

(16) o-seyuyū=ń á-kpasɩ́ a-gbɛ̌ nɩ́-m̄ CM-fruit=DEF SM-be.in CM-bowl DEF-inside ‘The fruit is in the bowl.’ (TRPS. 2)

Among others, nouns in the o- class include part-whole relational terms, like human and animal body parts, parts of plants and buildings, parts of the day, elements of the environment, some minerals, some animals, and some food items.

3.1.4 The i-class. The i- class is made up of the plural forms of nouns in the o- class. Depending on the [ATR] value of the first vowel of the noun root, nouns in this class take i/ɩ- as prefix, as in (17a) and (17b), and as subject-verb agreement markers, as illustrated in (18a) and (19a). They have (l)i/(l)ɩ- as their pronominal subject forms, as shown in (18b) and (19b). Depending on the individual speaker, the onset /l/ may not be realised.9

(17) a. i-nugbū ‘mouths’ b. ɩ-tɔmɩ́ ‘beards’ i-lí ‘necks’ ɩ-ma ‘towns’ i-sí ‘trees’ ɩ-shɛ̌̃ ‘branches’ i-yĩtsí ‘hawks’ ɩ-ɖá ‘metals, iron’ i-ní ‘soups’ ɩ-bhɩa ‘friends’

(18) a. i-seyuyū=ń ɩ́-sɔmɩ́ o-sí=ń í-shú CMPL-fruit=DEF SM-hang CM-tree=DEF CM-body ‘The fruits are hanging on the tree.’ (TPRS. 45) b. (l)ɩ-sɔmɩ́ o-sí=ń í-shú 3PL-hangCM-tree=DEF CM-body ‘They are hanging on the tree.’

9 The use of ɩ- and i- as subject pronouns is not common in recent years. 280 MERCEY LAMPTEY BOBUAFOR

(19) a. ɩ-kʊkɔnye=ń í-tsiri ’e-sukpogunū nɩ-abhā CMPL-cock=DEF SM-be.on CM-tree_stump DEF-top ‘The cocks are on the tree stump.’ b. (l)i-tsiri ’e-sukpogunū nɩ-abhā 3PL-be.on CM-tree_stump DEF-top ‘They are on the tree stump.’ c. i-seyuyū=ń á-kpasɩ́ a-gbɛ̌ nɩ́-m̄ CMPL-fruit=DEF SM-be.in CM-bowl DEF-inside ‘The fruits are in the bowl.’

It should be noticed that (19c) represents an example of the generalisation of the subject-verb agreement marker of the ba(a)- class to other noun classes.

3.1.5 The ki- class. The ki- class has nouns which take ki/kɩ- prefix. This class includes some nouns without a plural form, which are the ones given in (20c).

(20) a. kí-wī ‘day’ b. kɩ́-gɔ̄ ‘year’ c. kɩ́-lɛ̄ ‘wind’ kí-kū ‘yam’ kɩ-zɔ̄ ‘housefly’ ki-wí ‘sun’ ki-dri ‘wall’ kɩ-kana‘crab’ kɩ-bʊɩ́ ‘dew’

The subject-verb agreement marker for this class is ki/kɩ- which is also the pronominal form that occurs in subject position.

(21) a. ki-wí=ń í-tsú a-dzɩ=̄ ń CM-thorn=DEF SM-prick CM-woman=DEF ‘The thorn pricked the woman.’ b. kɩ-zɔ=̄ ń ɩ́-kpasɩ ’u-ní nɩ́-m̄ CM-housefly=DEF SM-be.in CM-water DEF-inside ‘The housefly is in the water.’

(22) a. ki-tsú a-dzɩ=̄ ń. 3SG-prick CM-woman=DEF ‘It pricked the woman.’ b. kɩ-kpasɩ ’u-ní nɩ́-m̄ 3SG-be.in CM-water DEF inside ‘It is in the water.’ NOUN CLASSES IN TAFI 281

As was explained earlier, most often, speakers tend to delete initial consonants in less careful speech. In the examples in (21), the subject-verb agreement markers are represented by the prefixes i- in (21a) and ɩ- in (21b), instead of ki- and kɩ-. Also, the initial consonant of the bu- prefix of bu-ní ‘water’ is not realised in (21b) and (22b). When I once tried to produce the same sentences with the complete prefixes, my language consultants reacted saying “if you speak like that, everybody will know you are a learner of the language or a stranger”. Moreover, if the full form of the bu- prefix of bu-ní ‘water’ is used in sentences like (21b) and (22b), the native speakers’ judgment is that ‘the sentence is not wrong but that it is not how we say it’. The generalisation of the subject-verb agreement marker of the a1- class is exemplified in the sentence (22c). This was a report made to one my language consultants by her niece who was asked to cook yam for lunch. In this sentence, the subject-verb agreement marker used is a- instead of kɩ-.

(22) c. kí-kū=n á-tsyɔ́mɔ.̄ CM-yam=DEF SM-rot ‘The yam is rotten.’

The ki- class contains semantically diverse nouns including some items whose shape is flat and long, like road, lizard, and spear; certain items with a round shape, like egg, gourd and stone; some abstract nouns, emo- tions, like fear; most body parts including head, brain, hair, thigh, bone, vagina, testicles; some household items, farm products, some insects, and some water-related animals, like frogs, among others.

3.1.6 The a2- class. The a2- class has been identified as a separate class, because it comprises plural nouns whereas the a1- class contains singular nouns. It pairs with the ki- class which is a singular class. Moreover, the pronominal agreement form also differs from that of the a1- class. Noun roots with a [-ATR] initial vowel occur with the a- prefix, as in (23a), and those with [+ATR] initial vowel with the prefix e-, as in (23b).

(23) a. á-gɔ̄ ‘years’ b. e-wí ‘days’ a-zɔ̄ ‘houseflies’ é-kū ‘yams’ a-kana ‘crabs’ e-dri ‘walls’ 282 MERCEY LAMPTEY BOBUAFOR

The subject-verb agreement marker of this class is the same as the prefixes and their subject pronominal form is la/le-. In the examples in (24), the full form of ’ifū ‘fear’ is kífū, of which the initial consonant is not realised.

(24) a. bé-fleté=ń a-ńflɔ̄ e-te-kú mɩ ’i-fū CMPL-leopard=DEF CMPL-claw SM-NEG-reach 1SG CM-fear ‘I am not afraid of the leopard’s claws.’ Lit.: ‘The leopards’ claws do not reach my fears.’ b. le-te-kú mɩ ’i-fū 3PL-NEG-reach 1SG CM-fear ‘I am not afraid of them’. (i.e., the leopards’ claws) Lit.: ‘They do not reach my fears.’

3.1.7 The ka- class. Nouns belonging to the ka- class take a ka- or ke- prefix. Noun roots with an initial [-ATR] vowel take a ka- prefix while those with an initial [+ATR] vowel take a ke- prefix. The subject-verb agreement marker is ka/ke- as shown in (25a) and (26a). In these examples also, the initial consonant of the subject-verb agreement markers is absent. The corresponding subject pronouns are also ka/ke-, as shown in (25b) and (26b).

(25) a. ka-hlɔɛ̃́=ń é-tsí b. ke-tsí CM-antelope=DEF SM-die 3SG-die ‘The antelope died.’ ‘It died.’

(26) a. ke-plukpá=ń á-wɔ́lɩ̄ b. ka-wɔ́lɩ̄ CM-book=DEF SM-fall 3SG-fall ‘The book fell.’ ‘It fell.’

Nouns in this class include a few human terms referring to age, like ke- bité ‘young woman’; some body parts, such as kahʊkpɔ́ ‘hand’; wild and domestic animals including kehúí ‘antelope’ and kakudzɔgɛ̌ ‘dog’; birds, farm implements and other entities, such as ground, land or parts of a tree. NOUN CLASSES IN TAFI 283

1 1 3.1.8 The bʊ - class. Nouns in the bʊ - class take a bʊ/bu- prefix. The pronominal forms of these nouns are bʊ/bu-, just as the prefixes as illustrated in (28b). The nouns in this class represent the plural forms of the nouns in the ka- class and a subset of the ki- class. Examples of such nouns are:

(27) a. bʊ-hlɔ ‘antelopes’ b. bu-plukpá ‘books’ bʊ-pā ‘hoes’ bu-tsé ‘monkeys’ bʊ-tsrʊkpɔ́ ‘feet’ bu-tsikpǐ ‘pots’ bʊ-dzɩmɩ̌ ‘young girl’ bu-bité ‘young women, maidens’ (28) a. bu-kplukpá ʊ́-lɩ́lɩ́nɩ́ pétéé tã dzí á-fɔɛ́ CMPL-book AM-DEM all burn become CMPL-charcoal ‘All those books burnt into ashes.’ b. bʊ-ʊ́-lɩ́lɩ́nɩ́ pétéé tã dzí á-fɔɛ́ 3SG-AM-DEM all burn become CMPL-charcoal ‘All those ones burnt into ashes.’

As shown in (28a), the subject is not cross-referenced on the verb. It appears that the nouns in this class are not cross-referenced on verbs. This is one of the signs that the noun class system is gradually decaying. ʊ́ is an agreement marker prefixed to the demonstrative -lɩ́lɩ́nɩ́ ‘that’, if it oc- 1 curs with nouns of the bʊ - class.

2 2 3.1.9 The bʊ - class. The nouns belonging to the bʊ - class take a bʊ/bu- prefix. The nouns in this class belong to a small set, consisting of body-parts like bʊ́-nyɩ́nyɩ́ ‘penis’ and paired body-parts like bʊ-tɔ́ ‘ear’, bʊ-zhɔ10̄ ‘cheek’, and bʊ́-wʊǹdɔ̄ ‘shoulder; non-count nouns like bú- hihē ‘sweat’, bu-ní ‘water’ and bʊ-tɔ́ ‘ashes’. Some others are bu-tú ‘mountain’, bú-vū ‘building’ and bʊ-yā ‘farm’. The pronominal form for nouns in this class is bʊ/bu- as shown below.

(29) a. bʊ-mwɩ̄ kpe faánɔ̄ nɩ́ o-ní nɩ́-m̄ CM-salt be.plenty much LOC CM-soup DEF-inside ‘There is too much salt in the soup.’

10 The IPA symbol for ‘zh’ is /ʒ/. ‘zh’ is the orthographic representation. 284 MERCEY LAMPTEY BOBUAFOR

b. bu-kpe faánɔ̄ nɩ́ o-ní nɩ́-m̄ 3SG-be.plenty much LOC CM-soup DEF-inside ‘It is too much in the soup.’

1 Even though the bʊ - class takes the same noun prefix and pronominal 2 form as the bʊ - class, in the former class the form marks the plural of nouns that belong to the ka- class and a subset of the ki- class whereas in the latter class it prefixes to singular and non-count nouns. Example (29a) is yet another example showing that the noun class system is in the process of decaying. In this sentence, the verb kpe ‘be plenty’ occurs without any subject-verb agreement marker. 2 The class prefix bʊ is prefixed to some verbs to form gerunds as illustrated below:

(30) a. bu-yí ‘killing’ ← yí ‘kill’ bʊ-pʊɩ̄ ‘scattering’ ← pʊɩ̄ ‘scatter’ bʊ́-gā ‘walking’ ← gā ‘walk’ bu-búkú ‘intoxicating’ ← búkú ‘intoxicate’

2 The bʊ - prefix can also be added to adjectives to derive nominals. The following are examples.

(30) b. bhulíyī ‘very small’ > bú-bhulíyi ̄ ‘smallness’ dɩdamā ‘tall’ > bʊ́-dɩdamā ‘tallness’ gbʊgblǎ ‘big’ > bʊ́-gbʊgblǎ ‘bigness’ ‘ kpukpomlī ‘short’ > bú-kpukpomlī ‘shortness

3.1.10 The tɩ- class. The tɩ- class nouns take the tɩ/ti- prefix. This class contains collective nouns which are mainly non-liquid such as:

(31) tɩ-wʊlɛ́ ‘intestines tɩ-pɩ́ ‘excrement’ tɩ-wá ‘grass’ tɩ-wʊlɔ̄̃ ‘rubbish’ tɩ́-kā ‘hair’ tɩ-rá ‘sleep’ tɩ-sɩ́ ‘feather’ ti-sí ‘clay, soil’

The subject-verb agreement marker and the subject pronominal form for this class is ki/kɩ-. In (32a), the initial consonant of the subject-verb agreement marker is not realised. NOUN CLASSES IN TAFI 285

(32) a. tɩ-wá=ń í-í-shē11 ziaziā kóŋ́ CM-grass=DEF SM-PRSPROG-grow RED-fast very_much ‘The grass is growing very fast.’ b. kí-í-shē ziaziā kóŋ́ 3SG-PRSPROG-grow RED-fast very_much ‘It is growing very fast.’

The prefix tɩ- is also used for the nominalisation of verbs, as the following examples show:

(33) a. tí-yíyékē ‘swelling’ ← yékē ‘swell’ tɩ́-zɩzā ‘staying, being’ ← zā ‘stay, be’ tɩ́-ŋɩŋā ‘eating’ ← ŋā ‘eat’ tɩ́-vɩvɩ ‘going’ ← vɩ ‘go’

In the process of nominalisation, the initial (or only) syllable of the verb is reduplicated before the prefix is added, while the reduplicated part contains a vowel pre-specified as [+high], which adapts its ATR value to the [ATR] value of the initial vowel of the verb root. Moreover, tɩ/ti- is the prefix attached to language names. For example,

(33) b. Ti-trugbu ‘the language of the people of Nyagbo’ Ti-yĩgbē ‘Ewe’ Ti-mũí ‘Akan’ Tɩ-frańsɩ̄̃ ‘French’

In the foregoing, I have discussed the way in which Tafi organises its noun classes. I have also looked at the pronominal forms of the nouns in the various classes and the subject-verb agreement markers which are used to cross-reference these nouns on the verbs they occur with. It has become apparent that, though the prefixes of some of the classes, their pronominal forms, or subject-verb agreement markers have the same form, they perform different functions. For these reasons, I have pro-

11 The IPA symbol for ‘sh’ is /ʃ/. ‘sh’ is the orthographic representation. 286 MERCEY LAMPTEY BOBUAFOR posed a system of noun classes that contains ten different categories, some of which are based on the shape and use of the prefix in singular nouns, some others on the shape and the use of the prefix in plural forms.

3.2 Loanwords It is well known that languages in contact influence one another. As a minority language in the region, Tafi commonly borrows from Ewe, Akan, and English. In view of the fact that the contact between the Tafis and especially the Ewes has been intense, the flow of borrowings from Ewe seems to be greater than from Akan and English. Most often, a borrowed form or pattern diverges from its shape in the source language. For example, the Tafi words a-kana ‘crabs’ and ɔ- ma ‘town’ are borrowed from Ewe agala ‘crab’ and Akan ɔmán ‘town’, respectively. It is also possible that in the borrowing process a semantic or functional change takes place, for instance e-nyi ‘cow’ in Ewe becomes ‘elephant’ in Tafi. Other loans are integrated without modification, as the Ewe words a-sra ‘tobacco’ and vudǒ ‘well’. Thus, loanwords vary in terms of their degree of phonological and morphological integra- tion into the borrowing language. When formal changes occur, these are usually determined by the structure of the borrowing language. In this part of the paper I will address the question of how borrowed nouns are incorporated into the noun class system of Tafi. Heine (1968) mentions three methods of integrating these nouns: (i) automatic, (ii) phonological and (iii) semantic allocations. I. In many noun class languages, loanwords are put in only one or few of the available classes. In Tafi, for instance, the default classes are the a1- and ba(a)- classes. In a count conducted in 2007 in the then available corpus of nouns, the greater number of loan words is found in these classes. Out of a total of 112 nouns that are part of the a1- class, about 29 are loanwords; some of these nouns form their plural in the ba(a)- class; 23 out of about 131 nouns that are part of the ba(a)- class are borrowed. Some of the nouns that Tafi borrowed do not have nominal class prefixes. Examples of such nouns include sɩká ‘money’, borrowed from Akan and kɔɖú ‘banana’ borrowed from Akan (kwadu) via Ewe (akɔɖú). These loanwords are prefix-less and their singular and plural variants are the same. The subject-verb agreement marker these nouns take is a/e- depending on the initial vowel of the verb, which signals that NOUN CLASSES IN TAFI 287 they are functioning like a1- class nouns as illustrated by the examples in (34). (34) kɔɖú=ń é-békē banana=DEF SM-finish ‘The banana is finished.’

II. Some borrowed nouns are allocated to certain noun classes, because their initial syllable bears phonological resemblance to the singular prefix of a particular class, as in (35a-c).

Singular Plural (35) a. a-gbeɖi ‘cassava’ ba-gbeɖī ‘cassava’ < Ewe agbeli id. b. a-wī ‘dress’ ba-wī ‘dresses’ < Ewe awu id. c. o-dotí ‘cotton’ i-dotí ‘cotton’ < Gã, odóntí id.

The loanwords in examples (35a)-(35c) are interpreted in Tafi as consisting of a prefix and a root. Thus, the initial vowels of these loanwords are identified with the singular prefix of the a1- class, as in examples (35a) and (35b), and of the o- class, as in example (35c).

III. Some nouns that Tafi has borrowed from other languages are integrated in the class system on the basis of their semantics. Nouns referring to animate entities are allocated to the a1- and ba(a)- classes, which mainly contain animate nouns. Some examples are given below:

a1- class ba(a)- class (36) a. a-dzyatá ‘lion’ ba-dzyatá ‘lions’ < Ewe, dzyatá b. memě ‘ant’ be-memě ‘ants’ < Ewe, memē c. tédzyí ‘donkey’ be-tédzyí ‘donkeys’ < Ewe, tédzyí

Although one might argue that some of the loanwords above were allocated to these classes on phonological grounds, examples (36b) and (36c) disprove this argument. As one observes, these loanwords are prefix-less in the singular, while their plural forms have been assigned to the ba(a)- class, which is the plural class for the a1- class. Also, the subject-verb agreement marker is a/e- as in (36d), in which a- is used as the subject marker on the verb tã ‘chew’ for both e-pidzyā ‘goat’ and tédzyí ‘donkey. 288 MERCEY LAMPTEY BOBUAFOR

(36) d. epidzyā=ń á-tã̄ a-wʊlakpá=ń kɩlɩ́ tédzyí nɩ́ tsyí CM-goat=DEF SM-eat CMPL-leaf=DEF CONJ donkey DEF too á-tã̄ tɩ-wá=ń SM-eat CM-grass=DEF ‘The goat ate the leaves while the donkey also ate the grass.’

There are some loanwords that are allocated to a given noun class for both semantic and phonological reasons. For example, the words in example (37a) and (37b) were allocated to the ki- and a2- classes, because these classes contain, among others, water-related animals like ‘fish’ and ‘crab’, whereas those in (37c) and (37d) have also been assigned to the a1- and ba(a)- classes, which mainly consist of animate nouns.

(37) a. kɩ-kpã̌ ‘fish’ a-kpã̌ ‘fishes’ < Ewe akpā b. kɩ-kana ‘crab’ a-kana ‘crabs’ < Ewe agalã c. a-dzyramʊá ‘cat’ ba-dzyramʊá ‘cats’

Phonologically, the plural forms of the loanwords in (37a) and (37b) are considered to belong to the a2- class, because their a- prefix is the same as the plural prefix of this class. Consequently, their singular forms are assigned the kɩ- prefix. The singular forms of the loanwords in (37c) and (37d) have been allocated to the a1- class, because their a- prefix is the same as the singular prefix of this class and their plural forms have been allocated to the ba(a)- class. The match between the singular and plural classes in Tafi and how they interrelate is represented by Figure 1. As shown in the figure, 1 2 the ba(a)- class is made up of plurals of nouns in the a - and bʊ - classes 1 and a subset of the ka- and ki- classes. Also, the bʊ - class comprises most of the plural forms of the ka- class and a subset of the ki- class. The tɩ- class contains non-count nouns so it has no plural counterpart. NOUN CLASSES IN TAFI 289

Singular Non-count Plural a1- class ba(a)- class o- class i- class ki- class a2- class 1 ka- class bʊ - class 2 bʊ -class tɩ-class Figure 1

In the next section, I will look at agreement within the noun phrase, to support the prefix classes so far identified as well as distinguish between the classes that have identical forms but are treated as different classes.

4.Agreement within the NP A noun phrase in Tafi consists of a head noun followed by attributive modifiers. The modifiers can be any of the following occurring in the following order: adjective, quantifier, interrogative qualifier, determiner, and intensifier. Some of these modifiers can be made to function as heads of the noun phrase when they are nominalised. Some modifiers agree with the head noun. However, adjectives, interrogative qualifiers and intensifiers do not agree with the head noun. Ordinals, unlike other quantifiers, behave like qualifiers in the sense that they do not show agreement with the head noun. For quantifiers, for example, the agreement prefix is a quantifier prefix t- and a concord vowel that agrees with the vowel of the noun prefix. This is a form of vocalic concord (Schuh, 1995). Example (38) illustrates a noun phrase in Tafi.

(38) be-tú dɩdamā tɩa-tá á-lɩ́lɩ́n pétéé CMPL-mountain tall AM-three AM-DEM all ‘All those three tall mountains’

In the above sentence, the head noun is be-tú ‘mountains’ while dɩdamā ‘tall’ which is an adjective qualifying the noun does not agree with the head noun; tɩa-tá ‘three’ is a numeral. tɩa- is the agreement prefix showing concord with the head noun; á-lɩ́lɩ́n ‘those’ is a demonstrative with an agreement prefix which is the vowel of the ba(a)- class to which the head noun belongs; and pétéé ‘all’ is an intensifier. 290 MERCEY LAMPTEY BOBUAFOR

4.1 Head noun and adjectives In Tafi, adjectives do not show any concord with the head noun. As shown in the (a) sentences below, the adjectives gbɩgblǎ ‘great’ (39a) and dzyrodō ‘long’ (40a) and (41a) have no prefix. Adjectives can be nominalised by the addition of the nominal prefixes, except for the tɩ- prefix. The tɩ- prefix cannot be added to any modifier in order to nominalise it. Rather, it is its subject pronoun form that is prefixed to the modifier as shown in (41b). When the adjectives are nominalised, they control agreement, as illustrated in (39b), (40b) and (41b). As expected, the [ATR] value of the vowel in the prefixed pronoun depends on the [ATR] value of the initial vowel in the adjectival base as in the following examples.

(39) a. bá-nyɩ́ gbɩgblǎ tɩa-lɩ́ á-lɩ́lɩ́ń pétéé é-nū CMPL-men great AM-four AM-DEM all SM-be bá-gbalɩgbā CMPL-bachelor ‘All those four great men are bachelors.’ b. ba-gbɩgblǎ tɩa-lɩ́ á-lɩ́lɩ́ń pétéé é-nū bá-gbalɩgbā CMPL-great AM-four AM-DEM all SM-be CMPL-bachelor ‘All those four great ones are bachelors.’

(40) a. bú-vu kɩ-kpa dzyrodō=ń 3PL-catch CM-fish long=DEF ‘We caught the long fish.’ b. ki-dzyrodō=ń í-tsirí kplɔ̌̃ nɩ́-ábhā CM-long=DEF SM-be.on table DEF-top ‘The long one is on the table.’

(41) a. tɩ-sɩ́ dzyrodō=ń í-dzyínī CM-feather long=DEF SM-break ‘The long feather is broken.’ b. ki-dzyrodō=ń í-dzyínī CM-long=DEF SM-break ‘The long one is broken.’

In (39b), and (40b), the nominal prefixes of bá-nyɩ́ ‘men’ and kɩ-kpa ‘fish’ have been attached to the adjectives gbɩgblǎ ‘great’ and dzyrodō NOUN CLASSES IN TAFI 291

‘long’, respectively, whereas in (41b), it is the subject pronoun for the tɩ- class that has been prefixed to the adjective dzyrodō ‘long’. The ATR value of the vowel in the prefix of ki-dzyrodō ‘long one’ in (40b) and (41b) is harmonic with the initial vowel of the adjective. In these examples, ba-gbɩgblǎ ‘great ones’ and ki-dzyrodō ‘long one’ are pronomi- nals, and can function as nouns.

4.2 Head noun and numeral In Tafi, the counting forms of cardinal numbers are different from the form they take in attributive position. In the counting form, the words for six, seven, and nine have a zero prefix.

(42). o-lí - ‘one’ ɩ-bhā - ‘two’ ɩ-tá - ‘three’ ɩ-lɩ̃́ - ‘four’ i-tí - ‘five’ holō - ‘six’ géné - ‘seven’ a-sʊɩ̄ - ‘eight’ zhitá or hitá12 - ‘nine’ kɩ́-fɔ̄ - ‘ten’ kɩ́-fɔ n’t-olí13 - ‘eleven’ (ten and one) kɩ́-fɔ n’tɩa-bhā - ‘twelve’ (ten and two) kɩ́-fɔ n’tie-tí - ‘fifteen’ (ten and five) á-ftabha - ‘twenty’ á-ftabha n’t-olí - ‘twenty-one’ (twenty and one) á-ftabha n’tɩa-bhā - ‘twenty-two’ (twenty and two) á-ftabha n’tie-tí - ‘twenty-five’ (twenty and five)

The counting forms of the numbers ‘eleven’, ‘twenty-one’, ‘thirty-one’, etc., are formed with the words for ‘ten’ or its multiples, the linker nɩ ‘and’ whose vowel is deleted, and the quantifier prefix t-, attached to the counting form of ‘one’. The numerals ‘twelve’ to ‘nineteen’, ‘twenty- two’ to ‘twenty-nine’, thirty-two’ to ‘thirty-nine’, etc., are formed in the same way, except that the roots of the numerals ‘two’ up to ‘nine’ are prefixed with tɩa- or tie-, depending on the [ATR] value of the vowel in

12 Zhitá and hitá are used interchangeably. 13 The mid tone on the final syllable of kɩ́-fɔ̄ ‘ten’ becomes low, because the following word nɩ ‘and’ has a low tone. 292 MERCEY LAMPTEY BOBUAFOR the root. The numerals ‘hundred’ and ‘thousand’ are expressed with alafá and akpé respectively which are loanwords from Ewe. In attributive position, the numerals, especially ‘one’ to ‘nine’, take a prefix made up of the quantifier prefix t- and a concord vowel that agrees with the vowel of the noun prefix14. Thus, agreement is shown on the numbers ‘one’ to ‘nine’, depending on the noun class. The number ‘ten’ has a form different from the counting form also, which is lɔ́-fɔ.̄ However, this functions as a default form, which does not agree with the noun it modifies, so it goes with all the classes. The numbers ‘twenty’, ‘thirty’, etc. do not agree with the head noun either. They maintain the same form when they occur with a noun from any of the noun classes. As regards the numbers ‘eleven’, ‘twenty- one’, ‘thirty-one’, etc., their attributive forms are identical to their counting forms, besides ‘eleven’, which is formed by using lɔ́-fɔ,̄ the modifier form of ‘ten’, as shown in Table 3 below. Moreover, the attributive forms of the numerals ‘twelve’ to ‘nineteen’, ‘twenty-two’ to ‘twenty-nine’, thirty-two’ to ‘thirty-nine’, etc., are also formed by using the ‘words for ‘ten’ or its multiples, the linker nɩ ‘and’, without its vowel and the attributive forms of ‘two’ to ‘nine’. Thus, for instance, if we have the number ‘twenty-five’ modifying a noun, it is only the number ‘five’ that agrees with the head noun whereas the form for ‘twenty’ remains the same, as illustrated in Table 3.

Class Concord 1 a - e-kusí té-ɖíkpó ‘one chief’ ba(a)- bee-kusí tɩ́á-bhā ‘two chiefs’ bee-kusí lɔ́-fɔ n’t-o-lí ‘eleven chiefs’ bee-kusí á-ftabha n’tie-tí ‘twenty-five chiefs’ o- o-nugbū to-ɖukpó ‘one mouth’ i- i-nugbū tɩ-bhā ‘two mouths’ i-nugbū lɔ́-fɔ n’t-o-lí ‘eleven mouths’ i-nugbū á-ftabha n’ti-tí ‘twenty-five mouths’

14 For the time being, the distinction between the form the numeral ‘one’ takes in modifier position (i.e., teɖikpó, trikpó etc.) and the form it takes after ‘ten’ and its multiples (i.e., t-o-lí) is not clear. Also, the numeral modifiers of the ba(a)- class maintain the counting forms of ‘two’ to ‘nine’ which occur after ‘ten’ and its multiples. NOUN CLASSES IN TAFI 293

ki- kɩ-tɔkpú tríkpó ‘one head’ 2 a - a-tɔkpú tá-bhā ‘two heads’ a-tɔkpú lɔ́-fɔ n’t-olí ‘eleven heads’ a-tɔkpú á-ftabha n’te-tí ‘twenty-five heads’ ka- ká-pā tie-ɖikpó ‘one hoe’ 1 bʊ - bʊ-pā tʊ-bhā ‘two hoes’ bʊ-pā lɔ́-fɔ n’t-o-lí ‘eleven hoes’ bʊ-pā á-ftabha n’tu-tí ‘twenty-five hoes’ 2 bʊ -15 bʊ-tɔ́ trúkpó ‘one ear’ tɩ- There are no count nouns in this class. Table 3: Cardinal numerals

The examples in Table 3 are an illustration of the concord between nouns in the various classes and the numerals they occur with.

4.3 Head noun and ordinals Ordinal numbers in the language are formed by adding the formative -mɩ̄ ‘ORD’ to the roots of numbers, except for tʊtɔpʊ16̄ ‘first’, bhlā ‘second’ and gʊgɔɛ́17 ‘last’. To form the ordinal ‘tenth’, the ordinal suffix is added to the word for ‘ten’ kɩ́-fɔ.̄ The root of ‘two’ has /l/ inserted after its initial consonant, while those of ‘three’ and ‘five’ have /r/ inserted after their initial consonant. The roots of ‘four’ and ‘seven’ are reduplicated. The final vowel of the root of ‘eight’ is elided, when the ordinal suffix is added to it. When the ordinals are used in isolation without modifying a noun, the form the tenth ordinal takes is kɩ́-fɔ-̄ mɩ̄ as in (43a), however when it modifies a noun, the form it takes is lɔ́-fɔ-̄ mɩ̄ which comprises the modifier form of ‘ten’ lɔ́-fɔ̄ and the ordinal suffix -mɩ̄as in (43f). Or- dinal numbers in the language do not take prefixes to mark agreement with the head noun, as shown in (43b) and (43d). The ‘first’ to ‘tenth’ and ‘twentieth’ ordinals are illustrated in (43a). These ordinals may, however, be nominalised in the same way as the adjectives as shown in

15 2 The plural forms of nouns in the bʊ - class belong to the ba(a)- class. 16 tʊtɔpʊ̄ is derived from the verb putɔ̄ ‘lead’. 17 gʊgɔɛ́ is derived from the verb gɔ́ ‘remain’. 294 MERCEY LAMPTEY BOBUAFOR

(43c) and (43e). When they are nominalised, they control agreement as in (43e).

(43) a. tʊtɔpʊ̄ - ‘first’ gi-géné-mɩ̄ - ‘seventh’ bhlā - ‘second’ sʊ-mɩ̄ - ‘eighth’ trǎ-mɩ̄ - ‘third’ zhitá-mɩ̄ - ‘ninth’ lɩ̃-lɩ̃-mɩ̄ - ‘fourth’ kɩ́-fɔ-̄ mɩ̄ - ‘tenth’ trǐ-mī - ‘fifth’ á-ftabha-mɩ̄ - ‘twentieth’ hlō-mɩ̄ - ‘sixth’ b. a-nʊ́vɔ̄ lɩ̃-lɩ̃-mɩ=̄ ń CM-child RED-four-ORD=DEF ‘The fourth child’ c. á-lɩ̃-lɩ̃-mɩ=̄ ń CM-RED-four-ORD=DEF ‘The fourth one’ (referring to the child) d. o-sí lɩ̃-lɩ̃-mɩ̄ =ń ó-dzyínī CM-tree RED-four-ORD=DEF SM-break ‘The fourth tree broke.’ e. ɔ-lɩ̃-lɩ̃-mɩ=̄ ń ó-dzyínī CM-RED-four-ORD=DEF SM-break ‘The fourth one broke.’ (referring to the tree) f. ɔ-lɔ́-fɔ-̄ mɩ=̄ ń ó-dzyínī CM-Pfx-ten-ORD=DEF SM-break ‘The tenth one broke’ (referring to the tree).

To form the ordinals ‘eleventh’ to ‘nineteenth’, the modifier form of the numeral ‘ten’, lɔ́-fɔ,̄ is used, followed by the linker nɩ ‘and’ (without its vowel) and the nominalised forms of the ‘first’ to ‘ninth’ ordinals for the various noun classes, as illustrated in example (43g). The ordinals ‘twenty-first’ to twenty-ninth’, thirty-first’ to ‘thirty-ninth’, etc., are formed in the same way by using the words for the multiples of ‘ten’, as in (43i). Examples (43h) and (43j) are pronominalised forms of (43g) and (43i), respectively.

(43) g. o-sí lɔ́-fɔ̄ n’ ɔ-trǎ-mɩ=̄ ń CM-tree PFX-ten LINK CM-three-ORD=DEF ‘The thirteenth tree.’ NOUN CLASSES IN TAFI 295

h. lɔ́-fɔ̄ n’ ɔ-trǎ-mɩ=̄ ń PFX-ten LINK CM-three-ORD=DEF ‘The thirteenth one.’ (referring to the tree) i. ka-kudzɔgɛ̌ aftabha nɩ ka-tʊtɔpʊ=̄ ń CM-dog twenty LINK CM-first=DEF ‘The twenty-first dog.’ j. áftabha nɩ ka-tʊtɔpʊ=̄ ń twenty LINK CM-first=DEF ‘The twenty-first one.’ (referring to the dog)

4.4 Head noun and indefiniteness marker The indefiniteness marker in Tafi has the root -lí ‘a, a certain, some’, which is the same form used for ‘one’ in counting. When this marker and other numerals modify a noun, they take the quantifier prefix t- and a concord vowel which corresponds to the noun class prefix vowel except for the ka- class. This concord vowel harmonises with the [ATR] value of the stem-initial vowel. The prefix vowel of the indefiniteness marker is illustrated in Table 4.

Class Concord 1 a - a-nʊ́vɔ̄ t-e-lí ‘a, some, a certain child’ ba(a)- baa-nʊ́vɔ̄ t-e-lí ‘some children’ o- o-nugbū t-o-lí ‘a, some mouth’ i- i-nugbū t-i-lí ‘some mouths’ ki- kɩ-tɔkpú t-i-lí ‘a, some head’ 2 a - a-tɔkpú t-e-lí ‘some heads’ ka- ke-dē t-ie-lí ‘a, some back’ 1 bʊ - bu-dē t-u-lí ‘some backs’ 2 bʊ - bʊ-tɔ́ t-u-lí ‘an, some ear’ tɩ- tɩ-wʊlɛ́ t-i-lí ‘some intestines’ Table 4: The Indefiniteness Marker

4.5 Head noun and demonstrative In Tafi, demonstratives, like the indefiniteness marker, are formed from stems that have to be marked for agreement with the noun head. The demonstrative stems express the relative distance of the referent from the 296 MERCEY LAMPTEY BOBUAFOR

deictic centre. There are two demonstratives, proximal -nɩ́18 ‘this’ and distal -lɩ́lɩ́nɩ́19 ‘that’. They agree with the head noun. The agreement markers are vowel prefixes which correspond to the noun class prefix vowel. These concord vowels are always [-ATR], because the vowels in 1 2 the roots of the demonstratives also are. Agreement in the bʊ - and bʊ - classes is also dependent on the height of the initial vowel of the noun root. When the initial vowel of the noun root is [+high], the agreement marker of the demonstrative is the [+back], [+high], [-ATR] vowel /ʊ/. When it is [-high], the agreement marker is the [+back], [-high], [-ATR] vowel /ɔ/. The demonstratives are illustrated in Table 5.

this that Examples with -nɩ́ Examples with -lɩ́lɩ́nɩ́ 1 a - a-nɩ́ á-lɩ́lɩ́nɩ́ ekusí a-nɩ́ ‘this chief’ ekusí á-lɩ́lɩ́nɩ́ ‘that chief’ ba(a)- a-nɩ́ á-lɩ́lɩ́nɩ́ beekusí a-nɩ́ ‘these chiefs’ beekusí á-lɩ́lɩ́nɩ́ ‘those chiefs’ o- ɔ-nɩ́ ɔ́-lɩ́lɩ́nɩ́ onugbū ɔ-nɩ́ ‘this mouth’ onugbū ɔ́-lɩ́lɩ́nɩ́ ‘that mouth’ i- ɩ-nɩ́ ɩ́-lɩ́lɩ́nɩ́ inugbū ɩ-nɩ́ ‘these mouths’ inugbū ɩ́-lɩ́lɩ́nɩ́ ‘those mouth’ ki- ɩ-nɩ́ ɩ́-lɩ́lɩ́nɩ́ kíkū ɩ-nɩ́ ‘this yam’ kíkū ɩ́-lɩ́lɩ́nɩ́ ‘that yam’ 2 a - a-nɩ́ á-lɩ́lɩ́nɩ́ ékū a-nɩ́ ‘these yams’ ékū á-lɩ́lɩ́nɩ́ ‘those yams’ ka- a-nɩ́ á-lɩ́lɩ́nɩ́ kedē a-nɩ́ ‘this back’ kedē á-lɩ́lɩ́nɩ́ ‘that back’ 1 bʊ - ɔ-nɩ́ ɔ́-lɩ́lɩ́nɩ́ bʊ́gbɔ́ ɔ-nɩ́ ‘these squirrels’ bʊ́gbɔ́ ɔ́-lɩ́lɩ́nɩ́ ‘those squirrels’ ʊ-nɩ́ ʊ́-lɩ́lɩ́nɩ́ budē ʊ-nɩ́ ‘these backs’ budē ʊ́-lɩ́lɩ́nɩ́ ‘those backs’ 2 bʊ - ɔ-nɩ́ ɔ́-lɩ́lɩ́nɩ́ bʊtɔ́ ɔ-nɩ́ ‘this ear’ bʊtɔ́ ɔ́-lɩ́lɩ́nɩ́ ‘that ear’ ʊ-nɩ́ ʊ́-lɩ́lɩ́nɩ́ búvū ʊ-nɩ́ ‘this building’ búvū ʊ́-lɩ́lɩ́nɩ́ ‘that building’ tɩ- ɩ-nɩ́ ɩ́-lɩ́lɩ́nɩ́ tɩsɩ́ ɩ-nɩ́ ‘these feathers’ tɩsɩ́ ɩ́-lɩ́lɩ́nɩ́ ‘those feathers’ Table 5: Demonstratives

Furthermore, demonstratives can be used substantively when formally marked, as shown in (44b), (44d) and (44f). The demonstrative pronouns are formed by attaching the class prefixes (column 2 on Table 2) to the

18 The proximal demonstrative -nɩ́ ‘this’ is identical in form with the definiteness marker nɩ́ ‘the’. However, the former always have to take a prefix to show agreement with the head noun whereas the latter is a clitic which does not require a prefix because it does not show concord with the head noun. 19 Most often, the final vowel of -lɩ́lɩ́nɩ́ ‘that’ is deleted, however, in formal speech it is maintained. NOUN CLASSES IN TAFI 297 modifying form of the demonstratives, which comprises the prefix and the root. However, the demonstrative pronoun for the tɩ- class is formed by attaching the subject pronoun of this class to the demonstrative stems. In the case of the demonstrative -nɩ́ ‘this’, its root is reduplicated20. The initial syllable of the reduplicated root bears a low tone.

(44) a. bee-kusí-a-nɩ́ ‘these chiefs’ b. ba-á-nɩnɩ́/ ba-á-nɩň ‘these ones’ c. bee-kusí á-lɩ́lɩ́nɩ́ ‘those chiefs’ d. ba-á-lɩ́lɩ́nɩ́ ‘those ones’ e. bú-vū-ʊ-nɩ́ ‘this building’ f. bʊ-ʊ́-nɩnɩ́/ bʊ-ʊ́-nɩň ‘this one’

4.6 Head noun and interrogatives In this section, I look at two interrogative modifiers, one is a quantifier, the other is a qualifier.

4.6.1 The interrogative quantifier -shɩ̄̃. The interrogative quantifier -shɩ̄̃ ‘how many/much’ agrees with the noun it occurs with. Its agreement prefix has the form tVV- where t- is the quantifier prefix and the VV in the prefix represents the lengthened copy of the prefix vowel of the head noun, except for the ba(a)- class prefix. The prefix of the agreement marker for the ba(a)- class is the same as the one used for the numerals ‘two’ to ‘nine’ when conjoined to ‘ten’ and its multiples (see section 4.2). For semantic reasons, -shɩ̄̃ ‘how many/much’ is used only for plural and uncountable nouns. When it occurs with countable nouns it is interpreted as ‘how many?’, and when it occurs with uncountable nouns, as ‘how much?’ Also, when the head noun ends in a final high tone, this tone spreads rightwards to all but the last syllable of the quantifying word.

20 In informal speech, the final vowel of the reduplicated form -nɩnɩ́ ‘this’ is elided. The resultant form is -nɩn.̌ The rising tone on the vowel in -nɩň suggests that the high tone on the elided final vowel has been attached to the low tone of the preceding syllable. The full form -nɩnɩ́ is used especially in emphatic speech. 298 MERCEY LAMPTEY BOBUAFOR

Class Concord ba(a)- bee-kusí tɩ́á-shɩ̄̃? ‘how many chiefs?’ i- i-nugbū tɩɩ-shɩ̄̃? ‘how many mouths?’ 2 a - a-tɔkpú táá-shɩ̄̃? ‘how many heads?’ 1 bʊ - bʊ-pā tʊʊ-shɩ̄̃? ‘how many hoes?’ tɩ- tɩ-wʊ̃́lɔ̄̃ tɩɩ-shɩ̄̃ / taashɩ̄̃? 21 ‘how much rubbish?’ Table 6: Concord in the interrogative quantifier

Moreover, interrogative pronouns are formed by free pronouns of the classes modified by the interrogative quantifier. The agreement relation is marked in the same way as between a nominal head and a modifying quantifier. Such interrogatives have two forms – a long form consisting of a prefix and a root and a short form in which the root of the free pronoun is elided and compensated for by lengthening the vowel of the prefix, as shown in Table 7. According to one of my language consultants, the full forms are used more often in formal speech or for emphatic reasons, whereas the reduced forms are used in informal speech. The tone on the prefix of the interrogative quantifier depends on the tone of the final syllable of the pronoun, just as with the nouns.

Class Concord ba(a)- ba-lɩ́ tɩ́á-shɩ̄̃? / baá-tɩ́á-shɩ̄̃? ‘how many (of them)?’ i- ɩ-lɩ́ tɩɩ́ ́-shɩ̄̃?/ ɩɩ́-tɩɩ́ ́-shɩ̄̃? ‘how many (of them)?’ 2 a - a-lɩ́ táá-shɩ̄̃? / aá-táá-shɩ̄̃? ‘how many (of them)?’ 1 bʊ - bʊ-lɩ́ tʊʊ́ ́-shɩ̄̃? / bʊʊ́-tʊʊ́ ́-shɩ̄̃? ‘how many (of them)?’ tɩ- kɩ-lɩ́ tɩɩ́ ́-shɩ̄̃? / kɩɩ́-tɩɩ́ ́-shɩ̄̃?/ ‘how much (of them)?’ a-lɩ́ táá-shɩ̄̃? / aá-táá-shɩ̄̃? Table 7: Interrogative pronouns (quantifiers)

4.6.2 The interrogative qualifier ɖúwē. Ðúwē ‘which’ is an interrogative qualifier and like the adjective and the ordinal, it does not show concord with the head noun, as is shown in (45) below:

21 Both tɩɩ-shɩ̄̃ / taashɩ̄̃? ‘how much’ are used for the tɩ- class. According to one of my consultants, taashɩ̄̃ is used more often. NOUN CLASSES IN TAFI 299

(45) a. a-nʊ́vɔ̄ ɖúwē? ‘which child?’ b. baa-nʊ́vɔ̄ ɖúwē? ‘which children?’ c. ɔ-ma ɖúwē? ‘which town?’ d. ɩ-ma ɖúwē? ‘which towns?’ e. ka-kudzɔgɛ̌ ɖúwē? ‘which dog?’ f. bu-kudzɔgɛ̌ ɖúwē? ‘which dogs?’

However, it can be used to form an interrogative pronoun by the addition of the nominal prefixes, except for the tɩ- class, which requires the prefixation of the subject pronoun. The vowels in all the prefixes are [+ATR] because those in the interrogative qualifier have the same [ATR] value.

Class Concord 1 a - e-ɖúwē? ‘which one?’ ba(a)- be-ɖúwē? ‘which ones?’ o- o-ɖúwē? ‘which one?’ i- i-ɖúwē? ‘which ones?’ ki- ki-ɖúwē? ‘which one?’ 2 a - e-ɖúwē? ‘which ones?’ ka- ke-ɖúwē? ‘which one?’ 1 bʊ - bu-ɖúwē? ‘which ones?’ 2 bʊ - bu-ɖúwē? ‘which one?’ tɩ- ki-ɖúwē? ‘which ones?’ Table 8: Interrogative pronouns (qualifiers)

4.7 The modifier ká-nyáá-ká Tafi has a negative modifier ká-nyáá-ká ‘none, nobody, nothing’ which occurs in the following frame:

(46) [Head nominal + ká-nyáá-ká] ká-nyáá-ká does not show concord with the noun, but it can be nominalised in the same way as the adjective and the ordinal. In (46a), we have the noun sɩkā ‘money’ occurring in the head slot, so the resultant structure becomes sɩkā ká-nyáá-ká ‘no money’. Sɩkā ‘money’ belongs to the 300 MERCEY LAMPTEY BOBUAFOR a1- class, so the a1- class prefix is added to ká-nyáá-ká to derive the pronominal form á-ká-nyáá-ká ‘none’. The third column of Table 9 illustrates the pronominal forms for all the classes with ká-nyáá-ka ‘none’. The derivation of the form ká-nyáá-ká ‘none’ is based on the complex NP structure [NPi nyáá NPi]. When this structure is used in affirmative sentences, it means ‘every, each, any NP’; when used in negative sentences, it takes on a negative meaning such as ‘none, no NP or not any NP’. Consider the following examples:

(47) a. e-yī-nyéé-é-yí á-klʊ́ā a-kā CM-child-every-CM-child SM-insult CM-father ‘Any child insults the father.’ b. e-yī-nyéé-é-yí á-tɩ́-klʊ́ā a-kā CM-child-every-CM-child SM-NEG-insult CM-father ‘No child insults the father.’

Agreement within this complex NP between nyáá and the nouns is indicated by a vocalic concord on nyáá, as shown in the second column of Table 9. If the vowel of the noun prefix is /a/, the vowels in nyáá stay the same. For this reason, all the vowels in the ká-nyáá-ká word are /a/. When the vowel in the noun class prefix is [+front], the nyáá word is realized with a front mid vowel depending on its [ATR] value. When the vowel in the noun class prefix is a high back vowel then the concord vowel is /ɔ/. Finally, when the noun class prefix vowel is mid back, the agreement marker is the corresponding [ATR] mid back vowel. A high tone on the final syllable of the first nominal root in the complex NP becomes a mid tone, while the high tone on the last vowel of nyáá spreads rightwards onto the following noun.

Class NPi nyáá NPi ‘every, each, any NP’ ká-nyáá-ká ‘none, nobody, or nothing’ 1 a - e-kusī-nyéé-é-kúsí ‘every, each chief’ a-ká-nyáá-ká ‘none, nobody, or nothing’ ba(a)- bee-kusī-nyéé-é-kúsí ‘any/ all chiefs’ ba-ká-nyáá-ká ‘none, nobody, or nothing’ o- o-nugbū-nyóó-ó-núgbú ‘every, each ɔ-ká-nyáá-ká ‘none’ mouth’ NOUN CLASSES IN TAFI 301

i- i-nugbū-nyɛɛ́ ́-ɛ́-núgbú ‘any/ all mouths’ ɩ-ká-nyáá-ká " ki- kɩ-tɔkpū-nyɛɛ́ ́-ɛ́-tɔ́kpú ‘every, each head’ kɩ-ká-nyáá-ká " 2 a - a-tɔkpū-nyáá-á-tɔ́kpú ‘any/ all heads’ a-ká-nyáá-ká " ka- ke-dē-nyéé-é-dé ‘every, each back’ ka-ká-nyáá-ká " 1 bʊ - bʊ́-gbɔ-̄ nyɔɔ́ ́-ʊ́-gbɔ́ ‘any/ all squirrels’ bʊ-ká-nyáá-ká " 2 bʊ - bʊ-tɔ-̄ nyɔɔ́ ́-ʊ́-tɔ́ ‘every, each ear’ bʊ-ká-nyáá-ká " bú-vū-nyɔɔ́ ́-ú-vú ‘every, each building’ bʊ-ká-nyáá-ká " tɩ- tɩ-wʊlɛ-̄ nyɛɛ́ ́-tɩ́-wʊ́lɛ́ ‘any/ all intestines’ kɩ-ká-nyáá-ká " Table 9: NPi nyáá NPi and ká-nyáá-ká

As shown by the examples in the second column, apart from the tɩ- prefix, the initial consonant of the CV prefixes of the second noun is not realised. The vowels of the kɩ-, and tɩ- prefixes of the second noun in the ki- and tɩ- classes are assimilated to the agreement vowel of nyáá. Simi- larly, the vowel prefix of the i- class is assimilated to the vowel of nyáá.

5. Pronouns Tafi has a set of pronouns that match the respective classes. The pronouns in the second column of Table 10 are substituted for nouns that have a subject function in the clause. Also, the subject pronoun for the tɩ- class is attached as a prefix to adjectives, ordinals, and the interrogative qualifier ɖúwē ‘which’, when they are substituted for nouns in this class. The pronouns in the third column of the table are used as free forms and as object pronouns. When these pronouns are used as free forms their full form is always used but when they are used as object pronouns the initial consonant of their CV prefixes is most often elided.

Class Subject Pronoun Free form 1 a - a/e- yɩ́ ba(a)- ba/be- (b)a-lɩ́ o- (l)ɔ/(l)o- ɔ-lɩ́ i- (l)ɩ/(l)i- ɩ-lɩ́ ki- kɩ/ki- (k)ɩ-lɩ́ 2 a - la/le- a-lɩ́ kaka/ke- (k)a-lɩ́ 302 MERCEY LAMPTEY BOBUAFOR

1 bʊ - bʊ/bu- (b)ʊ-lɩ́ 2 bʊ - bʊ/bu- (b)ʊ-lɩ́ tɩ- kɩ/ki- (k)ɩ-lɩ́22 Table 10: Pronouns

The (b) sentences in the examples below illustrate subject pronouns of the nouns in the (a) sentences. Most of these sentences are repetitions of earlier sentences. (6a) is repeated as (45a), (10a) and (10b) as (46a) and (46b), (13a) and (13b) as (47a) and (47b), (17a) and (17b) as (48a) and (48b), (20a) as (49a), (21a) as (49b), and (23a) and (23b) as (50a) and (50b).

(45) a. á-nyɩ́nʊ́vɔɛ̄ =̄ ń á-mɔ o-zī to-lí nɩ́ o-sí CM-boy=DEF SM-see CM-hole AM-INDEF LOC CM-tree gbɩgblǎ nɩ́-m̄ big DEF-inside ‘The boy saw a hole in the big tree’ (FS). b. á-mɔ o-zi to-lí nɩ́ o-sí gbɩgblǎ nɩ́-m̄ 3SG-see CM-hole AM-INDEF LOC CM-tree big DEF-inside ‘He saw a hole in the big tree.’

(46) a. ba-wi=ń á-sɔmɩ́ kí-húí nɩ́-ábhā CMPL-dress=DEF SM-hang CM-drying_line DEF-top ‘The clothes are hanging on the drying line.’ (TPRS. 37) b. bá-sɔmɩ́ kí-húí nɩ́-ábhā 3PL-hang CM-drying_line DEF-top ‘They are hanging on the drying line.’

(47) a. o-seyuyū=ń ɔ́-sɔmɩ́ o-sí=ń í-shú CM-fruit=DEF SM-hang CM-tree=DEF CM-body ‘The fruit is hanging on the tree.’ (TPRS. 27)

22 According to one of my consultants who is over 60 years, he and others of his age, heard their parents and grandparents using tɩ-lɩ́ as the object pronoun of the tɩ- class. NOUN CLASSES IN TAFI 303

b. (l)ɔ-sɔmɩ́ o-sí=ń í-shú 3SG-hang CM-tree=DEF CM-body ‘It is hanging on the tree.’

(48) a. i-seyuyū=ń ɩ́-sɔmɩ́ o-sí=ń í-shú CMPL-fruit=DEF SM-hang CM-tree=DEF CM-body ‘The fruits are hanging on the tree.’ (TPRS. 45) b. (l)ɩ-sɔmɩ́ o-sí=ń í-shú 3PL-hang CM-tree=DEF CM-body ‘They are hanging on the tree.’

(49) a. ki-wí=ń í-tsú a-dzɩ=̄ ń CM-thorn=DEF SM-prick CM-woman=DEF ‘The thorn pricked the woman.’ b. ki-tsú a-dzɩ=̄ ń 3SG-prick CM-woman=DEF ‘It pricked the woman.’

(50) a. ka-hlɔɛ=ń é-tsiri ’e-sukpogunū nɩ-abhā CM-antelope=DEF SM-be.on CM-tree_stump DEF-top ‘The antelope is on the tree-stump.’ b. ke-tsiri ’e-sukpogunū nɩ-abhā 3SG-be.on CM-tree_stump DEF-top ‘It is on the tree-stump.’

The object pronouns are exemplified in Table 11.

Class Á-mɔ….. hãa? Á-mɔ……. ‘Did he see …..? ‘He saw it/ them.’ 1 a - e-pidzyāń yɩ́ ‘the goat’ ba(a)- be-pidzyāń (b)a-lɩ́ ‘the goats’ o- o-síń ɔ-lɩ́ ‘the tree’ i- i-síń ɩ-lɩ́ ‘the trees’ ki- kɩ-shã́ń (k)ɩ-lɩ́ ‘the horn’ 2 a - a-shã́ń a-lɩ́ ‘the horns’ ka- ke-tséń (k)a-lɩ́ ‘the monkey’ 304 MERCEY LAMPTEY BOBUAFOR

1 bʊ - bu-tséń (b)ʊ-lɩ́ ‘the monkeys’ 2 bʊ - bú-vūń (b)ʊ-lɩ́ ‘the building’ tɩ- tɩ-wʊ̃́lɔ̄̃ń (k)ɩ-lɩ́ ‘the rubbish’ Table 11: Examples of object pronouns

6. Possessives Possessed pronouns in Tafi also establish agreement relations with the head noun. They are made up of a personal pronoun, a concord vowel and the 3SG pronoun yɩ́. The concord vowel added to the 3SG yɩ́ is determined by the class of the possessed noun. Except for the ka- class, as is shown in Table 12, the concord vowels correspond to the vowels of the class prefixes (Col 2, Table 2) and they agree in ATR value with the vowel of the 3SG pronoun yɩ́. The concord vowels for the ka- class are the same as those in the prefix of the numeral ‘one’ when it modifies nouns in this class. Furthermore, the ba(a)- and a2- classes have a plural form of the possessed pronoun. It is formed with a double marking on the 3SG form yɩ́. The double marking consists of the 3PL free pronoun form with its initial consonant elided yielding -alɩ́23 ‘they’. When the plural suffix -alɩ́ is attached, the vowel of the 3SG pronoun yɩ́ is deleted. The high tone of the vowel of yɩ́ ‘3SG’, however, remains and spreads right- ward to the initial vowel of the plural suffix. The 1SG and 3SG possessed pronoun forms are illustrated in Table 12.

Class 1SG 3SG 1 a - m-a-yɩ́ ‘mine’ y-a-yɩ́ ‘his’ ba(a)- m-aa-yɩ́/ m- ‘mine’ y-aa-yɩ́/ y-a-y-alɩ́ ‘his’ a-y-álɩ́ o- m-ɔ-yɩ́ ‘mine’ y-ɔ-yɩ́ ‘his’ i- m-ɩ-yɩ́ ‘mine’ y-ɩ-yɩ́ ‘his’

23 There is a variation among speakers in the use of the plural suffix -alɩ́. Some speakers apply it to human nouns and others apply it as well to nouns which are of interest to humans such as ba-fʊkpǎ ‘shoes’. Some speakers also tend to use this plural suffix for the a2- class as well probably because of its phonological shape. NOUN CLASSES IN TAFI 305 ki- m-ɩ-yɩ́ ‘mine’ y-ɩ-yɩ́ ‘his’ 2 a - m-a-yɩ́/ m-a- ‘mine’ y-a-yɩ́/ y-a-y-álɩ́ ‘his’ y-álɩ́ ka- m-ɩa-yɩ́ ‘mine’ y-ɩa-yɩ́ ‘his’ 1 bʊ - m-ʊ-yɩ́ ‘mine’ y-ʊ-yɩ́ ‘his’ 2 bʊ - m-ʊ-yɩ́ ‘mine’ y-ʊ-yɩ́ ‘his’ tɩ- m-ɩ-yɩ́ ‘mine’ y-ɩ-yɩ́ ‘his’ Table 12: Possessed pronouns

7. Conclusion The purpose of this paper has been to investigate the noun class system of the Tafi language. Tafi has an overt noun class system (e.g. Corbett 1991) which makes use of prefixes as noun class markers. I have classified nouns into ten individual classes on the basis of the prefixes they take, the agreement patterns within constructions such as the noun phrase, and the subject-verb clauses. I have also shown that the noun class system is a “crossed” one (Heine 1982: 197) where two or more singular classes can have one plural correspondent and where several plural classes can have one singular correspondent (see Figure 1). It has been demonstrated that there is agreement between the head noun and modifiers, the qualifiers and intensifiers excepted. The choice of agreement markers within the noun phrase is determined by the noun class to which the head noun belongs. Adjectives and ordinals may be nominalised, by the addition of a class prefix. The way borrowed nouns are integrated into the noun class system was also examined. It was shown that the a1- and ba(a)- classes serve as the default classes for loanwords. Other loanwords may be allocated to other classes based on phonological and/or semantic criteria.

Acknowledgements The research on Tafi is funded by a grant from the Netherlands Organisa- tion for Scientific Research (NWO). I am very grateful to my Tafi language consultants especially Mrs. Beatrice Adoboe, Nelson Agbley, Simon Amoadza, Patricia Amoadza and Diana Dzandu. I am also in- debted to Felix Ameka, Thomas Payne and an anonymous reviewer for 306 MERCEY LAMPTEY BOBUAFOR their valuable comments. I also thank Leo Wetzels for his editorial advice and patience. I take sole responsibility for any remaining errors.

Abbreviations AM = Agreement marker ATR = Advance Tongue Root C = Consonant CM = Noun class marker CMPL = Class marker plural CONJ = Conjunction DEF = Definiteness marker DEM = Demonstrative FS = Frog Story FUT = Future tense INDEF = Indefiniteness marker INT = Interrogative LOC = Locative L = Low tone LINK = Linker MRK = Marker NEG = Negative NP = Noun Phrase NUM = Numeral ORD = Ordinal 3PL = 3 Person Plural PRON = Pronoun PRSPROG = Present progressive PST = Past PSTPROG = Past progressive PFX = Prefix QTF = Quantifier RED = Reduplicative 1SG = 1 Person singular 2SG = 2 Person singular 3SG = 3 Person singular SM = Subject marker V = Vowel

References Ameka, Felix K., Carlien de Witte & David P. Wilkins (1999). Picture Series for Positional Verbs: Eliciting the verbal component in locative descriptions. In David P. Wilkins (ed.), A manual for the 1999 field season, Version 1.0, 48-56. Nijmegen: Max Planck Institute for Psycholinguistics. Blench, Roger (2001). What have we learnt since Heine? Paper presented at the 32nd Annual Conference on African Linguistics and subsequently revised. Berkerley, 23rd-25th March 2001. Bowerman, Melissa & Eric Pederson (1993). Topological Relations Picture Series. In Eve Danziger & Deboreh Hill (eds.), Manual for Space Stimuli kit 1.2, 40-50. Nijmegen: Max Planck Institute for Psycholinguistics. Corbett, Greville (1991). Gender. Cambridge: Cambridge University Press. Dakubu, M.E. Kropp & Kevin C. Ford (1988). The Central Togo Languages. In M.E. Kropp Dakubu (ed.), The Languages of Ghana, 119-153. London: Kegan Paul International. Dorvlo, Kofi (2008). A Grammar of Logba (Ikpana). PhD Dissertation, University of Leiden. (LOT Dissertation Series 183). NOUN CLASSES IN TAFI 307

Dorvlo, Kofi (this volume). Noun class system and agreement patterns in Logba. Essegbey, James (to appear). Noun Classes in Nyagbo: A preliminary analysis. Journal of West African Languages. Ford, Kevin C. (1973). On the loss of cross-height vowel harmony. Research Review Supplement 4, 49-80. Funke, Emil (1910). Die Nyangbo-Táfi Sprache: Ein Betrag zur Kenntnis der Sprachen Togos. Mitteilungen des Seminars für orientalische Sprachen, 13(3):166-201. Gordon, Raymond G. Jr. (2005). Ethnologue: Languages of the World, 15th edition. Dallas: Texas: SIL International. Greenberg, Joseph H. (1966). The Languages of Africa. The Hague: Mouton. Heine, Bernd (1982). African noun class systems. In Hansjakob Seiler & Christian Leh- mann (eds.), Apprehension: das sprachliche Erfassen von Gegenstanden, v. 1, 189-216 Tuebingen: Gunter Narr Verlag Heine, Bernd (1968). A preliminary survey of the noun classes of Basila. Journal of African Languages 7(1), 1-13. Heine, Bernd (1968). The allocation of loan-words within the nominal class systems of some Togo Remnant Languages. Journal of African Languages, Volume 7(2): 130-139. Mayer, Mercer (1969). Frog, where are you? New York: Dial Books. Ring, James Andrew (1995). Lelemi tone. Papers from GILLBT’S seminar week, Jan- uary 30-February 3 1995, Tamale, 16-26. Tamale: GILLBT Press. Ring, James Andrew (2000). Languages as resource: the least studied languages of Ghana. Paper presented at Linguistics Association of Ghana meeting, University of Ghana, Legon. Schuh, Russell G. (1995). Avatime noun classes and concord. Studies in African Lin- guistics 24.2, 123-148 Stewart, John M. (1989). Kwa. In J. Bendor Samuel (ed.), The Niger-Congo languages, 217-245 Lanham: University Press of America. Struck, Bernhard (1912). Einige Sudan-Wortstamme Zeitschrift fur Kolonialsprachen 2: 233-253, 309-323 Welmers, William E. (1973). African Language Structures. Berkeley: University of California Press. Westermann, Diedrich & Margaret A. Bryan (1952). Languages of West Africa. Handbook of African Languages 2. Oxford: Oxford University Press for Inter- national African Institute. Williamson, Kay & Roger M. Blench (2000). Niger-Congo. In Bernd Heine & Derek Nurse (eds.), African Languages: An Introduction, 11-41. Cambridge: Cam- bridge University Press.

Part III: Middle East and East-Timor

Nominalization in Soqotri, a South Arabian language of Yemen

Masayoshi Shibatani (Rice University) and Khaled Awadh Bin Makhashen (Hadhramout University and Universiti Sains, Malaysia)

1. Introduction In his recent article Deutscher (2009: 199) suggests that “nominalization is an unsung hero in the story of subordination”. He goes on to say that “[t]he ability to derive a noun from a verb, that is, to reify a verbal predicate and present it as a nominal argument or modifier, is the core of subordination”. While Deutscher (2009) believes that the relevant nominalization for his study is “derivational nominalization”, which derives nouns from verbs, we find his characterization of nominalization as an unsung hero to be quite fitting not only in the study of subordination but also in other related phenomena, which we wish to characterize as the case of “grammatical nominalization”. Unlike derivational or lexical nominalization of the employ-er/employ-ee type, which supplies the lexicon with new lexical items belonging to the noun class, grammatical nominalization yields nominal expressions but which are not nouns. We shall explicate the difference between lexical nominalization and grammatical nominalization below, from which various properties of grammatical nominalizations follow. The issues addressed in this paper are those nominal expressions that have been discussed under the heading of “noun phrases without nouns” by Matthew Dryer in his recent review of noun phrase structure (Dryer 2007), where he surveys noun phrases without a nominal head such as adjectives functioning as noun phrases by themselves (see (1) below), possessor phrases without a noun head (see (2) below), and so- called headless relative clauses (see (3a) below):

(1) Nkore-Kiga (Dryer 2007:194 quoting Taylor 1985) [omuto] aka-gamba na-anye young 3SG-REM.PAST-speak with-me ‘the young one spoke with me’ 312 MASAYOSHI SHIBATANI & KHALED AWADH

(2) Your car is nice, but John’s is nicer.

(3) Miya (Dryer 2007:197 quoting Schuh 1998) a. má ráɓaza (headless RC) REL.FEM.SG wet ‘the one (feminine, singular) that is wet’ b. kàba [má ráɓaza] (headed RC) gown REL.FEM.SG. wet ‘the gown that is wet’

While Dryer (2007) does not invoke the notion of grammatical nominalization for any of these constructions above, some recognize the involvement of nominalization in some of them. For example, Comrie & Thompson (2007: 378, 379) recognize what they regard as “a somewhat more rare function of nominalization: as a relative clause modifying a head noun…” and make the puzzling “claim that in certain languages relativization is indistinct from nominalization”, as if relative clauses and nominalizations were two separate constructions that happen to be structurally indistinct in some languages. A less misleading way to characterize the situation that Comrie & Thompson (2007) observe is to make a clearer distinction between form and function, and to consider that relativization and nominalization are two different functions – modification for the former and referential for the latter –, and then to recognize that one and the same form plays both functions in some languages. Besides Dryer a countless number of grammarians characterize forms such as Miya example (3a) above as headless relative clauses as if they were derivatives of (headed) relative clause constructions. Indeed, the most popular analysis of such forms is a deletion analysis, where the relevant forms (e.g. 3a) are derived from headed relative clauses (e.g. 3b) via deletion of the head nominal. We consider this kind of analysis mis- guided and instead regard forms such as (3a) as simply nominalizations. We consider relative clauses of the type seen in (3b) above to be not independent constructions which happen to be structurally indistinct from nominalizations but that they are nominalizations themselves that play a modification function. This position has recently been expounded in Shi- batani (2009), where it is shown that a nominalization functioning as a relative clause is not at all rare, contrary to the suggestion made by Com- rie & Thompson (2007; see above), and that it is widely seen around the NOMINALIZATION IN SOQOTRI 313 globe, and even in familiar European languages. Our attempt here is to show that nominalizations are involved not only in relative clauses (of both headed and headless types) but also in other types of noun phrases surveyed by Dryer (2007). The relevant data derive from the endangered language Soqotri of Yemen, which displays a wide array of nominalization phenomena, ranging from lexical to grammatical nominalization, and from argument nominalization to different types of event nominalization. Our primary focus in this paper is the phenomenon of grammatical argument nominalization.

2. Soqotri The Soqotri language is one of six pre-Islamic languages that form a group called Modern South Arabian languages. This group, which includes Soqotri, Mehri, Jibbali, Bathari, Harsusi and Hobyot, is spoken in parts of Oman particularly in Dhofar and south east of Yemen in Mahrah Governorate and Soqotra Island. Modern South Arabian languages belong to the southern branch of the western Semitic languages. The other Semitic languages grouped in this branch are the Semitic languages of Ethiopia and the extinct inscription languages, which are also called Epi- graphic Old South Arabian languages. Soqotri is the native language of the Island of Soqotra located in the east of Aden Gulf about 300km south of the Arabian Peninsula. It is also spoken in two small nearby islands called the Island of Abdul kuri and the Island of Samha. The number of the Soqotri speakers is estimated to be around 50,000. The island inhabi- tants depend on fishing, growing palm dates trees or rearing camels, cows and goats. The Soqotri language and culture is under a great influence of the dominant Arabic language and culture, and it is regarded as an endangered language (Naumkin 1998, Simeone-Senelle 2003). Many Arabic-speaking Yemenis have settled in the Soqotri territory perma- nently. Arabic has become the official language in the island, and it is used in the Soqotri schools as medium of instruction. The Soqotri stu- dents are prohibited from using their mother tongue while they are at school. Any Soqotran seeking employment must master Arabic before landing a reasonable job. The young Soqotrans, fluent in Arabic, prefer it to their mother tongue, which they learn imperfectly, mixing many Ara- bic words in it. They cannot recite or understand any piece of Soqotri oral literature. 314 MASAYOSHI SHIBATANI & KHALED AWADH

Soqotri is primarily a head-initial language with VS/SV and VSO/ SVO word orders seen equally frequently. It has prepositions, and adjectives and relative clauses as well as genitives (typically with a full possessor noun) follow the head noun they modify, while demonstratives, numbers, and genitives (typically with a pronominal possessor) precede the head noun. The modifiers agree in number and gender with the head noun, as shown below:

(4) -    - - DEM-DU.F woman-DU.F tall-DU.F here REL.DU.F come-3DU.F.PER    now my aunt-DU.F ‘These two tall women who came just now are my aunts.’

3. Relative clauses and other types of noun modification Of the several types of noun modification in Soqotri, the patterns we are interested in here are relativization, adjectival modification and genitive constructions, all of which share what we claim to be the basic nominalization mechanism of the language. As seen in (4) above, Soqotri relative clauses are marked by what is identified as the “relative particle” by Simeone-Senelle (1997). It inflects for number and gender in accordance with these categories of the head noun.

(5) a.   -   man REL-SG.M 3SG.M.IMP-love children come-3SG.M.PER ‘A man who loves children came.’ b.      man-DU.M REL DU.M 3DU.M.IMP-love children come-3DU.M.PER ‘Two men who love children came.’ c.   -   men REL-PL 3PL.M.IMP-love children come -3PL.M.PER ‘Men who love children came.’

(6) a. -     saw-1SG.M.PER woman REL SG.F 3SG.F.IMP-love children ‘I saw a woman who loves children.’ NOMINALIZATION IN SOQOTRI 315

b. -     saw-1SG.M.PER woman-DU REL.DU.F 3DU.F.IMP-lovechildren ‘I saw two women who love children.’ c. -     saw-1SG.M.PER women REL.PL 3PL.F.IMP-love children ‘I saw women who love children.’

Keenan (1985:149) notes that relative pronouns are “typically the same as, or morphologically related to, the demonstrative pronouns”. This is true for the Soqotri “relative particles”, which are identical with the demonstrative pronouns. The Soqotri demonstratives combine demonstrative pronouns and locative adverbials  (SG/DUAL) and  (PL) ‘here’ for proximal and   (SG/DUAL) and  (PL) ‘there’. Demon- strative pronouns may also be followed by these locative adverbials.

Number Masculine Feminine Singular … ‘this’ … ‘this’ Dual … ‘these’ … ‘these two’ Plural … ‘these’ … ‘these’ Table 1: Proximal demonstratives

Number Masculine Feminine Singular ...   ‘that’ ...’ ‘that’ Dual …’ ‘those’ …’ ‘those two’ Plural … ‘those’ … ‘those’ Table 2: Distal demonstratives

(7) a.  ()    DEM here COP my child ‘This (here) is my child.’ b.  -  ’ want-3SG.IMP 1SG.SUB-buy DEM.PL there.PL ‘I want to buy those (FEM/MAS).’ c.       DEM boy here COP my child ‘This boy is my child.’ 316 MASAYOSHI SHIBATANI & KHALED AWADH

d.  -   ’ want-3SG.IMP 1SG.SUB-buy DEM.PL cow-PL.F there.PL ‘I want to buy those cows’

What can be characterized as headless relative clauses in Soqotri show formal resemblance to headed relative clauses as in other languages pos- sessing these two types of constructions, as comparison of the following with the corresponding relative clauses above indicates.

(8) a.  -   REL.SG.M 3SG.M.IMP-love children come-3SG.M.PER ‘The one who loves children came.’ (cf. 5a) b.     REL.DU.M 3DU.M.IMP-love children come-3DU.M.PER ‘The ones who love children came.’ (cf. 5b) c.  -   REL.PL 3PL.M.IMP-love children come-3PL.M.PER ‘The ones who love children came.’ (cf. 5c)

(9) a. -     saw-1SG.M.PER REL.SG.F 3SG.F.IMP-love children ‘I saw the one who loves children.’(cf. 6a) b. -     saw-1SG.M.PER REL.DU.F 3DU.F.IMP-love children ‘I saw the ones who love children.’ (cf. 6b) c. -     saw-1SG.M.PER REL.PL.F 3PL.F.IMP-love children ‘I saw the ones who love children.’ (cf. 6c)

The reason we enclosed the term “relative particles” in the quotation marks above is that the forms dε, dε, etc. that introduce relative clauses are used in other constructions besides relative clauses and the so-called headless relative clauses seen above. They are used in adjectival attributive constructions and genitive, or more broadly noun modifier, constructions. Like English, adjectives in Soqotri can modify nouns directly or in a form similar to relative clauses, as below: NOMINALIZATION IN SOQOTRI 317

(10) a.   woman beautiful ‘a beautiful woman’ or ‘The woman is beautiful.’ b.    woman REL.SG.F beautiful ‘a beautiful woman’ or ‘the woman who is beautiful’

Apparently (10b) above allows both attributive and restrictive relative readings. Adjectives differ from verbs in that while the former can modify a noun directly, as in (10a), the latter cannot. For example, (11a) below cannot mean ‘a running woman’.

(11) a.   b.    woman 3SG.F.IMP-run woman REL.SG.F 3SG.F.IMP- run ‘The woman runs.’ ‘the woman who runs’

Genitive constructions in Soqotri come in three types. In expressing kinship relations, the kin terms are simply marked by bound object pronouns (12). The other two involve the marking relevant to our main discussion. One of them involves the “possessive particle”  and free pronouns or nouns, as in (13), and the other the prefix m- and bound object pronouns, as in (14).

(12) a.  - b.  - brother-1SG.M.OBJ.SUF brother-SG.M.OBJ.SUF ‘my brother ‘your brother’ c.  -  brother-3.SG.F.OBJ.SUF ‘her brother’

(13) a.  f b. f   c.    my face face POSS. man daughter POSS Ali ‘my face’ ‘the man’s face’ ‘Ali’s daughter’

(14) a.  f b.  f c.  f my face your face his face ‘my face’ ‘your face’ ‘his face’ 318 MASAYOSHI SHIBATANI & KHALED AWADH

The genitive constructions involving the “possessive particle” seen in (13) are in fact general noun modifier constructions used in the modification of a noun by another noun, as in the following forms:

(15) a.    b.    teacher POSS school book POSS mathematics ‘school teacher’ ‘mathematics book’ c.    d.    meat POSS sheep rain POSS night ‘sheep meat’ ‘night rain’

While the “possessive particle” involved here is obviously related to the “relative particles” seen earlier, it behaves somewhat differently. Namely, the “possessive particle”, unlike the “relative particle”, does not inflect according to the head noun. The form  is the masculine singular form, which is also used as a default form when the gender and number of the referent is indeterminate (see (23) below). The reason why the “possessive particle” does not inflect is not obvious at this point. Besides the formal resemblance between the “relative” and the “possessive” particle, there is some indication that the relevant constructions involve the same mechanism. Namely, the m- possessive prefix seen in (14) appears to be the same prefix used in lexical argument nominalization that yields forms like the following (see below for more examples):

(16) a.   (N) ‘fisherman’ b. ’ (N) ‘prayer/person who prays’ c.  (N) ‘sharpener’ d. ’ (N) ‘oar’

This sharing of the marker m- by both argument nominalization and genitive constructions is indicative of the fact that nominalization is involved in the relevant genitive construction. Along the same vein, we can safely hypothesize that both relative clause constructions and the other modification patterns observed above involve nominalization resulting in -marking seen in all these noun modification constructions. This hypothesis above is bolstered by the similarity between relative clauses and the adjectival and noun modifier (including the genitive) constructions that goes beyond the sharing of the -marking. Just as NOMINALIZATION IN SOQOTRI 319 relative clauses have “headless relative clause” counterparts, both adjectival and noun modifier constructions allow “headless” expressions, as seen below:

(17) “Headless adjectival modifiers” a.     man REL.SG.M black.3SG.M came.3SG.M.PER ‘A black man came.’ a.    REL.SG.M black.3SG.M came.3SG.M.PER ‘A black (one) came.’ b.   () -- marry.3SG.M.PER woman REL.SG.F-beautiful-3SG.F ‘He married a beautiful girl.’ b.  () -- marry.3SG.M.PER REL.SG.F-beautiful -3SG.F ‘He married a beautiful one.’

(18) “Headless noun modifiers” a.  ’    ’ DEM.SG.F house here COP my house ‘This house is my house.’ a.  ’   DEM.SG.F house COP mine ‘This house is mine.’ b.       ’    DEM.SG.F COP my daughter and DEM there COP his daughter ‘This is my daughter and that is his daughter.’ b.       ’   DEM.SG.F COP my daughter and DEM there COP his ‘This is my daughter and that is his.’

All this consideration suggests reconsidering the treatment of -marking above, where it has been sometimes glossed as REL as if it were a relativ- izer, following Simeone-Senelle’s (2003) labeling of it as the “relative particle” or as POSS, following her labeling of it as the “possessive particle”, as if it were a form separate from what has been glossed as REL.A more serious problem arises in identifying the function of -marking in 320 MASAYOSHI SHIBATANI & KHALED AWADH

(8-9) and (17-18), where it occurs in constructions without an obvious head.

4. Are these all headless constructions? It is clear why constructions like (8) and (9) are given an oxymoronic appellation of “headless relative clauses” or “free relatives” by those who have been concerned with the typology of relative clauses such as Keenan (1985), Dryer (2005, 2007), Andrews (2007) and Huang (2008); across languages these constructions are (nearly) identical in form with the ones used to modify a nominal head, i.e., the regular relative clauses (e.g., (5) and (6) above). The basic question never raised by those treating constructions like (8) and (9) as headless relative clauses is whether the relevant constructions fit the functional definitions given to relative clause constructions. Consider the following definitions offered for relative clause constructions:

(19) “[a relative clause construction] specifies a set of objects (perhaps a one-member set) in two steps: a larger set is specified…and then restricted to some subset of which a certain sentence, the restrict- ing sentence, is true.” (Keenan & Comrie 1977: 63-64)

“a construction consisting of a nominal…and a subordinate clause interpreted attributively modifying the nominal. The nominal is called the head and the subordinate clause the RC. The attributive relation between head and RC is such that the head is involved in what is stated in the clause.” (Lehmann 1986: 664)

It is clear that the constructions in (8) and (9) and similar ones in other languages do not really function to specify “a subset of objects in two steps” as stated in the definition by Keenan and Comrie or to modify a nominal element, as in Lehmann’s definition. Indeed, we claim that the constructions in (8) and (9) are not modifiers at all. Rather they are nominal expressions in their own right, i.e., nominalizations, that refer to an entity characterized in terms of the event in which it has crucial relevance, e.g., being involved as a subject or an object (see next section on the nature of the argument nominalizations alluded to here). NOMINALIZATION IN SOQOTRI 321

The traditional way of fitting so-called headless relative clauses to the mold of the standard relative clause construction is in terms of a deletion analysis, in which these forms are derived from headed relative clauses by deleting the head nominals (see Adams (1972) on Ancient Greek, Weber (1983) on Quechua, Sneddon (1996) on Indonesian1, Huang (2008) on Qiang, Wrona (2008) on Old Japanese, etc., etc.). We consider this approach to so-called headless relative clauses to be wrong- headed and totally unmotivated in some cases. Treating the so-called headless relative clauses as a species of relative clauses is obviously due to the wrong perspective that the linguists such as those named above have had about these constructions, namely viewing them from the perspective of relative clauses, which contain them. For more than thirty years, relative clauses have been a center piece in both Generative Grammar, where they motivated syntactic rules of movement/extraction and deletion, and typological studies, where cross-linguistic patterns of relativization have been pursued from several different angles, including the most influential seminal work in this area by Keenan & Comrie (1977). We consider this situation similar to the effort of a Japanese person trying to characterize tuna fish in terms of sashimi, as if tuna fish were a derivative of sashimi rather than the other way around. Our position, as suggested above, is that the so-called headless relative clauses such as those seen in (8) and (9) are not relative clauses at all. They are nominalizations that are independent of relativization, but which can be used in the relativization context as a modifier of a nominal head. In many other languages, similar nominalizations are used in other contexts, such as noun and verb complements and other types of subordinate constructions, besides the relativization context, just as tuna is used to make other dishes besides sashimi. We also advance the same nominalization analysis for the other “headless constructions” examined in the preceding section. At the empirical front, there are several nominalizations that cannot be derived from headed relative clause constructions via deletion of the head. For example, several European languages have nominalizations that do not occur in the relativization context (see Shibatani (2009 for the

1 Sneddon (1996) at least recognizes what is derived from deleting the head nominal as a nominalization. 322 MASAYOSHI SHIBATANI & KHALED AWADH relationship between nominalization and relativization in European languages and elsewhere). For example:

(20) Russian a. (tot,) [kto vymyl ruki], mozhet nachatj jestj (that) who washed hands can start eat ‘The one who has washed his hands can start eating.’ b. *Maljchik, [kto vymyl ruki], mozhet nachatj jestj boy who washed hands can start eating ‘The boy who has washed his hands can start eating.’

(21) German a. Ich empfange, [wer (auch) morgen kommt] I receive who (also) tomorrow comes ‘I receive who(ever) comes tomorrow.’ b. *Ich empfange den Mann, [wer morgen kommt] I receive ART man who tomorrow comes ‘I receive the man who comes tomorrow.’

(22) Spanish a. Veré a [quien viene mañana]. I.will.meet to who comes tomorrow ‘I will meet the one who comes tomorrow.’ b. *Veré al hombre [quien viene mañana]. I.will.meet to.the man who comes tomorrow ‘I will meet the man who comes tomorrow.’

There are also cases of nominalization where there is no motivation for positing a specific lexical head since the relevant expressions refer to a non-specific or generic entity, as in the following Soqotri examples:

(23) a.  [- ’’ ] - NMZ.SG.M 3SG.M.IMP-speak Soqotri well 3SG.M.IMP-win ‘The one who speaks Soqotri well will win (the contest).’ b. -  [ ’] 3SG.M.IMP-see NMZ.SG.M 3SG.M.IMP-come tomorrow ‘I will see who(ever) comes tomorrow.’ NOMINALIZATION IN SOQOTRI 323

c.   -  [  ] person not 3 SG.M.IMP-eat NMZ.SG.M fall-3SG.M.PER on earth ‘One shouldn’t eat what has fallen on the floor.’

A possible response to these criticisms against the deletion analysis is to posit an empty pronominal head such as the arbitrary PRO, which has no phonetic substance but which refers to an arbitrary entity as in (23), or the anaphoric pro referring to some antecedent for the form seen in (24) below, for example. Both deletion and empty pronoun analysis appear motivated because the kind of constructions under consideration typically occur in the context where there in an “antecedent”, as in the following example, where the “antecedent” is understood, or where they refer to a generic or an arbitrary entity as in (23) above.

(24)  -’     ’ Ali 3SG.M.IMP-read book NMZ.SG.M his father 3SG.M.PER-give     ’  [ him (Ali) it (book) and I read NMZ.SG.M my  ’  ] teacher 3SG.M.PER-give me it ‘Ali is reading the book which his father gave him, and I am reading the one that my teacher gave me.’

Similar deletion and empty pronoun analyses can be advanced for those “headless” constructions of other noun modification constructions such as adjectival and noun modifier constructions seen above. For example, Matisoff (1973), while arguing against the deletion analysis of the “headless relative clauses”, subscribes to the deletion analysis of the “headless” genitive constructions of the Tibeto-Burman language Lahu. Ad- vancing a deletion or an empty pronominal analysis for these “headless” constructions because they typically occur in the environments noted above is like calling whales fish because they live in the ocean. The crux of the problem is arriving at a proper understanding of nominalizations that would explain why they typically occur in those contexts, just as a proper understanding of whales would explain why they live in the ocean despite the fact they are not fish. Indeed, the past works (e.g., Noonan (1997) and Wrona (2008)) on nominalizations fail to explicate what nominalization really is, concentrating more on their distribution pattern and historical development. Our position is that once we come to a 324 MASAYOSHI SHIBATANI & KHALED AWADH proper understanding of the relevant nominalizations, their distributional and historical development patterns also yield to a natural understanding. The nominalization analysis of the relevant “headless” constructions advanced here regards these constructions to have no other head, which is deleted or which takes empty pronominal forms. Instead, the nominalizations themselves are heads in these constructions.

5. What is nominalization? The essence of nominalization is creation of a referring expression, hence its sharing this essential nominal function with nouns, which refers to a state of affairs characterized by an event denoted by the clause (event nominalization), to an entity characterized in terms of the denoted event in which it has crucial relevance (argument nominalization), or to an entity having crucial relevance to the referent of a noun (“genitive” nominalization). There are important distinctions between what can be termed “lexical nominalization” and “grammatical nominalization”. While both lexical and grammatical nominalization processes yield referring expressions and the both types of nominalization fill the syntactic function as arguments or predicate nominals just like simple nouns, the former creates nouns (e.g., employ-er/employ-ee) that belong to the noun class of the lexicon, and the latter creates nominal expressions that have no lexical status (e.g., the equivalents to the English expressions (the one) who employs/(the one) whom someone employs). Secondly, lexical nominalizations, like other simple nouns, identify or put a label on the referents and classify them as belonging to a particular class of entities denoted by the label, grammatical nominalizations lack identification or names. Thus, the English expression what I bought yesterday or its equivalents in other languages, for example, characterizes an entity referred to in terms of an event of my buying it yesterday, but its identity is not specified – it could be a book, a newspaper, a hamburger, an um- brella, or any other things that could be bought. The properties and their distribution of grammatical argument nominalizations follow from this process of creating a variety of new referring expressions pertaining to a limitless range of everyday events that vary in their content. First, argument nominalization is associated with a presupposition that an event characterizing the entity referred to has taken place. The expression what I bought yesterday presupposes that I NOMINALIZATION IN SOQOTRI 325 bought something yesterday. This property of argument nominalization carries over to its relativization function such that relative clauses represent presuppositions, as the book which I bought yesterday presupposes that I bought a book yesterday. Thus, neither in the nominalized form nor in its relative clause function, is there assertion of the state of affairs denoted in the presupposition even if the form may contain a finite verb form as in the English examples here. This is an important distinction between sentences and nominalized forms, and which allows the latter to be embedded or subordinated into a main assertive clause. Since argument nominalizations are typically created for the nonce, they are often indefinite, and this accounts for the use of indefinite interrogative pronouns such as what and who as a place holder for the gap created by the nominalization process, as in a number of European languages. Since nominalizations are nominal entities, they often call for an article or other types of determiners just like regular nouns, as also in many European languages.2 This is what we see in Soqotri, where nominalizations are marked by demonstratives. That is, what has been identified as “relative” and “possessive” particles by Simeone-Senelle (1997) arose from demonstratives marking nominalizations as nominal entities. While Soqotri demonstratives and definite articles used in other languages normally mark definite nominals, their use in marking nominalizations has resulted in the grammaticalization of them as nominalization markers such that they no longer mark only definite nominals. This is what we see in forms such as (23) above. Further indication of this is that even in relative clause constructions, some speakers do not inflect the nominalization marker and simply use the default (masculine) form . In other words, the best analysis of the so-called “relative” and “possessive” particles in Soqotri is to treat them as nominalization markers, similar to the nominalization markers no in Japanese and de in Chinese, which also head a relative clause (Chinese) and mark nominal modifiers in genitive and other noun modifier constructions (Japanese and Chi- nese), and gloss it as NMZ, as we did in (23), in all its occurrence in the relevant constructions. Because the identity of what are referred to by argument nominalizations is unspecified, the distribution of argument nominalizations is constrained in such a way that they typically occur in those contexts

2 See (20-22) above and Shibatani (2009) for further examples. 326 MASAYOSHI SHIBATANI & KHALED AWADH where the entity identification is provided or sought – in (a) relative clause constructions, where the head supplies the identification, (b) wh- questions that seek the identity of the entity referred to in the equation format (see (25a) below), (c) cleft-type constructions, in which the entity identification is made in the equation format (see (25b) below), (d) where the entity identification can be made from the context (see example (24) above), or (e) the entity referred to is generic or arbitrary (see the examples in (23) above).

(25) a.  [  ] who NMZ.SG.M come-3SG.M.PER yesterday ‘Who is the one who came yesterday?’ b.  [ ]   NMZ.SG.M. come-3SG.M.PER yesterday COP Ali ‘The one who came yesterday is Ali.’

In formal analysis, these constructions are often treated as involving extraction of an element from the clause – a wh-word in (25a) and the identifier noun phrase in (25b). There is really no basis for such analysis. These constructions are basically equational constructions of the type [A is B] involving two nominals, just like Who is this? and This is Ali. The only special feature with the expressions in (25) is that they carry the presupposition associated with the nominalization involved, namely that someone came yesterday, contrasting with the following where there is no such presupposition.

(26) a.    who come-3SG.M.PER yesterday ‘Who came yesterday?’ b.   Ali come-3SG.M.PER yesterday ‘Ali came yesterday.’

6. From apposition to modification Of the nominalizations discussed in this paper, perhaps the least obvious is the “genitive” nominalization, which creates a nominal out of a (pro)noun. But the wide use of a nominalization marker as a genitive NOMINALIZATION IN SOQOTRI 327 marker, as in Chinese and Japanese, has been noted by Matisoff (1972). Essentially, what the “genitive” nominalization does is to create another noun referring to an entity “pertaining to” to what is referred to by a (pro)noun. Thus, Soqotri expression dε  means “an entity pertaining to Ali”, where the precise interpretation of “pertaining to” is determined by context. The expression such as    is likely to have arisen from an appositive construction of the form [] + [ ] with the meaning like ‘the book, the thing pertaining to Ali’. When these two nominals are unified into a noun phrase unit as [[][ ]]NP, the non-head element takes on a modification function resulting in the meaning such as ‘the book pertaining to Ali’. Still the meaning of ‘pertaining to’ is indeterminate and the whole expression can mean the book possessed by Ali, the book written by Ali, the book about Ali, etc. The same applies to the modification constructions involving regular nouns such as [  ] ‘mathematics book’ and [  ] ‘night rain’. The development of modification function in a nominal element is seen in other [nominal + nominal] constructions as in noun compounds such as [[song][bird]] and [[night][train]] and phrasal forms such as [[bird] [singing in the bush]] and [[singing][bird]]. The development of the modification function here is perhaps due to the constraint that a single noun phrase refers to only a single entity. We would analyze Soqotri relative clauses and adjectival modification constructions the same way, namely modification by the nominalizations below is the function of the NP constructions into which they enter:

(27) a. [ [ ]]NP NMZ.SG.F 3SG.F-love children ‘(feminine) one who loves children’ b. [ [ [ ] ]]NP woman NMZ.SG.F 3SG.F-love children ‘the woman who loves children’

(28) a. [ []]NP NMZ.3SG.M black.3SG.M ‘one that is black’ b. [ [ []]]NP man NMZ.SG.M black-SG.M ‘the black man’ 328 MASAYOSHI SHIBATANI & KHALED AWADH

7. Lexical nominalization and grammatical nominalization As mentioned above, Soqotri shows some measure of productivity in lexical argument nominalization, which derives nouns from perfective verb forms referring to different types of semantic roles.

(29) Lexical nominalization a. Agent/subject nominalization []V → []N ‘fished’ ‘fisherman’ [’ ]V → [’]N ‘prayed’ ‘prayer/one who prays’ [’ ]V → [’]N ‘learned’ ‘learner’ [’]V → []N ‘went mad’ ‘madman’

b. Patient/object lexical nominalization []V → [ ]N ‘killed’ ‘one who was killed’ [ ]V → [ ]N ‘jailed’ ‘one who was jailed/jail bird’ [] → []N ‘beat’ ‘one who was beaten’

c. Instrumental nominalization [ ]V → [ ]N ‘sharpened’ ‘sharpener’ [’ ]V → [’]N ‘rowed’ ‘oar’ []V → [] N ‘took water ‘a big glass’ ‘with a glass’

While lexical nominalization of the above type involves change in vocalic melody morphemes, grammatical nominalization does not – it simply marks what is nominalized by the nominalization marker dε. Gram- matical nominalization also yields argument nominalization as lexical NOMINALIZATION IN SOQOTRI 329 nominalizations. Formal treatment of this process has not been studied, but we may characterize the process in the following way:

(30) Grammatical nominalization a. Agent/subject nominalization [  ] →  [Øi  ]i man buy.3SG.M.PER goat NMZ buy.3SG.M.PER goat ‘The man bought a goat.’ ‘the one who bought a goat’ [ ] →  [Øi  ]i man black-SG.M NMZ black.SG.M ‘The man is black.’ ‘a black one’

b. Patient/object nominalization [  ] →  [  Øi]i man buy.3SG.M.PER goat NMZ man buy.3SG.M.PER ‘The man bought a goat.’ ‘the one that the man bought’

c. “Genitive” nominalization [] →  []i [] →  []i she NMZ she night NMZ night ‘she’ ‘hers’ ‘night’ ‘of the night’

The essential point is that grammatical nominalization produces a nominal, which is coindexed with the argument nominalized (argument nominalization) or which simply takes on a new referential status (“genitive” nominalization), and which, together with the nominalization marker , forms a noun phrase or D(eterminer) P(hrase), referring to an entity that is characterized by the role of the derived nominal as an agentive or patientive event participant, or an entity pertaining to what is denoted by the noun being nominalized. In other words, these expressions do not involve a deleted lexical head or an empty pronominal head as in the widely practiced conventional analysis (see above). As in other languages, some grammatical nominalizations have been lexicalized. For example,   can mean ‘a killer’ in addition to the grammatical nominalization reading of ‘a one who killed’. The unmarked patient/object nominalization forms in (29b) can be preceded by the nominalization marker  and become indistinguishable from grammatical nominalizations of the same form. It thus appears that the distinc- 330 MASAYOSHI SHIBATANI & KHALED AWADH tion between lexical and grammatical nominalization is blurred in certain expressions, indicating that a fair number of  grammatical nominalizations have become lexicalized. If this conjecture on the current state of the  nominalization in Soqotri is correct, we can further surmise that lexical agentive/instrumental nominalization marked by m- or the unmarked patientive nominalizations may have been once productive grammatical nominalizations, as the parallel involvement of m- in agentive/instrumental nominalization and genitive nominalization indicates (see the earlier remark on this). Thus, the current grammatical nominalizations marked by dε – at least those that refer to what are deemed name- worthy entities – may become lexicalized eventually, when a new grammatical nominalization is likely to be invented. Soqotri possesses several other types of nominalization used as verb complements and other types of subordinate expressions, including some types of participial expressions. These and other possible types of nominalization require a separate treatment from the analysis advanced above. But the essential point of this paper, namely that (grammatical) nominalization plays a very important role in grammar extends beyond the scope of this paper.

8. Conclusion Both descriptive and theoretical linguists have tended to be preoccupied with form and have often failed to recognize some salient phenomena simply because they are not accompanied by some formal marking. In the case of nominalization, grammarians’ preconception centers on derivational morphology. Even when clausal nominalization is talked about, some non-finite features are sought after such as participial or gerundive forms of verbs and/or genitive marking of the subject nominal. Indeed, the typical initial reaction of most practicing linguists is negative when asked whether or not the complement clause in John thinks that Bill is honest is nominalized. Nominalization, much like other notions such as relativization and causation, is a functional, not a morphological or formal, notion, referring to the creation of referring expressions. Grammati- cal nominalization we discussed in this paper creates nominal expressions referring to entities characterized in terms of an event in which they are involved as an agent/subject, patient/object, etc. Since the entities referred to are characterized in terms of events, the process of grammatical NOMINALIZATION IN SOQOTRI 331 nominalization is perfectly compatible with tense/aspect marking and other finiteness features, since the speaker may want to characterize the relevant entity in terms of its having bought a book yesterday, for example, e.g., as in (the one) who bought a book yesterday. Ironically, the type of grammatical nominalization we examined in this paper involves a clear formal marking, namely the particle  and its variants that occur in front of the nominalized form. The problem here and elsewhere (see Shibatani 2009) was the wrong perspective both ty- pologists and theoretical linguists have had about grammatical nominalizations; namely, looking at them from the point of view of relativization, as if the relevant nominalizations were derivatives of relative clauses. In this paper, we have endeavored to clarify these two problems inherent in the past studies of nominalization and relativization, arguing for the nominalization-based analysis of relativization and other types of noun modification constructions. A proper understanding of the nature of nominalization also explains why nominalizations are used in relativization and occur in other environments such as cleft-constructions and cleft-based wh-questions.

References Adams, Douglas Q. (1972). Ancient Greek. In Paul M. Peranteau, Judith N. Levi & Gloria C. Phares (eds.), The Chicago Which Hunt: Papers from the Relative Clause Festival, 9-22. Chicago: Chicago Linguistic Society. Andrews, Avery (2007). Relative clauses. In Timothy Shopen (ed.), Language Typol- ogy and Syntactic Description. Volume 2, (Second edition), 206-236. Cam- bridge: Cambridge University Press. Comrie, Bernard & Sandra A. Thompson (2007). Lexical Nominalization. In Timothy Shopen (ed.), Language Typology and Syntactic Description, 334-381. Volume 3, (Second edition). Cambridge: Cambridge University Press. Deutscher, Guy (2009). Nominalization and the origin of subordination. In T. Givón & Masayoshi Shibatani (eds.), Syntactic Complexity: Diachrony, Acquisition, Neu- ro-cognition, Evolution, 199-214. Amsterdam: John Benjamins. Dryer, Matthew S. (2005). Order of relative clause and noun. In Martin Haspelmath, Matthew S. Dryer, David Gil & Bernard Comrie (eds.), The World Atlas of Language Structures, 366-367. Oxford & New York: Oxford University Press. Dryer, Matthew. S. (2007). Noun phrases. In Timothy Shopen (ed.), Language Typol- ogy and Syntactic Description, 151-205. Volume 2, (Second edition). Cam- bridge: Cambridge University Press. Huang, Chenglong (2008). Relativization in Qiang. Language and Linguistics 9.4: 735- 768. 332 MASAYOSHI SHIBATANI & KHALED AWADH

Keenan, Edward L. (1985). Relative clauses. In Timothy Shopen (ed.), Language Typ- ology and Syntactic Description, Vol. 2: Complex Constructions, 141-170. Cam- bridge: Cambridge University Press. Lehmann, Christian (1986). On the typology of relative clauses. Linguistics 24: 663- 680. Makhashen, Khaled Awadh (2009). Lexical and Syntactic Aspects of the Soqotri Dialect of Galansiyah. PhD dissertation, Universiti Sains, Malaysia. Makhashen, Khaled Awadh, Munir Shuib & Salasiah Che Lah (2008). The noun phrase structure in Soqotri. Paper presented at the International Conference on Mori- bund Languages andCultures 14-16 October 2008, University of Malaya. Matisoff, James A. (1972). Lahu nominalization, relativization, and genitivization. In John P. Kimball (ed.), Syntax and Semantics, 237-257. Volume 1. New York: Academic Press. Matisoff, James A. (1973). The Grammar of Lahu. Berkeley: University of California Press. Naumkin, V.V. (1998). Personal pronouns in Socotran folklore texts. In the proceedings of the seminar for Arabian studies 28, 229-232. Noonan, Michael (1997). Versatile nominalizations.In Joan Bybee, John Haiman & Sandra Thompson (eds.), Essays on Language Function and Language Type. In Honor of T. Givón. Amsterdam/Philadelphia: John Benjamins. Shibatani, Masayoshi (2009). Elements of complex structures – where recursion isn’t: The case of relativization. In T. Givón & Masayoshi Shibatani (eds.), Syntactic Complexity: Diachrony, Acquisition, Neuro-cognition, Evolution, 163-198. Am- sterdam: John Benjamins. Simeone-Senelle, Marie-Claude (1997). The Modern South Arabian languages. In Robert Hetzron, (ed.) The Semitic languages, 378-423. London: Routledge. Sneddon, James N. (1996). Indonesian: A Comprehensive Grammar. London: Rout- ledge. Weber, John D. (1983). Relativization and Nominalized Clauses in Huallaga (Hua- nuco) Quechua. Berkeley: University of California Press. Wrona, Janick (2008). The Nominal and Adnominal forms in Old Japanese: Conse- quences for a reconstruction of pre-Old Japanese syntax. In Bjarke Frellesvig & John B. White (eds.), Proto-Japanese, 199-221. Amsterdam:John Benjamins. On derivational processes in Fataluku, a non- Austronesian language in East-Timor

Aone van Engelenhoven (Leiden University Centre for Linguistics)

1. Introduction1 With its 30,000 speakers, Fataluku is the fourth language of the republic of East-Timor. It genetically belongs to the non-Austronesian Timor- Alor-Pantar branch that is a member of the Trans-New-Guinea phylum (van Engelenhoven 2006, Donohue & Schapper 2007). Whereas Hull still analyses Oirata as a ‘Fataluku dialect characterized by certain archa- isms’ (Hull 2005: 1), van Naerssen (2007), in an unpublished paper, convincingly argues that two separate languages should be distinguished. One is Oirata, spoken on the island of Kisar in Southwest Maluku (Indo- nesia) characterized by its conservative morphology, the other Fataluku exclusively spoken in the nearby Lautem District, which is in the eastern tip of the republic of East-Timor. Neighbouring languages of Fataluku are Makalero, which is confined to the Iliomar Subdistrict that borders on the Vikeke District (Huber 2008), and two severely endangered Makasai dialects – Sa’ani and Naini –, spoken in the Luro Subdistrict that borders on Baukau Dis- trict. They belong to the same Timor-Alor-Pantar subgroup as well. At least two other languages used to be spoken in the region: Ma- kuva and Rusenu, or Nisa. Makuva is an Austronesian language of the East Subgroup in Extra-Ramelaic, next to the offshore isolects of Lu- angic-Kisaric in Southwest Maluku and the Karui-Waimaha-Midiki- Naueti dialect chain in the Manatutu, Baukau and Vikeke Districts. It is confined to the Tutuala Subdistrict and was long considered to be nearly extinct. However, it turned out to be subjected to a process of ‘language concealment’ in order to prevent it from disappearing (van Engelenhoven

1 This paper was written within the framework of the Fataluku Language Pro- ject in the Endangered Languages Programme of NWO (project no. 256-70- 560). We want to thank Juliette Huber, Paulino dos Santos, Hein Steinhauer, Ruben Stoel and Justino Valentim for their advice and support. 334 AONE VAN ENGELENHOVEN and Valentim 2006, van Engelenhoven 2009). Ongoing research suggests that Rusenu or Nisa was formerly spoken in the Lautem Subdistrict where it became extinct in the first half of the twentieth century (Andrew McWilliam, p.c.). A final semispeaker of the language was found in January 2007.2 The data thus far indicate that Rusenu/Nisa was closely related to Fataluku. Fataluku today has five mutually intelligible dialects. The North and Northwest dialects are spoken in and around Lautem village and in the region at the northcoast that borders on Baukau District. The Central and South dialects are spoken in the Lospalos Subdistrict, of which the latter is confined to the Lorehe municipality.3 The East dialect is exclusively spoken in the Tutuala Subdistrict. An exclusive characteristic of this dialect is that it sometimes has a mid-back vowel [o] in final position where the other dialects have a back high vowel [u] and an initial voiced palatal occlusive [j] that corresponds with a voiced apico-dental fricative [z] in the other dialects. The North and Northwest dialects distinguish themselves by the retention of the post-alveolar occlusive [] where the Central, East and South dialects feature a voiceless palatal occlusive [c], and the glottal stop that is absent in East and obsolete in the other dialects. The South dialect in Lorehe is mainly distinguished by long vowels that are either absent or obsolete in the other dialects. It shares with the East dialect a palatal glide [j] that corresponds to a voiced apico-dental fricative [z] in the other dialects. A characteristic that spreads from the Central dialect to the others is the tendency of deleting final vowels, as for example nope ‘tomorrow’ that frequently becomes nop.

2 Since no linguistic analyses have been published on this language, I refer to the interviews in the Dutch media that can be retrieved through http://www. fataluku.com/staff/ interviews/. 3 This classification follows Hull (2005) in great outline. Valentim (2001) prefers to combine the North and Northwest dialects into one dialect, and distinguishes the lect of Kakavem near the border of the subdistricts of Lospalos and Iliomar as a separate dialect. NOMINALIZATION IN FATALUKU 335

Northwest North Central South East ‘deed’             ‘horse’                           ‘sleep’                              ‘wife’                                  ‘exist’                                  Table 1: Sound correspondences in the Fataluku dialects

Local oral tradition has it that, except for two clans, Kati ratu and Tutu- ala ratu, all clans in the district originate from outside Timor and brought along their own ancestral language (Gomes 1972). Makuva was the original language of the two indigenous clans. Two immigrant clans, Cailoro ratu and Latuloho ratu, were more or less responsible for the dissemination of Makuva and the predecessor of Fataluku, respectively. After centuries of warfare, both clans and their tributary clans united under “one correct speech” (Fatalukunu) into a uniform society. It was Latuloho ratu’s own language that was chosen as the “correct speech”. Nevertheless did Makuva manage to survive as a spoken language till the early sixties in the twentieth century. This is generally explained by the fact that the villages where it was spoken, Loikero, Porla- mano and Pitileti, were, because of their remoteness, isolated from the rest of the district. This language shift scenario explains why the Fataluku are generally reluctant to learn other languages, albeit that Makasai-Fataluku and Makalero-Fataluku bilingualism among ethnic Fataluku has been attested in the Serelau and Lorehe municipalities, repectively.4 It also suggests intensive language contact between the linguistically diverse clans before they were unified. Ongoing research confirms this hypothesis. The most salient evidence of language contact is the extensive lexical parallellism in Fata- luku oral traditions, which appears nearly fully compatible with the lexical pairs found in the Austronesian offshore languages in Southwest Maluku. Fox (2005) attests the same phenomenon in the non-Austro-

4 Personal communication by Paulino dos Santos (Lospalos) and the late Prof. Henri Campagnolo (email message October 25th, 2004). 336 AONE VAN ENGELENHOVEN nesian Bunak language and its Austronesian neighbour Kemak and analyses it as an explicit influence of the latter on the first. As in most languages in the region, Fataluku oral traditions make use of clichés or proverbial sayings that epitomize a history. Sometimes, however, these cli- chés are in a language that is not Fataluku and consequently no longer understood by the audience.5

2. Typological sketch Hull (2001, 2004) explains how Fataluku and its co-geners adapted to the typology of the Timor Sprachbund and as such have diverted from the Trans-New-Guinean language type that features elaborate morphological systems. From a phonological point of view, Fataluku has a segment inventory that is even simpler than most other languages in East-Timor6. Its only voiced occlusive is either the post-alveolar [] in the North and Northwest dialects, or the voiced palatal [j] in the East dialect. Of the three eastern languages, Makasai is completely isolating. Makale- ro is typologically somewhere in between Makasai and Fataluku (Huber 2008). Fataluku, like its genetic neighbours, has a SOV word order. Loca- tions and directions can be encoded as postpositions or as prefixes on the verb. The latter strategy is preferred. This is exemplified in (1) below, where the locational ‘on’ is prefixed to the verb nate ‘stand’ in the relative clause (indicated by square brackets) and the directional ‘up’ is prefixed to the main verb moi ‘ascend’.

5 Elsewhere (van Engelenhoven 2008), I elaborate that their bleaching semantic character makes these proverbial sayings rather unsuccesful instruments in the memorization of narratives. This is especially salient in Josselin de Jong’s (1937) transcription of an Oirata myth that contains small songs in a language that nobody understands. Valentim (2002: 114) accidentally mentions an example in his dictionary: ‘Woro-konai: a word that one pronounces when one tells stories from the past (we do not know its meaning).’ The fact that Valentim is a member of the Cailoro ratu clan that still uses Makuva for its rituals, and the form of the word suggest that it is actually Makuva. 6 The elaborate consonant systems of the Karui-Waimaha-Midiki-Naueti dialect chain is a striking exception (Hajek & Bowden 2002, Hull 2002). NOMINALIZATION IN FATALUKU 337

(1) 7 ... malai piti-ne le afa Ø mica-nat-e hi’a-moi... foreigner white-VEB house[1plinc.SØ on-stand-VEB] up-ascend ‘... a white stranger climbed the house we were standing on ...’

Equally in line with the SOV word order is its possessive phrase, in which the possessor precedes the possession noun. Fataluku has a special set of pronouns to encode either object arguments or possessors. For example, “our (inclusive) house” in Fataluku is afi le (1plinc house). In example (1) above ‘1plinc’ afa ends in a low vowel and as such is marked as the subject argument. Conform the pattern found elsewhere in the Timor Sprachbund, Fataluku also uses the third person singular marker i as a possessive marker between possessor and possession nouns. This is exemplified in (2) where (a) displays the Fataluku construction and (b) its counterpart in Tetum, the Austronesian language that is one of East- Timor’s official languages.

(2) a. nami i ca’u-taka (Fataluku) b. mane nia xapeu (Tetum) man 3sg head-cover man 3sg hat ‘the hat of the man’ ‘the hat of the man’

Like many other languages in the region, Fataluku displays extensive clause-combining and verb serialisation (van Engelenhoven, in press). Its verbal morphology itself, however, rather sets it apart from the other languages. Campagnolo (1973) observed that many verbs display initial consonant mutation of /f/ to/p/, and /t, h, s/ to/c/, when prefixed by either a verbal or locational/directional morpheme. This is exemplified in Table 2a below. f → p fa’i → -pa’i ‘make, do’ fetil → -petil ‘stumble’ fulutete → -pulutete ‘spit’ t → c taia → -caia ‘sleep’ teku → -ceku ‘stir’ tomok → -comok ‘be soft’

7 For the abbreviations, refer to the list at the end of the paper. 338 AONE VAN ENGELENHOVEN h → c her → -cer ‘dry’ hina → -cina ‘plait’ hura → -cura ‘take (cooked food) from the fire’ s → c sil → -cil ‘bind’ Table 2a: Verbs displaying initial consonant mutation

Another exclusive feature that distinguishes Fataluku from all other languages in the region is the prothesis of a vowel to a transitive verb that anaphorically refers to the ‘object’8 in a previous clause. If the first syllable of the root is a mid vowel, the prefixed vowel is nearly always a copy. In other cases, the prefixed vowel is unpredictable: /i/, /u/ or /a/. This is exemplified in Table 2b.

#iC ← #C icuar ← cuar ‘sit (on) (pl)’ imir/amir ← mir ‘sit (on) (sg)’ ipah ← pah ‘disseminate’

#uC ← #C unam ← nam ‘catch in the hands’ ulavere ← lavere ‘clothe’ ulur ← lur ‘sweep’

#eC ← #C eme ← me ‘take’ eler ← ler ‘read’ ekeh ← keh ‘count’

#oC ← #C okolev ← kolev ‘embrace’ onof ← nof ‘die’

#aC ← #C ahar ← har ‘send’ ape ← pe ‘move (on sth.)’ aruka ← ruka ‘multiply’ Table 2b: Verbs displaying vowel prothesis

8 I use here the term ‘object’ because of lack of a more appropriate one. Vowel prothesis also occurs on certain verbs that classify as prototypical monovalent verbs, as for example imir ~mir ‘sit’ and onof ~nof ‘die’ in Table 2b. NOMINALIZATION IN FATALUKU 339

Campagnolo (1973) distinguishes two separate verb classes that display one of these morphological processes. To these classes he adds a third, where the verbs display the effect of both processes.9 This is shown in Table 2c.

#VC ← #C ofot ← fot → pot ‘cut up’ f → p ufal ← fal → pal ‘grab’ ufan ← fan → pan ‘feed’

#VC ← #C utu’ ← tu’ → cu’ ‘plant’ t → c oto’ ← to’ → co’ ‘be inside (sg)’ itih ← tih → cih ‘cook anew’

#VC ← #C h → c ihir ← hir → cir ‘wait’ aha’ ← ha’ → ca’ ‘warp’ uhuleve ← huleve → culeve ‘spoon up’

#VC ← #C isil ← sil → cil ‘bind’ s → c Table 2c: Verbs displaying initial consonant mutation and vowel prothesis

Ongoing research suggests another morphological feature in Fataluku reminiscent of Jonker’s (1906) observation of the ‘anorganic consonants’ or final consonant suffixes in Austronesian languages of the region. This is nicely exemplified by the following verbs that suggest a lexical root

9 In fact, Campagnolo (1973: 110-139) did not acknowledge separate verb classes, but rather distinguished ten morpheme classes with different phonological alternations in the stem. The first three classes he mentions, however, are all verbs. A consequence of this approach is that another class, the verbs that do not add an initial vowel or mutate an initial consonant, is overlooked. Rather, the data suggest that Campagnolo’s three verb classes are subgroups of one major class where the verb changes its phonological make-up, the other major class being the one where the verb form remains unchanged. 340 AONE VAN ENGELENHOVEN lau: lau-’ ‘dissolve’, lau-k ‘putrify’, lau-s ‘be mixed with something cold’.10 The fact that the dictionaries of Nacher (2002/2003, 2004) and especially Valentim (2002) list these and other comparable words as separate lemmas is a strong indication that this prefixation is no longer a productive device in the language. It does, however, imply that Hull’s (2004) thesis of Fataluku having adapted to the isolating typology of the Timor Sprachbund requires to be revisited. Unlike its neighbours Maka- sai, which indeed is fully isolating, and Makalero, Fataluku still displays agglutinative morphology. This is elaborated in the next paragraphs that focus on nominalisation in Fataluku.

3. Nominalisation versus verbalisation: /-(n)u/ 3.1 Abstract, product, mass and ‘additive’ nouns

This paragraph focuses on the suffix /-(n)u/, which is the main nominalising device in the language. Campagnolo (1973: 132-134) describes how the /-nu/ allomorph attaches to morpheme-final vowels.11 This is displayed in the following table.

‘see’ aci aci-nu ‘vision’ ‘lack’ eru eru-nu ‘shortage’ ‘be mute’ kolo kolo-nu ‘muteness’ ‘see’ toto toto-nu ‘view’ ‘be lean’ lika lika-nu ‘leanness’ Table 3a: Nominalisation with /-nu/: abstract nouns

The /-u/ allomorph attaches to morpheme-final consonants. However, because Fataluku phonotactics disallow utterance-final consonants, nouns that are pronounced in isolation also receive an /-u/ vowel if their roots end in a consonant. This phenomenon seems reminiscent of Stre-

10 As far as we know no such phenomenon has been attested in the Timorese co-geners of Fataluku. 11 Hull (2005: 12) overlooked this fact and thought /-nu/ was confined to verbal roots ending in /i#/, NOMINALIZATION IN FATALUKU 341 semann’s (1927) and Mills & Grima’s (1980) observation of ‘echo vowels’ in Central and Southwest Malukan languages, respectively, that are added to morphemes with final consonants. The fact that /-(n)u/ is mutually exclusive with its verbalising counterpart /-(n)e/ gives way to formations like tupur-u ‘woman’ versus tupur-e ‘be a woman, feminine’. The following table displays examples of /-u/ that are opposed to their counterparts marked with the verbalising suffix.

‘chop’ (N) lama’-u lama’-e ‘chop’ (V) ‘crumb’ lamak-u lamak-e ‘crumble’ ‘vomit’ (N) i’is-u i’is-e ‘vomit’ (V) Table 3b: Nominalisation with /-u/ versus verbalisation with /-e/

Hull (2005: 12) observes that /-u/ is mutually exclusive with /-e/. He fails, however, to recognise the latter as a verbalising morpheme. Instead, he considers /-e/ an intrinsic element of the verb. As such, he describes the function of /-u/ with verbs as creating nouns ‘with a concrete or nearly concrete sense’, which is also shown in Tables 3a and b. We suggest, however, to re-analyse the examples in Table 3b rather as product nominalisations in which the noun profiles the product and the verbal counterpart profiles the corresponding action. This is also shown by the examples in Table 3c. Observe, however, that whereas the first three sets project scenes in which a product evolves from the verbal act by an agent, the last set rather projects a performance by the referent of the nominal morpheme. 342 AONE VAN ENGELENHOVEN

‘sign (N)’ catan-u catan-e ‘sign (V)’ ‘orchard’ laman-u laman-e ‘make an orchard’ ‘chop (N)’12 lapar-u lapar-e ‘chop (V)’ ‘tuber’ apat-u apat-e ‘emerge from the soil’ Table 3c: Product nouns versus production verbs marked with /-e/

It is very difficult to determine on formal grounds whether the root morpheme is to be classified as a noun or a verb. It is the context, which determines whether such a root morpheme – if it occurs without /-u/ or /e/ – has to be interpreted nominally or verbally. As such, this type of ‘final consonant roots’ are reminiscent of what is often called ‘precategorial’ in Austronesian: they do not belong to any word class, but become a member of one within a certain morphological or syntactic context.13 Purely from a cognitive point of view, the sets above suggest that the root is intrinsically a noun, because it refers to a ‘time-stable concept’ (Payne 2006: 33). This is also displayed in the following table where from a logical point of view the root is nominal and the verb refers to the action of adding or placing what is profiled by the root on to something. In the last three examples the nouns rather profile the instrument of the verbal act.

‘lime’ aur-u aur-e ‘put lime on s.th.’ ‘salt’ asir-u asir-e ‘put salt on s.th.’ ‘saddle’ kosin-u kosin-e ‘saddle (a horse)’ ‘nautilus’ lilir-u lilir-e ‘decorate (a house) with nautilus shells’

12 Valentim (2002: 36) describes it as follows: laparu- alivanatu harak hai me eme a pelere i ta’a place SEQ cut PERF take it.take LOCremove DEM say ‘chop- means a place where a cut has been taken out’ 13 See, however, Himmelmann (2005) who convincingly invalidates precate- gorials for Philippine languages. NOMINALIZATION IN FATALUKU 343

‘file’ limar-u14 limar-e ‘to file’ ‘hammer’ leler-u leler-e ‘to hammer’ (South dialect) ‘sword’ hit-u hit-e ‘hit with a sword’ Table 3d: ‘Additive’ nouns and their verbal counterparts

The example of tupur-u ‘woman’ ~ tupur-e ‘be feminine’ mentioned above seems copied after mass nouns with final consonants whose verbal counterparts are marked with /-e/. This is exemplified by the nouns in Table 3e.

‘sand’ inik-u inik-e ‘be sandy’15 ‘stone’ matar-u matar-e ‘be stoney’ ‘gold’ karas-u karas-e ‘be goldlike’ ‘mud’ lumuk-u lumuk-e ‘be muddy’ Table 3e: Mass nouns and their verbal counterparts

With other roots, however, classification on semantic grounds is less evident. Some of them are displayed in Table 4a.

k.o. illness ifi-nu ifi ‘maggot, to have maggots in one’s wounds’ ‘gambling (N)’ taru-nu taru16 ‘gamble (N), to gamble’ ‘plaster’ lapa-nu lapa k.o. medical plaster17 ‘wounded manu-nu manu ‘wound (N)’ condition Table 4a: Double class morphemes and their nominalised counterparts

Informants automatically classify the root morphemes as nouns, despite the fact that this root type also occurs unmarked in predicate position.

14 < Portuguese limar ‘to file’. 15 Campagnolo (1973: 134) and Valentim (2002: 24) also mention a metonymi- cal extension of the initial meaning of ‘sandy’: ‘to winnow’ and ‘make “very lime-like” (fa’i tu na’u aure)’.Observe that aur-e (lime-VEB) exemplifies as a verb of ‘adding’ in Table 3d. 16 < Tetum taru ‘to gamble’. 17 Made of ripped bamboo (Nacher 2002/3: 186). 344 AONE VAN ENGELENHOVEN

The following examples (3a) and (b) respectively show a Fataluku sentence and its colloquial Indonesian counterpart where the same phenomenon occurs. In the co-official language Tetum, however, it is less normal (3c).18

(3) a. Alfredo hai n=umu, presidente hala manu. (Fataluku) Alfredo PERF LNK=die president only wound. b. Alfredo sudah mati, presiden luka saja (coll. Indonesian) Alfredo PERF dead president wound only c. Alfredo mate tiha ona, presidente isin-kanek de’it. (Tetum) Alfredo die PERF president body-wound only a-c: ‘Alfredo is already dead, the president is only wounded.’

3.2 Cardinal numerals Fataluku has a decimal numeral system. Except for ‘1’, ‘2’, ‘3’, ‘8’ and ‘10’, all basic cardinal numerals clearly are Austronesian loans. Except for the numbers ‘7, ‘8’ and ‘9’ all numerals are marked as verbs by the /-e/ suffix19, which Klamer (2002) mentions as a typical Austronesian, albeit marked, feature of languages in Central and Eastern Indonesia. In- terestingly, the numbers marked as verbs receive a nominalising suffix /- u/ when they function as an ordinal numeral and are preceded by the ‘3sg’ pronoun, which also functions as a possessive marker (see example (2a) above). Whereas Campagnolo (1973) simply refers to this marker as a ‘determination marker’ (p. 132), Hull (2005: 29) specifically labels it a possessive pronoun. This will be elaborated in the section on the suffixes /-‘ina/ and/ -(n)ana/. Table 4b below displays the cardinal numerals and

18 As far as we know now the phenomenon is absent in the languages surrounding Fataluku. As such it may be an example of grammatical influence of Indo- nesian on Fataluku. Hull (2005), however, reports that Fataluku has many lexical loans from local Malay (= dialectal Indonesian) but does not mention grammatical influences. 19 E.g. fitu ‘7’ (< Proto-Austronesian *pitu) ~ i fitu ‘seventh’, kafa ‘8’(< *kapaw, according to Hull (2004: 86), who reconstructed this form on the basis of the related kapa ‘8’ in Oirata; the Proto-Austronesian original, however, is *walu. Consequently, the Fataluku (and Makasai and Makalero) variant may be linked to a non-Austronesian original. NOMINALIZATION IN FATALUKU 345 their ordinal counterparts. Observe that the possessive marker is the only formal indication for the ordinal counterparts of the cardinal numerals that are not marked by the verbalising /-e/. Morphemes with initial vowels receive a prothetic /h/ in combination with the possessive marker /i/. For comparative purposes, the protoforms of the Austronesian originals are added where applicable. first i h-ukani / i h-emer-ana20 ← ukani 1 second i h-ec-u ← ec-e 2 third i hutu’-u ← hutu’-e 3 fourth i fat-u ← fat-e (< *Sempat) 4 fifth i lim-u ← lim-e (< *lima) 5 sixth i nem-u ← nem-e (< *enem) 6 tenth i ta’an-u ← ta’an-e 10 hundredth i rah-u ← rah-e (< *Ratus) 100 thousandth i rehun-u ← rehun-e (< *Ribu) 1000 Table 4b: The nominalising suffix to mark ordinal numbers and their verbal counterparts marking cardinal numbers

The attributive use of the nominalising suffix is generally confined to the numerals. However, Campagnolo (1973: 134) mentions one other case in the South dialect where a noun, koco ‘bamboo wall’, is used attributively with a nominal suffix and preceded by the possessive marker /i/: i koco- nu (3sg bamboo.wall-NOM) ‘the one with walls’21. We are not aware of the existence of a verb *koco-ne (bamboo.wall-VEB), which suggests that ‘bamboo wall’ is an exclusive case within the double class morphemes mentioned in table 4a.

3.3 Adverbs A productive process in Fataluku is the use of nominalised state verbs functioning as adverbs to a following verb. Interestingly, neither Cam- pagnolo (1973) nor Hull (2005), nor the lexicographers Nacher (2002/3, 2004) or Valentim (2002) have noticed this fact, maybe because in these

20 I h-emer-ana is actually a deverbal attribute meaning ‘(being the) first’. See also the section on /-‘ina/ and /-(n)ana/. 21 The original French translation is “celle (la maison) dont les murs sont cons- truits”. 346 AONE VAN ENGELENHOVEN adverb-verb constructions, the vocalic ending of the nominalising suffix generally disappears in running speech, as is displayed in example (4a). In our discussions with school teachers on Fataluku spelling, the hypercorrect counterpart in (4b) came up, showing that the adverb modifier is identified as a noun.22 Example (4c) shows that on verbal roots with a final vowel it is the /n/ a of the nominal /-nu/ allomorph that evidences this nominalisation strategy.

(4) a. ... muha hi’a rane tava emer mohe. (running speech) few up adorn 3sg first run ‘(but then) a few dressed up ones ran up to him first.’ b. ... muha hi’a rane tava emer-u mohov-e (hypercorrect sp.) few up adorn 3sgfirst-NOM run-VEB ‘idem’ c. Afa valale-n ete kaur-e akam sala 1plinc.S quick-NOM off leave-VEB NEG23 wrong ‘If we leave quickly we are not wrong’ (question 361, Fataluku Catechism).

The nominal affix /u/ is systematically present on adverbially used state verbs that contain an offglide.24 In this phonotactic context it is the offglide in the root rather than the nominalizing /u/ that is deleted. This is exemplified by masu (< mais-u ‘right-NOM) in (5a).

22 Fataluku speakers in general are obsessed with the correctness of their speech. Fataluku itself means ‘to speak correctly’. Although it does not have a written tradition, there is a general concensus on how Fataluku words should be written as became very apparent during the three workshops held in 2005, 2006 and 2007. Whether or not to write the final /u/ of these preverbal adverbs is still a matter of discussion, however, because they are not regularly pronounced. The purpose for the topic of this paper is that speakers identify the adverb and its ‘deleted final vowel’ as nominalisations. 23 In most dialects predicate initial negation is a petrified combination of aka ‘NEG’ + the m allomorph of ‘take’. Aka ‘NEG’ only survived in the East dialect (see van Engelenhoven, in press). 24 Albeit not expressed specifically as such, we understand from Campagnolo’s (1973) and Stoel’s (2007) syllable analyses that the [j] and [w] are allophones of the high vowel phonemes /i/ and /u/, when they occur in the syllable’s mar- gin. NOMINALIZATION IN FATALUKU 347

(5) a. ... lau i tahin-u lavere ho tapi masu (< mais-u) cloth 3sg beautiful-NOM clothe and very right-NOM lauh-e... live-VEB ‘... (who) wear expensive cloths and live in luxury...’ (Gospel of St. Luke 7: 25)

Elsewhere (van Engelenhoven, in press), we observed that this type of deverbal adverbialisation creates serial verb constructions in which the deverbal adverb fills the object slot of the bivalent verb. The noun referring to the instrument, product, or theme participant of the scene projected by the verb is encoded as the object of ‘take’ in the directly preceding clause. This is exemplified in (5b) where tour-u tu relijaun i telira nere (mass-NOM SEQ religion 3sg different follow) ‘people who adhere to a different religion’ is encoded as the object of me ‘take’ whereas it profiles the perceived participant in the scene of ‘seeing’.25

(5) b. Afa ... tour-u tu relijaun i telira nere 1plinc.S mass-NOM SEQ religion 3sg different follow me ranu (

Lau i tahin-u (cloth 3sg beautiful-NOM), literally meaning ‘beautiful cloth’ in example (5a) above taken from the Fataluku translation of St. Luke’s Gospel, shows that nominalised verbs by means of /-(n)u/ are also

25 Thilo Schadeberg in a personal communication during Ruben Stoel’s presentation about Fataluku prosody on December 7th, 2007 at Leiden University pointed out that Fataluku seems to have a ‘1 argument only rule’ that ‘pushes’ prototypical object candidates to a ‘take’ clause in case of nominal (deverbal) adverbs. Further research is required to assess what the position is of intransitive constructions as in (5a) where a nominal adverb masu (< mais-u ‘right- NOM’) modifies a monovalent verb lauh-e (live-VEB) and itself is also modified by the adverb tapi ‘very’. 348 AONE VAN ENGELENHOVEN used to translate adjective-like modifiers in foreign languages.26 In fact, this construction is ambiguous in that it also can be translated as ‘the cloth’s beauty’. Valentim (2002: 95) explicitly indicates that nouns referring to ‘things’ (nacunu) receive tahin-u with a nominal ending,whereas humans, for example tupur ‘woman’ rather require tahin-ana: tupur i tahin-ana (woman 3sg beautiful-ANA) ‘beautiful woman’. As has been said above, this use of i ‘3sg’ creates possessive constructions in which the possessor noun is placed to the left of the pronoun and the possession noun to its right. From a syntactic point of view, then, the noun lau ‘cloth’ would be the modifier in a possessive construction (‘the cloth’s beauty’), but the head in a NP containing an adjectival modifier (‘beautiful cloth’). This will be elaborated further in the following paragraph. Table 5 provides a few examples from a ‘rest group’ where the nominalising suffix creates nouns that cannot be categorised with the ones discussed above.

‘citron sap’27 aka-akah-u akah-e ‘be sour’ k.o. bird of prey hoil-u hoil-e ‘cut in wedges’ ‘nail’ keles-u keles-e ‘pinch’ ‘twig’ ketel-u ketel-e ‘meagre’ k.o.bird kokir-u kokir-e ‘collect’ ‘fenced place’ laca-nu laca ‘fence (V)’ ‘broom (N)’ lu-lur-u lur-e ‘broom (V)’ Table 5: other nouns by /-(n)u/ and their verbal counterparts

4. Nominalisation or adjectivisation: /-(’)ina/ and /-(n)ana/ This section discusses the suffixes /-(’)ina/ and /-(n)ana/. Like the nominalising suffix discussed in the previous section, the adjective-forming morphemes have two allomorphs of which those with initial /‘/ and /n/ occur after roots with a final vowel, while the other allomorph occurs

26 The Indonesian original from which it was translated is ber-pakai-an begitu (VEB-[wear-NOM such]) ‘wear such clothing’ in which ‘such’ refers to ‘beautiful’ in the previous clause. The square brackets above intend to show that in Indonesian it concerns an NP being verbalised by the prefix ber-. 27 Explained in Valentim (2002: 3) as “something that makes (food) sour”. NOMINALIZATION IN FATALUKU 349 after roots with a final consonant. Whereas Campagnolo (1973), Hull (2005), Nacher (2002/3, 2004), and Valentim (2002) all mention the suffix /-(n)ana/, only Campagnolo (1973) explicitly mentions /-ina/. Valen- tim (2002) only mentions one instance of /-ina/: po’otina ‘ill (person)’. The dictionary of Nacher (2002/3, 2004) and the Fataluku wordlists of Fataluku.org, on the other hand do mention several other instances of /-(’)ina/ that are nevertheless outnumbered by instances of /-(n)ana/. There are many instances in Fataluku of morphemes ending in /-(’)ina/ that have no verbal counterpart. Some of these are displayed in Table 6a. Observe that the notion of ‘extend’ distinguishes between a singular and a plural form, which is typical for a subset of verbs in Fata- luku, Makalero and Makasai. If it is true that the forms in Table 6a contain a petrified synchronically opaque suffix, then the kusina~kusevere opposition implies that these forms are verbal.

‘firm’ akina ‘end(N)’ hufina ‘drug’ humina ‘extend.SG’ kusina kusevere ‘extend.PL’ ‘iron’ mumina ‘termite queen’ parakina ‘anvil’ pikorina ‘roof’ pui‘ina ‘lowland’ puicina ‘inside of a corner’ valikina valiku ‘corner, elbow’ Table 6a: Opaque instances of /-(’)ina/-derivations

An interesting case is represented by the last example of Table 6a, for which the counterpart of valikina ‘inside of a corner’ is the noun valiku ‘corner’. Based on the discussion in section 3, one would surmise that the latter has a nominal suffix /-u/,which is mutually exclusive with /-(’)ina/. The verbal counterpart ‘to be in a corner’ is valiku-ne (corner-VEB) instead of *valik-e.28

28 The verb acaku-ne (match-VEB) is the only other example we are aware of where the final /u/ is deleted when /-ina/ is added, but remains when the nominal suffix /-nu/ is added. An ad-hoc solution might be that both verbs acaku-ne (match-VEB) and valiku-ne (corner-VEB) derive from the deverbal and de- 350 AONE VAN ENGELENHOVEN

Table 6b provides two more examples that are clearly derived. Like valikina ‘inside of a corner’, cai’ina looks nominal. It refers to a permanent dry rice field. It corresponds to the verb cai-ne (flint-VEB) ‘light something’29, which itself has a nominal root. On the other hand, Acu’ina ‘fit’ seems rather ‘adjectival’ or ‘verbal’ in that it occurs as a predicate.30 Its verbal counterpart is acu-ne (be.over-VEB) that itself is a verbalisation of the postposition acu ‘over’.

‘fit’ acu‘ina acu-ne ‘be over s.th.’ acu ‘over (PP)’ k.o ricefield cai‘ina cai-ne ‘light s.th.’ cai ‘flint’ Table 6b: /-‘ina/ derivations with verbal and other counterparts

Since the roots in table 6b end in a vowel, the glottal stop /‘/ is part of the derivational suffix. The phonotactic constraints are the same as for the nominalising suffix mentioned in the previous section; the allomorph with the initial glottal stop occurs after roots with final vowels, the allomorph without the glottal stop occurs after roots ending in a consonant. The following table displays the frequent most instances of /-(’)ina/. It has only been attested with verbal roots.

nominal postpositionals aca “at” and vali “beside”, respectively. This would suggest that /-ku/ be an example of a nominalising ‘anorganic suffix’ /k/ plus an epenthetic /u/ (see section 2). This seems to be borne out by the full derivational paradigm of aca “arrive at/reach”: aca-ne (arrive.at-VEB) “suffice” : aca-nu (arrive.at-NOM) “perfect(ion, as in cele acanu ‘maize that is ready for roast- ing”, Nacher 2002/3: 141)”: aca-nana (arrive.at-ANA) “sufficient”; aca-k-e (arrive.at-K-VEB) “concern (V)”: aca-k-u (arrive.at-K-NOM) “concern (N)” (Nacher 2002/3: 140) : aca-k-u-ne (arrive.at-K-NOM-VEB) “match (V)” : aca- k-ana (arrive.at-K-ANA) “matching, match (N)”. The term ‘anorganic suffix’ was introduced into Timorese linguistics by J.C.G. Jonker (1906) in his authori- tative paper on consonant suffixes in the Austronesian languages of Timor and surroundings that were not part of the lexical morpheme. 29 The scenario of a permanent dry rice field is that in order to create one, a part of the forest is cut and burnt down first. 30 For example: A leura fa’a te acu-‘ina? (2sg meat part where over-INA) “How big is your share (of the meat)?” (Fataluku Community Internet wordlist 2 at www.fataluku.org). NOMINALIZATION IN FATALUKU 351

‘burning’ ho’-ina ho’-e ‘burn’ ‘sweet’ lemes-ina lemes-e ‘be sweet’ ‘right’ mais-ina mais-e ‘be right’ ‘s.th. stolen’ o’os-ina o’os-e ‘steal’ ‘rotten’ opos-ina opos-e ‘rot’ ‘ill (person)’ po’ot-ina po’ot-e ‘be ill’ ‘ripe’ poros-ina poros-e ‘be ripe’ ‘dry’ sasak-ina sasak-e ‘be dry’ ‘boil’ so-sok-ina sok-e ‘have boils’ ‘lid’ sul-ina sula-ne ‘stop, close’ ‘s.th. severed’ tot-ina31 tot-e ‘sever’ ‘froth’ ufut-ina ufut-e ‘seethe’ Table 6c: Derivations with /-(’)ina/ and their verbal counterparts

A quick glance at table 6c tells us that seven out of twelve roots contain the back mid vowel /o/. Eight roots end in one of the dental consonants /t/ or /s/ . The glosses of five of them, ‘sweet’, ‘right’, ‘ill’, ‘ripe’ and ‘dry’ suggest to the reader an adjectival interpretation. However, Cam- pagnolo (1973) classifies /-(’)ina/ as a suffix nominalising verbs, but systematically translates his examples into French by way of relative clauses, for example: i h-ufut-ina (3sg LNK-seethe-ina) “celle (l’eau, etc. ...) qui bout”: “ the one (water, etc. ...) that boils” (Campagnolo 1973: 135). In other words, the nominal glosses of ‘froth’ and ‘lid’ can be also paraphrased as ‘something that seethes/fizzes’ and ‘something that closes something’, respectively. Hence, ho’-ina (burn-ina) ‘burning’ can be paraphrased as ‘something that burns’. Formulated differently, these three /-ina/ derivations seem to refer to the agent participant in the scene profiled by the verb. Because Fataluku lacks a specific adjective word class, the same analysis applies to the ‘adjectives’ mentioned above. Tot- ina (sever-ina) ‘something (that is) severed’ and o’os-ina ‘something (that is) stolen’, however, rather refer to the patient participant in the scene profiled by the verb. This is further elaborated below where we discuss the suffix /-(n)ana/.

31 This derivation is mainly known as the Fataluku name of the island of Jako off the eastern tip of East Timor. Oral tradition has it that Jako used to be part of the mainland as the head of the country, mu’a ca’u (land head) but was severed from it by a boat sailing through the part that connected Tutuala with Jako. 352 AONE VAN ENGELENHOVEN

Compared with the /-(’)ina/ derivations above, there are very few instances of opaque /-(n)ana/ derivations. The most salient example is alivana ‘place’, which suggests a root aliv that is not attested in the language. Both Campagnolo (1973) and Hull (2005) classify the suffix /-(n)ana/ as nominal. Whereas the use of /-(’)ina/ is quite confined, a quick glance at Valentim’s (2002) dictionary shows that /-(n)ana/ is even more productive than the nominalising /-(n)u/. Also, /-(n)ana/ derivations always have a verbal counterpart. Table 7a shows that the /-(n)ana/ derivation refers to an entity conceptually projecting those features that we recognise as elements of the state or condition profiled by the root.

‘s.th. being swollen’ akat-ana akat-e ‘be swollen’ ‘s.th. being rough’ as-ana as-e ‘be rough’ ‘s.th. being fat’ auh-ana auh-e ‘be fat’ ‘s.o. being tired’ kaiar-ana kaiar-e ‘be tired’ ‘s.o. being beautiful’ tahin-ana tahin-e ‘be beautiful’ Table 7a: /-(n)ana/ derivations of state verbs

Table 7b, on the other hand, displays monovalent action verbs whose /-(n)ana/ derivation refers to the acting participant in the profiled scene or state.

‘s.th. growing’ erek-ana erek-e ‘grow (of weeds)’ ‘s.th. falling’ ican-ana ican-e ‘fall’ ‘s.th. descending’ isi-nana isi ‘descend’ ‘s.th. floating’32 palis-ana palis-e ‘float’ Table 7b:/-(n)ana/ derivation of monovalent action verbs

In case of monovalent verbs expressing sounds, the /-(n)ana/ derivation ambiguously refers to either the participant making the sound, or the sound itself. This is displayed in Table 7c.

32 Nacher (2004: 142) translates this word as ‘island’. Interestingly, Fataluku does not seem to have a single lexeme for the concept of ‘island’. In both Valentim (2002) and the Internet wordlists of Fataluku Community it is paraphrased as mu’a-ca’u-vele je cau(-nu) (soil-head-skin [= world] self put (- NOM)) ‘soil that is set apart’. NOMINALIZATION IN FATALUKU 353

‘s.th. sounding/ sound(N)’ cu’ut-ana cu’ut-e ‘sound (V)’ ‘s.th. wailing/wail (N)’ kikir-ana kikir-e ‘wail (of dogs)’ ‘s.th. snoring/snore (N)’ kofor-ana kofor-e ‘snore (V)’ ‘s.o. grumbling/grumble (N)’kohot-ana kohot-e ‘grumble (V)’ ‘s.o. singing/song’ kol-ana kol-e ‘sing’ ‘s.o. shouting/shout (N)’ kul-ana kul-e ‘shout (V)’ Table 7c: /-(n)ana/ derivation of verbs referring to sound production

We know of three sound production verbs that do not follow this pattern. Koso ‘shout’ uses the nominal suffix /-(n)u/ to refer to the sound produced: koso-nu (shout-NOM) ‘shout (N)’, while the /-(n)ana/ derivation profiles the participant that shouts: koso-nana ‘s.o. shouting/ shouter’. Katih-e (yell-VEB), on the other hand, only derives a production noun by means of /-(n)ana/: katih-ana (yell-ANA) ‘yell (N)’. The concept of cry- ing has only been attested profiled by a verb, arur-e (cry-VEB). Bivalent verbs with a /-(n)ana/ suffix refer to the non-active participant in the scene, which in most cases is the patient or the product of the action profiled. This is displayed in Table 7d.

‘s.th. sliced/slice (N)’ afil-ana afil-e ‘slice (V)’ ‘s.th. spun/thread’ aik-ana aik-e ‘spin’ ‘cooked food’ culu-nana culu ‘cook’ ‘wish, want(N)’ eluh-ana eluh-e ‘want’ ‘payment’ hiar-ana hiar-e ‘pay’ ‘s.th. hunted/prey’ hoil-ana hoil-e ‘hunt’ ‘nail (N) ruh-ana ruha ‘nail (V)’ Table 7d: /-(n)ana/ derivation of bivalent action verbs

The final example, ruh-ana (nail-ANA) ‘nail (N)’ ~ ruha ‘nail (V)’ is the only example we are aware of in which a final vowel of a verbal root is replaced by the /-ana/ suffix. Ruha ‘nail (V)’ itself, however, occurs in the Austronesian languages off shore in Southwest Maluku as a noun meaning ‘nail’, for example Meher: ruha and Leti: rusa. We therefore suppose it to be a loan rather than original Fataluku, albeit that it remains unexplained why the /-(n)ana/ derivation is ruhana rather than *ruha- nana. As may have become clear from the foregoing discussion, we prefer not to analyse the /-(’)ina/ and /-(n)ana/ derivations as nominal. Our 354 AONE VAN ENGELENHOVEN reluctance is also based on the fact that Valentim’s dictionary (2002) lists many instances where the /-(n)ana/ derivations are themselves nominalised by means of /-(n)u/. This suggests that the derivation without the nominal suffix is not a noun, but something ‘verbal’. Furthermore, the internet wordlists of Fataluku Community provide some accidental examples of /-(’)ina/ derivations being nominalised by /-(n)u/. In both Valentim’s dictionary and the Internet wordlists, the nominal suffix derives nouns that profile abstract notions. These are displayed in Table 7e.

‘s.th. flying’ ipil-ana ipil-ana-nu ‘flying behaviour’ ‘s.th. superfluous’ lam-ana lam-ana-nu ‘condition of being superfluous’ ‘s.th. sorted’ lamil-ana lamil-ana-nu ‘situation ofbeingsorted’ ‘s.th. dissolved’ lau’-ana lau’-ana-nu ‘state of being dissolved’ ‘s.th. living’ lauh-ana lauh-ana-nu ‘life’ ‘s.th. cut’ tot-ina tot-ina-nu ‘stateofbeingcut up(?)’33 Table 7e: Nominalisation of /-(n)ana/ and /-(’)ina/ derivations

We will consequently not follow Campagnolo (1973) and Hull (2005), who analyze the /-(n)ana/ and /-(’)ina/ derivations as nominalising opera- tions. We rather follow the suggestion contained in Campagnolo’s translations according to which these derivations are modifiers, when following a noun. This interpretation is corroborated by the Fataluku description for the concept ‘human being, man’ versus the concept ‘animal’. The following sentence in (6a) has been extracted from Valentim’s (2002) dictionary.

(6) a. ... lau-lauh-ana unu i tu ma’ar lauh-ana apiapar-e tu RED-live-ana one DEM SEQ man live-ana care-VEB SEQ navar-u cura tu le hi’a-lauh-e una. know-NOM mouse SEQ house up-live-VEB eat “an animal (= something that lives) that a human (= man that lives) takes care of in order to eat mice that live in the house.” (= cat, Valentim 2002: 74)

33 The example given in the Fataluku Community Internet wordlists is mu’a tot- ina-nu (earth sever-INA-NOM), which is used to refer to the English concept ‘archipelago’. NOMINALIZATION IN FATALUKU 355

In fact, ma’ar lauh-ana (man live-ANA) is a fixed construction in the Fa- taluku lexicon that designates the concept of ‘human being’. Otherwise, constructions containing a /-(n)ana/ derivation require the 3sg possessive marker i, as was discussed in section 2. This is exemplified in (6b), where the construction ma’ar lauh-ana is extended with another /-(n)ana/ derivation meaning ‘first’.

(6) b. Ma’ar lauh-ana i h-emer-ana umani i? man live-ANA 3sg LNK-first-ANA who DEM ‘Who is the first human being?’ (Fataluku Catechism, question 32)

Campagnolo’s (1973) consistent translations of /-(n)ana/ derivations as relative clauses suggest that the /-(n)ana/ derivation, and therefore im- plicitly also the /-(’)ina/ derivation, is comparable to a participle. Like a participle, lauh-ana (live-ANA) modifies a noun, ma’ar ‘man’. Another problem is the possessive marker i. In fact, it is because of this marker that Hull (2005) analyses the /-(n)ana/ derivation as a noun referring to possession while the preceding noun refers to the possessor. This is exemplified by (7a) and (b).

(7) a. ocava i pala b. mu’a i co-nana master 3sg field earth 3sg far-ana ‘the master’s field’ ‘a far-off land’ (Hull 2005: 16) (Hull 2005: 34)

The possessive relation between the referents of both nouns in (7a) is obvious. Following the regional typology, Fataluku possessor nouns precede possession nouns. In other words, the possession noun, pala ‘field’, is the head of the construction. If (7b) were a possessive construction too, co-nana (far-ANA) ‘far-off’ would be the head also. This, however, seems counter-intuitive to us. We suggest that (7a) and (b) display the multifunctionality of i as a possessive marker and as a relative marker, respectively.34 Whereas possessive i has an emphatic counterpart hin,

34 Two other functions of i are deictic modification (as in: le i [house 3sg] ‘this house’) and marking nominal predications (as in: ... eri ta paca lafai ... hai=n=i=i. [sg.EMPH SEQ commoner big ... PERF=LNK=3sg=3sg] ‘you ... 356 AONE VAN ENGELENHOVEN relative i does not. Consequently, if i in both examples above is replaced by hin, the construal of (7b) changes into a possessive construction, as is exemplified by (7c) and (d).

(7) c. ocava hin pala d. mu’a hin co-nana master POS field earth POS far-ANA ‘the master of the field’ ‘the distance to the land’

For the reasons discussed, we propose to label the /-(n)ana/ and /-(’)ina/ derivations as adjectives. Adjectives can function as nominal heads like nouns, which explains why in the Fataluku Community’s Internet wordlists many English items for which there are no Fataluku words are translated with a /-(n)ana/ or /-(’)ina/ derivations. This is exemplified by (8a) below where the /-(n)ana/derivation functions as a noun modified by the demonstrative ia.

(8) a. [I h=iti-jet-ana] ia umani=t taratori me. 3sg LNK=deep-fall-ANA DEM who=SEQ tractor take ‘The issue (literally: “what falls down”) was who is to use the tractor.’ (Fataluku Community, nd)

Taking into account that Fataluku does not have a copula, adjectives – exemplified by kapar-ana (bad-ANA) in (8b) below – can function as predicates, like verbs.

(8) b. Kaka i sorot ara en-en i kapar-ana. older.brother 3sg book base RED-DEM 3sg bad-ANA ‘This dictionary of yours is no good’ (Fataluku Community, nd)

are the great commoner.’ (Campagnolo 1973: 208). The term ‘relative marker’ is somewhat problematic. Fataluku lacks true relative clauses in that it prefers clause combinations with the sequential marker tu SEQ as is shown in the initial part of example (6a) above: cura [tu le hi’a-lauh-e] una (mouse SEQ house up-live-VEB eat) ‘eat a mouse [that lives in the house]’. The i construction rather creates phrases and these lack the intrinsic features of a clause. The example in (6a) can not be transferred into *cura i le hi’a-lauh-ana (mouse 3sg house up-live-VEB-ANA). NOMINALIZATION IN FATALUKU 357

Interestingly, there is a general dislike in Fataluku to construct NPs with attributive adjectives. Speakers prefer full verbs over adjectival constructions, as can be observed in the following sentence, where the object noun vahin ‘tooth’ is followed by a verb (marked by /-ne/) rather than by an adjective (marked by /-nana/). Further research is required in order to determine when attributive constructions are exactly required.

(8) c. Doutor Asumoto i vahin naka-ne tu tali doctor Asumoto 3sg tooth stained-VEB SEQ very kapar-e hai n-apam-e bad-VEB PERF LNK-pull-VEB ‘The doctor pulled out Asumoto’s stained tooth that was very bad.’ 35

5. Remaining noun and adjective creating mechanisms The nominalising and adjectivising suffixes described in sections 3 and 4, respectively, and the verbal suffix /-(n)e/ are in fact the only productive derivational morphemes in Fataluku. Campagnolo (1973: 135) mentions a verbal suffix -eke, of which he acknowledges two nominal counterparts, -aka and -uka. Nacher (2002/3, 2004) furthermore mentions five cases with a possible suffix -aka. Only mimiraka ‘red(ness)’, the example Campagnolo (1973) mentions, has a verbal counterpart mimireke ‘be red’. Gomes (1972: 26, footnote 1) mentions ipi as an alternative for ipinaka.

35 Cf. Asumoto i vahin ?i naka-nana Asumoto 3sg tooth 3sg stained-ANA ‘Asumoto’s stained tooth’ The Fataluku speech community, as can be surmised through its name fata luku (correct speak) ‘speak correctly’, is obsessed with the correctness of its speech. The rejection of the adjectival construction may very well be based purely on stylistic grounds. In elicitation sessions where a certain construction is assessed by a native speaker the usual comment is: Ana nere navare po kapare : ‘I understand it, but it is ugly’. 358 AONE VAN ENGELENHOVEN

‘palm (of hand)’ alaka ‘star’ ipinaka ipi ‘star’36 ‘red(ness)’ mimiraka mimireke ‘be red’ ‘brancard’ sapelaka ‘cave’ veraka Table 8a: Unproductive nominal -aka suffix

On the same page, Campagnolo (1973: 135) mentions one example of -uka, susuka ‘fork’37 that corresponds with a verb ususeke ~ suseke ‘make into pins’. Nacher (2002/3, 2004) mentions two other cases, aruka ‘arch’ and tupuka ‘hill’. A verbal counterpart of the latter is accidentally provided by Valentim (2002: 104): tupeke ‘be mountainous’. Whereas reduplication has been reported in many languages in the region as a means to nominalise verbs into instrument nouns (van Enge- lenhoven 2004), this phenomenon has not been attested in Fataluku. On the other hand, compounding appears to be quite productive. Table 8b provides examples of noun-verb combinations, /-(n)ana/ derivations with ‘prefixed postpositions’ and compounds with alivana ‘place’, which profile the place of what the preceding verb designates.

36 Gomes (1972: 26) footnote 1. 37 This is actually the translation of the Fataluku Community’s Internet wordlists. Campagnolo describes it as a “fine pointed bamboo stalk functioning as pin”. NOMINALIZATION IN FATALUKU 359

‘orphan’ nalu-pali (mother-not.exist) ‘fisherman’ api-na-na-me (fish-RED-LOC-take) ‘sinner’ fa’a-kail-u-fa’i (very-sin-NOM-make) ‘prophet’ ina-hara-nu (eye-light-NOM)

‘example’ na-na-ce’-ana (RED-LOC-point-ANA) ‘servant’ aca-nat-ana (arrive.at-stand-ANA)

‘bed’ taia-n alivana (sleep-NOM place) ‘bin’ luratu h-alivana (garbage LNK-place)38 ‘freezer’ jelu alivana (ice place) Table 8b: Compounding

The nominalising and verbalising suffixes /-(n)u/ and /-(n)e/ are very productive and confirm the agglutinative character of Fataluku morphology. Huber (2005: 14), nevertheless mentions a bisyllabic unproductive suffix /-ini/ in Makasai, meaning “anything to do with X” in which “X” refers to the meaning of the verb, as for example: pa’ ‘sew’ → pa’-ini ‘everything to do with sewing’. She hypothesises that this suffix is related to the verb gini ‘do, make’. Unfortunately, there is no indication in Fataluku that /-(n)ana/ and /-(’)ina/ are derived from a verb. Indirect indications for their verbal origin39 are the fact that they are suffixes and because they may be input to nominalisation by the /-(n)u/ suffix (Table 7e above). The nominalisation option implies that the morpheme – whether it is complex or simple – to which the nominalising suffix is added is basically not a noun. Their bisyllabic character surmises that they are originally lexical morphemes. This is also visible in the way postpositionals that are in majority bisyllabic are docked onto verbs as a kind of applica-

38 The /h-/ attached to alivana ‘place’ suggests that this compound actually is a possessive phrase from wich the possessive marker i has been deleted : luratu i h-alivana ‘garbage’s place’. 39 An important remark here must be that there seems to be no formal indication to distinguish a special adjective class from the verbal class. Further research is required to elaborate on this phenomenon. In Indonesian where a similar situation exists, one can point out morphological – but not necessarily syntactic – features with which both word classes can be distinguished (Steinhauer 2008). 360 AONE VAN ENGELENHOVEN tive prefix (section 2). The fact that /-(’)ina/ and /(-(n)ana/ are suffixes relates to the final position of the predicate in a Fataluku clause, imply- ing that they are orginally verbs. Juliette Huber (p.c.) proposes a comparable scenario for the postverbal intention marker na’a~ana in Makalero. Hull (2004) and van Naerssen (2007) acknowledge the genetic relationship between Fataluku and the languages of the Trans-New-Guinea phylum. Its typological divergence from most other languages in this phylum is explained by Hull (2004) as an adaptation to the isolating typology of the Austronesian languages in the Timor Sprachbund. The phenomenon of the ‘anorganic’ verbal suffixes indeed resembles what has been described for Rotinese and Dawanese by Jonker (1906). Since no such phenomenon has been reported for Makalero and Makasai, one is tempted to analyse these suffixes as an Austronesian influence on Fa- taluku. The SOV word order of the non-Austronesian languages in a scenario of ‘a drift towards an agglutinative typology’ can indeed enable predicatively used verbs40 to turn into markers as in Makalero or even into a derivational suffix as in Makasai and Fataluku. However, together with the phenomenon in Fataluku and Makalero (Huber, personal communication) of initial consonant mutation in verbs, they convincingly show that these languages do not even tend towards an isolating typology as has been described for their Austronesian neighbours on Timor. The genetic and typological exclusivity of these languages, therefore, remains unexplained for the time being.

Abbreviations 1plinc first person plural inclusive NOM nominaliser suffix 3sg third person singular PERF perfect marker pl plural POS possessive marker sg singular PP postposition sth something RED reduplication DEM demonstrative SEQ seq. conj. marker LNK linker V verb LOC locative marker VEB verbaliser suffix N noun

40 Or auxiliaries, which are prototypically expected in clause final position in an SOV word order. NOMINALIZATION IN FATALUKU 361

References Campagnolo, Henri (1973). La Langue des Fataluku de Lórehe (Timor Portugais), PhD Thesis, Université René Descartes, Paris. Campagnolo, Henri [et al.] (1997). Lexique thématique plurilingue de trente-six langues et dialectes d´Asie du Sud-Est insulaire. Laboratoire Asie du Sud-Est et Monde Austronesien (LASEMA), Centre National de la Recherche Scientifique (CNRS). Paris: L´Harmattan. Donohue, Mark & Antoinette Schapper (2007). Towards a morphological history of the languages of Timor, Alor and Pantar. Paper presented at the 5th International East Nusantara Conference, Kupang: 1-3 August, 2007. Engelenhoven, Aone van (2004). Leti, a language of Southwest Maluku, VKI 211, Leiden: KITLV Press. Engelenhoven, Aone van (2006). Ita-nia nasaun oin-ida, ita-nia dalen sira oin-seluk, “Our Nation is One, Our Languages Are Different”. Language Policy in East Timor. In Paulo Castro Seixas & Aone van Engelenhoven (eds.), Diversidade Cultural na Construção da Nação e do Estado em Timor-Leste, 104-132, Porto: Publicações UFP. Engelenhoven, Aone van (2008). Ktunu: Clues in the quest of the Sailfish: linguistic insights in southwest Malukan narratives. In Y. Lander & A. Ogloblin (eds.), Lan- guage and text in the Austronesian World, 311-325, Munich: Lincom. Engelenhoven, Aone van (2009). The position of Makuva among the Austronesian languages in East Timor and Southwest Maluku. In K. A. Adelaar & A. Pawley (eds.), Austronesian historical linguistics and culture history: a festschrift for Bob Blust, 411-428, Canberra: Pacific Linguistics. Engelenhoven, Aone van (in press). Verb serialization in Fataluku, in: Frankfurter Afri- kanistische Blätter, special publication, edited by Amha Azeb and Christian Rapold. Engelenhoven, Aone van & Justino Valentim Cailoro (2006). The Makuva Enigma: Locating a Hidden Language in East-Timor. Paper presented at the Second Con- ference on Austronesian Languages and Linguistics, Oxford, June2-3, 2006. Fataluku Community (not dated). Fataluku and English wordlists, retrievable at http://www.fataluku.org. Fataluku Language Project (not dated). Fataluku Online Dictionary, retrievable at http://www.fataluku.com. Fox, James J. (2005). Ritual languages, special registers, and speech decorum in Austronesian languages. In K. A. Adelaar & P. Himmelmann (eds.), The Austro- nesian languages of Asia and Madagascar, 87-109. London: Curzon Press. Gomes, Francisco de Azevedo (1972). Os Fataluku, PhD Thesis at Universidade Técnica de Lisboa, Instituto Superior de Ciências Sociais e Política Ultramarina. Hajek, John & John Bowden (2002). A Phonological Oddity in the Austronesian Area: Ejectives in Waimoa. Oceanic Linguistics 41(1): 222-224. Himmelmann, Nikolaus P. (2005). The Austronesian languages of Asia and Mada- gascar. In K. A. Adelaar & N. P. Himmelmann (eds.), The Austronesian languages of Asia and Madagascar, 110-181. London: Curzon Press. Huber, Juliette (2005). First steps towards a grammar of Makasae, MA Thesis, Zürich University. 362 AONE VAN ENGELENHOVEN

Huber, Juliette (2008). Preliminary remarks on object-marking in Makalero. In Marjo van Koppen & Bert Botma (eds.), Linguistics in The Netherlands 2008, 87-96, Amsterdam: John Benjamins. Hull, Geoffrey (2001). A Morphological Overview of the Timoric Sprachbund. Estudos de Línguas e Culturas de Timor-Leste 4: 98-205. Hull, Geoffrey (2002). Waimaha (Waima’a), East Timor Language Profiles 2, Dili: Instituto Nacional de Linguística/Universidade Nacional de Timor-Leste. Hull, Geoffrey (2004). The Papuan Languages of Timor. Estudos de Línguas e Culturas de Timor-Leste 6: 23-99. Hull, Geoffrey (2005). Fataluku, East Timor Language Profiles 7, Dili: Instituto Naci- onal de Linguística/Universidade Nacional de Timor-Leste. Jonker, J.C.G. (1906). Over de eind-medeklinkers in het Rottineesch en het Timo- reesch. Bijdragen tot de Taal-, Land- en Volkenkunde van Nederlandsch-Indië (zevende volgreeks, vijfde deel): 263-343. Josselin de Jong, J.P.B. de (1937). Studies in Indonesian Culture I: Oirata a Timorese Settlement on Kisar. Amsterdam: Noord-Hollandsche Uitgeversmaatschappij. Klamer, Marian (2002). Typical Features of Austronesian Languages in Central/Eastern Indonesia. Oceanic Linguistics 41(2): 363-383. Klamer, Marian (2004). East Nusantara: Genetic, Areal, and Typological Approaches. Oceanic Linguistics 43(1): 240-244. McWilliam, Andrew (2007). Austronesians in linguistic disguise: Fataluku cultural fusion in East Timor. Journal of Southeast Asian Studies 38(2): 355-375. Mills, Roger F. & John Grima (1980). Historical developments in Lettinese. In Paz Buenaventura Naylor (ed.), Austronesian Studies. Papers from the second eastern conference on Austronesian languages, Michigan Papers on South and Southeast Asia No. 15, 273-283. Ann Arbor: The University of Michigan. Naerssen, Maaike van (2007). Klankverschuivingen van drie non-Austronesische talen op Oost-Timor: Fataluku, Oirata en Makasai. Unpublished manuscript. Nacher, Alfonso Maria, NSB (2002/3). Léxico Fataluco-Português (primeira parte). Estudos de Línguas e Culturas de Timor-Leste 5: 135-196. Nacher, Alfonso Maria, NSB (2004). Léxico Fataluco-Português (segunda parte). Estu- dos de Línguas e Culturas de Timor-Leste 6: 119-177. Payne, Thomas (2006). Describing Morphosyntax, a guide for field linguists. Cam- bridge: Cambridge University Press. Stoel, Ruben (2007). The prosody of Fataluku. Paper presented at the Workshop on the Languages of Papua, August 8-10 Universitas Negeri Papua, Manokwari. Steinhauer, Hein (2008). Does Indonesian have adjectives? Keynote paper presented at the 12th International Symposium on Malay/Indonesian Linguistics, Leiden, 26- 27 June, 2008. Stresemann, Erwin (1927). Die Lauterscheinungen in den Ambonischen Sprachen. Ber- lin: Dietrich Reimer. Valentim, Justino (2002). Fata-lukunu i disionariu. Dili: Timor Lorosa’e – Nippon Cul- ture Center. Index

Aari 191 affirmative interrogative 195, 209 Abbott, M. 95fn1, 96, 115, 118, 119, African languages 244, 272 121 Afroasiatic 200 ablative 178fn11, 200 agentivo 18, 20ff, 32, 43, 51fn23, abstract nouns, see nouns 53fn25 acción nominalizada 17, 20, 22, 40, agglutinative languages 189, 189fn2, 41 340, 359, 360 acusativo 67, 85, 88, 89 agreeing modifier 251 Adams, D. 321 agreement 9, 10, 200, 202, 203, 204, additive nouns, see nouns 207, 211, 229, 256ff, 268, 272ff Adelaar, W. 3, 7, 171ff agreement features 256ff Adelaar, W. & S. van de Kerke agreement markers 208, 211, 212, 177fn10 214, 215, 258ff, 271, 273ff Adelaar, W. with P. Muysken 172fn2, agreement patterns 243ff 178fn12 Aikhenvald, A. 79fn18, 177, 252 adjectival 5, 96, 100, 116, 119ff, 128, Akan 244, 286, 288 226, 290, 314ff, 323, 327, 348, Akawaio 95fn1, 109ff, 126 350f, 357, 357fn35 alativo 21, 51ff adjective 5, 10, 95ff, 223, 226, 227, Algonquian 179 243, 257, 258, 264, 271, 284, 289, Allan, E. 233 290f, 293, 298f, 301, 305, 311, allomorph 190, 225, 340, 346, 314, 316, 317, 348, 351, 356, 357, 346fn23, 348, 350 359fn39 Amazonia 5, 16fn3, 42fn17, 54fn26, adjetivo 15, 17, 21, 27fn10, 41, 66f 63, 79fn18, 83fn22, 90, 144, 145, adjetivo verbal 21, 27 147 adjunct 116, 123, 123fn8, 124, 247, Ameka, F. 235, 238fn3, 244, 264 248, 271 Ameka, F., C. de Witte & D. Wilkins adverbs 5, 11, 95ff, 136, 137, 142, 268 143, 165, 167, 174, 201fn3, 217, Amha, A. 8, 189, 193, 199ff, 229 225, 315, 345ff Amodio, E. & V. Pira 95fn1 ‘adjectival’ adverbs 125 Andean languages 7, 171ff adverbial complements 109, 110, Andrews, A. 320 111, 113 anorganic consonants 339 subclass (of adverbs) 5, 103, 105, anorganic suffix 350fn28, 360 119, 121, 122ff antropónimos 68 adversativo 22, 41 Apalaí 96, 107, 119 affirmative 8, 144, 190, 193ff, 202, Arabic 10, 313 204ff, 300 Araucanian 178fn12, 182 364 INDEX

Huilliche 179, 182, 182fn17 bound morpheme 189fn2, 200, 201, Arawak 5, 68, 177 206, 217, 218 Arawakan 135, 149 Bowerman, M. & E. Pederson 268 argumento nuclear 16 Bresnan, J. 32fn12 argumento único 17fn4, 18, 19, 26ff, Bresnan, J. y J. Mugane 33fn13 58 Bunak 336 Arnold, J. 172, 179 Bybee, J. & P. Hopper 215 Asli, G. & C. Kerslake 189 aspect 7, 8, 97, 101, 149, 168, 189ff, Cabrera, G., C. Franky & D. Mahecha 200, 204ff, 223, 247ff, 331 64fn3, 66fn7 aspecto nominal 41, 45 Cacua 64, 64fn4, 65, 88, 89 assertion 142, 216, 325 caducidad 22, 44 asymmetry 126, 222, 232, 240 Campagnolo, H. 335fn4, 337, 339, atributiva 17ff, 38, 42, 43, 57, 58 339fn9, 340, 343fn15, 344, 345, attributive 9, 108, 109, 113ff, 120, 346fn24, 349, 351, 352, 354, 355, 126f, 131, 193, 221, 223ff, 239, 356fn34, 357, 358, 358fn37 289, 291ff, 316, 317, 320, 345, 357 Cañaris 176 Augusta, F. 182 cardinal numbers 243, 257f, 291, 345 aumentativos 68, 71 Carib 118, 121 Austronesian 335, 336, 337, 339, 342, pre-proto-Carib 126fn9, 131 344, 345, 350fn28, 353, 360 Cariban 5, 6, 95ff, 135, 135fn1, 137, non-Austronesian 11, 333ff 138fn2, 141, 142, 147, 149 Proto-Austronesian 344fn19 Carlin, E. 5, 6, 95fn1, 96, 103, 119, Avatime 264, 267 135ff Aymara 177, 184 casos locativos 41, 50, 55 Aymaran languages 173, 184 Cassirer, E. 146 categorías mixtas 32fn12 Bagbɔ 267 causative 168, 206 Baker, M. 102fn3 Cerrón-Palomino, R. 172fn2 Bantu languages 243, 267 Chadic 7, 200 Bara 4, 64 Chappell, H. & W. McGregor 233, basic constituent order 247f 235, 240 Beachy, M. 235 Chinese 10, 325, 327 Belle, W. van & W. van Langendonck clasificadores 63, 79fn18, 82, 83, 230 83fn22, 84, 88, 89 Bench 227, 229 class marker 257, 305 Bender, M. 191, 222, 239 classification nominale 163, 165fn7 Berber 7, 200 Claudi, U. & F. Serzisko 222, 231fn1, Bertho, R. 251 233, 235, 240 bivalent action verbs, see verbs clausal negation 8, 201, 218 Blench, R. 244, 244fn1, 267 clause-combining 337 body part nouns, see nouns cláusula matriz 20, 26, 51, 58 Boons, J.-P. 164 cláusulas relativas 16, 27ff borrowed nouns, see nouns cláusulas subordinadas 47, 55 INDEX 365 clichés 336 definiteness 222, 223, 226, 228, 229, Collberg, S. & G. Hakansson 189 231, 232, 239, 240, 271, 274fn7, comitativo 37, 67, 85, 86, 89 295, 296fn18 comparative construction 116, 121 DeLancey, S. 27fn11 complement clauses 181, 225, 330 demonstrative 9, 223, 229, 243, 256, complemento 16, 18, 21, 22, 26, 28, 257, 259, 260, 261, 271, 283, 289, 35, 44, 46, 47, 56, 57, 58 295, 296, 314, 325, 356 complex predicates 8, 202, 204ff, 218 demonstrative pronouns 136, 296f, compounded plurals 252 315 compounding 358f dependent clauses 8, 201, 216ff, 223 Comrie, B. 320 depictive 140, 142 Comrie, B. & S. Thompson 15fn1, Derbyshire, D. 95fn1, 96, 99fn2, 118, 312 119ff concord 9, 10, 17fn4, 47, 74, 243, Derbyshire, D. y D. Payne 83fn22 244, 261ff, 268, 272, 289, 290ff derivación agentiva 21 condicional 22, 55, 57f derivativos 16, 41 conditional mood 180 descriptive nouns, see nouns consonant mutation 337ff, 360 determiner 4, 257, 258, 260, 289, 325 construcción clivada 24 Deutscher, G. 311 construcción finita 47 Dime 7, 8, 189ff, 227 construcción no finita 47 diminutivos 68, 71 construcciones nominales 20 direct voice 179 controller 256, 259 Dixon, R. 5, 95, 96, 118, 121, 126, copula 8, 99, 100f, 109ff, 128, 130, 235, 254 190, 192, 193, 194, 210, 216, 232f, Dizi 8, 9, 222, 231fn1, 233, 234, 235, 356 240 cópula 41 Donohue, M. & A. Schapper 333 Corbett, G. 256, 305 Dorvlo, K. 9, 243ff, 272 co-referencial 24, 27, 31, 32, 51 double class morphemes 343, 345 count nouns, see nouns double negation 189 Courtz, H. 118, 120 Drude, S. & S. Meira 131 Croft, W. 96, 130fn10 Dryer, M. 190, 311ff, 320 cualificativos 33, 34, 45 dual 178, 179, 181, 182, 256, 315 Cubeo 71, 71fn11, 88, 88fn27 Cushitic 7, 199fn1, 200 Egyptian 7, 200 emphasis 142, 237, 238, 240 dativo 17, 67, 85, 86, 88, 89 emphatic 135, 137ff, 142ff, 148ff, dativo existencial 22, 51 297fn20, 298, 355 Dâw 4, 65, 65fn5, 69, 87, 88, 89, 90 emphatic denial 216 Dawanese 360 empty pronominal head 323, 329 declarative 8, 177, 178, 192, 195ff, Engelenhoven, A. van 11, 333ff 201, 203ff, 218, 223 Engelenhoven, A. van & J. Valentim 333f 366 INDEX

English 9, 102fn3, 121, 123, 125, 128, Funke, E. 268, 272 129, 130, 202, 247, 268, 286, 316, 324, 325, 354fn33, 356 Gamo 199, 200 entrenchment 135, 147, 148, 149 Ganta 200 Ephrem, D. 197 Garbantsa 200 Epps, P. 65, 65fn5, 73fn14, 79fn18, Gbe 244 79fn19, 83fn21, 84, 87, 88, 89, 90 gender 4, 9, 194, 207, 221ff, 226ff, equative 109, 193f 239, 255, 314, 318 ergativo 18, 20, 24, 48 generó 67, 89 Essegbey, J. 272 genitive 11, 200, 223, 235, 314, 316ff, eventivo 21, 41, 46 324ff, 330 evidential 6, 136, 137, 149f genitivo 20, 67, 85, 89 Ewe 9, 235, 238fn3, 244, 250ff, 267, German 102fn3, 322 268, 268fn3, 275, 276, 286ff, 292 Germanic languages 152 existential 8f, 109, 193, 203, 222, 229, Ghana Togo Mountain (GTM) lan- 230f, 239f guages 9, 243f, 264, 267f existential verb, see verbs Central-Togo languages 267 existential-predicative (auxiliary) Na and Ka-subgroups 243f, 264, verbs, see verbs 267 Extra-Ramelaic 333 Nyo branch 244 Togo Remnant Languages, facsimile 6, 137ff, 149 Togorestsprachen 243, 267 Fataluku 11, 333ff Gibbs, R., D. Beitel, M. Harrington & Fataluku dialects 334ff P. Sanders 160 figure 165, 167, 168 Gildea, S. 5, 95ff finalidad 22, 41, 48, 56, 57 Girón, J. 3, 15ff, 65, 73fn14, 87, 88 Fleming, H. 200, 222, 229 Girón, J. & W.L. Wetzels 16fn2 focalisation 248 Givon, T. 67, 215 focalización 18 Gomes, F. 335, 357, 358fn36 focus 178, 200, 203, 204, 210, 211, Gomez-Imbert, E. 84fn24, 88 218, 248 Goodman, N. 146, 147 fonológico 66, 66fn7 Gordon, R. 267 Ford, K. 268 grammaticalization 140, 221, 227, 325 fourth person 173 Greek, Ancient 321 Fox, D. 95fn1 Greenberg, J. 243, 267 Fox, J. 335 frase verbal finita 33 habitual aspect 204, 247 Frawley, W. 67 Haegeman, L. 102fn3 free pronoun 273, 298, 304 Hajek, J. & J. Bowden 336fn6 free relatives 320 Hamar 189 Freeman, D. 199 Hardman, M. 137 French 121, 140, 345fn21, 351 Haro 193, 200, 215 Canadian French 215 Haspelmath, M. 33fn13 frustrative 6, 135, 137ff, 144ff Hawkins, R. 96, 119 INDEX 367

Hayward, R. 215 intensifier 138, 143, 243, 257, 260, head noun, see nouns 264, 289 headless relative clauses, see relative intensity 142 clauses interdental 180fn15 Heine, B. 221, 231, 232, 244, 264, interrogative 8, 9, 165, 195ff, 201, 267, 272, 286, 305 203, 207ff, 218, 260, 325 Hellenthal, A.C. 8f, 221ff interrogative qualifier 289, 298f, 301 Hengeveld, K. 1f interrogative quantifier 297f Herslund, M. & I. Baron 221, 222, inverse 7, 171ff 230, 231, 240 inverse markers 171ff hierarchy of grammatical persons 172ff Japanese 10, 189, 325, 327 Himmelmann, N. 342fn13 Old Japanese 321 Hixkaryana 95ff Jaqaru 184 Hoff, B. 118, 120, 121, 127 Jê 131 Hopper,P. y S. Thompson 67 Jonker, J. 350fn28, 360 hortative-imperative 180, 181 Josselin de Jong, J. 336fn5 Hough, K. 145fn3 Huang, C. 320, 321 Kari’nya 127 Huber, J. 333, 336, 359, 360 Karui-Waimaha-Midiki-Naueti 333 Hull, G. 333, 334fn3, 336, 336fn6, Katamba, F. 73fn13 340fn11, 341, 344, 344fn18, Keenan, E. 315, 320 344fn19, 345, 349, 352, 354, 355, Keenan, E. & B. Comrie 320, 321 360 Kemak 336 Hungarian 189fn2 Koehn, E. & S. Koehn 96, 119 Hup 4, 65, 65fn5, 73fn14, 83fn21, 84, Koorete 200, 215 87, 88, 89, 90 Koptjevskaja-Tamm, M. 16fn3 Kropp-Dakubu, M. & K. Ford 267 imperative 97, 176, 178, 180, 181, Kwa 243f, 267 193, 197, 223 Proto-Kwa 244 imperative/optative mood 8, 201, 208, 211ff Lahu 10, 323 imperativos negativos 50 Lakoff, G. 254 imperfective 190, 191, 194, 197, 204, Lamptey, M. 10, 267ff 206, 216 Langacker, R. 67 Incahuasi 176 language concealment 333 inclusive 173ff, 212, 337 language contact 149, 335 inclusive person 173ff language shift 335 indefiniteness marker 295 Lehmann, C. 320 individuated nouns, see nouns Lemmens, M. 152, 152fn3 Indonesian 344, 344fn18, 348fn26, Leti 353 359fn39 Levinson, S. 165 infinitivo 46 lexical parallelism 335 lexical stem 245 368 INDEX lexicalización 41 Mayer, M. 268 lexicalization 125, 130 McWilliam, A. 334 Likpe 244, 264 Meher 353 liquid nouns, see nouns Meira, S. 5, 23fn8, 95ff Lithuanian 189 Mendisu, B. 215 loanwords 271, 275f, 286ff, 305 Miestamo, M. 203 locative 3, 4, 109, 116, 165, 167, Mills, R. & J. Grima 341 193f, 222, 229, 231, 233, 235, 247, Miya 312 271, 315 modal 7, 18, 24, 31, 40, 45, 49, 201, locative constructions 127, 247, 271 206, 213, 223, 241 locativo 21, 35, 41, 50, 53, 55,67, 76, modalité 160, 163, 169 77, 85, 89 Modern South Arabian languages 10, Logba 9, 10, 243ff, 267, 272 313 Peki dialect 255 modification 108ff, 286, 312, 314, Luangic-Kisaric 333 318, 323, 326, 327, 331 Lydall, J. 189 modifier 9, 10,11,101, 108,114, 116, 120, 122, 124, 127, 131, 223, 224, Maale 189, 193, 197, 216, 229 251, 258, 271, 272, 289, 290, Macuna 71, 71fn11 292ff, 305, 311, 314, 316, 318ff, Mahecha, D. 4, 63ff 346, 348, 354 main clause 8, 201, 202, 206ff, 218, módulo verbal 16, 18, 20, 24, 37, 38, 223 40, 46, 51, 56 Makalero 333, 335, 336, 340, monovalent action verbs 352 344fn19, 349, 360 mood 8, 97, 173ff, 200f, 206, 209, Makasai 333, 335, 336, 340, 344fn19, 213, 215, 218, 223 349, 359, 360 morfema derivativo 16 Makhashen, K.A. 10, 311ff morfología 15, 17, 66, 67, 82, 89, 90 Makú Occidental 65fn5 morphological concords 9, 244, 272 Makú-Puinave 3, 15fn2, 63, 64, 65, morphological reduction 203 84, 87, 88, 89 morphologization 201 Makushi 95ff morphology 7, 10, 33fn13, 97ff, 126, Makuva 333, 335, 336fn5 131, 135, 139ff, 149, 171, 183, Malay 344fn18 184, 189, 190, 223, 229, 330, 333, Mäle 200 337, 340, 359 Mapuche 7, 171ff Morse, N., J. Salser & N. Salser Mapudungun 178fn12 71fn11 marcadores de persona 18, 30, 40, 51 Mulugeta, S. 8, 189ff, 227 Martins, S. 65, 87, 88, 89 Martins, V. 65, 65fn5 Nkak 3, 4, 63ff Martins, V. & W.L. Wetzels 65fn5 Nacher, A. 340, 343fn17, 345, 349, Mason, A. 64 350fn28, 352fn32, 357, 358 mass nouns, see nouns Nadahup 65fn5 Matisoff, J. 323, 327 Nadeb 4, 65, 65fn5 Mawayana 149, 150 Naerssen, M. van 333, 360 INDEX 369

Naini 333 nominal root 126f, 273, 300, 350 nasal prefix 244, 248, 249, 250, nominales 15ff 261fn6, 263, 264 nominalización atributiva 26, 31 nasalidad 66 nominalización de acción 53, 58 native speakers 251, 254, 281, nominalización de argumento 22, 36 357fn35 nominalización de habituativo 34 Naumkin, V. 313 nominalización de instrumento 37 Nayi 8, 197 nominalización de la negación verbal néerlandais 160fn5, 168 33 negation 8, 103fn4, 144, 189ff, 199ff, nominalización de lugar 35 346fn23 nominalización locativa 35, 43, 51 negative copula 8, 193 nominalizador 17, 20, 21, 22, 24, 35, negative declarative 192, 204, 205, 37, 38, 41, 42, 51, 59, 63, 72 206, 207, 213, 215, 216, 219 nominalization 3f,10f,101,105, 107, negative imperatives 8, 189, 192, 197, 110, 118, 125, 180, 181, 226, 227, 211, 213, 214, 223 228, 285, 311ff, 340ff negative interrogative 194, 195, 197, agent/subject nominalization 328f, 203, 207ff 330 negative marker 8, 189, 191, 193, 194, agentive/instrumental nominaliza- 196, 197, 201, 205, 213, 214, 216, tion 330 248f argument nominalization 10f, 313, negative morphemes 8, 189, 190, 197, 318, 320, 324f, 328f 205 derivational nominalization 330 negative nominal clause 8, 193f event nominalization 10, 313, 324 negative verbs 8, 191, 193, 201, 205ff, genitive nominalization 11, 324, 211ff 326f, 329f Niger-Congo 9, 243f, 267 grammatical nominalization 311f, Nisa 333, 334 313, 324, 328ff Nkore-Kiga 311 lexical nominalization 311, 324, nombre 4, 15, 15fn1, 20ff, 33, 35, 38, 328 40, 41, 43, 46, 63ff patient/object nominalization 329 nombre de acción 38 patientive nominalization 330 nombres compuestos 73, 74, 78, 80 nominalized relative clause, see rela- nombres de época 70f tive clauses nominal 4, 6, 8, 15ff, 63, 67, 74, 79, non-Austronesian, see Austronesian 83ff, 89f, 100, 102, 105fn5, 109, non-count nouns, see nouns 110, 113, 114, 126, 127, 128, 136, non-speech act participant 178 137, 139, 140, 141, 145, 148, 149f, nonverbal predicates 108, 109ff, 127, 173, 177, 200, 264, 277, 290, 298, 128, 131 299, 300, 311, 312, 320ff, 341ff Noonan, M. 16fn3, 27fn11, 323 nominal class 137, 244, 268, 286 Nootka 125, 130 nominal clauses 8, 190, 192ff noun class system, noun classes 9, 10, nominal modification 114, 128 243ff, 267ff, 311, 324 nominal modifier 271, 325 370 INDEX

semantic basis (of noun classes) numbers, numerals 9, 115, 243, 257ff, 249, 252 263, 271, 289, 291ff, 304, 314, noun classification 243 344f noun modification 314ff, 323, 331 número 17, 29, 30, 33, 41, 55fn27, 67, Noun Phrase (NP) 9, 103, 108, 114, 89 116, 120, 131, 221ff, 228, 230, Nyagbo 267, 272 231, 234ff, 243, 248, 257ff, 263f, 271ff, 289, 300, 305, 311, 313, oblicuo 22, 25, 26, 51, 56 326, 327, 329, 348, 348fn26 oblique case 180 NP subject 243 obviative 179ff noun phrase structure 311 Oirata 333, 336fn5, 344fn19 noun prefixes 246, 251, 256, 261, 263, Oliveira, C. 131 284, 289, 292, 300 Omotic 7, 8, 190fn3, 191, 193, 197, nouns 199, 200, 215, 216, 218, 221, 222, abstract nouns 281, 340 227, 229, 239, 241 additive nouns 340, 343 East Ometo 200, 215 body part nouns 9, 222, 231, North Ometo 8, 200, 215 233ff, 240, 253, 255, 263, 279, Proto-Omotic 229, 239 281, 282, 283 South Ometo 190fn3, 216 borrowed nouns 10, 273, 275, 286, West Ometo 215 287, 305 optative 8, 193, 201, 208, 211ff, 218, count nouns 259, 260, 293 223 descriptive nouns 96, 121f orden rígido AV 18 head noun 9f, 221, 223, 224, 243, ordinals 293, 298, 299, 344f 257ff, 268, 271, 289ff, 312, 314, Oré, L. 176 318 orthography 177, 247, 267fn1, individuated nouns 251 269fn4, 269fn5, 283fn10, 285fn11 liquid nouns 248f, 259, 260, 263 Ortiz, S. 64 mass nouns 243, 244, 248ff, 259ff, Ospina, A.M. 65, 68fn10, 69, 71, 263, 343 73fn13, 79fn19, 83fn21, 84, 87, 88, non-count nouns 250, 255, 261, 89 272, 283, 284, 288 Overing, J. 145f plural nouns 249ff, 259, 261, 263, 277, 281 palabra verbal eventiva 48f possessed noun 257, 304 partículas adverbiales 17, 22, 41, 50, product nouns 342 55, 59 zero prefix nouns 249 pasado nominal 22, 26fn9, 32, 41, 44, number 4, 97, 98, 99, 101, 115, 173, 45 173fn4, 178, 179, 181, 182, 194, past progressive 203, 207, 247, 271 207, 244, 259, 263, 315, 318 patron prosódico 66, 72, 73fn14 number agreement 258 Payne, D. 114 number differentiation 244, 263 Payne, J. 189 Payne, T. 74fn15, 109, 196, 229, 234, 342 INDEX 371 perfective 190, 191, 194, 197, 216, prédication de localisation 167 328 prédication locative 165, 167 person 7, 97, 98, 101, 104, 136, 171ff, prefixes 9, 10, 97, 98, 103, 104, 105, 189, 193ff, 207, 211ff, 223, 337 107fn6, 136, 200, 223, 237, 241, personal pronoun 177, 304 243ff, 268ff, 317, 318, 336, 337, personal reference 7, 171ff 338, 340, 348fn26, 358, 360 marking 7, 172ff presupposition 324f, 326 Philippine languages 342fn13 product nouns, see nouns plural morpheme 250 production verbs 342, 353 plural nouns, see nouns pronominal forms 10, 272ff, 300, 324 polar interrogatives 190, 197, 207 pronouns 136, 142, 177, 178, polar negative 194, 195 180fn15, 181, 203, 212, 214, 215, Poplack, S. 215 261, 270, 271, 273, 274, 279fn9, portmanteau 174, 189, 197 282, 290, 291, 296ff, 301ff, 315, Portuguese 11, 343fn14 317, 323, 325, 337, 344, 348 posesión 85, 89 properties 96, 97ff possessed noun, see nouns property concepts 5, 95ff possession property nouns 100, 114, 118, 126, attributive possession 9, 221, 127, 130 223ff, 239 prothesis 338f inalienable possession 9, 221, 222, proximate 179ff 233, 235, 240 puinave 3, 15ff, 65, 65fn5 possessed pronouns 304f Puquina 7, 171ff possessor ascension 9, 221, 222, purposive adverbial clauses 217 231, 233ff Pustet, R. 109 possessor promotion 233, 235, 240 predicative possession 8, 9, 221, Qiang 321 222, 231ff quantifier 128, 201fn3, 257, 261, 289, possessive 9, 98, 99, 100, 104, 130, 291, 292, 295, 297ff 140,145, 173,193, 194,221,222,, Quechua 7, 171ff, 321 224, 225ff, 239, 257, 271, 304ff, Ayacucho Quechua 173fn5 317ff, 325, 337, 344, 345, 348, Argentinean Quechua 7, 172fn2, 355, 356, 359fn38 173fn4 possessive constructions 221, 225, Central Peruvian Quechua 7, 172, 233, 234, 235, 271, 273, 348, 355f 172fn2 possessor 8f, 98, 99, 140, 145, 173, Ecuadorean Quechua 7, 172fn2 221, 222, 223, 230ff, 257, 311, Northern Junín Quechua 173fn5 314, 337, 348, 355 Pacaraos 173, 173fn3, 174 post-verbal object 247 Proto-Quechua 172, 174fn7 precategorial 342, 342fn13 Quechua I 174 predicado de cláusulas intransitivas 29 Quechua II 174, 174fn7, 175fn9 predicate types 8, 201, 202, 205, 218 Southern Bolivian Quechua 7, predication 7, 8, 108, 109, 115, 127, 172fn2, 173fn4 128, 142, 222, 232, 239, 355fn34 Queixalós, F. 6, 131, 151ff 372 INDEX

Rapold, C. 227, 229 Smothermon,J. y J. Smothermon Raposo, C. 95fn1 71fn11 reconstruction 125, 172 Sneddon, J. 321, 321fn1 reduplication 107, 107fn6, 358 Soqotri 19f, 311ff relative clauses 4, 10f, 181, 210, 211, Spanish 3, 171fn1, 322 216, 223, 311, 312, 314ff, 331, state verbs 11, 345, 346, 352 336, 351, 355, 356fn34 Steinhauer, H. 359fn39 headless relative clauses 10, 311, Stoel, R. 346fn24, 347fn25 312, 316, 319, 320ff Struck, B. 243, 267 nominalized relative clauses 210, subject 7, 9, 10, 99, 100, 104, 109, 211, 216 130fn10, 171ff, 194, 195, 200, relative pronouns 315 202ff, 207ff, 222, 230ff, 236, 243, relativization 312, 314, 321f, 325, 246ff, 261ff, 270ff, 297, 299, 330, 331 301ff, 305, 320, 328, 329, 330, 337 resultativo 22, 37, 41, 43 subject agreement 202ff, 262f Ring, J. 243, 267 subject cross reference 273 Rivet, P. y C. Tastevin 64 subject pronouns 212, 215, 270, Rivière, P. 145 273, 279fn9, 282, 290f, 297, 299, Rodrigues, A. 131 301ff Rotinese 360 subject-verb agreement 10, 271, Rusenu 333, 334 272, 275, 277ff Russian 322 subordinación 51 subordinadas 16, 22, 26, 41, 43, 47, Sa’ani 333 48, 50ff Salas, A. 172, 178fn12, 179, 182 subordination 311 Sataré-Mawé 131 suffixing morphology 171, 176, 217, Schachter, P. & T. Shopen 122 251 Schadeberg, T. 347fn25 suppletive forms 202, 252 Schuh, R. 244, 264, 272, 289, 312 Swahili 189fn2 Semitic languages 7, 200, 313 syllabic nasals 246, 249, 270 serial verb constructions 11, 337, 347 symbol 141, 146, 147, 149 Sheko 7, 8, 9, 221ff symbolization 147 Shibatani, M. 10, 311ff sikuani 6, 151ff Tafi 9, 10, 267ff Simeone-Senelle, M. 313, 314, 319, target 230, 256, 259, 264 325 Tatuyo 84fn24 similative 6, 135, 137, 141 Tavares, P. 96, 119 singular 9, 173, 178ff, 207, 223, 228, tense 8, 136, 137, 173ff, 191, 194, 243, 244, 249ff, 268, 272, 275, 197, 201ff, 213, 217, 218 277, 278, 281, 284ff, 305, 312, tense-aspect 8, 97,101,191,200, 204, 315, 318, 337, 349 206, 209, 211, 218, 331 sintaxis 18, 33, 89, 90 Tetum 11, 337, 343fn16, 344 site 165, 167, 168 Tɩgbɔ 267, 268fn3 Smeets, I. 178fn12, 182fn16 Timor Sprachbund 336, 337, 340, 360 INDEX 373

Timor-Alor-Pantar branch 11, 333 verbo intransitivo estativo 23fn7, Tiriyó / Trio 23fn8, 95fn1, 96, 98ff, 27 141 verbs 4, 6, 7, 8, 10, 11, 95, 96, 97ff, tone 200, 200fn2, 222ff, 227, 246f, 99, 101, 102, 107, 118, 128, 130, 249, 270f, 274fn7, 291fn13, 297, 130fn10, 131, 136, 141, 143, 171, 297fn20, 298, 300, 304 172, 173, 174, 177, 178, 180, 181, tonal patterns 246 191ff, 200ff, 223, 226, 258, 268, topicalizado 25, 31 270, 273, 274, 278, 283, 284, 285, topicalization 248 311, 317, 330, 337ff, 358, 359, 360 Torero, F. 172, 176fn10 bivalent action verbs 353 transitional endings 171, 173ff existential verb 191, 202, 203, Trans-New-Guinea phylum 11, 333, 205, 206, 216 336, 360 verbs referring to sound production Tucano 71, 84fn24, 89, 90 353 TuKaJê 131 vowel harmony 269 Tupían 5, 131 ATR harmony 269 Tupí-Guaranían 131 root controlled 269 Turkish 189 vowel system 245, 269 typology 5, 8, 95, 96, 102fn3, 126, 127, 128, 130fn10, 135, 136, 171, Waiwai 96, 119 190, 197, 201, 214, 221, 222, 235, Wánsöhöt / wãnsöjöt 3, 4, 15ff, 65, 239, 245, 252, 268, 320, 321, 331, 73fn14, 87, 88 336, 340, 355, 360 Wayana 5f, 107, 119, 135ff typological convergence 171 Weber, J. 321 typological tendencies 201, 214 Westermann, D. 251, 264 Westermann, D. & M. Bryan 243, 267 Valdivia, L. 182 Wetzels, W.L. 3ff Valentim, J. 334, 334fn3, 336fn5, Wittgenstein, L. 147 340, 342fn12, 343fn15, 345, 348, Woldemariam, H. 193, 215 348fn27, 349, 352, 352fn32, 354, word classes 10, 33fn13, 95, 96, 97ff, 358 122, 124, 125, 128, 130, 136, 137, Van Valin Jr., R. & R. LaPolla 23fn7, 342, 351, 359fn39 27fn10 Wrona, J. 321, 323 verbal inflection 189, 193, 211, 215 verbalisation 340ff, 350 Yimam, B. 200 verbes de posture 151, 160, 161, 163, Ylikosky, J. 46fn19 165, 168 Yujup 4, 65, 65fn5, 69, 73fn14, verbes distensifs 168 79fn19, 83fn21, 84, 87, 89 verbos 17, 18, 20fn5, 23fn6, 23fn7, 23fn8, 27fn10, 33fn14, 35, 27fn15, Zargulla 7, 8, 199ff 38, 39, 40, 44, 50, 51, 67, 73, 77 Zayse 200, 215 verbo intransitivo activo 27 zero prefix nouns, see nouns Zuñiga, F. 85fn26, 88, 90