The Expression of Possession
≥
The Expression of Cognitive Categories ECC 2
Editors
Wolfgang Klein Stephen Levinson
Mouton de Gruyter Berlin · New York
The Expression of Possession
edited by
William B. McGregor
Mouton de Gruyter Berlin · New York
Mouton de Gruyter (formerly Mouton, The Hague) is a Division of Walter de Gruyter GmbH & Co. KG, Berlin.
앝 Printed on acid-free paper which falls within the guidelines 앪 of the ANSI to ensure permanence and durability.
Library of Congress Cataloging-in-Publication Data The expression of possession / edited by William B. McGregor. p. cm. ⫺ (The expression of cognitive categories ; 2) Includes bibliographical references and index. ISBN 978-3-11-018437-2 (hardcover : alk. paper) 1. Grammar, Comparative and general ⫺ Possessives. I. McGregor, William, 1952⫺ P299.P67E925 2009 415⫺dc22 2009038372
Bibliographic information published by the Deutsche Nationalbibliothek The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data are available in the Internet at http://dnb.d-nb.de.
ISBN 978-3-11-018437-2 hb ISBN 978-3-11-018438-9 pb 쑔 Copyright 2009 by Walter de Gruyter GmbH & Co. KG, D-10785 Berlin. All rights reserved, including those of translation into foreign languages. No part of this book may be reproduced in any form or by any means, electronic or mechanical, including photocopy, recording, or any information storage and retrieval system, without permission in writing from the publisher. Cover design: Frank Benno Junghanns, Berlin. Printed in Germany.
Contents
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 William B. McGregor English possessives as reference-point constructions and their function in the discourse . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13 Peter Willemse, Kristin Davidse and Liesbet Heyvaert On the co-variation between form and function of adnominal possessive modifiers in Dutch and English . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51 Jan Rijkhoff Is possession mere location? Contrary evidence from Maa . . . . . . . . . . . . 107 Doris Payne Learning to encode possession . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143 Sonja Eisenbeiß, Ayumi Matsuo and Ingrid Sonnenstuhl Plain vs. situated possession in Czech: A constructional account . . . . . . 213 Mirjam Fried Attributive possessive constructions in Oceanic . . . . . . . . . . . . . . . . . . . . . . 249 Frantisek Lichtenberk Possessive clauses in East Nusantara, the case of Tidore . . . . . . . . . . . . . . 293 Miriam van Staden Possessive expressions in the Southwestern Amazon . . . . . . . . . . . . . . . . . 343 Hein van der Voort Possession in the visual-gestural modality: How possession is expressed in British Sign Language . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 389 Kearsy Cormier and Jordan Fenlon Contributors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 423 Index of subjects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 425 Index of languages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 430 Index of persons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 432
Introduction William B. McGregor
The nine papers constituting The Expression of Possession deal with a range of issues concerning the expression of the relation of possession in human languages. It is not intended to present a comprehensive overview of what is known about the topic, but rather to provide some flavour of what is interesting descriptively and theoretically about the expression of possession, to present fresh perspectives on this well-researched topic, and to suggest viable prospects for future research. Thus most of the papers show a strong descriptive orientation, and present detailed and nuanced accounts of possessive constructions in particular languages or areally/genetically constituted groups of languages. Many of them also address issues of current relevance, and/or question widely presumed knowledge. If we are to investigate the ways of expressing possession we are immediately confronted with the question: What is the possessive relation? A good deal has been written on this topic – see for instance, Seiler (1983); Taylor (1989: 202–203); Tsunoda (1995); Heine (1997: 3–6, 33–41); Herslund and Baron (2001: 1–4). For present purposes it is sufficient to say that it is a relational concept that potentially covers a wide range of conceptual relations between entities, including, for human beings, between persons and their body-parts and products, between persons and their kin, between persons and their representations (e.g. names, photographs), between persons and their material belongings (animate and inanimate items they own), between persons and things that they have usership-rights to or control over, between persons and cultural and intellectual products, and so on. For other animates and inanimates a more restricted range of conceptual relations is generally available. Most linguists – including the present author and the contributors to this volume – would probably agree that the definition should be couched in terms of linguistic factors, rather than purely conceptually. If this approach is adopted, it becomes apparent that different notions of possession are generally invoked in different linguistic constructions. For instance, a number of languages show different constructions depending on how ‘close’ the possessive relation is. Many languages distinguish between inalienable possession and alienable possession, where the former is associated with
2 William B. McGregor the ‘closest’ and most inherent relations (e.g. body parts, kin), the latter with less close, less inherent relations (e.g. owned material objects). (See further Chappell and McGregor 1995.) To assist the the reader track their way through the volume and to identify the main issues, we now provide detailed summaries of each of the papers. It is hoped that these summaries will whet the reader’s apetite for the detailed discussions and arguments in the individual papers, which of necessity can be at best hinted at here. But before we begin, it is necessary to establish some basic notions and terminology. Throughout the book the term possessum (abbreviated PM) is used in reference to that which is possessed; it is also sometimes used of the linguistic expression that denotes this item. Correspondingly, the term possessor (abbreviated PR) is used in reference to the person, animal, or whatever, that possesses the PM. Three primary and general types of possessive construction are usually distinguished, attributive, predicative, and external. As it is usually used, the term attributive possession refers to constructions in which the PM and PR expressions form an NP, as in my dog, the king of France’s bald pate, and Cliff’s ankle. These constructions are also termed adnominal possession. By contrast, predicative possession is used of constructions in which the possessive relation is expressed in a predicate, often by a possessive verb, as in I have a dog and The king of France has a bald pate. External possession constructions (EPCs) are constructions in which the possessive relation is not specified either by the lexical verb or within the NP – the PM and PR expressions, that is, do not belong to an NP – but rather at the level of a clausal construction, as in The dog bit Cliff on the ankle (see further Payne and Barshi 1999). Sometimes the term internal possession construction (IPC) is used instead of adnominal possession, especially when a contrast is being drawn with EPCs. In the first paper, English possessives as reference-point constructions and their function in the discourse, Peter Willemse, Kristin Davidse and Liesbet Heyvaert enquire into the information status of the PM referent in English adnominal possessive constructions with prenominal possessors (PRs), as in Greta Garbo’s knickers. Based on corpus data, they argue that the standard analyses of possessives as mere definite NPs (as per e.g. Quirk et al. 1985; Lyons 1999; Rosenbach 2002), or as NPs presupposing the identifiability of their referent (Du Bois 1980; Martin 1992) are problematic. At the same time, Taylor’s (1996) contrary prediction – based on the theory of possessive NPs as reference-point constructions (as per e.g. Langacker 1995; Taylor 1996) – that PM referents will be overwhelmingly discourse-
Introduction
3
new and anchored to a given PR, is also problematic: in most instances PM referents turn out to be discourse-given at least to some extent. Willemse, Davidse and Heyvaert argue that a taxonomy of discourse statuses must be recognised ranging from fully discourse-given, through text reference, inferable, ‘anchoring’, ultimately to fully discourse-new. They show that PM referents of adnominal possessive NPs in English may have discourse statuses at any point on this taxonomy. This paper adds a needed discourse dimension to the reference-point theory of adnominal possessive NPs. It demonstrates that it is not sufficient to study these constructions in isolation; account must be taken of the discourse context in which they occur. More generally, the need for discourse studies of adnominal possessive constructions in other languages is indicated. Jan Rijkhoff’s On the co-variation between form and function of adnominal possessive modifiers in Dutch and English deals with adnominal possessive modifiers of common nouns denoting concrete objects that are introduced by van ‘of’ in Dutch and of in English. He argues that these adnominal possessives can fill three of the five modifier functions distinguished in his Functional Grammar-inspired layered model of the NP (Rijkhoff 2002). In particular, they can serve a classifying function (indicating the kind of entity being referred to, as in a man of prayer), a qualifying function (indicating a property of the entity, as in a woman of great beauty), and a localizing function (indicating a location of the entity either in physical or conversational space, as in the bicycle of his father). Possibly in some languages adnominal possessives can serve discourse referential functions as well (e.g. if the third person singular possessive pronoun is used as a definiteness marker); however, it seems that in no language do they fill the fifth, quantifying function. Rijkhoff identifies three grammatical parameters with respect to which van ‘of’ and of adnominal possessives vary: modification (whether or not the possessive can be modified), predication (whether the possessive can occur in predicate position), and reference (whether the possessive can be referential). He argues, on the basis of usage data garnered from the internet, that the variation in grammatical and semantic properties found in expressions involving these markers correlates with the grammatical role of the modifier in the NP. It emerges clearly from Rijkhoff’s contribution that adnominal possessive constructions marked by van ‘of’ in Dutch and by of in English do not serve a unique modifying function in either language. An important consequence is that attributive possession is not – at least in Dutch and English – a unitary
4 William B. McGregor category, but embraces a range of emically distinct subtypes. The extent to which this applies in other languages as well demands investigation – cf. below on van Staden’s chapter for a different problem. In her contribution Is possession mere location? Contrary evidence from Maa, Doris Payne addresses the question of whether possession is conceptually identified with location, as has been widely presumed in typological and cognitive linguistic literature: the “possession-is-location” view, as she dubs it. She argues – on the basis of elicited and corpus examples in the Nilo-Saharan language Maa (Tanzania and Kenya) – that (at least in Maa) possession is cognitively distinct from location, and consequently that Possessor and Locative are distinct roles. Payne demonstrates that the verb tii ‘be at’ has locational and existential uses in particular constructions, while the verb ata ‘have’ has possessive and existential uses in particular constructions. However, there is no evidence that tii ‘be at’ can be used to predicate possession, or that ata ‘have’ can be used to predicate location. The non-overlap in locative and possessive senses for these two roots would be surprising if possession and location were conceptually indistinguishable. What the Maa verbs tii ‘be at’ and ata ‘have’ share is use in existential constructions. This leads one to suspect that in languages in which possession and location are represented by the same verb or construction the cognitive connection might be indirect, via the existential sense. This indicates the need for careful typological and historical investigations to test whether there might be closer connections between possession and existence, and location and existence, than between possession and location. Recent research on Nyulnyul (non-Pama-Nyungan, Australia) lends plausibility to the hypothesis. Nyulnyul has two formally similar negative constructions, a negative possessive and a negative existential/presentative; there is no evidence of any diachronic link via a locative construction, sense or use. Payne accepts that the fact that various languages use a single lexical item to express both possession and location is indicative of the existence of conceptual connections between the two domains; this does not, however, imply that they are conceptually identical. Indeed, in many languages where the same lexical item is used for expressing both location and possession, different constructions are employed, as Payne shows for Jakaltec and Amharic. This observation further underlines the need to go beyond mere lexical identity, and to recognize the relevance of constructions, a point made in other contributions to this volume. Learning to encode possession, by Sonja Eisenbeiß, Ayumi Matsuo, and Ingrid Sonnenstuhl deals with the acquisition of the expression of possession. Its main focus is on the acquisition of adnominal, predicative, and ex-
Introduction
5
ternal possession in German, although English and other languages are also mentioned. In contrast to many acquisition studies, this one focusses on the encoding of possessive relations, rather than merely uses examples of possessive constructions as instances of morpho-syntactic phenomena to be acquired. It is based on a corpus of children acquiring German monolingually, which includes data elicited utterances by stimuli designed to elicit possessive relations; child-directed speech is also included. It is shown that the target constructions emerge step by step, and that similarly the range of possessive functions encoded in the constructions increases over time. In the earliest stages, adnominal possession constructions appear not to exist. In the first stages, possessive relations need not be spoken of at all; when they subsequently are, it may begin with single word utterances that just identify the PR, even if the child is in the two-word stage. Prepositional possessive constructions involving von ‘of’ emerge later, the –s genitive construction even later. One of the interesting issues in the acquisition of the genitive –s concerns the acquisition of language specific constraints: in German, its restriction to unmodified PR nouns. The child acquiring German appears not to generalize the genitive to nominals with modifiers; instead, they adopt a strategy such as the omission of a required modifier (even where otherwise they would use the modifier), or omission of the marker itself. The developmental stages may overlap: constructions involving target morphemes obligatory in the adult language may alternate with constructions lacking them in the child’s speech. Moreover, there are often lexical restrictions, whereby the morphemes are initially restricted to particular lexical nouns, and only later generalize. Less well studied is the acquisition of predicative possession constructions and EPCs, especially the latter, and comparatively little data is available. This paper thus presents significant new data on the acquisition of these constructions. Eisenbeiß, Matsuo, and Sonnenstuhl show that HAVEconstructions emerge prior to BELONG-constructions, and show fewer deviations from the adult norm, consistent with the notion that HAVEconstructions are less marked. The German child language data shows EPCs only rarely, and quite late. In many circumstances in which adult speakers prefer a dative EPC, children often employ IPCs. Where they do produce something different, it often differs from the target dative EPC. Interestingly, Japanese children show no tendency to extend the Japanese dative construction to EPCs. How and when the double subject and double object EPCs of Japanese (e.g. Tsunoda 1995) are acquired is not known. More generally, research on the acquisition of predicative possession and EPCs in a wider sample of languages is called for.
6 William B. McGregor In Plain vs. situated possession in Czech: a constructional account, Mirjam Fried discusses the motivation for use of IPCs rather than the dative EPC in Czech. Two genitive IPCs are distinguished in Czech which show possessive marking of the PR, but differ in the relative order of the PM and PR expressions. In the EPC, the PM and PR are denoted by syntactically separate NPs, with dative marking of the PR NP. Fried argues that the genitive IPCs and the dative EPC contrast not just formally, but also semantically and pragmatically. She shows, using corpus data, that IPCs can be used for virtually any possessive relation, while the possessive relation in EPCs is more constrained. EPCs, she argues, are strongly associated with the inalienablility of the PM and the affectedness of the PR (hence her label “affected PR”), as is widespread crosslinguistically (e.g. Payne and Barshi 1999). More precisely, Fried shows that the PM in an EPC is an entity associated with the personal domain of the PR, as per Bally (1926/1995). The possessive relation between the PM and PR in EPCs cannot be pinned down precisely in terms of the semantic features of the PM and PR; rather, what is relvant is the way each is construed in context – hence Fried’s label “situated possession”, in contrast with “plain possession” of the IPCs. Genitive IPCs and dative EPC, Fried argues, encode distinct conceptualizations of possession. The formal, semantic and pragmatic differences between the genitive IPCs and the dative EPC indicate that they represent separate constructions in the Construction Grammar sense (e.g. Goldberg 1995). Fried proposes Construction Grammar analyses of the constructions which represent their formal characteristics and meanings. The dative EPC is situated with respect to other nearby constructions, including the genitive IPCs and the Dative of Interest construction, thus demonstrating that they occupy overlapping domains in the functional space of attributive possession. The paradigmatic relations among the constructions is relevant to their meanings and uses. Frank Lichtenberk’s Attributive possessive constructions in Oceanic shows that Oceanic languages typically distinguish two main types of attributive possession construction which are semantically and pragmatically distinct. In one type, the direct type, the PM carries an affix cross-referencing the PR; in the other type, the indirect type – which falls into a number of subtypes – the cross-referencing affix is carried by a possessive classifier element. In both types the affix cross-referencing the PR is most commonly a suffix, and the PR may be also realized by a full NP in addition; in some languages cross-referencing is reduced to an invariant third person singular affix if the PR is also represented by a full NP. (Further
Introduction
7
complexities are identified and discussed in Lichtenberk’s article; these need not concern us here.) As expected given its morphological simplicity (see e.g. Chappell and McGregor 1989, 1995), the direct type of possession construction strongly tends to be used in expressing inalienable possession; the indirect types strongly tend to be used for alienable possession. In the indirect types, the possessive classifier indicates what category of possession the PM represents. As Lichtenberk observes, the possessive classifier categorises the PM in respect to the relation of possession it exhibits to the PR. Possessive classifier systems show properties characteristic of classifier systems generally, including fluidity (the possibility of assignment of items to more than one category with accompanying meaning differences), and the existence of unpredictable classifications (exceptions to regularity in meaning, where choice of possessive classifier is semantically and/or pragmatically unexpected or irregular). A system of three possessive classifiers can be reconstructed for protoOceanic, distinguishing PMs for eating, drinking, vs. general. In some languages (particularly Micronesian languages) the system has been elaborated to make more distinctions. In other languages it has remained a ternary system, or has been reduced to a binary system. In a few languages, the possessive classification system has eroded completely, leaving a single possessive construction marked by a reflex of a former classifier (which may or may not contrast with a direct construction). Lichtenberk argues that the classifier system arose in the context of alienable possession due to their wide range of possible interpretations. For inalienable possession, by contrast, the relation between PR and PM is far more predictable, and there is thus less motivation to distinguish categories. In Possessive clauses in East Nusantara, the case of Tidore, Miriam van Staden shows that the Papuan language Tidore is typologically unusual in that it does not distinguish between attributive and predicative possession, contra the widespread belief that the distinction is universally maintained (e.g. Heine 1997: 26). All five possessive constructions in Tidore behave not as NPs, but as clauses. Four of them express predicative possession exclusively, while one construction, the focus of the paper, expresses etically both predicative and attributive possession. This rather simple construction consists of a bare noun expressing the PM and a prefix specifying the PR in terms of person and number; an NP denoting the PR may also occur initially. Tidore is fairly typical of the languages of the East Nusantara region: in both Austronesian and Papuan languages the possessive constructions in-
8 William B. McGregor volve verbal elements and retain some clausal characteristics. Thus in some languages the possessive construction involves a ligature deriving from a ‘have’ verb, while in others the possessive construction involves attachment of a marker cross-referencing the PR to the PM nominal, this marker being identical or almost identical with the corresponding verbal subject or object marker. Languages differ, however, in the extent to which the construction is grammaticalised; in some languages the possessive marker is clearly verbal, in others it is a ‘ligature’. The word order in possessive constructions is PR–PM rather than PM–PR as elsewhere in Indonesia and throughout the Austronesian world; this is also a reflection of the verbal/ clausal nature of the construction. Showing as it does that the distinction between predicative and attributive possession need not always be systematically maintained, van Staden’s paper nicely complements Jan Rijkhoff’s paper, which, it will be recalled, argues that adnominal attributive possession in Dutch and English do not constitute a single (emic) construction type. In Possessive expressions in the Southwestern Amazon, Hein van der Voort discusses adnominal possession constructions in eight unrelated languages of the linguistically diverse southwest Amazon region: three language isolates and five languages belonging to different genetic families. van der Voort identifies two general types of adnominal possessive construction (each of which occur in a range of subtypes): Type I in which a general possessive marker is added to the PR; and Type II in which the PM hosts PR agreement or person marking morphemes. In Kwaza and Aikanã (both isolates), both types of marking are found, though Type I predominates, and Type II is reduced and used only with third person singular PRs. Baure is also unusual in that it has a full set of PR marking morphemes, as well as a general possessive marker that is attached to optionally possessed nouns (but not to obligatorily possessed or unpossessable nouns), and not to the PR as in Type I. The main focus of van der Voort’s paper is on Kwaza. Kwaza is interesting not just because it shows both types of possessive construction, but also because of the way Type I possession is marked, by the form -dyh. The analysis of this form is somewhat uncertain. The balance of evidence seems to indicate that it is synchronically unanalysable. However, it is possible that it is analysable, at least diachronically, into a possessive marker -dy(homophonous with a range of morphemes with related meanings, including causative/benefactive) and a nominalising derivational morpheme -h, which is also used as a neutral classifier (see van der Voort 2004 for discussion of the classifiers). Interestingly, in possessive constructions -h can
Introduction
9
be replaced by a specific classifier. (This situation is not to be confused with the type of possessive classifiers found in Oceanic languages which categorise the PM in regard to the possessive relation itself, as discussed in Lichtenberk’s contribution to this volume – see above. In Kwaza, the classifier categorises the PM independently of the possessive relation, and is not restricted to possessive constructions.) Another unusual feature of Kwaza is that there is a semantically empty dummy lexeme e- which is either homophonous with a ‘have’ verb, or can be used used to express predicative possession (van der Voort inclines to the second possibility). Such semantically empty lexemes are found in other languages of the region, though not with the ‘have’ sense. The final paper deals with a topic that has not received a great deal of attention in the literature on possession, the expression of possession in sign languages. In this paper, Possession in the visual-gestural modality: how possession is expressed in British Sign Language, Kearsy Cormier and Jordan Fenlon focus on the expression of attributive and predicative possession in British Sign Language (BSL), and make some comparative remarks on similarities to and differences from other sign languages. Attributive possession is expressed in two main ways. In pronominal possession it is indicated by a set of possessive pronouns that differ in form from the cardinal possessive pronouns, and which precede the PM expression. In nominal possession PR and PM NPs are linked by a free possessive pronoun serving as a possessive copula, a pattern also found in spoken languages (see e.g. McGregor 2001). Interestingly, the PR may be omitted if it is inferable (see also Fried’s paper in this volume). Older signers also use a finger-spelled S, borrowed from signed English systems, as a clitic at the end of nominal PR expressions. This mode of expression, obsolescent in BSL, is still viable in Australian Sign Language (Auslan) and American Sign Language (ALS). Cormier and Fenlon also examine whether the expression of possession in sign languages is comparable with its expression in spoken languages. They draw out a number of commonalities, including the distinction between attributive and predicative possession, the existence of an optional alienable-inalienable contrast in BSL, and the existence of both HAVE and BELONG verbs. They also observe that there are connections between possessive expressions and locative expressions, and a tendency for possessive morphemes to grammaticalise, which has happened to a limited extent to possessive pronouns in BSL. As this paper makes clear, sign languages must be taken into account in serious typological investigations of possession.
10 William B. McGregor As the above summaries indicate, a diverse range of genetically and typologically different languages are discussed in the volume; these languages are also quite widely spread geographically, languages from Europe, South America, the Pacific region, and Africa being represented. Familiar languages such as English, Dutch, and Czech are included as well as less well known languages such as BSL and Kwaza. The papers dealing with the former set reveal that we still have a lot to learn about possession even in the best described languages. The papers of this volume are all strongly empirically oriented, and many are based squarely on usage data, including (where possible) instances from corpora. Constructed examples also play an important role in many of the papers, in some cases by necessity (e.g. where the bulk of the available data consists of elicited utterances), in some cases to fill gaps in usage data or to identify what is grammatically impossible. The orientation to usage is augmented by deep concern with meaning. Thus contributions are on the whole not content with just identifying formal construction types, but also make serious attempts to determine what differences in meaning might be encoded or implicated by the alternative modes of expression. The empirical orientation of the contributions does not mean that theory is eschewed. Indeed, the majority of papers in the volume are concerned in one way or another with linking empirical observations with theoretical concerns. Thus some papers explicitly confront theory with empirical data; some use theory in an attempt to explain empirical observations; and some use theory to account for the semantics and/or pragmatics of possessive constructions. Granted the concern with meaning and usage, it is not surprising that theories from the functionalist end of the spectrum are most strongly represented; however, the contribution by Eisenbeiß, Matsuo, and Sonnenstuhl shows that there may be some chance of approachment between formalist and functionalist theories – and that in some domains at least, similar predictions are being made. The need for further study of the means used by languages to encode relations of possession is manifest: we are a long way from having a complete knowledge and understanding of the formal modes of expression. In his contribution van der Voort mentions that the Amazonian isolate Movima expresses inalienable and predicative possession by reduplication (Haude 2006: 238ff.); one wonders what other unusual modes of expression might remain undetected in the thousands of undescribed languages of the world. It is hoped that the papers in this volume will stimulate readers to investigate possession in more languages, spoken and signed. It is also hoped that they will stimulate further research on the well described languages.
Introduction
11
References Bally, Charles 1926 / The expression of concepts of the personal domain and indivisibility 1995 in Indo-European languages. In The Grammar of Inalienability: A Typological Perspective on Body Part Terms and the Part-Whole Relation, Hilary Chappell and William B. McGregor (eds.), 31–61. Berlin /New York: Mouton de Gruyter. Chappell, Hilary and William B. McGregor 1989 Alienability, inalienability and nominal classification. Berkeley Linguistics Society Proceedings 15: 24 –36. Chappell, Hilary and William B. McGregor 1995 Prolegomena to a theory of inalienability. In The Grammar of Inalienability: A Typological Perspective on Body Part Terms and the Part-Whole Relation, Hilary Chappell and William B. McGregor (eds.), 3–30. Berlin /New York: Mouton de Gruyter. Goldberg, Adele E. 1995 Constructions: A Construction Grammar Approach to Argument Structure. Chicago and London: University of Chicago Press. Du Bois, John W. 1980 Beyond definiteness: The trace of identity in discourse. In The Pear Stories: Cognitive, Cultural and Linguistic Aspects of Narrative Production, Wallace L. Chafe (ed.), 203–274. Norwood: Ablex. Heine, Bernd 1997 Possession: Cognitive Sources, Forces and Grammaticalization. Cambridge: Cambridge University Press. Haude, Katharina 2006 A Grammar of Movima. PhD thesis, Radboud Universiteit Nijmegen. (available on http://webdoc.ubn.ru.nl/mono/h/haude_k/gramofmo.pdf) Herslund, Michael and Irène Baron 2001 Introduction: Dimensions of possession. In Dimensions of Possession, Irène Baron and Michael Herslund (eds.), 1–25. Amsterdam: John Benjamins. Langacker, Ronald W. 1995 Possession and possessive constructions. In Language and the Cognitive Construal of the World, John R. Taylor and R. E. MacLaury (eds.), 51–79. Berlin: Mouton de Gruyter. Lyons, Christopher 1999 Definiteness. Cambridge: Cambridge University Press. Martin, James R. 1992 English Text: System and Structure. Amsterdam: John Benjamins.
12 William B. McGregor McGregor, William B. 2001 Non-verbal predicative possession in Nyulnyulan languages. In Forty Years on: Ken Hale and Australian Languages, Jane Simpson, David Nash, Mary Laughren, Peter Austin and Barry Alpher (eds.), 337– 352. Canberra: Pacific Linguistics. Payne, Doris L. and Immanuel Barshi (eds.) 1999 External Possession. Typological Studies in Language, 39. Amsterdam: John Benjamins. Quirk, Randolph, Sidney Greenbaum, Geoffrey Leech and Jan Svartvik 1985 A Comprehensive Grammar of the English Language. London: Longman. Rijkhoff, Jan 2002 The Noun Phrase. Oxford: Oxford University Press. Rosenbach, Annette 2002 Genitive Variation in English. Berlin /New York: Mouton de Gruyter. Seiler, Hansjakob 1983 Possession as an Operational Dimension of Language. Tübingen: Narr. Taylor, John R. 1989 Linguistic Categorization: Prototypes in Linguistic Theory. Oxford: Clarendon Press. Taylor, John R. 1996 Possessives in English. Oxford: Oxford University Press. Tsunoda, Tasaku 1995 The possession cline in Japanese and other languages. In The Grammar of Inalienability: A Typological Perspective on Body Part Terms and the Part-Whole Relation, Hilary Chappell and William B. McGregor, (eds.), 565–630. Berlin /New York: Mouton de Gruyter. Voort, Hein van der 2004 A Grammar of Kwaza. Mouton Grammar Library 29. Berlin /New York: Mouton de Gruyter.
English possessives as reference-point constructions and their function in the discourse Peter Willemse, Kristin Davidse and Liesbet Heyvaert
1. Introduction In this article we will be concerned with the discourse status of the possessum (PM) referent of prenominal 1 possessive NPs such as his car, John’s car.2 With regard to this issue we find two opposed, indeed paradoxical, claims in the literature: (a) possessive NPs are definite NPs (e.g. Quirk et al. 1985; Lyons 1999; Rosenbach 2002) or they presume the identifiability of the PM referent (e.g. Du Bois 1980; Martin 1992), and (b) possessive NPs introduce overwhelmingly new, previously unknown, PM referents into the discourse by linking them to typically given possessor (PR) referents (Taylor 1996). In this article we will argue, basing ourselves on real usage data,3 that the analysis of possessive NPs as either mere definite NPs, or as NPs which typically introduce new referents in the discourse, cannot be maintained. We will propose that (a) possessive NPs have an identification mechanism different from that found in NPs with definite articles or demonstratives, and (b) the question of the discourse status of PM referents of possessive NPs cannot be reduced to a binary distinction between new or given in the discourse. On the basis of a qualitative and quantitative analysis of a corpus of possessive NPs in extensive discourse contexts, we will argue that many PM referents have a discourse status in between fully given and fully new. For this range of discourse statuses we will propose a continuum-like classification. 1
2 3
As per Taylor (1996: 2), we use the term prenominal possessives to refer to possessive NPs in which a genitive or a possessive determiner precedes the head noun, as opposed to ‘postnominal’ possessives such as a friend of John’s. This excludes NPs with ‘classifying’ or other non-determining genitives like the lion’s share, a mother’s boy. The data we used were extracted from the COBUILD Bank of English corpus (examples are marked ‘CB’) and the Lancaster-Oslo-Bergen corpus (examples are marked ‘LOB’). The data from the COBUILD corpus are reproduced with the kind permission of HarperCollins Publishers Ltd.
14 Peter Willemse, Kristin Davidse and Liesbet Heyvaert 2.
English prenominal possessives in the literature
2.1. Alleged definiteness of possessive NPs and presupposed identifiability of their referents Possessive NPs have predominantly been associated with definiteness in the literature. Abbott (2004: 122), in her chapter on definiteness in the Handbook of Pragmatics, states that possessive NPs “are almost universally considered to be definite”. Specifically with regard to English, many descriptive grammars classify possessive NPs as belonging to the paradigm of definite NPs, implying a far-reaching parallelism between possessive determiners and genitives on the one hand, and definite determiners such as the definite article on the other. Quirk et al. (1985: 326) analyze “the genitive construction [i.e. the prenominal possessive, PW/KD/LH] as a noun phrase embedded as a definite determinative within another noun phrase”. Biber et al. (1999: 271) claim that “possessive determiners make NPs definite”. Huddleston (1988: 90–91) classifies possessives as “determiners that mark the NP as definite”. Besides grammars of English, other accounts have classified possessive NPs as definite. Lyons (1999: 23ff.) also holds that in English, “possessives render the noun phrase which contains them definite”. He supports this claim with the argument that it is generally possible to paraphrase possessive NPs with NPs marked by the definite article, for instance: (1) my cousin the son/daughter of my aunt and uncle the man next door’s car the car belonging to the man next door Rosenbach (2002: 14) proposes a similar analysis of possessive NPs as definite on the basis of definite paraphrases, and claims that this holds true even if the genitive has itself indefinite marking, as illustrated by the following examples: (2) the king’s daughter the daughter of the king a king’s daughter the daughter of a king In addition to claims classifying possessive NPs paradigmatically with definite NPs, some scholars have linked possessives to the concept of ‘presupposed/presumed identifiability’ of referents, which is generally assumed to be the meaning signalled by the definite article (see e.g. Chafe 1976; Gundel, Hedberg and Zacharski 1993, 2001; Langacker 2002: 33). Du Bois (1980: 218), whose early analysis of possessive NPs has been rather influ-
English possessives as reference-point constructions
15
ential, proposes that “[p]ossessive noun phrases (…) presuppose identifiability”.4 In his characterization of NPs in the ‘Pear Stories’ data, Du Bois characterizes several examples of possessive NPs as definite. The NP his hat in the following example is described as a definite initial mention of a referent: (3) … when he turns around his hat flies off.
(Du Bois 1980: 243)
It is added that “[h]is is similar to the in that it demands (presupposes) identifiability, but different in that it supplies some extra information that may help make the identification possible” (Du Bois 1980: 243). Martin (1992: 132) situates possessive NPs within the general class of ‘phoric’ NPs, i.e. NPs coding their referent as in some way retrievable. He bases his claims about possessive NPs on Du Bois (1980) as well as on Halliday and Hasan (1976: 70), who analyze possessives as realizing a type of specific deixis similar to that construed by the definite article and the demonstratives this/that/these/those. Importantly, the identifiability of the referent is considered by Martin (1992) to be coded by the possessive NP; in other words, there is something about the use of a possessive NP to refer to a referent that signals its identifiability. The key to the identification of the PM referent is said to be the PR: “[t]his is after all literally what the grammar of the English nominal group argues: ‘recover the identity of the possessed participant here through its possessor’” (Martin 1992: 133). Martin also pays explicit attention to the questions how the referents of possessive NPs are embedded in the discourse and how they participate in the reference chains which are construed by NPs with anaphoric and cataphoric deictics. He recognizes that possessive NPs have two discourse referents, the possessed and the PR, and he considers the question whether the possessed has an identifiability status of its own. However, following Du Bois’ (1980: 243– 245) claim that “a frog of his or a friend of John’s do not alternate with his frog or John’s friend to introduce participants” (Martin 1992: 132–133), he concludes that “possessive deictics are the deixis of the participants they possess” (Martin 1992: 132). Consequently he (1992: 133) proposes that “possessive nominal groups … only be coded once for phoricity”. In his actual text analyses of phoric relations, he systematically analyzes possessive deictics as referring back to the PR, leading to reference chains such 4
It is not fully clear whether Du Bois uses “presupposed identifiable” in the technical sense of ‘already available in the discourse context as a singled-out entity’ (Langacker 2002: 33).
16 Peter Willemse, Kristin Davidse and Liesbet Heyvaert as: the cat she her dinner (1992: 143), the boy his frog (1992: 144). His phoric chains, in other words, track the identifiability status of the PR, not of the PM referent. We can now sum up the main elements in the tradition that views possessive NPs as definite and their referents as identifiable, and formulate our main criticisms with regard to them. Firstly, linguists adhering to this analysis of possessive NPs often invoke systematic alternation of possessive NPs with definite NPs as a grammatical argument for according definiteness to the former. However, there is textual evidence that this alternation is not as systematic as claimed. Possessive NPs in predicative copular sentences, for instance, do alternate with indefinite NPs, e.g. You’re my friend – You’re a friend of mine (see also Declerck 1986: 32). Likewise, NPs with indefinite genitives such as a friend’s friend in (4) can, against Rosenbach’s (2002: 14) claim, alternate with indefinite NPs, e.g. a friend of a friend in (5). (4) It is not to be wondered at [...] that a friend’s friend, described by letter, should turn out an unrecognizable stranger. (CB) (5) Jimmy was a referral, a friend of a friend.
(CB)
Our data contained other examples in which possessive NPs alternate with indefinite NPs, such as the following: (6) On Monday, Christie’s in New York is to sell Greta Garbo’s knickers. (CB) In this example, the prenominal possessive Greta Garbo’s knickers alternates with indefinite (a pair of) knickers of Greta Garbo. In view of this, it is hard to maintain that English prenominal possessives code definiteness, since a subset, viz. those alternating with indefinite article + noun + of + PR, are functionally indefinite. In this context, it can be recalled that other languages, such as Italian and Spanish, code (in-)definiteness and prenominal possession separately, making a distinction between, for instance, il mio libro (‘the my book’) and un mio libro (‘a my book’) (Lyons 1999: 24). In English prenominal possessives, the contrast definite – indefinite remains covert, but it can be made explicit in the corresponding complex NPs in which the PR is expressed by postmodifier of + NP. Secondly, if a functional definition of possessive NPs is given, it is observed that the PM is retrievable through the PR (Martin 1992: 133). This explanation refers to the identification mechanism internal to possessive
English possessives as reference-point constructions
17
NPs. However, this NP-internal identifying relation has to be distinguished from the external relations which the two discourse referents may maintain with other elements in the surrounding discourse. If we look more closely at the latter, two distinct perspectives can be taken. On the one hand, as illustrated by Martin’s (1992) text analyses, possessive NPs can be viewed as partaking in reference chains keeping track of the PR referent. As the PR is typically given, co-reference to the PR will account for a large part of the reference chains construed by possessive NPs. However, in some – admittedly fewer – cases, the PM referent may also be coreferential with a discourse referent, as in the following example, where Fleming’s cardboard booby refers back to ‘Bond’. (7)
Goldfinger, the third Bond movie, was released in December of that year, and with it was founded an industry that would turn Fleming’s cardboard booby into a product that 30 years later rivals Mickey Mouse in terms of global penetration. (CB)
In other words, the PM referent can already be present as a singled-out instance in the discourse, and need not be discourse-new. This shows that the two referents of possessive NPs insert themselves into the discourse with distinct given-new statuses, which have to be studied in actual discourse. Martin’s (1992) observation that the PM referent is recoverable through the PR hints at the internal identifying relation present in possessive NPs, but he does not explore it in more detail. The internal identification mechanism of possessive NPs and its ‘anchoring’ of referents has been at the core of the analysis of possessives as reference-point constructions, developed within the framework of Cognitive Linguistics by Langacker (1993, 1995) and Taylor (1996). Taylor (1996) also discusses certain properties of both the PR and the PM referent from a discourse perspective. We will turn to this account in the next section. 2.2. Alleged newness of PM referents as targets of reference-point constructions Langacker (1993, 1995) proposes that possessive NPs should be analyzed as reference-point constructions, i.e. as constructions motivated by the relation between two entities, one of which functions as the ‘reference point’ giving mental access to the other. More specifically, in possessive NPs, the PR functions as a reference point for the identification of the PM, which func-
18 Peter Willemse, Kristin Davidse and Liesbet Heyvaert tions as the ‘target’ needing identification. For instance, in the possessive NP Sarah’s car, the PR, ‘Sarah’, functions as the reference point for the identification of the ‘car’ which is ultimately being referred to; when processing this NP, the addressee will initially make mental contact with the entity ‘Sarah’ and subsequently identify the ‘car’ in question as the one associated with Sarah (e.g. the one she owns or drives). Since the reference point serves to give mental access to the target, a reference point is normally cognitively more easily accessible than the target that is being linked to it; an entity is, then, chosen as a reference point for another on the basis of the fact that it has “a certain cognitive salience, either intrinsic or contextually determined” (Langacker 1993: 6). The reference-point analysis thus offers a detailed account of the identification mechanism set up within possessive NPs. However, the PR (reference point) and PM (target) are themselves also discourse referents embedded in the discourse in which the possessive NP is used. Taylor (1996), who further develops Langacker’s (1993, 1995) reference-point analysis specifically in relation to prenominal possessives, formulates a number of predictions for PR referents as well as for PMs with regard to their givenness or newness in the discourse. In order to make these predictions, he is led by what he perceives to be the inherent logic of the reference-point relation. Since the aim of using a reference-point construction is to make a target entity more accessible by tying it to a reference point, Taylor argues, it is to be expected that the reference point should be more easily mentally accessible than the target, as “it would be perverse indeed to invoke a less accessible entity to aid the identification of a more accessible entity” (Taylor 1996: 210). Conversely, “were the target as easily accessible as the reference point, there would be no point in using the reference point for its identification” (ibid.). This line of reasoning leads Taylor (1996: 218) to posit that PRs and PMs are “maximally differentiated” in terms of their (typical) discourse properties. PRs, on the one hand, overwhelmingly have ‘given’ status in the discourse, since given entities (i.e., entities already present in the discourse and known to the addressee) are more cognitively accessible than newly introduced entities. More specifically, Taylor (1996: 212) predicts that PRs will typically be “entities mentioned in recently preceding discourse”, that discourse or text topics will frequently occur as PRs and that PR nominals will frequently be definite. By contrast, as entities that need anchoring to a reference point for identification, “possessees overwhelmingly introduce new, previously unnamed entities into the discourse” (Taylor 1996: 217).
English possessives as reference-point constructions
19
2.3. Overview and research questions To sum up, existing accounts of English prenominal possessives have tended not to fully recognize their complexity as constructions referring to two discourse referents between which an NP-internal relation of identification is set up and which at the same time maintain external relations with the surrounding discourse context. The discourse status of the PR is relatively straightforward, since it either (in the case of a genitive) has explicit definite or indefinite marking or (in the case of a possessive determiner) is realized by an inherently definite pronominal form. New PR referents are therefore normally realized by an indefinite genitive,5 whereas given PRs have definite marking or a pronominal form. What is more, the PR referent has convincingly been established to be overwhelmingly discourse-given (see Taylor 1991, 1996). By contrast, the discourse status of the PM referent is controversial and underresearched – both conceptually and empirically. On the one hand, within the tradition that includes possessive NPs in the paradigm of definite NPs, the PM referent tends to be viewed as ‘presumed identifiable’. On the other hand, within the reference-point approach to possessive NPs, Taylor (1996) assumes that PM referents, as the target of identification, are overwhelmingly new. Given such opposed claims, closer investigation of the discourse status of PM referents imposes itself. Since the PM referent does not have direct definite or indefinite marking, its discourse status cannot simply be deduced from the form of the NP. In order to uncover it, possessive NPs must be studied in extensive discourse contexts, which make it possible to trace the PM referent’s givenness or newness throughout the discourse context preceding the possessive NP (Willemse 2005: 106–133). For the present study, 400 instances of possessive NPs have therefore been studied in extensive discourse contexts. The aim of the analysis was to investigate the ‘external’ relations of the PM referent to the surrounding discourse context, i.e. to determine whether the possessive NP was used to introduce a new referent into the discourse or to refer to a given discourse referent. Carrying out this analysis involved two steps. Firstly, a theoretical-descriptive framework had to be set up to account in a precise and systematic way for the various degrees of discourse givenness/newness displayed by the PM referents. Secondly, this description had to be applied to the data-base and the results quantified. Analytically, these are two distinct steps, but in practice the descriptive categories were 5
It should be noted that indefinite grounding genitives are by no means anomalous; see Willemse (2005: 183–200).
20 Peter Willemse, Kristin Davidse and Liesbet Heyvaert set up not only with reference to the literature, but also by shunting between the description and the data to ensure optimal coverage of the patterns emerging from the latter. The remainder of this article is structured as follows. In section 3, we will look more closely at functional equivalents of (in)definiteness discussed in the literature. We will focus in particular on the important distinction between recoverability (relevant to the reference point analysis of possessive NPs) and that of first – subsequent mention (relevant to the discourse status of the PM referent). In section 4, we will present the qualitative and quantitative results of the data analysis. In the first place, this involves setting out our descriptive classification of the different discourse statuses, ranging from fully given (coreferential), over a number of statuses in between given and new, to fully new. Secondly, we also report on the quantitative results of Willemse’s (2005) corpus study, i.e. on the relative frequency with which these different statuses occur in the data-base. Finally, we will also situate our descriptive analysis in a broader discourse perspective: possessive NPs turn out to serve a variety of specific discourse functions such as reiteration of the PR as a central discourse participant or reclassification of the PM referent, which cautions against viewing them only as referential identification mechanisms along a strict reference point logic. In section 5, we will summarize our main findings and point to their implications for the analysis of possessive NPs as reference point constructions and as so-called ‘definite’ NPs. 3. The functional dimensions of definiteness and possessive NPs As observed by Lyons (1999), there is less unanimity about the functional definition of definiteness than is often assumed. In his formally-oriented study of definiteness, Lyons (1999: chapter 1) includes two main components in its semantic definition. The first one is identifiability, understood as an extended version of the so-called ‘familiarity hypothesis’. The use of a definite article signals that the hearer should be able to identify the referent of the NP it occurs in. This may mean that the hearer is familiar with the actual referent or that he can be directed to it via other mechanisms such as anaphora and bridging or associative uses, as in a taxi – the driver. The second meaning component is uniqueness: the definite article signals that there is just one entity corresponding to the description given, relative to a particular context. Following Hawkins (1978), uniqueness is extended to inclusiveness to include plural and mass NPs, for which the definite article
English possessives as reference-point constructions
21
signals that reference is made to the totality of objects or mass in the context satisfying the description. ‘Inclusive’ reference is treated as secondary to identifiability by Lyons and as an implicature, rather than a meaning component, of definiteness by Hawkins (1978) and Declerck (1986). Martin (1992) uses the notion of recoverability, rather than identifiability, which more clearly suggests that other processes than actual knowledge of the referent may be involved. According to Martin (1992: 98) “every time a participant is mentioned, English codes the identity of that participant as recoverable from the context or not”. When the grammatical resources used in a specific NP signal that the identity of its referent is in some way recoverable, the NP is said to be phoric. Phoric NPs are, consequently, different types of definite NPs (proper names, pronouns, NPs grounded by the definite article or by a demonstrative) which all embody “directives indicating that information is to be retrieved from elsewhere” (Halliday and Hasan 1976: 31). Like Lyons, Martin also stresses that ‘this information’ may pertain to the referent itself or to another referent with which it is indirectly associated. It is in accordance with this logic that Martin (1992: 133), despite recognizing that possessive NPs refer to two referents, codes them only once for phoricity, as the PM referent is recoverable through the PR referent. Whilst this is reminiscent of the reference point analysis, it has to be pointed out that Martin’s recoverability analysis is restricted to functionally definite possessive NPs, and does not apply to examples such as (4) and (6) above. By contrast, the reference point analysis is broader – at least as a theoretical concept, rather than in Taylor’s (1996) interpretation of it in relation to possessive NPs. It foregrounds the conceptual relation between reference point and target and can also be applied to functionally indefinite possessive NPs. As recoverability of a referent via association or bridging is not discussed much in general reference works, we will consider the specialized literature on it more closely. When a bridging relation holds between two elements, it may be motivated by one of several types of conceptual relations, i.e., the ‘source’ of bridging may differ. Generally speaking, the basis for bridging is some sort of strong associational relationship between entities, one which is strong enough to allow the identification of one entity on the basis of an earlier mention of the other, associated entity. This translates to several more specific relation types, the most important one of which is the part-whole relation, e.g. (8): (8) Peter has bought a new car. There is much more room in the boot than there was in his old car.
22 Peter Willemse, Kristin Davidse and Liesbet Heyvaert In example (8) the part ‘boot’ can be marked as recoverable on the basis of its relation with the whole that it forms part of, ‘car’. The processing of bridging reference thus requires some inferencing on the part of the addressee, who needs to retrieve the conceptual relation which forms the basis for the identification of the referent from background knowledge (cf. Ariel 1990: 185), while the role of the immediate discourse context is to provide an element which activates this inference (a ‘trigger’, Hawkins 1978: 123). Besides part-whole relations, other relations of strong and habitual association between two entities may be strong enough to motivate bridging, as in (9). (9) He was very interested in buying that old house up the road, but the owner wouldn’t sell. Although an owner is of course not a part of a house, it is an entity typically associated with a house, which can hence be introduced with the definite NP the owner. Brown and Yule (1983: 257) emphasize the similarities between part-whole relations and relations of strong, stereotypical association by classifying both under the heading of relations of ‘having’ (compare: ‘a car has a boot’ – ‘a house has an owner’). Both part-whole relations and relations of strong association are relations between entities. Definite reference to an entity based on its relation to an event, activity or situation described in the preceding context occurs frequently as well. For instance, in the following example, the referent ‘killer’ is realized with a definite NP, since the event of a murder involves a killer as the agent of the described action: (10) It was dark and stormy the night the millionaire was murdered. The killer left no clues for the police to trace. (Brown and Yule 1983: 258) In the literature, the invoking of events to make a referent recoverable has been discussed in terms of notions such as frames, scenarios and schemata. With reference to Chafe (1972), Du Bois (1980: 235ff.) develops the concept of event-frames, which are “composed of a network of related actions, along with the people and objects involved in those actions” (Du Bois 1980: 246), and are therefore rather hard to delineate. He notes that definiteness can be used as an indication to decide what elements make up a particular frame. The notion of schema is used by Chafe (1996) to capture the way in which a referent may be inferred from contextual information about events. He gives the example of someone who is describing the advantages of eating
English possessives as reference-point constructions
23
cream cheese out of a carton and introduces the referent ‘spoon’ with a definite NP. (11) and if you just sort of rinse the spoon off afterwards, you don’t really have to wash dishes […]. (Chafe 1996: 39) The treatment of ‘spoon’ as an identifiable referent, Chafe argues, is due to the fact that it was “indirectly shared because of its association with the eating-out-of-the-carton schema” (Chafe 1996: 39). The notion of scenarios, finally, was developed by Sanford and Garrod (1981, 1998). Scenarios are defined as representations of situations and of the roles involved in them, which are activated by text input (i.e. by explicit ‘clues’ in the text) and retrieved from long-term memory. Entity tokens (i.e. discourse referents) are mapped onto the roles which are part of the invoked scenario in a specific context. Thus, when an appropriate scenario has been activated in a specific context, discourse referents filling role ‘slots’ in the scenario can be referred to with a definite NP. Sanford and Garrod (1981: 112ff.) support this analysis with psycholinguistic evidence that reading times for sentences containing a definite NP which introduces an entity in the discourse are not longer when an appropriate scenario has been invoked than when the entity has been introduced explicitly in the preceding discourse (i.e. when the definite NP is an anaphor). The question of which types of relations and associations are apparently strong enough to motivate bridging has been speculated on a lot in the literature. Chafe (1972: 63), for instance, suggests that in the case of partwhole relations, the obligatoriness of a part influences the possibility of bridging – i.e., the more optional the part is, the less likely it is to be coded definitely. Ariel (1990: 184 –185) points to a number of factors which may govern the ‘inferability’ of an entity, such as the stereotypical nature of the entity to another entity in the context and the prominence of the ‘antecedent’ in the discourse. However, she also remarks that much of what determines the inferability of specific (types of) entities is probably governed by language-specific conventions. Returning to the functional definition of definiteness, then, a third basic functional dimension of definiteness has been identified in the text-based studies of Du Bois (1980) and Fraurud (1990), viz. that of first versus subsequent mentions. Du Bois (1980: 220ff.) and Fraurud (1990: 413ff.) both point out that it is traditionally assumed that indefinite NPs involve the first mention of a referent: they are said to establish a new discourse referent and to instruct the hearer to open a new file in their consciousness. Definite NPs,
24 Peter Willemse, Kristin Davidse and Liesbet Heyvaert by contrast, are thought of as involving the retrieval of previously established discourse referents, i.e. non-initial mentions, and as instructing the hearer to update an old file. Du Bois and Fraurud both also show that this assumed correlation between indefiniteness – first mention and definiteness – subsequent mention is confirmed only to a certain extent by actual text analysis. In their data, there are considerable portions of definite first mentions and (smaller) sets of indefinite subsequent mentions. The first-subsequent mention distinction applies only to NPs designating specific referents (Du Bois 1980: 207); it does not apply to generic NPs and to proper names. Du Bois discusses three main types of definite first mentions: (i) ones marked by an unstressed demonstrative; (ii) ones containing identifying information in a postmodifying relative clause; (iii) ones due to association with a frame. In Du Bois’ spoken data unstressed demonstratives this/these were commonly used for initial mentions, and by most speakers even restricted to first mentions, as in he could possibly see this little boy [1st] coming on a bicycle (Du Bois 1980: 219). The clash that we find here between the definite determiner and the introduction of a new discourse referent has generally been noted. NPs containing specific and new information in the presupposed format of a restrictive relative clause may also make a definite initial mention “acceptable”, as Du Bois (1980: 223) puts it, e.g. she knocks the hat that he’s wearing [1st] off on the ground (Du Bois 1980: 222). That the identifying information provided by defining relative clauses may motivate the use of a definite article for the whole NP is, again, well-established. Finally, a larger whole or a specific activity may serve as the ‘frame’ enabling the definite initial mention of referents typically associated with them, as with living room – the wall (Du Bois 1980: 233), sell – the money (Du Bois 1980: 215). Here we are up against the mechanism of bridging again. However, whereas bridging has generally been discussed in the context of explaining the definite form of the NP and glossing its meaning as ‘recoverable’, Du Bois also stresses that, in spite of the definite form, we have initial mentions here. Given Du Bois’ incisive comments about the first mention status of referents of bridged NPs, it is perhaps surprising that he does not extend this observation to PM referents of possessive NPs, which, in a considerable number of cases, are also first mentions with inferential relations to the context. As we will see in section 4.4, possessive NPs may designate instances that, in a strict sense, are mentioned for the first time in the discourse, even though they are indirectly related to other elements in the discourse, from
English possessives as reference-point constructions
25
which they are inferable such as the handbrake to the Mitsubishi Starwagon in (12). (12)
Police prosecutor Snr-Sgt Geoff Jackson told the court Pizzino was one of eight passengers in the Mitsubishi Starwagon which crashed on Robina Parkway at the Gold Coast about 5.10 am on Wednesday. Snr-Sgt Jackson said Pizzino activated the vehicle’s handbrake, causing the driver to lose control. (CB)
In fact, in many cases the prenominal possessive can be replaced by the definite article, showing that the conceptual relation between the frame evoked by the context and the inferred referent is the same as that underlying bridging anaphora, e.g. (12’) […] Snr-Sgt Jackson said Pizzino activated the handbrake, causing the driver to lose control. Moreover, PM referents may also be ‘more new’ to the discourse than inferable first mentions: they may have some link to a conceptual ‘anchor’ in the discourse that makes their occurrence not entirely unpredictable, but is not strong enough to intrinsically convey uniqueness/inclusiveness on them so as to make them ‘recoverable’. This can be illustrated with possessive NPs referring to clothes and other alienable possessions, which have a certain conceptual link to their ‘owners’. If the possessions referred to are contextually unique or include all the instances in the given context, then such NPs are functionally definite. In (13) the detective is wearing only one hat, and (14) is concerned with all of Paxman’s ties. (13) An Irish detective arrested a wanted criminal in a Dublin street. Just as he was about to slap the handcuffs on him, a gust of wind blew the detective’s hat down the street. “Shall I go and fetch it for you?” asked the criminal. (CB) (14) Most of Jeremy Paxman’s ties don’t go with his shirts.
(CB)
However, if this is not the case, then they are functionally indefinite, as in example (6) above, Christie’s to sell Garbo’s knickers, and example (15), in which Ruth has been stealing some, but not all (*the), clothes, jewellery and accessories of Elizabeth’s. These examples show that the conceptual relation of alienable possession is not in itself strong enough to make a referent inferable.
26 Peter Willemse, Kristin Davidse and Liesbet Heyvaert (15) Raven-haired Ruth is a statuesque woman consumed with envy for the blonde Elizabeth, who Ruth’s parents took in after she was left orphaned. All her life Ruth has been secretly stealing Elizabeth’s clothes, jewellery and accessories and dressing up as the girl she alternately idolises and hates. (CB) Finally, some PM referents are ‘fully new’ in the discourse, unpredictable from any elements in the context. This is, for instance, the case with actions (and aspects of actions such as results) expressed by nominalizations or deverbal nouns, which can scarcely be predicted from the link to their agent or patient in the preceding text, as in (16), in which Milner’s prints refers to ‘prints of Milner’s’. (16)
Point two, Milner was the only person who entered Stevens’s apartment the night of the murder – he was positively identified by the hallman. Point three, we found Milner’s prints [sic] four places in the apartment including the library. (CB)
From this brief discussion of first-mention PM referents it transpires that different degrees of discourse newness can be distinguished for them. As we will see, we can also observe a difference in the degree of discourse givenness for PM referents that are subsequent mentions. The typology of discourse statuses of PM referents that we will set out in the next section will therefore be couched in terms of degrees of discourse-newness and discourse-givenness. 4. Data analysis We pointed out in section 2 that one of the most important remaining questions in the literature on possessive NPs pertains to the discourse status of their PM referent. In section 3, we linked up the notion of discourse status with the discussion of initial-subsequent mentions, but we differentiated this contrast into degrees of discourse newness and givenness. Thus, we are now in a position to tackle the discourse status of PM referents in real usage. We compiled a set of 400 instances of possessive NPs, composed as listed in Table 1.
English possessives as reference-point constructions
27
Table 1. Overview of the data base NP type
# of instances
Genitive + N Total Genitive
200 200
its + N
50
their + N my + N her + N Total Possessive Det.
[30] + [20] 50 50 200
Total
400
Source COBUILD Bank of English (CB) Lancaster-Oslo-Bergen corpus (LOB) [LOB] + [CB] CB CB
The central question of our analysis was whether the PM entity is a subsequent mention of a discourse referent or a first mention. Seeing the importance of distinguishing a range of discourse statutes, we propose five categories which form a continuum from discourse-given (coreferential possessive NP) to discourse-new (possessive NP introducing a new referent). Table 2. Discourse statuses of the PM referent COREFERENTIAL
the PM referent has been mentioned in the preceding discourse and is referred back to
TEXT REFERENCE
the PM referent is a text referent which is construed on the basis of the preceding discourse
INFERABLE
the PM referent is inferable from an associated referent or a scenario in the preceding context
ANCHORED
the PM referent is ‘anchored’ to (an) element(s) in the preceding discourse, which reduces its ‘newness’
NEW
the PM referent is newly introduced by the possessive NP
Table 2 gives an overview of these five discourse statuses, which, in the rest of this section, will be discussed in more detail along with the quantitative results of the corpus study for each of these categories.6 6
From the genitive sample, a few tokens had to be removed because the PM was a proper name, e.g. The electro sound, which grew out of Bambaataa’s pioneering
28 Peter Willemse, Kristin Davidse and Liesbet Heyvaert 4.1. The possessive NP is COREFERENTIAL with another NP in the preceding discourse As we saw in section 1, Taylor (1996: 217) proposed that PM referents are overwhelmingly new entities that are being introduced into the discourse. Our data show, however, that in a relatively small, though not insignificant, number of cases, the PM is a referent which has been mentioned in the preceding discourse and is referred back to by the possessive NP. In such cases, the PM is, in Chafe’s (1994) terminology, an ‘active’ referent, i.e. a referent which is given in the discourse and therefore readily available in the addressee’s consciousness. The possessive NP is thus, in such cases, coreferential with another NP and may even form part of a longer reference chain. In addition to the PR referent, which, in case of a definite genitive or a possessive pronoun, is explicitly marked as being retrievable, the PM may thus be a subsequent mention as well. Table 3 shows how frequent coreferential or given PM referents were in our data.
‘Planet Rock’, kept the street end of rap cooking (CB). As noted in the previous section, proper names do not construe the contrast between initial – subsequent mention. Another small set of tokens of the dataset was left out of the classification, because the possessive NP in these cases functioned as part of a fixed expression (often a prepositional phrase), e.g. (i) If Laurie noticed she chose not to comment, but she was curious in her way. (CB) (ii) In a stern reply, he reminded him that ‘Our troops are elated and confident; (CB) those on the enemy’s side cannot but be depressed.’ These expressions are not processed compositionally and, as a result, they do not have a distinct PM referent with a discourse status of its own. For each category, the number of data left out of the classification was the following: genitive (15); its (2); their (4); my (4); her (4). The percentages in the tables are calculated on the basis of the number of tokens included in the classification.
English possessives as reference-point constructions
29
Table 3. Results for the category coreferential data set
# of tokens
percentage
Total Genitive
28 /185 28 /185
15.14 % 15.14 %
Total Poss. Det.
1 / 48 4 / 46 8 / 46 3 / 46 16 / 186
2.08 % 8.70 % 17.39 % 6.52 % 8.60 %
44 / 371
11.86 %
Genitive [CB] its + N their + N my + N her + N
[LOB] [LOB] [CB] [CB]
Grand total
Coreferential possessives account for about 10% of the total data. Still, in a comprehensive account of the discourse status of possessive NPs, this makes them a non-negligible category. Let us consider some examples: (17)
Mr Ashby, a former name who suffered substantial losses at Lloyd’s, had sought damages over an article in January last year alleging that he shared a double bed with another man on holiday in Goa. And when the jury found against him in a majority verdict, he put his head in his hands and wept. […] After the verdict, the newspaper’s solicitor, Alistair Brett, said he would expect the present editor, John Witherow, to see the case as ‘a tragic family problem’ and be sensible about what to do now. […] Senior Tories expressed their determination to help him after the verdict and launched a campaign ‘to keep him buoyant’ that was immediately evident in his reception in the Commons. […] He has 28 days to appeal against the jury’s verdict and it is then likely to take up to nine months for his costs to be determined by taxation proceedings. (CB)
The referent ‘verdict’ is first introduced by the NP a majority verdict and subsequently referred back to twice with the definite NP the verdict, before finally being taken up again by the possessive NP the jury’s verdict. The possessive NP is in this case clearly part of a reference chain in the discourse and the PM referent is not a new referent. The relation between the PM and the PR is usually already established in the preceding context, before the two are actually constructionally linked up in the possessive NP. This is the case in (17), where the link between
30 Peter Willemse, Kristin Davidse and Liesbet Heyvaert ‘verdict’ and ‘jury’ is made clear in the preceding context (the jury found against him in a majority verdict), as well as in the following example: (18)
Britain’s leading arms manufacturers secretly liaised with ministers, civil servants and the CIA on ways to silence the Saudi opposition leader Muhammed al-Masari, it was claimed last night. […] Lawyers acting for Dr. al-Masari were preparing yesterday to appeal against his removal. […] Dr. al-Masari’s lawyers also allege that the home secretary decided to push ahead with the removal of the dissident in spite of a written pledge by the Home Office that his application to stay in Britain would be considered substantively. (CB)
The indefinite NP marked in bold type introduces the referent ‘lawyers’ while the postmodifier describes the link with the PR ‘Dr. al-Masari’. In some cases, the direct antecedent of the possessive NP is itself a possessive NP, as in the following example: (19)
[…] Her salvation was also a cause for celebration among the 60 volunteers and 12 staff who run Britain’s only national charity set up to find missing people. This Christmas, the runaway was one of 14,000 people on the charity’s computer database, which is housed in a spartan, donated office above a supermarket in East Sheen, southwest London. […] A lot of adolescent girls aged around 14 and 15 do not get on with their parents. […] Some fall prey to prostitution and others, among the most urgent on the charity’s database, become caught up in paedophile rings. (CB)
In such cases, the reference chain which the possessive NP forms part of consists of several possessive NPs, in which not only the PR (in this case ‘charity’) is linked up with all its previous mentions, but the PM (in this case ‘(computer) database’) as well (this use of possessive NPs is mentioned by Barker (2000: 214)). The coreferential use of possessive NPs has important implications for the reference-point model in its application to possessive constructions, and underlines the importance of studying their discourse context, instead of treating them as isolated syntagms and looking at their internal identification mechanism only. In actual usage, the reference-point mechanism can be adapted to specific discourse needs and used for particular ‘rhetorical’ purposes. Thus, the use of a possessive NP to refer to a given referent appears to be often motivated by the desire to ensure non-ambiguity of the referent in contexts where there is potential confusion. For example, in (17),
English possessives as reference-point constructions
31
the possessive construal makes it clear that reference is to those lawyers acting for al-Masari, and not to lawyers working for someone else. Another ‘rhetorical’ use of coreferential possessive NPs involves employing a different lexical classification to refer to a given referent. This may be a synonymous classification or it may (and often does) entail a more or less drastic recategorisation of the referent. Such cases are treated as coreferential because, despite the use of a different lexical classification, or ‘type specification’ as Langacker (1991: 144–148) calls it, the same referent is referred to. Blanche-Benveniste and Chervel (1966) have described such cases as anaphore infidèle (‘unfaithful anaphora’), as opposed to anaphore fidèle (‘faithful anaphora’), the latter being restricted to cases in which the referent is not recategorized. It is interesting to note that when a different type specification is used, it often incorporates additional contextually specified information about the referent. For instance, in (20), the underwear being referred to is first described as knickers whereas in the possessive NP, the type specification silk panties is used, which incorporates the information that the garment in question is made of silk, as indicated in the preceding context. (20) On Monday, Christie’s in New York is to sell Greta Garbo’s knickers. They are described with proper dignity. ‘A pair of silk, creamcoloured ladies’ briefs […] In fact they are a souvenir of what Christie’s delicately call ‘a night of romance’ between Garbo and the Mexican star Roland Gilbert. […] When they parted, Roland gave Garbo the gold ring he was wearing and was given the actress’s silk panties in exchange. (CB) Such incorporation of contextual information confirms the idea that reference is not only about referring ‘back’ to the previous textual mention of a referent, but rather about activating a mental representation of the referent, which naturally evolves and is enriched as the discourse progresses and new information is added (see, among others, Emmott (1997: chapter 7) and Brown and Yule (1983: 201–204)).7 Additionally, the use of a different 7
Brown and Yule (1983: 202) give the following example (involving pronominal reference): (i) Kill an active, plump chicken. Prepare it for the oven, cut it into four pieces and roast it with thyme for 1 hour. In such cases, they argue, although the identity of the referent ‘chicken’ does not change, its description does: “A reader who simply went back up the endophoric
32 Peter Willemse, Kristin Davidse and Liesbet Heyvaert type specification to refer to a referent which is already present in the discourse can also be used to categorize that referent according to the subjective evaluation of the speaker. This is the case in (21), where the referent ‘Bond’ is classified as a cardboard booby: (21)
With the Bond books, as his friend Ernest L. Cuneo wrote, Fleming’s ‘objective was the making of money’ and he succeeded. But it wasn’t until after his death in August 1964 that Bondmania erupted and the money really began to flow. Film was responsible. Goldfinger, the third Bond movie, was released in December of that year, and with it was founded an industry that would turn Fleming’s cardboard booby into a product that 30 years later rivals Mickey Mouse in terms of global penetration. (CB)
4.2. The PM referent is a TEXT REFERENT construed from the preceding discourse context As observed by Halliday and Hasan (1976: 52), anaphoric retrieval relations are not restricted to “an entity that is encoded linguistically as a participant” but may also involve “any identifiable portion of text”. Possessive NPs naming a referent which has been construed over a preceding stretch of text can also realize text reference, as in the following examples: (22) One bae insider said last week that there could be no formal discussions until Daimler and the Dutch government had resolved the problems of Fokker, the ailing short-haul aircraft maker that will collapse unless it receives an emergency cash injection of almost £ 1.4 billion. […] But Fokker’s crisis is only one contributor to the problems of Daimler-Benz Aerospace […] (CB) (23) Students may soon be offered American-style loans at cheap rates by leading banks and building societies, after the government promised lenders generous subsidies to enter the student-loan market. The plan […] aims to shift most of the cost of financing student loans on to high-street lenders. The Department for Education and Employment chain and substituted the expression an active plump chicken for it in the last clause would, in a significant sense, have failed to understand the text” (Brown and Yule 1983: 202).
English possessives as reference-point constructions
33
is urging financial institutions to tender ‘up to four’ licences to lend to students in return for loan finance at discounted rates of interest. The government would also pay the lenders a share of running costs and a percentage of unrecoverable debts. […] But the banks and building societies are unimpressed. None has said it would take up the government’s offer to lend to students. (CB) Examples (22) and (23) illustrate a distinction between two kinds of text reference in relation to possessive NPs. In one type, illustrated by (23), the possessive NP contains a metatextual noun, that is, the possessive NP summarizes and categorizes something which has been described in the preceding discourse as a semiotic phenomenon, i.e. a symbolically processed phenomenon. The second type, illustrated by (22), summarizes and categorizes something which has been described in the preceding discourse as a non-semiotic phenomenon, i.e. as a phenomenon (event, state, activity, etc.) in reality. Following Takahashi (1997: 63), we will call this the summative type. Table 4 shows how many of the possessive NPs in our database have text reference, either metatextual or summative. Table 4. Results for the category text reference data set
# of tokens
Genitive [CB] Total Genitive its + N their + N my + N her + N
[LOB] [LOB] [CB] [CB] Total Poss. Det.
Grand total
Percentage
metatextual T.S.
summative T.S.
metatextual T.S.
summative T.S.
10 / 185 10 / 185
12 / 185 12 / 185
5.41 % 5.41 %
6.49 % 6.49 %
0 / 48 1 / 46 1 / 46 4 / 46 6 / 186
0 / 48 3 / 46 0 / 46 1 / 46 4 / 186
0% 2.17 % 2.17 % 8.70 % 3.22 %
0% 6.52 % 0% 2.17 % 2.10 %
16 / 371
16 / 371
4.31 %
4.31 %
While the number of tokens for this category is relatively small in our corpus, the cases which occur shed an interesting light on possessive NPs, as they reveal their potential to categorize complex referents built up over longer stretches of text. Again, this demonstrates the rhetorical versatility of possessive NPs in interaction with the surrounding discourse.
34 Peter Willemse, Kristin Davidse and Liesbet Heyvaert Possessive NPs containing a metatextual type specification typically contain ‘semiotic nouns’, i.e. nouns which designate linguistically processed phenomena, such as decision, claim, remark, etc. Some examples are the following: (24) The Government has called on housing associations to raise more of their funds through the private sector and announced that housing association grants will be cut next year. John Battle, Labour’s Shadow Housing Minister, warned that Black pensioners will be badly hit by increasing housing association rents if Government grant cuts go ahead. […] Mr Battle, who addressed the National Conference on Housing Black and Minority Ethnic Elders last week, said added rents will soar, hitting pensioners already facing huge hikes in their fuel bills as a result of the VAT increases already announced in the Budget. A National Federation of Black Housing Organisations spokeswoman echoed Mr Battle’s sentiments. (CB) (25) Thanks to the federal prosecutor in Munich, Compuserve subscribers no longer have access to 200 dubious and distinctly sad Internet newsgroups catering for people who think sex is an activity that can be pursued through a mouse and a modem. The Germans, in a fit of prudishness, told Compuserve it would be prosecuted if it did not stop allowing their citizens to leapfrog through the commercial service into the internet which is quite beyond the company’s control and read the poison on their pcs. […] The prosecutor’s decision was plain stupid. (CB) In (23), offer in the possessive NP the government’s offer refers back to the special advantages and conditions offered by the British government to banks and building societies willing to grant loans to students, which have been described in the preceding discourse. In (24), sentiments in the possessive NP Mr Battle’s sentiments refers back to the opinions and ideas of John Battle which have been represented explicitly through indirect speech in the preceding discourse. Note that the possessive NP not only refers back to what has been stated in the text, but also categorizes (parts of) the preceding text as a text referent. Thus, a possessive NP realizing text reference not only refers back to a preceding stretch of discourse, but also actively construes a text referent in the sense that it categorizes and ‘labels’ the text it refers to. This is also the case in (25), where more inferencing is required from the addressee than in (23). The decision referred to by the possessive
English possessives as reference-point constructions
35
NP the prosecutor’s decision is not directly represented, but rather indirectly rendered through the description thanks to the federal prosecutor in Munich, Compuserve subscribers no longer have access (…). From this it can be inferred that what the prosecutor decided was to order Compuserve to remove offensive newsgroups from their network, and it is this decision which is referred to by the possessive NP. Possessive NPs with a summative type specification generally require some inferencing on the part of the addressee as well. Consider examples (26), (27) and (22) above: (26) A little over a month ago the Federal Government announced a $ 17.8 million grant to Indonesia to help combat a HIV/AIDS epidemic which may infect 2.5 million people by 2000. […] I spent a few weeks in Thailand in 1989 and am convinced bureaucrats misled the population into thinking the virus did not infect Asians, largely to protect their rich international sex-tour industry. If Thailand was lulling the population and the rather stupid sex tourists into a false sense of security, why not other Asian countries? So now we have to cough up $ 17 million to Indonesia alone trying to hold back the scourge. The irony is that a month after our generous grant was announced, Indonesian President Suharto’s son, Mr Tommy, bought a majority stake in Italy’s glamour sports car maker, Lamborghini. […] What he paid is anyone’s guess, a trifle more than our $ 17 million, I would think. Sounds mighty like he is in a position to match our generosity in his country’s fight against AIDS, does it not? (CB) (27) Among the regions, London heads the list as the centre with the lion’s share of venture-capital deals; 35 of companies involved in raising money last year were based in the capital. Scotland showed a high level of activity with 13 of the deals, closely followed by northwest England. The northeast, despite its high levels of inward investment, is not, on these statistics, generating much fresh entrepreneurial activity; it accounted for 2 of the deals. The capital’s dominance is confirmed by the regional breakdown for flotations. (CB) In (26), fight against AIDS summarizes in a relatively straightforward manner information given in the preceding discourse, indicated in bold type in the example. In (27), more inferencing is needed: the preceding text contains the statement that London heads the list [among the regions] and explicitly elaborates on the number of companies involved in venture-capital
36 Peter Willemse, Kristin Davidse and Liesbet Heyvaert deals based in London in comparison with such companies based elsewhere in Britain. This information is summarized by dominance, the interpretation of which at the same time requires the addressee to infer that the information given earlier implies that London is dominant in this particular area. In (22), finally, the information that Fokker needs a vital cash injection in order not to collapse, is categorized as a crisis, which requires the addressee to make the inference that for a business, such a situation qualifies as a crisis. As has become clear from the examples we have discussed, text reference involves a fair amount of inferencing on the part of the addressee. It is therefore a separate category in the classification of discourse statuses, distinct from coreferential possessives in which the PM refers back to a preceding nominal realization of a referent. On the continuum of discourse statuses text reference is right next to coreferential possessives, however, since it involves a PM referent corresponding directly to a stretch of text in the preceding discourse.
4.3. The PM referent is INFERABLE from the preceding discourse Table 5 lists the number of cases in which the PM referent is inferable from a referent or another element in the preceding discourse. Table 5. Results for the category inferable data set Genitive [CB] Total Genitive its + N their + N my + N her + N
[LOB] [LOB] [CB] [CB] Total Poss. Det.
Grand total
# of tokens
percentage
37 /185 37 /185
20.00 % 20.00 %
23 / 48 12 / 46 12 / 46 16 / 46 63 /186
47.92 % 26.09 % 26.09 % 34.78 % 33.87 %
100 /371
26.95 %
The conceptual relations causing inferability of a PM referent overlap to a considerable degree with those enabling bridging or associative anaphora, coded by NPs with definite articles (see section 3). The latter involve definite reference to a referent (signalling its recoverability) based on an indirect anaphoric relationship, i.e. a relation with another element (and not a
English possessives as reference-point constructions
37
previous mention of the same referent) in the preceding discourse. This preceding element is thus an ‘indirect antecedent’, three subtypes of which were discussed in section 3. Firstly, the antecedent can be another entity referent (i.e. a ‘thing’), which allows the inference of another entity on the basis of a part-whole relation (e.g. a car the steering wheel) or a more general association (e.g. a house the owner). The antecedent may also be an event or activity, from which an entity typically involved in it can be inferred (e.g. he has been murdered the police haven’t found the killer yet). Finally, the ‘antecedent’ may be a scenario or frame, evoking entities fulfilling specific roles in it (e.g. [eating at a restaurant] the waiter forgot to bring us the menu). In the case of bridging, the definite article signals that the referent of the NP is presumed to be recoverable by the addressee, and for the addressee to retrieve it, the link with the indirect antecedent needs to be grasped. The same conceptual relations of inferability between referents and discourse contexts can also motivate the use of a possessive NP, which, however, makes explicit that there is a link between two entities (the PR and the PM referent), and is thus in a sense an alternative for the construal with a NP with definite determiner.8 Thus, while in (28a), the relationship between ‘book’ and ‘pages’ is made explicit through the use of the possessive determiner its (referring back to the book), in (28b) the connection has to be inferred by the addressee in order to resolve the reference of the definite NP the pages. (28) a. As she sprang to her feet and ran to Alistair, the book fell to the (CB) floor, face downwards, its pages doubling up in disorder. b. […] the book fell to the floor, face downwards, the pages doubling up in disorder. If we want to recognize different degrees of discourse-newness of possessive NPs, PM referents inferable from other entity referents, activities or larger scenarios evoked by the context, can be viewed as constituting the first, lowest degree of discourse newness. The possibility of designating such inferable referents with NPs containing a definite article is evidence of this relatively low degree of newness, and it can also be used as a formal indication of which referents to include in this category. This can be illustrated with the following example: 8
We are not claiming that there is free variation between these two possible realizations of the inferable referent. Different semantic and pragmatic restrictions apply to NPs with prenominal possessives and definite determiners, and many instances of bridging have no possessive alternative and vice versa.
38 Peter Willemse, Kristin Davidse and Liesbet Heyvaert (29) The Leviathan of Parsonstown, said to be the world’s largest telescope, was built in 1845 by the 3rd Earl of Rosse at Birr Castle, Co Offaly. It will form the centrepiece of Ireland’s new Historic Science Centre which will be based at the castle. […] The rest of the scientific centre, which will include an exhibition of the astronomical work of the 3rd and 4th Earls and scientific galleries in the castle’s moat, will be completed by the end of the century. (CB) The PM referent of the possessive NP, ‘moat’, even though strictly speaking new, has a relation of strong association with the referent ‘castle’ which is mentioned in the preceding context (a castle is very often surrounded by a moat). On the basis of this conceptual relation the PM referent ‘moat’ is immediately recoverable in the sense that it can be inferred from the referent ‘castle’. In fact, in this example, the ‘moat’ could also have been referred to with a definite NP: (29’) […] an exhibition of the astronomical work of the 3rd and 4th Earls and scientific galleries in the moat. In cases like these, the possibility of replacing the prenominal possessive by a definite article can be used as formal test to establish the inferability of the PM referent. A question which immediately springs to mind when one is confronted with such examples is why the speaker chooses to use a possessive NP, rather than an NP with a definite article, to refer to a referent which is inferable from the discourse context. There are a number of possible pragmatic factors which motivate the use of a possessive in such contexts. In general, these have to do with the speaker aiming at more clarity: the inferential relation or ‘bridge’ that is there in the discourse is made explicit by the possessive NP. Speakers thus seem to choose the possessive in order to avoid potential confusion about the antecedent, as illustrated by (29), in which there are four clauses between the nearest mention of the referent ‘castle’ and the introduction of the referent ‘moat’, so that the link between the two might not be sufficiently salient at the moment when ‘moat’ is mentioned. Moreover, the referent ‘castle’ is not topical at the moment when ‘moat’ is introduced, since the four intervening clauses deal with the Victorian telescope located at the castle. Non-topicality of the indirect antecedent is indeed a second factor which influences the choice of a possessive rather than a definite NP in cases where bridging is possible. This also seems to be the main reason for the use of a possessive in (30):
English possessives as reference-point constructions
39
(30) Josephina Zacaroli-Walker is a businesswoman who became the star of a comic strip in a Japanese magazine. The strip’s originator simply fell for the blonde and blue-eyed charms of the woman who has been the public relations face of Land Rover abroad, and who is now promoting Rover’s saloons and hatchbacks as well as the new MGF around the world. (CB) Distance between antecedent and associated entity is in this case clearly not the problem, as the referent ‘(comic) strip’ is mentioned right before the associated referent ‘originator’. However, the referent ‘comic strip’ is syntactically embedded in an of-phrase in the preceding clause, which reduces its accessibility. Consequently, although an NP with definite article is possible (The originator simply fell for the blonde and blue-eyed charms…), the speaker opts for a possessive realization of the referent. However, the data show that a clear motivation for the use of a possessive NP to refer to a referent which could have been realized by a bridging NP is certainly not always present. Consider, for instance, the following example: (31) But even this hardly prepared one for the spectacle that the house itself presented on closer view. It stood, as it were, knee-deep in weeds – like some forlorn prehistoric creature in an edible pasture. Its grey surfaces were flaked and cracked; its woodwork was denuded of paint; many of the lower windows showed tattered curtains pulled awry, and some of the upper ones lacked entire panes of glass. […] If challenged to date it, Appleby would have said 1718; if challenged to the name of the builder, he would have said James Gibbs. But now it spoke either of madness – which, indeed, was what was attributed to its owner – or of penury. (LOB) The inanimate referent ‘house’ is clearly the topic of this passage. Moreover, all of the underlined NPs can be analyzed as realizing bridging reference, since they name parts (surfaces, woodwork, windows) or associated human referents (builder, owner). However, two of them are anaphorically bridged, whereas the others are realized by means of a possessive NP. The antecedent is clear for all of these referents, and no obvious other differences between e.g. lower windows (referred to by means of bridging) and woodwork (referred to with a possessive NP) or between builder (bridging) and owner (possessive) can account for their different realization. It seems, then, to be a matter of personal choice on the part of the speaker for one construal or the other in many cases.
40 Peter Willemse, Kristin Davidse and Liesbet Heyvaert The possibility of replacing the prenominal possessive by a definite article is an important indication of the inferability of the referent but, as a formal test, it has to be handled with some caution. It seems to apply pretty systematically when the indirect antecedents are inanimate ‘wholes’. For instance, its (example 31) and their referring to inanimate PRs (the books their covers) can generally be replaced by the. However, the picture is somewhat more complicated with possessive NPs referring to body parts of human PRs. In English, body parts of humans cannot generally be introduced with a definite article-NP based on the preceding mention of the ‘possessor’ (cf. Du Bois 1980: 241); realization with a possessive NP is often obligatory. This is illustrated by the following examples: (32) a. She is on sick leave because she has broken her leg . b. *She is on sick leave because she has broken the leg . (33) a. During cross-examination the woman said her eyes were not open all the time, but she was sure she did not fall asleep. (CB) b. *During cross-examination the woman said the eyes were not open all the time, but she was sure she did not fall asleep. However, certain constructions – namely external possession constructions – do allow for a body part to be introduced with a definite NP: (34) The burglar hit John on the head with a baseball bat. (35) She took me by the hand and led me to the study. In these examples, reference to the body parts by means of a definite NP is constructionally determined and a considerable degree of idiomaticity is involved, especially in (35). One is reminded here of Ariel’s (1990: 184) point that the coding of inferable referents is often governed by “languagespecific conventions”. Still, such examples indicate that body part-relations are formally treated as inferable (allowing bridging reference) in English in a number of cases. This is not surprising, since body parts stand in a conceptually close, ‘inalienable’ relation to their PRs (see Chapell and McGregor 1996). Another indication of the inferable nature of body parts is that when the ‘possessor’ is non-human, and particularly when the ‘possessor’ is a dead animal, bridging to a body part is possible, as illustrated by the following examples:
English possessives as reference-point constructions
41
(36) When hounds had run a stag to a standstill, the stag would turn to defend itself. And few dogs fancied first toss on the antlers, so would stand baying for the huntsmen to come in for the kill. (CB) (37) Gut the mackerel, slit down the belly and open them out kipper-style, or skewer fillets, season and grill. (CB) Because of the conceptually inalienable nature of the relation between body-parts and their PRs and the fact that they can be referred to by NPs with the in a number of cases, we have classified them with the inferable PM referents. What is the import of the analysis proposed here to the overall analysis of the discourse status of PM referents? We argue that when the PM referent is inferable from the surrounding discourse, it cannot be analyzed as entirely new to the discourse. For instance, although strictly speaking, the referent ‘moat’ in (29), for instance, is mentioned for the first time, it is inferable on the basis of its conceptual (part-whole) relation with ‘castle’. However, since it has no direct antecedent but only an indirect one, it is not strictly discourse-given (i.e. coreferential) either. Moreover, in the hierarchy of discourse statuses it is below text reference (i.e., closer to ‘newness’ than text reference). The latter type of reference involves a referent which directly corresponds to and categorizes a preceding stretch of text. Inferable referents, by contrast, involve a relation that holds between two different referents, between an event or activity and a referent which it is involved in it or between a scenario or frame and a referent which fulfils a role in it.
4.4. The PM referent is ‘ANCHORED’ to an element in the preceding discourse This fourth category is transitional between the previous category of inferable PM referents and the next one of new PM referents. It contains cases of first mention PM referents that have a conceptual relation to other referents in the discourse but one that is not strong enough to make them precisely inferable. Looking at these conceptual relations prospectively, one could say that the element to which the PM referents are related does not predict them as a unique instance or inclusive set. As a consequence, PM referents of this category do not systematically alternate with definite NPs with possessive postmodifiers, but can also alternate with indefinite NPs + of + PR. In other words, in this category we find functionally indefinite
42 Peter Willemse, Kristin Davidse and Liesbet Heyvaert possessive NPs, as illustrated by examples (4), (6) and (15) above. We will refer to this class as ‘anchored’ PM referents: they are conceptually anchored to other elements in the discourse – to which, indeed, possessive NPs ‘anchor’ them via their internal reference point mechanism – and, as a consequence, cannot be classified as entirely new. At the same time, they differ from the inferable PM referents. As we saw in section 4.3, these have a stronger conceptual relation to their antecedents in the discourse, which inherently guides the person processing the text to unique instances or to inclusive sets of instances, and excludes functionally indefinite NPs. Specifically, two main kinds of anchoring relations can be distinguished: on the one hand, kinship and other interhuman relations and, on the other hand, alienable possessions. Table 6 gives an overview of the number of cases in which the PM maintains a relation of ‘anchoring’ with the discourse context. Table 6. Results for the category anchoring relation ‘interhuman’
‘alienable possession’
‘interhuman’
‘alienable possession’
11 / 185 11 / 185
26 / 185 26 / 185
5.94 % 5.94 %
14.05 % 14.05 %
0 / 48 2 / 46 13 / 46 8 / 46 23 / 186
14 / 48 9 / 46 5 / 46 3 / 46 31 / 186
0.00 % 4.35 % 28.26 % 17.39 % 12.37 %
29.17 % 19.57 % 10.87 % 6.52 % 16.67 %
34 / 371
57 / 371
9.16 %
15.36 %
The first main type of anchoring relation is formed by kinship and similar interhuman relations. They are evoked by certain nouns with an inherently relational meaning (see Barker 1991 [1995], 2000; Fraurud 1996). Taylor (1989: 675) gives a list of nouns which in addition to kinship nouns “invoke, in their semantic structure, various other kinds of interpersonal relationships”, including friend, fiancée, colleague, guest, fellow student, competitor, confidant, etc. Taylor (1989) discusses such relational nouns in terms of how they ‘steer’ the interpretation of the relationship between PR and PM by evoking an unprofiled relationship in their semantic structure, an element of which is elaborated by the PR nominal. Possessive NPs whose PM is a relative, friend or associate can be functionally indefinite, as illustrated by (4), a friend’s friend, described by letter, should turn out an unrecognizable stranger, above, (38) and (39).
English possessives as reference-point constructions
43
(38) Created in 1968 and based on a true story, Ashton depicts Elgar’s friends visiting him after he had sent his score of his Variations to the famous Viennese composer Ritcher, to try and interest him in the work. (CB) (39) Dwight wrote Victor Serge that subscriptions were pouring in and “all kinds of people were being stimulated to write for it […]”. […] Dwight’s associates and contributors wrote long and constructive criticisms. (CB) The covert indefiniteness of the possessive NPs in these examples is made overt if we recode them in their periphrastic form with of: a friend of a friend, friends of Elgar, associates and contributors of Dwight (in these last two examples, reference does not appear to be made to all of Elgar’s friends or all of Dwight’s associates and contributors). However, such NPs are functionally definite if there is only one instance (or only one in the given context) corresponding to the relation, e.g. (40), or if all contextually available instances are referred to, as in (41). (40) In the months leading up to the killing, Avent’s wife became pregnant and he lost his job as a supermarket shelf stacker because he stole $300. (CB) (41) Mack’s friends were all sixteen or seventeen. (CB) In informal spoken English, unique or inclusive reference to close of kin can even be realized by bridging NPs, as illustrated by (42) and (43). This suggests that the distinction inferable – anchored referent being proposed here is of a continuum-like nature. (42)
H h how do you get to the Fab Club? <M01> Well FX c erm the d the daughter comes and picks us up (CB) or we get taxis.9
9
The codes in this example are part of the transcription of spoken language used in the COBUILD corpus. and <M01> indicate the two participants in the dialogue; and indicate a repetition of (parts of) words, and ‘FX’ and ‘MX’ replace, for privacy reasons, female and male proper names mentioned in the dialogue.
44 Peter Willemse, Kristin Davidse and Liesbet Heyvaert (43) When a man loses his job, it deals a sickening blow to his ego and sense of identity. Feelings of failure and rejection strain family relationships to breaking point and it’s often the wife who’s left to pick up the pieces. (CB) Alienable possessions form the second main type of anchored PM referents. Possessive NPs designating such referents can also be either functionally indefinite or definite. Thus, (44) and (45) are functionally indefinite: they do not refer to a uniquely identified possession or to an inclusive set of belongings. (44) No one checked up on us. At 14, I was cutting school in the lunch hour and getting drunk in the afternoons usually at a friend’s house. (CB) (45) She smiles with some regret when the fans rip Brett’s shirts to shreds during The Drowners. He’s lost so many beautiful clothes; […] (CB) By contrast, (46) and (47) are functionally definite, as they refer to the one car driven by Rachelle in (46), and to all contextually given instances of garments of Christ in (47). (46) […] “I saw Rachelle’s car slow in the final and figured she must have struck problems,” Cowin said. (CB) (47) There is a long-established anecdote concerning the two angels who are holding Christ’s garments in Andrea del Verrochio’s painting, The Baptism of Christ. (CB) In sum, possessive NPs referring to kinship and interpersonal relations, and to alienable possessions mark the point on the continuum at which the PM referent is no longer intrinsically inferable from the preceding discourse. Even though this category is not very frequently attested by possessive NPs, they still should not be brushed under the carpet. The occurrence of functionally indefinite possessive NPs in this category is one of the main arguments against the analysis of possessive NPs as definite NPs. 4.5. The PM referent is a NEW referent The quantitative results for the category of new PM referents in our data are given in Table 7.
English possessives as reference-point constructions
45
Table 7. Results for the category NEW data set Genitive [CB] Total Genitive its + N their + N my + N her + N
[LOB] [LOB] [CB] [CB] Total Poss. Det.
Grand total
# of tokens
percentage
61 / 185 61 / 185
32.97 % 32.97 %
10 / 48 15 / 46 7 / 46 11 / 46 43 / 186
20.83 % 32.61 % 15.22 % 23.92 % 23.12 %
104 / 371
28.03 %
These results show that although it is true that possessive NPs newly introduce referents into the discourse in a number of cases (28% of our corpus), new referents certainly do not represent the majority of PM referents. This last category forms the ‘new’ end of the continuum of discourse statuses that we propose. It contains cases in which the PM is a new referent in the discourse, not only in the sense of not having been mentioned as such in the preceding discourse, but also in the sense of not being inferable or predictable from elements in the preceding discourse. In many of these cases, the PM is realized by a nominalization or deverbal nouns, as in (48) or (49) below. (48) [about the Turnable Emergency Non-capsizable Triangular System] It can survive punctures in two of its surfaces and still remain afloat. Hunter has produced two prototypes and is in talks with a lifeboat manufacturer that could lead to the system’s launch in the spring of next year. (CB) In these examples, the PR nominal designates the subject of the action described by the nominalization. While the PR referent typically maintains an anaphoric relation with the preceding discourse, the action designated by the ‘PM noun’ is not inferable from the preceding text. In some of these examples, there is some sort of general semantic relation between the PM referent and the preceding discourse context, but this relation is not strong enough to confer (even partial) givenness on the referent. This is for instance the case in example (49), in which the PM referent fits in with the scenario that is being evoked in the context, without being a predictable part of this scenario. That no ‘givenness’ of the PM referent is present in
46 Peter Willemse, Kristin Davidse and Liesbet Heyvaert such cases is confirmed by the fact that an introduction of the referent by means of a definite article-NP is excluded. (49) Queensland cricket captain Stuart Law yesterday warmed up for this week’s Sheffield Shield battle against New South Wales with a mighty double century for Valley in a XXXX Brisbane club cricket game yesterday. […] In a punishing display, Law smacked 32 boundaries including two sixes, and faced 266 balls. He came to the wicket when the score was eight after opener Michael Ephraims departed, and immediately went on the attack. […] There was little respite for University’s bowling attack which included Sheffield Shield bowlers Michael Kasprowicz and Peter Jackson. (CB) In (49), the context evokes the scenario of a cricket game, in which a ‘bowling attack’ is possible, though by no means necessary; consequently, the scenario does not lead to identifiability of the referent ‘bowling attack’, which is being introduced (for the sake of clarity, the bowling attack referred to is not the attack implied earlier (Law… immediately went on the attack), since it is being executed by the team playing against Law’s team). With regard to this category of the ‘newest’ PM referents, the reflection imposes itself that such nominalizations, despite being commonly used with specific textual goals in certain registers such as newspaper and scientific language, are not usually considered very typical possessive NPs. 5. Conclusion In this paper we set out to investigate the underresearched issue of the discourse status of the PM referent of possessive NPs in English. We investigated the ways in which this referent can be an initial or non-initial mention, paying attention to how it interacts with previously given information in extensive discourse contexts. We demonstrated that a binary distinction between given and new does not suffice to capture its discourse status, and proposed instead a continuum of discourse statuses, ranging from fully discourse-given to fully discourse-new over a number of in-between statuses in which the PM is only partially given. Our analysis has the following major implications. Firstly, our analysis reviews and refines the description of English possessive NPs in a number of ways. We have shown that possessive NPs cannot be analyzed as mere definite NPs which presuppose the identifiability
English possessives as reference-point constructions
47
of their referent. A number of possessive NPs are functionally indefinite, which argues for a covert definite-indefinite contrast in English. Building on earlier analyses of possessive NPs as complex NPs with two distinct referents among which an NP-internal relation of identification exists (cf. Langacker 1993, 1995; Taylor 1996; Martin 1992), we have shown that this NPinternal identification mechanism is distinct from the external discourse functioning of each of the two referents of the possessive NP. Whereas the discourse status of the PR referent is straightforward (it is given with possessive determiners and with definite genitives, and new with indefinite genitives), the discourse status of the PM referent is controversial and has remained underresearched empirically. Against the view of Taylor (1996) (who, projecting the asymmetric reference-point relation internal to NPs onto their discourse behaviour, predicts that PMs are overwhelmingly discourse-new and anchored to a typically given PR), we have shown that fully discourse-new PM referents represent only a relatively small portion (28%) of the data. PM referents may also be ‘anchored’ to elements in the preceding text, which reduce its newness but do not make it recoverable in a strict sense. In the majority of the cases, however, the PM is given at least to some extent. It may be retrievable from extended descriptions, quotes or reports in the surrounding discourse, and it may be inferable from a given referent or other element in the preceding discourse. In a small but significant number of cases (11% of our data), finally, the PM turned out to be fully given, in the sense of being coreferential with a given discourse referent. The analysis of possessive NPs proposed in this paper has theoretical implications as well. In particular, it adds a systematic discourse perspective to the theory of reference-point constructions (Langacker 1993, 1995; Taylor 1996), which has always focused on the NP-internal reference-point relation. We have shown that possessive NPs cannot be studied in isolation, without taking account of the discourse context in which they occur. By studying possessives in extensive discourse contexts, we have revealed that the reference-point mechanism may be employed in various ways, adapted to specific discourse purposes. For instance, possessive NPs may be used to guarantee referential clarity by their explicit link to the PR or to reactivate the PR as topic, and they may be employed to recategorize existing referents, or to label text referents. In this way, our analysis has shown that the possessive is a versatile and complex NP type, whose identificatory potential as a reference-point construction can only be fully uncovered by taking its discourse functions into account.
48 Peter Willemse, Kristin Davidse and Liesbet Heyvaert Acknowledgements We would like to thank the Research Council of the University of Leuven for the postdoctoral grant PDM/05/43 given to Peter Willemse for the academic year 2006–2007. We also greatfully acknowledge the support by grant no. HUM2007-60706/FILO of the Spanish Ministry of Education and Science and the European Regional Development Fund. This article is based on a chapter from Peter Willemse’s doctoral dissertation (Willemse 2005: 59–136), which was supervised by Kristin Davidse and Liesbet Heyvaert. We owe a debt of gratitude to John Taylor, Theo Janssen and Kurt Feyaerts, who acted as members of the doctoral examination board, for useful comments and suggestions. Our particular thanks go to Bill McGregor for his very generous and incisive comments which identified some crucial points to be addressed. Needless to say, we are the only ones responsible for remaining errors of thought in the final version. References Abbott, Barbara 2004 Definiteness and indefiniteness. In The Handbook of Pragmatics, Laurence R. Horn and Gregory Ward (eds.), 122–149. Oxford: Blackwell. Ariel, Mira 1990 Accessing Noun-Phrase Antecedents. London: Routledge. Barker, Chris 1991 Possessive descriptions. PhD thesis, University of California Santa Cruz [published 1995, Stanford: CSLI]. Biber, Douglas, Stig Johansson, Geoffrey Leech and Randolph Quirk 1999 Longman Grammar of Spoken and Written English. London: Longman. Blanche-Benveniste, Claire and André Chervel 1966 Recherches sur le syntagme substantif. Cahiers de Lexicologie IX (2): 3–33. Brown, Gillian and George Yule 1983 Discourse Analysis. Cambridge: Cambridge University Press. Chafe, Wallace L. 1972 Discourse structure and human knowledge. In Language Comprehension and the Acquisition of Knowledge, Roy O. Freedle and John B. Carroll (eds.), 41– 69. Washington: Winston. Chafe, Wallace L. 1976 Givenness, contrastiveness, definiteness, subjects and topics. In Subject and Topic, Charles N. Li (ed.), 25–55. New York: Academic Press.
English possessives as reference-point constructions
49
Chafe, Wallace L. 1994 Discourse, Consciousness, and Time. Chicago: The University of Chicago Press. Chafe, Wallace L. 1996 Inferring identifiability and accessibility. In Reference and Referent Accessibility, Thorstein Fretheim and Jeannette K. Gundel (eds.), 37–46. Amsterdam: John Benjamins. Chappell, Hillary and William B. McGregor 1996 The Grammar of Inalienability: A Typological Perspective on Body Part Terms and the Part-Whole Relation. Berlin /New York: Mouton de Gruyter. Declerck, Renaat 1986 Two notes on the theory of definiteness. Journal of Linguistics 22: 25–39. Du Bois, John W. 1980 Beyond definiteness: the trace of identity in discourse. In The Pear Stories: Cognitive, Cultural and Linguistic Aspects of Narrative Production, Wallace L. Chafe (ed.), 203–274. Norwood: Ablex. Emmott, Catherine 1997 Narrative Comprehension: A Discourse Perspective. Oxford: Oxford University Press. Fraurud, Kari 1990 Definiteness and the processing of noun phrases in natural discourse. Journal of Semantics 7: 395–433. Fraurud, Kari 1996 Cognitive ontology and NP form. In Reference and Referent Accessibility, Thorstein Fretheim and Jeannette K. Gundel (eds.), 65–87. Amsterdam: John Benjamins. Gundel, Jeannette K., Nancy Hedberg and Ron Zacharski 1993 Cognitive status and the form of referring expressions in discourse. Language 69: 274–307. Gundel, Jeannette K., Nancy Hedberg and Ron Zacharski 2001 Cognitive status and definite descriptions in English: why accommodation is unnecessary. Journal of English Language and Linguistics 5: 273–295. Halliday, Michael A.K. and Ruqaiya Hasan 1976 Cohesion in English. London: Longman. Hawkins, John A. 1978 Definiteness and Indefiniteness: A Study in Reference and Grammaticality Prediction. London: Croom Helm. Huddleston, Rodney 1988 English Grammar: An Outline. Cambridge: Cambridge University Press.
50 Peter Willemse, Kristin Davidse and Liesbet Heyvaert Langacker, Ronald W. 1991 Foundations of Cognitive Grammar, Volume 2: Descriptive Application. Stanford: Stanford University Press. Langacker, Ronald W. 1993 Reference-point constructions. Cognitive Linguistics 4: 1–38. Langacker, Ronald W. 1995 Possession and possessive constructions. In Language and the Cognitive Construal of the World, John R. Taylor and R. E. MacLaury (eds.), 51–79. Berlin /New York: Mouton de Gruyter. Langacker, Ronald W. 2002 Remarks on the English grounding systems. In Grounding: The Epistemic Footing of Deixis and Reference, Frank Brisard (ed.), 29–38. Berlin /New York: Mouton de Gruyter. Lyons, Christopher 1999 Definiteness. Cambridge: Cambridge University Press. Martin, James R. 1992 English Text: System and Structure. Amsterdam: John Benjamins. Quirk, Randolph, Sidney Greenbaum, Geoffrey Leech and Jan Svartvik 1985 A Comprehensive Grammar of the English Language. London: Longman. Rosenbach, Annette 2002 Genitive Variation in English. Berlin /New York: Mouton de Gruyter. Sanford, Anthony J. and Simon C. Garrod 1981 Understanding Written Language: Explorations of Comprehension beyond the Sentence. Chichester: John Wiley and Sons. Sanford, Anthony J. and Simon C. Garrod 1998 The role of scenario mapping in text comprehension. Discourse Processes 26 (2 /3): 159–190. Takahashi, Hidemitsu 1997 Indirect anaphors: definiteness and inference. Leuvense BijdragenLeuven Contributions in Linguistics and Philology 86: 53–80. Taylor, John R. 1989 Possessive genitives in English. Linguistics 27: 663–686. Taylor, John R. 1991 Possessive genitives in English: a discourse perspective. South African Journal of Linguistics 9 (3): 59–63. Taylor, John R. 1996 Possessives in English. Oxford: Oxford University Press. Willemse, Peter 2005 Nominal reference-point constructions: possessive and esphoric NPs in English. PhD thesis, University of Leuven.
On the co-variation between form and function of adnominal possessive modifiers in Dutch and English Jan Rijkhoff
1. Introduction This contribution is concerned with Dutch and (to lesser extent) English possessive modifiers introduced by the preposition of (Dutch van), as in a woman OF INFLUENCE or (Dutch) de auto VAN MIJN BROER (the car OF MY BROTHER) ‘my brother’s car’. The main goal of this paper is to demonstrate that the remarkable variation in the grammatical properties of this possessive construction directly correlates with the kind of modifier function it has in the noun phrase. It is first shown that lexical possessive modifiers with van /‘of’ (“adnominal possessives” for short) are used to express most of the modifier functions recognized in a semantic, five-layered model of the noun phrase (section 2). I will then argue that the values for certain grammatical parameters (here subsumed under the labels MODIFICATION, PREDICATION, REFERENCE) are determined by the kind of modifier function the adnominal possessive has in the noun phrase (section 3 and 4); a tentative explanation is given in section 5. The more general point this paper wants to make is that functional modifier categories like CLASSIFYING MODIFIER, QUALIFYING MODIFIER, and LOCALIZING MODIFIER can be characterized in grammatical terms, which makes it possible to capture grammatical differences between members of the same form class (such as prepositional phrases, as shown in Table 3) and grammatical similarities between members of different form classes (e.g. adjectives and prepositional phrases, see Tables 1a–b) within and across languages. This paper is restricted to possessive modifiers of nouns that denote concrete objects.
52 Jan Rijkhoff 2.
Layering and adnominal possessives
2.1. The layered organization of linguistic structures Layered representations of clausal structures, which are designed to reflect scopal differences amongst modifiers categories, have been proposed in various theories of grammar (for a short overview, see Butler 2003, Part 1: 239–249). In Dik’s Functional Grammar (FG), layering was introduced by Hengeveld (1988, 1989, 1990), whose proposals were immediately incorporated into the general FG framework (Dik 1989, 1997). Subsequently Rijkhoff (1988, 1990, 1992, 2002) put forward a layered model of the noun phrase (NP), arguing that, to some extent, NPs and clauses can be analyzed in a similar fashion. The most recent version of this layered NP/clause model is shown in Figure 1. Grammatical modifiers (e.g. cardinal numeral or number marker/ω2, demonstrative/ω3, (in)definite article/4) are also called OPERATORS and in the NP they are symbolized by ω (Representational Level) or (Interpersonal Level); clausal operators (such as aspect/ o, tense/ 3, (ir)realis/4 , mood/5 markers) are symbolized by or . Lexical or phrasal modifiers, which are also called SATELLITES, typically involve content words (Verb, Noun, Adjective, Adverb). NP satellites are represented by the symbols τ (Representational Level) or (Interpersonal Level), which may stand for adnominal adjectives (a BIG car), NPs (with or without an adposition or case marker – the vase ON THE TABLE), or relative clauses (the man WHO STOLE MY BICYCLE), clausal satellites take the form of various kinds of adverbs and adverbials and are symbolized by σ or .1 Subscripts 0–6 indicate the various layers at which modifiers are specified. The layers are organized hierarchically, which means that illocutionary modifiers 6/6 have the widest scope (the whole clause), whereas the scope of classifying modifiers (ω0/τ0 in the NP, 0/σ0 in the clause), which only further specify the kind of thing or event referred to, is restricted to the head constituent of the NP or clause (typically the head noun and the verb respectively). Figure 2 shows more clearly that (up to a point, at least) NPs and clauses can be analyzed along the same lines. For example, adnominal demonstratives (ω ω 3) specify the location of an entity (thing) in space, whereas tense markers (3) specify the location of an entity (event) in the temporal dimension. The various boxes indicate differences in semantic scope. 1
See Rijkhoff (2002) for details; the most recent accounts are Rijkhoff (2008a, b, d).
On the co-variation between form and function of adnominal possessives 53 INTERPERSONAL LEVEL (‘LANGUAGE AS EXCHANGE’) At the Interpersonal Level, modifiers are concerned with the Interpersonal Status of four kinds of entities in the World of Discourse: [i] clauses (or rather the messages contained in the clauses), [ii] propositions, [iii] events and [iv] things. MODIFIERS IN THE NOUN PHRASE
MODIFIERS IN THE CLAUSE
5. Proposition modifiers (5, 5) inform A about S’s personal assessment of / attitude towards a proposition Xi as regards the probability, possibility or desirability of the actual occurrence of event ei. 4. Discourse-Ref. modifiers (4, 4) 4. Discourse-Ref. modifiers (4, 4) specify the existential status of thing xi or event ei in the World of Discourse.
Scope increase
6. Illocutionary modifiers (6, 6) inform Addressee about the illocutionary status of the clause (Decl, Int, …).
REPRESENTATIONAL LEVEL (‘LANGUAGE AS CARRIER OF CONTENT’) At the Representational Level, modifiers specify properties of spatiotemporal entities (things, events) in the World of Discourse in terms of the notions Kind (Class), Quality, Quantity, and Location. 3. 2. 1. 0.
Localizing modifiers Quantifying modifiers Qualifying modifiers Classifying modifiers
(ω3, τ3) (ω2, τ2) (τ1) (ω0, τ0)
3. 2. 1. 0.
Localizing modifiers Quantifying modifiers Qualifying modifiers Classifying modifiers
( 3, σ3) ( 2, σ2) (σ1) ( 0, σ0)
Figure 1. Layers of modification in the noun phrase and in the clause (DiscourseRef. = Discourse-Referential; Greek characters symbolize grammatical [ω/ , /] and lexical [τ /, σ /] modifiers in the NP and the clause).
54 Jan Rijkhoff Clause operators ( /)
Clause satellites ( /) 4. Discourse-Referential 3. Location 2. Quantity 1. Quality 0. Kind EVENT clause (head: verb/main predicate
NP (head: noun) THING 0. Kind 1. Quality 2. Quantity 3. Location 4. Discourse-Referential NP operators (/ )
NP satellites (/ )
Figure 2. Layers of modification: parallels between the layered organization of NPs and clauses
As can be seen in Figures 1 and 2, there is a certain asymmetry in the distribution of operators and satellites in that QUALIFYING OPERATORS (ω1, 1) are not deemed to exist (Rijkhoff 2008a,d). By definition, members of a grammatical word class (OPERATORS) constitute a smallish, closed set of items, capturing only a limited number of crucial, absolute distinctions (Dik 1997 Part 1: 160). For example, if Definiteness (4) is a grammatical category in some language, the only two available choices are +Definite or – Definite (indefinite). Since qualities are gradable properties (e.g. big/bigger/ biggest; fairly/rather/very big), they can only be expressed by satellites (lexical modifiers): a house can be rather big but not rather SINGULAR. Even though there can be certain form-function correlations (for instance, if a language has a distinct class of adjectives, they are typically used as qualifying satellites), there is no strict one-to-one relationship between the formal category of a noun modifier (e.g. adjective, prepositional phrase) and its modifier function in the noun phrase (Figure 2). This holds in particular for satellites (cf. Dik 1997 Part 1: 208). CLASSIFYING SATELLITES in the NP (τ0) further specify what KIND of entity (thing, event) is being referred to by the speaker (e.g. a corporate lawyer, a man of prayer), whereas QUALIFYING SATELLITES (τ1) specify more or less inherent properties such as size (a big house, telescopes of enormous size), value or quality
On the co-variation between form and function of adnominal possessives 55
(a cheap suit, wine of an incredible richness), age (a young child, youths under age 16), or color (blue curtains, a Jovian moon of incredible redness). These examples illustrate, once again, that the same function (CLASSIFYING MODIFIER) can be performed by members of different form classes (here: adjectives and prepositional phrases or PPs with ‘of’). QUANTIFYING SATELLITES (τ2) in the NP specify the cardinality of an entity through phrasal or lexical means (as in example (2) from Samoan) and LOCALIZING SATELLITES (τ3) are concerned with (broadly speaking) locative properties of the referent in conversational space, as in the book on the table, but also the book that you just gave me or my book. A LOCALIZING SATELLITE basically serves to make the referent of the matrix NP (book) locatable and consequently identifiable for the addressee. DISCOURSEREFERENTIAL SATELLITES (4), finally, are not so much concerned with notions like Class, Quality, Quantity or Location, but rather with the pragmatic or interpersonal status of the entity, specifying whether or not some entity already exists in the world of discourse (Definite vs. Indefinite), or whether the speaker refers to the same or a different token of a certain type (the same book, another book).2 The fact that members of the same formal modifier category (adjective, NP, PP, relative clause) can be used in different modifier functions in the NP is also exemplified below, where a relative clause serves as a qualifying, a quantifying or a localizing satellite. Localizing satellites typically contain reference to a familiar entity, which makes it possible to identify the referent of the matrix NP in the shared world of discourse, as in the following example (with the referential anchor underscored; more on this in section 3.4). (1)
Could you pass me the book that is lying next to you?
In Samoan a special kind of relative clause is used to specify cardinality. In (2) the numeral appears as the head of a relative clause construction that is introduced by the general tense-aspect-mood marker [GENR] e if the NP has specific reference.
2
Synchronic and diachronic relationships between modifiers of the clause and the NP are discussed in Rijkhoff and Seibt (2005) and Rijkhoff (2008a, d).
56 Jan Rijkhoff
(2)
Samoan Sa fau=siae e Tagaloaalagi fale e tolu … PAST build=ES ERG Tagaloaalagi house GENR three … ‘Tagaloaalagi built three houses …’ (Mosel and Hovdhaugen 1992: 318)
Galela (a Papuan language spoken in Halmahera, Indonesia) uses participial forms of stative verbs to express quality concepts.
(3)
Galela awi d ohu i lalamo his foot it be.big.PRT ‘his big foot’
(van Baarda 1908: 35)
Possessives, too, can be used in a variety of modifier functions. This is discussed in the remaining sections. 2.2. Adnominal possessives Possessive constructions or “genitives” are notorious for the wide range of meanings they cover (Williams 1981: 89). The paperback edition of Longman’s Dictionary of Contemporary English alone lists twenty-five different senses of the possessive preposition of. The Dutch possessive preposition van ‘of’ is at least as polysemous as its English counterpart and is currently even used as a quotative marker (Hengeveld 1994): (4)
Ik dacht van ga hem eens huren. I thought of go him once rent ‘I thought: let’s just rent it [i.e. a video – JR]’ (cf. Ik dacht: “Ga hem eens huren”)
There is no consensus as to why the possessive construction is so extremely versatile within and across languages, but it seems this is at least partly due to the fact that the wide ‘belonging to’ relation covered by possessives plays a central and fundamental role in human cognition (Seiler 1983, Heine 1997). So as to restrict to domain of research to a manageable size, I will mostly be concerned with lexical adnominal possessives in Dutch and English that are introduced by the preposition van /‘of’ and that modify a common noun
On the co-variation between form and function of adnominal possessives 57
denoting a concrete physical object (such as car, dog or tree). Here are some examples (all non-asterisked examples were found on the Internet using Google or Ask):3 CLASS (see section 3.2) (5)
een man van God a man of God ‘a man of God’ (i.e. a priest, a clergyman)
(6)
een man van het toneel a man of the stage ‘an actor’ (especially a dedicated actor)
(7)
een man van de geest (the article is optional; see note 4) a man of the mind ‘A philosopher’ (or simply an intellectual or religious person)4
(8)
een meisje van plezier a girl of pleasure ‘a girl of pleasure’ (i.e. a prostitute)
QUALITY (see section 3.3) (9)
een huis van steen (material) a house of stone ‘a stone house’
(10) een kind van twee [jaar oud] (age) a child of two [year old] ‘a two-year old child’ (11) een biljet van 10 [euro] (value) a banknote of 10 euro ‘a ten-euro note’
3 4
Rosenbach (2006) is a recent study of genitives in English. The same text also contains an example without a definite article: Julius Stinde is niet alleen een man van geest, maar ook een man van het ware midden. Hij weet maat te houden. ‘Julius Stinde is not just a man of mind/spirit, but also a man of the true middle. He knows how to contain himself’.
58 Jan Rijkhoff (12) een man van contradicties (human propensity) a man of contradictions ‘a man of contradictions’ (13) een meloen van twee kilo [zwaar] (weight) a melon of two kilo [heavy] ‘a melon that weighs two kilos’ (14) een muur van twee meter hoog (height) a wall of two meter high ‘a two-meter high wall’ (15) een berg van de eerste categorie (height) a mountain of the first category ‘a mountain of the first category’ (16) een fraaie gevel van 2.44 meter breed (width) a nice front of 2.44 meter wide ‘a nice front that is 2.44 meters wide’ LOCATION (see section 3.4 as to why these possessives are regarded as localizing satellites) (17) de hoed van opa the hat of grandfather ‘grandfather’s hat’ (18) de trein van 8 uur the train of 8 o’clock ‘the 8 o’clock train’ (19) de krant van gisteren the newspaper of yesterday ‘yesterday’s paper’ (20) de fiets van mijn vader the bicycle of my father ‘my father’s bicycle’ Examples (5)–(20) illustrate that adnominal lexical possessives can occur as CLASSIFYING (τ0), QUALIFYING (τ1), OR LOCALIZING MODIFIERS (τ3) in the Dutch NP (Figure 1). As to QUANTIFYING MODIFIERS (τ2), we do find NPs such as kinderen van drie ‘children of three’ but here the numeral does not specify the number of children (a quantity) but their age (i.e. a quality:
On the co-variation between form and function of adnominal possessives 59
‘three year old children’). As a matter of fact, I am not aware of any language that uses adnominal possessives to express the cardinality of the referent of the NP (Rijkhoff 2002: chapter 5).5 What we find instead are instances of so-called dependency reversal (Malchukov 2000): the numeral word is formally expressed as the head of the construction with the noun denoting the counted entity displaying features of a possessive dependent. We will return to this gap in the functions of the adnominal possessive in section 3.2 below. Dutch does not use lexical possessives as DISCOURSE-REFERENTIAL SATELLITES (T4) either, but notice that in some languages a (bound) possessive pronoun has been claimed to mark definiteness. If this is indeed the case, the (erstwhile) possessive pronoun would serve as a grammatical DISCOURSE-REFERENTIAL GRAMMATICAL MODIFIER (cf. Siewierska et al. 1998: 811–812, notes 14 and 33). For example, Comrie (1988: 465), referring to Tauli (1966: 148), writes that some Uralic languages “use the third person singular possessive suffix as a general marker of definiteness” (see also Englebretson 2003 on Indonesian –nya and Guillemin 2007 on Mauritian Creole so). Similarly, in his monograph on the Turkic languages Menges (1968: 113) states that “the possessive significance of the suffix of the 3rd person can completely recede when it defines or determines a noun”, in which case the suffix conveys “a determinative idea as expressed by the article in Indo-European or Semitic languages”. Komi, Southern Permyak dialect (21) et-piri see woktis ru. ru-is ig. once then came fox. fox-POS.3SG hungry ‘Once the fox came that way. The fox was hungry’ (Fraurud 2001: 252; taken from Rédei 1978: 474) To what extent the possessive affix in these languages has actually grammaticalized into a bound marker of definiteness is still a matter of debate, as it seems that the change in meaning has not (yet) clearly resulted in a category switch (from suffixed possessive pronoun to suffixed definite article; Fraurud 2001). Furthermore, in Turkish the possessive suffix does not seem to occur on nouns. 5
Cardinality is indicated by an adnominal possessive in e.g. een group van 20 (mensen) ‘a group of 20 (persons)’ but this seems only possible when the phrase is headed by a collective noun like ‘group’ (i.e. not a noun denoting an individual).
60 Jan Rijkhoff Turkish (22) Bun-u iste-mi-yor-um, baka-sın-ı ver this-ACC want-NEG-PRES-1S other-POS.3SG-ACC give ‘I don’t want this (one), give me the other one’ (Gerjan van Schaaik, pers. comm.) As this contribution is mainly concerned with lexical possessives in Dutch, the issue will not be pursued here. Before we proceed, it is perhaps worth emphasizing, once again, that this chapter is restricted to possessive modifiers of common nouns denoting a concrete physical object (cf. section 1). Hence I will ignore adnominal possessives headed by (with the head noun underscored): (a) a proper name (Anna’s Peter, Rembrandt van Rijn) (b) a mass noun (the emperor’s gold) or a collective noun (the victim’s family) (c) a noun denoting a location (the place of destination, the city of Amsterdam), an event (the celebration of her birthday, the trial of the century) or some other higher order entity (the fairy tale of the princess and the pea) (d) an abstract noun (the theory of Functional Grammar, metaphors of power) (e) NPs expressing a part-whole relation or inalienable possession (the top of the mountain; see also e.g. Willemse 2006). Furthermore I have only taken into account instances in which the adnominal possessive modifier is marked by a special form (such as the -s suffix or the preposition van ‘of’) or is a member of the special set of possessive pronouns), thus ignoring possessives that are unmarked (as in the Nasioi example below) or not formally distinguished from other modifier categories. Nasioi (23) Máteasi bauran Mateasi daughter ‘Mateasi’s daughter’
(Rausch 1912: 120)
The results of this investigation indicate that adnominal possessives can occur in most of the modifier functions recognized in the layered model of the NP outlined above (examples (5)–(20)). The functional modifier categories mentioned in Figures 1 and 2 must, however, be regarded as focal
On the co-variation between form and function of adnominal possessives 61
points on a scale of scope increase (from CLASSIFYING SATELLITE/τ0 to DISCOURSE-REFERENTIAL SATELLITE /T4) rather than discrete classes of modifiers with sharp boundaries between them (section 4). Thus Dutch in Dutch cheese can refer to the country of origin (location), but also to a certain kind of cheese. Similarly, depending on the context, in the Dutch NP een huis van steen ‘a house of stone’ the possessive van steen ‘of stone’ can be interpreted as a classifying or as a qualifying satellite (see also note 8 and section 3.4 on localizing satellites).6 2.3. A note on the grammaticalization of of: scope increase or scope decrease According to the LOCALIST HYPOTHESIS, language is built on a spatial metaphor in that spatial expressions are thought to be more basic than various kinds of non-spatial expressions (Lyons 1977: 718; Lakoff and Johnson 1980: 14f., 56f.; but cf. Heine et al. 1991: 114–118). On the assumption that the locative/possessive meaning precedes all other (non-spatial) meanings in the diachronic development of a polysemous item like van ‘of’, one could hypothesize the following scenario (Heine and Kuteva 2007: 280– 283). When the Dutch preposition van ‘of’ became more and more polysemous, acquiring new non-spatial meanings along the way, it also appeared in qualifying and classifying adnominal modifiers. The scope of qualifying and classifying satellites is, however, more restricted than the scope of localizing satellites (Figures 1 and 2). (24) SEMANTIC INCREASE AND SCOPE DECREASE: THE DIACHRONIC DEVELOPMENT OF NON-SPATIAL MEANINGS OF DUTCH VAN ‘OF’ AND THE SCOPE OF LEXICAL ADNOMINAL POSSESSIVES IN DUTCH (cf. Figures 1 and 2): narrow scope HEAD CLASSIFYING QUALIFYING [Quantifying NOUN SATELLITE τ0 SATELLITE τ1 satellite τ2]
wide scope LOCALIZING SATELLITE τ3
historic development • PREPOSITION van ‘of’ (the part): SEMANTIC INCREASE (polysemy) • ADNOMINAL MODIFIER introduced by van ‘of’ (the whole): SCOPE DECREASE
6
On gradience, see e.g. Aarts (2004a/b, 2006), Rijkhoff (2008c), Sorace and Keller (2005).
62 Jan Rijkhoff Thus, paradoxically, according to this scenario INCREASE in meaning of the preposition van ‘of’ must have gone hand in hand with DECREASE in semantic scope of the new modifier category it could occur in (from outer to inner layer or from periphery to center). It may be interesting to add that grammatical modifiers in the NP (operators ω/ ) tend to develop in the opposite direction diachronically, i.e. from inner to outer layer or from center to periphery. For example, a demonstrative (Location) may turn into a definite article (Discourse-Referential), and the numeral one (Quantity) may become an indefinite article (DiscourseReferential). Similarly, at the level of the clause a perfective marker (Kind) may ultimately become a past tense marker (Location), and a future tense marker (Location) may become a mood marker, which is represented at the Interpersonal Level (Figure 1). The only counterexamples seem to involve ambiguous categories such as the PERFECT (between tense and aspect), which can be source of a past tense marker (Location), as in the Germanic languages, or a perfective marker (Kind), as in the Romance languages (Heine and Kuteva 2002: 231–232; Rijkhoff 2008d).7 3. Adnominal possessive modifiers in a layered model of the noun phrase This section investigates properties of adnominal possessives with van/‘of’, showing that each modifier function of this construction discussed in section 2 (Classifying, Qualifying, Localizing) correlates with a different set of values for the parameters MODIFICATION, PREDICATION, and REFERENCE. Special attention is given to the fact that adnominal possessives do not seem to occur as quantifying satellites (section 3.3). 3.1. Possessives as classifying satellites Classifying modifiers specify what KIND of entity is being denoted by the head noun; some common examples of classifying satellites in English are the adjectives annual in annual report, presidential in presidential election, electric in electric train, or social in social security. They differ from quali7
See Hengeveld (1989: 142) on the diachronic development of clausal operators and Song (2005) for a recent discussion of scope increase and scope decrease in grammaticalization.
On the co-variation between form and function of adnominal possessives 63
fying satellites (section 3.2.) in that they do not specify some more or less objective property of an entity (round table) or the speaker’s subjective attitude towards the entity (beautiful picture), but rather a particular subclass of the category denoted by the head noun. The semantic range of classifying satellites is rather broad and may include such categories as material, purpose and function, status and rank, origin, and mode of operation (Halliday 2004: 320). Essentially they relate to any feature that may serve to classify entities into a system of smaller sets. In English classifying satellites may take the form of an adjective (25a), but also of a possessive construction (25b–c): (25) a. a corporate lawyer b. a boy’s shirt 8 c. a house of worship Classifying satellites seem to be grammatically restricted in various ways. For example, a classifying adjective does not admit an intensifier (26), comparison (27), or predicative position (28):9 (26) a. an electric train b. *a very electric train [intensifier]10 (27) a. a medical examination b. *a more medical examination [comparison] (28) a. the presidential election b. *the election is presidential [predicative position] An adnominal possessive that serves as a classifying satellite is also called a NON-REFERENTIAL or NON-DETERMINER GENITIVE, to distinguish it from its referential counterpart (see also section 3.2). Thus the classifying possessive 8
9 10
Actually a boy’s shirt is potentially ambiguous, meaning either ‘the shirt of an unidentified boy’ (non-classifying) or ‘a particular kind of shirt’ (classifying), see on this difference e.g. Taylor (1996: 665) and Willemse (2005); see also section 3.4. Cf. Quirk et al. (1985: 1339); on non-predicative adjectives, see also e.g. Farsi (1968) and Levi (1973). One does find e.g. quick divorce lawyer, where quick modifies divorce, but in such cases we seem to be dealing with a fixed expression (for example, divorce cannot be modified by slow or short).
64 Jan Rijkhoff woman’s in a woman’s hat does not refer to any particular woman but merely serves to specify for the hearer that the speaker is talking about the kind of hat that is worn by women. Furthermore, as in the case of classifying adjectives, classifying possessives such as woman’s in (29a) and (30a) cannot be modified (29b) or separated (30b) from the head noun by other modifiers, at least not without changing its function:11 CLASSIFYING POSSESSIVE
(29) a. the pretty [woman’s hat] CLASSIFYING POSSESSIVE
(30) a. the blue [woman’s hat]
NON-CLASSIFYING POSSESSIVE b. the [pretty woman’s] hat NON-CLASSIFYING POSSESSIVE b. the [woman’s] blue hat
Classifying possessives cannot be used as predicates either, compare: CLASSIFYING POSSESSIVE
(31) a. a woman’s hat
b. *that hat is a woman’s12
CLASSIFYING POSSESSIVE
(32) a. a man of prayer
b. * this man is of prayer
The same goes for these Dutch examples, compare: (33) a. De priester is dus een man God-s, een man van gebed, een man the priest is thus a man God-GEN, a man of prayer a man van de Kerk, een herder voor de mensen. of the Church, a shepherd for the people ‘so the priest is a man of God, a man of prayer, a man of the Church, a shepherd for the people’ b. *deze man is van gebed 13 this man is of prayer
11 12 13
Gunkel and Zifonun (2008) is a detailed study of one particular subcategory of classifying satellites (‘relational adjectives’) in English, German and French. Notice that it is possible to have a possessive predicate in e.g. That hat is Mary’s (That hat belongs to Mary). The same goes for een man Gods ‘a man of God’ and a man van de Kerk ‘a man of the Church’.
On the co-variation between form and function of adnominal possessives 65
(34) a. Een mooi citaat is een diamant aan de vinger van een man a nice quotation is a diamond on the finger of a man en een kei in de hand van een dwaas van geest of mind/spirit and a rock in the hand of a fool ‘a nice quotation is a diamond on the finger of a wise man and a rock in the hand of a fool’ b. *die man is van geest that man is of mind/spirit (35) a. Kijk in de agenda van een meisje van plezier look in the calender of a girl of pleasure ‘Look in the calendar of a girl of pleasure’ b. *dit meisje is van plezier this girl is of pleasure Classifying possessives can also take the form of a case marked modifier, as in these examples from Swedish and Lithuanian: Swedish (36) en folk-et-s teater a:C people-DEF.C-GEN theatre ‘a theatre for the people’ (Koptjevskaja-Tamm 2003a: 539–540) Lithuanian (37) duon-os peilis bread-GEN knife ‘a bread knife’
(Koptjevskaja-Tamm 2002: 155)
The difference between classifying and non-classifying modification may also manifest itself in the presence or absence of a possessive suffix: Roviana (38) a. mamalaengi barikaleqe voice woman ‘a woman’s voice/a female voice’ (classifying/non-referential) b. mamalaengi-na [barikaleqe hoi] that] voice-3SG.POSS [woman ‘that woman’s voice’ (referential/non-classifying) (Corston-Oliver 2002)
66 Jan Rijkhoff The English translation (‘bread knife’) of Lithuanian duonos peilis (bread.GEN knife) shows that it may be difficult to draw the line between a noun plus classifying satellite and compounds or quasi-compounds, especially when (as in ‘bread knife’) there is no sign of a dependency relation between the two elements (on this problem see e.g. Bauer 1998, Giegerich 2005).14 3.1.1. Classifying possessives: modification, predication, reference It turns out that prepositional phrases with van/‘of’ behave differently with respect to the following three grammatical tests, depending on the kind of modifier function they perform in the NP: (a) whether or not the noun can be modified (internal MODIFICATION); (b) whether or not the phrase can occur in predicate position (PREDICATION); (c) whether or not the phrase is referential (REFERENCE). The data presented above indicate that classifying possessives have negative scores on all counts: Table 1a. Classifying adnominal possessives POSSESSIVES
MODIFICATION
PREDICATION
REFERENCE
CLASSIFYING
—
—
—
Een meisje van plezier
*een meisje van veel plezier
*het meisje is van plezier
*het meisje van zijn plezier
‘a girl of pleasure’
‘a girl of much pleasure’
‘the girl is of pleasure’
‘the girl of his pleasure’
As we saw earlier, the same restrictions hold for classifying adjectives:
14
Apart from the close connection between compounding and classification, there is also a relation between inalienability and classification (as in a mountain top vs. the top of a mountain; for discussion see e.g. Chappell and McGregor 1989, 1996).
On the co-variation between form and function of adnominal possessives 67 Table 1b. Classifying adjectives ADJECTIVES
MODIFICATION
PREDICATION
CLASSIFYING
—
—
een medisch onderzoek
*een erg medisch onderzoek
*het onderzoek is medisch
‘a medical investigation’
‘a very medical investigation’
‘the investigation is medical’
REFERENCE
(not applicable)
3.2. Possessives as qualifying satellites Qualifying modifiers relate to qualitative, more or less inherent properties of the entity designed by the noun. Dixon (1982: 16) recognized seven core semantic domains that are typically covered by adjectives (if a language has them): DIMENSION (big, long, wide, thin etc.), PHYSICAL PROPERTY (hard, heavy, smooth, hot, sweet etc.), COLOR (black, white, green etc.), HUMAN PROPENSITY (jealous, sad, happy, crazy, rude etc.), AGE (new, young, old etc.), VALUE (good, bad, excellent, atrocious etc.) and SPEED (slow, fast etc.). These properties can also be specified by possessives headed by abstract nouns (denoting shape, size, age, volume, weight etc.): Hausa alheri / arzaki / hankali (39) mutum mài person having kindness / prosperity / intelligence ‘a kind/prosperous/intelligent person’ 15 (Schachter 1985: 15) Here are some adnominal possessives with van ‘of’ exemplifying some of the major adjectival categories distinguished above. DIMENSION (40) kabeljauwen van enorme grootte cod:fish(PL) of enormous size ‘cod fish of enormous size’ (41) een oudere man van geringe lengte an older man of short length ‘an elderly man of short stature’ 15
mài (also: màasú) is glossed as ‘owner, possessor of’ in Newman (1987: 721).
68 Jan Rijkhoff (42) een tank van 50 of 100 liter (inhoud) a tank of 50 or 100 liter (volume) ‘a tank with a capacity of fifty or a hundred liters’ (43) een kerstkraam van 3 meter lengte en 1 meter diepte a Christmas:stand of 3 meter length and 1 meter depth ‘a Christmas stand that is 3 meters long and 1 meter deep’ PHYSICAL PROPERTY (44) Dymfna was … een christelijke moeder van grote schoonheid Dymfna was … a Christian mother of great beauty ‘Dymfna was … a Christian mother of great beauty’ COLOR16 (45) een Ferrari F40 van rode kleur a Ferrari F40 of red color ‘a red Ferrari F40’ HUMAN PROPENSITY (46) Hij was een man van grote charme17 he was a man of great charm ‘He was a very charming man’ AGE (47) mensen van middelbare leeftijd people of medium age ‘middle-aged persons’ 16
17
In English the color term can occur by itself, i.e. without the abstract noun kleur ‘color’ (as in a Jovian moon of incredible redness, from the book Cosmos by Carl Sagan), but in Dutch this seem to be restricted to NPs headed by nouns denoting non-spatial entities: het effect van rood ‘the effect of red’, het karakter van rood ‘the character of red’, de richting van de snelheid ‘the direction of the speed’, de effecten van snelheid ‘the effects of velocity’. There is a wide variety of qualities that can be expressed this way (here all with the adjective grote ‘great’): een man van grote spontaniteit en een ontwapenende openhartigheid ‘a man of great spontaneity and disarming sincerity’, een man van grote gaven ‘a man of great gifts’, een man van grote wijsheid, geduld en begrip ‘a man of great wisdom, patience and understanding’, een man van grote kracht ‘a man of great power’, een man van grote, bombastische theorieën ‘a man of big, bombastic theories’.
On the co-variation between form and function of adnominal possessives 69
(48) een man van 40 a man of 40 ‘a 40-year old man’ VALUE (49) een parel van grote waarde a pearl of great value ‘a very valuable pearl’ (50) een huis van 4,5 ton of meer a house of 4.5 TON or more (a ‘ton’ = 100.000 euros) ‘a house that is worth 450.000 euros or more’ SPEED (51) The Boots of Great Speed (title of a story by R. P. Barnett) Thus it appears that the core adjectival categories are also fairly well covered by adnominal possessives, but the Internet search with Google and Ask also seems to point to certain gaps, at least in Dutch and English. For example, it turned out to be extremely difficult to find examples of adnominal possessives expressing notions concerning certain physical properties (e.g. of softness, of smoothness), speed (e.g. of slowness/of slow speed), or taste (e.g. of sweetness/of sweet taste). Such properties tend rather to be expressed with the entity denoting noun in the dependent (possessive) construction as in the seductive softness of the Sheridan Ultrasoft towel, the redness of berries, the slowness of the Concorde or the sweetness of honey (see also below on Dependency Reversal) or in combination with a noun denoting a higher order entity (de perceptie van bittere smaak ‘the perception of bitter taste’, de bron van bittere smaak ‘the source/cause of bitter taste’ – see also note 16). Although one could construe more or less acceptable Dutch NPs like een appel van bittere smaak ‘an apple of bitter taste’, such adnominal possessives do not seem to occur in the actual (written) language.18 The precise reason for the scarcity of certain adnominal posses18
One can say een man van goede smaak ‘a man of good taste’, but here smaak ‘taste’ is used metaphorically, i.e. it does not mean that the man tastes good but rather that he HAS good taste. The fact that adnominal possessives are not used to express the notion of speed is probably at least partly due to the fact that SPEED is not an inherent property but rather a potential property of certain movable objects, like cars and trains.
70 Jan Rijkhoff sives remains obscure, but there is little doubt that the observed gaps are at least partly due to the fact that the current study is limited to properties of concrete objects, since we saw above that one can easily find examples outside this restricted ontological domain: tekenen van roodheid ‘signs of redness’, de weg van zachtheid ‘the road of softness’, oorzaken van zuurheid ‘causes of sourness’ (cf. also English the benefits of softness, a touch of softness, areas of softness, the concept of softness, a feeling of softness; a touch of green, shades of green; the Well of Sweetness, acts of sweetness, scoops of sweetness, the perils of sweetness; the taste of sweetness; the discovery of slowness, in praise of slowness, a philosophy of Slowness, etc.). 3.2.1. Qualifying possessives: modification, predication, reference Qualifying possessives appear to display certain differences in the grammatical behaviour with respect to the three parameters mentioned in section 3.1.1: MODIFICATION, PREDICATION, and REFERENCE. We saw in section 3.1 that classifying possessives cannot be used as predicates: (52) a. een kerel van stavast a fellow of stand.firm ‘a hefty/plucky fellow’ b. *die kerel is van stavast that fellow is of stand.firm (53) a. een man van de wereld a man of the world ‘a man of the world’ b. *die man is van de wereld 19 that man is of the world In either case we are dealing with non-referential possessives, whose head constituent cannot be modified: (54) *een kerel van enorme stavast a fellow of enormous stand.firm 19
This sentence is acceptable in Dutch if there is a qualifier such as totaal ‘totally’ or helemaal ‘completely’, but this results in an idiomatic expression with the meaning ‘to be strange, confused’ or even ‘unconscious’.
On the co-variation between form and function of adnominal possessives 71
Certain qualifying possessives do not appear as predicates either, but they can be modified (QUALIFYING TYPE A): 20 (55) a. een man van (groot exegetisch) gezag 21 a man of (great exegetical) authority ‘a man of (great exegetical) authority’ b. *die man is van (groot exegetisch) gezag that man is of (great exegetical) authority (56) a. een man van (grote) tegenstrijdigheden a man of (great) contradictions ‘a man of great contradictions’ (see also example (12) above) b. *die man is van (grote) tegenstrijdigheden that man is of (great) contradictions With some qualifying possessives of type A, however, the modifier is compulsory (QUALIFYING A+): (57) a. Prokofiev: een man van meerdere werelden Prokofiev: a man of multiple worlds ‘Prokofiev: a man of many worlds’ b. *Prokofiev: een man van werelden Prokofiev: a man of worlds c. *die man is van (meerdere) werelden that man is of (multiple) worlds We also find examples of qualifying possessives that can be modified and occur as predicates (QUALIFYING B): (58) a. een kroon van (zuiver) goud a crown of (pure) gold ‘a crown of pure gold’ b. het hoofd van het beeld was van (zuiver) goud the head of the statue was of (pure) gold ‘the head of the statue was made of (pure) gold’ 20
21
In some cases (e.g. with possessives of age or measure, i.e. height, length, weight) the predicative variant is only possible without the possessive marker in Dutch: een muur van twee meter hoog (lit. ‘a wall of two meter high’) vs. die muur is twee meter hoog (lit. ‘that wall is two meter high’); see also note 22. I will say more on the classifying flavor of some of these examples in section 4.1.
72 Jan Rijkhoff As in the case of type A qualifying possessives, some qualifying possessives of type B are only acceptable with a modifier, both with the possessive in adnominal and predicative position (QUALIFYING B+): (59) a. deze Egyptische beelden (zijn) van grote kwaliteit these Egyptian statues (are) of great quality ‘these Egyptian statues (are) of high quality’ b. *deze Egyptische beelden van kwaliteit these Egyptian statues of quality c. *deze Egyptische beelden zijn van kwaliteit these Egyptian statues are of quality Apart from possessives like van grote kwaliteit ‘of high quality’, this group seems to be confined to possessives specifying some measure (of age, size, length, width, height, depth, volume, weight etc.).22 (60) a. twee zusjes van geringe lengte two sisters of short length ‘two sisters of short stature’ b. veel nieuwkomers zijn van geringe lengte en omvang many newcomers are of short length and size ‘many newcomers are of short stature and size’ Apparently the same holds for English: (61) a. Sperm whale: a very large toothed whale (Physeter macrocephalus), having a head of enormous size b. The head of the sperm whale is of (enormous) size [constructed – JR] Note, however, that the predicative variant is not possible when the measure is given a specific value. In such cases, a non-possessive predicate is used (headed by an adjective rather than a noun), as in (62c).
22
In Dutch, there is also the variant with an adjectival form (instead of an abstract noun), as in een boot van zeventig meter langA (a boat of seventy meter long) ‘a boat of seventy meters’; see also example (62c).
On the co-variation between form and function of adnominal possessives 73
(62) a. twee koperdraden van 1 meter lengteN two copper.wires of 1 meter length ‘two 1-meter copper wires’ b. *twee koperdraden zijn van 1 meter lengteN two copper.wires are of 1 meter length c. twee koperdraden zijn 1 meter lang A [constructed – JR] two copper.wires are 1 meter long ‘two copper wires are 1 meter long’ Qualifying possessives of type B+ can also be used referentially, but notice that in such cases they do not refer to an entity but rather to a PROPERTY of an entity, typically specifying the degree to which some property (quality, size, color, etc.) applies to that entity. (63) Kinderen van deze leeftijd zijn uitgesproken nieuwsgierig children of this age are particularly curious ‘Children of this age are particularly curious’ (64) Munten van deze kwaliteit zijn zeer kwetsbaar coins of this quality are very vulnerable ‘Coins of this quality are very vulnerable’ (65) Vissen van deze grootte hebben nauwelijks vijanden fish(PL) of this size have hardly enemies ‘Fish of this size hardly have enemies’ (66) Ik vind het dus veel geld voor een lens van zulke kwaliteit I find it thus much money for a lens of such quality ‘So I think it is a lot of money for a lens of such quality’ These adnominal possessives can also occur with an adjectival modifier or as a predicate: (67) een van de laatste Attische grafmonumenten van deze hoge kwaliteit one of the last Attic burial monuments of this high quality ‘one of the last Attic burial monuments of this high quality’ (68) De meeste stenen … waren van deze kwaliteit. the most stones … were of this quality ‘Most of the (precious) stones … were of this quality’
74 Jan Rijkhoff At this point the classification looks as follows: Table 2. Properties of classifying and qualifying possessives ADNOMINAL POSSESSIVES
MODIFICATION
PREDICATION
REFERENCE
de beelden zijn van grote kwaliteit ‘the statues are of high quality’
beelden van deze kwaliteit ‘statues of this quality’ (REFERENCE TO A PROPERTY)
een kroon van (zuiver) goud ‘a crown of (pure) gold’
de kroon is van (zuiver) goud the crown is of (pure) gold ‘the crown is made of (pure) gold’
—
een man van vele g ezichten ‘a man of many faces’
—
—
—
—
—
—
QUALIFYING B+ (modifier compulsory) beelden van beelden van grote kwaliteit grote kwaliteit ‘statues of great ‘statues of high quality’ quality’
QUALIFYING B een kroon van goud ‘a crown of gold’
QUALIFYING A+ (modifier compulsory) een man van vele gezichten ‘a man of many faces’
QUALIFYING A een man van gezag een man van ‘a man of authority’ (groot) gezag ‘a man of (great) authority’ CLASSIFYING een man van de — wereld ‘a man of the world’
On the co-variation between form and function of adnominal possessives 75
Table 2 shows that both classifying and qualifying possessives of type A/A+ and B are characterized by the fact that they have no referring potential. Thus one could say that nouns heading these adnominal possessives are deprived of what may be regarded as the most characteristic feature of the prototypical noun: its potential to be used in a referring expression (using the same nouns as in the possessives in Table 2: our world, Hillary’s authority, John’s face, the king’s gold, Lone’s qualities. The same is true for these Swedish examples of non-determiner swear genitives, whose nouns appear to have lost most or all of their nominal properties (Koptjevskaja-Tamm 2003a: 519). Swedish (69) a. en satan-s kärring a:C Satan-GEN crone ‘a damned crone’ oväsen b. ett herra-n-s a:N lord-DEF.C-GEN noise ‘a hell of a noise’ In some languages, such as English, we see that the same property (e.g. ‘(to be) rich, richness’) can be expressed in the form of an adjective or in the form of a possessive construction (Quirk et al. 1987: 1278, 1286): (70) ADNOMINAL ADJECTIVE a. a rich man b. an influential woman c. a courageous woman
vs.
ADNOMINAL POSSESSIVE
a man of riches a woman of influence a woman of courage / valor
Now compare: 23 (71) a. A Man of Riches Built His Team From a Trust Fund b. A Rich Man Built His Team From a Trust Fund (constructed example – JR) There is no consensus on the exact difference between sentences like (71a) and (71b), but the a-sentence seems to create some kind of distancing effect, which is iconically reflected in the fact that in a man of riches the 23
“A Man of Riches Built His Team From a Trust Fund” is the title of an article by Richard Justice about Jack Kent Cooke (The Washington Post, April 7, 1997).
76 Jan Rijkhoff preposition of appears between the head noun (man) and the abstract noun (riches). More specifically, whereas rich seems to refer to an actual property, of riches seems to specify a property of a man for whom being rich is a more characteristic or permanent feature, which may come with a set of other features (the money was acquired in a proper manner, the man has class, etc.). One could even say that the NP containing the possessive of riches has a distinct classificatory flavor in that it refers to a certain kind of man (as opposed to rich in a rich man). This is confirmed by the way NPs that contain a qualifying adnominal possessive are used in actual speech (see section 4.1. below). Both in English and in Dutch we also find instances of adnominal possessives (sometimes referred to as METAPHORICAL CONSTRUCTIONS in Dutch grammar) in which the ‘qualifying’ lexeme seems to serve as the head of the construction:24 Dutch (72) a. een boom van een vent a tree of a man ‘a very strong man’ b. een schat van een kind a treasure of a child ‘a very sweet child’ (73) a. a hell of a guy b. a whopper of a flash drive At least in Dutch these are rather idiomatic expressions, with limited possibilities for morphosyntactic variation (*een hoge boom van een aardige vent lit. ‘a tall tree of a kind man’ or *een grote schat van een Nederlands kind lit. ‘a big treasure of a Dutch child’ are ungrammatical). Semantically, such metaphorical constructions typically have a rather strong emotional value and indicate a property in excess. A similar construction is attested in a number of Austronesian languages of the Western Oceanic group. In these languages an NP such as the big house is ostensibly expressed as ‘the house’s big(ness)’. Ross (1996) ex24
Compare on this construction Paardekooper (1956), Quirk et al. (1987: 1279, 1284–1285), Everaert (1992), Foolen (2003), Plank (2003); also e.g. Ross (1998 a,b), Koptjevskaja-Tamm (2003a,b) and Lichtenberg (2005) and in particular Keizer (2007: chapter 5). See Malchukov (2000) on DEPENDENCY REVERSAL.
On the co-variation between form and function of adnominal possessives 77
plicitly compares such possessive constructions with the ones exemplified in (72) and (73): […] I have come to the conclusion that Western Oceanic adjective structures do not reflect historically e.g. “the bigness of the house” (which would be exclamatory) but rather “a/the big one of a house” (in my dialect of English “a whopper of a house”), where the “possessor” is coded as non-specific.
3.3. Possessives as quantifying satellites Possessive modifiers are also found in NPs that contain quantifying expressions, but such NPs typically seem to involve constructions in which the quantifying expression displays at least some nominal properties (such as plural marking) and takes the quantified entity as its argument. (74) a. hundreds of people b. millions of dollars This kind of construction is common in the Slavic languages. For example, in Russian numerals higher than one in the nominative or the so-called inanimate accusative case appear with the NP referring to the quantified entity in the genitive of quantification (for details see e.g. Comrie 1981: 101–104): Russian (75) pjat’ main five:NOM car:GEN.PL ‘five cars’ It is not clear why there should be this gap in the modifier functions of a possessive, but it may be relevant to point out that cardinal numerals have been found to display certain properties that sets them apart form other noun modifiers. Firstly, whereas other grammatical modifier categories of the noun have a strong tendency to (further) grammaticalize and increase their scope (‘from inner to outer layer’), numerals tend to be rather resistant to such diachronic processes. For example, collective markers (classifying operators/ω ω 0) are a common source of plural markers (quantity/ω 2) and in many languages demonstratives (ω 3) have grammaticalized into definite articles (4). But as far as I know, only the lowest numerals are involved in grammaticalization processes. For example, in many languages the numeral ‘one’
78 Jan Rijkhoff is the diachronic source of the indefinite article (Givón 1981) and it has been established that at least in some languages the numeral ‘two’ is the diachronic source of the dual marker (Dixon 1980: 323). Secondly, whereas in many languages certain words from the spatial domain may come to be used for temporal distinctions (see e.g. Haspelmath 1996), this is not so for numerals (Rijkhoff 2008d). For example, von Garnier (1909) showed that there is clear diachronic relation between markers of (spatial) collectivity and (temporal) perfectivity in languages of the IndoEuropean family. Thus in Dutch and other Germanic languages the prefix ge- still has a collectivizing meaning in nouns like gebroeders ‘(collection of) brothers’ and gebergte ‘(collection of) mountains’ (modern German Gebrüder and Gebirge respectively). At some point in the history of Germanic this prefix came to be used with certain imperfective verbs to express the notion of completeness (i.e. perfectivity) and ultimately it became associated with the past participle form of the verb (Kirk 1923: 65), as in: Dutch (76) Heb jij dat ge-daan? have you that done ‘Did you do that?’ To give another example of the change from Space to Time, in Panare (a Cariban language) two tense-marking auxiliaries are derived etymologically from demonstrative pronouns (Gildea 1993: 53). In the case of number markers there is, however, no consensus in the direction of change.25 For example, Newman (1990: 118) has argued that verbal plural markers in Chadic developed from the nominal plural markers, whereas Frajzyngier (1977) has claimed that verbal plural markers are the source of nominal plural markers in the Chadic languages (see also Mithun (1988) on this development in some North American languages).26 In sum, whereas in the case of classifying, qualifying and localizing operators (Figure 1) there appears to be clear evidence for unidirectionality in language change (‘from space to time’; Rijkhoff 2008d), there is no clear 25
26
Gil (1993: 281) claims that there is asymmetry between nominal and verbal quantification: “…, while nominal quantifiers can show up on the verb, …, verbal quantifiers can never show up on a noun.” As to the origin of number markers, Frajzyngier (1997) argued that both nominal and verbal plural markers developed from the same source: a set of deictics, determiners and anaphors (see on this subject also Lehmann 1982).
On the co-variation between form and function of adnominal possessives 79
evidence for such a development in the case of quantifying operators, which might be at least partly due to the fact that numeral concepts (singular/plural number, cardinality) are beyond the dimensions of space and time.27 Notice furthermore that all languages, in one way or other, seem to allow their speakers to verbally specify properties concerning class, quality, location, or discourse status of an entity, but that not all languages allow their speakers to specify the exact number of entities, which also points to the special status of quantifying modifiers compared to other modifier categories. Recent studies investigating the relationship between language and arithmetic in human cognition indicate that there is a distinction between a nonverbal system of number approximation and a language-based counting system for exact number and arithmetic.28 The latter system seems to be absent in speakers of Australian and Amazonian languages (Bill McGregor pers. comm.; Everett 2005). Evidence for the nonverbal representation of number in human cognition is also offered in recent studies of infants (Jordan and Brannon 2006; Berger et al. 2006). 3.4. Possessives as localizing satellites Localizing modifiers specify the location of the referent in time or space in the world of discourse. In definite NPs these modifiers typically provide addressees (hearers or readers) with an entity through which they can identify the otherwise unidentifiable referent of the matrix NP (Prince 1981, Rijkhoff 1989, Hawkins 1991, Haspelmath 1999). In Dutch the clearest examples of localizing satellites are probably adpositional NPs with a spatial prepositions such as op ‘on’, onder ‘under, below’ or in ‘in’ (as before all examples were collected with Google or Ask): (77) De man op de fiets is Dirk Mol. the man on the bicycle is Dirk Mol. Op de derde foto zien we … On the third photo see we … ‘The man on the bicycle is Dirk Mol. In the third picture we see …’ 27 28
Notice that ordinal numerals are often expressed in terms of spatial concepts (Rijkhoff 2004: 167). Evidence for the nonverbal representation of number in human cognition is also offered in recent studies of infants (Jordan and Brannon 2006; Berger et al. 2006).
80 Jan Rijkhoff (78) Daar riep de taxichauffeur naar de voorbijgangers dat ze there called the taxi driver to the passers-by that they moesten tegenhouden. de man op de fiets the man on the bicycle must stop ‘There the taxi driver called to the passers-by that they had to stop the man on the bicycle’ One could say that an adpositional modifier such as op de fiets ‘on the bicycle’ makes the referent of the matrix NP de man ‘the man’ locatable and hence identifiable for the addressee in the world of discourse by establishing a semantic relationship between referent of the matrix NP (the man) and the referent of the adnominal possessive phrase (the bicycle). Notice that the notion LOCATION is also closely related to the notion EXISTENCE: if an entity has a location in the world of discourse it exists it that world of discourse, and vice versa (cf. Lyons 1977: 718–724; Bugenhagen 1986: 127). The reason for treating (definite) possessives as localizing satellites has to do with the fact that they, too, typically serve to ‘ground’ the (otherwise unidentifiable) referent of the matrix NP: they license the definiteness of the matrix NP by providing a referential anchor for the referent of that matrix NP. In other words, without this modifier the addressee could not locate or identify the referent of the matrix in the shared world of discourse. This can even be illustrated with these more or less isolated sentences. The first example concerns a 71-year old man who in his younger days used to participate in bicycle races): (79) hij verzorgt zijn lichaam heel goed. Geen wonder dat men he takes.care.of his body very well. No wonder that people hem vroeg of dat de fiets van zijn vader was! him asked if that the bicycle of his father was ‘He takes very good care of his body. No wonder that they asked him if that was his father’s bicycle!’ Since de fiets ‘the bicycle’ is linked up with referent of the localizing satellite (i.e. an identifiable entity: the father of the 71-year old man), the addressee has no problem identifying the referent of the bicycle in the shared world of discourse. By contrast, without the adnominal possessive van zijn vader ‘of his father’ the sentence would be pragmatically marked, as there is no way in which the addressee could infer the existence or location of de fiets ‘the bicycle’ in conversational space. The same goes for the next sen-
On the co-variation between form and function of adnominal possessives 81
tences, which also contain the phrase de fiets van zijn vader ‘the bicycle of his father’. (80) Als we er voorbij liepen, wist hij zonder nadenken when we there past walked, was.able he without thinking de fiets van zijn vader aan te wijzen. the bicycle of his father to.point.out ‘When we walked past it, he could point out his father’s bicycle immediately’ The following example is taken from an account of an event that happened during World War II, when a man called Tinus Veenstra and his companion were forced to give their bicycles to some German soldiers. (81) Beiden konden lopend terug, terwijl Tinus drie dagen later ook both could walking back, while Tinus three days later also de fiets van zijn vader diende af te staan. the bicycle of his father must give.up ‘Both were allowed to walk back, but three days later Tinus also had to give up his father’s bicycle’ Restrictive relative clauses are also typically used to provide the addressee with a referential anchor for the identification of the referent of the matrix NP. In fact there are languages in which relative clauses are almost exclusively attested in definite NPs. Consider in this context, for example, these remarks taken from Lehmann’s monograph on relative clauses (see also Moravcsik 1969: 167, 170; Bach 1974: 192, 272; Givón 1990: 645ff.). Mit einem Relativsatz kann man leicht einen bestimmten Gegenstand durch Spezifikation der Situation, an der er teilhat, identifizieren. So erklärt es sich, daß die typische Relativkonstruktion von einem Definitum begleitet ist, wiewohl das natürlich prinzipiell nicht notwendig ist [With a relative clause one can easily identify a certain object by specifying the situation in which it is involved. This explains that the typical relative construction co-occurs with a determiner, although in principle this is not necessary, of course – JR]. (Lehmann 1984: 402). Das Adjektiv dient mehr der Begriffsbildung, der Relativsatz mehr der Gegenstandsidentifikation [The adjective primarily adds to the meaning, the relative clause typically serves to identify an object – JR] (Lehmann 1984: 405).
82 Jan Rijkhoff As in the case of possessives, reference must be made to an identifiable entity in the modifier. Compare these constructed examples: (82) a. The police have arrested the man. [What man?] b. The police have arrested the man who stole my car a couple of days ago. In (82a) the man has not been mentioned before and one is inclined to ask what man the speaker is referring to. The response to the b-sentence is probably different, for example “I did not know you had a car” or “I had no idea your car was stolen”. The definiteness marking of man in (82b) does not depend on whether or not the addressee knows the referent of the NP the man, but on the hearer’s ability to identify the speaker’s car referred to in the relative clause. Due to presuppositions contained in this clause (‘the speaker has a car and it was stolen a couple of days ago’), the addressee has no problem identifying the referent of the NP the man, whose existence and location in the world of discourse can be inferred on the basis of its semantic relation to the identifiable entity in that world of discourse referred to in the adnominal modifier (‘my car’ i.e. the speaker’s car; in fact, ultimately just the referent of ‘my’, i.e. the speaker). Relationships between locative, possessive, as well as existential constructions have been observed and discussed in many studies. Lyons (1967) was among the first to point out that these constructions are related, both synchronically and diachronically. Clark (1970, 1978) investigated the nature of this relation in more detail and demonstrated on the basis of a sample of 65 languages that these constructions are systematically connected in terms of word order and patterns of verb use. She interprets possessors as locations, saying that, cognitively, possessed items can be argued to be located ‘at’ the possessor (Clark 1970: 3): psychologically it would appear quite plausible to argue that if an object is in some place, and the ‘place’ is actually an animate being, then the object is possessed by the ‘place’. In other words, it is the [+Animate] feature added to the locative phrase that transforms it into a Possessor-nominal.
Lyons and Clark were only concerned with sentential constructions, but obviously the localist account of possession (see section 2.3) also holds for adnominal possessive modifiers. An example of a locative element that has developed into a marker of possessorship can be found in the following example from Ewe (Niger-Kordofanian), where “the relational noun φé ‘place’ was used as a vehicle to denote possession and developed into a general
On the co-variation between form and function of adnominal possessives 83
marker of nominal possession” (Claudi and Heine 1986; Heine and Kuteva 2002):29 Ewe (83) fofó nye φé x father my place house ‘my father’s house’ (‘the house at my father’s place’) (Claudi and Heine 1986: 316) 3.4.1. Definiteness and indefinites in NPs with localizing possessives So far we have been dealing with adnominal localizing possessives that provide a referential anchor to ground the referent of the matrix NP, i.e. we have been concerned with definite, referential possessives as adnominal modifiers in a definite NP. This is in fact how localizing possessives are commonly used. Haspelmath (1999), who did a text count in several languages (English, Italian, Modern Greek), found that adnominal possessives serve as (what I would call) localizing satellites in approximately 95% of the definite NPs. His results are corroborated by a random Internet search, in which I checked all definite/indefinite combinations of the matrix NP headed by the noun fiets (de fiets ‘the bicycle’ and een fiets ‘a bicycle’) and definite and indefinite adnominal possessives headed by the nouns father, mother, husband, wife, child, son, daughter, man, woman, boy, or girl. So as to put an upper limit on the number of NPs, I only searched for examples with an indefinite (Dutch een ‘a/an’) or definite article (de ‘the:C’ is the common article, het ‘the:N’ its neuter counterpart) or the possessive pronoun mijn ‘my’. Consequently I ignored NPs with other determiners (such as the Dutch equivalents of this, that, your or his) or indefinite quantifiers (like enige ‘some’). It turns out that localizing (definite) adnominal possessives are almost exclusively attested in definite matrix NPs: 23,957 out of 23,988 cases (i.e. 99.87%).30
29
The locative sense of Dutch van ‘of ’ is still present in expressions specifying a source, as in Hij komt van buiten ‘He comes from outside’ or De jongens stalen een fiets van het meisje ‘The boys stole a bike from the girl’ (where van het meisje ‘from the girl’ is an oblique argument of the verb stelen (van) ‘to steal (from)’.
84 Jan Rijkhoff DEFINITE + DEFINITE
(84) a. de fiets van mijn vader b. c. d. e. f. g. h. i. j. k. l.
de fiets van de fiets van de fiets van de fiets van de fiets van de fiets van de fiets van de fiets van de fiets van de fiets van de fiets van
mijn man mijn moeder mijn vrouw mijn kind mijn zoon mijn dochter de man de vrouw het kind de jongen het meisje
number of hits* (lit. ‘the bicycle of my father’) (lit. ‘the bicycle of my husband’) (lit. ‘the bicycle of my mother’) (lit. ‘the bicycle of my wife’) (lit. ‘the bicycle of my child’) (lit. ‘the bicycle of my son’) (lit. ‘the bicycle of my daughter’) (lit. ‘the bicycle of the man’) (lit. ‘the bicycle of the woman’) (lit. ‘the bicycle of the child’) (lit. ‘the bicycle of the boy’) (lit. ‘the bicycle of the girl’) *Google, Sept. 06
1,310 821 1,380 1,050 247 1,040 1,070 4,250 3,830 649 4,030 4,280 23,957
Notice that the figures in (84) are slightly exaggerated because I did not quite catch the odd doublet and all the (rare) instances in which the vanphrase happened to follow the NP de fiets ‘the bicycle’ as a prepositional phrase at the level of the clause (i.e. not as a modifier) specifying a source, as in this example from a published police report (see also note 29): (85) Een 31-jarige fietsendief en een 23-jarige heler zijn donderdagmiddag in de Oude Ebbingestraat aangehouden ‘A 31-year old bicycle thief and a 23-year old fence were arrested in the Oude Ebbingestraat on Thursday afternoon’ De 23-jarige man kocht voor f 12,– een fiets van de man the 23-year.old man bought for f 12,– a bicycle from the man ‘The 23-year old man bought a bicycle from the man for 12 guilders’ In this sentence, the buying event involves three parties (and the price of the bicycle, of course: 12 Dutch guilders): the buyer (Agent), the item bought (Patient), and the seller (Source), who in (85) is being referred to in the prepositional phrase van de man ‘from the man’. By contrast, the Internet search produced only 31 examples (31/ 23988 = 0.13% of the total number of instances) in which the matrix NP or the localizing possessive modifier is INDEFINITE. 30
A detailed study on definite and indefinite genitives in English (such as the/a girl’s name) is Willemse (2005); see also Willemse et al. (this volume).
On the co-variation between form and function of adnominal possessives 85
INDEFINITE + DEFINITE [4 instances] (86) a. een fiets van mijn vader/man/moeder/vrouw/kind/zoon[2]/dochter a bicycle of my father/husband/mother/wife/child/son/daughter ‘a bicycle belonging to my father/husband/mother/wife/child/ son/daughter’ b. een fiets van de man[1] / de vrouw[1] / het kind / de jongen / a bicycle of the man / the woman / the child / the boy / het meisje the girl ‘a bicycle belonging to the man/the woman/the child/the boy/the girl’ DEFINITE + INDEFINITE [20 instances] (87) a. de fiets van een vader / moeder / kind/zoon/dochter the bicycle of a father/husband/mother/wife/child/son/daughter ‘the bicycle of a father/husband/mother/wife/child/son/daughter’ b. de fiets van een man[4]/vrouw[5]/kind[3]/jongen[4]/meisje[4] the bicycle of a man/woman/child/boy/girl ‘the bicycle of a man/woman/child/boy/girl’ INDEFINITE + INDEFINITE [7 instances] (88) a. een fiets van een vader/ moeder /kind[2]/zoon /dochter a bicycle of a father/husband/mother/wife/child/son/daughter ‘a bicycle belonging to a father/husband/mother/wife/child/son/ daughter’ b. een fiets van een man / vrouw[1]/kind/jongen[2]/ meisje[2] a bicycle of a man / woman/child/boy/girl ‘a bicycle belonging to a man / woman/child/boy/girl’ Interestingly, 17 of the 31 cases were found in published police reports [14; but see also (94)], film scripts [2] or jokes [1], discourse genres in which persons and things are often referred to in general terms, because (i) they must remain anonymous, (ii) their identity does not matter as they are merely non-specific props figuring in the background:
86 Jan Rijkhoff DEFINITE + INDEFINITE (de fiets van een man / vrouw/ kind/ jongen /meisje) From police reports: (89) Op maandag werd de fiets van een vrouw gestolen. on Monday became the bicycle of a woman stolen ‘Last Monday the bicycle of a woman was stolen’ (first sentence of the report) (90) Bijna twee jaar nadat de fiets van een vrouw uit almost two year after the bicycle of a woman from Veenendaal werd gestolen, heeft zij haar fiets terug. Veenendaal was stolen, has she her bicycle back ‘Nearly two years after the bicycle of a woman from Veendendaal was stolen, she has got her bicycle back’ From film scripts: (91) en een tiener-meisje dat achterop de fiets van een man zit and a teenage-girl who on.the.back the bicycle of a man sits ‘and a teenage girl who is sitting on the back of the bicycle of a man’ (92) Richard ziet hoe Wesley bijna in elkaar getrapt Richard sees how Wesley almost in each.other kicked wordt omdat hij naar de fiets van een jongen kijkt becomes because he at the bicycle of a boy looks ‘Richard sees how Wesley almost gets beaten up because he is looking at the bicycle of a boy’ INDEFINITE +INDEFINITE (een fiets van een man / vrouw/kind/ jongen/meisje). From police reports: (93) Daders pakten tevens een fiets van een jongen af perpetrators took also a bicycle of a boy away ‘(the) perpetrators also stole a bicycle of a boy’ (94) Een politieman mocht echter een fiets van een meisje a policeman was.allowed however a bicycle of a girl gebruiken use ‘a policeman was, however, allowed to use a bicycle of a girl’
On the co-variation between form and function of adnominal possessives 87
INDEFINITE +INDEFINITE (een fiets van een man /vrouw/ kind/ jongen/meisje). From a joke (95) snel pakte de man een fiets van een vrouw quickly grabbed the man a bicycle of a woman ‘Quickly the man grabbed a bicycle of a woman’ Notice also that a NP like a bicycle of a woman can also be interpreted as ‘a woman’s bicycle’, i.e. with the possessive modifier woman’s as a classifying modifier (see also note 8 and the end of section 3.4.1). Let us now turn to the remaining 14 cases where we do not find a DEFINITE + DEFINITE combination involving (what seems to be) a localizing possessive modifier of the noun. The only instance of an INDEFINITE + DEFINITE combination in a police report is used when damaged items are listed: INDEFINITE + DEFINITE (een fiets van de man /de vrouw/ het kind/de jongen /het meisje). From a police report: (96) De twee, een 13-jarige jongen en een 34-jarige man, hadden the two a 13-year.old boy and a 34-year.old man had een tuinverlichting en een fiets van de vrouw vernield en a garden_lights and a bicycle of the woman destroyed and haar verbaal bedreigd. … her verbally threatened ‘the two, a 13-year old boy and a 34-year old man, had destroyed garden lights and a bicycle of the woman and verbally threatened her’ Emphasis falls naturally on the two indefinite NPs een tuinverlichting and een fiets (van de vrouw), which indicates that this particular construction (INDEFINITE + DEFINITE) is used to highlight the listed items. This is also true for one of the other three instances of the combination INDEFINITE + DEFINITE. This example occurs in a book report from a teenage boy, after he mentioned other items the man was not supposed to have in his possession:
88 Jan Rijkhoff INDEFINITE+DEFINITE (een fiets van de man / de vrouw/ het kind/de jongen /het meisje). (97) Ze vonden ook nog een fiets van de man they found also still a bicycle of the man ‘Additionally they found a bicycle of the man’ The remaining 12 cases can be divided in two main groups. In one group, the adnominal possessive contains a modifier that specifies a property that is relevant for the further development of the story: (98) De jongen die de fiets van een man zonder rechterbeen the boy who the bicycle of a man without right.leg had gejat, had stolen ‘The boy who had stolen the bicycle of a man whose right leg was missing’ (About a boy who steals a bicycle of a one-legged man, which only has a pedal for the left foot; ironically in his attempt to steal this bicycle, the boy breaks his right leg.) (99) … de fiets van een man met een vies gele broek. … the bicycle of a man with a dirty yellow trousers Hij kent die broek … He knows those trousers ‘… the bicycle of a man with dirty yellow trousers. He has seen those trousers before.’ The second group has adnominal possessives that mention highly nonspecific entities (like the cases attested in the police reports). The NPs in question typically occur in texts from chat sites or web-logs: From a chat site: (100) een vriend van me heeft eens, toen hij 6 was ofzo, … a friend of me has once, when he 6 was or.so … getrapt een fiets van een meisje in elkaar a bicycle of a girl in each.other kicked ‘when he was around 6 years old, a friend of mine once destroyed a bicycle of a girl’
On the co-variation between form and function of adnominal possessives 89
Thus, there appears to be a rather strong connection between the value of the discourse-referential operator ±Definite and the function of certain modifiers (cf. Quirk et al. 1987: 1276). Whereas in a definite NP like the ship in the ship’s funnel or the funnel of the ship serves as a localizing satellite, in an indefinite NP such as a ship’s funnel (or the funnel of a ship) the possessive serves as a classifying satellite (or is at least ambiguous between the two modifying functions; see also note 8 on a boy’s shirt). 3.4.2. Localizing possessives: modification, predication, reference Localizing possessives are the most unrestricted NP satellites: they can be modified (101b), used as predicates (101c) and are fully referential (101d): (101) a. de fiets van mijn vader the bicycle of my father ‘my father’s bicycle’ [+ MODIFICATION] b. de fiets van mijn oude vader the bicycle of my old father ‘my old father’s bicycle’ [+ PREDICATION] c. die fiets is van mijn (oude) vader the bicycle is of my (old) father ‘that bicycle belongs to my (old) father’ [+ REFERENCE] d. de fiets van mijn/jouw/etc. (oude) vader the bicycle of my/your/ etc. (old) father ‘the bicycle of my/your/etc. (old) father’ This is summarized in Table 3, which also shows that the layered model (Figures 1 and 2) seems to iconically reflect the degree of grammatical independence of NP satellites (‘iconicity of distance’): the further away the modifier is from the head noun in the model (i.e. the wider the scope of the modifier) the less restricted it is with respect to MODIFICATION, PREDICATION and REFERENCE.
90 Jan Rijkhoff Table 3. Properties of classifying, qualifying and localizing possessives in Dutch ADNOMINAL POSSESSIVES
MODIFICATION
PREDICATION
REFERENCE
LOCALIZING de fiets van mijn vader ‘the bike of my father’
de fiets van mijn (oude) vader ‘the bike of my (old) father’
die fiets is van mijn (oude) vader the bike is of my (old) father ‘the bike belongs to my (old) father’
de fiets van Peters vader ‘the bike of Peter’s father’ (REFERENCE TO AN ENTITY)
QUALIFYING B+ beelden van grote kwaliteit ‘statues of great quality’
beelden van grote kwaliteit ‘statues of high quality’
de beelden zijn van verschillende kwaliteit ‘the statues are of varying quality’
beelden van deze kwaliteit ‘statues of this quality’ (REFERENCE TO A PROPERTY OF AN ENTITY; SECTION
4.2) QUALIFYING B een kroon van goud ‘a crown of gold’
QUALIFYING A+ een man van vele gezichten ‘a man of many faces’ QUALIFYING A een man van gezag ‘a man of authority’
een kroon van (zuiver) goud ‘a crown of (pure) gold’
de kroon is van (zuiver) goud the crown is of (pure) gold ‘the crown is made of (pure) gold’
—
— een man van vele gezichten ‘a man of many faces’
—
—
—
—
—
een man van (groot) gezag ‘a man of (great) authority’ (SEE SECTION 4.1)
CLASSIFYING — een man van de wereld ‘a man of the world’
On the co-variation between form and function of adnominal possessives 91
4. On the gradual nature of modifier categories It was mentioned earlier that it is sometimes rather difficult to draw a hard and fast line between the various types of adnominal possessives (section 3.2.1). For that reason it seems better to regard the types distinguished in Table 3 as reference points on a scale of noun modification rather than distinct modifier classes. Section 4.1 discusses the fuzzy boundary between classifying and qualifying satellites of type A; section 4.2 is concerned with qualifying possessives of type B, which closely resemble localizing possessives (see Table 3). 4.1. Between Classifying and type A Qualifying adnominal possessives We saw above that adnominal QUALIFYING possessives like of (great) stature or of many faces resemble adnominal CLASSIFYING possessives such as of pleasure, the only difference being that qualifying possessives may or must contain a modifier (selected from a rather small paradigm, e.g. small, great) which is not true for classifying possessives, which resist internal modification (Table 3). The close relationship between classifying possessives and qualifying possessives of type A also manifests itself in the fact that NPs with a qualifying possessive tend to be used in grammatical contexts where they have a strong classifying flavor. Qualifying possessives of type A can occur as a modifier in subject or object NPs, as in (102)–(104), but this does not happen very often. Subject (102) There is photographic evidence that native Australians of small stature […] lived in northern Queensland during the 19th and early 20th … (103) Man of many pursuits insists arms trade isn’t among them. Object (104) The past couple of weeks we’ve been honored to have had riders of such incredible stature racing their bikes once again in T-town.31
31
The phrase of stature is particularly frequent in biblical texts. For more examples, see: http://concordance.biblebrowser.com/s/stature.htm.
92 Jan Rijkhoff It turns out that type A qualifying possessives are almost exclusively attested in indefinite NPs that occur in predicate position as well as in captions, headings and book titles. From a grammatical perspective, titles are rather similar to non-verbal clauses, except for the fact that non-verbal clauses must contain a copula (at least in Dutch and English), which is typically absent in book titles or article headings (instead we tend to find a colon). Below are examples of the string ‘… man of many …’ in the contexts mentioned above (all found on the Internet using the Ask search engine). Fridtjof Nansen: Man of many facets. … His gentler qualities … came perhaps from his quieter, more ascetic father, a lawyer of repute and a man of unswerving integrity; Benjamin Franklin: A Man of Many Talents; David A. Poulsen: A Man Of Many Hats; Alec Guinness – A Man of Many Parts; Bryan Adams is a man of many faces; Nansen: man of many talents; A man of many strong opinions … Carl Everett is a man of conviction; Maseda is a man of many talents; Russell B. Farr: a man of many colors; Hip-Hop Artist Madlib, Man of Many Names; Thomas Merton: Man of Many Journeys; Man Of Many Voices – The Official Johnny Vallis Web Site; A Man of Many Firsts – Otellini Finds New Ways to Reach Consumers; A man of many parts; Ebbers is a man of many contradictions; Bruno Kirby: A Man of Many Scenes; Ed Case is a Man of Integrity; Gideon Omnibus, a collection of fictional stories about a Scotland Yard senior superintendent of enormous stature and presence, …; Mhanda is a man of stature. … This is a man of stature; And he slew an Egyptian, a man of great stature, five cubits high.
The string ‘… woman of many …’ did not yield as many results, but here are some examples: A woman of many parts; Woman of many faces: Isabelle Huppert; (on Christina Ricci) She is a woman of many talents; Maria Skobtsov, Woman of Many Faces; Sheila McKenna, woman of many faces, is PG’s Performer of the Year; Kate: woman of many names; Cassie aka Abigail Miller, a woman of many faces. … Cassie is a woman of many facets, not unlike any other gem; New queen a woman of many talents; A woman of stature: artist Aleta Hayes portrays Sophocles heroine in solo piece.
These examples involve NPs with a type A qualifying possessive, but notice that in each case it is the whole NP has a strong classifying flavor, not just the possessive modifier. This ties in nicely with the fact that indefinite NPs in predicate position are typically used to express class inclusion (Shelley is
On the co-variation between form and function of adnominal possessives 93
a communist – Dik 1997 Part 1: 205; cf. Hengeveld 1992: 89, Keizer 1992: ch. 6).32 (105) Aaron Sorkin is [a man of many words] CLASSIFYING PREDICATE NP Thus, whereas in these examples above the adnominal possessive by itself can be categorized as a QUALIFYING satellite of type A, the NP in which it occurs has a CLASSIFYING function. 4.2. Between qualifying adnominal (Type B) and localizing satellites We saw in section 3.4 that referential possessives typically occur as localizing satellites, but that there is also one kind of qualifying possessive modifier that can be used referentially. This involves qualifying possessives of type B+, which refer to a property (rather than an entity, as in the case of localizing possessives): (106) Wie een paar keer koffie van deze kwaliteit gedronken heeft, … Who a couple.of time coffee of this quality drunk has, … ‘Someone who has tasted coffee of this quality a couple of times, …’ (107) Heb je geen kaarten van deze kleur … Have you no cards of this color … ‘If you have no cards of this color …’ This may seem as if we cannot really distinguish between localizing possessives and qualifying possessives of type B+, but there is an important difference. Whereas localizing possessives typically occur in a definite NP, qualifying satellites of type B+ are normally attested in indefinite NPs, where they do not serve to make the referent of the matrix NP identifiable but specify the degree to which some property (QUALITY, COLOR, AGE, 32
By contrast, when a DEFINITE NP occurs in predicate position, it is typically employed for identification purposes (Suzy is the singer). This is also true for predicate NPs with a qualifying adnominal possessive of Type A, as in: Dutch (i) Huub is [de man van de computers]IDENTIFYING PREDICATE NP in ons kantoor Huub is the man of the computers in our office ‘In our office, Huub is the man of the computers’
94 Jan Rijkhoff SIZE,
etc.) applies to an entity. One could say that type B+ qualifying possessives, which refer to PROPERTIES, form a bridge between the more typical qualifying possessives, which cannot be used referentially, and localizing possessives, which refer to ENTITIES and which makes it possible for the addressee to identify the referent of the matrix NP. These data indicate that the boundaries between the various modifier categories (CLASSIFYING, QUALIFYING, LOCALIZING) are gradual rather than distinct. Qualifying possessives of type A occur between other types of qualifying possessives and classifying possessives, because the NPs in which they occur often signal class inclusion, as in Ed Case is [a Man of Integrity], i.e. Ed Case belongs to the class of men who are honest and strong about what they believe to be right; section 4.1). Qualifying possessives of type B+ are situated between other types of qualifying possessives and localizing possessives because they are the only qualifying possessives that can be used referentially (but recall that they refer to properties rather than entities). 5. A tentative explanation How can we explain the distribution of values for the three parameters in Table 3? The short answer is: the different values reflect the functional requirements of the various modifier functions in which the adnominal possessive is used. To put it differently, the possessive construction has a wide range of semantic and syntactic possibilities that are exploited in different degrees (depending on the modifier function it serves in the NP) by the language user. From a formal perspective the possessive construction is simply a phrasal structure with (potentially at least) the full array of grammatical features associated with the three major lexical word class Adjective (MODIFICATION), Verb (PREDICATION), and Noun (REFERENCE): – MODIFICATION (i.e. internal modification): the head of the possessive phrase can be modified itself (… van mijn/jouw/etc. vader ‘… of my/ your/ etc. father’); – PREDICATION: the possessive phrase can occur in predicate position (… is van mijn vader, lit. ‘… is of my father’, meaning ‘… (that) belongs to my father’); – REFERENCE: the possessive phrase can be used to pick out an individual in the world of discourse (… van die man daar ‘… of that man over there’).
On the co-variation between form and function of adnominal possessives 95
The values for the three parameters in Table 3 seem to suggest that there are no grammatical restrictions when the adnominal possessive with van ‘of’ serves as a LOCALIZING SATELLITE (section 3.4), providing the addressee with an entity (a referential anchor) that makes the referent of the matrix NP identifiable in the shared world of discourse. There is, however, one pragmatic constraint: since the referential anchor must be an identifiable entity, the localizing adnominal possessive phrase must be definite (a marginal number of apparent counter examples were discussed in the previous section). When the possessive phrase is used as a QUALIFYING SATELLITE (section 3.2), it specifies a more or less characteristic property of the referent (color, value, size, etc.), a function typically associated with adjectives (if a language has them). Therefore there is no need to fully exploit the referring potential of the possessive phrase. In fact, there is only one type of qualifying possessives (type B+) which has restricted referential potential in that it can be used to refer to the extent to which some property (size, age etc.) applies to an entity (as in munten van DEZE kwaliteit ‘coins of this quality’). Since qualities are gradable properties (a rather/very /incredibly small table), it is possible to modify the head of the qualifying adnominal possessive (i.e. modification), as in een parel van g rote waarde ‘a pearl of great value’. Qualities are also properties that can be predicated of an entity, therefore possessives used as qualifying satellites may also appear as predicates (just like adnominal possessives used as localizing satellites), as in een paspoort is van g rote waarde ‘a passport is of great value’. The fact that type A qualifying possessives cannot occur as predicates (which puts them in between classifying possessives and qualifying possessives of type B) nicely illustrates the fuzzy boundary between qualifying and classifying satellites (see note 8 and section 4). CLASSIFYING POSSESSIVES further specify what kind of entity is being denoted by the head noun (section 3.1), e.g. steam train, electric train, express train all specify a particular kind of train. It seems, however, that one can only linguistically predicate a property of an entity that is conceptually complete. Thus, the reason why classifying satellites cannot serve as a sentence predicate is precisely that the subject entity does not constitute a conceptually complete entity yet to predicate a property of, as the classifying satellite actually helps to define that entity.33 In other words, one cannot say 33
Alternatively one could say (Bill McGregor, pers. comm.): the reason why classifying possessives do not occur as predicates is that a classifying modifier highlights a (proto)typical feature of a set of entities (a TYPE), and such features cannot be predicated of any particular instance (a TOKEN).
96 Jan Rijkhoff that *a hero is urban or that *a man is of the cloth, because there first has to be ‘an urban hero’ or ‘a man of the cloth’ before one can predicate some property of these entities (as in, for example, the urban hero is a myth).34 It is basically for the same reason that classifying possessives cannot be modified themselves. Since the classifying modifier and the head noun constitute a tight unit, not just conceptually (as we just saw) but also syntactically, one can only modify the combination of classifying modifier-plusnoun. In other words, the strong conceptual and syntactic bond between classifying modifier and head makes it impossible to single out the classifying possessive for modification or predicative purposes. 6. Conclusion We have seen that adnominal possessives with van/‘of’ can occur in three of the four descriptive modifier functions distinguished in the layered model of the NP presented in Figures 1 and 2: they can be used as CLASSIFYING, QUALIFYING, and LOCALIZING NP satellites. They are, however, not attested as quantifying NP satellites, which may have to do with the special role of numerical concepts in human cognition (section 3.3). Furthermore there is a positive correlation between the kind of modifier function of the adnominal possessive in the NP and its grammatical properties. CLASSIFYING ADNOMINAL POSSESSIVES (section 3.1), which only have the noun in their scope, are severely restricted with respect to the three grammatical parameters Modification, Predication, and Reference. By contrast, LOCALIZING ADNOMINAL POSSESSIVES, which have the widest scope, are not restricted with respect to these parameters, and the various types of QUALIFYING ADNOMINAL POSSESSIVES occupy intermediate positions between the two extremes (section 3.2):
34
Notice that of the cloth does not refer to a particular cloth, i.e. the presence of the definite article in of the cloth does not mean that the classifying possessive is used referentially here.
On the co-variation between form and function of adnominal possessives 97 Table 4. Classification of adnominal possessive modifiers with van ‘of’ (A+/ B+ = internal modifier obligatory, as in ‘of enormous length’). FUNCTION
MODIFICATION
PREDICATION
REFERENCE
LOCALIZING MODIFIER
+
+
+ (reference to an entity)
QUALIFYING MODIFIER B+
+
+
+ (reference to a property)
QUALIFYING MODIFIER B
+
+
—
QUALIFYING MODIFIER A/A+
+
—
—
CLASSIFYING MODIFIER
—
—
—
The more general conclusions to be drawn from this contribution are, firstly, that the term “attributive possession” (cf. Heine 1997 and many others) is too blunt and secondly, that there is no such thing as ‘the function’ of a particular form or construction: members of the same formal category (here: the adnominal possessive phrase with van/’of’) can be used in different modifier functions, and, vice versa, the same modifier function can be fulfilled by members of different form categories (illustrated, for example, in section 3.1). Thirdly and perhaps more importantly, the results of this study indicate that functional categories can be characterized in grammatical terms, making it possible to capture grammatical differences between members of the same form class (Table 3) and grammatical similarities between members of different form classes (e.g. adjectives and prepositional phrases; section 3.1) within and across languages.35
35
See McCawley (1985: 675) and Dik (1986) for a critical discussion of the ‘one form – one function’ approach in linguistics (Newmeyer 1983).
98 Jan Rijkhoff Acknowledgements I am grateful to Bill McGregor and Lærke Munkholm Molbech for helpful comments. Earlier versions of this paper were presented on three occasions: [i] the 40th Annual Meeting of the Societas Linguistica Europaea (September 2007; University of Joensuu, Finland), [ii] the symposium ‘Nouns – Cross-linguistically’ (June 2007, Campobasso, Italy), organized by the MaxPlanck-Institute for Evolutionary Anthropology (Leipzig, Germany) and the Università del Molise (Italy), [iii] the workshop ‘Classification in Cognition’ (October 2007; University of Aarhus, Denmark), part of COST A31 (“Stability and adaptation of classification systems in a cross-cultural perspective”) financed by the EU Research Framework Program. I am grateful to the audiences for interesting discussion, in particular Kristin Davidse and Lutz Gunkel. References Aarts, Bas 2004a Aarts, Bas 2004b Aarts, Bas 2006
Modelling linguistic gradience. Studies in Language 28 (1): 1–49. Conceptions of gradience in the history of linguistics. Language Sciences 26 (4): 343–389.
Conceptions of categorization in the history of linguistics. Language Sciences 28 (4): 361–385. Aertsen, Henk, Mike Hannay and Rod Lyall (eds.) 2004 Words in their Places: A Festschrift for J. Lachlan Mackenzie. Amsterdam: Free University, Faculty of Arts. Bach, Emmon 1974 Syntactic Theory. New York: Holt, Rinehart & Winston. Barton, Irène, Michael Herslund and Finn Sørensen (eds.) 2001 Dimensions of Possession. Amsterdam: Benjamins. Bauer, Laurie 1998 When is a sequence of two nouns a compound? English Language and Linguistics 2: 65–86. Bennis, Hans and Jan W. de Vries (eds.) 1992 De Binnenbouw van het Nederlands: Een Bundel Artikelen voor Piet Paardekooper. Dordrecht: Foris. Berger, Andrea, Gabriel Tzur and Michael I. Posner 2006 Infant brains detect arithmatic errors. Proceedings of the National Academy of Sciences 103 (33): 12649–12653.
On the co-variation between form and function of adnominal possessives 99 Bugenhagen, Robert D. 1986 Possession in Mangap-Mbula: its syntax and semantics. Oceanic Linguistics 25: 124 –166. Butler, Christopher S. 2003 Structure and Function: A Guide to Three Major Structural-Functional Theories. Part 1: Approaches to the Simplex Clause; Part 2: From Clause to Discourse and Beyond. Amsterdam/Philadelphia: Benjamins. Chappell, Hillary and William B. McGregor 1989 Alienability, inalienability and nominal classification. In Hall et al. (eds.), 24–36. Chappell, Hilary and William B. McGregor (eds.) 1996 The Grammar of Inalienability: A Typological Perspective on Body Part Terms and the Part-Whole Relation. Berlin /New York: Mouton de Gruyter. Clark, Eve V. 1970 Locationals: a study of the relation between ‘existential’, ‘locative’, and ‘possessive’ constructions. Working Papers in Language Universals 3 (Stanford University), L1– L36 + xiii. Clark, Eve V. 1978 Locationals: existential, locative, and possessive constructions. In Greenberg et al. (eds.), 85 –126. Claudi, Ulrike and Bernd Heine 1986 On the metaphorical basis of grammar. Studies in Language 10 (2): 297–335. Cole, Peter (ed.). 1981 Radical Pragmatics. New York: Academic Press. Comrie, Bernard 1981 Language Universals and Linguistic Typology. Oxford: Blackwell. Comrie, Bernard 1988 General features of the Uralic languages. In Sinor (ed.), 451–477. Comrie, Bernard (ed.) 1987 The World’s Major Languages. London: Croom Helm. Connolly, John H. and Simon C. Dik (eds.) 1989 Functional Grammar and the Computer (Functional Grammar Series 10). Dordrecht: Foris. Corston-Oliver, Simon 2002 Roviana. In Lynch et al. (eds.), 467–497. Corum, Claudia, T. Cedric Smith-Stark and Ann Weiser (eds.) 1973 You Take the High Node and I’ll Take the Low Node. Papers from the 9th Regional Meeting of the Chicago Linguistic Society, April 13–15, 1973. Chicago: Chicago Linguistic Society. Dik, Simon C. 1986 On the notion ‘functional explanation’. Belgian Journal of Linguistics 1: 11–52.
100 Jan Rijkhoff Dik, Simon C. 1989 The Theory of Functional Grammar (Functional Grammar Series 9). Dordrecht: Foris. Dik, Simon C. 1997 The Theory of Functional Grammar (2nd revised edition, edited by Kees Hengeveld). Part 1: The Structure of the Clause. Part 2: Complex and Derived Constructions. Berlin/New York: Mouton de Gruyter. Dixon, Robert M.W. 1980 The Languages of Australia. Cambridge: Cambridge University Press. Dixon, Robert M.W. 1982 Where have all the Adjectives Gone? And other Essays in Semantics and Syntax. Berlin /New York: Mouton de Gruyter. Englebretson, Robert 2003 Searching for Structure: The Problem of Complementation in Colloquial Indonesian Conversation [Studies in Discourse and Grammar 13]. Amsterdam /Philadelphia: Benjamins. Everaert, Martin 1992 Nogmaals: Een schat van een kind. In Bennis and de Vries (eds.), 45–54. Everett, Daniel L. 2005 Cultural constraints on grammar and cognition in Pirahã: another look at the design features of human language. Current Anthropology 46 (4): 621–634 (plus comments and a reply from the author, 635– 646). Foolen, Ad 2004 Expressive binominal NPs in Germanic and Romance languages. In Radden and Panther (eds.), 75–100. Farsi, A. A. 1968 Classification of adjectives. Language Learning 18: 45–60. Frajzyngier, Zygmunt 1997 Grammaticalization of number: from demonstratives to nominal and verbal plural. Linguistic Typology 1 (2): 193–242. Fraurud, Kari 2001 Possessives with extensive use: a source of definite articles? In Barton et al. (eds.), 243–267. García Velasco, Daniel and Jan Rijkhoff (eds.) 2008 The Noun Phrase in Functional Discourse Grammar [Trends in Linguistics. Studies and Monographs 195]. Berlin /New York: Mouton de Gruyter. Giegerich, Heinz J. 2005 Associative adjectives in English and the lexicon-syntax interface. Journal of Linguistics 41 (3): 571–591. Gil, David 1993 Nominal and verbal quantification. Sprachtypologie und Universalienforschung (STUF) 46 (4): 275–317.
On the co-variation between form and function of adnominal possessives 101 Gildea, Spike 1993 The development of tense markers from demonstrative pronouns in Panare (Cariban). Studies in Language 17 (1): 53–73. Givón, Talmy 1981 On the development of the numeral ‘one’ as an indefinite marker. Folia Linguistica Historica 2 (1): 35–55. Givón, Talmy 1990 Syntax: A Functional-Typological Introduction. Volume II. Amsterdam: Benjamins. Greenberg, Joseph H., Charles A. Ferguson and Edith A. Moravcsik (eds.) 1978 Universals of Human Language. Volume 4: Syntax. Stanford: Stanford University Press. Guillemin, Diana 2007 A look at so in Mauritian Creole: From possessive pronoun to emphatic determiner. In Huber and Velupillai (eds.), 279–296. Amsterdam /Philadelphia: Benjamins. Gunkel, Lutz and Gisela Zifonun 2008 Constraints on relational-adjective noun constructions: a comparative view on English, German and French. Zeitschrift für Anglistik und Amerikanistik 56: 283–302. Hall, Kiria, Michael Meacham and Richard Shapiro (eds.). 1989 Proceedings of the Fifteenth Annual Meeting of the Berkeley Linguistics Society, February 18–20, 1989. General Session and Parasession on Theoretical Issues in Language Reconstruction. Berkeley: Berkeley Linguistics Society. Halliday, Michael Alexander Kirkwood 2004 An Introduction to Functional Grammar (third edition, revised by Christian M.I. M. Matthiessen). London: Arnold. Hammond, Michael and Michael Noonan (eds.). 1988 Theoretical Morphology: Approaches in Modern Linguistics. New York: Academic Press. Haspelmath, Martin 1996 From Space to Time: Temporal Adverbials in the World’s Languages (LINCOM Studies in Theoretical Linguistics 03). München: Lincom Europa. Haspelmath, Martin 1999 Explaining article-possessor complementarity: economic motivation in noun phrase syntax. Language 75 (2): 227–243. Hawkins, John A. 1991 On (in)definite articles: implicatures and (un)grammaticality prediction. Journal of Linguistics 27: 405–442. Heine, Bernd 1997 Possession: Cognitive Sources, Forces, and Grammaticalization (Cambridge Studies in Linguistics 83). Cambridge: Cambridge University Press.
102 Jan Rijkhoff Heine, Bernd, Ulrike Claudi and Friederike Hünnemeyer 1991 Grammaticalization: A Conceptual Framework. Chicago: The University of Chicago Press. Heine, Bernd and Tania Kuteva 2002 World Lexicon of Grammaticalization. Cambridge: Cambridge University Press. Heine, Bernd and Tania Kuteva 2007 The Genesis of Grammar: A Reconstruction. Oxford: Oxford University Press. Hengeveld, Kees 1988 Layers and operators. Working Papers in Functional Grammar 27. Hengeveld, Kees 1989 Layers and operators in Functional Grammar. Journal of Linguistics 25 (1): 127–157. Hengeveld, Kees 1990 The hierarchical structure of utterances. In Nuyts et al. (eds.), 1–23. Hengeveld, Kees 1992 Non-Verbal Predication: Theory, Ttypology, Diachrony. Berlin /New York: Mouton de Gruyter. Hengeveld, Kees 1994 Ik heb zoiets van: “Ze bekijken het maar!”. In F.G. van Werkgem (ed.), 8–12. Huber, Magnus and Viveka Velupillai (eds.) 2007 Synchronic and Diachronic Perspectives on Contact Languages [Creole Language Library 32]. Amsterdam /Philadelphia: Benjamins. Jordan, Kerry E. and Elizabeth M. Brannon 2006 The multisensory representation of number in infancy. Proceedings of the National Academy of Sciences 103 (9): 3586 –3489. Josephson, Folke and Ingmar Söhrman (eds.). 2008 Interdependence of Diachronic and Synchronic Analyses [Studies in Language Companion Series 103]. Amsterdam /Philadelphia: Benjamins. Keizer, M. Evelien 1992 Reference, predication and (in)definiteness in functional grammar. A functional approach to English copular sentences. Ph.D. dissertation, Vrije Universiteit Amsterdam. Keizer, M. Evelien 2007 The English Noun Phrase – The Nature of Linguistic Categorization. Cambridge: Cambridge Universtity Press. Kirk, Arthur 1923 An Introduction to the Historical Study of New High German. Manchester: The University of Manchester Press. Koptjevskaja-Tamm, Maria 2002 Adnominal possession in the European languages: form and function. Sprachtypologie und Universalienforschung (STUF) 55 (2): 141–172.
On the co-variation between form and function of adnominal possessives 103 Koptjevskaja-Tamm, Maria 2003a A woman of sin, a man of duty, and a hell of a mess: non-determiner genitives in Swedish. In Plank (ed.), 515–558. Koptjevskaja-Tamm, Maria 2003b Possessive noun phrases in the languages of Europe. In Plank (ed.), 621–722. Lehmann, Christian 1984 Der Relativsatz: Typologie seiner Strukturen, Theorie seiner Funktionen, Kompendium seiner Grammatik. Tübingen: Narr. Levi, Judith N. 1973 Where do all those other adjectives come from? Papers from the Ninth Regional Meeting of the Chicago Linguistic Society, April 13– 15. In Corum et al. (eds.), 332–345. Lichtenberg, Frantisek 2005 On the notion “adjective” in Toqabaqita. Oceanic Linguistics 44 (1): 113–144. Lynch, John, Malcolm Ross and Terry Crowley (eds.). 2002 The Oceanic Languages. Richmond: Curzon. Lyons, John 1967 A note on possessive, existential and locative sentences. Foundations of Language 3: 390–396. Lyons, John 1977 Semantics (2 volumes). Cambridge: Cambridge University Press. Malchukov, Andrej L. 2000 Dependency Reversal in Noun-Attribute Constructions: Towards a Typology (LINCOM Studies in Language Typology 03). München: Lincom Europa. McCawley, James 1985 Review of Newmeyer 1983. Language 61: 668–679. Menges, Karl Heinrich 1968 The Turkic Languages and Peoples: An Introduction to Turkic Studies. Wiesbaden: Harrassowitz. Mithun, Marianne 1988 Lexical categories and the evolution of number marking. In Hammond and Noonan (eds.), 211–234. Moravcsik, Edith A. 1969 Determination. Working Papers on Language Universals 1: 64 –130. Mosel, Ulrike and Even Hovdhaugen 1992 Samoan Reference Grammar. Oslo: Universitetsforlaget AS. Newman, Paul 1987 Hausa and the Chadic languages. In Comrie (ed.), 705–723. Newman, Paul 1990 Nominal and Verbal Plurality in Chadic (Publications in African Languages and Linguistics 12). Dordrecht: Foris.
104 Jan Rijkhoff Newman, Paul and Roxana Ma Newman (eds.). 1977 Papers in Chadic Linguistics. Leiden: Afrika-Studiecentrum. Newmeyer, Frederik J. 1983 Grammatical Theory: its Limits and Possibilities. Chicago: University of Chicago Press. Nuyts, Jan, A. Machtelt Bolkestein and Co Vet (eds.). 1990 Layers and Levels of Representation in Language Theory: A Functional View. Amsterdam: Benjamins. Paardekooper, Petrus C. 1956 Een schat van een kind. De Nieuwe Taalgids 49: 93–99. Plank, Frans 2003 Double articulation. In Plank (ed.), 337–395. Plank, Frans (ed.). 2003 Noun Phrase Structure in the Languages of Europe (Empirical Approaches to Language Typology/Eurotyp 20-7). Berlin /New York: Mouton de Gruyter. Prince, Ellen F. 1981 Toward a taxonomy of Given-New information. In Cole (ed.), 223– 255. Quirk, Randolph, Sidney Greenbaum, Geoffrey Leech and Jan Svartvik 1985 A Comprehensive Grammar of the English Language. London: Longman. Radden, Günter and Klaus-Uwe Panther (eds.). 2004 Studies in Linguistic Motivation. Berlin / New York: Mouton de Gruyter. Rausch, P. J. 1912 Die Sprache von Südost-Bougainville, Deutsche Salomonsinseln. Anthropos 7: 105 –134, 585 – 616, 964 – 994. Rédei, Károly 1978 Zyrian Folklore Texts (translated by I. Gombos). Budapest: Akadémia Kiodó. Rijkhoff, Jan 1988 A typology of operators: toward a unified analysis of terms and predications. Working Papers in Functional Grammar 29. Rijkhoff, Jan 1989 The identification of referents. In Connolly and Dik (eds.), 229–246. Rijkhoff, Jan 1990 Toward a unified analysis of terms and predications. In Nuyts et al. (eds.), 165–191. Rijkhoff, Jan 1992 The noun phrase: a typological study of its form and structure. Ph.D. dissertation, University of Amsterdam. Rijkhoff, Jan 2002 The Noun Phrase. Oxford: Oxford University Press.
On the co-variation between form and function of adnominal possessives 105 Rijkhoff, Jan 2003 When can a language have nouns and verbs? Acta Linguistica Hafniensia 35: 7–38. Rijkhoff, Jan 2004 Iconic and non-iconic word order patterns: on symmetry in the NP and counter examples to Universal 20’. In Aertsen et al. (eds.), 169– 180. Rijkhoff, Jan 2008a Layers, levels and contexts in functional discourse grammar. In García Velasco and Rijkhoff (eds.), 63–115. Rijkhoff, Jan 2008b Layering and iconicity in the noun phrase: descriptive and interpersonal modifiers. Linguistics 46 (4): 789–829. Rijkhoff, Jan 2008c On flexible and rigid nouns. Studies in Language 32 (3): 727–752 (Special issue, U. Ansaldo, J. Don and R. Pfau (eds.), Parts of Speech: Descriptive Tools, Theoretical Constructs). Rijkhoff, Jan 2008d Synchronic and diachronic evidence for parallels between noun phrases and sentences. In Josephson and Söhrman (eds.), 13–42. Rijkhoff, Jan and Johanna Seibt 2005 Mood, definiteness and specificity: a linguistic and a philosophical account of their similarities and differences. Tidsskrift for Sprogforskning 3 (2): 85–132. http://ojs.statsbiblioteket.dk/index.php/tfs/issue/archive. Rosenbach, Anette 2006 Descriptive genitives in English. English Language and Linguistics 10 (1): 77–118. Ross, Malcolm 1996 Adjectives with possessor nouns. LINGUIST List: Vol-7-1678, Subject 7.1678. Ross, Malcolm 1998a Proto-Oceanic adjectival categories and their morphosyntax. Oceanic Linguistics 37: 85–119. Ross, Malcolm 1998b Possessive-like attribute constructions in the Oceanic languages of northwest Melanesia. Oceanic Linguistics 37: 234 –276. Schachter, Paul 1985 Parts-of-speech systems. In Shopen (ed.), 3–61. Seiler, Hansjakob 1983 Possession as an Operational Dimension of Language. Tübingen: Narr. Shopen, Timothy (ed.). 1985 Language Typology and Syntactic Description. Volume I: Clause Structure. Cambridge: Cambridge University Press.
106 Jan Rijkhoff Siewierska, Anna, Jan Rijkhoff and Dik Bakker 1998 Appendix: 12 word order variables in the languages of Europe. In Siewierska (ed.), 783–812. Siewierska, Anna (ed.) 1998 Constituent Order in the Languages of Europe (Empirical Approaches to Language Typology/Eurotyp 20-1). Berlin /New York: Mouton de Gruyter. Sinclair, John (ed.) 1990 Collins COBUILD English Grammar. London: Harper Collins. Sinor, Denis (ed.). 1988 Handbuch der Orientalistik, Achte Abteilung; V-1: The Uralic Languages – Description, History and Foreign Influences. Leiden: Brill. Song, Jae Jung 2005 Grammaticalization and structural scope increase: possessive-classifier-based benefactive marking in Oceanic languages. Linguistics 43 (4): 695–838. Sorace, Antonella and Frank Keller 2005 Gradience in linguistic data. Lingua 115: 1497–1524. Tauli, Valter 1966 Structural Tendencies in the Uralic Languages (Indiana University Publications, Uralic and Altaic Series 17). Bloomington: Indiana University. Taylor, John R. 1996 Possessives in English. Oxford: Oxford University Press. van Baarda, M. J. 1908 Leiddraad bij het bestuderen van ’t Galela’sch dialekt, op het eiland Halmaheira [Manual for the study of the Galela dialect, on the island of Halmahera]. The Hague: Nijhoff. van Werkgem, Fienie van (ed.). 1994 Dubbel Nederlands – 23 Opstellen voor Simon C. Dik. Amsterdam: IFOTT / Dept. of Linguistics, University of Amsterdam. von Garnier, Katharine 1909 COM– als perfektierendes Praefix bei Plautus, SAM– im Rigveda, CYN– bei Homer. Indogermanische Forschungen 25: 86–109. Williams, Edward 1981 Argument structure and morphology. The Linguistic Review 1: 81– 114. Willemse, Peter 2006 Esphoric the N of a(n) N-nominals: forward bridging to an indefinite reference point. Folia Linguistica 40 (3–4): 319–364. Willemse, Peter, Kristin Davidse and Liesbet Heyvaert this vol. English possessives as reference-point constructions and their function in the discourse.
Is possession mere location? Contrary evidence from Maa Doris L. Payne
1. Introduction A claim sometimes found in typological and cognitive linguistic literature is that the concept of predicative “possession” is essentially no different than the concept of location,1 and hence that “Possessor” as a semantic role does not exist (Baron and Herslund 2001; DeLancey 2002; Freeze 2001; Sørensen 2001, inter alia and section 2 below). An opposing view is that Possessor is a bona fide semantic role, distinct from Locative. My goal in this paper is empirical as well as conceptual: I will examine Maa (Maasai and Il-Chamus) elicited and corpus data to evaluate the extent to which certain verb roots, and certain constructions involving those roots, are used for both location and possession. To the extent that there may be overlap in senses for a single root or root-plus-construction, it supports the view that possession is merely location with something like an animate Locative. To the extent that this does not happen, it provides support for the cognitive distinctness of possession from location. I will conclude that there is solid evidence for the latter position, at least for Maa. This paper is concerned with the type of cognitive category or categories expressed in predicative possession. By “predicative possession” I have in mind the type of construction in which possession is related to the meaning of the predicate (e.g. have) or to the meaning of an entire clausal construction (e.g. certain copular constructions). However, I exclude here External Possession which is also clausal in nature (cf. Payne 1997, Payne and Barshi 1999), and exclude the type of predicative possession that involves a 1
I capitalize semantic role labels (Locative, Theme, etc.), but I use lower case (location, possession, etc.) when representing the view that there is not necessarily a theoretically relevant semantic role involved. When in quotes, “possession” indicates the view that “Possessor” is a spurious semantic role and no different from Locative or the predication of location. I will use the abbreviations PR (possessor) and PM (possessum) only when I am not particularly concerned one way or the other with the theoretical status of such notions as semantic roles.
108 Doris L. Payne genitive construction in the predicate, i.e. constructions like This book is mine/John’s). The use of NP-internal devices to express possession is only tangentially mentioned. I basically adopt a Construction Grammar approach, in part because it is too simplistic to say that all semantic role information can be accounted for solely by properties of lexical predicates (cf. Goldberg 1995). It is also too simplistic to address verb senses without simultaneously considering the constructions that verbs occur in. I will first present two views on the relationship between possession and location (sections 2 and 3), and lay out definitions of key terms as used in this paper (section 4). Basic morphological and clause-level constructions of Maa are then introduced (section 5). Against this background, we can see that the Maa verb tii ‘be at’ appears to have locational and existential uses in particular constructions, while the verb ata ‘have’ has possessive and existential uses in particular constructions (sections 6 and 7). What is less clear from simple elicited data is whether the roots tii and ata overlap at all in predicating location and possession. Thus, a corpus study is undertaken (section 8). The elicited and text data converge to support the conclusion that speakers specialize root+constructional combinations for predicating possession versus location. The only plausible conclusion must be that human cognizers can and do recognize sufficiently-important conceptual distinctions between possession and location, both of which are relational – hence, semantic role-like – conceptualizations obtaining between two entities. On the other hand, the fact that tii is used for both locational and existential predications and that ata is used for both possession and existential predications suggests that where languages do show single forms or partially shared constructions for predicating both location and possession, we should further explore whether the conceptual link is not via an existential stage. That is, if we have a root that historically first predicates location, is this then extended to predicating existence; and once its use for predicating existence is established, is it then perhaps the existential which is extended for predicating possession? 2
2
William McGregor (pers. comm.) notes that in Nyulnyul and related Australian languages, a negative existential and a negative possessive construction share highly similar syntax that must be related historically. Significantly, there is no formally related negative locative construction to provide any link between the two – i.e. the possessive and existential constructions are directly connected.
Is possession mere location? Contrary evidence from Maa 109
2. The “possession is location” view The view that “possession” is just location with a (likely animate) Locative is well attested in the literature (though held to varying degrees of strength). I will refer to this as the “possession-is-location” view. For just a few examples, Baron and Herslund (2001) argue that the fundamental meaning of have is that one object is simply located with respect to another, though additional part-whole or other meanings may be superimposed onto this fundamental meaning. DeLancey (2002: 8) suggests that constructions which predicate possession and existence/location have the “same underlying structure, and differ only in the relative salience, inherent or contextuallydetermined, of the two arguments”. Additional claims in this direction are found in Gruber (1965), Lyons (1967, 1968), Clark (1978), Jackendoff (1983), inter alia. These sorts of claims are generally grounded in the view that the semantic role(s) of both putative “Possessor” and Locative derive from one-and-the-same conceptual or cognitive model of [x BE.AT y], where x is a Theme and y is Locative. In general, Theme and Locative derive only from either BE.AT or GO.TO types of abstract predicate notions, which may be understood in either literal or metaphorical ways. Agent derives from a CAUSE (or DO) predicate notion. All more lexically-specific predicates can be subsumed to one of these, and hence all core semantic roles in any clause whatsoever can be subsumed to Agent, Theme or Locative.3 DeLancey (2002) argues that Theme and Locative are grounded in – if not exactly equivalent to – the cognitive distinction between Figure and Ground (cf. Wertheimer 1923). In cognitive studies, Figure designates an entity which perceptually stands out against a Ground, it is the focus of attention, and it is seen in detail. A Ground is that part of the perceptual field against which the Figure stands out, it is in the periphery of attention, lacks
3
In this paper I cannot attempt to address the vast literature on theories or systems of semantic roles. DeLancey (2002) is concerned with developing a constrained theory of semantic roles that can account for surface case marking of core arguments of predicates, and specifically not with oblique cases. He does not deny the existence of semantic categories like Beneficiary or Instrument for obliques, but denies they are ever warranted as roles for core arguments. My own approach is to posit whatever is empirically driven, and believe the cross-linguistic evidence supports a somewhat larger inventory than DeLancey posits (cf. Comrie and van den Berg (2006) on case marking evidence for core Experiencier distinct from core or any type of oblique Locative).
110 Doris L. Payne detail and, according to perceptual psychologists, is not usually even perceived as an object. Particular lexemes may combine CAUSE and GO.TO/BECOME predicate notions, resulting in all three roles being present in a sentence. For example, with the predicate provide, as in This document provides you with advice, the Agent is this document, the Locative is you, and the Theme is advice. Also, a core role may sometimes be expressed in the verbal lexeme itself. In Mary baked the cookies, Mary is Agent, the cookies is Theme, and the metaphorical Locative, i.e. the condition of being baked, is lexicalized into the verb. In Susan kissed him, Susan is Agent, him is Locative, and the Theme “kiss”, which is transferred from one participant to the other, is lexicalized into the verb. A strength of the general framework surrounding the possession-islocation view is that it is highly constrained. Faced with a new sentence with unique verb and novel NPs, one cannot lightly suggest there must be new roles present just because it somehow “feels” different; properties of particular referents cannot be misconstrued as necessarily implying a different semantic role. Thus, in The storm blew the house down, the storm does not have a distinct role from Agent merely because it is inanimate. Rather, because it is linguistically presented as the CAUSE of a change, it is necessarily the Agent. With regard to possession, if y happens to be human or is otherwise well-individuated, and has potential power over the disposition of x, one might be tempted to say – naively, the possession-is-location model would hold – that the relationship between x and y is one of “possession”. But analogously to the case of the storm, the naiveté turns on mistaking features inherent in the nature of particular referents for semantic roles, when the latter must derive strictly from the nature of predicates. Returning to our focus on possession versus location, in this model meaning differences between predicates like have/own versus be (at)/reside (in) are considered irrelevant for determining semantic roles, as all these predicates subsume to the [x BE.AT y] cognitive model. This interpretation is argued for “Primarily on the basis of gross parallels in the organization of the syntactic and lexical expression of possession and location” in multiple languages (DeLancey 2002: 8, drawing on Gruber, Lyons and Jackendoff; see especially Clark 1978). Consider The garden has lots of bees and I have ten dollars, both of which employ the lexical verb have. Under the possession-is-location view, the garden and I are both Locatives, while lots of bees and ten dollars are Themes conceptualized as being at those Locatives. The fact that one of the places happens to be a well-individuated animate entity that can control and even dispose of the money, while the
Is possession mere location? Contrary evidence from Maa 111
other is an extended physical space, are irrelevant conceptual differences for an inventory of semantic roles. Similarly, French Ce livre est a Marie ‘Marie has a book’ and Pierre est a Paris ‘Pierre is in Paris’, share the structure [NP est a NP]. Such pairs transparently suggest that a PR (like Marie) may be conceptualized as a Locative (like Paris) where a Theme (PM) is located. In Jakaltek (Jacaltec; Craig 1977: 19–21), existence, location, and possession can all be predicated with the same copula ay: 4
(1)
(2)
(3)
Existential Ay anma yul conob mach scuy yuninal yin abxubal. exist people in town not teach their.children in Jacaltec ‘There are people in town who do not teach their children Jacaltec.’ Location Ay-c’oj ha mam. exist-DIR your father ‘Is your father here?’ Possession Ay no’ hin txitam. exist CL my pig ‘I have a pig.’
In some very important historical and cognitive work, Heine (1997, 2001) argues that there are eight conceptual sources of predicative possession constructions, only one of which he calls locational. An ACTION conceptual 4
Abbreviations are as follows: A – most agent-like argument of canonical transitive verb, ACC – accusative, ANTIP – antipassive, ASSOC – associative, CL – noun class, CN – discourse connective, COMP – complementizer, COP – copula, DAT – dative, DEF – definite, DIR – directional, DSCN – discontinuous discourse connective, EP – epenthetic, F – feminine, IMPF – imperfective, INF – infinitive, LOC – locative, MID – middle, NEG – negative, NMLZ – nominalizer, NOM – nominative, NPF – non-perfect/non-perfective, OBJ – object, OBL – oblique, P – most patient-like argument of canonical transitive verb, PASS – (impersonal) passive, PF – perfect/perfective aspect, PL – plural, PM – possessum, POSS – possessive, PR – possessor, PRF – perfect, PRT – particle, REL – relative, S – single argument of canonical intransitive verb, SBJ – subject, SBJV – subjunctive, SG – singular, VENT – ventive. Maa examples taken from texts are tagged with a text name and line number.
112 Doris L. Payne source would be something like She grabbed/took/kept/held the book where a verb like grab/take historically develops a specifically possessive sense (which is the historical story of English have). The other seven are LOCATION, GOAL, COMPANION, GENITIVE, SOURCE , TOPIC(-comment), and EQUATION; these are illustrated in (4). (Word order in (4) is irrelevant and artificially falls out as it does only because I am trying to express these conceptual sources in sensible English.) Possession-is-location adherents would aver that essentially all of the sources in (4) are locational, as they can be subsumed to a BE.AT predicate type, as indicated by the headings in (4). Further, in many languages a “Possessor” requires an adposition that doubles as a Locative marker in clear locative phrases (Clark 1978) – which would seem to be true in at least Heine’s LOCATION, GOAL , and SOURCE scenarios. (4)
Conceptual sources of predicative possession constructions, viewed according to the [Theme BE.AT Locative] model LOCATION GOAL GENITIVE SOURCE EQUATION
THEME The book The book Her book The book The book
BE.AT is is exists exists is
LOCATIVE at her. for/to her. (at her). from her. hers.
COMPANION TOPIC
LOCATIVE She As for her,
THEME with the book the book
BE.AT is. exists.
The only one of Heine’s schemas that at first glance might seem problematic to the possession-is-location framework is ACTION because she would be AGENT in She grabbed/took the book (i.e. these could well stand as an answer to What did Mary do? She grabbed the book.) However, the fullyspecified event chain underlying Mary grabbed the book would be [Mary CAUSE [ [book GO.TO Mary ] & [book BE.AT Mary] ] ]. Furthermore, since the preceding schemas are presented as historical sources of possession constructions, the possession-is-location adherent simply notes that as CAUSE or “doing” semantics is eventually bleached out of a verb like grab and it comes to have an inherently stative meaning, the descendent verb is no longer a CAUSE type of predicate (nor even a GO.TO predicate), but simply a BE.AT predicate.
Is possession mere location? Contrary evidence from Maa 113
3. The “possession is more-than-location” view In some contrast to the preceding view is, I believe, the view of possession articulated by Langacker (1993) and Velazquez-Castillo (1996), as well as scholars who treat Posessor and Locative as distinct semantic roles (e.g. Kemmer 2002; Tham 2005; Van Valin and LaPolla 1997: 125–127).5 Langacker and Velazquez-Castillo point out how possession involves the cognitive notion of singling out one entity, the “target” (or PM), for individual conscious awareness in terms of a more salient entity which is the “reference point” (or PR). According to Langacker (1993: 6), a reference point construction includes not only the “target” and the “reference point”, but the relevant construction is applied within a particular contextually-determined “dominion” (or set of entities) which the reference point enables access into. Reference points are determined by ease of identifiability on the part of the conceptualizer (e.g. speaker). Ease of identifiability can be context dependent, but in some way the reference-point entity has greater perceptibility – whether because it is inherently more salient, or because it is an entity for which the conceptualizer has greater empathy, familiarity, acquaintance, etc. Strikingly, such features correspond to cognitive Figures, as opposed to their Grounds; i.e., a Figure is conceptually the most salient entity. Thus, conceptually the salient PR is naturally a primary Figure, which stands out against its background. (If anything corresponds to Ground in Langacker’s scenario, it would seem to be the “dominion”.) Thus, if Theme and Locative are grounded in or equivalent to the conceptual notions of Figure and Ground, overall this should lead to the designation of PRs as being tantamount to Themes – contrary to what is assumed in the possession-is-location view. Now, of course, there can be hierarchical layers of what is Figure and what is Ground – one Figure could be a Ground for yet another Figure as the mind’s eye ranges across a conceptual scene – and this is doubtless part of what is conceptually operative in possessive constructions. But very importantly, this potential nesting of a Figure against Figure-as-relativeGround does not destroy the insight that PRs are highly perceptible and sa5
Van Valin and LaPolla’s (1997) multiple semantic roles are differentiated from their small set of macro-roles, and their Locative and Possessor are claimed to pattern together in various ways along with Experiencier, Attributant, etc. Kemmer (2002) presents typological evidence for distinct conceptualization of Possessor from Locative.
114 Doris L. Payne lient. Crucially, they may be highly perceptible and salient relative to the PM. Thus, predicating possession is different from just predicating location. If possession is “location” at all, it is localizing something relative to a highly perceptible, typically individuated, Figure-like entity. We might say that the PR is a “big” Figure, relative to a “small” Figure that is the PM. Note that these are relational (hence semantic or cognitive role-like) notions, regardless of whether they are instantiated in a verbal or clausal predicate or within an NP structure. That is, Figure and Ground mutually co-define each other, and cannot be reduced to the features of solitary referents in-andof themselves.6 I will refer to this as the “possession-is-more-than-location” view. All this leads to the question: Is there empirical lexical and/or morphosyntactic evidence that languages treat predicative possession as a significant, even basic, concept relevant to the structure of human language? That is, is there evidence supporting a distinct Possessor role (and possibly also Possessum role) related to certain predicates, different from Locative and Theme – at least in certain languages? Or can lexical and constructional behaviour be fully accounted for just in terms of the three core roles posited by the possession-is-location theory? My answer is “I don’t think that possession-is-location accords with all the linguistic data.” Indeed, this answer should not be surprising given the prima facie ramifications that understandings of possession have in ordinary social relations.7 The most obvious empirical evidence is that many languages do have distinct syntactic structures for predicating location versus possession. Indeed, even in many languages where a single predicate may be used for possessive as well as locative and/or existential meanings, there are additional required grammatical features, besides just the predicate morpheme, which distinguish the meanings. This can be seen in the 6
7
One reason a Possessor may seem different from a Locative is because a Possessor may often be topical. Proponents of the possession-is-location view would presumably say that participant topicality has no direct bearing on the existence of any particular semantic role in a clause, as topicality is a discourse or other cognitive property not based in the relational nature of the predicate. Indeed, referring to Lyons (1967: 392), DeLancey (2002: 9) suggests that physical possession is the more primitive concept in a developmental sense. Neither DeLancey nor Lyons pursues empirical acquisition research; but if DeLancey’s suggestion is true, it would seem reasonable to hypothesize that Possessor ought to be the more basic role, and “location” just a derivative or metaphorical extension of that.
Is possession mere location? Contrary evidence from Maa 115
Jakaltek examples (1)–(3) above: though they share the same copular root ay, they require distinct constructions – and something about human cognition is certainly driving the development of distinct constructions taken as a whole. In her focus on what is typologically common across constructions that predicate possession, existence, and location, Clark’s (1978) exposition does not particularly highlight how common differences are among these constructions, even if they might happen to share a particular predicate, a locative phrase at some point in history, or some other morphosyntactic element. However, differences are evident from a closer examination of the grammars of at least some languages on which she originally based her conclusions. For just one example, Amharic is one of the 30 languages in her sample. In Amharic, the same root may be used in existential (5) and predicative possessive (6) clauses, but this hardly means that the constructions are the same: the existential is intransitive, while the possessive is transitive, as shown by argument marking on the verb (examples are from Michael Ahland, pers. com.; see also Ahland 2009). (5)
Existential a. (Be-t’erapeza laj) ms’ haf-ot t all-u. LOC-table top book-PL exist.PRF-3PL.SBJ ‘There are books/some specific books (on the table).’ Existential/Locative (vague) b. Ms’ haf-ot t be-t’erapeza laj all-u. book-PL LOC-table top exist.PRF-3PL.SBJ ‘The/Some specific books are on the table.’
(6)
Possession L d -itwa ms’ haf-ot t all-u-at. child-DEF.F book-PL exist.PRF-3PL.SBJ-3F.OBJ ‘The girl has books/some specific books.’
Example (5b) is technically vague between predicating location and existence, depending on particular discourse context for clarification and on how the predication of location versus the predication of existence is defined – and it is not the case that these can be differentiated just on the basis of identifiability of participants (contra Clark’s somewhat circular methodology). However, according to both Ahland (pers. comm.) and Clark (1978), the predication of location in Amharic is primarily done via a different root, not shared with the predication of possession:
116 Doris L. Payne
(7)
Location Ms’haf-ott b-t’rapeza laj n-attw. book-PL LOC-table top COP.IMPF-3PL ‘The/some specific books are on the table.’
If one were to argue that the “Possessor” in (6) is somehow a Locative, akin to the Locative in (7), it certainly is not the case that the constructions as a whole are the same: first, (6) is transitive while (7) is intransitive; second, the order of Themes and putative Locatives in (6) and (7) are reversed; and third, the predicates are not the same.8 Clark (1978) goes on to argue that a single root (nw) is shared both for predicating location (7) – in (7) this is shortened to n – and for what she calls “Possession 2”, i.e. a construction that involves a genitive NP in a copular construction akin to The book is John’s. But when we compare the Locative in (7) with Clark’s Possession-2 structure, the constructions are still not the same because the Possession-2 structure necessarily requires a genitive NP, i.e. something like John’s or John’s thing – the possessive sense does not emerge simply from the copular construction alone.9 In sum, I do not dismiss the clear linguistic evidence that human beings can and very often do see conceptual connections between predicating location of an object, possession of an object, and existence of an object – else why would a Jakaltek or an Amharic root used at one point in time to predicate one of these notions be adapted to predicate another of these notions? I do, however, believe it is too reductionist to say that they are simply the same. My concern in this paper is to explore usage patterns of lexical and constructional forms for expressing location and possession, so as to bring further empirical evidence to bear on our understanding of how distinct these conceptions may be for human cognizers. I conclude that the Maa evidence points towards location being closely linked conceptually to existence; and of possession being somewhat closely connected to existence. But the data examined here present no evidence that possession and location are directly linked conceptually. In Maa, they do not share any lexical or constructional 8
9
Ahland (2009) presents arguments that the PR in the Amharic structure in (6) developed out of a formal locative phrase at some point in history. This does not destroy the clear fact that speakers perceived the meaning of possession to be sufficiently distinct from location that they did develop distinct constructions. Clark (1978) does not give a full Amharic clause illustrating what she calls Possession-2, as she is concerned with reporting results from a large language sample.
Is possession mere location? Contrary evidence from Maa 117
forms in common beyond general clause-level constructions that all types of predicates employ. Thus, it is problematic to say that Possessor is (just) Locative.
4. Key definitions Further definitions and elaboration of key terms are necessary in order to proceed. These definitions are generally worded in terms of speech acts on the supposition that, in making predications, speakers are primarily performing actions. Predicative possession. In this paper, I basically assume the referencepoint arguments presented by Langacker (1993). However, as there are multiple reference-point constructions (in Langacker’s treatment, e.g., possessive constructions, topic-comment constructions, metonymy), possession involves more than just identifying one participant with reference to another salient participant. I suggest that the central meaning components of predicative possession are (a) to predicate an intimate relationship between the primary reference point and another participant; and/or (b) to predicate control by the reference point participant over another participant. The first underlies the phenomenon of inalienable possession, and the second underlies alienable possession in languages which grammaticalize the difference – a difference which can show up in predicative as well as NP possession.10, 11 The meaning components in (a) and (b) just above cover a range of more specific subcases (or subsenses). The senses in (8) through (13) ap10
11
Though there is important conceptual overlap, Kuna (2003) observes that predicative possession is more precise or constrained semantically, while NP possession is polysemous. In a corpus study of Guaraní, Velazquez-Castillo (1996: 70–77) compared two predicative structures used for expressing possession. She found that the primary usage of a non-verbal predicate construction was to express intimate relationships like ‘I have arms’; she also found that no instances of the construction were used to designate spatial relations. A second verbal predicate construction with reko ‘have’ was used to express that a PR is consciously in charge of or in control of the PM. She argues that the latter structure can sometimes also place emphasis on an unusual location when used with a body part, but this is not a consistent meaning to the structure; rather, conscious control is the consistent meaning. See McGregor (2001a, b) for similar distinctions in Nyulnyul. The Maa ata Possessive construction described further below apparently covers both types of meanings.
118 Doris L. Payne pear to be characteristic (based on a limited Google search for English have).12 I am not suggesting that the subsenses surveyed below are discrete, nor that particular examples might not express more than one subsense at a time. (Also, sometimes examples appear to be ambiguous between a possession sense, and an existential function.) (8)
Predicate control over an entity; ownership a. b. c. d. e.
(9)
I have a set of Will Rogers bookends. The car I have now has a lot of things wrong with it. Do I have the right team? [directed to business managers] I have the job I want. I cursed my fate because I had no shoes … [perhaps also existential]
Predicate potentially legal rights over something (e.g. intellectual property) a. If you have intellectual property to protect, but you lack the resources of a major company, then The Patent Process is for you. b. Who has the most effective way for consumers to reduce the amount of energy used by industry? c. Many, perhaps most, of the people who read my site have weblogs of their own.
(10) Predicate a mental or other experience of a participant a. How can I have the most memorable pan-American journey possible? b. I have the feeling about 60 percent of what you say is crap. c. Should I have the maternal serum triple or quadruple test? d. I have had on occassion [sic] a few calls stating that the client has stated that the information I have provided has been incorrect. e. I have three boys all about to have birthdays. 12
Other scholars have come up with other lists, with differing numbers, of possessive subsenses (e.g. Langacker 1995; Heine 1997). My point here is just to sketch some of the typical semantics as a basis for later evaluating the hypothesis that Maa ata codes possession. The English predicate own is semantically much narrower than have, being mostly restricted to subsenses (8) and (9). I have excluded aspectual and modal auxiliary uses of have. Though the modal and aspectual senses are historically connected to the possessive sense, the former are not central to the core possession meaning.
Is possession mere location? Contrary evidence from Maa 119
(11)
Predicate a characteristic of a participant a. b. c. d. f.
Do I have the ability to connect an external antenna to my eTrex? Do I have the right kind of faith? I have the A Negative blood type. I have the fear of being afraid. Do I have the right to receive information about how my child is doing in school? g. I have the honour to introduce the combined second, third, fourth and fifth periodic report on the implementation of the Convention on the Elimination of All Forms of Discrimination against Women on behalf of the Malawi Government. h. The car I have now has a lot of things wrong with it. (12)
Predicate a part-whole relationship. (In their context of use, all the following are likely more concerned with predicating characteristics of the possessors, than on the part-whole relationship.) a. b. c. d. e.
(13)
I have a herniated disk. … and then I met a man who had no feet. The better half thinks I have awful feet. I have eyes with two different colors. Do I have a big nose?
Predicate kinship (or other close human relationship) a) I have the most fantastic daughter who ever lived. b) I have three boys all about to have birthdays.
What holds all the preceding together is the predication of a control relationship (in (8)–(9)) or particularly intimate relationship (in (10)–(13)) between two entities. In English predicative possession with have, these two entities are generally expressed as Subject (PR) and Direct Object (PM). Possessor. Given the preceding exposition, I define Possessor (for predicative possession) as the argument of a predicate or clause (a) presented as a salient reference point relative to which the speaker identifies another participant, and (b) which predicates it as having an intimate and/or controlling relationship to another participant. Agent is the argument of a predicate or clause which the speaker presents as doing or causing something; it corresponds to x in “What did x do?” (Chafe 1970; Van Valin and LaPolla 1997).
120 Doris L. Payne Theme is the argument of a predicate or clause which the speaker presents as existing (or not), being in a location or state, or changing from one location or state to another. In this paper, I will not focus on the possible distinction between PM and Theme, taking the PM to be a kind of subordinate and subsidiary Figure, relative to the Possessor as primary Figure. Location is the (often diffuse and not particularly-well individuated) argument of a predicate or clause at which the speaker predicates the Theme as being. We may note that English have can participate in predicating location – but typically this requires a locative preposition, as in How would I be able to turn on the laptop if I have it under the seat? I consider this example to conform to a distinct construction from the have-NP construction illustrated in all the example sets (8)–(13) above. Finally, an Existential predication is a predicate or clause which the speaker primarily uses to assert the existence of an argument. 5. A primer on Maa morphosyntax Maa (Eastern Nilotic, Nilo-Saharan) is the language of the Maasai, Samburu, il-Chamus, and possibly other ethnicities in Kenya and Tanzania. There is dialect variation across the Maa-speaking territory, but it is not clear that any dialect variation affects the issues explored here. In order to understand the examples and corpus data which follow, I first introduce basic facts of Maa grammar. Maa is a verb-initial, mixed head/dependent-marking language. As a head-marking language, an inflected verb usually can constitute a complete sentence. Table 1 presents basic allomorphs of singular person-number prefixes on both Transitive and Intransitive verbs.13
13
See Payne, Hamaya and Jacobs (1994) for further discussion of the InverseDirect nature of the person prefixes, and Rasmussen (2002) for tonal behaviour of these prefixes.
Is possession mere location? Contrary evidence from Maa 121 Table 1. Maa singular person prefixes for verbs Transitive INVERSE P
A
1SG
1SG 2SG 3
k áà-
Intransitive DIRECT
2SG
3
S
áá-
a -
a -
k -
Maa uses tone to mark two case distinctions on nouns and modifiers of nouns in a marked-nominative typological pattern (König 2006). There are multiple tone classes for nouns and adjectives, but each lexeme has two tonal case forms. Postverbal subjects, objects of the oblique preposition t, and vocatives occur in what Tucker and Mpaayei (1955) called the “Nominative” tone. Other post-verbal NPs, all preverbal NPs regardless of grammatical relation, objects of the oblique Associative preposition , nouns and adjectives in their citation form, and NP-internal PRs occur in their “Accusative” tone pattern. Though phrases can occur before the verb for specific pragmatic reasons, the most neutral order is for clauses to begin with a verb. Thus, a basic transitive or ditransitive clause will have the following structure (numeric indices simply refer to relative order): [V (NP1) (NP2) (NP3)]. Because NPs are marked for case in postverbal position, subject and object may vary in order. In this structure, at most one NP can be Nominative, and this could be either either NP1 or NP2/3. Invariably, NP1 is more discourse-topical than NP2/3 (Payne, Hamaya and Jacobs 1994). Text data show that order of nominals does not correspond to identifiability (definiteness) per se. Compare: (14)
-wá nk-àjí nk-ár. 3-take.PF FSG-house.ACC FSG-water.NOM ‘The house has been consumed by fire.’ (lit: “Water took the house”; an idiomatic expression) (Mol 1996: 31 and confirmed by additional speakers)
(15)
-wá nk-ár n-cán -s nyáí pk . 3-take.PF FSG-water.NOM F.PR.PRT F-rain MSG-sand.ACC all.ACC ‘The rainwater has washed off (taken) all the sand.’ (lit: ‘Water of rain took all the sand.’)
122 Doris L. Payne Tests for transitivity of a stem include its ability to take an Inverse prefix, and Antipassive and Middle suffixes (which then derive an intransitive stem). Thus, the following show that the root du is lexically transitive. (16)
Áà-dû.
3>1SG-cut ‘He/she/ they will cut me.’
Á-dú-ò
(17)
(nàn ). 1SG-cut-MID.NPF 1SG.NOM ‘I am cut’ (in my flesh).
(18)
È-dú- íshó nkráí (*nkn ). 3-cut-ANTIP FSG-child.NOM strap.ACC ‘The child is cutting something’/‘The child is able to cut without using anything.’
Basic intransitive clauses have the structure [V (NP)]. Evidence that a stem is intransitive includes its inability to take an Inverse prefix, a Middle, or an Antipassive suffix. Compare: (19)
K--d ò-sòít. DSCN-3-be.red MSG-rock.NOM ‘The rock is red.’
(20) *Áà-d. 3>1SG-be.red (21) *È-du-ísh ór-ó. 3-cut-ANTIP-MID.NPF Another construction that will figure in our discussion is the Impersonal (or “Impersonal Passive”). As Greenberg (1959) demonstrated, the Maa Impersonal developed historically out of a third person plural suffix - . This apparently satisfied the subject requirement of the verb, and hence no free Nominative NP could (nor can) co-occur in the sentence. Also, an Agent cannot be expressed in the clause in an oblique phrase. Certainly in the Maasai variety of Modern Maa, the Impersonal can occur with both transitive and intransitive stems.
Is possession mere location? Contrary evidence from Maa 123
(22) *Áà-dù-í t n nc. 3>1SG-cut-PASS OBL 3PL.NOM (‘I will be cut by them.’ / They will cut me.’) (23) Áà-dù-í (nán ). 3>1SG-cut-PASS 1SG.ACC ‘I will be cut.’ (24) -d- . 3-be.red-PASS ‘People [in general, nonspecific] are red’ / ‘Being red happens’. Whether in transitive or intransitive clauses, NPNOM and NPACC can each express a variety of semantic roles. In post-verbal position, I know of no instances where NPACC expresses Agent. But beyond that (and certainly once the Instrumental and Dative applicatives – each of which has a range of meaning – are considered), postverbal NPACC can express a wide variety of roles including Theme/ Patient, Recipient, Benefactive, Instrument, Causee, Locative, Associative, and Source (cf. also Lamoureaux 2004). The same is true for NPNOM, even though there is no promotional passive (there is a promotional Middle). Briefly consider the following; note that out of context any of the NPs below could be equally well translated with an English definite or indefinite article: (25) È-dú nk-áyíóní èn-kíné. 3-cut FSG-boy.NOM FSG-goat.ACC NOMINATIVE = AGENT ACCUSATIVE = THEME ‘The boy cut the goat.’ (26) E-tí í nk-áyíóní ò-rèyíét . 3-be.at FSG-boy.NOM MSG-river.ACC NOMINATIVE = THEME ACCUSATIVE = LOCATIVE ‘The boy is at the river.’ (27) -átá nk-áyíóní èn-kíné. 3-have FSG-boy.NOM FST-goat.ACC NOMINATIVE = POSSESSOR/LOCATIVE ACCUSATIVE = THEME ‘The boy has the goat.’ We will return below to what is the appropriate semantic role of ‘boy’ in (27).
124 Doris L. Payne With this background, we now turn to the senses of tii and ata, which cannot really be considered apart from the constructions that the roots occur in. Thus, I will review senses of each of these roots according to construction.
6.
Senses of Maa tii
We will see below that tii can occur in both the simple transitive and intransitive syntactic constructions. However, it cannot take the Antipassive (28a), Middle (28b), or Inverse (29) affixes. These properties argue that it is not a lexically transitive root. (28) a. *È-tíi-íshó. 3-be.at-ANTIP
b. *È-tíí-ò. 3-be.at-MID.NPF
(29) a. *Áà-tìì. 3>1SG-be.at (‘It is at/ near/ on me.’)
b. *Áà-tìì èn-kínè. 3>1SG-be.at FSG-goat.NOM (‘The goat is near/ at me.’)
6.1. tii in the [V NP1 NP2] construction In the [V NP1 NP2] construction, tii primarily has a locative sense, possibly an existential sense, and in some contexts is possibly in the process of grammaticalization as an imperfective aspect marker. Tii most commonly predicates location of objects at literal physical places. In (30), the river is Ground (or Locative) and the boy is Figure (or Theme). This is a prototypical example of the predication of location. Certainly neither the river nor the boy have discretionary control over each other. Additional examples follow, demonstrating a range of locative situations including metaphorical places. Note how relative order of Nominative and Accusative phrases can vary. (30) È-tí í nk-áyíóní l-kj . 3-be.at F.SG-boy.NOM MSG-river.ACC ‘The boy is at the river.’
Is possession mere location? Contrary evidence from Maa 125
(31) nk-l n-á-tìí -mányátá FSG-day.NOM F.REL-FSG.REL.NOM-be.at FSG-settlement.ACC ìl-mórûâk . MPL-elders.NOM ‘a day when the elders are in the settlement’ (murrano.0105a) (32) È-tí í òl-tóò nk-ár à-bòré. 3-be.at MSG-barrel.ACC FSG-water.NOM INF.SG-be.full ‘The barrel is full of water.’ (lit: ‘Water is at the barrel to be full.’)
Tii is strongly dispreferred, if not ungrammatical, if something like 1st or 2nd person is the Locative (33). Additionally, the best kind of Theme with tii is one that is capable of moving or being moved from one location to another. Thus, ‘mountain’ is not a preferred Theme with this verb (e.g., as in something like ‘The mountain is in the forest.’). (33) *Áà-tí í èn-kínè. 3>1-be.at FSG-goat.NOM (‘The goat is at/near me.’) An infrequent use of tii in the [V NP1 NP2] construction is to predicate existence. Though (34) does have a locative nà-kp ‘that land’, the primary purpose of the clause in its original context was probably to posit the existence of òl-tíkàná ‘East Coast Fever’.
àpá
nà-kp (34) Náà k-é-tí í and DSCN-3-be.at formerly that.F.ACC-land.ACC òl-tíkàná. MSG-East.Coast.Fever.NOM ‘And in that place there is the East Coast Fever.’
(emutata.012)
The following is similar. Here the protagonist has traveled for some distance to see some cattle, eventually locating the herd. The last line of this excerpt states that a striped ox (which had consumed multiple people) is located / exists within this herd:
126 Doris L. Payne (35) ‘…until he has gone to see bulls far away. There he is, until he has gone where they are,’ e-tí í k nâ kíshú l-k t 3-be.at these.FPL.ACC cattle.ACC MSG-cow.NOM ó-írìm-ò … MSG.REL.NOM-make.spots-MID.NPF ‘in this herd of cows there is a striped ox …’ (enamuke1.0012) Another shade of meaning for tii in the [V NP1 NP2] construction is the predication of metaphorical location. This is infrequent in the corpus data – and it is in this point only, it seems, where the use of tii comes closest to anything that could be thought of as possession. For example, in (36) one might opine that the Maasai character “has” the good feature of generosity. Note that the free translation below, using an English locative (not an English possessive) was provided by a highly fluent bilingual speaker, perhaps suggesting that he conceptualized it as closer to a locative than to English has. (36) Náà
órè -l-ón náà nàbô bá now FSG-be.generous-NMLZ FOCUS one.F.ACC matter.ACC sídáí àpá n-à-tí í àtûâ good.ACC formerly REL.F-FSG.REL.ACC-be.at inside l-kúààk l- l-máásâ . MSG-character.ACC M.PM-PL.PR .ACC MPL-Maasai.PL.ACC ‘And generosity was one of the good things that is within the Maasai character.’ (elengon2.013) FOCUS
Finally, an incipient (infrequent) use of tii in the transitive [V NP1 NP2] construction may involve imperfective aspect. In the following the discontinuous expression “education with the elders” is perhaps a metaphorical Locative, and involves a nominalization of the verb root sma ‘read’. (37) n-k smá è-tí í ó l-mórùàk. FSG-education.ACC 3-be.at ASSOC MPL-elders.ACC ‘They are learning with the elder.’ (lit: ‘They are at education with the elders.’) (bulunoto.070b) If the Accusative referent is completely identifiable and very topical in discourse, then it may be possible to omit the Accusative NP (i.e. the reference to the location or Locative is via a “definite null” zero form). For example,
Is possession mere location? Contrary evidence from Maa 127
in one written story, the writer recounted a famous event of when a British District Commissioner (DC) seized cattle from a warrior and fought over a particular bull that belonged to the warrior: (38) a. -shm l dìsí à- b
l-móí 3-go.PF that.M.NOM DC.NOM SG.INF-seize MPL-oxen.ACC l-nyná t èn-gólòn à-lòt-ú M-3SG.ACC.PR.of.PL.PM OBL FSG-strength.NOM SG.INF-go-VENT ‘that DC went to take his oxen by force’ b. à-m r t l-m nàndà t Mórìjò SG.INF-sell OBL MSG-market.place.NOM OBL Morijo.NOM Lóítà. Loita.NOM ‘to come to sell them in the market place in Morijo Loita.’ c. N-é-m-é-tí í àpá
l m rràní
nâ CN-NEG-3-be.at long.ago that.M.NOM warrior.NOM that.F.ACC kátá time.ACC ‘And that warrior was not (there) [in the market] at that time.’ (DC 6–8) In line (c) of the preceding excerpt, there is no pronominal demonstrative or lexical NP reference to ‘there’ or to ‘the market’; but this location is completely known from the context. It is unlikely that line (c) of the preceding has an existential function, as the warrior was introduced two lines prior to the beginning of this excerpt. Thus, I still treat line (c) as expressing a location, and as a transitive construction with a definite null accusative phrase. 6.2. Locative in an oblique phrase With the root tii, it is rare to express the Locative in an oblique t phrase, rather than in a simple Accusative NP; indeed, this has been rejected in elicited material. However, it is attested and three such instances occurred in the 130-item sample from the corpus (section 8). In at least the following example, the oblique t with ‘our home’ is conceivably motivated by the fact that a Comitative oblique occurred before the Locative phrase, making it more difficult for a comprehender to know how to interpret an otherwise bare NP so structurally distant from the verb tii.
128 Doris L. Payne (39) è-yíéún-òtó nk-áí pèê kì-tìí FSG-want-NMLZ F.PR.PRT FSG-God.ACC so.that SBJV.1PL-be.at táàtá t nébò
l-mn t -ná now OBL together.NOM MPL-visitors.ACC OBL-this.FSG.NOM â á home.NOM our.F.NOM ‘it is God’s will that we are together with our visitors here in our home’ (camus1.080) 6.3. tii in the [V NPNOM] construction We now turn to use of tii in what I believe is a genuinely intransitive construction [V NPNOM]. That is, the absence of an Accusative NP cannot be interpreted as a “definite null” form. The first sense of tii in this construction is to posit the existence of the referent coded as the Nominative NP.14 (40) K-é-tí í dòí
n-krâ n-áà-bìk DSCN-3-be.at indeed FPL-children.NOM REL.F-3FPL.REL.NOM-wait -m--shm. until-SBJV-3-go.SBJV ‘There are children that stay (i.e. remain without being shaved) until they are able to walk.’ (eishoi.024a)
è-òkí. (41) A-é-m-é-tí í CN-EP-NEG-3-exist FSG-sin.NOM ‘And there is no sin/offence.’
(embul.198)
(42) a. N-é-tìí àpá l tm ó ènkìtòjó, CN-3-be.at long.ago this.M.NOM elephant.NOM ASSOC FSG-hare.ACC ‘Long time ago there was this elephant and/with a hare,’ l tm na á n ny b. órè DSCN this.M.ACC elephant.ACC FOCUS 3SG.ACC -sàp k t n-kítòjó MSG.REL.ACC-big.NOM OBL FSG-hare.NOM ‘and/now this elephant was bigger than the hare.’ (elephare.001–002) 14
Contra Gutierrez (2006) and others, existential constructions do not always “require locative complementation” in any overt way – clearly they do not in Maa.
Is possession mere location? Contrary evidence from Maa 129
There is a second closely associated construction which “inherits” the preceding construction, but is invariably in third person form, carries the discourse connective n- at the beginning of the verb form, includes àpá ‘formerly, before’, and most typically also has a pause after àpá. Consider the following example:
àpá, l-m rràní ób ó, (43) a. N-é-tìí CN-3-be.at long.ago MSG-warrior.NOM one.NOM ‘Long ago, there was a warrior,’ b. órè
l m rránì, náà k--p àpá òlê DISCN this.M.ACC warrior.ACC FOCUS DSCN-3-brave long.ago very ‘and/now this warrior, he was very brave.’ (arinkoi.001– 002)
The comma in (43a) reflects a pause in the original recording. In contrast, one native speaker listening to the following recorded story with no pause after apá commented that it did not sound like the beginning of the story, but rather as if we were in the middle of a story. He commented that if the story instead began with [nét] (which is the phonological tone pattern that could occur just before a pause), it would sound more like the beginning. (44) N-é-tìí àpá l-páyìàn -yàm-á CN-3-exist long.ago MSG-elder.NOM MSG.REL.NOM-marry-PF è-sìànkí kì. FSG-bride.ACC ‘A long time ago there was an old man who married a young woman. (divorce 001) A short time later, the same consultant listened to the following, and said that this was a good way to begin a story, involving repetition of nétìí. With tone sandhi and vowel coalesence, the first phrase sounds like [nétìápà]. (45) N-é-tìí àpá, n-é-tìí ná kítòjó CN-3-exist long.ago CN-3-exist this.FSG.NOM hare.NOM ASSOC l-j n. MSG-hyena.ACC ‘A long time ago, there was this hare and a hyena.’ (enkitojo.001) In the corpus study reported further below, if the sequence nétìí àpá appeared to have strictly a formulaic story-opening function as in (45), it was
130 Doris L. Payne counted as such for the study. Otherwise, as in (43) and (44) above, it was counted as an existential. Clearly, however, these two functions and constructions shade into each other historically. 7. Senses of Maa a ta Unlike tii, ata is completely grammatical with Inverse prefixes and certainly can take 1st and 2nd person objects. But like tii it does not take Antipassive or Middle suffixes (46). The fact that it can take Inverse prefixes suggests it is lexically (more) transitive. One hypothesis as to the reason why the Antipassive and Middle suffixes are not acceptable with ata may be because the Aktionsart is stative, but this needs further investigation. (46) a. *-át á- sh 3-have-ANTIP
b. *-átà-à 3-have-MID.NPF
7.1. ata in the [V NP1 NP2] construction In contrast to tii, ata does predicate possession in the [V NP1 NP2] construction. It is attested with a range of possessive subsenses, as described in Section 4 above. In the following examples, the woman is an individuated human who can exercise control over the Accusative NPs, to the extent of selling or killing them. (47) -átà èn-kít ók èn-kíné. 3.have FSG-woman.NOM FSG-goat.ACC ‘The woman has a goat.’ (48) N-é-l ót-ú n-tásât à-nà- CN-3-go-VENT FSG-old.person.NOM INF.SG-throw.down-VENT ò-rèyíét -àtá l-k d ó MSG-river.ACC 3-have.while MSG-tail.ACC ASSOC è-sòsiân. FSG-cleaning.stick.ACC ‘The woman came down the valley having a fly whisk and a cleaning stick.’ (emutata.038b)
Is possession mere location? Contrary evidence from Maa 131
The construction can express part-whole relationships; in (49) ‘ritual hair’ is the hair on the child’s head which has been allowed to grow without cutting or shaving.
-átâ l-màsí … (49) N--tn n-kráí CN-3-sit FSG-child.NOM SBJV.3-have MSG-ritual.hair.ACC ‘the child will stay with the ritual hair …’ (eishoi.021b) The construction predicates characteristics of participants:
-s màsh òlê. (50) N--átá CN-3-have FSG-hunger.ACC very ‘And they were very hungry.’
(divorce 009b)
tìtó n-c pâî. (51) -átà ná 3-have this.NOM girl.NOM FSG-happiness.ACC ‘This girl is happy.’ k -àtà dòí -nyàmal í (52) … áà-jò-kì INF.PL-tell-DAT 1PL-have indeed FSG-problem.ACC tèné-wúéjî. in.this-place.NOM ‘… tell that we have a problem in this place.’ (ilomon.0297) The construction can also be used for predicating the experience of an event or situation, relative to a participant.
l-k sh r l-shr. (53) K--átà DSCN-3-have MPL-Kishuru.NOM MSG-senior.elder.initiation.ACC ‘The Ilkishuru age-set has a senior-elder ceremony.’ (i.e., to make them senior elders) (54) Orè
òshî -tn m--t a- DSCN normally SBJV.3-sit NEG-3-have-PL n-k páátá k--d ár àké FSG-circumcision.ceremony.ACC DSCN-3-shout just l-áyíóní ób ó. MSG-boy.NOM one.M.NOM ‘Now normally before the initial ceremony/they have the ceremony, one boy first shouts. (bulunoto.049)
132 Doris L. Payne The closest example I have encountered of where the predicate ata could be argued to express location of an object is the following elicited clause, where the barrel could not (in most worlds) be said to have control over the water; though ‘having water’ could certainly be a way of predicating a particular characteristic of a barrel. However, out of any particular discourse context it is hard to say whether the speaker’s communicative intent would be predication of a characteristic or of a location. (55) -átà òl-tóô nk-ár. 3-have MSG-barrel.NOM FSG-water.ACC ‘The barrel has (contains) water.’ Finally, marginal uses of ata may include indicating something like value or need, i.e. the possible beginning of modal sense. The data are so sparse on this point that I have nothing further to say about it here. As with tii, the Accusative argument of ata can be expressed with a definite null when it is clearly known from the context. In the following, the understood Accusative is ‘the story about the strap to heaven’, which was mentioned two clauses previously. The person posing the question comes from a different region of the Maa-speaking territory than the addressee, and is asking whether people in the il-Chamus region also ‘have’ this story: (56) M-à-yíólò tnáà -átá-tà òshî
n t á . NEG-1SG-know if 2-have-PL normally you.PL.NOM ‘I do not know if you have it yourselves.’ (camus1.030a) For an example like (57), both arguments would be known from context. I consider all such definite-null examples to also be instances of a transitive construction. (57) K-áà-àtà. DSCN-3>1SG-have ‘He owns me.’
Is possession mere location? Contrary evidence from Maa 133
7.2. ata in the [V- NPACC] Impersonal construction The Impersonal [V- NPACC] construction with ata predicates existence of an Accusative argument. (58) K--át á- òshî ìn-tòkìtín DSCN-3-have-PASS always FPL-things.ACC n-àá-j-î nk-áshê nk-án y t … REL.F-FPL.REL.ACC-say-PASS FSG-calf.ACC FSG-respect.ACC ‘There is something called the heifer of manners …’ (enkashe.001a) (59) N--àtá- aké n ny òl-òtùnó. CN-3-exist-PASS still emphatic MSG-leader.of.age.set.ACC ‘There is still the leader of the age-set.’ [called Olotuno, the-onewho-planted]. (aibartisho.004b) (60) a m k-é-yíólò áà-jò k--a t á- nk-áí … because DSCN-3-know INF.PL-say DSCN-3-have-PASS FSG-God.ACC ‘because they know there is God …’ (enkai.004) 7.3. ata in the intransitive [V NPACC] Existential construction The corpus study reveals a construction which is unusual for Maa, and to my knowledge unique to the verb root ata. This is an intransitive clause with a single Accusative argument, but without the Impersonal suffix -
(which otherwise is the only known way of creating an intransitive clause with a single Accusative NP). This construction exclusively has the sense of predicating existence. Impressionistically, it is common to use this construction when the existence of something is negated; but corpus data show that it is not at all exclusive to negative existence. (61) àm m--tà l-t ánì ò-yíólò à-jó because NEG-3-exist MSG-person.ACC MSG.REL.ACC-know COMP è-tù-búl-ú-á dúóó, 3-PF-grow-VENT-PF RELEVANT.INFORMATION ‘because there is no person who knows when he grows …’ (embul.005–006)
134 Doris L. Payne (62) àm k--ák m--tà [l-t ánì because DSCN-3-become NEG-3-exist MSG-person.ACC ò-ìtókì á kàtà à-ló nk-á M.REL.SG.ACC -repeat other.time INF.SG-go FSG-home.ACC ny.] 3SG.POSS.ACC ‘because there will not be anybody who will go to (his) home again.’ (elengon2.009) This construction is unusual because it is the one and only morphologically underived SO construction found in Maa (i.e. where the single argument of an intransitive construction is coded like the object of a transitive construction). 8. Corpus results The table incorporated into Figure 1 reports raw numbers based on a sample of 130 tokens of ata and 130 tokens of tii, extracted from a larger text corpus of approximately 20,000 clauses (containing mostly spoken but also some written material, and a variety of genres). The original methodological intent was to simply consider the absolute first 130 instances of each of these roots; but due to some false starts or other lack of clarity among the first 130 instances, some of those examples were discarded and subsequent instances were considered until a data set of 130 instances of each was compiled. I will henceforth refer to these 260 clauses as “the corpus”. In the corpus, 47% of tii instances expressed existence, 47% expressed location, and 6% expressed other senses including stative aspect; see Figure 1. There is no evidence whatsoever from this corpus that tii predicates possession. In the corpus, 65% of ata instances expressed existence, 34% expressed possession, and 1% arguably expressed modal senses; see Figure 1. There is no evidence from the corpus study that ata is used for predicating location.15
15
There is also essentially no evidence from elicitation that ata predicates location, but recall the discussion of (55) above.
Is possession mere location? Contrary evidence from Maa 135
Figure 1. Corpus distribution of ata and tii by sense (N = 130 for each root)
Figure 2 shows the constructions used for predicating location. The vast majority are the [V NP1 NP2] transitive construction with tii. The few instances of the [V NPNOM] construction have definite null locations.
tii-TRANSITIVE
tii-NOM
tii-OBL
Figure 2. Corpus distribution of predicative location constructions
Figure 3 shows that the dominant construction for predicating possession is ata in a transitive clause (ata -Transitive). Depending on discourse context, this could be a [V NP1 NP2] construction where both Theme and PR are expressed by overt NPs, with PR in the Nominative and Theme in the Accusa-
136 Doris L. Payne
ata-TRANSITIVE
ata-NOM
ata-
Figure 3. Corpus distribution of predicative possession constructions
tii-TRANSITIVE
Figure 4.
tii-NOM
ata-
ata-ACC
Corpus distribution of Existential constructions
tive case; or it could be that the Nominative PR was discourse-identifiable and sufficiently topical that it was mentioned only via a bound pronominal element, without additional use of an NP; in this case, it was counted together with the ata-Transitive category. Similarly, if the Accusative referent was referred to with a definite null because it is so clearly identified in discourse, it was counted in the ata-Transitive category. More rarely an ata clause with overt Nominative NP but no overt Accusative NP occurs with a possession meaning, as in (56) above. For the sake
Is possession mere location? Contrary evidence from Maa 137
of consistency across surface constructional patterns with the two different roots (tii vs. ata), these were counted separately from the ata -Transitive constructions. It is my clear understanding, however, that clauses like the (56) have “definite null” accusative arguments and are also transitive. Only twice did the Impersonal/Plural - suffix occur with ata to predicate possession, in contiguous clauses: (63) N--ítàyù dúóó n-kráí tànáà k--a t á- tànáà CN-3-give indeed FSG-child.NOM if DSCN-3-have-PL if k--át á- dúóó nnâ krà n-é-ítàyì-tuô … DSCN-3-have-PL indeed those children.ACC CN-3-give-PL.PF ‘The child gives, if they have any, if they have those children they have already given …’ (ilomon 0258) Here the - suffix might be interpreted as referring to a plural possessive subject, but the sense is almost existential. In sum, my conclusion is that there is really only one predicative construction for indicating possession: the ata-Transitive clause construction. Where the two roots tii and ata do overlap is for predicating existence. Figure 4 shows that the intransitive [tii NPNOM] (with no definite null locative referent), and the intransitive Impersonal [ata - NPACC] are the constructions of choice for predicating existence. Though the intransitive [ata NPACC] construction without the Impersonal suffix - is robust, it is clearly less frequent. Strikingly, for predicating existence the use of a surface transitive-like construction with overt expression of a location (even as a placeholder of sorts) is almost non-existent in the Maa corpus. This is notable, given the clear attestation in so many other languages of the use of a locative phrase in existential constructions – even if it is a semantic dummy, as in There is an idea that democracy is the preferred political system, where there does not refer to any location whatsoever. In Maa the near absence of any such phrase is possibly because an abstract notion of Locative is lexicalized into the root tii; so this language may ultimately not be particularly exceptional. 9. Conclusions Under the possession-is-location view, ata and tii have to be analyzed as having mirror image patterns of how semantic roles map to grammatical cases. With two-participant ata, Locative would be Nominative and Theme would be Accusative; while with two-participant tii, Theme is Nominative and Locative is Accusative.
138 Doris L. Payne The corpus study, however, argues that ata and tii are in complementary distribution for predicating possession versus location – unlike English have or the French locative copular construction. The corpus data are supported by the general study of senses-by-construction which also included elicited material. If “Possessor” were always just a metaphorical extension of Locative, the non-overlap of these two roots for predicating location and possession would be surprising. In particular, if all “possession” is cognitively the predication of (e.g., animate) Locative (Clark 1978: 89), then why would ata have zero attestation for spatial locational uses? And if all “possession” is cognitively the predication of location, then why would tii have zero attestation for possessive uses? The wheels of grammatical and lexical change are by all accounts slow, so finding doubling of a form across two functions (or senses) is normal and expected in the process of change. The lack of doubling – either for the roots or for constructions – suggests the lack of cognitive “sameness” or identity about possession and location. The apparent non-overlap between these two roots suggests that Possessor is indeed a salient relation that Maa speakers distinguish from Locative. Even where there are conceptual linkages that motivate sharing certain lexical roots or pieces of the morphosyntax (in constructional terms, “inheritance” of subconstructions; cf. Heine 1997), distinct constructions do emerge via the accretion and stablization of extra morphemes, and become grammaticalized. The concept of possession (and hence Possessor) is essential in explaining the specialization of both dedicated linguistic constructions and senses of forms. Finally, what merits closer cross-linguistic and historical investigation is whether human beings in general do not find closer conceptual links between possession and existence, and between location and existence, than between possession and location. Clark (1978: 113) herself noted that location and existence more often share the same verb root, than does either existence with possession, or location with possession. With reference to Maa, it could be that further usage-based research might show overlap in sense between tii and ata. However, since the [tii NPNOM] existential construction is intransitive, I would not expect it to directly develop into a possessive construction since possession necessarily involves predicating a relationship between two entities. Similarly, given that both the [ata-
NPACC] and [ata NPACC] existential constructions are effectively one-argument predicates, I would not expect them to develop into new locationals unless they are incorporated into larger constructions which somehow add a further argument.
Is possession mere location? Contrary evidence from Maa 139
Acknowledgements I am grateful to Leonard Ole-Kotikash, Keswe Ole-Mapena, Sarah Tukuoo, and multiple other Maa speakers for work with Maa texts, on which this study is based. The data in this study come primarily from Maasai and ilChamus varieties of Maa. I am also grateful to Tom Payne and Michael Ahland for discussion of issues in this paper, though they are not responsible for any of my shortcomings in understanding or interpretation. Data collection for the current study was partially supported by NSF grant Grant SBR9616482, and under Kenya research permit #OP/13/001/23C28. References Ahland, Michael 2009 From topic to subject: grammatical change in the Amharic possessive construction. Studies in Language 33 (3): 685–717. Baron, Irene and Michael Herslund 2001 Semantics of the verb HAVE. In Dimensions of Possession, Irene Baron, Michael Herslund and Finn Sørensen (eds.), 85–98. Amsterdam: John Benjamins. Clark, Eve V. 1978 Locationals: existential, locative and possessive constructions. In Universals of Human Language, Vol. 4: Syntax, Joseph H. Greenberg (ed.), 85–126. Stanford: Stanford University Press. Chafe, Wallace 1970 Meaning and the Structure of Language. Chicago: University of Chicago Press. Comrie, Bernard and Helma van den Berg 2006 Experiencer constructions in Daghestanian languages. In Semantic Role Universals and Argument Linking: Theoretical, Typological, and Psycholinguistic Perspectives, Ina Bornkessel, Matthias Schlesewsky, Bernard Comrie and Angela Friederici (eds.), 127–154. Berlin /New York: Mouton de Gruyter. Craig, Colette Grinevald 1977 Jacaltec: The Structure of Jacaltec. Austin: University of Texas Press. DeLancey, Scott 2002 The universal basis of case. Logos and Language 1 (2): 1–15. Freeze, Ray 2001 Existential constructions. In Language Typology and Language Universals: An International Handbook, Martin Haspelmath (ed.), 941– 953. Berlin /New York: Mouton de Gruyter.
140 Doris L. Payne Goldberg, Adele 1995 Argument Structure Constructions. Chicago: University of Chicago Press. Greenberg, Joseph 1959 The origin of the Masai passive. Africa 29: 171–176. Gruber, Jeffrey 1965 Studies in lexical relations. Ph.D. dissertation. MIT. [Reprinted 1976, as part of Lexical Structures in Syntax and Semantics. Amsterdam: North Holland.] Gutierrez Morales, Salome 2006 Morphosyntactic expressions of possession and existence in Sinhala. In Proceedings from the Workshop on Sinhala Linguistics (Santa Barbara Papers in Linguistics 17), Robert Englebretson and Carol Genetti (eds.), 20 –28. Heine, Bernd 1997 Possession: Cognitive Sources, Forces, and Grammaticalization. Cambridge Studies in Linguistics. Cambridge: Cambridge University Press. Heine, Bernd 2001 Ways of explaining possession. In Dimensions of Possession, Irene Baron, Michael Herslund and Finn Sørensen (eds.), 311–328. Amsterdam: John Benjamins. Jackendoff, Ray 1983 Semantics and Cognition. Cambridge, MA: MIT Press. Kemmer, Suzanne 2002 Human cognition and the elaboration of events: some universal conceptual categories. In The New Psychology of Language: Cognitive and Functional Approaches to Language Structures, Michael Tomasello (ed.), 89 –118. Mahwah, NJ: Lawrence Earlbaum. König, Christa 2006 Marked nominative in Africa. Studies in Language 30: 655 –732. Kuna, Branko 2003 Izmedu atributne i predikatne posvojnosti. (Between attributive and predicative possession.) Rasprave Instituta za Hrvatski Jezik i Jezikoslovlje 29: 157–171. Lamoureaux, Siri van dorn 2004 Applicative constructions in Maasai. MA thesis, University of Oregon. Langacker, Ronald 1993 Reference-point constructions. Cognitive Linguistics 4: 1–38. Langacker, Ronald 1995 Possession and possessive constructions. In Language and the Cognitive Construal of the World, John R. Taylor and E. MacLaury (eds.), 51–79. Berlin /New York: Mouton de Gruyter.
Is possession mere location? Contrary evidence from Maa 141 Lyons, John 1967 A note on possessive, existential, and locative sentences. Foundations of Language 3: 390 –396. Lyons, John 1968 Introduction to Theoretical Linguistics. Cambridge: Cambridge University Press. McGregor, William B. 2001a Non-verbal predicative possession in Nyulnyulan languages. In Forty Years On: Ken Hale and Australian Languages, Jane Simpson, David Nash, Mary Laughren, Peter Austin and Barry Alpher (eds.), 337–352. Canberra: Pacific Linguistics. McGregor, William B. 2001b The verb HAVE in Nyulnyulan languages. In Dimensions of Possession, ed. by Irène Baron and Michael Herslund, 67–84. Amsterdam: John Benjamins. Mol, Fr. Frans 1996 Maasai Language and Culture Dictionary. Limuru, Kenya: Kolbe Press / Maasai Centre Lemek. Payne, Doris L. 1997 The Maasai external possessor construction. In Essays on Language Function and Language Type, Joan Bybee, John Haiman and Sandra Thompson (eds.), 395–422. Amsterdam: John Benjamins. Payne, Doris L. and Immanuel Barshi (eds.) 1999 External Possession. Amsterdam: John Benjamins. Payne, Doris L., Mitsuyo Hamaya and Peter Jacobs 1994 Active, passive, and inverse in Maasai. In Voice in Discourse, T. Givón (ed.), 283–315. Amsterdam: John Benjamins. Rasmussen, Kent 2002 Tone in il-Keekonyokie Maa. MA Thesis, University of Oregon. Sørensen, Finn 2001 Possession spaces in Danish. In Dimensions of Possession, Irene Baron, Michael Herslund and Finn Sørensen (ed.), 57– 66. Amsterdam: John Benjamins. Tham, Shiao Wei 2005 Representing possessive predication: semantic dimensions and pragmatic bases. Ph.D. dissertation, Stanford University. Tucker, Archibald N. and John T. Mpaayei 1955 Maasai Grammar with Vocabulary. London: Longmans, Green & Co. Van Valin, R. and R. LaPolla 1997 Syntax: Structure, Meaning and Function. Cambridge: Cambridge University Press. Velazquez-Castillo, Maura 1996 The Grammar of Possession: Inalienability, Incorporation and Possessor Ascension in Guarani. Amsterdam: John Benjamins.
142 Doris L. Payne Wertheimer, Max 1923 Untersuchungen zur Lehre von der Gestalt II. Psycologische Forschung 4: 301–350. [Translation published in W. Ellis. 1938. A Source Book of Gestalt Psychology, 71–88. London: Routledge and Kegan Paul.]
Learning to encode possession Sonja Eisenbeiß, Ayumi Matsuo and Ingrid Sonnenstuhl
1. Introduction Learning and talking about their own possessions and the possessions of their peers and caretakers plays a central role in children’s daily life. It is unsurprising then that relationships between possessors and their possessions are amongst the first relationships that children encode when they start to string words together (see e.g. Brown 1973); and it is no wonder that many psycholinguists have made use of this rich data source to address questions about the mechanisms that drive children’s linguistic development. However, most of the available studies of the acquisition of possessive constructions that we will discuss have investigated only one or two possession-encoding constructions in an individual language. Moreover, the focus has typically not been on the encoding of the possessive relation itself, but on other aspects of the respective possessive construction. For instance, possessive -s markers in German and English (e.g. Susi-s Huhn ‘Sue’s chicken’) were analysed in studies that investigated whether the syntactic categories of the target language were already present in early child grammars (e.g. Eisenbeiß 2000; Marinis 2002, 2003; Radford 1990). In these studies, possessive markers were simply treated as morpho-syntactic realisations of syntactic categories; and semantic aspects were largely ignored. Similarly, possessive constructions with two-place verbs like have and belong were investigated in studies of the acquisition of syntax-semantic mappings, but these constructions were just treated as one type of twoargument construction and not compared to other constructions encoding possession (see e.g. Bowerman 1985; Pinker 1984). To our knowledge, no study has yet provided a comprehensive cross-linguistic overview that focuses on the different ways in which possessive relationships are encoded linguistically. In order to fill this gap, we will provide a cross-linguistic overview of studies of children’s acquisition of the constructions that their target language employs to encode possession. In addition, we will present new data from German child language and child-directed speech, and discuss the im-
144 Sonja Eisenbeiß, Ayumi Matsuo and Ingrid Sonnenstuhl plications for theoretical linguistics and language acquisition research. Our focus will be on three ways of encoding relationships between PRs and possessed entities (see Heine 1997; Baron, Herslund and Sørensen 2001 for overviews): – adnominal possession: Both Possessor (PR) and Possessum (PM) are
encoded within the same noun phrase (e.g. my/daddy’s chickens, the chickens of our neighbours, … ); – predicative possession: The possessive relationship is encoded by a twoplace predicate such as have, own or belong or by be (e.g. I have a dog. The dog belongs to me. This dog is mine); – “external possession”: the PR and the PM are realised as arguments of a verb whose lexical meaning does not involve the notion of possession (e.g. I tapped him-PR on the shoulder-PM). We will first show how studies of children’s possession constructions can help us to evaluate models of children’s linguistic development. Against this background, we will present studies of the acquisition of adnominal, predicative and external possession constructions (EPCs). For each of these construction types, we will provide a brief overview of possession constructions in adult German and contrast them with possessive constructions in other languages for which acquisition studies are available. This will allow us to discuss empirical findings from earlier studies and our own analysis of German child data. Finally, we will compare the development of the three types of possession constructions and discuss the implications of our findings for theoretical linguistics and models of children’s linguistic development. In particular, we will show how the available empirical findings about the acquisition of possession constructions can be captured in approaches that try to integrate core insights from current generative and usage-based approaches. 2. Theoretical issues in acquisition research While concepts of ownership and possession seem to be part of all cultures, there is considerable variation with respect to (i) the legal norms for establishing, maintaining and negotiating ownership and (ii) the linguistic means to encode ownership and other types of possessive relationships. Hence, when they learn to talk about ownership and possession children have to acquire both cultural and linguistic knowledge. In this study, we will focus
Learning to encode possession 145
on linguistic aspects, i.e., we will investigate how children acquire the possessive constructions of their target languages – and what this can tell us about the mechanisms that drive children’s linguistic development. Current research on the mechanisms underlying children’s language acquisition is characterised by an opposition between generative approaches (see Eisenbeiß 2009 for overview) and functionalist or usage-based approaches (Bates and MacWhinney 1987; Tomasello 2003, 2006; Goldberg 2006). Moreover, acquisition researchers have provided competing accounts for the time course of linguistic development and for the orders in which children acquire the properties of their target language. In the following, we will first provide an overview of these debates before we present data on the acquisition of possession constructions that can help us evaluate competing models.
2.1. The logical problem of language acquisition The opposition between generative and usage-based accounts of language acquisition arose from discussions about the so-called “logical problem of language acquisition” (Baker 1979; Bertolo 2001; Pinker 1989, and the special issue of The Linguistic Review 19, 1–2). Children only hear a limited sample of their target language, and hence have to generalise over individual input utterances in order to comprehend and produce new utterances. If children’s hypothesis space for these generalisations were completely unconstrained, children might make incorrect generalizations and one would have to explain how they would ultimately reject them. For example, German children frequently hear the possessive marker -s on nouns that encode PRs (e.g. Susi-s Huhn ‘Susi’s chicken’, Oma-s Haus ‘granny’s house’). This might lead them to use this marker with any PR noun or noun phrase. However, this would not be the appropriate generalisation for German: -s can only be affixed to proper names like Susi and a few kinship terms such as Mama ‘mommy’ that can function as proper names (see Harbert 2007: 161ff. and the discussion below). Moreover, even with this restricted set of nouns, -s cannot appear when the noun is modified, e.g. by a determiner or possessive pronoun (*meine Mamas Huhn ‘my mommy’s chicken’). As we will show below German children do not always restrict the use of -s to unmodified proper names and kinship terms, but overgeneralise it to unmodified count nouns such as Affe ‘monkey’, which cannot be combined with -s in the target language (Mills 1985; Eisenbeiß 2000). Thus, one has to explain why children produce such non-target-like combinations, how
146 Sonja Eisenbeiß, Ayumi Matsuo and Ingrid Sonnenstuhl they overcome these “errors”, and how they learn to use morpho-syntactic forms appropriately. One could argue that children “unlearn” errors on the basis of negative evidence, i.e. information about the ungrammaticality of their utterances. However, many studies have shown that explicit corrections are not systematically available to all children at all developmental stages (Marcus 1993). Moreover, even for explicit corrections such as You can’t say that, it is not obvious whether the correction refers to the phonological or morphosyntactic structure, the use of lexical elements, or the appropriateness of the utterance in the social context. In addition, children do not always take up corrections – and even if they do seem to take them up, they might later go back to their non-target like structures; see Marcus (1993) for an overview and the following example from Simone Miller (2;4,1 see Miller 1976; Eisenbeiß 2003: 45):2 (1)
Father:
Wem gehört der Löffel? whom-DAT belongs the-NOM spoon? ‘To whom does the spoon belong?’
Simone: Ich. (correct: Mir) I-NOM (correct: me-DAT ) ‘I’ (correct: ‘to me’) Father:
Wem gehört der Löffel? whom-DAT belongs the-NOM spoon? ‘To whom does the spoon belong?’
Simone: Ich. Ja. I-NOM. Yes ‘I’ (correct: ‘to me’)
1 2
Age information is provided in the following format: Year; Month. We have used the following abbreviations and glosses: [#] – pause; [ /] – interruption; ACC – accusative; DAT – dative; D-elements – case /gender /numbermarked articles, possessive pronouns, demonstratives, quantifiers; EPC – external possession construction (I hit him on the head); FEM – feminine; GEN – genitive; IPC – internal possession construction (I hit his head); MASC – masculine; NEUT – neuter; NOM – nominative; PART – particle (note that as it is difficult to provide exact translations for German focus and other particles, particles will simply be glossed as PART, without further information); PM – Possesssum; PR – Possessor; SG – singular; and TAG – tag question.
Learning to encode possession 147
Father:
Simone:
Father:
Simone:
Father:
Simone:
Father: Simone: Father:
Simone:
Wem gehört der Löffel? whom-DAT belongs the-NOM spoon? ‘To whom does the spoon belong?’ Ich. I-NOM ‘I’ (correct: ‘to me’) Mir. Wem gehört der Löffel? me-DAT . Whom-DAT belongs the-NOM spoon? ‘To me. To whom does the spoon belong?’ Mir. me-DAT ‘To me’ Wem gehört der Löffel? whom-DAT belongs the-NOM spoon? ‘To whom does the spoon belong?’ Mir. me-DAT ‘To me’ Mir. Und das bist Du. ne? me-DAT . And that is you. TAG? Ja. Gehört mir. [...] yes. belongs me-DAT Wem gehört der Löffel? whom-DAT belongs the-NOM spoon? ‘To whom does the spoon belong?’ Ich. I-NOM. ‘I’ (correct: ‘to me’)
Other types of negative evidence do not seem to provide sufficient substitutes for explicit corrections: indirect negative evidence, i.e. the lack of a particular form or construction in the input, might be informative, but taken on its own, it is not reliable enough. There are many types of structures that children are never exposed to (e.g. sentences with long extractions), but are still not considered ungrammatical by adults. This suggests that children do not simply stop adding possessive -s to common nouns like Affe ‘monkey’ because they never hear possessive constructions like that. Some acquisition researchers have argued that certain types of parental responses – e.g.
148 Sonja Eisenbeiß, Ayumi Matsuo and Ingrid Sonnenstuhl repetitions with reformulations – occur more frequently after errors than after correct utterances (see Marcus 1993; Cowie 1999 for references and discussion). However, even such “noisy” negative evidence is not provided systematically enough and it is not clear how exactly children could make use of it (Marcus 1993). Hence, acquisition researchers have to explain how children can generalise beyond individual input utterances, but recover from non-target-like generalisations even though they cannot rely on explicit, indirect or noisy negative evidence. Faced with this “logical problem of language acquisition”, generative linguists have postulated that children’s language acquisition is guided by innate well-formedness constraints that apply to all grammatical structures of human languages. These innate constraints are assumed to restrict children’s hypothesis space so that they can only make target-like generalizations or generalizations that can be rejected without explicit correction. In the early days of generative grammar, the universal well-formedness constraints were formulated as domain-specific principles; i.e. they were considered to be specifically targeted to the domain of language (Chomsky 1965, 1981). Thus, the child’s language acquisition mechanism was viewed as a domain-specific mechanism that cannot be derived from other cognitive or social skills. Recently, generative linguists in the minimalist framework (e.g. Chomsky 1995, 2001) have started to derive the domain-specific principles of early generative grammar from more general cognitive principles, for instance, economy principles according to which grammatical operations (e.g. movement) are only allowed if they are required to fulfill other wellformedness constraints (see Chomsky 1995, 2001; Eisenbeiß 2003, 2009 for overviews). This shows some convergence with the functionalist or usage-based approaches to language acquisition, which assume that general cognitive and socio-pragmatic principles suffice to constrain children’s hypothesis space because children’s input provides rich, structured information about the target language (see e.g. Bates and MacWhinney 1987; Tomasello 2003, 2006). In particular, most generative and usage-based approaches to human languages and their acquisition now assume some version of the Specificity Principle, a general principle that requires operations with more specific outputs or inputs to have precedence over operations with less specific outputs or inputs (see Eisenbeiß 2003, 2009 for discussion). This principle has been formulated as a domain-specific Elsewhere Principle (Kiparsky 1982) or Blocking Mechanism (Marcus et al. 1992) and as a pragmatically based general Principle of Contrast (Clark 1987); and the function of the Specificity Principle has been investigated in connectionist simulations (Corina
Learning to encode possession 149
1994). Whatever its precise formulation, such a general Specificity Principle can, for instance, explain why an adult who has acquired the specific irregular past tense form went would not use overgeneralised forms like *go-ed: the presence of the specific irregular form would block or pre-empt the application of the general past tense -ed-affixation. Moreover, acquisition researchers have suggested that a Specificity Principle could help children overcome morphological overgeneralisations such as the overgeneralisation of the plural -s to irregularly inflected nouns (e.g. *mouses instead of mice; see e.g. Bates and MacWhinney 1987; Braine and Brooks 1995; Clark 1987; Eisenbeiß 2003, 2009; Marcus et al., 1992; Tomasello 2003, 2006). For such morphological overgeneralisations, where a specific form is competing with an overgeneralised form, any use of the appropriate specific adult form in the child’s input could provide a competitor for the child’s non-target-like general form – and thus lead to the child abandoning the non-target-like overgeneralised form. Recognising adult morphological forms and their contrast to the nontarget-like forms is supported by several characteristic properties of children’s input. In particular, cross-linguistic and cross-cultural studies have shown that child-directed speech is rich in so-called “variation sets”, i.e. sequences of utterances with a constant communicative intention, but minimal variations in word order, lexical choice or inflectional form (Eisenbeiß 2003; Küntay and Slobin 1996, 2002; Slobin et al. submitted). For instance, the variation set in (2a) contains both the singular and the plural form of mouse, which highlights these forms and the paradigmatic contrast between them. In this way, variation sets with appropriate adult forms might make competitors for the child’s non-target-like forms more salient and support their acquisition. Other properties of child-directed speech that highlight contrasts between adult and non-target-like child forms are reformulations such as (2b), where the adult form is directly contrasted with the child’s incorrect form. Such reformulations seem to be frequent in Western cultures, but it is not clear whether they are common across the world (see Chouinard and Clark 2003 for discussion). Nevertheless, where they occur, reformulations might offer additional support for children’s acquisition of morphological forms. (2)
a. Look there are so many mice in this picture: There is one mouse under the table and one mouse under the chair and two mice under the bed and three mice in the corner. b. Child: There are the two mouses again! Adult: Yes, we have seen these two mice before.
150 Sonja Eisenbeiß, Ayumi Matsuo and Ingrid Sonnenstuhl Note, however, that reformulations do not necessarily indicate that the child’s utterance is incorrect. For instance, if children use my to refer to their own possession and the parent replies using your, this should not be taken as a correction of the child’s use of my, which was appropriate, but spoken from a different perspective than the mother’s utterance. Thus, in sum, any use of the appropriate adult form that occurs in the child’s input could provide a competitor for a child’s overgeneralised inflected form. In addition, variation sets or reformulations might highlight these competitors and their contrast to the non-target-like form and thus provide further support. Appendix A provides a transcript from a German family that shows how a short everyday conversation can provide children with many models and contrasts for adnominal possessive constructions (A1, A2, A10, A11, A12, A14, A16), predicative possessive constructions (A18, A20, A21) and EPCs (A22, A23). However, more work on the role of contrasts in the child’s input is needed. Note, however, that the mechanisms that might help children to drop overgeneralisations such as *mouses from their language cannot help them to overcome overgeneralisations of German possessive -s to common nouns such as Affe ‘monkey’: in contrast to the domain of morphology, the domain of syntax allows more variation. For instance, in adult German, there are several alternative adnominal possessive constructions: the possessive -s construction (Mama-s Huhn ‘mommy’s chicken’), the prepositional construction (das Huhn von der Mama [the chicken of the mommy] ‘mommy’s chicken’) and the rather formal and much rarer genitive construction (das Huhn der/meinerGEN Mama [the chicken the/my-GEN mommy] ‘the/my mommy’s chicken’). Thus, when children hear prepositional possessive constructions like das Huhn von der Mama, they should not stop using Mamas Huhn ‘Mommy’s chicken’ as this -s possessive is actually correct in adult German. Consequently, hearing die Banane von dem Affen ‘the banana of the monkey’ should not directly drive out an error like *Affes Banane ‘monkey’s banana’. Thus, we will have to determine how children can learn how to use alternative possessive constructions and handle the constraints for their use in the target language. At the same time, we will have to explain why some errors appear in children’s data, while other potential errors have not been observed. For instance, we will show that though German children overgeneralise -s to unmodified common nouns that are incompatible with this marker, children do not seem to overgeneralise -s to full phrasal PRs (e.g. *meine Mamas Huhn ‘my mother’s chicken’). This requires an explanation as -s overgeneralisations are documented in German child language and phrasal possessive
Learning to encode possession 151
markers are not uncommon in natural languages – what is more, they occur in English, which has adnominal possessive constructions that are otherwise very similar to the corresponding German constructions. Thus, when we discuss the acquisition of possessive constructions in the following sections, we will address the logical problem and explain how children manage to avoid or overcome deviations from the target language without recourse to reliable negative evidence. In addition, we also have to capture the time-course of children’s linguistic development, i.e. explain why children acquire particular properties of their target language at a particular time and in a particular order. 2.2. The time-course of linguistic development Most acquisition researchers today, whether they subscribe to a generative or a functionalist view, agree that the mechanisms that drive language acquisition do not change qualitatively over time (see e.g. Pinker 1984; Eisenbeiß 2009; Tomasello 2003 for overviews and discussion). Moreover, there seems to be an emerging consensus that children acquire the most basic generalisations of their target language very early in their linguistic development (Guasti 2001; Tomasello 2006). This has for instance been shown for the basic word order patterns or the ways in which motion events are encoded (see e.g. Slobin et al. submitted). The early adaptation to the target language can be captured in generative and in functionalist approaches alike, though the explanations differ: generative authors highlight the role of innate constraints (e.g. Crain 1991), others point out children’s powerful predispositions for pattern detection and analogy formation (e.g. Tomasello 2006), while others refer to “helpful” input properties such as variation sets (Slobin et al. submitted). Thus, if our investigations should show early adaptations to the target language for the domain of possession constructions, this would be compatible with a broad range of theoretical approaches to language acquisition and provide further evidence for them. In addition, investigating children’s “errors” in the acquisition of possession constructions can contribute to the ongoing debate about children’s grammatical representations at different stages of development. Proponents of generative Full-Competence approaches claim that children’s morphosyntactic representations are adult-like as soon as they start to combine words (e.g. Hyams 1996; Lust 1994; Rizzi 1993/1994, 2000; Wexler 1998). Thus, they cannot view omissions of possessive markers or other deviations from the target language as results of non-adult grammatical representations.
152 Sonja Eisenbeiß, Ayumi Matsuo and Ingrid Sonnenstuhl Rather, any deviations from the target must be attributed to other factors; and Full-Competence proponents have suggested a number of such factors, e.g. the late maturation of general cognitive constraints, problems with the morphological or phonological realization of unstressed morphemes and underdeveloped pragmatic knowledge (see Guasti 2002; Eisenbeiß 2009 for overviews). In contrast to such Full-Competence approaches, Structure-Building approaches do not assume adult-like representations for the early two-word stage. Hence, they must account for children’s early grammatical representations and explain how children acquire adult-like representations. Most current Structure-Building approaches adopt versions of the Lexical Learning Hypothesis (see Pinker 1984; Eisenbeiß 2007, 2009 for overviews; see Radford 1990 for an earlier maturational approach). According to this hypothesis, children have adult-like categorisation abilities, but they still need to determine the grammatical features and properties of the input elements they encounter. Recent versions of Lexical-Learning approaches assume that this process is incremental: grammatical distinctions are acquired one by one, lexeme by lexeme, and with initial restrictions of inflections to individual lexemes that frequently appear with these markers in the input (see e.g. Eisenbeiß 2003, 2007). For instance, possessive markers should initially be restricted to individual words – and only later generalised to all words or phrases that can carry this marker in the target language (Eisenbeiß 2000). Usage-based approaches make similar predictions as Lexical-Learning approaches, though on the basis of slightly different assumptions. According to them, adults grammars are based on schemas or constructions, i.e. interrelated form/meaning pairs that are characterized by various degrees of abstractness, ranging from idioms with concrete lexical items (e.g. kick the bucket ‘die’), to abstract templates characterized by grammatical roles such as subject-predicate (e.g. Goldberg 1995, 2006). Children are assumed to acquire such templates step by step; beginning with limited generalisations that are centred on individual words or phrases and then gradually extending these generalizations by analogy. For instance, Tomasello (2003, 2006) argues that grammatical morphemes such as case markers and agreement markers are initially associated with individual (high frequency) verbs. Only when a critical mass of such “verb-islands” is learned do children acquire more general constructions, e.g. the transitive construction. Thus, taken together, both generative and usage-based approaches predict that children acquire the general properties of the possessive constructions that they hear in their input already in the two-word stage. Moreover, structure-building approaches as well as usage-based approaches specifi-
Learning to encode possession 153
cally predict an incremental acquisition of the specific properties of individual possessive constructions and markers, with initial lexical restrictions of possessive markers. By contrast, capturing such a developmental path would require additional assumptions in a full-competence approach, for instance, reference to the interaction between linguistic development and the learning of cultural norms for possessive relations and negotiations about them. With respect to the order in which linguistic constructions are acquired, most current approaches to language acquisition only make very general predictions that are not directly relevant for the acquisition of possessive constructions.3 However, most acquisition researchers agree that conceptual complexity may influence acquisition orders (see Eisenbeiß 2006, 2009 for discussion). In the following, we will discuss several factors that affect conceptual complexity and could thus influence developmental orders: in the section about attributive possession, we will explore the conceptual differences between possessive relations that involve ownership, kinship relations, body-part relations, etc. Here, we will argue that more prototypical types of possession, which involve physical control and proximity (Heine 1997; Seiler 1983) are acquired earlier than more abstract notions of possession. In the section on predicative possesion, we will discuss the observation that possession constructions have often developed on the basis of older locative constructions that encode a more concrete and “visible” relationship; and we will investigate whether we can observe a similar pattern in children’s linguistic development. Moreover, we will study when children start to produce utterances that involve different types of possessive relations, e.g. ownership vs. current physical proximity and control in an utterance like I have my mommy’s glasses now. In the section on external possession, we will investigate EPCs like I hit him on the head and internal possession constructions (IPC) like I hit his head. Both constructions involve an Agent-Patient relation between the hitter and the hittee as well as a possessive relation between the Patient and 3
For instance, Radford (1996) argues that morpho-syntactic realizations of the functional category COMP (complementizers, wh-elements, etc.) are acquired later than realizations of the functional category INFL (tense and agreement inflections, etc.). Moreover, some generative psycholinguists argue that the ability to produce adult-like passive sentences only develops around the fourth birthday, due to neural maturation of the underlying mechanisms (e.g. Borer and Wexler 1987). See Eisenbeiß (2009) and Tomasello (2003) for critical discussion.
154 Sonja Eisenbeiß, Ayumi Matsuo and Ingrid Sonnenstuhl their body part. However, in the EPC, both relations have to be mapped onto a single argument hierarchy as both PR and PM are realised as verb arguments. By contrast, in IPCs, the Agent-Patient relation is realised on the sentential level while the possessive relationship is encoded within the Patient noun phrase. We will argue that this is conceptually simpler than the integration of agentive and possessive relations in the EPC – which should be reflected in acquisition orders and possibly also in deviations from the target language. 3. The German data Our study of German child language is based on 64 recordings from 7 monolingual German children, aged 1;11 to 3;6. The data come from the Clahsen corpus, the LEXLERN corpus (see Clahsen, Vainikka, and YoungScholten 1990) and the Wagner corpus (1985; see Appendix B for details). Some of the LEXLERN recordings involved games in which possessions were compared or exchanged (see Eisenbeiß 1994, 2003). Therefore, they are comparatively rich in possessive constructions and some aspects of possessive constructions have already been studied in these corpora (Eisenbeiß 2000). In order to analyse the course of development and to create comparable data sets from children at similar stages of development, we assigned the individual recordings to four developmental stages. We did not use chronological age to group the data as age is not considered to be a good indicator of linguistic development. MLU (mean length of utterance) is also not sufficient as some of the later recordings have MLUs of more than 3 words per utterance; and MLU-values in that range are not as indicative of linguistic development as lower values. Thus, we had to find another way of grouping the recordings for comparisons. Noun-phrase development is obviously crucial for adnominal possession constructions; and the case markers in noun phrases are central for predicative possession constructions and EPCs. Therefore, we decided to group our recordings on the basis of the four-stage model of noun-phrase development that Eisenbeiß (2000, 2003) had proposed in her analysis of German child language data and applied to our data set (see Appendix B). STAGE I
Children frequently omit possessive markers and other morphological markers. They also omit D-elements, i.e. function words in the noun
Learning to encode possession 155
phrase (e.g. case/gender/number-marked articles, possessive pronouns, demonstratives, quantifiers). At the same time, the occurrence of Delements is restricted to potentially formulaic combinations with a small set of high-frequency predicates and nouns (e.g. das-is-ein-X ‘that-is-aX’, die-mama ‘the mommy’). STAGE II
The percentage of D-elements falls to even lower levels. STAGE III
The percentage of D-elements increases again. Formulaic utterances become less frequent and children start to combine D-elements more freely with a broader range of predicates and nouns. STAGE IV
Children typically produce D-elements whenever they are required. Taken together, this U-shaped development in the provision of function words and the early distributional restrictions for these elements suggest that these elements are initially part of unanalysed chunks that are later reanalysed. Hence, it is crucial not to mix data from the different stages for analysis. 4.
Adnominal possession
In the following, we will first give an overview of adnominal possession constructions in adult German and contrast them with the corresponding constructions in other languages for which acquisition studies are available – English, Japanese, Greek, and Hebrew. Based on this, we will discuss empirical findings from available acquisition studies and our own study. 4.1. Adnominal possession in the adult language In order to illustrate the typological differences with respect to adnominal possession constructions and their implications for language acquisition, we will first look at two closely related languages, namely German and English. In both languages, PRs can either be referred to by pronominal elements as in (3a, b) or by non-pronominal noun phrases as in (3c–i). Moreover, both pronominal and non-pronominal PRs can be combined with a preposition as in (3b, d–f) or they can appear without it as in (3a–c, g–i). The prepositional
156 Sonja Eisenbeiß, Ayumi Matsuo and Ingrid Sonnenstuhl PR phrase has to follow the PR in English as in (3b, d), and it tends to appear in the same position in German as in (3b, d). However, in German, at least some speakers find it acceptable to position the prepositional PR phrase to the left of the PR for emphasis as in (3e) or to extract the prepositional PR phrase in a question – see (3f). Note, however, that structures where the prepositional phrase does not follow the PR are highly marked – if not unacceptable. In addition to the adnominal possession constructions in (3a) to (3f), German has two constructions that do not have an equivalent in English. The first one is a genitive construction, where the PR noun phrase typically follows the PM as in (3g), but may also precede it as in (3h). This genitive construction is more characteristic of formal and written German, has so far not been observed in the speech of pre-school children (Clahsen et al. 1994; Eisenbeiß 2000; Mills 1985), and does not occur in the corpora we analysed. Note that the genitive construction is different from the possessive -s construction though this is not immediately obvious for masculine nouns that take -s as their genitive ending (see (3g, h)). The difference between the -s possessive and the “real” genitive can be seen when one looks at masculine nouns with a different genitive ending or femine nouns, which do not carry any overt marker in the genitive (see (3i)). The second German possessive construction without an English equivalent appears in some spoken variants of German and involves a PM that is preceded by a dative-marked PR and a resumptive possessive pronoun, as in (3j). Due to its regional character, this construction is rarely discussed in the acquisition literature (see Penner and Weissenborn 1994 for some initial observations), and we only found three instances of this type in the data of the German boy Carsten that we analysed. Therefore, we will not discuss adnominal genitive and dative constructions in a lot of detail. (3)
a. sein Freund his friend ‘his friend’ b. ein Freund von ihm a friend of his ‘a friend of his’ c. Pauls Freund Paul’s friend ‘Paul’s friend’
Learning to encode possession 157
d. ein Freund von Paul/seinem Vater a friend of Paul /his father ‘a friend of Paul’s’ / ‘a friend of his father’ e. Das ist bestimmt VON PAUL der Freund that is surely OF PAUL the friend ‘That is surely Paul’s friend’ f. Von wem hast du den Vater gesehen? of whom have you the father seen? ‘Whose father have you seen?’ g. ein Freund seines Vaters a friend his-GEN father’s ‘a friend of his father’ h. Pauls / seines Vaters Freund Paul’s / his-GEN father’s friend ‘Paul’s friend’ / ‘his father’s friend’ i. eine Freundin meiner Mutter / des Jungen a friend my-GEN mother’s / the-GEN boy’s ‘a friend of my mother’ / ‘a friend of the boy’ Paul / Vater sein Freund j. dem the-DAT Paul / father his friend ‘Paul’s friend’ / ‘the father’s friend’ Given the range of constructions mentioned above, both German and English children have to learn (i) when to use pronominal or a non-pronominal constructions and (ii) when to use a construction with a non-prepositional PR rather than a construction with a PR-PP. Moreover, German children have to acquire adnominal constructions with genitive PRs or a combination of dative PRs and possessive pronouns – though we won’t be able to investigate this aspect of the acquisition process due to a lack of relevant data. With respect to the choice of pronominal vs. non-pronominal constructions, German and English are similar: possessive pronouns are preferred when the PRs can be identified on the basis of contextual or discourse information. As this is particularly easy for speakers and hearers, adults tend to use 1st and 2nd person possessive pronouns rather than names when they talk to other adults (e.g. mein-e/dein-e Henne ‘my/your chicken’). However, in children’s early language and in language directed at young children, we can often observe the use of names and kinship terms instead of 1st and 2nd person possessive or personal pronouns (e.g. Ruff 2000). For instance, a mother might tell her daughter Jane: Das ist Jane-s Auto; und das ist
158 Sonja Eisenbeiß, Ayumi Matsuo and Ingrid Sonnenstuhl Mama-s Auto ‘this is Jane’s car; and this is mommy’s car’. Thus, we will have to investigate when and for which types of PRs children use possessive pronouns. Whether a prepositional or a non-prepositional construction is chosen, depends on a range of factors. The first factor is the syntactic status of the PM: if it is an unmodified noun such as the proper name Paul, it can either appear with a prepositional PR as in (3d) or with a non-prepositional PR as in (3c). However, possessive pronouns as well as -s-marked PRs induce a definite reading and cannot co-occur with determiners or wh-elements (*ein / dieser/welcher Pauls/ mein Freund ‘a/ this / which Paul’s / my friend). Thus, the use of a wh-question or an intended indefinite reading may require a prepositional construction. We will now look at factors that play a role if the choice of construction is not already determined by the type of PR. For pronominal PRs, English or German speakers have a choice between a possessive pronoun construction as in (3a) and a prepositional construction with a pronoun as in (3b); the possessive pronoun construction is typically preferred as PR phrases tend to proceed the PR, especially when they are short, animate and topical and the PR is not used contrastively, see e.g. my friend vs. ?the friend of mine. For non-pronominal PRs, English and German differ with respect to the factors that determine their choice of construction. In English, -s can attach to non-pronominal PR phrases of any syntactic complexity (e.g. Jack’s/the old farmer’s chicken farm). Similarly, prepositional constructions can involve simple PR nouns or complex PR noun phrases (e.g. a teacher of Jane/ my little daughter). Thus, for non-pronominal PR phrases, English speakers have a choice between -s and of. This choice is determined by the animacy, topicality and syntactic weight of the PR and by the type of possessive relationship: -s is preferred for prototypical and inalienable possessive relations and when the PR is animate, topical and short (e.g. Sue’s eyes) – whereas of is preferred when the PR is inanimate, not topical and syntactically modified and the relationship between PR and possesum is not a close and prototypical possessive relation (e.g. the fumes of a shabby old car; see e.g. Rosenbach 2002, 2005, 2008; Jäger and Rosenbach 2006; Denison, Scott, and Börjars 2008 and references cited there for the discussion of these factors and their interaction). Note, however, that while these factors can conspire to make one construction highly preferable over the other, none of these factors on its own can determine the choice of construction. By contrast, German exhibits a constraint for the use of possessive -s that cannot be violated and hence can uniquely determine the choice of non-pronominal possessive construction: -s can only be combined with PR
Learning to encode possession 159
nominals that lack an article or any other modifiers. Typically, these nouns are proper names like Susi, but some kinship terms can also be used without a determiner and can thus be combined with -s (e.g. Mamas/Papas Auto ‘mommy’s / daddy’s car’). This has led some linguists to distinguish possessive -s markers from genitive markers by calling them “proper name possessive markers” (see e.g. Harbert 2007: 161ff.). We will argue that the constraint on the use of -s is a syntactic constraint and not a restriction of -s to a particular semantic class of nouns. In many regional variants of colloquial German, kinship terms or proper names appear with determiners (e.g. die/ eine Mama ‘the/a daddy’ or die Emma ‘the Emma’). However, when such phrases are used as PRs, -s cannot be used (e.g. *der Mamas/ Emmas Auto ‘the mommy’s/Emma’s car’). Rather, a prepositional construction is chosen (e.g. das Auto von der Mama/ Emma ‘the car of the mommy/ Emma). Thus, it is not the type of noun per se that determines whether -s can appear, but the lack vs. presence of determiners or other modifiers. Given the syntactic constraints for -s, German speakers can only choose between -s and the prepositional von-construction when the PR is an unmodified proper name such as Emma or name-like kinship term such as Mama ‘mommy’. As the referents of these nouns are all animate and the length of the PR phrase is limited to one word, animacy or syntactic weight cannot determine the choice between an -s construction like Annas Auto ‘Anna’s car’ and a prepositional construction like das Auto von Anna ‘the car of Anna’. However, the type of possessive relationship might play a role, with -s being preferred for closer and more prototypical possessive relationships. To our knowledge, this has not yet been investigated thoroughly. To summarise, German and English both have possessive constructions with pronominal and non-pronomial PRs; and the choice of pronominal vs. non-pronominal PRs is determined by similar pragmatic factors. Moreover, in both languages, possessive pronouns are preferred to prepositional phrases with pronominal PRs – unless the PR requires a modifier. However, in German and English non-pronominal possession constructions, additional factors play a role: the choice between -s and prepositional constructions for non-pronominal PRs is determined by semantic and discourse factors in English, whereas the use of the German possessive marker is restricted to particular syntactic environments – i.e. unmodified nominals. Thus, a choice between -s and the prepositional construction is only available for proper names and a few kinship terms that can appear without a determiner. Not all languages show such a competition between a prepositional construction and a construction with a possessive marker. For instance, in Japa-
160 Sonja Eisenbeiß, Ayumi Matsuo and Ingrid Sonnenstuhl nese, all PR phrases, whether they are personal pronouns, unmodified nouns or more complex noun phrases, are marked by the postposition no and precede the PM (e.g. watashi/Toshiko no kuruma ‘I/Toshiko’s car). In Hebrew, the PR is marked by the preposition shel and follows the possesum (e.g. pe shel buba ‘(the) mouth of (the) doll’). The PR < PR order can also be observed in constructions with possessive pronouns (e.g. hasefer sheli ‘the book (of) my/mine’; see e.g. Armon-Lotem, Crain, and Varlokosta 2005). Even in a language with only one possessive marker for non-pronominal constructions, children may have to acquire different word order patterns: For instance, in Standard Modern Greek, the genitive-marked PR is compatible with determiners for the PR and can either precede the PM as in (4a) or it can follow it as in (4b); see Marinis (2002, 2003): (4)
a. Pira tu nikous took the-GEN Nikos-GEN ‘I took Niko’s book.’ b. Pira to vivlio took the-ACC book-ACC ‘I took Niko’s book.’
to vivlio. the-ACC book-ACC tu Niku. the-GEN Nikos-GEN
Thus, when children acquire the adnominal possession construction of their target language, they have to learn the different markers and the syntactic, semantic, and discourse constraints on their use. They also have to determine the word order options of their target language.
4.2. Adnominal possession in child language In the following, we will first focus on the developmental problem and investigate when children start to produce the adnominal possessive constructions of their target language and whether their earliest uses of these constructions are restricted to particular lexical items – as predicted by structure-building and usage-base approaches. Then, we will study whether children show early sensitivity to the language-specific constraints that govern the choice of construction. Against this background, we will then try to provide an account for the order in which adnominal possessive constructions are acquired and used to encode different types of possessive relations.
Learning to encode possession 161
4.2.1. The course of development Table 1 gives an overview of adnominal possession constructions in our German child language data. Table 1. Adnominal possesssive constructions Pronominal PR
Non-Pronominal PR
Stage Child
possessive pronoun
von + pronoun
I
Ann Han Leo Mat Total
6 – 0 3 9 (30%)
– – – – 0 (0%)
1 – 17 3 21 (70%)
II
Ann Han Leo Mat Total
3 – 1 – 4 (33%)
– – – – 0 (0%)
III
And Ann Han Leo Mat Total
63 15 – 17 7 102 (77%)
IV
Ann Car Han Leo Mat Sve Total Total
von + NP
Total
– – – – 0 (0%)
– – – – 0 (0%)
7 0 17 6 30
1 – 6 – 7 (58%)
– – 1 – 1 (8%)
– – – – 0 (0%)
4 8 – 12
– – – – – 0 (0%)
2 3 – 2 – 7 (5%)
2 – 1 20 – 23 (17%)
– – – – – 0 (0%)
67 18 1 39 7 132
25 105 4 26 31 75 266 (86%)
– 1 – – – 1 2 (<1%)
– – – – – – 0 (0%)
– – 1 7 5 17 30 (10)
– 4 1 – 1 5 11 (4%)
25 110 6 33 37 98 309
381 (79%)
2 (<1%)
35 (7%)
54 (11%)
11 (2%)
483
PR+PR w/o -s
PR+PR w. -s
This tabulation suggests that there might be an early stage without adnominal possessive constructions: Hannah does not produce any of these constructions in stages I and II and only one proper name possessive construction in stage II. In stage I, she mostly labels or points out objects and actions or asks for things, but does not talk about possessions. From stage II on, possessive relations become a talking point for her and the possessive
162 Sonja Eisenbeiß, Ayumi Matsuo and Ingrid Sonnenstuhl pronoun mein(s) ‘my/mine’ appears, but only in one-word utterances, not in combination with a PM noun. This seems to be a precursor for adnominal possessive constructions that emerge in stage III (see Ruff 2000 for similar observations). When adnominal possessive constructions emerge, they seem to do so incrementally: prepositional constructions can only be found in data from stage IV – and even then only Carsten and Svenja produce more than one of these constructions, see e.g. (5a,b). This suggests that prepositional constructions are a late acquisition, though they seem to be acquired earlier than the more formal genitive construction, which does not occur in our data at all. Similar observations have been made by Ruff (2000) in her longitudinal study with 7 German children (age range at beginning of recordings: 2;0 –2;9; age at the end: 2;4 –3;2). (5)
a. mama und wo is der kopp hier mommy and where is the head here von den junge von diesen? of the boy of this-one? ‘Mommy, and where is the head of this boy?’ b. ich bin der hund von Sewenja (= Svenja). I am the dog of Svenja. ‘I am Svenja’s dog.’
(Carsten / 3;6)4
(Svenja/14 / 3;2)
Both proper noun possessive constructions as in (6 a) and possessive pronoun constructions as in (6b) can already be found in stage I, though proper noun possessives seem to emerge earlier. In stage I, Leonie only produces 17 proper noun possessives, but no possessive pronoun, and in stage II she uses 7 proper noun PRs, but only one possessive pronoun. And when Hannah starts to use adnominal possessive constructions, her first construction contains a proper noun PR. Moreover, we can observe a shift towards pronominal PRs: In stages I and II, only 31% (13 /42) of all adnominal possessive constructions involve a pronominal PR, compared to 84% (370 /441) in stages III and IV. In stage IV, each of the children uses predominantly pronominal PRs. Carsten does not use proper name possessive constructions at all, but this is probably due to the fact that he uses the dative possessive construction, which is a regional variant, instead (see (6c)). However, even this construction only occurs three times. 4
Recording information is provided in the following format: (Name/number of recording /age).
Learning to encode possession 163
(6)
a. S.E.:
das is mamas koffer. ne? (Leonie/ 01/01;11) this is mommy’s suitcase. TAG? und das da? and that there? ‘This is mommy’s suitcase, isn’t it? And that one over there (is)?’ Leonie: mann koffer. man(’s) suitcase ‘(the) man(’s) suitcase’
b. mein puppe i weg. my doll is gone. ‘My doll is gone.’ c. das is oma ihr lesezeichen mama that is granny her bookmark, mommy ‘That is granny’s bookmark, mommy.’
(Annelie/02 /2;5)
(Carsten /3;6)
The observed shift towards pronominal PRs cannot simply be due to children initially preferring their own name for self-reference: of the seven children, only Annelie, Hannah, and Leonie ever employ their own name as PRs; and Annelie and Leonie use both the possessive pronoun and their own name as soon as they talk about themselves as PRs. Hannah produces her name only once in an adnominal possessive construction (stage II) before she starts using both the possessive pronoun and her own name (stage III). But recall that she already used the possessive pronoun in one-word utterances in stage II. Thus, those children who use their own name for selfreference in adnominal possessive constructions do not seem to do so because they have not yet acquired pronouns. Rather, they exhibit an alternation between pronouns and their names in these constructions. As we do not have video data available to study the details of the situational context, we cannot investigate all factors that determine when a possessive pronoun is chosen. However, an alternation between pronominal and non-pronominal references to the speaker as PR has also been observed by Ruff (2000) who argues that possessive pronouns are initially used in demands and conflict situations, whereas non-pronominal PRs appear in descriptive utterances. Moreover, Ruff reports that German children use pronouns more frequently when the speaker is the PR than when the addressee is the PR. Taken together, we observe an incremental development of possessive constructions, with a late acquisition of prepositional constructions and a
164 Sonja Eisenbeiß, Ayumi Matsuo and Ingrid Sonnenstuhl shift towards pronominal constructions. While capturing these observations would require additional assumptions in full-competence approaches, it is explicitly predicted by both usage-based approaches and generative structure-building approaches. These approaches are also compatible with an early stage during which possessive markers are omitted; and they predict initial restrictions to particular lexical elements once they do occur. For possessive -s markers, Eisenbeiß (2000, 2003) has already demonstrated that the German children we are investigating go through an early stage without -s (see also Table 1). Note that in stages I and II, -s is even omitted when the target form occurred in the preceding discourse, see e.g.: (7)
S.E.:
Und Papas Hose brauchen wir noch and daddy’s trousers need we still ‘and we still need daddy’s trousers’ (Leonie/ 03 / 2;1)
Leonie: da papa hose there daddy(’s) trousers. ‘there (are) daddy(’s) trousers.’ An early stage with omissions of possessive markers has also been observed for other German children and in studies of the corresponding possessive markers in English, Greek, Hebrew and Japanese (Clahsen, Eisenbeiß and Vainikka 1994; Penner and Weissenborn 1996; Brown 1973; Radford 1990; Radford and Galasso 1998; Marinis 2002, 2003; Berman 1985; Armon-Lotem 1998; Clancy 1985). Proponents of full-competence approaches have tried to show that possessive markers appear early even if they are not always realised. For instance, Bohnacker (1997) argued that possessive markers appeared in the data of a young Swedish child (Embla; age: 1;8–2;1). However, the entire corpus only contained 14 possessive markers; and none of these occur at the beginning of the recording period (Eisenbeiß 2003). Still, it is not clear whether all children in all languages go through a stage in which possessive markers are omitted. Note, however, that structure-building and usage-based approaches do not in principle rule out the possibility that possessive markers appear early – they simply allow for the gradual and incremental acquisition of such markers and the possibility of an early, marker-less stage. In fact, for languages where the possessive marker is frequent, obligatory, salient (e.g. syllabic) and not homonymous with any other marker, proponents of structure-building and usagebased approaches would expect this marker to be acquired quite early.
Learning to encode possession 165
4.2.2. Lexical restrictions of possessive markers and pronouns With respect to the early use of possessive markers, studies of German, English, Japanese, and Hebrew have demonstrated that adnominal possessive constructions with target-like markers can temporarily co-occur with constructions that lack such markers (e.g. Brown 1973; Berman 1985; Clancy 1985; Peters/Menn 1993; Clahsen, Eisenbeiß and Vainikka 1994, 2000; Penner and Weissenborn 1996; Armon-Lotem 1998; Radford and Galasso 1998; Ruff 2000). For instance, Clahsen et al. (1994) showed that the German child Simone provided the -s only in 33 of the 49 proper noun possessives she produced between 2;0,25 and 2;2,21.; and the English child Nicholas used ’s only in 14 of 60 obligatory contexts between 3;2 and 3;6 (Radford and Galasso 1998). A similar stage can be observed in the data of Leonie and Andreas who produced -s only in 8 out of 29 and 2 out of 4 obligatory contexts in stage IV (see Table 1 and Eisenbeiß 2000). Independently of the language under investigation, the early examples of possessive markers that are reported in the literature typically involve kinship terms like dad or the name of the child or close friends and relatives (e.g. Clancy 1985: 458; Mills 1985: 185; Radford 1990: 89; Peters and Menn 1993: 757ff.; Clahsen, Eisenbeiß and Vainikka 1994: 97 ff.; Stenzel 1994: 196f.; Radford and Galasso 1998: 37). This is compatible with the assumption that the use of possessive markers is initially lexically restricted to individual nouns that children hear frequently with this marker. Further evidence for this hypothesis comes from the observation that the Englishspeaking child Daniel combined -s with the name Mike, then with the kinshp term mommy and only later with other PRs (Peters and Menn 1993). Similarly, Leonie goes through a stage where she uses -s with familiar names and kinship terms as in (8a), but not with the name Sonja, which she had not been familiar with before the start of recordings as in (8b). Even though she heard the form Sonjas during the recordings, she does not use this marker herself until the last recording in stage II. Similarly, in the recording we investigated Andreas consistently uses -s with papa ‘daddy’ as in (8c), but always omits it with mama ‘mommy’ as in (8d). (8)
a. is mamis is mommy’s ‘(This) is mommy’s.’
(Leonie/ 05/2;3)
b. S.E.:
(Leonie/ 06 / 2;3)
Und welches ist Sonjas Auto? and which-one is Sonja’s car? ‘and which one is Sonja’s car?’
166 Sonja Eisenbeiß, Ayumi Matsuo and Ingrid Sonnenstuhl Leonie: Sonja autos Sonja cars ‘Sonja(’s) cars’ (correct: ‘car’) c. hm papas gürtel hm papa’s belt ‘papa’s belt’
(Andreas /2;1)
d. e mama ticktack is(t) das e mommy ticktack is that ‘That is mommy’s clock.’
(Andreas/2;1)
One could attribute the early omissions of possessive markers and the initial lexical restrictions to non-syntactic factors, e.g. to the fact that possessive markers are typically unstressed and that some combinations of PRs and possessive markers might be more difficult to hear or pronounce. However, Eisenbeiß (2000, 2003) has argued that phonological factors alone cannot account for the distribution of possessive markers in the corpora we analysed. For instance, the word Papa ‘daddy’, which Andreas uses with -s, and the word Mama ‘mommy’, which appears without -s, have the same syllable structure and the same final vowel. Moreover, a distributional analysis of the possessive pronouns in our corpora offers further support for the assumption of initial lexical restrictions. In stages I and II, only Annelie and Mathias produce more than one possessive pronoun in an adnominal possessive construction; and the 9 instances of possessive pronouns in Annelie’s early data only involve 5 highfrequency nouns (puppe ‘doll’, mama ‘mommy’, schuhe ‘shoes’, zimmer ‘room’, and bilder ‘pictures’). Similarly, in stages I and II, Mathias only produces 3 types of adnominal possessive constructions with the possessive pronouns. Finally, all possessive pronouns in stages I and II are forms of mein ‘my’; forms of dein ‘your’, unser ‘our’ or sein ‘his’ as in (9) do not occur until stages III and IV. (9)
da da is doch seine mama. there there is PART his mommy ‘there is his mommy’
(Svenja/ 08 / 3;0)
At the same time, agreement with the PR is not target-like: except for one use of the correct plural form meine in stage II (meine bilder ‘my pictures’), Annelie only uses the uninflected form mein, which leads to agreement errors for feminine and plural nouns that require the form meine (e.g. *mein mama ‘my mommy’). Mathias produces the uninflected form mein once, in
Learning to encode possession 167
an appropriate context. The remaining two possessive pronouns are incorrectly inflected (*meiner buch ‘my book’, *meiner lappen ‘my rag’). Thus, both the range of possessive pronouns and their combinations are limited initially. This is in line with studies that observed that English-speaking and German children initially only used possessive pronouns when the PR was the speaker (Tomasello 1998; Ruff 2000). 4.2.3. Sensitivity to language-specific constraints While the previous section showed quite a few studies of the time course of the acquisition process in the domain of adnominal possessive constructions, there are fewer studies of the constraints that govern the use and choice of these constructions, which is somewhat surprising given the rich theoretical literature on the topic. We have already discussed the choice between pronominal and non-pronominal constructions above (see Ruff 2000 for more details). With respect to non-pronominal possessives, we observed earlier that the possessive marker -s is restricted to unmodified PR nouns in German – typically proper names like Susi, and some kinship terms (e.g. Mamas/Papas Auto ‘mommy’s/daddy’s car’). Thus, for modified PR nouns (e.g. das kleine Huhn ‘the (little) chicken’), German speakers have to use a prepositional or genitive construction (e.g. der Kopf von dem kleinen Huhn or der Kopf desGEN kleinenGEN HuhnsGEN ‘the head of the little chicken’). This raises questions with respect to German child language: as we observed earlier, prepositional and genitive constructions do not appear in our data in stages I– III and we only found a few prepositional constructions in some files from late stage IV. At the same time, noun-modifying determiners emerge in stage III and appear in nearly all obligatory contexts in stage IV. Thus, what do young German children do when the PR is a common noun that requires a determiner or some other modifier – and is thus incompatible with -s? In principle, they could refrain from using any non-pronominal adnominal possessive constructions with PRs that require a modifier until they have acquired the prepositional or genitive constructions of their target language, but – as we will see – that does not seem to be what happens. Thus, children could do one of three things. The first option is to use an -s possessive but to omit the required noun modifier. This would not violate the German constraint for -s, though it would mean treating a common noun like a proper noun. As reported in Eisenbeiß (2000), two of the children we investigated exhibited this option: Svenja produced one and Leonie pro-
168 Sonja Eisenbeiß, Ayumi Matsuo and Ingrid Sonnenstuhl duced two overgeneralizations to an unmodified PR in an adnominal possessive construction, as in (10a–c). (10) a. S.E.:
Das is? that is? ‘That is (what)?’ Leonie: affes banane monkey’s banana ‘the monkey’s banana’
(Leonie/ 07/ 2;4)
b. clowns hut clown’s hat ‘the clown’s hat’
(Leonie/11/ 2;7)
c. das is junges gürtel this is boy’s belt ‘this is the boy’s belt’
(Svenja/13 / 3;2)
While this is clearly a very small set of examples, more can be found in the literature: see (11) for -s overgeneralizations reported by Mills (1985: 185). What is striking about this set of examples is that it contains a plural noun, which can definitely not be interpreted as a proper name. Thus, children do not seem to violate the constraint on the affixation of -s to modified nouns. They DO violate the requirement to use determiners with count nouns. Note, however, that some German speakers who we showed the examples to commented that adnominal possessive constructions such as affes banane are acceptable if one analyses the respective PR noun as a proper name, which would not require a determiner in German. (11) a. das is männers wagen that is men’s car ‘That is the men’s car’
(Scupin 3;1)
b. hier is männers wohnung here is men’s apartment ‘Here is the men’s apartment’
(Scupin 4;3)
c. da tut männers bauch weh there does men’s tummy ache ‘The men’s tummies are aching here.’
(Scupin 4;4)
d. an elefantes zähne on elephant’s teeth ‘on the elephant’s teeth’
(Scupin 5;8)
Learning to encode possession 169
The second option for children faced with the dilemma of either violating the -s or the determiner constraint would be to use a modified PR and omit -s, thus producing a construction that exhibits a modifier, but does not show the required morphological marking. We do not have a lot of evidence for this option, only one example, which we did not include in our counts as the interruption in it makes it difficult to interpret: (12) darf ich mein mama [/ ] mein mama bademütze haben? may I my mommy [/ ] my mommy swimming.cap have? ‘May I have my mommy’s swimming cap?’ (Annelie/04/2;7) What is striking, however, is that we did not find any evidence for the third option: overgeneralisation of -s to a modified PR noun, i.e. a violation of the constraint on -s that would avoid treating the PR noun like a proper noun. That is, we did not observe any “English-style” constructions such as mein Mamas bademütze ‘my mommy’s bathing cap’. This is quite surprising as children DO obviously overgeneralise -s. Moreover, most of the nouns that the children in our sample correctly use as -s-PRs could be modified in the target language because kinship terms and even proper nouns may be combined with determiners or some other modifiers in most varieties of spoken German – even though they do not HAVE TO. And indeed all children in our sample that produce -s-PRs provide modifiers for the nouns they use as -s PRs when they appear in other contexts: Andreas affixes -s to papa ‘daddy’ when it is an unmodified PR (as in (8c) above), but he combines this noun with a possessive pronoun in other contexts (as in (13a)). Hannah only uses her own name in an -s possessive construction, but at the same time she mostly combines her name with determiners in other contexts, see e.g. (13b). Mathias only uses unmodified proper nouns with -s, but he produces combinations of these nouns with determiners in other contexts (e.g. (13c)). Leonie produces -s overgeneralisations from stage III on, but the PR nouns she uses with -s appear without a determiner, though they are used with modifiers in other contexts, as in (13d). The PR nouns that appear with an overgeneralized -s marker in stage III and IV, occur with determiners during the same stage, as in (13e). Svenja also combines -s with nouns that she otherwise uses with articles and other modifiers, see e.g. (13f), including the common noun junge ‘boy’, which she incorrectly affixes with -s. Moreover, Svenja is one of those children that already show a contrast between the -s construction and the possessive construction; and we find minimal pairs with an unmodified proper name or kinship term in the -s construction and the same name with a modifier in the prepositional construction of (13g, h).
170 Sonja Eisenbeiß, Ayumi Matsuo and Ingrid Sonnenstuhl (13) a. mein Papa macht wieder drauf. my daddy makes again onto-this. ‘My daddy puts (it) on that again.’
(Andreas/2;1)
b. da is die Hannah. there is the Hannah. ‘There is Hannah.’
(Hannah/06/2;6)
c. der Daniel hat das puttmach. the Daniel has that broken. ‘Daniel has broken that.’
(Mathias / 21/ 3;0)
d. un(d) de (/) die große Sonja. and the (/) the tall Sonja ‘And the tall Sonja.’
(Leonie/ 11/ 2;7)
e. da ist wieder der clown mit den kalten füßen, ne? there is again the clown with the cold feet, TAG? ‘There is the clown with the cold feet again, isn’t he?’ (Leonie/ 09 / 2;6) f. nein das is doch kein papa! (Svenja/ 05 / 2,10) no that is PART no daddy! ‘No, that isn’t a daddy!’ g. Saschas hut (Svenja/13 /3;2) Sascha’s hat ‘Sascha’s hat’ h. von die (corr: dem) Sascha. of the-NOM /ACC.FEM.SG (corr:DAT.MASC.SG.) Sascha ‘Sascha’s’ (Svenja/13 / 3;2) Taken together, these findings suggest that children may overgeneralise -s to common nouns, but they do not violate the constraint that prohibits the combination of -s with modified nouns. They do not even use a modifier for proper nouns and kinship terms in possessive -s constructions when they use these nouns with modifiers in other contexts. This supports the assumption that children are sensitive to a syntactic constraint on the use of -s as soon as we find this marker in their data, while they do not show a general restriction of -s to proper nouns and kinship terms. This is in line with our claim that the constraint on -s is a syntactic and not a semantic constraint in the target language. Children’s early sensitivity to the syntactic constraint on -s raises the question why they do not make the incorrect assumption that the constraint on the use of -s is semantic. That is, we have to explain why they do not
Learning to encode possession 171
restrict the use of -s to proper names and kinship terms and overgeneralise -s to modified proper names and kinship terms. We argue that children who hear -s with single nouns would only assume that -s can also be combined with more complex noun phrases if they found positive evidence for this in the input. Thus, children acquiring English would learn to use -s with complex noun phrases because they hear such combinations, whereas German children would never have any evidence that would lead them to extend the use of -s in this way. This means that German children do not have the means to produce any adnominal possessive constructions with modified PR nouns before they have acquired prepositional or genitive constructions. This would leave them only one option: to treat the common PR nouns they want to use as if they were proper names. Note that treating a common noun like a proper name is not that uncommon for nouns with animate referents. For instance, we know quite a few adults who use words such as baby and cat as proper names. Thus, children’s overgeneralisations are not completely outside the limits of the target language, children are simply stretching these limits when they are faced with a dilemma: violating the syntactic constraint on -s or violating the requirement to use a determiner or other modifier with a count noun. Once prepositional or genitive constructions are acquired, there is no pressure to affix -s to common nouns that require a modifier. The idea that children need positive evidence to assume that -s can be used for modified PRs can be captured straightforwardly: one could simply assume that children’s unmarked expectation would be that bound morphemes will concern the word to which they are attached, rather than the phrase. One could also capture this idea in feature-based underspecification models of morpho-syntactic development – such as the one assumed by Eisenbeiß (2003; see Eisenbeiß 2009 for a summary). In such models, grammatical features are only integrated into morphological representations if there is positive evidence in the input that requires them. For instance, children only integrate number features into lexical entries for nouns when they are confronted with contrasts between singular and plural forms. The difference between a simple noun and a complex noun phrase is captured by additional features for phrasal projections in different syntactic frameworks (see e.g. Grimshaw 1994). Thus, one could assume that English children include such a phrase-feature in their input specification for ’s because they find positive evidence in their input that supports this. In contrast, German children do not find such positive evidence and thus do not extend the range of elements that can be affixed by -s.
172 Sonja Eisenbeiß, Ayumi Matsuo and Ingrid Sonnenstuhl Interestingly, there is some initial evidence for such a conservative approach in Carsten’s data. The three adnominal possessive constructions with a dative PR that Carsten produces all involve an unmodified noun, even though the noun Kind ‘child’ requires a determiner in adult German and Carsten uses it with modifiers in other contexts, as in (14a), and produces determiners in almost all obligatory contexts. This is compatible with the assumption that he initially restricts the use of this construction to unmodified nouns. Similarly, Penner and Weissenborn (1996) provide a few examples for the emergence of dative possessive constructions in Swiss German and the earliest examples with an overt possessive pronoun also involve unmodified PR nouns, as in (14b). However, the data base is currently too limited to warrant a stronger conclusion. Moreover, children hear adult-like constructions with modified dative PRs when they learn a variety of German that allows them. Thus, they might acquire the appropriate generalisation early on. (14) a. wi(ll) das kind viele autos? wants the child many cars? ‘Does the child want many cars?’ b. Nadaw sis Ue (=uhr) Nadaw his watch ‘Nadaw’s watch’
(Carsten / 3;6)
(J 1;10 ;19)
We already observed that the syntactic factors that determine the choice of possessive construction in German are not relevant for English, where factors such as animacy play a crucial role. So far, no published study seems to have systematically varied all the factors discussed in the literature on adult English, but Armon-Lotem et al. (2005) carried out an elicited production study with English-speaking children (3;2–6;3). These children used ’s with count nouns but not with mass nouns; and they used ’s with animate PRs 90% of the time to encode a part-whole relation (the cowboy’s arm), but less than 50% of the time for inanimate PRs (the tractor’s wheel). With respect to the acquisition of the target word order, some studies explicitly mention that possessive pronouns are correctly positioned (see e.g. Ruff 2000 for German) and we are not aware of any reports that children who acquire the languages under study ever incorrectly position the possessive pronoun after the PR noun. Similarly, studies of German -s possessives, English ’s-possessives, and Japanese no-possessives report that children show the PR-initial target word order from the beginning (Brown
Learning to encode possession 173
1973; Clahsen, Eisenbeiß and Vainikka 1994; Clancy 1985; Eisenbeiß 2000; Radford 1990; Radford and Galasso 1998; Ruff 2000). Both observations also hold for our German data. As we mentioned above, the PR typically follows the PM in German prepositional possessive constructions; and prepositional constructions where the PR precedes the PM are highly marked. However, of the 13 prepositional possessive constructions that we have found in our German data, 3 exhibit the marked PR-initial order (15a–c) and one has a fronted prepositional phrase with a wh-element (15d): (15) a. aber das is nich von wurst die pelle. but that is not of sausage the skin. ‘But that is not the skin of the saussage’
(Carsten / 3;6)
b. ob deiner auch von diese malers das anspitzt. whether yours also of these crayons that sharpen ‘(Let’s see) whether yours also sharpens that (tip) of these crayons?’ (Carsten / 3;6) c. das is vo von de von de Sascha fahrrad. (Svenja/13 / 3;2) that is of of the of the Sascha bike. ‘That is Sascha’s bike.’ d. von wem hast du die schuhe da an? (Svenja/08 /3;0) of whom have you the shoes there on? ‘Whose shoes have you got on, there?’ Note that the examples where the prepositional PR phrase precedes the PM come from different children and corpora and similar examples can easily be found even in a cursory search of the German child corpora in the CHILDES database (MacWhinney 2000); see e.g. the following examples from the corpus provided by Wagner (1985): (16)
aber das sind von Pfe[#] von der Reitschule Pferde. but that are of hor[#] of the riding:school horses. ‘But that are the horses of the riding school.’ (Roman /9;2)
For Greek, which exhibits two different orders for PM and PR for non-pronominal adnominal possessive constructions, we cannot draw any firm conclusions about the early stages, due to a scarcity of early data (see Eisenbeiß 2003 for a discussion of Marinis 2002, 2003). For Hebrew, where the PR follows the PM in constructions with the marker shel, Berman (1985) and Armon-Lotem (1998) argue that children produce possessive constructions
174 Sonja Eisenbeiß, Ayumi Matsuo and Ingrid Sonnenstuhl with the non-target-like order PR < PM before they start to use the correct order and then finally the posssssive marker shel. Taken together, these observations tentatively suggest a tendency to place the PM before the PR, independently of the target language and construction. However, further research is required in this domain as the findings for Hebrew are based on a rather small set of examples and thus have to be taken with caution. Moreover, more systematic corpus searches for prepositional constructions in German child language are required.
4.2.4. Acquisition orders and conceptual complexity Our discussion so far suggests that children acquire the adnominal possessive constructions of their target language incrementally, but show early sensitivity to the constraints on the use of these constructions in their target language. Now, we will have a closer look at the order in which children start to encode different types of possessive relationships. Table 2 gives an overview of the the types of PRs in adnominal possessive constructions (including Carsten’s three dative possessive constructions, which were not included in Table 1). As can be seen in Table 2, all PR (pro)nouns in stages I and II refer to the children themselves or to other people. In stage III, we find a nonhuman, but animate PR as in (10a), but inanimate PRs only appear once in Svenja’s data and four times in Carsten’s data from stage IV as in (17) and (18). Note, however, that in example (18), where the referent of the possessive pronoun is a vehicle, this PR is construed as having “animate” properties. (17)
die wand von’n (corr: ’m) fenster seh ich. the wall of’the-ACC (corr: DAT) window see I. ‘I see the wall of the window.’ (Svenja/15 /3;2)
(18)
dann schleudert er doch nich mehr sein räder then flings he PART not anymore his wheels ‘Then he doesn’t fling his wheels anymore.’ (Carsten / 3;6)
The observed incremental extension in the range of PR types is consistent with the assumption that children start out with more prototypical adnominal possessive constructions that involve a human PR and move on to less prototypical ones with inanimate PR. In order to further evaluate this claim,
Learning to encode possession 175 Table 2.
The animacy of PRs in adnominal possessive constructions
Stage
Child
I
Ann Han Leo Mat Total
II
Ann Han Leo Mat Total
III
And Ann Han Leo Mat Total
55 11 1 20 5 92 (70%)
IV
Ann Car Han Leo Mat Sve Total
Total
Self 7 – – 3 10 (33%)
Human
Other Animate
Inanimate
Total
– – – – 0 (0%)
– – – – 0 (0%)
7 – 17 6 30
– – – – 0 (0%)
– – – – 0 (0%)
4 0 8 0 12
12 7 – 18 2 39 (30%)
– – – 1 – 1 (<1%)
– – – – – 0 (0%)
67 18 1 39 7 132
20 77 5 24 24 47 197 (63%)
5 32 1 9 13 50 110 (35%)
– – – – – – 0 (0%)
– 4 – – – 1 5 (2%)
25 113 6 33 37 1 312
305 (63%)
175 (36%)
1 (<1%)
5 (1%)
486
4 – 2 – 6 (100%)
– – 17 3 20 (67%) – – 6 – 6 (100%)
we will look at the types of (possessive) relations involved in adnominal possessive constructions. Table 3 shows the types of relationships that adnominal possessive constructions encode in our German child data. Note that we have used a wide sense of “ownership” here that covers temporary possessions as well as legal and habitual possession relations. We will try to distinguish between these later. As can be seen in Table 3, children’s adnominal possessive constructions in stages I and II only involve ownership (as in (6a)) or kinship relations as in (19a)). Constructions encoding relationships between a body part and its owner emerge in stage III, as in (19b), and relationships between objects and their parts only appear in the stage IV data of Svenja and Carsten (see e.g. (15a,b) above). None of the children ever uses possessive pro
176 Sonja Eisenbeiß, Ayumi Matsuo and Ingrid Sonnenstuhl Table 3.
Types of possessive relations in adnominal possessive constructions
Stage
Child
I
Ann Han Leo Mat Total
6 – 17 6 29
1 – – – 1
– – – – 0 (0%)
– – – – 0 (0%)
7 – 17 6 30
II
Ann Han Leo Mat Total
3 – 8 – 11 (97%)
1 0 0 0 1 (3%)
– – – – 0 (0%)
– – – – 0 (0%)
4 0 8 0 12
III
And Ann Han Leo Mat Total
50 12 1 36 7 106 (80%)
12 1 – 1 – 14 (11%)
5 5 – 2 – 12 (9%)
– – – – – 0 (0%)
67 18 1 39 7 132
IV
Ann Car Han Leo Mat Sve Total
18 91 6 27 34 71 247 (79%)
5 7 – 3 – 10 25 (8%)
2 11 – 3 3 16 35 (11%)
– 4 – – – 1 5 (2%)
25 113 6 33 37 1 312
393 (81%)
41(8%)
47 (10%)
5 (1%)
486
Total
Ownership
Kinship
Body Part
Part of Object
Total
nouns, -s possessives or prepositional constructions for non-prototypical possessive relations (the dog’s shadow, the size of the block, the state of the car, etc.) Thus, the earliest adnominal possessive constructions only involve prototypical possessive relations. (19) a. mein mama my mommy ‘my mommy’ b. mein nase läuft noch mehr my nose runs even more ‘My nose is running even more.’
(Annelie/02/2;5)
(Annelie/ 04 / 2;7)
Learning to encode possession 177
So far, we have not distinguished between legal ownership and temporary possession, which is simply based on current physical control and proximity. This is difficult to do on the basis of spontaneous speech samples because, in everyday situations, the person who has physical control over something or is close to it, is often also the owner. Thus, we did not attempt to distinguish between legal ownership and physical control for all adnominal possessive constructions. Instead, we searched for utterances in which two possessive relations were encoded in the same utterance – one by a possessive predicate and one within an adnominal possessive construction. We found 10 interpretable utterances of this type in the data of Andreas, Carsten, Leonie, Mathias, and Svenja. In these examples, we can observe a clear distinction between the legal or habitual ownership relation, which is encoded noun-phrase internally, and a temporary ownership or physical control relation, which is encoded at the sentential level. For instance, Andreas has a water ball. In (20), he states that another person, namely Anette has temporarily taken over control of the ball. (20) da Annette hat mei(nen) Wasserball da. there Anette has my water.ball there. ‘Anette has got my water ball over there’
(Andreas/2;1)
Note that 2 of the 10 utterances that encode two different possessive relations come from Andreas (stage III) and all the others occur in stage IV. One might attribute this to the fact that the combination of two possessive relations in the same utterance requires a certain sentence length. However, some of these sentences are actually only 4 or 5 words long; and in each of stages I and II, we find more than a hundred utterances that are longer than 3 words (see e.g. (6a)). Thus, the appearance of combinations such as (20) in stage III might not be an artifact of increasing sentence length. It is at least in line with the general observation that children in stages III and IV extend the range of PRs and possessive relations from ownership relations with human PRs via body part relations to part-whole relations for inanimate objects.
4.3. Summary Our analysis of the German child data and our literature review lead to the following generalizations:
178 Sonja Eisenbeiß, Ayumi Matsuo and Ingrid Sonnenstuhl – At least some children do not seem to produce adnominal possessive
–
– –
–
–
–
–
–
constructions in the early two-word stage, but only precursors, such as single-word utterances that consist of the PR’s name or a possessive pronoun. German adnominal possessive constructions emerge incrementally, in the order proper name possessives possessive pronouns < prepositional constructions < genitive constructions. As children get older, the proportion of pronominal PRs increases. An early stage where possessive markers are omitted could be observed in our own data and in studies of the corresponding possessive markers in English, Greek, Hebrew and Japanese. Studies of German, English, Japanese, and Hebrew child language have demonstrated that adnominal possessive constructions with target-like markers can temporarily co-occur with constructions that lack such markers. In our German data, we observed initial lexical restrictions of possessive pronouns and possessive markers. To our knowledge, detailed analyses of such lexical restrictions have not been carried out for other languages, but the available empirical results we discussed for English possessive ’s are compatible with the assumption of initial lexical restrictions. Children show early sensitivity to the constraints on the use of adnominal possessive constructions in their target language: German children overgeneralise the -s marker to common nouns that do not take this marker in the target language. However, children do not violate the syntactic constraint that prohibits the use of -s for modified PRs – not even for those PRs that they themselves use with modifiers in other contexts. An experimental study of English child language also showed early sensitivity to some of the factors that determine the choice of adnominal possessive construction, in particular animacy and the count/mass distinction. Children also seem to adapt to the target word order early on, though we found a few fronted prepositional PRs in our own data and there are some reports about non-target-like PR-initial constructions in child Hebrew. The German children we observed showed an incremental extension in the range of PRs and possessive relations that were encoded in adnominal possessive constructions – from ownership relations with human PRs via body part relations to part-whole relations for inanimate objects.
Taken together, these empirical generalisations suggest that children show early sensitivity to the constraints of their target language. Thus, the analysis of adnominal possessive constructions provides additional evidence for the
Learning to encode possession 179
emerging consensus about early adaptations to the target language. However, our observations tentatively suggest a tendency to place the PM before the PR, independent of the target language and construction – though further research is required to establish whether this is a reliable finding. With respect to the time-course of development, we found evidence for an incremental extension both in the range of constructions and in the range of possessive relations that are encoded by these constructions. This can be captured straightforwardly in lexical learning and usage-based approaches as they explicitly predict that possessive constructions are acquired incrementally. By contrast, full-competence approaches would need to make additional assumptions to account for these observations. In addition, fullcompetence approaches do not directly predict the initial lexical restrictions that we observed for possessive markers or function words such as possessive pronouns. Such initial lexical restrictions are, however, to be expected if one adopts a Lexical learning or usage-based approach as these approaches assume that children’s early generalisations are limited and linked to particular lexical elements. 5. Predicative possession In contrast to the acquisition of adnominal possessive constructions, the acquisition of predicative possession constructions has not been studied extensively. We are only aware of a few studies which have investigated the acquisition of possessive constructions with have and belong to evaluate approaches to syntax-semantic mappings. We will discuss these studies briefly below, though they are inconclusive as the data base is quite limited and there is considerable controversy about the availability of constraints on the mapping of PR and PM to argument structure positions (see e.g. Bowerman 1990; Pinker 1984). Therefore, we will focus on empirical generalisations about predicative possessive constructions that were established in cross-linguistic studies and can inform acquisition studies of predicative possessive constructions. In what follows, we first overview these generalizations and their application to German before we discuss our German child language data and relevant findings for English child language. 5.1. Predicative possession in adult language The background for all discussions of predicative possessive constructions is the realization that the possessive relation they encode is a binary and
180 Sonja Eisenbeiß, Ayumi Matsuo and Ingrid Sonnenstuhl “symbiotic” relationship between two entities that only receive their semantic interpretation by virtue of one another: a possessive relation is “a relation between two entities, a PR and a PM, such that one, the PR, is seen as being in some way related to the other, the PM, as having it near or controlling it” (Baron, Herslund, and Sørensen 2001: 2). However, there is no PR without a PM and a PM is not a PM without a PR. Possessive relations share this property with locative constructions and experiencer constructions: something is only conceived of as a Location if something is located there and this element is only considered to be a Locatum if it has a Location. Similarly, an Experiencer requires a Stimulus to experience something, just as a stimulus cannot be construed as a Stimulus unless it causes a sensation in an Experiencer. By contrast, Agents do not necessarily need Patients for their activities – we can talk about people who kick, dance, and write without ever specifying who they kick or what they dance or write. Given this inherent co-dependency between the entities involved in a possessive, locative or experience relationship, it is difficult to rank these entities according to their agentivity or their control over the relationship – i.e. with respect to the variables that (co-)determine which of them is linguistically realized as the topic or grammatical subject of a construction. Hence, it is unsurprising that we can observe a large amount of inter- and intra-language variability in the syntactic realization of Experiencers and Stimuli, Locata and Locations, and PRs and PMs (see e.g. Chappel and McGregor 1996; Heine 1997; Seiler 1983). An Experiencer can be realized as topic or subject (Most farmers fear angry roosters) or it can be encoded as an object (Angry roosters frighten most farmers) – and the same is true for the Locatum (The pencils are (lying) in the box vs. The box contains some pencils). Similarly, cross-linguistic studies of predicative possession constructions frequently make a distinction between HAVE-constructions, where the PR is realized as topic and subject (I have/own a car), and BELONGconstructions, where the PM appears in this role (The car belongs to me. The car is mine/ Peter’s; see e.g. Baron, Herslund, and Sørensen 2001 for overview). Some authors have claimed that this distinction can be observed in all human languages (Heine 1997: 33), though this is controversial. HAVE and BELONG constructions do not just differ with respect to the mapping of arguments onto grammatical roles; they can also differ with respect to the definiteness of their arguments. In particular, the PM is typically defninite in German BELONG constructions, but indefinite in German HAVE constructions (Heine 1997: 30f., see e.g. Das Huhn gehört mir ‘The chicken belongs to me’ vs. Ich habe ein Huhn). Note, however, that it is
Learning to encode possession 181
possible to combine a definite PM with haben ‘have’, especially when a contrastive reading is intended (Ich habe den linken Schuh, aber nicht den rechten ‘I have the left shoe, but not the right one’). By contrast, the restriction against combining gehören ‘belong’ with an indefinite PM seems quite strict. Whether children are sensitive to the definiteness contrast between HAVE and BELONG constructions has – to our knowledge – not been investigated so far and we try to fill this gap. Both HAVE- and BELONG-constructions are not restricted to encoding ownership; they can also be used to express locative relations (Boston has many great buildings. This district has always belonged to London, etc.). However, HAVE constructions can express a wider set of meanings (kinship, abstract possession as in She has a cold, etc.), while BELONG constructions tend to encode more concrete relations. Therefore, many linguists consider HAVE constructions the unmarked option (see Baron, Herslund, and Sørensen 2001 for discussion). However, in his discussion of learnability issues in first language acquisition, Pinker (1984) argues that the unmarked or “canonical” mapping is the one where the Locatum Argument becomes the subject. This would make HAVE-constructions the marked alternative. Bowerman (1990) argues that this would lead to the prediction that children should initially make “default marking errors” and produce HAVE-constructions with PM subjects and PR objects. However, such errors have not been observed yet – and in later work, Pinker no longer makes the assumption of canonical mappings (Pinker 1989). In the following, we will investigate whether we find any evidence for non-target-like mappings of PR and Possessum onto grammatical functions – and whether they follow the HAVEor the BELONG-pattern. Moreover, we will investigate the relationship between the encoding of locative and possessive relations. According to Heine, possession exhibits various structural similarities with domains such as location, experience and existence and predicative possessive constructions draw on different preexisting constructions or “source schemata” (see Baron et al. 2001 for discussion). The most commonly cited ones are the action schema, which underlies HAVE-constructions, and the location schema, which is the basis for BELONG-constructions. These links raise the question of whether children show any evidence for the primacy of location for verbs that are taken to be derived from the locative source schema – i.e., whether they use verbs such as belong or be first with a locative or existential reading (Das gehört hier hin ‘This belongs here’) and only later with a possessive reading.
182 Sonja Eisenbeiß, Ayumi Matsuo and Ingrid Sonnenstuhl Note, however, that some linguists argue that possession is independent of location – hence they would not assume that locative constructions form the source of possessive constructions (see e.g. Payne this volume).
5.2. Predicative possession in child language In the following, we will look at HAVE- and BELONG-constructions with the words haben ‘have’, gehören ‘belong’ and sein ‘be’ in our German child language data. Other possession verbs – such as besitzen ‘own’ – do not occur in our data. We will investigate (i) when the different types of predicative possessive constructions appear in our data, (ii) whether children are sensitive to the fact that HAVE- and BELONG-constructions impose different constraints on the definiteness of the PM, (iii) whether we can find any mapping errors in children’s predicative possessive constructions, and (iv) whether children show any evidence for the primacy of location for verbs that are taken to be derived from the locative source schema. In Table 4, we can see how often children used haben ‘have’ and gehören ‘belong’ with a definite PM or an indefinite PM. Moreover, as discussed above, we found 10 instances of haben with a possessive noun phrase (see (20) above) and we listed these utterances in a separate column. As Table 4 demonstrates, predicative constructions with haben ‘have’ appear before constructions with gehören ‘belong’. However, we can see differences between the individual children: Svenja uses both verbs, whereas Hannah, Carsten and Andreas use haben, but not gehören. Annelie used haben from stage I, but gehören ‘belong’ only appears from stage II. Leonie and Mathias do not exhibit any predicative possessive constructions in stages I and II. In stage III, haben occurs quite frequently and we find one instance of locative gehören in Leonie’s data. In stage IV, both children use gehören. Table 4 also shows that children are sensitive to the definiteness constraint for gehören and do not use any indefinite noun phrases as PM for this verb. Instead, they produce definite articles – either used as pronouns or in combination with nouns as in (21a, b, c) and demonstrative pronouns as in (21d–g). In addition, we found one unclassifiable utterance with a self-interruption that we excluded from the analysis. By contrast, haben predominantly occurred with indefinite noun phrases (342 / 480 =71%, see e.g. (22a), though we also found PM noun phrases with definite articles (22b) and the 10 noun phrases with possessive pronouns that we discussed above – see (20).
Learning to encode possession 183 Table 4. Types of PM Noun Phrases co-occurring with haben ‘have’ and gehören ‘belong’
Indef
I
II
Ann Han Leo Mat Total Ann Han Leo Mat Total
gehören possessive (‘this belongs to me’)
haben
Child
a
Def
1 4 – – 5
– 1 – – 1
(83%)
(17%)
7 2 – – 9
Poss Total
gehören locative (‘this belongs here’)
Indef
Def
Total
Indef
Def
– – – – 0
1 5 – – 6
– – – – 0
– – – – 0
– – – – 0
– – – – 0
– – – – 0
– – – – 0
– – – – 0
– – – – 0
7 2 – – 9
– – – – 0
– – – – 0
– – – – 0
– – – – 0
4 – – – 4
4 – – – 4
18 2 – 11 – 31
2 – – – – 2
74 35 – 23 7 139
– – – – – 0
– – – – – 0
– – – – – 0
– – – – – 0
(22%)
(1%)
11 9 4 11 12 48 95
– 2 – 2 2 2 8
(100%)
III
And Ann Han Leo Mat Total
54 33 – 12 7 106 (76%)
IV
Total a
Ann Car Han Leo Mat Sve Total
12 45 9 47 28 81 222
Total
(100%)
(68%)
(29%)
(6%)
342
127
10
(71%)
(26%)
(2%)
– – – 1 – 1
– – – 1 – 1
(100%)
23 56 13 60 42 131 325
– – – – – – 0
– – – 1 1 14 16
480
0
16
– – – 1 1 14 16
– – – – – – 0
1 – – 6 – 7 14
16
0
19
(100%) (100%)
1 – – 6 – 7 14
(100%)
19
(100%)
Def: Definite NP, Indef: Indefinite NP, Poss: adnominal possessive construction
Note that all of the utterances with haben are target-like with respect to definiteness and with respect to the mapping of arguments onto grammatical roles. By contrast, we find mapping problems with gehören. Svenja never uses target-like dative marking for the PR. In example (21b), she
184 Sonja Eisenbeiß, Ayumi Matsuo and Ingrid Sonnenstuhl produces a prepositional phrase with von ‘of’ instead of a dative PR noun phrase. In (21c–g), she produces nominative forms. Similar non-target-like mappings have been observed in the example by Simone that we discussed in the introduction. Simone’s resistance to corrections for these deviations from the target suggests that such “errors” are not simply slips of the tongue. Rather, at least some children seem to initially overgeneralise the form-meaning mapping of the HAVE-construction to the BELONGconstruction. This is in line with the assumption that HAVE is the default, but it would surprising if the HAVE-pattern were marked (Pinker 1984). (21) a. der (ge)hört mir. this-one belongs me-DAT ‘This one belongs to me.’
(Mathias / 26 / 3;5)
b. der gehört von den (corr: dem) (\) this-one belongs of this-one-ACC (corr: DAT) (Svenja/13 / 3;2) ‘This one belongs to this one’ c. wer (ge)horn die füße? who-NOM belong the-NOM/ACC feet? ‘To whom do the feet belong?’
(Svenja/16 / 3;3)
d. denn kuck mal hier wer das d(=g)ehört. then look PART here who-NOM that-NOM/ACC belongs ‘Then just look here, to whom that belongs.’ (Svenja/13 / 3;2) e. der (corr: dem) (ge)hört das ne? this-one-NOM (corr: DAT ) belongs that-NOM/ACC TAG? ‘That belongs to this one, doesn’t it?’ (Svenja/16 / 3;3) f. der (corr: dem) (ge)hört das. (Svenja/16 / 3;3) this-one-NOM (corr:DAT) belongs that-NOM/ACC ‘That belongs to this one.’ g. dann (g)ehört das (Svenja/ 04 / 2;9) then belongs that-NOM/ACC niemand [#] keiner. nobody-NOM (corr:DAT) [#] no-one-NOM (corr:DAT). ‘Then that belongs to nobody.’ (22) a. jetz ham (= haben) wir eine. now have we one. ‘Now, we have one.’
(Andreas/2;1)
Learning to encode possession 185
b. die schere hat Julia. the scissors has Julia ‘Julia has the scissors.’
(Mathias/17/2;9)
For gehören, we observe more than just mapping problems. The data in Table 4 also suggest that locative BELONG is a pre-cursor for possessive BELONG: of the 3 children who use gehören more than once, only Svenja, the most advanced child, uses both type of constructions. Recall, however, that she shows mapping problems. She also uses the proximity preposition bei ‘by’ in locative BELONG-constructions and not the target preposition zu ‘to’, which is directional, as shown by (23a,b). These utterances are somewhere between a locative and a possessive construction as they seem to focus on placing an object in the proximity of an animate PR. In Annelie’s data we only find 5 cases of the locative variant (24); and Leonie uses the locative variant once in stage III and 6 times in stage IV (25a), but only produces one instance of possessive gehören (25b). (23) a. das schiff gehört bei dir. the ship belongs at you. ‘The ship belongs to you.’
(Svenja/05 /2;10)
b. die (ge)hört [#] ei [#] die (ge)hört bei (?) dir. this-one belongs [#] ??? [#] this-one belongs at (?) you. ‘She belongs to you.’ (Svenja/04 /2;9) (24) so (ge)hört das hin. so belongs that hither. ‘That belongs here, like that.’
(Annelie/03/2;6)
(25) a. das (ge)hört da drauf. that belongs there onto-this. ‘That belongs up there.’
(Leonie/15 / 2;11)
b. Sassa (ge)hört das. Sascha belongs that. ‘That belongs to Sascha.’
(Leonie/ 09 / 2;6)
In order to further explore the relationship between possession and other semantic domains, we have compared different uses of sein ‘be’: utterances that focus on location (and existence), as in (26); predication, as in (27); combinations of location and possession, as in (28); and utterances with an adnominal possessive as a subject and a further predication, i.e. utterances
186 Sonja Eisenbeiß, Ayumi Matsuo and Ingrid Sonnenstuhl where something is predicated about something that is said to belong to someone, as in (29). We have also looked at sein in utterances that were used to establish possession, as in (30). In these utterances a PR was either used on its own or in a noun phrase with a PM. We have not distinguished between these two types any further as they fulfill a similar function and appeared with very similar frequencies in each stage. Table 5 provides an overview. (26) da is die mama there is the mommy ‘There is mommy’
(Annelie/02 /2;5)
(27) is en löwe. (Hannah/01/2;0) is a lion. ‘(This) is a lion.’ (28) da is Klaras. there is Klara’s. ‘There is Klara’s.’
(Leonie/ 09 / 2;6)
(29) daniels kopf is härter. Daniel’s head is harder. ‘Daniel’s head is harder.’
(Mathias/26 /3;5)
(30) das is doch Julias schiff. that is PART Julia’s ship. ‘But that is Julia’s ship.’
(Mathias/22 /3;1)
As Table 5 shows, sein is predominantly used for predictation and location, though we also find one possessive use and a few utterances where the child talks about the location of a possession. In stage III, possessive uses become more frequent – except for Hannah. Thus, we can see a shift towards more sein utterances involving possessive constructions. However, in contrast to gehören ‘belong’, utterances with sein involve expressions of possession from early on. But only in stage IV do we find more than one utterance where the subject involves an adnominal possessive construction and a further predication is added. This is in line with the observation that combinations of haben ‘have’ with adnominal possessive constructions only appear in stages III (Andreas) and IV (Carsten, Leonie, Mathias, and Svenja). The only child to combine both haben and sein with adnominal possessive constructions in stage III is Andreas, who is quite advanced for this stage and already on the verge to stage IV (see Eisenbeiß 2000, 2003).
Learning to encode possession 187 Table 5. Child
Types of sein ‘be’ Location
Predication
Location + Possession
Possession
Predication + Possession
Total
I
Ann Han Leo Mat Total
24 – – 10 34 (61%)
7 1 2 8 18 (32%)
1 – 1 2 4 (7%)
– – – – 0 (0%)
– – – – 0 (0%)
32 1 3 20 56
II
Ann Han Leo Mat Total
29 2 1 12 44 (68%)
15 – 1 3 19 (29%)
1 – – – 1 (2%)
– – 1 – 1 (2%)
– – – – 0 (0%)
45 2 3 15 65
III
And Ann Han Leo Mat Total
82 31 1 5 6 125 (56%)
34 22 1 3 16 76 (34%)
10 1 – 2 – 13 (6%)
3 4 – 2 1 10 (4%)
1 – – – – 1 (<1%)
130 58 2 12 23 225
IV
Ann Car Han Leo Mat Sve Total
42 64 20 83 34 275 518 (42%)
43 135 29 83 53 227 570 (47%)
6 21 1 6 9 7 50 (4%)
16 15 4 4 8 34 81 (7%)
– 1 – 3 1 1 6 (<1%%)
107 236 54 179 105 544 1225
721 (46%)
683 (43%)
68 (4%)
92 (6%)
7 (<1%)
1571
Total
5.3. Summary Our analysis of predicative possessive constructions in the German child data has led to the following generalizations: – HAVE-constructions appear before BELONG-constructions. – As soon as children start to use HAVE- and BELONG-constructions,
they seem to be sensitive to the different definiteness constraints for these constructions: they only use gehören ‘belong’ with a definite PM, but combine haben ‘have’ with both definite and indefinite phrases.
188 Sonja Eisenbeiß, Ayumi Matsuo and Ingrid Sonnenstuhl – While HAVE-constructions show target-like mappings of arguments
onto grammatical roles, systematic overgeneralizations of the HAVEpattern to the BELONG-verb gehören are observed. – When gehören emerges, it is initially restricted to its locative use (This belongs here). Moreover, sein ‘be’ is initially mostly used for predication and location, though the possessive use of this verb precedes the use of gehören. – Combinations of haben and and sein with adnominal possessive constructions (X has Y’s Z, Y’s Z is X) only appear in data from late stage III and stage IV. Similar to the empirical generalisations about adnominal possessive constructions, these generalisations suggest that children become sensitive to the constraints of their target language very early on. Thus, they are also in line with the emerging consensus about early adaptation to the target language. Moreover, just as in the adnominal domain, we observe an incremental extension of forms and functions. In the case of predicative possessive constructions, we can observe a primacy of locative relations, which is in line with some of the typological literature on possession (see e.g. Baron, Herslund, and Sørensen 2001). Such an incremental developmental path can be captured straightforwardly in lexical learning and usage-based approaches as they explicitly predict that possessive constructions are acquired incrementally. By contrast, full-competence approaches would need to make additional assumptions to account for these observations. Our findings about the acquisition of predicative possessive constructions cannot only contribute to acquisition research; they can also help us evaluate claims made in theoretical linguistics. In particular, the early and error-free acquisition of HAVE constructions and the observed overgeneralisaton of the HAVE-pattern to the BELONG-verb gehören support the assumption that the HAVE-construction is the unmarked construction. This is in line with a common view in the typological literature on possession (see Baron, Herslund, and Sørensen 2001 for discussion), but it is incompatible with Pinker’s earlier analysis. Recall that the observation that combinations of haben ‘have’ and sein ‘be’ with adnominal possessive constructions (X has Y’s Z, Y’s Z is X) appear comparatively cannot be simply attributed to the required sentence length. Note that some of the utterances of this type actually only consist of 4 or 5 words, which is an utterance length that can be observed quite frequently in stages I and II. As we argued above, the combination of different types of possessive relations (temporary physical control vs. long-term
Learning to encode possession 189
ownership) might be the reason for the late acquisition of combinations of adnominal possessive constructions with haben. The observation that adnominal possessive constructions also appear late in combination with sein, could suggest a more general explanation: both the sein and the haben constructions that appear late involve adnominal possessive constructions that encode a presupposition about possession, which is then combined with another predication. We will further explore this idea below, where we discuss constructions that encode both possessive relations and Agent-Patient relations.
6.
External possession
In this section we focus on constructions that involve both possessive relations and Agent-Patient relations. This can be achieved by realizing the possessive relation within a noun phrase and encoding the Agent-Patient relation at the sentential level. For instance, in Jack tapped Frank’s shoulder, there is a possessive relationship between Frank and his shoulder, which is captured within the adnominal possessive construction. In addition, there is an Agent-Patient relation between Jack on the one side and Frank and his body part on the other. This relation is mapped onto the subject-object relation. Alternatively, the PR can be realized as an independent verb argument in a so-called EPC (Neumann 1995; Payne and Barshi 1999; Shibatani 1994). For example, in the sentence I tapped Frank on the shoulder; the PR Frank and its PM shoulder are realized as independent phrases. In what follows, we focus on the realisation of possessive relations and Agent-Patient relations in EPCs or adnominal possessive constructions in German and Japanese child language data. 6.1. External possession in the adult language EPCs are very common in German and occur in everyday language addressed to young children (see Neumann 1995; Appendix A, A23). Here, the PR is typically realized as a dative-marked “extra” argument that is not subcategorized by the verb. This extra dative argument refers to the PR of the entity that is the Patient of the Action, while this PM is encoded as the direct accusative object (die Haare ‘the hair’). This possessive relation makes it possible that the dative noun phrase is integrated into the construction as an argument with all the morpho-syntactic characteristics of a sub-
190 Sonja Eisenbeiß, Ayumi Matsuo and Ingrid Sonnenstuhl categorised indirect object. The dative PR can be combined with a large variety of intransitive and transitive verbs, but not with verbs that already have a dative argument; see e.g.: (31) a. Das Zebra beißt dem Hasen in den Schwanz the zebra bites the-DAT hare- into the tail ‘The zebra bites the hare’s tail’ b. Der Mann waescht dem Baeren die Pfote the-ACC paw the man washes the-DAT bear ‘The man washes the bear’s paw’ c. Der Junge legt dem Affen den Hut auf den Ruecken the boy puts the-DAT monkey the-ACC hat on the back ‘The boy puts the hat on the monkey’s back’ d. *Der Junge gibt dem Vater dem Kind das Buch the boy gives the-DAT father the-DAT child the-ACC book’ ‘The boy gives the child’s father the book’ The dative PRs in these constructions are not adjuncts with a semantic case that is solely determined by their thematic role and unaffected by any syntactic processes. Rather, we observe the same syntactically determined case alternations for dative PRs (as in (32)) as for indirect dative objects of three-place verbs (as in (33)). In so-called recipient passive sentences with the auxiliary kriegen/bekommen ‘get/become’ (e.g. (32b, 33b)), the dative noun phrase of the corresponding active sentence (e.g. (32a, 33a)) carries a nominative marker: (32) a. Der Mann wäscht dem Hund den Kopf. the-NOM man washes the-DAT dog the-ACC head ‘The man washes the dog’s head.’ b. Der Hund kriegt den Kopf gewaschen the-NOM dog gets the-ACC head washed ‘The dog gets its head washed.’ (33) a. Der Mann gibt dem Baeren den Honig. the-ACC honey the-NOM man gives the-DAT bear ‘The man gives the bear the honey.’ b. Der Baer kriegt den Honig gegeben the-NOM bear gets the-ACC honey given ‘The bear is given the honey.’
Learning to encode possession 191
Thus dative PRs behave differently from dative adjuncts with semantic case, e.g. the so-called ethical datives in (34), which express the attitude the referent of this noun phrase has to the action encoded in the verb. In contrast to dative PRs, ethical dative adjuncts do not exhibit any case alternation in recipient passives. That is, the ethical dative pronouns keep their dative marking, as in (34b, 35b) while external PRs appear with nominative marking in recipient passives, as in (35b) – just as indirect dative objects, as in (34b). Moreover, ethical datives can be combined with both indirect objects, as in (34), and dative PRs, as in (35), which suggests that they occupy a different syntactic position, while indirect objects and dative PRs have the same status. (34) a. Dass du mir der-DAT Oma that you-NOM me-DAT the-DAT granny nicht wieder die falschen Tabletten gibst! not again the-ACC wrong-ACC tablets give! ‘I hope that you won’t give granny the wrong tablets again!’ b. Dass mir die Oma nicht wieder that me-DAT the-NOM granny not again die falschen Tabletten gegeben kriegt! the-ACC wrong-ACC tablets given gets! ‘I hope that granny won’t be given the wrong tablets again!’ (35) a. Dass du mir nicht wieder that you-NOM me-DAT not again jemandem auf die Füe trittst! someone-DAT on the feet step! ‘I hope that you won’t step on someone’s feet again!’ b. Dass mir nicht wieder jemand that me-DAT not again someone-NOM auf die Füe getreten kriegt! on the feet step gets! ‘I hope that nobody gets stepped on their feet again!’ While German exhibits external PR constructions with dative PRs as in (36a), it is also possible to realize the possessive relation within an adnominal possessive construction and the Agent-Patient relation at the sentential level as in (36b):
192 Sonja Eisenbeiß, Ayumi Matsuo and Ingrid Sonnenstuhl (36) a. Sue legt dem Pferd den Sattel auf den Ruecken. Sue lays the-DAT horse the-ACC saddle on the back. ‘Sue puts the saddle on the horse’s back.’ b. Sue legt den Sattel auf den Ruecken von dem Pferd. of the horse Sue lays the-ACC saddle on the back ‘Sue puts the saddle on the horse’s back.’ Which of these constructions is selected is determined by semantic factors (see e.g. Neumann 1995; Payne and Barshi 1999). For instance, the PR in EPCs must be perceived as beneficially/adversely affected by an action by virtue of being the PR of the entity which is directly affected by this action. Moreover, the stronger the effect on the PR and the more intimate the possession relation is, the more likely it is that the PR will be encoded as an extra argument. Thus the inalienable possession relation between a person and a body part is more likely to be encoded in an EPC than the relation between a person and an alienable possession (e.g. a car). While German shows parallels between indirect dative objects and external dative PRs, the presence of indirect dative objects and other dative arguments (e.g. Experiencers) in a language does not imply that this language will also exhibit dative-marked external PRs. For instance, Japanese has indirect dative experiencers and a range of dative constructions that are very similar to German. It also has “double subject” and “double object” EPCs (Tsunoda 1995). However, Japanese lacks external dative PRs. In the translation equivalents of the German examples (31a–c), both PM and PR have to be encoded within the same noun phrase (e.g. Watashi-wa kuma-no te-o aratta ‘(I) the bear’s the paw washed’). Note that the presence of such an IPC does not rule out external PRs per se as EPCs and IPCs can co-exist in languages such as German. Thus, if Japanese children ever started to use datives for external PRs, it would be difficult to overcome this deviation from the target language without recourse to reliable negative evidence.
6.2. External possession in child language As far as we are aware, there are no systematic studies of EPCs in child language. We will therefore first look at our German corpus data and then present some preliminary resuls from elicitation games with German and Japanese children. In our German data, we searched for all utterances that involved an Agent-Patient relationship where the PR of the Patient was
Learning to encode possession 193
mentioned: (i) as an extra argument, (ii) within the same noun phrase as the Patient or (iii) both. Table 6 provides an overview. External and adnominal possessive constructions
– – – – – – – – – 1 – 1 2
2 – – – – 2 1 4 – 1 – 1 7
1
36
1
38
2
9
Total
2 – – – – 2 2 13 – 2 5 14 36
Both
– – – – – – – – – – – 1 1
IPC
IPC
2 – – – – 2 2 12 – 2 5 13 34
EPC
EPC
– – – – – – – 1 – – – – 1
Total Total
Total
Total
Both
IV
IPC
III
And Ann Han Leo Mat Total Ann Car Han Leo Mat Sve Total
Body-Part
EPC
Ownership
Both
Table 6.
– – – – – – – 1 – – – 1
2 – – – – 2 1 5 – 2 – 2 10
– – – – – – – 1 – 1 – 1 3
4 – – – – 4 3 16 – 3 5 14 41
– – – – – – – 1 – – – 1 2
4 – – – – 4 3 18 – 4 5 16 46
1
12
3
45
2
50
As indicated in Table 6, Hannah does not produce any combinations of possessive relations and Agent-Patient relations. For the other children, such combinations only appear in stage III (Andreas) or IV (for the other children; see the examples below) – i.e. at the same time when we can observe the first combinations of haben ‘have’ and adnominal possessive constructions (see the previous section) and the first combinations of adnominal possessive constructions with a further prediction introduced by sein ‘be’. This is in line with our idea that the combination of possessive relations and other relations is a later development. Again, the fact that combinations of possessive relations with another relation appear compartiavely late cannot simply be attributed to sentence length as four or five words would be sufficient for such constructions. Moreover, EPCs are very rare in our child data. We only found three examples; see e.g. (37). Children tend to use constructions where possessive relations are encoded within a noun-phrase – even when most adults would probably prefer an EPC to encode events of taking something away from someone, grooming and getting dressed, see e.g. (38). This suggests that the
194 Sonja Eisenbeiß, Ayumi Matsuo and Ingrid Sonnenstuhl integration of both possessive and Agent-Patient relations into one single representation for verb arguments poses particular problems. This is supported by the fact that we also found two utterances where the PR appeared both as a pronominal dative argument (mir ‘to-me’) and as a pronominal possessive pronoun (meine ‘my’) – as if the occurrence of the extra dative argument was insufficient to encode both the fact that the PR is affected by the action and the fact that the speaker is the PR of the shoes/hands. (37) jetz hau ich dir aber aber ganz xxx den popo. now hit I you-DAT PART PART really xxx the-ACC botty. ‘Now I will really hit you on the botty.’ (Svenja/15 / 3;2) (38) a. mama hat meine hose weggemommt (=weggenommen) mommy has my trousers away-taken. ‘Mommy has taken away my trousers.’ (Annelie/05 /2;8) b. jetz zieh i(ch) wieder mein(e) schuhe an. now put I again my shoes on. ‘Now I am putting on my shoes again.’
(Svenja/14 / 3;2)
(39) a. mama du solls(t) mir jetz meine hände aber waschen. hands PART wash. mommy you shall me-DAT now my ‘Mommy, you ought to wash my hands now.’ (Carsten / 3;6) b. dann dann zieh ich mir then then put I me-DAT meine meine schusche (=schuhe) an. my my shoes on. ‘Then I will put on my shoes.’
(Svenja/10 / 3;1)
Clearly, further studies of external possession with older children are necessary. Moreover, in order to raise the number of potential contexts for external possession contexts, elicitation stimuli or games might be needed (see e.g. the picture book stimulus by Eisenbeiß and McGregor 1999). We have obtained some preliminary evidence from an elicitation study with Japanese and 20 German and Japanese children (2–6 yrs; Eisenbeiß and Matsuo 2003, 2005). This study made use of the puzzle tasks where children are asked to describe events depicted on a puzzle board to obtain puzzle pieces with matching pictures. The individual pictures on the puzzle board differ minimally from each other, so that children have to express the differences verbally in order to clearly identify the puzzle piece they want. For the external possession study, three different types of events with af-
Learning to encode possession 195
fected body parts were depicted and the pictures differed with respect to the participants and body-parts, see Figure 1.
Figure 1.
Sample elicitation materials for External Possessive Constructions (Eisenbeiß and Matsuo 2003, 2005)
In order to obtain the desired puzzle piece, children had to mention Agents, Patients and those body parts that were affected by the action. The German target sentences with a dative PR involved three different constructions, one with a nominative subject and a prepositional phrase for the body part as in (40), one with a nominative subject and a direct accusative object as in (41), and one with a subject, a direct accusative object and a prepositional phrase for the body part as in (42). (40) Das Zebra beißt dem Hasen in den Schwanz the zebra bites the-DAT hare into the tail ‘The zebra bites the hare’s tail.’ (41) Der Mann waescht dem Hund den Schwanz the man washes the-DAT dog the-ACC tail ‘The man washes the dog’s tail.’
196 Sonja Eisenbeiß, Ayumi Matsuo and Ingrid Sonnenstuhl (42) Der Junge legt dem Pferd den Sattel auf den Ruecken the boy puts the-DAT horse the-ACC saddle on the back ‘The boy puts the saddle on the horse’s back.’ An initial analysis of the data showed early adaptations to the target language: Japanese children consistently realized the possessive relationship noun-phrase-internally, whereas German children used EPCs in more than 80% of the utterances that referred to a body part and its PR. However, many utterances produced in the elicitation game were not target-like. When children produced adnominal possessive constructions instead of external PR constructions (which are not strictly ungrammatical, just dispreferred), they sometimes exhibited the -s overgeneralizations to single PR nouns that we described above, leading to non-target-like utterances such as (43). When they attempted to produce external PR constructions, children often do not manage to realize the PR as a case-marked dative noun phrase, but use a prepositional phrase instead, as in (44). Note that these utterances are not always target-like. If they use the preposition von ‘of’ for the PR, they should use it within an adnominal possessive construction, following the PM. Instead, they sometimes position the von-phrase to the left of the PM, sometimes with other elements in between. This leads to utterances that are not acceptable for the adult speakers we consulted, as in the case of (44a,b). We also found utterances where the child used two prepositional phrases with directional prepositions – one for the body part and one for its PR, as in (44c). Thus, both were mentioned as the endpoint of the caused motion event, but in a parallel fashion. This is also not an option in adult German. Note that the deviations from the target we observed in the elicitation games do not seem to be an artifact of the elicitation technique. Recall that we found -s overgeneralizations and fronted prepositional phrases for PRs in different sets of corpus data. We will argue that the deviations from the target in these utterances result from difficulties in the intergration of possessive and Agent-Patient relations. This is supported by the observation that they were more likely to use adult-like external possessive constructions for two-place verbs such as bite than for three-place verbs such as put. (43) a. Da legt der junge die leine auf katzes hals there puts the boy the-ACC leash on cat’s neck ‘There the boy puts the leash around the cat’s neck.’ b. Da legt der junge den hut auf affens bauch there puts the boy the-ACC hat on monkey’s tummy ‘There the boy puts the hat on the monkey’s tummy.’
Learning to encode possession 197
(44) a. da v [/] legt der junge vom pferd auf’n ruecken there [/] puts the boy of:the horse on:the back den sattel the-ACC saddle ‘There the boy put the saddle on the back of the horse.’ b. von kuh der junge in den (corr: das) bein beissen of cow the boy into the-ACC.MASC (corr:NEUT) leg bite ‘The boy bites the cow’s leg.’ c. da legt der junge auf den pferd auf den kopf there puts the-NOM boy on the horse on the head den sattel. the-ACC saddle. ‘There, the boy puts the saddle on the horse’s back.’
6.3. Summary Clearly, more studies are required for EPCs. However, our analysis of predicative possessive constructions in the German child data has led to some preliminary generalizations and insights: – In spontaneous speech from German children, we did not observe any
EPCs in stages I– III and only a few examples in stage IV. – The German children from our corpus study sometimes produced con-
structions where possessive relations are encoded within a noun-phrase – even when most adults would probably prefer an EPC to encode events of taking something away from someone, grooming and getting dressed. The older children in our elicitation study preferred EPCs to adnominal possessive constructions for events where the body part of an animate PR is affected physically, but also sometimes chose to realize the possessive relation noun-phrase internally where an adult would produce an extra dative PR. – We found a few external possessive constructions where the PR was realized both as a dative argument and as a possessive pronoun within the possessum noun phrase. – External PR constructions contained more deviations from the targetlanguage for three-place verbs like legen ‘put’ than for two-place verbs like beissen ‘bite’.
198 Sonja Eisenbeiß, Ayumi Matsuo and Ingrid Sonnenstuhl – Common error types were the use of prepositional PR phrases instead of
dative PRs; and some of these prepositional phrases were fronted, which is highly marked (or even non-target-like for many native speakers). – By contrast, Japanese children do not overextend the use of the datives they hear in their input to produce EPCs with dative PRs, which they do not encounter in their target language. Taken together, these observations suggest children adapt to the properties of their target language quite early, but the integration of both possessive and Agent-Patient relations into one single representation for verb arguments poses problems. Moreover, we found additional evidence for a tendency to position the PR before the PM – even if this leads to non-targetlike utterances. 7. Implications for theoretical linguistics and models of language acquisition When we compare the results for adnominal, predicative and external possessive constructions, we observe incremental development in all three domains: the constructions of the target language emerge step by step. At the same time, the range of functions that are encoded by possessive constructions increases over time. For adnominal possessive constructions, the range of relations expands from prototypical possessive relations via body part relations to relationships between objects and their parts. In the case of predicative possessive constructions, we can observe a primacy of locative relations in early stages, which is in line with the typological literature on possession. In addition to incremental development, we found initial lexical restrictions for both possessive -s-markers and possessive pronouns in adnominal possessive constructions. As discussed above, both incremental development and initial lexical restrictions are explicitly predicted by structurebuilding approaches as well as usage-based approaches. By contrast, capturing such a developmental path would require additional assumptions in a full-competence approach, for instance, reference to the interaction between linguistic development and the learning of cultural norms for possessive relations and negotiations about them. While children may not use the full range of forms and functions in the early two-word stage, they seem to adapt to the core properties of their target language early on. For instance, children overgeneralise the German -s-
Learning to encode possession 199
marker, which is restricted to unmodified proper names and kinship terms, to count nouns that cannot be used in this way. However, they do not overgeneralise -s to modified nouns. Thus, they respect the syntactic constraint of the adult language and do not adopt the incorrect generalization that -s can be affixed to any names or kinship terms, whether they are modified or not. We have argued that children do not make the non-target-like semantic generalization because they would need positive evidence to assume that -s can also be combined with more complex noun phrases – and they do not find any such evidence in their input. In particular, we suggested that this could be captured in feature-based underspecification models of morphosyntactic development – such as the one proposed by Eisenbeiß (2003; see Eisenbeiß 2007, 2009 for summaries). In such a model, the difference between a simple noun and a complex noun phrase could be captured by additional features for phrasal projections and children would only include such a phrase-feature in their input specification for -s if they find positive evidence for it in their input – as they do in English – but not in a language like German. We also observed some overgeneralisations of the HAVE-pattern of mapping arguments onto grammatical functions to BELONG-verbs. This is surprising if one assumes that the BELONG-pattern is a “canonical” pattern, as the analysis by Pinker (1984) suggests. However, if one assumes that the HAVE-pattern is the default pattern (see Baron, Herslund and Sørensen 2001 for discussion), such overgeneralisations are to be expected. However, further cross-linguistic research is required to determine whether children ever produce the reverse overgeneralisations and whether these overgeneralisations could simply be attributed to the high frequency of the HAVE-pattern. Similarly, we would like to see more studies of the acquisition of word order in possession constructions to determine whether the observed tendency to position the PR before the PM is a more general phenomenon. BELONG-constructions and PM
200 Sonja Eisenbeiß, Ayumi Matsuo and Ingrid Sonnenstuhl Appendix A: A family discussion of possession (A1) Mother: du isst noch dein eis? you eat still your ice-cream? ‘You are still eating your ice-cream?’ (A2) Liam:
issn [?] ich noch mein eis. eat I still my ice-cream ‘I am still eating my ice-cream’
(A3) Mother: ja, iss mal schoen weiter. yes eat PART nicely further. ‘Yes, just go on eating’ (A4) Mother: lass dir ruhig zeit. leave yourself quietly time ‘Give yourself some time’ (A5) Mother: womit isst du das eis denn? what-with eat you the icecream PART? ‘With what are you eating the icecream?’ (A6) Liam:
xxx mit Ole, hm mit Leon, mit xxx. with Ole, hm with Leon, with xxx ‘With Ole, hm with Leon, with xxx.’
(A7) Mother: aha. Ah ‘Ah’ (A8) Mother: womit [?]. what-with ‘With what?’ (A9) Liam:
mit du [?] (=dir) (corr: DAT ) with you-NOM ‘With you.’
(A10) Mother: womit [?] du [?] und isst du das [//] what-with [?] du [?] and eat you that ist das dein messer, is that your knife oder womit isst du das? or what-with eat you that? ‘With what [?] you [?] and you eat that [//] is that your knife or with what are you eating that?’
Learning to encode possession 201
(A11) Liam:
mein loeffel. my spoon ‘My spoon.’
(A12) Mother: das is(t) dein loeffel? that is your spoon ‘That is your spoon? (A13) Liam:
ja. yes ‘Yes.’
(A14) Mother: nein, das is(t) mamas loeffel. no, that is mommy’s spoon ‘No, that is mommy’s spoon.’ (A15) Liam:
na. na ‘No.’
(A16) Mother: aber mamas eis. but mommy’s ice-cream ‘but (it is) mommy’s ice-cream.’ (A17) Liam:
na. na ‘No.’
(A18) Mother: mir gehoert das eis. me-DAT belongs the ice-cream ‘The ice-cream belongs to me.’ (A19) Liam:
nea. nea ‘No.’
(A20) Mother: ach dir gehoert das eis. oh you-DAT belongs the-NOM ice-cream. ‘oh, the ice-cream belongs to you.’ (A20) Mother: na gut, dann is(t) gut. oh well, then is good ‘That is ok, then.’ (A21) Mother: oder gehoert das der Luna? or belongs that the-DAT Luna? ‘Or does it belong to Luna?’
202 Sonja Eisenbeiß, Ayumi Matsuo and Ingrid Sonnenstuhl (A22) Lenny:
der Leon hat sich nich(t) die zaehne the Leon has himself not the-ACC teeth geputzt, mama. brushed, mommy ‘Leon has not brushed his teeth, mommy.’
(A23) Mother: Leon, putz dir bitte die zaehne, ja? Leon, brush you-DAT please the teeth, yes? ‘Leon, brush your teeth, yes?’ This transcript comes from the L-family corpus, a corpus collected by Sonja Eisenbeiß (http://corpus1.mpi.nl/ds/imdi_browser/, see MPI corpora > L1 Acquisition). The collection and archiving of this corpus has been funded by the Max-Planck-Society; and we would like to thank the family for their participation in the project and all the members of the Max-PlanckInstitute who were involved in the collection, digitization and archiving of the data. The transcription of the recordings was funded by a Research Promotion Grant of the University of Essex, awarded to Sonja Eisenbeiß. The video recordings were transcribed by two native speakers and conflicts were either resolved or – if this was not possible – the relevant (part of the) utterance was classified as unintelligible (Eisenbeiß and Sonnenstuhl 2007). The ages at the time of recording are: Lenny (7;9), Leon (4;7), Liam (2;5). Appendix B: Stages of Noun-Phrase Development utterances child Filea age 1–2 Ann
3 4 5–6 1–2
Han
3–4 5 6–8
MLU
2;4– 2;5 2;6 2;7 2;8– 2;9
2,01– 2,11 2,53 2,61 2,54– 3,07
2;0– 2;1 2;2– 2;3 2;4 2;6– 2;8
1,18– 1,23 1,23– 1,38 1,65 2,45– 2,85
n
anal.
stage
D-contexts
predicate +Dformula
n
n
+D
%
remaining D+N combinations
% of tokens +D
types per file
651
473
I
71
33 46 16
48
17
#1: 7, #2: 7
438 589
340 490
II III
83 93
35 42 70 75
5 2
14 3
30 68
#3: 25 #4: 56
744
674
IV
133
125 94
9
7
116
552
442
I
33
19 58 14
74
5
#1: 3, #2: 2
499
355
II
50
10 20
6
60
1
#3: 1, #4: 3
96
54
III
10
7 70
0
0
7
831
538
IV
136
131 96
0
0
131
#5: 5 #6: 23, #7: 47, #8: 31
#5: 37, #6: 47
Learning to encode possession 203
utterances child Filea age 1–2 3–4 Leo
5–8 9– 15
Mat
1;11 –2;0 2;1– 2;2 2;3– 2;5 2;6– 2;11
MLU
n
1,57– 576 1,67 1,60– 659 1,66 1,86– 1587 2,08 2,12– 3157 3,06
9– 17
2;3– 1,25– 1210 2;9 2,11
18 19– 21
2;11 2,62 2;11 2,62– –3;0 2,65
22– 27
3;1– 2,24– 3;6 3,51
anal.
stage
n
% of tokens +D
types per file
12 44
8
67
4
#1: 2, #2: 2
414
II
32
11 34
7
64
4
#3: 3, #4: 1
1166
III
183
131 72 21
16
110
2462
IV
545
499 92 33
7
466
901
I
178
22
12
0
0
22
123
115
II
18
4
22
0
0
4
264
243
III
67
58
87
0
0
58
804
719
IV
204 184
90
0
0
184
2,44
2344
1450
III
3,6
4,22
2314
1795
IV
b
%
+D
27
2,1
1
n
remaining D+N combinations
I
b
Car
predicate +Dformula
341
Andb 1
a
D-contexts
#5: 10, #6: 26, #7: 26, #8: 17 #9: 36, #10: 38, #11: 58, #12: 54, #13: 42, #14: 47, #15: 52 #9: 0, #10: 2, #11: 0, #12: 0, #13: 2, #14: 0, #15: 3, #16: 2, #17: 10 #18: 4, #19: 21, #20: n.a., #21: 23 #22: 37, #23: 6, #24: 23, #25: 27, #26: 35, #27: 28
The column “file” provides the number of the respective recording. The column MLU shows the mean length of utterance for this recording. The columns under “utterances” provide the total number of utterances and the total number of analyzable utterances (i.e. utterances that were intelligible and not simple yes/no answers or formulas such as hallo ‘hello’). Under “D-contexts” are: (i) the number of contexts in which an adult native speaker would have produced a D-element, i.e. a nominal function word (determiner, possessive pronoun or quantifier), (ii) the number of overt D-elements and (iii) the percentage of D-contexts where a D-element was used. The column predicate+D-formula shows how many of the D-elements that children produced were found in potentially formulaic combinations with a small set of high-frequency predicates (e.g. das-is-ein-X ‘that-is-a-X’, diemama ‘the mommy’). The following columns show how many tokens and different types of D-element-noun combinations remained after we excluded the predicate+D formulas from the total number of D-elements in D-contexts. For the cross-sectional data from the advanced stages no quantitative analyses of predicate+D formula and D+N combinations were carried out as both Andreas and Carsten used D-elements with broad ranges of nouns and predicates in obligatory contexts.
204 Sonja Eisenbeiß, Ayumi Matsuo and Ingrid Sonnenstuhl Acknowledgements We would like to thank the children and the families who provided us with data and Harald Clahsen for letting us use the Lexlern data. We are grateful to Bill McGregor, Martin Henson and Bettina Landgraf for valuable comments on earlier drafts and to Anette Rosenbach and Nikola Koch for many discussions about possessive constructions. The research reported in this article was supported by a grant from the Research Promotion Fund, University of Essex.
References Ambridge, Ben, Julian M. Pine, Caroline F. Rowland and Chris R. Young 2008 The effect of verb semantic class and verb frequency (entrenchment) on children’s and adults’ graded judgments of argument structure overgeneralization errors. Cognition 106: 87–129. Armon-Lotem, Sharon 1998 Mommy sock in a minimalist eye: On the acquisition of DP in Hebrew. In Issues in the Theory of Language Acquisition, Norbert Dittmar and Zvi Penner (eds.), 15–36. Bern: Lang. Armon-Lotem Sharon, Stephen Crain and Spyridoula Varlokosta 2005 Interface conditions in child language: a crosslinguistic look at some aspects of possession. Language Acquisition: A Journal of Developmental Linguistics 12 (2): 171–217. Avrutin, Sergey 1999 Development of the Syntax-Discourse Interface. Dordrecht: Kluwer. Baker, Carl L. 1979 Syntactic theory and the projection problem. Linguistic Inquiry 10: 533–581. Baron, Irene, Michael Herslund and Finn Sørensen (ed.) 2001 Dimensions of Possession. Amsterdam /Philadelphia: Benjamins. Bates, Elizabeth and Brian MacWhinney 1987 Competition, variation and language learning. In Mechanisms of Language Acquisition, Brian MacWhinney (ed.), 157–194. Hillsdale, NJ: Erlbaum. Berman, Ruth A. 1985 The acquisition of Hebrew. In The Crosslinguistic Study of Language Acquisition. Vol. 1, Dan I. Slobin (ed.), 255–371. Hillsdale, NJ: Erlbaum. Berman, Ruth A. 1988 Word class distinctions in developing grammars. In Categories and Processes in Language Acquisition, Yonata Levy, Izchak M. Schlesinger, and Martin D. S. Braine (eds.), 45–72. Hillsdale, NJ: Erlbaum.
Learning to encode possession 205 Bertolo, Stefano (ed.) 2001 Language Acquisition and Learnability. Cambridge: Cambridge University Press. Bohnacker, Ute 1997 Determiner phrases and the debate on functional categories in early child language. Language Acquisition 6: 49–90. Borer, Hagit and Kenneth Wexler 1987 The maturation of syntax. In Parameter Setting, Thomas Roeper and Edwin Williams (eds.), 123–172. Dordrecht: Reidel. Bowerman, Melissa 1985 Mapping thematic roles onto syntactic functions: are children helped by innate linking rules? Linguistics 28: 1253–1289. Braine, Martin D. S. and Patricia J. Brooks 1995 Verb argument structure and the problem of avoiding an overgeneral grammar. In Beyond Names for Things: Young Children’s Acquisition of Verbs, Michael Tomasello and Walter E. Merriman (eds.), 352– 376. Hillsdale, NJ: Erlbaum. Brown, Roger 1973 A First Language: the Early Stages. London: Allen and Unwin. Chappell, Hilary and William B. McGregor, (eds.) 1996 The Grammar of Inalienability: A Typological Perspective on Body Part Terms and the Part-Whole Relation. Berlin and New York: Mouton de Gruyter. Chomsky, Noam 1965 Aspects of the Theory of Syntax. Cambridge, MA: MIT Press. Chomsky, Noam 1981 Lectures on Government and Binding. Dordrecht: Foris. Chomsky, Noam 1995 The Minimalist Program. Cambridge, MA: MIT Press. Chomsky, Noam 2001 Beyond explanatory adequacy. MIT Occasional Papers in Linguistics 20. Cambridge, MA: MIT Press. Chouinard, M. M. and Eve V. Clark 2003 Adult reformulation of childerrors as negative evidence. Journal of Child Language 30, 637–669. Clahsen, Harald and Martina Penke 1992 The acquisition of agreement morphology and its syntactic consequences: new evidence on German child language from the SimoneCorpus. In The Acquisition of Verb Placement: Functional Categories and V2 Phenomena in Language Acquisition, Jürgen M. Meisel (ed.), 181–224. Dordrecht: Kluwer. Clahsen, Harald, Sonja Eisenbeiß and Anne Vainikka 1994 The seeds of structure. A syntactic analysis of the acquisition of case marking. In Language Acquisition Studies in Generative Grammar,
206 Sonja Eisenbeiß, Ayumi Matsuo and Ingrid Sonnenstuhl Teun Hoekstra and Bonnie D. Schwartz (eds.), 85–118. Amsterdam / Philadelphia: Benjamins. Clahsen, Harald, Anne Vainikka, and Martha Young-Scholten 1990 Lernbarkeitstheorie und Lexikalisches Lernen. Eine kurze Darstellung des LEXLERN-Projekts. Linguistische Berichte 130: 466–477. Clancy, Patricia M. 1985 The acquisition of Japanese. In The Crosslinguistic Study of Language Acquisition. Vol. 1, Dan I. Slobin (ed.), 373–524. Hillsdale, NJ: Erlbaum. Clark, Eve V. 1978 Ergativity. In Syntactic Typology: Studies in the Phenomenology of Language, Winfred P. Lehmann (ed.), 329–394. Hassocks, Sussex: Harvester Press. Clark, Eve V. 1985 The acquisition of Romance with special reference to French. In The Crosslinguistic Study of Language Acquisition. Vol. 1, Dan I. Slobin (ed.), 687–782. Hillsdale, NJ: Erlbaum. Clark, Eve V. 1987 The principle of contrast: a constraint on language acquisition. In Mechanisms of Language Acquisition, Brian MacWhinney (ed.), 1–33. Hillsdale, NJ: Erlbaum. Corina, David P. 1994 The induction of prosodic constraints: implications for phonological theory and mental representation. In The Reality of Linguistic Rules, Susan D. Lima, Roberta L. Corrigan and Gregory K. Iverson (eds.), 115–145. Amsterdam /Philadelphia: Benjamins. Cowie, Fiona 1999 What’s Within: Nativism Reconsidered. New York: Oxford University Press. Crain, Stephen 1991 Language acquisition in the absence of experience. Behavioral and Brain Sciences 14: 597– 611. Denison, David, Alan Scott and Kersti Börjars 2008 What’s wrong with possessive ’s? Presentation at ISLE1, University of Freiburg, 8 –11 October 2008. (Available online at http://llc.stage.manchester.ac.uk/subjects/lel/staff/ david-denison/papers/thefile,143616,en.pdf.) Eisenbeiß, Sonja 1994 Elizitation von Nominalphrasen und Kasusmarkierungen. In Elizitationsverfahren in der Spracherwerbsforschung: Nominalphrasen, Kasus, Plural, Partizipien. (Arbeiten des Sonderforschungsbereichs 282, Nr. 57), Sonja Eisenbeiß, Susanne Bartke, Helga Weyerts and Harald Clahsen (eds.), 1–38. Düsseldorf: Heinrich-Heine-Universität.
Learning to encode possession 207 Eisenbeiß, Sonja 2000 The acquisition of the Determiner Phrase in German child language. In The Acquisition of Syntax: Studies in Comparative Developmental Linguistics, Marc-Ariel Friedemann and Luigi Rizzi (eds.), 26–62. London: Longman. Eisenbeiß, Sonja 2003 Merkmalsgesteuerter Grammatikerwerb. Doctoral dissertation, University of Düsseldorf. (Available online at http://docserv.uniduesseldorf.de/servlets/DerivateServlet/Derivate-3185/1185.pdf.) Eisenbeiß, Sonja 2006 Documenting child language. In Language Documentation and Description. Vol. 3, Peter K. Austin (ed.), 106 –140. London: SOAS, The Hans Rausing Endangered Languages Project. Eisenbeiß, Sonja 2007 The lexical learning hypothesis. Essex Research Reports in Linguistics 54: 1– 4. (To appear in: The Cambridge Encyclopedia of the Language Sciences.) Eisenbeiß, Sonja 2009 Generative approaches to language learning. Linguistics 47 (2): 273– 310. Eisenbeiß, Sonja and Ayumi Matsuo 2003 External and internal possession – a comparative study of German and Japanese child language. Poster presented at the 28th Annual Boston University Conference on Language Development, Boston University, USA. Eisenbeiß, Sonja and Ayumi Matsuo 2005 Eliciting language production data from young children. Presentation at the Xth International Congress for the Study of Child Language, Berlin, Germany. Eisenbeiß, Sonja and William B. McGregor 1999 The circle of dirt. Ms. Max-Planck-Institute for Psycholinguistics, Nijmegen. Eisenbeiß, Sonja and Ingrid Sonnenstuhl 2007 Transcription conventions for the L-family corpus. Ms. University of Essex. Goldberg, Adele 1995 Constructions. Chicago: University of Chicago Press. Goldberg, Adele 2006 Constructions at Work: the Nature of Generalization in Language. Oxford: Oxford University Press. Grimshaw, Jane 1994 Minimal projection and clause structure. In Syntactic Theory and First Language Acquisition: Cross-linguistic Perspectives, Barbara Lust, M. Suner and J. Whitman (eds.), 75–83. Hillsdale, N.J.: Lawrence Erlbaum Associates.
208 Sonja Eisenbeiß, Ayumi Matsuo and Ingrid Sonnenstuhl Guasti, Maria Teresa, Anne Christophe, Brit van Ooyen and Marina Nespor 2001 Pre-lexical setting of the head: Complement parameter through prosody. In Approaches to Bootstrapping: Phonological, Lexical, Syntactic and Neurophysiological Aspects of Early Language Acquisition, Vol. 1, Jürgen Weissenborn and Barbara Höhle (eds.), 231–248. Amsterdam /Philadelphia: Benjamins. Harbert, Wayne 2007 The Germanic Languages. Illustrated Edition. Cambridge: Cambridge University Press. Heine, Bernd 1997 Possession: Cognitive Sources, Forces, and Grammaticalization. New York: Cambridge University Press. Hyams, Nina M. 1996 The underspecification of functional categories in early grammar. In Generative Perspectives on Language Acquisition. Empirical Findings, Theoretical Considerations and Crosslinguistic Comparisons, Harald Clahsen (ed.), 91–127. Amsterdam/ Philadelphia: Benjamins. Gerhard Jäger and Annette Rosenbach 2006 The winner takes it all – almost. Cumulativity in grammatical variation. Linguistics 44 (5): 937–971. Kiparsky, Paul 1982 From cyclic phonology to lexical phonology. In The Structure of Phonological Representations. Part 1, Harry van der Hulst and Norval Smith (eds.), 131–175. Dordrecht: Foris. König, Ekkehard and Martin Haspelmath 1998 Les constructions à possesseur externe dans les langues d’Europe. In Actance et Valence dans les Langues d’Europe, Jack Feuillet (ed.), 525– 606. Berlin and New York: Mouton de Gruyter. Küntay, Aylin and Dan I. Slobin 1996 Listening to a Turkish mother: some puzzles for acquisition. In Social Interaction, Social Context, and Language, Dan I. Slobin, Julie Gerhardt, Amy Kyratzis and Jiansheng Guo (eds.), 265–286. Mahwah, NJ: Erlbaum. Lust, Barbara 1994 Functional projection of CP and phrase structure parametrization: an argument for the strong continuity hypothesis. In Syntactic Theory and First Language Acquisition: Cross-Linguistic Perspectives. Vol. 1, Barbara Lust, Margarita Suner and John Whitman (eds.), 85–118. Hillsdale, NJ: Erlbaum. MacWhinney, Brian 2000 The CHILDES Project: Tools for Analyzing Talk. 3rd Edition. Vol. 2: The Database. Mahwah, NJ: Erlbaum. Marcus, Gary F. 1993 Negative evidence in language acquisition? Cognition 46: 53–85.
Learning to encode possession 209 Marcus, Gary F., Steven Pinker, Michael Ullman, Michelle Hollander, John T. Rosen and Fei Xu 1992 Overregularization in language acquisition. Monographs of the Society for Research in Child Development. Vol. 57 (Serial No. 228), 1–165. Marinis, Theodoros 2002 Acquiring the possessive construction in Modern Greek. In Proceedings of the 1999 GALA Conference, Ingeborg Lasser (ed.), 57–80. Frankfurt: Lang. Marinis, Theodoros 2003 The Acquisition of the DP in Modern Greek. Amsterdam /Philadelphia: Benjamins. Miller, Max 1976 Zur Logik der frühkindlichen Sprachentwicklung. Stuttgart: Klett. Mills, Anne E. 1985 The acquisition of German. In The Crosslinguistic Study of Language Acquisition. Vol. 1, Dan I. Slobin (ed.), 141–254. Hillsdale, NJ: Erlbaum. Neumann, Dorothea 1995 The dative and the grammar of body parts in German. In The Grammar of Inalienability: A Typological Perspective on Body Part Terms and the Part-Whole Relation, Hilary Chappell and William B. McGregor (eds.), 745–779. Berlin /New York: Mouton de Gruyter. Payne, Doris L and Immanuel Barshi (eds.) 1999 External Possession. Amsterdam /Philadelphia: Benjamins. Penner, Zvi, Manuela Schönenberger and Jürgen Weissenborn 1994 The acquisition of object placement in early German and Swiss German. Linguistics in Potsdam 1: 93–108. Penner, Zvi and Jürgen Weissenborn 1996 Strong continuity, parameter setting and the trigger hierarchy: On the acquisition of the DP in Bernese Swiss German and High German. In Generative Perspectives on Language Acquisition. Empirical Findings, Theoretical Considerations and Crosslinguistic Comparisons, Harald Clahsen (ed.), 161–200. Amsterdam/Philadelphia: Benjamins. Peters, Ann M. and Lise Menn 1993 False starts and filler syllables: ways to learn grammatical morphemes. Language 69: 742–777. Phillips, Colin 1996 Order and structure. PhD thesis, Cambridge, MA: MIT. Pinker, Steven 1984 Language Learnability and Language Development. Cambridge, MA: Harvard University Press. Pinker, Steven 1989 Learnability and Cognition: The Acquisition of Argument Structure. Cambridge, MA: MIT Press.
210 Sonja Eisenbeiß, Ayumi Matsuo and Ingrid Sonnenstuhl Radford, Andrew 1990 Syntactic Theory and the Acquisition of English Syntax: The Nature of Early Child Grammars of English. Oxford: Blackwell. Radford, Andrew 1996 Towards a structure-building model of acquisition. In Generative Perspectives on Language Acquisition. Empirical Findings, Theoretical Considerations and Crosslinguistic Comparisons, Harald Clahsen (ed.), 43–90. Amsterdam /Philadelphia: Benjamins. Radford, Andrew and Joseph Galasso 1998 Children’s possessive structures: a case study. Essex Research Reports in Linguistics 19. Essex: University of Essex, 37–45. Rizzi, Luigi 1993/94 Some notes on linguistic theory and language development: The case of root infinitives. Language Acquisition 3: 371–393. Rizzi, Luigi 2000 Remarks on early null subjects. In The Acquisition of Syntax: Studies in Comparative Developmental Linguistics, Marc-Ariel Friedemann and Luigi Rizzi (eds.), 269–292. London: Longman. Rosenbach, Anette 2002 Genitive Variation in English: Conceptual Factors in Synchronic and Diachronic Studies (Topics in English Linguistics 42.) Berlin /New York: Mouton de Gruyter. Rosenbach, Anette 2003 Aspects of iconicity and economy in the choice between the sgenitive and the of-genitive in English. In Determinants of Grammatical Variation in English, Günter Rohdenburg and Britta Mondorf (eds.), 379–411. (Topics in English Linguistics 43) Berlin/New York: Mouton de Gruyter. Rosenbach, Anette 2005 Animacy versus weight as determinants of grammatical variation in English. Language 81: 613–644. Rosenbach, Anette 2008 Animacy and grammatical variation – findings from English genitive variation. Lingua 118: 151–171. Ruff, Claudia 2000 Besitz und Besitzer in der Sprachentwicklung von deutschen und italienischen Kindern. PhD thesis, TU Braunschweig, Biowissenschaften, Psychologie. (Available online at http://deposit.ddb.de/cgibin/dokserv?idn=961129425.) Scott, Alan, David Denison and Kersti Börjars 2007 Is the English possessive ’s truly a right edge phenomenon? Handout of paper presented at The Second International Conference on the Linguistics of Contemporary English (ICLCE2), University of Toulouse-le Mirail, 2nd July 2007.
Learning to encode possession 211 Seiler, Hansjakob 1983 Possession as an Operational Dimension of Language. Tübingen: Narr. Shibatani, Masayoshi 1994 An integrational approach to PR raising, ethical datives, and adversative passives. BLS 20: 461– 486. Slobin, Dan I., Melissa Bowerman, Penelope Brown, Sonja Eisenbeiß and Bhuvana Narasimhan in prep. Putting things in places: developmental consequences of linguistic typology. Stenzel, Achim 1994 Case assignment and functional categories in bilingual children: Routes of development and implications for linguistic theory. In Bilingual First Language Acquisition: French and German Grammatical Development, Jürgen M. Meisel (ed.), 161–208. Amsterdam/Philadelphia: Benjamins. Stephany, Ursula 1997 The acquisition of Greek. In The Crosslinguistic Study of Language Acquisition. Volume 4, Dan I. Slobin (ed.), 183–333. Mahwah, NJ: Erlbaum. Stephany, Ursula 1998 A crosslinguistic perspective on the category of nominal number and its acquisition. In Studies in the Acquisition of Number and Diminutive Marking, Steven Gillis (ed.), 1–23. (Antwerp Papers in Linguistics 95) Antwerpen: Universitaire Istelling Antwerpen. Tomasello, Michael 1998 One child’s early talk about possession. In The Linguistics of Giving, John Newman (ed.). Amsterdam/Philadelphia: John Benjamins. Tomasello, Michael 2003 Constructing a Language: A Usage-Based Theory of Language Acquisition. Cambridge, MA: Harvard University Press. Tomasello, Michael 2006 Acquiring linguistic constructions. In Handbook of Child Psychology, D. Kuhn and R. Siegler (eds.), 1–101. New York: Wiley. Tsunoda, Tasaku 1995 The possession cline in Japanese and other languages. In The Grammar of Inalienability: A Typological Perspective on Body Part Terms and the Part-Whole Relation, Hilary Chappell and William B. McGregor (eds.), 565–630. Berlin /New York: Mouton de Gruyter. van der Linden, E. and A. Blok-Boas 2005 Exploring possession in simultaneous bilingualism: Dutch/French and Dutch/Italian. In EUROSLA Yearbook, S. Foster-Cohen, M. d. P. García-Mayo and J. Cenoz (eds.), 103–135. Amsterdam /Philadelphia: Benjamins.
212 Sonja Eisenbeiß, Ayumi Matsuo and Ingrid Sonnenstuhl Wagner, Klaus R. 1985 How much do children say in a day? Journal of Child Language 12: 475– 487. Wexler, Kenneth 1998 Very early parameter setting and the unique checking constraint: a new explanation of the optional infinitive stage. Lingua 106: 23–79.
Plain vs. situated possession in Czech: A constructional account Mirjam Fried 1. Introduction The focus of this paper is the kind of possession that can be identified broadly as a time-stable relation that is presupposed (not asserted) between two entities, a possessor (PR) and a possessum (PM). It is well known that this relation can be expressed in a number of ways even in a single language, let alone cross-linguistically; what still remains to be worked out in sufficient detail is the exact nature of the variation and the relationships among the variants. My purpose here is (i) to identify the factors that help differentiate between syntactically distinct expressions of this kind of possession, and (ii) to propose a way of representing the patterns in a network of grammatical constructions organized around their shared features. The illustrative material comes from authentic Czech usage, both written and spoken, as attested in the Czech National Corpus (CNK).1 Heine (1997: 143) labels the presupposed, time-stable possession as “attributive possession”, so categorized in contrast to “predicative possession”, and this categorization includes characteristic differences in syntax as a crucial criterion: phrasal syntax for the former, clausal for the latter. This distinction, though, is too general to give us a realistic picture of the variety of possessive expressions. The non-clausal expressions tend to come in several distinct formal variants which represent a rather diverse set, both in their form and their function. I will examine a subset of such variants in Czech and offer a functionally and cognitively oriented analysis that is motivated by three general questions: (i) What kind of linguistic knowledge is necessary for native-like production and comprehension of these possessive patterns? (ii) How can the speakers’ knowledge and understanding be adequately represented? And (iii) can the representation help us make more precise claims about attributive possession as a linguistic category, with implications beyond accounting for the Czech facts? 1
The material used here comes from roughly 400 million-word written corpora (SYN2000, SYN 2006PUB), supplemented by a sample from Czech language and literature (LITERA, SYNEK), and a 2.5 million-word spoken corpora (PMK, BMK, ORAL2006).
214 Mirjam Fried Some attributive possessive expressions are exemplified in (1)–(3); all of these patterns are familiar and found in many languages. The examples in (1) are two variants of an adnominal PR in a NP headed by the PM.2 (1a) is an agreeing modification structure, in which the PR is marked by possessive morphology and agrees with the noun in case, gender, and number, like any pre-nominal modifier in Czech. In (1b), the PR is in the genitive and obligatorily follows the head noun. Conceptually, both of them could be treated as cases of “possessor specification” (Heine 1997: 167), but I will refer to them collectively as Genitive PR (GP), choosing a more traditional and familiar label. In (2), both the PM and the PR are syntactically independent NPs and the PR status is associated with the dative marking on the PR. This pattern is known as “external possessor” (e.g. Payne and Barshi 1999) and is typically associated with PR affectedness; I will, therefore, call it Affected PR (AP). The example in (3) shows a strategy in which the PR is not explicitly identified.3 For easier orientation, the possessive expressions will be in bold throughout the paper.4 (1)
2
3
4
5
Genitive PRs (GP)5 a. Pre-nominal Natali-in-y rodie ty maj Natalie-PR.F-NOM.PL parents.NOM they have.PRES.3PL furt {ákej státní svátek , tak sou doma poád.} at.all.times ‘Natalie’s parents, they have {state holidays all the time, so they’re always at home}’ [PMK137;148304] The distribution of these two forms is conditioned morphologically: the postnominal genitive is obligatory if the PR is morphologically neuter (1b), or a multiword NP, or a plural noun, or is modified by a relative clause. In other cases, the agreeing pre-nominal form is the neutral choice. There are additional variants, including a morphological adjective as an expression of PR status. I will not be concerned with these marginal cases here but the general approach provides ways for incorporating such patterns as well. A note on presenting the data: when the surrounding text is necessary for clearer understanding, it will be enclosed in curly brackets {} and left without interlinear glossing. Abbreviations in examples: ACC ‘accusative’, AP ‘affected PR’, AUX ‘auxiliary’, DAT ‘dative’, F ‘feminine’, GEN ‘genitive’, GP ‘genitive PR’, IMP ‘imperative’, INF ‘infinitive’, INS ‘instrumental’, LOC ‘locative’, M ‘masculine’, N ‘neuter’, NEG ‘negative’, NOM ‘nominative’, PRES ‘present tense’, PST ‘past tense’, RF ‘reflexive’, PR ‘possessor’, SG/PL ‘singular/plural’, and SPRL ‘superlative’.
Plain vs. situated possession in Czech: A constructional account 215
b. Post-nominal Ohneme no iky dítte v kolenou bend.PRES.1PL leg.ACC.F.PL child.GEN.SG.N in knee.LOC.PL ‘We’ll bend the baby’s legs at [the] knees’ [SYN2000;34388742] (2)
(3)
Affected PR (AP) spravil nám i auta fix.PST.SG.M 1.PL.DAT also car.ACC.PL ‘he even fixed our cars’
[SYNEK;1598227]
Implicit PR: ten pán ztratil brejle {a nemohl je najít} that man.NOM.SG.M lose.PST.SG.M glasses.ACC ‘that man lost [his] glasses {and he couldn’t find them}’ [LITERA;1680536]
My primary concern will be the relationship between (1) and (2); reference to the Implicit PR variant is necessary mainly as a background against which the explicit patterns can be studied. The example in (3) is simple: even the minimal context given here is enough to suggest that the owner of the lost glasses is the subject referent, rather than some other person. The preferred interpretation has to do with what we know about glasses as common personal possessions. Other times the implicitness allows a greater range of possible interpretations, especially if the sentence contains more than one potential candidate for the PR status, including direct discourse participants, or if the PM is something other than a body part. Speakers know to infer the appropriate configuration based on conventional expectations about possible possessive relationships vis-à-vis particular context. However, if it is the case that a possessive relation can be (and in Czech very often is) left implicit, then we have to ask what reasons speakers might have for choosing the explicit options and particularly, what – if anything – conditions the choice of (1) vs. (2). I will be concerned only with the second (easier) question here, pursuing the hypothesis that the differences should revolve primarily around the (in)alienability of the PM and the affectedness of the PR as two independent, competing factors; the hypothesis is motivated by the patterning found in many other languages. However, the analysis will demonstrate that the interaction between affectedness and (in)alienability is systematically more complex than this: specifically, in cases of conflict, affectedness takes precedence over inalienability in licensing the AP form. This leads, among
216 Mirjam Fried other things, toward an absolute prohibition on PM (of any kind) as a transitive subject and almost equally strong prohibition on subjects of active intransitive verbs. Given that similar syntactic restrictions on the grammatical role of the PM are cross-linguistically common in external possessor constructions (Payne and Barshi 1999), the present analysis highlights a semantic explanation that may apply well beyond the Czech facts. Finally, mapping out the interaction between affectedness and (in)alienability also provides a generally applicable approach to incorporating peripheral cases in which a non-possessive expression may acquire a possessive construal. In order to keep the analysis reasonably focused, I will take as my starting point a relatively narrow definition of the concept of possession. I exclude the broad sense of belonging in part-whole relations in which the whole is not an animate entity (Heine’s 1997: 35 “inanimate possession”), but my conception is not quite as narrow as Taylor’s (1989: 202–203) notion of prototypical ownership. Like Taylor, I will take possession to be an “experiential gestalt”, which presupposes that the PR is a sentient being and has specific reference. However, I will consider a broader range of possessa than his prototype allows (“specific concrete things”). Finally, I will also work with the notion of possessibility hierarchy. There seems to be no firm consensus about the exact shape of a universally applicable hierarchy; the main points of variation seem to be the relative placement of kin relations, inherent attributes, and clothing, and their status evidently depends on particular languages. For the purposes of this study, I find it sufficient to assume a very general scale along the lines of Payne and Barshi 1999: body parts > kin > close alienable > distant alienable. However, a close analysis of the Czech AP (section 3) will suggest a more refined scale, partially overlapping with Tsunoda’s (1995: 576) Possession Cline. By examining the range of factors that motivate the distribution of the GP and AP patterns, I will argue that the distinction goes far beyond reducing the issue to treating the two strategies as two formal variants of a conceptual schema called “possessive specification” (Heine 1997: 167), each motivated by a particular instantiation of an existential event schema – Goal vs. Genitive (p. 47).6 Both of these variants merely assert the existence of a presupposed possessive relation, each in a different syntactic form. The Genitive schema is associated with the pattern [PR’s PM]), while the Goal schema, here corresponding to the AP pattern, takes the form [PM (exists) 6
Only these two schemas in Heine’s typology are relevant here. Unlike, for example, Russian or Romanian, Czech does not have any locative-type patterns [PM (is) at PR] expressing possession.
Plain vs. situated possession in Czech: A constructional account 217
for/to PR]. Both schemas are said to correlate with permanent inalienable possession. Neither of these broad schemas thus helps explain the fact that the Czech GP and AP have very different distributions in actual discourse and that the difference in form between (1) and (2), as well as some special syntactic constraints associated with (2) but not with (1), follow from a number of co-occurring properties: the inherent meaning of the PM and the PR, the involvement of the PR in the event expressed by the clause, verb semantics, and information flow. I will also suggest that the presence or absence of some of these properties correlates with particular speech situations. The two forms will thus be best treated as grammatical constructions in the sense of Construction Grammar (see especially Fillmore 1989; Croft 2001; Fried and Östman 2004), and I will also draw on the notion “cognitive frame” (Fillmore 1982) for incorporating the possessive relation in the constructional representations. I will argue that each possessive variant constitutes a semantically and pragmatically distinct pattern – a conventional cluster of semantic, pragmatic, and morpho-syntactic properties – and that each pattern represents a functional prototype within a network of possibilities for expressing attributive possession. The paper is organized as follows. The semantic and pragmatic properties of the PR and the PM are discussed in sections 2 and 3, respectively, addressing the issue of a possessive prototype. Section 4 studies the interaction between the possessive relation and the structure and meaning of the sentence in which it occurs. All of this comes together in section 5, in which I present the patterns (1) and (2) as two grammatical constructions that occupy overlapping domains within the functional space of attributive possession. I illustrate the ways in which the constructions in the network may interact both with each other and with the possessive prototype, suggesting a more systematic account of the ways in which the prototype can be extended to more peripheral instances. Section 6 briefly concludes the paper. 2. Semantic and pragmatic features of the PR The inherent semantics of a quintessential PR does not predict, by itself, anything about the linguistic form and behavior of the noun phrase that encodes the PR; the referents in (1)–(3) are all specific human beings. Moreover, both AP and GP are sometimes possible for a given possessive relation, as shown in (4); note, however, that they are not quite synonymous.
218 Mirjam Fried (4) a. Ped pti lety nám zemel otec. before 5 years 1PL.DAT die.PST.SG.M father.NOM.SG.M {Matka se zhroutila. Z nemocnice ji propustili, ale není v poádku.} ‘Five years ago, father died. {Mother fell apart. She’s back from the hospital, but she’s not in good shape.}’ [SYNEK;2279741] b. Ped pti lety zemel ná otec. before 5 years die.PST.SG.M 1PL.PR.NOM.SG.M father.NOM.SG.M ‘Five years ago, our father died.’ Other times, though, only one form is available. For example, the AP in (5a) cannot be replaced by GP (5b) and the examples in (6) demonstrate the reverse: (5)
a. V u ích nám hvízdal vítr in ear.LOC.PL 1PL.DAT whistle.PST.SG.M wind.NOM.SG.M ‘Wind was whistling in our ears.’ [SYN2000;2062266] b. *V na ich u ích hvízdal in 1PL.PR.LOC.PL ear.LOC.PL whistle.PST.SG.M vítr wind.NOM.SG.M
(6)
a. Z njakého […] dvodu ho from some.GEN reason.GEN 3SG.M.ACC její vlasy nesmírn vzru ovaly. 3SG.F.PR.NOM.PL hair.NOM.PL immensely excite.PST.PL ‘For some [unknown] reason, her hair excited him enormously.’ [SYNEK;941353] b. *… vlasy jí ho nesmírn vzru ovaly hair.NOM.PL 3SG.F.DAT 3SG.M.ACC immensely excite.PST.PL
Similar effects have been noted for various languages (e.g. O’Connor 1994; Croft 1985; Berman 1982; Chappell and McGregor 1995; Manoliu-Manea 1995; Payne and Barshi 1999), including Czech (e.g. Zimek 1960; Pit’ha 1992; Fried 1999a, 2008), and the subsequent discussion will show that the contrast has to do both with the involvement of the PR in the reported event and the relative communicative prominence of the PR and PM in a given piece of discourse.
Plain vs. situated possession in Czech: A constructional account 219
2.1. Affectedness of PR The relationship exemplified in (5) illustrates the fact that inalienable possession overwhelmingly prefers AP in Czech, to the point of leaving it as the only expressive option in some cases (i.e., if the speaker has a reason to mention the PR explicitly in the first place, instead of opting for Implicit PR). The case in (6), then, appears to be an ‘exception’ since the only possibility there is GP, even though the PM is a body part like in (5).7 However, this apparent contradiction is fully motivated and I will revisit it in sections 4 and 5. First let us consider a less categorical situation: examples where AP appears to be replaceable by a GP form. The expressions in (a) are corpus attestations of AP, the (b) examples are their GP counterparts (constructed). (7)
a. {Pak pi la prvodí} a pro típla nám and punch.PST.SG.F 1PL.DAT lístky {a cesta doplynula ve v í pohod} [SYNEK;2261787] ticket.ACC.PL ‘{Then the conductor came} and punched our tickets {and the rest of the trip went all smoothly}. b. pro típla na e lístky punch.PST.SG.F 1PL.PR.ACC.PL ticket.ACC.PL
(8)
a. “vsate se, e se tomu ulpas-ovi policajtsk-ému bet.IMP RF that RF that.DAT nitwit-DAT cop.ADJ-DAT vysmolím ped dvee na schody poop.PRES.1SG in.front.of door onto staircase {a je t si s ním pitom budu povídat!}” [SYN2000;87610] ‘“you can bet that I’ll take a dump on the steps in front of the nitwit cop’s door {and keep chatting with him in the process, too!}”’ b. … e se vysmolím ped dvee toho ulpas-a policajtsk-ého door that.GEN nitwit-GEN cop.ADJ-GEN
7
Whatever observations are made in this paper about the behavior of body parts as the PM, they apply to all specific candidates for body part status. Czech does not make any grammar-coded distinctions between different types of body parts.
220 Mirjam Fried (9)
a. Kdy jim vzali peníze mnovou when 3PL.DAT take.PST.PL money.ACC currency.ADJ.INS .SG reformou, reform.INS.SG {prodávali koberce, perky a obrazy, jen aby mohla zstat doma.} ‘When [the government] took their money in the currency reform, {they kept selling rugs, jewelry and paintings, anything [to make it] possible for her to stay at home.} [SYNEK;653077] b. Kdy vzali jejich peníze when take.PST.PL 3PL.PR money.ACC
The issue in (7)–(9) is not the incompatibility between the PM and the GP form, as is the case in (5). The problem in (7)–(9) is the degree to which the PR is involved in the depicted events. The example in (7) is the most flexible one: GP (7b) implies that the conductor punched some tickets that belonged to us but were not necessarily related to our riding the train. The AP (7a), in contrast, presents the tickets as necessary for the ride, we were holding them in our hands, handed them over, and then took them back from the conductor, to hang onto them. The GP form is in principle possible but somewhat odd, given what we know about the usual ways in which passengers and conductors interact. A GP form in (8) would be even more problematic since its implication would be contradicted by the subsequent coordinated clause: GP (8b) would be felicitous whether the cop is present or not when the speaker is squatting in front of the cop’s door; the automatic interpretation actually would be that the cop is not around. Yet the point of the speaker’s bet is that he will be simultaneously having a conversation with the cop. The contextual incompatibility is still stronger in (9): GP (9b) would be possible whether or not the PR was alive during the currency reform, while AP (9a) is felicitous only if the PR was alive. The AP emphasizes this state of affairs, which makes it the only coherent choice for the follow-up about the consequences the PR suffered (selling off property to make ends meet). To summarize, the two forms display a systematic division of labor that can be related to Bally’s (1995/1926) notion of personal domain or (subjectively applied) indivisibility. GP expresses plain possession in the broadest sense, where the concept of indivisibility plays no role. By contrast, AP casts the possession relation as something that is relevant to the PR in a particular way, as something in his sphere of interest beyond just the fact of being owned. AP signals that the PR is being affected (positively or negatively)
Plain vs. situated possession in Czech: A constructional account 221
by something that affects the PM (cf. also Chappell and McGregor 1995). In this light, it is not surprising that certain types of possessa strongly prefer AP, in contrast to GP: the tighter the possessive relationship, the greater the chance that manipulating the PM will directly affect the PR. Body parts are the extreme on the continuum since they are truly inseparable from the PR, and hence the unacceptability of (5b). The GP form can only be interpreted in one way: the ears are not attached to the PR’s body (are not, that is, bona fide body parts) but are some ear-like objects, physically detached from any bodies. This extreme restriction applies only to the pre-nominal GP, though; as shown in (1b), the post-nominal genitive is sometimes attested with body parts and the form maintains the true body part reading. We thus have to conclude that the relative flexibility between allowing both AP and GP extends all the way to the top of the possessibility hierarchy. Nevertheless, a difference in form (AP vs. GP) always correlates with a shift in interpretation, no matter what the PM. For example, the choice of GP with a body part in (1b) is motivated by the type of discourse and the focus of attention. This token is taken from instructions about how to use a rectal thermometer to take a baby’s temperature. The communicatively relevant issue here is how to manipulate the baby’s position so that the person succeeds in what needs to be done; the author of the instructions is not concerned with what (dis)comfort it may bring to the baby, which would be the only possible reading if an AP form were chosen. A similar analysis applies to example (4), in which a kinship relation is cast in a different light depending on whether AP or GP is used. The GP in (4b) would imply that the speaker was estranged from, or at least not very close to, his father and therefore unaffected by his death, in contrast to (4a), which unequivocally implies that the father’s death had tangible consequences, as is clear from the subsequent text. Two generalizations emerge from these facts: (i) inherent (in)alienability is not a good predictor of AP vs. GP encoding (as also noted explicitly by Bally 1926/1995; Chappell and McGregor 1995) and (ii) the PR in the AP form plays a special role in the event expressed by the predicate, while no special status is associated with the GP variant. 2.2. Information structure As a syntactically independent NP, the dative PR is free with respect to its position in the clause relative to the PM. It can, therefore, participate in ar-
222 Mirjam Fried ticulating topic-focus relations within the possessive relation itself, and thereby accommodate the nuances of structuring information flow, which in Czech is expressed through word order (Firbas 1966; Dane 1974; Sgall et al. 1986; Grepl and Karlík 1998). On the other hand, the PR position within GP is restricted: the pre-nominal PR cannot be very easily separated from the PM (except under very special contextual circumstances) and the genitive is obligatorily fixed in the post-nominal position. Consequently, GP is dispreferred or outright impossible in certain contexts simply for reasons of information structure, even if it might be appropriate on semantic and syntactic grounds. Consider (10): (10) a. {Za nco se stydl,} bál se podívat lidem fear.PST.SG.M RF look.INF people.DAT do tváe, {kdy jim nco nabízel} in face.GEN.SG.F ‘{He felt embarrassed about something,} he was afraid to look people in their FACES {when he was selling them something}’ [LITERA; otapavel] b. bál se podívat do tváí lidí, fear.PST.SG.M RF look.INF in face.GEN.SG.PL people.GEN ‘…he was afraid to look in the faces of PEOPLE’ The example in (10a) gives discourse prominence to the faces, as indicated by the small caps in the English translation. The point of the sentence is to describe the subject’s reluctance to face people. Replacing AP by a GP expression (10b) would shift the focal status onto the genitive PR, simply by virtue of its phrase-final position, and this reconfiguration would sound very odd in the context of this passage. (10b) creates the expectation that lidí ‘of people’ is to be understood as contrastive focus: either in contrast to the faces of other entities mentioned earlier, or as the head of a restrictive relative clause that would have to follow (‘he was afraid to look in the faces of [those] people who he was selling something to, but not other folks’). Such readings are clearly unintended here, and using AP is a general strategy for making the PR and the PM available for structuring information flow independently of each other. The infelicitous character of (10b) cannot be easily attributed to the disembodiment reading associated with pre-nominal GP, for two reasons. First, the GP pattern as such does not imply disembodiment, as we see in (5a); in that sentence, the PR’s hair is unambiguously attached to her head.
Plain vs. situated possession in Czech: A constructional account 223
Moreover, since the post-nominal GP is obligatory with certain grammatical types of PRs, listed in footnote 1 (here the issue is the plural of the PR noun), the inherent semantics of the PM becomes irrelevant. And second, since the information structure value of the PR can be freely manipulated in the AP, it is possible to form a discourse-structure equivalent of (10b) as well, shown in (10c) below: (10) c. bál se podívat do tváe lidem fear.PST.SG.M RF look.INF in face.GEN.SG.F people.DAT ‘…he was afraid to look in the faces of PEOPLE’ This AP form is subject to exactly the same interpretive restrictions as (10b), namely, putting the people into a focus role, whether contrastive or plain. The only difference is that the post-nominal GP cannot be preposed, under any circumstances. To summarize, AP gives the PR a certain prominence that GP does not provide. However, this kind of prominence must be understood as an eventbased prominence (i.e., treating the PR as an affected event participant, as opposed to just an owner), independent of a discourse-based prominence (i.e., assigning information structure roles within the possessive relationship), which can be manipulated separately. I will revisit the implications of this feature in section 4. 3. Semantics of the possessum As illustrated in (11), the head noun in the GP pattern can be semantically anything: concrete, abstract, animate, inanimate. (11) Han-in-y ruce/sestry/kamarádky/knihy/názory/povinnosti Hana-POSS.F-NOM.PL hands/sisters/friends/books/opinions/duties This semantic freedom suggests that the GP pattern simply marks a very general relationship between two entities, which can be conceptualized as denoting a unit of sorts. This conceptualization accommodates possession (including inalienable) in the narrow, experiential sense as well, but the form evidently is not restricted to encoding truly possessive relations. The semantic and functional breadth is corroborated by the fact that the same form is found not only in combinations such as (11), all expressing (at least loosely understood) possession, but also in various common ‘genitive’
224 Mirjam Fried functions (not discussed here) or in the ‘syntactic’ function, i.e. encoding verbal arguments in nominalized expressions. The latter is exemplified in (12), where the pre-nominal possessive form Janovu ‘Jan’s’ marks the agent role of the noun pomoc ‘help’, not any PR of help. (12) dkoval mi za Jan-ov-u pomoc thank.PST.SG.M 1SG.DAT for Jan-POSS.M-ACC.SG.F help.ACC.SG.F ‘he thanked me for Jan’s help’ [LITERA;3101771] The AP pattern is clearly different from GP in that AP does not cover the syntactic function shown in (12); we cannot create an AP paraphrase of (12). However, paraphrasing the expressions in (11) is possible, at least in principle. Thus, considering the AP examples in (7)–(9) above and (13) below, AP might not appear to present a dramatically different picture from GP with respect to the kinds of possessa it permits, along the full possessibility hierarchy: (13) a. Komolil jména lidem pi nejrznj ích distort.PST.SG.M name.ACC.PL people.DAT at SPRL.various píle itostech [LITERA;375950] occasion.LOC.PL ‘He mangled people’s names on all kinds of occasions’ b. kdy jsme byly tomu strejkovi na tom when AUX.1PL be.PST.PL that.DAT uncle.DAT on that.LOC pohbu funeral.LOC ‘when we were at that uncle’s funeral’ [PMK441;786749] c. jak se nám rozpadá spolen as RF 1PL.DAT fall.apart.PRES.3SG common.NOM.SG.M stát state.NOM.SG.M ‘as our shared country is splitting [into two]’ [SYNEK;1953603] It has been noted in the external PR research that external PRs tend to cooccur with possessa at the high end of the possessibility hierarchy, with different cut-off points in different languages (Payne and Barshi 1999; Payne 1997a, b). We have now seen that the Czech AP roughly follows the same scale of preferences but it is worth checking this general tendency against the semantic range attested in actual discourse. The sample in (2), (7)–(9),
Plain vs. situated possession in Czech: A constructional account 225
and (13) can be easily extended by additional possessa found with AP in CNK: pytel ‘sack’, hraky ‘toys’, vodovod ‘water-pipe’, svt ‘world’, cesta ‘journey’, sebevdomí ‘self-confidence’, smlouva ‘contract’, ivot ‘life’, svatba ‘wedding’, reforma ‘reform’, králík ‘rabbit’, k ‘horse’, etc. We can see that even this fairly randomly assembled list covers the whole spectrum of the possessibility hierarchy and contains both concrete and abstract entities. At the same time, it must be acknowledged that abstract nouns generally do have limited currency in the AP pattern; it is impossible to either find or construct coherent examples with nouns such as hodnota ‘value’, krása ‘beauty’, kvalita ‘quality’, chyba ‘mistake’, moudrost ‘wisdom’, vmluva ‘excuse’, nadazenost ‘superiority’, while any of these would be perfectly compatible with the GP pattern. On a closer look, however, the possessa that are found in the AP pattern fall into several conceptual categories, which can be properly identified in relation to conventional cultural expectations, rather than in terms of general linguistic categories such as alienability, concrete/abstract, animate/inanimate, etc. The corpus material suggests a number of salient semantic classes of items possessible either individually or collectively by human beings and construable as such within the speakers’ cultural experience. The list in (14) does not necessarily represent a real hierarchy in Czech but it is of course possible to trace certain parallels with the linguistically defined possessibility hierarchy that is more explicitly attested in other languages (cf. Tsunoda 1995): (14) a. things that are part or features of self (body parts; name, title; speech; life; doubt, memory, intention, self-confidence, right to decide, etc.)8 b. members of ‘family’, understood broadly as a culturally established unit of shared domestic life (kinship relations; pets and other domestic animals) c. garments and their parts
8
This group can also be conceptually further decomposed into more specific categories, some of which are also suggested by Tsunoda (1995): parts of body, representations of self, cognitive and speech qualities, manifestations of selfawareness, etc. Such subcategorization, however, leaves the main point of the generalization in (14a) for Czech unaffected, since these additional conceptual distinctions seem to have no grammatical reflexes.
226 Mirjam Fried d. environment perceived as essential to our existence, including dwellings and their parts (world; house, door, plumbing, bathroom; prison cell; backyard) e. objects useful in an individual’s daily life (cars, toys, flashlights, money, tickets, guitar strings) f. common activities and established rituals (journey, wedding, funeral, graduation, education, vacation, holiday) g. social and / or political organization (state/country, constitution, reform) This list makes it evident that there is no blanket prohibition on the usage of abstract nouns. Moreover, the attested combinations cannot be simply replaced with GP without changing their meaning, as we have seen in section 2. It is important to note, though, that the abstract concepts that co-occur with AP tend to come from particular semantic domains and are experientially based: they have to do with mental or physical states and cognitive capacities typical of human beings (14a) or with personal and social rituals (14f). The latter can be easily extended further into more specialized contexts, in which a possessive construal can apply to abstract concepts that are inherent in various types of social institutions in general, such as political organization and public life (14g), and where the PR is, therefore, a collective, not an individual. Granted, the categories in (14f–g) do not represent the same sense of ownership/possession as the classes in (14a–e) and as such are somewhat removed from the central, prototypical definition of possession. Nonetheless, they are not the only source of abstract nouns in the AP pattern and so they do not invalidate the general observation about the possessibility of abstract entities. On the other hand, we also find a number of both abstract and, especially, concrete entities that do not seem to fit easily into any of the categories in (14) at all: branka ‘goal/score [in soccer]’, branká ‘goalie’, pacient ‘[hospital] patient’, tramvaj ‘street car’, devo ‘wood’, smlouva ‘contract’, etc. Taken out of context, these items seem entirely random and might suggest, yet again, that the AP pattern is not that different from GP in the range of possessa it permits. However, ‘context’ is the operative word here; these items are always invoked and interpreted within specialized contexts (e.g., medical care, transportation, commerce), sometimes even in distinct types of discourse (sports reporting), which frame possessive relations in ways specific to those contexts and establish the conditions that allow the PR to be cast as affected in particular ways. No such contextual framing is required for the use of the GP variants. And it is also worth pointing out that
Plain vs. situated possession in Czech: A constructional account 227
these context-sensitive possessive relations are, therefore, different from the possessa in (14f–g): those two categories are relatively independent of special contexts and can be understood as being inherent (and in some ways universal) parts of common human experience. By contrast, the contextdependent possessa are more transient and non-essential. We thus must conclude that the distribution of GP and AP in actual usage is subject to various preferences and sometimes even inviolable constraints with respect to the type of PM. Relevant generalizations can be formulated as follows: (i) body parts cannot occur in the pre-nominal GP pattern and, instead, require AP; (ii) the range of possessa that can naturally co-occur with AP is semantically more restricted than with GP; (iii) the restrictions can be meaningfully articulated only in terms of culturally based clusters of concepts, sometimes in combination with specific types of discourse contexts, not in terms of purely linguistic distinctions, as has been traditionally accepted; and (iv) the distribution of GP is more restricted in terms of information structure. 4.
Constraints on verb semantics and clause structure
4.1. Verb semantics The semantic and pragmatic differences between AP and GP outlined in the preceding sections have different consequences for the types of predicates each form co-occurs with. Predictably, the distribution of GP is entirely independent of any verb meaning, since GP expresses pure possession. By contrast, the dative PR’s involvement in the event depicted by the verb makes AP sensitive to the semantic type of the verb. The affectedness of the PR, which is tied to the affectedness of the PM, presupposes events that are semantically compatible with affectedness. This is, indeed, a correlation well attested with the AP pattern. AP strongly prefers ‘contact’ predicates, such as spravit ‘fix’ (2), pro típnout ‘punch through’ (7a), istit ‘clean’, postílet ‘shoot down’, narovnat ‘straighten’, roztrhnout ‘tear’, pivázat ‘tie to st.’, hodit ‘toss’, umt ‘wash’, as well as other affective verbs, such as komolit ‘distort’ (13a), niit ‘ruin’, osv it ‘refresh’, etc. The corpus material shows that another very well-represented class concerns verbs of removing, such as vzít ‘take away’ in (9), se rat ‘gobble up’, vylouit ‘expel’, krást ‘steal’, ztratit ‘lose’; these verbs presuppose a possessive relation and are thus particularly good candidates for accommodating affected PRs, by expressing possession-removal situations. Among intransitives, verbs of states or spontaneous processes without any identifiable
228 Mirjam Fried instigator are highly compatible with AP semantics and richly attested in the corpus as well: zemít ‘die’ in (4), rozpadat se ‘fall apart’ in (13b), zítit se ‘collapse’, rozbít se ‘break down’, padat ‘fall’, zmizet ‘disappear’, mrznout ‘freeze’, ze ílet ‘go crazy’, cukat ‘twitch’, tást se ‘shiver’, smrdt ‘stink’, etc. Predicates that do not fit these semantic profiles are very hard to come by. The following, showing the syntactically transitive verb vidt ‘see’, is only an apparent counterexample to this generalization: (15) {není to proto, e ti drbani jsou nadáni mimoádnou jasnozivostí a} vidí lidem do kapes see.PRES.3PL people.DAT into pocket.GEN.PL.F ‘{[if a businessman gets robbed…] it’s not because these lowlifes are endowed with extraordinary clairvoyance and} see into people’s pockets’ [LITERA; frybort2] First of all, the use of AP in (15) can be explained by appealing to information structure, just like we saw in (10): the use of the post-nominal GP kapes lidí ‘pockets of people’ (the only GP possibility here due to the plural of the PR noun) would shift the focus away from the pockets, which would make the form contextually problematic. The context clearly centers on what happens to the owners of the pockets; the passage gives explanations about the circumstances under which people get robbed. But equally important is the verb itself: it is not used here in its purely perceptual meaning of ‘having a visual experience’, which would require an accusative-marked perceptum. Instead, the directional phrase do kapes ‘into the pockets’ indicates a more active reading along the lines of ‘look inside’. The sense of pure perception is completely incompatible with AP, as shown in (16); this combination is incomprehensible and is equally impossible even with body parts as PM: (16) *vidí lidem kapsy see.PRES.3PL people.DAT pocket.ACC.PL.F ‘(s)he sees people’s pockets’ The corpus data thus confirm the conclusions of previous studies, namely, that non-affective verbs generally require a GP form. The following examples further illustrate the prohibition on APs with such verbs, in contrast to the attested GP, both with transitive (17) and active intransitive (18) predicates:
Plain vs. situated possession in Czech: A constructional account 229
(17) a. ted’ znám e t její rodie now know.PRES.1SG also 1SG.F.PR parents.ACC ‘now I know her parents too’ [PMK143;519767] b. *ted’ jí znám e t rodie now 3SG.F.DAT know.PRES.1SG also parents.ACC (18) a. {N.P. mla dokonce kurzy v jazykovce zadarmo,} kdy ta její máma tam when that.NOM.SG.F 1SG.F.PR mom.NOM.SG.F there pracovala. work.PST.SG.F ‘{N.P. could even take courses in the language school for free} since her mom was working there’ [oral2006; 175302] b. *kdy jí tam ta máma when 3SG.DAT.F there that.NOM.SG.F mom.NOM.SG.F pracovala. work.PST.SG.F The impossibility of the (b) forms cannot be explained by the semantics of the PM; recall that kinship relations preferentially select AP (examples (4a), (13b)). Instead, it must be attributed to the AP’s semantic incompatibility with predicates that do not allow the PR’s affected role in the depicted event to be properly integrated with the meaning of the predicate. GP, on the other hand, does not place any such requirement on the verb meaning and becomes, therefore, the only available alternative. Notice also that the affectedness must be facilitated by the predicate semantics, not just a potentiality motivated by the inferences suggested by the broad context. This difference is illustrated in (15) and (18) above. In (15), the relevant verb meaning is signaled overtly by the morphology of the second argument. The case in (18), with the intransitive verb pracovat ‘work’, is more subtle. An AP variant (18b) cannot be substituted even though the general context clearly supports the inference that the PR (i.e., the daughter, identified as N.P.) is positively affected by the fact that her mother works in a particular place; N.P.’s benefit is the central concern of this utterance. I will revisit this problem in section 5, to consider potential counterexamples. The co-occurrence patterns attested with AP vs. GP thus lead to two generalizations: GP is fully independent of verb semantics, while AP can only occur with predicates that are semantically compatible with affectedness.
230 Mirjam Fried The latter can be formulated in terms of a hierarchy of preferences as in (19), with a clear cut-off point (indicated by the double arrow >>) between active intransitive and weakly transitive predicates; the latter are fully excluded from the AP pattern: (19) {strongly transitive, non-active intransitive} > active intransitive >> weakly transitive (Vs of perception and cognitition) 4.2. Syntactic constraints on the possessum These verb-related semantic preferences alone cannot account for certain additional facts about the AP pattern; we still need to examine the cases exemplified by (6), here repeated as (20), which involve a particular syntactic constraint on the PM. In (20), the main predicate, vzru ovat ‘cause excitement’, is affective and the PR is a pre-nominal GP, which should not be possible with a body part PM; recall the discussion of (5). Yet, (20a) is the only grammatical way of expressing the possessive relation in this sentence and cannot be replaced by an AP under any circumstances (20b): (20) a. Z njakého […] dvodu ho její from some.GEN reason.GEN 3SG.M.ACC 3SG.F.PR.NOM vlasy nesmírn vzru ovaly. [SYNEK;941353] hair.NOM.PL immensely excite.PST.PL ‘For some [unknown] reason, her hair excited him enormously.’ b. *… vlasy jí ho nesmírn vzru ovaly hair.NOM.PL 3SG.F.DAT 3SG.M.ACC immensely excite.PST.PL Since GP is inside a self-contained NP, we can expect its distribution to be quite free across all syntactic slots. The AP, on the other hand, follows the cross-linguistically commonly attested pattern of excluding the PM of an external PR from certain syntactic functions. In Czech, the PM can be a transitive object (as in (2), (7a), (9a), (13a)), an oblique complement (as in (5a), (8a), (10a), (15)), and an intransitive subject (as in (4a), (13c)). However, we see in (20b) that AP cannot appear as a transitive subject, even if it is a body part. The same prohibition is evidenced in (1a) with a kinship PM, here repeated as (21a): GP is the only way to express the possessive relation here, while AP (21b) is ungrammatical. The same restriction extends to the subjects of active intransitive verbs, as we saw in (18) above.
Plain vs. situated possession in Czech: A constructional account 231
(21) a. Natali-in-y rodie ty maj furt Natalie-PR.F-NOM.PL parents they have.PRES.3PL all.the.time ákej státní svátek, some state holiday.ACC.SG.M [PMK137;148304] ‘Natalie’s parents, they have some kinda state holidays all the time’ b. *Natali-i maj rodie furt ákej Natalie-DAT have.PRES.3PL parents.NOM all.the.time some státní svátek state holiday.ACC.SG.M In order to motivate the syntactic restrictions on the PM, we have to appeal again to the affectedness requirement. The PM cannot appear in event roles that presuppose agentive participants. The conclusion we can draw from the syntactic behavior and the combinability with different predicate types is the following. There are two competing factors for choosing the AP vs. the GP strategy – PR affectedness and the (in)alienability of the PM. Generally, AP preferentially attracts inalienable possessa and for these, it is the only expressive possibility vis-àvis pre-nominal GP forms. However, the (verb-determined) affectedness of the PR evidently takes precedence over inalienability, leading toward absolute prohibition on possessa (of any kind) as transitive subjects and almost equally strong prohibition on subjects of active intransitive verbs. 4.3. Phrasal syntax and the attributive possession We have already noted the readily obvious syntactic difference between GP and AP: GP is so called because the PR is a modifier syntactically dependent on the head noun denoting the PM. The GP is thus a particular (semantic) variant of a more general Modification construction, which is a cluster of syntactic, morphological, and semantic properties defining a particular type of attributive relationship between a noun and its modifier, following a simple phrasal template: [ModPR NPM]NP for the pre-nominal form and [NPM ModPR,GEN]NP for the post-nominal GP. In contrast, it is hard to see on what definition of constituency we could argue that the Czech dative-marked PR forms a single syntactic constituent with the PM. The example in (22) makes it clear that the Czech AP is not an adnominal structure. The corpus example in (22a) shows GP used in an adposition to a noun phrase and (22b) demonstrates that the dative form is prohibited; (22b) is severely ungrammatical:
232 Mirjam Fried (22) a. {Ve védském Stokholmu ije od dtství, kdy tam} jeho maminka, sestra Jiiny, his mom.NOM.SG.F sister.NOM.SG.F Jiina.GEN.SG.F emigrovala. emigrate.PST.SG.F ‘{[He]’s been living in Stockholm, Sweden, since childhood,} when his mom, Jiina’s sister, emigrated.’ [SYNEK;11289608] b. *jeho maminka, sestra Jiin, his mom.NOM.SG.F sister.NOM.SG.F Jiina.DAT.SG.F emigrovala. emigrate.PST.SG.F The PR and the PM have to be analyzed as two autonomous NPs, both of them incorporated as ordinary complements into the structure and the meaning of a sentence. In fact, both the form (dative) and the semantics of affectedness place AP in the family of other dative complements.9 These include ‘thematic’ datives (required by the predicate as one of its arguments, namely one that is in some non-manipulative, perhaps ‘mental’ way affected by the event in which it participates) or adjuncts that are added to verbs that do not require a dative-marked argument but can accommodate an extra participant that can be construed as a beneficiary or a maleficiary of the expressed event. A thematic dative is exemplified in (23) with the verb pomoci ‘help’, showing also the fact that not all of these datives are restricted to animate referents. An adjunct dative is in (24), with the verb otevít ‘open’, and for these datives, the referent is necessarily human; the datives are in bold. (23) a. pomohl mnoha lidem a dtem help.PST.SG.M many.DAT people.DAT and child.DAT.PL ‘he helped lots of people and children’ [SYNEK;10733898] 9
In terms of grammatical roles, the dative complements could be classified as indirect objects, with the understanding that a two-place predicate in Czech can encode its arguments using the pattern [subject – indirect object], e.g. pomoci ‘help’ in (23). Since it is more informative to refer to these complements either through their case form (always dative) or their semantic role status, I choose not to label them in terms of grammatical roles here. The point is that Czech dative NPs are primarily motivated semantically, not syntactically in terms of grammatical roles (cf. also Fried 1994).
Plain vs. situated possession in Czech: A constructional account 233
b. trend, kterému pomohl pedev ím trend.NOM.SG.M which.DAT.SG.M help.PST.SG.M especially {siln píliv zahraniního kapitálu} ‘a trend which was helped particularly {by a strong flow of foreign investment} [SYNEK;8751696] (24) Tento balón otevel lidstvu oblohu. this balloon.NOM.SG.M open.PST.SG.M mankind.DAT.SG sky.ACC.SG.F ‘This balloon opened up the sky for mankind.’ [SYN2000;137864] This connection has been explored in detail elsewhere (Fried 1999a, b) and I will not revisit it here beyond summarizing that the AP shares the special affectedness with other dative-marked roles (experiencer, recipient, beneficiary/maleficiary) and particularly with the usage shown in (24), which I label Dative of Interest (DI). Both DI and AP require the referent to be human and in both cases, the dative NP is an extra element not required by the valence of the verb. However, the datives are incorporated into the sentence as if they were full-fledged arguments. But AP adds a particular feature that is absent in the other dative roles, including DI: the presupposed possessive relation between the referent of the dative and something else in the sentence. This is an important point that bears on the issue of framing possessive relations in terms of broad cognitive schemas, such as the Genitive vs. Goal-based schemas suggested in Heine’s (1997) typology. It is of course true that at some very abstract level, all of the datives, including AP, relate to the concept of goal-ness. But it is also very saliently the case that AP is distinct from other goal-expressing roles and that the distinction is not just a matter of interpretation but has systematic reflexes in semantic constraints and syntactic behavior. In the following section, I will suggest a way of reorganizing the conceptual space of attributive possession so that we can capture more accurately the relationships – commonalities as well as contrasts – between the relevant semantic categories. 5.
Czech GP and AP as distinct functional patterns
5.1. Plain possession vs. situated possession The syntactic behavior only confirms that GP and AP cannot be taken simply as two alternative and fully comparable expressions of attributive possession. To capture the essence of the distinction, we can label the patterns as Plain
234 Mirjam Fried Possession and Situated Possession, respectively. What they both share, necessarily, is a human PR and a pre-existing possessive relation. We could formalize this common background in the form of an interpretive frame, POSSESSION, roughly outlined in (25). The frame represents a conventionally established knowledge structure that schematizes the speakers’ understanding of prototypical ownership as a particular relation between two entities (PR and PM), each of which can be associated with various properties that are motivated by both individual and generally cultural experience. Minimally, the prototypical PR is schematized as a human being that is highly individuated and referentially specific, and the PM as fitting somewhere along the possessibility hierarchy (FE stands for ‘frame element’): (25) frame POSSESSION:
FE#1 Possessor [+human] FE#2 Possessum [on possessibility hierarchy]
For our purposes here, it is not crucial to dwell on all the additional details of this general frame (such as, perhaps, listing a set of preferred properties for each frame element). I will simply take (25) as a minimal way of representing speaker’s conceptual understanding of prototypical possession, which then is shaped into different instantiations by elaborating on the specific characteristics of the PR and/or the PM. Crucially, though, this shaping requires reference to both form and meaning, not just one or the other. The conceptual prototype organized in the background frame does not, by itself, say anything about the morphological or syntactic requirements associated with the morphosyntactic strategies for expressing the possessive relation. Those involve additional layers of constraints and I will argue that the best way to capture the nature of those expressive strategies is to treat each pattern as a grammatical construction in the sense of Construction Grammar, i.e. as a conventionally expected association between the elements of this frame and their linguistic expression. 5.2. Constructional organization of AP and GP properties Let us start by summarizing the properties of the two frame participants when used in the AP pattern, which is clearly the more constrained of the two forms; the items in the list below can be read as being in contrast with the corresponding features of the GP pattern:
Plain vs. situated possession in Czech: A constructional account 235
(26) Situated Possession (AP): Possessor – is a participant in the depicted event (i.e., sensitive to V semantics); – is (indirectly) affected by the event; – is distinct from other datives (semantically, syntactically); – can fully participate in information structure relations. Possessum – comes from a semantically defined and restricted class of items; – must be affected by the depicted event; – is prohibited in certain syntactic functions; – can fully participate in information structure relations. Much of the information summarized in (26) can be attributed to the affectedness requirement. For example, the semantics of the PM is predicted to include entities that can be manipulated and that are inherently relevant to human beings and the routines of their daily existence. It also follows that AP attracts particularly strongly items high on the possessibility hierarchy. At the same time, what we know about possessive relations, dative-marked affectedness, or the possessibility hierarchy does not lead to the prediction that any competition between the preference for inalienable possessa vis-àvis affectedness must be resolved, as a rule, in favor of maintaining a consistent affectedness status (hence the prohibition on possessa as transitive subjects and intransitive agentive subjects). Nor does it account for two additional features: the fact that both the PR and the PM can participate in information structure relations and that the dative form does not cast the PR in a role simply identical with the role of other dative nominals. There are thus properties of the AP pattern that have to be captured in some other way, not just as simple consequences of the PR affectedness. Which brings us to positing AP and GP as complex conventionalized clusters of specific syntactic, morphological, semantic, and pragmatic properties, i.e., as distinct grammatical constructions. Using the Construction Grammar formalism of Fried and Östman (2004), we can represent what speakers have to know about AP as in Figure 1. The inner box represents the head verb, indicating that AP is dependent on certain features of the verb: the verb is specified as coming from a semantic class broadly characterizable as ‘affective’ and as having at least one argument (labeled FE #2) that will be expressed in syntax (as the PM); the latter is indicated by the val(ence) statement. The rest of the information (i.e. everything in the outside, larger box) represents properties that are idiosyncratic to the AP pattern,
236 Mirjam Fried and these have to do with integrating the possessive relation with the semantic and syntactic structure contributed by the head verb. First off, it is the AP construction as a whole, not the head predicate, that supplies the link to the possessive relation, through the statement [frame POSSESSION]), which brings along the PR and the PM and all the background knowledge associated with the representation in (25). Second, the constructional val(ence) statement in the outer box captures the fact that the syntactic and semantic properties of the argument supplied by the head verb are constrained by the requirement that its event role be semantically non-agentive (thus ensuring the prohibition on the PM as a transitive or active intransitive subject). The AP construction does not specify anything about the PM’s semantic or syntactic role (both of these are determined by the semantics of the head verb). However, the co-indexing (#2) between the head verb, the constructional valence, and the POSSESSION frame indicates that whatever this argument is in terms of its semantic and syntactic role in the sentence, it will be interpreted as the PM. Third, the construction also gives the PR an independent information-structure status relative to the PM; this must be specified as the construction’s prag(matic) property. It follows from this feature that both the PR and the PM are subject to articulating regular information structure relations, which operate independently of this particular construction (cf. the discussion in section 2.2); put differently, both the PR and the PM can appear as either a topical or a focal element, independently of each other. The rest of the representation contains features that are shared across AP and DI, as is indicated by the inherit statement at the top; all the features that come from this relationship are printed in gray, to show that these specifications are not unique to the AP construction (strictly speaking, the inherit statement would be sufficient and all the remaining gray-colored information need not be spelled out). These features include the following. First, the construction is syntactically a verb-based pattern (the syn(tax) statement at the top). Second, the construction itself has additional valence requirements, namely, the PR (#1) is in the dative (case DAT) and is interpreted as other datives of interest (expressed through the rel(ation) statement specifying the semantic role, labeled ). And finally, it is the construction as a whole that carries the overall meaning of Situated Possession, spelled out in the sem(antics) statement, as a combination of the inherited DI semantics and the POSSESSION frame.
Plain vs. situated possession in Czech: A constructional account 237 Affected Possessor inherit Dative-of-Interest
syn
[ cat v, lex + ]
prag
[ 'greater discourse independence of Possessor vis--vis Possessum' ]
sem
[ 'circumstances described by the predicate have significant consequences for the interested party (#1), whose referent is not in control of the event' ] frame POSSESSION FE #1 Possessor [human] FE #2 Possessum [ ]
val
{ #1 [ rel [
'interest' ]] , #2 [ rel [
'non-agentive']]}
[ syn [ case DAT ]] syn sem val
[ cat v, lex + ] frame [affective] FE #2[ ] {...#2[ ]...}
Figure 1. Affected PR construction
When we try to represent GP as a construction in its own right, its differences from AP come into sharp relief. To save space, I will focus on the pre-nominal variant only. Figure 2 shows that the GP construction is a subtype of general Modification construction (through an inheritance link). Again, all the inherited information is in gray and is only included for clearer exposition: the GP construction is a phrasal structure of the category NP (indicated by the syn(tax) statement at the top of the constructional box) and consists of two constituents (the inner boxes), with the modifier preceding the head. The information specific to the GP construction concerns two properties. One is the mapping between the POSSESSION frame participants onto the two structural daughters, through the co-indexing: the PR is the modifier, the PM is the head. The other feature is the cat(egory) of the modifier; I have abbreviated this by simply giving a list of morphological classes (enclosed in curly brackets): the filler will be either a possessive pronoun or the special nominal form derived by the possessive suffixes –v and –in. Nothing more needs to be specified.
238 Mirjam Fried
syn
Genitive Possessor inherit Modification
[cat n]
prag ['restrict reference of the noun (#2) by property exrpessed in #1'] sem
frame POSSESSION FE #1 Possessor [human] FE #2 Possessum [ ]
#1 cat [ poss {pro; -UV / -IN }] role modification morphol. case #i [ ] number #j [ ] gender #k [ ]
#2
cat n role head morphol.
case #i [ ] number #j [ ] gender #k [ ]
Figure 2. Genitive PR construction (pre-nominal)
These representations express a hypothesis about the kind of knowledge speakers of Czech must possess in order to produce and interpret a variety of concrete linguistic expressions of attributive possession. One part of the hypothesis is the generalization that the speakers’ native-like understanding of such expressions involves the understanding of a rather intricate interplay between several layers of information, as shown in the figures. But the constructional analysis and representation can enhance also our insight into the way these grammatical patterns may be organized in larger networks of distinct but partially overlapping patterns. Such an organization, in turn, should allow us to be more precise about the circumstances under which the possessive prototype can be extended in various directions. I will suggest such a network and its implications in the next section. 5.3. Constructional network If we take the concept of attributive possession as a type of functional space that can be occupied by various expressions of this general possessive relation, we can organize all the features we have identified as relevant (semantic as well as grammatical) in a network that shows precisely which features are shared across individual patterns and which are specific to each pattern. This is what we see in Figure 3. The shaded area in the middle represents the frame POSSESSION, with its two crucial participants and the minimal constraints on their referents listed under each participant (human PR and the relevance of the possessibility hierarchy). This general possessive relation
Plain vs. situated possession in Czech: A constructional account 239
can be expressed as Plain Posession, in the upper part of the plane, or as Situated Possession, in the lower half; we could think of both as more specialized sub-frames, each of which is conventionally associated with a particular form – the GP and AP grammatical constructions, respectively. The solid-line rectangle in the upper part of the diagram represents the GP construction and its relationship to the background frame: the PR must be in the genitive form(s) and for the PM, we have to note that inalienable possessa are generally incompatible with, or at least strongly dispreferred in, this construction (the symbols ‘<’, ‘>’ indicate the direction of preference). The dashed-line rectangle represents the AP construction, which modulates the background frame by adding a number of features to both the PR and the PM, listed in the columns under each frame participant. With respect to the PM, the most important change consists in delimiting the semantic classes of permissible possessa (in the diagram indicated by reference to the classes indentified in (14) and in noting that the AP construction preferentially selects items from the inalienable end of the possessibility hierarchy, in contrast to GP. The dotted-line, rounded rectangles are included to indicate that both GP and AP constructions overlap in specific ways with other, non-possessive constructions. In the case of the Syntactic Genitive, the common feature with GP is just the Genitive form; the construction does not include any part of the possessive frame and also differs from GP in that the genitive is a participant in the event denoted by the head noun (agent or patient in nominalizations). AP, on the other hand, shares all four characteristics of the PR (animacy, event participant status, affectedness, and dative marking) with the DI construction, which excludes the possessive semantics of the dative nominal and the AP’s constraint on verb semantics. Instead, DI is related to other non-possessive dative constructions (the space labeled Affected datives). I will return in a moment to the significance of the three items that are underlined.10
10
Additional constructions expressing attributive possessive relations (such as adjectival, coming-into-possession patterns based on transfer, etc.) can be incorporated into this network, once their features are properly worked out.
240 Mirjam Fried Synt. genitive
(ex. 12)
event partic. Plain possession (GP)
frame POSSESSION:
Situated possession (AP)
'affective' verbs
GEN/poss. morphol.
inalienable < alienable
[ Possessor,
Possessum ]
human
possessibility hiearchy
DAT event partic. affected
inalienable > alienable 'salient in daily routine' (14)
Dat. of interest Affected datives
affected (ex. 24) (ex. 23)
Figure 3. (Partial) ‘attributive possessive’ network
The network thus captures important facts about the way constructions may interact and the ways in which linguistic expressions may stretch the properties of those constructions. I will now briefly comment on three such examples, all involving the AP construction. As already noted, a systematic account of possession always faces the question of what should count as possession and how inclusive or noninclusive we should be in defining the PR and the PM. Even with the relatively permissive prototype assumed in this paper (compared to Taylor’s), we still have to account for examples such as (27), in which the PM does not easily fit the categories suggested for AP in (14): (27) Ale jednou se nám ztratil jeden but once RF 1PL.DAT get.lost.PST.SG.M one.NOM.SG.M pacient patient.NOM.SG.M ‘But once one of our patients disappeared’ [SYNEK; Hrabal 1993] It is of course hard to argue that there is an ownership relation between the hospital personnel (here the speaker) and the patients. As discussed in section 3, the speaker’s choice to employ AP will be motivated by the discourse context and/or genre: one that sets up the plausibility of an affective
Plain vs. situated possession in Czech: A constructional account 241
possessive reading which involves the relationship between a human being and another entity that is in a contextually salient relation to it. Such cases, then, receive the reading of situated possession by virtue of being used in the AP construction, and we are not forced to relax the possessive prototype itself to account for them. Since the criteria for a plausible PM are evaluated on a sliding scale to begin with, there is perhaps always room for stretching the scale in a motivated way. A more dramatic departure from the possessive prototype is shown in (28), where an inanimate entity seems to be cast in a PR-like relation to another inanimate entity. Since the PR is defined as human (or at least animate), admitting (28) as a case of possession is a much more serious matter. (28) {Pokud by pda poklesla jen o padesát centrimetr,} uhnijí v em stromm koeny. rot.away.PRES.3PL all.DAT tree.DAT .PL.M root.NOM.PL.M ‘{Should the soil sink even by as little as 50 cm,} the trees will lose all their roots to rot’ (lit. ‘the roots of all the trees will rot away on them’) [syn2006publ; Respect 26 /1993] We could declare that (28) is not a case of possession but a simple partwhole relation. The advantage of such an analysis would be that it would preserve the concept of possession as an experiential gestalt, specific to human beings. This way we would also avoid the danger inherent in any prototype-based analysis: in order to account for every new deviation, we could, in principle, keep relaxing the prototype ad infinitum, which then amounts to justifying just about anything as an instance of the same concept, and the prototype loses its coherence as a tool of systematic analysis. However, a categorical exclusion of examples such as (28) leaves unanswered an obvious similarity between them and the AP pattern, both formally (the ‘whole’ being in the dative) and in the overall affective interpretation: when the roots die, the tree is certain to die as well. In order to account for these and similar extensions, the constructional approach offers an alternative that allows us to incorporate the full range of deviations that may arise in actual discourse, while at the same time preserving the possessive prototype as the conventional semantic basis. The AP construction expects the prototype – represented in the frame – to hold, but of course the match between the prototype and the lexical fillers of the constructional slots will not be always perfect, stretching the prototype to varying degrees. In (28), the stretching concerns the semantics of the PR, but at the same time, it is very close to the prototype in two ways: (i) the PM is (construable
242 Mirjam Fried as) inalienable in the same way body parts are and (ii) the mutual relationship between the whole and its part is fully compatible with the affective meaning of the AP construction: the whole is affected because its constitutive part is affected. It is also important to note that GP, shown in (29), does not evoke the situated possession reading but stays purely at the level of a part-whole relationship. In (29), the implication may very well be that the trees will somehow make it anyhow; in any case, (29) is not concerned with the fate of the trees, it is about the fate of the roots only. (29) uhnijí koeny v ech strom rot.away.PRES.3PL root.NOM.PL.M all.GEN tree.GEN.PL.M ‘the roots of all the trees will rot away’ I would argue that this difference is due precisely to the fact that the GP construction is not associated with a special meaning; the form can cover both possessive and non-possessive relations, and consequently cannot impose a possessive reading so easily on combinations that deviate from the prototype in a radical way (such as presenting an inanimate entity as an owner of anything). By contrast, if the semantic and pragmatic properties of the AP construction as a whole (especially the affectedness of the PR) are invoked we have a principled way of explaining what allows the stretch into domains in which we do not have real PRs but only a very specific (and tight) part-whole relationship. In other words, it is the use of the AP construction in encoding a (close) part-whole relation that allows a personification reading, i.e. a conceptualization which mimics a relationship between an animate PR and a body part. We could say that both of these cases (27)–(28) illustrate scenarios in which the AP construction, as a conventional grammatical pattern in its own right, facilitates manipulations of the possessive frame, sometimes also in an interaction with the closely related, but more general, PART-WHOLE frame.11 But the AP construction may also attract other patterns in the network and pull them into an AP reading because of certain shared constructional properties. Here I have in mind the issue of intransitive agents and their potential for compatibility with an AP interpretation, in an apparent 11
I have not provided the details of this frame or its place in the network, mostly for reasons of keeping the representation uncluttered and easily readable, given the focus of this paper (pure possession). But it is obvious that it must be part of a more complete representation of this functional space, particularly in working out the Genitive-related domain.
Plain vs. situated possession in Czech: A constructional account 243
contradiction to the participant hierarchy established in section 4.2. It might seem that cases such as (30), with the active intransitive verb utéct ‘run away’, are simply evidence that Czech does not, after all, restrict the possessa to non-agentive referents and instead excludes only agents/subjects of transitive verbs. (30) va emu tat’kovi utek ten your.DAT dad.DAT run.away.PST.SG.M that.NOM.SG.M keek taky, hamster.NOM.SG.M also {to byl první jeho keek. von byl stra n z toho smutnej. a tak sme mu museli koupit hned dal ího} ‘your dad’s hamster also ran away [in our backyard], {that was his first hamster and he was all so sad about it and so we had to buy a new one for him right away}’ [oral2006;1195644] Such an analysis, though, would be an oversimplification since this patterning, attested very rarely to begin with, is also highly restricted. First, the utterance has to be available for an affective construal, as is the case here; replacing the dative in (30) with GP would be incoherent in the light of the speaker’s subsequent elaboration about what effect the event had on the PR. And as noted in section 4.2, this availability has to be accommodated by the verb meaning. The fact is that only very few verbs appear to work semantically: recall that the quintessential active verb pracovat ‘work’ resists an AP reading (18). A better fit seems to come with verbs of removal (utéct ‘run away’, schovat se ‘hide’), which is not surprising, given the semantics of the AP construction as affecting pre-existing possession. Also verbs of eating or drinking are possible candidates, presumably because the result of such acts has the potential of entering the sphere of interest of people other than the eaters/drinkers.12 And second, the PM (necessarily animate) must be very high on the possessibility hierarchy. Examples such as (31) cannot be interpreted possessively, even though there are two ani12
While I have not run across such an example in the corpus, I can think of usages such as (i). The verb semantics found in these combinations is still waiting for more careful research. (i) Syn se jim opíjel. son.NOM.SG.M RF 3PL.DAT drink.to.excess.PST.SG.M ‘Their son drank heavily [to their worry/embarrassment].’
244 Mirjam Fried mate participants and the action of one (the police) clearly is intended to have consequences for the other (ostensibly mice, figuratively for the inhabitants): (31) {…obklopila na i vilu vozidla Státní bezpenosti a postupovali tak,} aby jim neutekla ani my so.that 3PL.DAT NEG.run.away.PST.SG.F not.even mouse.NOM.SG.F ‘{…cars of the Secret Police surrounded our house and they [=secret agents] proceeded [carefully]} so that not even a mouse [could] escape from them’ (lit. ‘to them’) [syn2006pub;LN 33/1992] We have to conclude that examples such as (30) are more plausibly analyzed as instances of the DI construction, such as exemplified in (24) and (31), in which any kind of verb, including all semantic types of intransitives, can be used, and which expresses a situation with an indirect effect on the ‘interests’ of an animate entity. DI of course overlaps with the AP construction to a great extent; outside of not having the possessive dimension, the DI only differs in that it places absolutely no constraints on the verb semantics. It is not a stretch, then, for a DI token to invite an AP reading, provided that certain features of that token coincide with a particular narrow set of features of the AP construction. This finally brings us to the significance of the three items that are underlined in Figure 3. The features human on the PR and inalienability on the PM are central to the notion of experientially defined possession. We can, therefore, expect that if an inherently non-possessive expression (such as DI) invites a possessive reading, it can be only at the level of the core possessive properties. But satisfying these two features of the possessive frame does not, by itself, guarantee a successful AP interpretation. The PR must also be construable as affected by the event expressed by the verb, which is a central property of the AP construction. This is, of course, related to the lexical meaning of the verb, which is constrained in AP, but that restriction is evidently not as rigid as the requirement that, whatever the verb, the PR must come out as an affected entity. Some verbs with active semantics are inherently better equipped for such a stretch (e.g. verbs of removal) than others (e.g. pracovat ‘work’), and that is what accounts for the relative (un)availability of DI tokens for an AP interpretation. The important point is that all these shifts, whether they involve extensions of the possessive prototype into broader semantic domains or, on the contrary, attracting tokens of non-possessive constructions, can all be explained by appealing to the same cluster of properties (the core features of
Plain vs. situated possession in Czech: A constructional account 245
the possessive prototype associated with AP) and to the AP pattern as a constructional gestalt: the situated possessive meaning of the whole construction is the ‘glue’ that holds all these seemingly disparate uses together. We could thus think of the three underlined items as having a privileged status within this network: they constitute the set of features that are instrumental in various partial shifts, giving rise to and at the same time constraining novel usages. 6. Conclusions Perhaps the central – and inherently thorny – issue in sorting out possessive expressions in any language is the question of how we define the conceptual category to begin with. The present analysis is based on the notion of possessive prototype understood as an experiential gestalt, which takes the PR to be necessarily an animate (human) entity and the PM to be placed somewhere along the possessibility hierarchy, without stating categorically what may or may not count (universally) as a PM. Based on the corpus attestations of the Czech patterns, we can draw at least two conclusions about possessibility. (i) Rather than relying on purely linguistic categories, such as (in)animacy, concreteness, control, etc., possessibility is best defined in terms of culturally determined clusters of concepts and expectations about what is conventionally construed as possessible; the conventional understanding can be then extended to cases where possessive construal is conditioned by the type of discourse or genre. And (ii) different grammatical forms expressing possession may interact with the possessibility hierarchy in different ways. These generalizations are based on a close study of two syntactic patterns that both express possession as a time-stable and presupposed relation. The analysis has established that the patterns are not equivalent semantically or pragmatically and, therefore, cannot be treated simply as structural variants of a single possessive schema. Each pattern encodes a distinct conceptualization of a possessive relation, compatible with different communicative contexts: plain possession is expressed by a Genitive PR (GP), which shares certain features with non-possessive genitives; situated possession – an idiosyncratic combination of possession and affectedness – is expressed by dative-marked Affected PR (AP), which shares certain features with nonpossessive dative-marked roles. It follows that an adequate representation of the speakers’ understanding of these patterns requires reference to several layers of information: semantic and structural limits on the PM (AP), affect-
246 Mirjam Fried edness of the PR (AP), semantic and morphological constraints on the PR (GP), relative discourse prominence of the PR vis-à-vis the PM (AP, GP), verb semantics (AP), and contextual compatibility (AP, GP). The analysis thus makes an argument for taking a Construction Grammar approach as a particularly useful way of framing our understanding of all the relevant issues. First, the clusters of conventionally co-occuring features are naturally captured through the notion of ‘grammatical construction’. Second, the two constructions constitute distinct pieces of a larger network of grammatical entities organized around shared features, both formal and semantic: AP and GP can be shown to occupy partially overlapping domains within the general functional space of attributive possession. And finally, organizing our knowledge about individual constructions in such a network provides us with a more refined map of criteria that can play a role in expressing possessive relations in general. Based on such a map, we can start articulating more systematic hypotheses about the paths along which the prototype might be extended into more peripheral instances; specifically, the extensions can be systematically motivated (and also constrained) by the ways in which the constructions in the network may interact both with each other and with the possessive prototype itself.
Acknowledgements I wish to thank all those colleagues whose questions, at various conference gatherings, have helped me articulate my thoughts about this topic. And I’m especially grateful to Bill McGregor and Elizabeth Traugott for very careful reading of the manuscript and helpful comments on the work. Data esk národní korpus (spoken corpora PMK, BMK, ORAL2006; written corpora SYN2000, SYN2006PUBL). Ústav eského národního korpusu FF UK, Praha. esk národní korpus – A sample of Czech language and literature (SYNEK, LITERA) [CD-ROM]. Ústav eského národního korpusu FF UK, Praha.
Plain vs. situated possession in Czech: A constructional account 247
References Bally, Charles 1995 The expression of concepts of the personal domain and indivisibility in Indo-European languages. In Chappell, Hilary and William B. McGregor (eds.), 31–61. [First puplished in 1926.] Berman, Ruth A. 1982 Dative marking of the affectee role: data from Modern Hebrew. Hebrew Annual Review 6: 35–59. Chappell, Hilary and William B. McGregor (eds.) 1995 The Grammar of Inalienability: A Typological Perspective on Body Part Terms and the Part-Whole Relation. Berlin /New York: Mouton de Gruyter. Croft, William 1985 Indirect Object ‘Lowering’. Berkeley Linguistic Society 11: 39–51. Croft, William 2001 Radical Construction Grammar. Syntactic theory in typological perspective. Oxford: Oxford University Press. Dane , Franti ek 1974 Functional sentence perspective and the organization of the text. In Papers on Functional Sentence Perspective, F. Dane (ed.), 106 –128. Praha: Academia. Fillmore, Charles J. 1982 Frame Semantics. In Linguistics in the Morning Calm, The Linguistic Society of Korea (ed.), 111–137. Seoul: Hanshin. Fillmore, Charles J. 1989 Grammatical Construction Theory and the familiar dichotomies. In Language Processing in Social Context, R. Dietrich and C. F. Graumann (eds.), 17–38. Amsterdam: North-Holland /Elsevier. Firbas, Jan 1966 On defining the theme in functional sentence perspective. Travaux linguistiques de Prague 2: 267–280. Fried, Mirjam 1994 Grammatical functions in case languages. Berkeley Linguistic Society 20: 184 –193. Fried, Mirjam 1999a From interest to ownership: a constructional view of external possessors. In External Possession, Doris L. Payne and Immanuel Barshi (eds.), 473–504. Amsterdam /Philadelphia: John Benjamins. Fried, Mirjam 1999b The ‘free’ datives in Czech as a linking problem. In Annual Workshop on Formal Approaches to Slavic Linguistics 7, K. Dziwirek, H. Coats, and C. Vakareliyska (eds.), 145–166. Ann Arbor: Michigan Slavic Publications.
248 Mirjam Fried Fried, Mirjam 2008 Dative possessor: delimiting a grammatical category based on usage. In Gramatika a Korpus / Grammar and Corpora, F. tícha and M. Fried (eds.), 51–64. Praha: Academia. Fried, Mirjam and Jan-Ola Östman 2004 Construction Grammar: a thumbnail sketch. In Construction Grammar in a Cross-Language Perspective, Mirjam Fried and Jan-Ola Östman (eds.), 11–86. Amsterdam: John Benjamins. Grepl, Miroslav and Petr Karlík 1998 Skladba spisovné e tiny [Syntax of literary Czech]. Praha: Votobia. Heine, Bernd 1997 Possession: Cognitive Sources, Forces, and Grammaticalization. Cambridge: Cambridge University Press. Manoliu-Manea, Maria 1995 Inalienability and topicality in Romanian: pragma-semantics of syntax. In Chappell, Hilary and William B. McGregor (eds.), 711–743. O’Connor, Mary Catherine 1994 The marking of possession in Northern Pomo: Privative opposition and pragmatic inference. BLS 20: 387–401. Payne, Doris L. 1997a The Maasai external possessor construction. In Essays on Language Function and Language Type, Joan Bybee, John Haiman and Sandra A. Thompson (eds.), 395–422. Amsterdam/ Philadelphia: John Benjamins. Payne, Doris L. 1997b Argument structure and locus of affect in the Maasai External Possession Construction. BLS 23 [Special Session on Syntax and Semantics in Africa]: 98–115. Payne, Doris L. and Immanuel Barshi (Eds.) 1999 External Possession. Amsterdam /Philadelphia: John Benjamins. Pit’ha, Petr 1992 Posesivní vztah v e tin. Praha: AVED. Sgall, Petr, Eva Haji ová and Jarmila Panevová 1986 The Meaning of the Sentence in its Semantic and Pragmatic Aspects. D. Reidel Publishing Company. Taylor, John R. 1989 Linguistic Categorization: Prototypes in Linguistic Theory. Oxford: Clarendon Press. Tsunoda, Tasaku 1995 The possession cline in Japanese and other languages. In Hilary Chappell, and William B. McGregor (eds.), 565–630. Zimek, Rudolf 1960 K chápání posesivnosti. Rusko- eské studie, Sborník Vysoké koly pedagogické v Praze, Jazyk a literatura II, 131–156.
Attributive possessive constructions in Oceanic Frantisek Lichtenberk
1. Introduction A feature characteristic of the Oceanic languages 1 is the existence of more than one type of attributive possessive construction. As a rule (there are a few exceptions), Oceanic languages have two basic types of possessive construction, one of which usually has two or more subtypes. Examples (1)–(3) from Manam illustrate. In (1) the possessive suffix, which indexes the PR, is attached to the PM noun, while in (2) and (3) the possessive suffixes are attached to two different possessive classifiers:2 (1)
(2)
1
2
ara-gu name-1SG:POSS ‘my name’
(Lichtenberk 1983a: 282)
pera ana-gu house POSS.CLF-1SG:POSS ‘my house’
(Lichtenberk 1983a: 292)
Oceanic is a subgroup within Austronesian. Oceanic languages are spoken in mainland New Guinea and neighbouring islands, Island Melanesia, Polynesia and Micronesia, but not all of the indigenous languages of New Guinea, Island Melanesia and Micronesia are Oceanic. For a detailed overview of the Oceanic family see Lynch et al. (2002). Besides the Leipzig Glossing Rules, the following abbreviations are used in glossing the examples: CONSTR – construct; NONSG – non-singular; NUM – numeral marker; PC – paucal; REAL – realis; SV – stem vowel; THC – thematic consonant. The glossing conventions are – by and large – those of the sources. In some cases the glosses have been adjusted for the sake of uniformity. The inclusive forms are not considered here a subtype of the first person but a category of its own (Daniel 2005; Lichtenberk 2005a), hence the absence of specification of person. Stress marking has been omitted from the Manam examples. The Toqabaqita data come from my own field notes.
250 Frantisek Lichtenberk (3)
asi ne-gu bushknife POSS.CLF-1SG:POSS ‘my bushknife’
(Lichtenberk 1983a: 401)
As discussed in section 3, the formal differences among the possessive construction types are associated with semantic/pragmatic differences: they signal different types of relation between the PM and the PR, although there are language-specific exceptions. The terms “PM” and “PR” will be used here in two senses: first, they will signify the two constituents in a possessive construction: the PM is the head and the PR its modifier; and second, they will signify their referents. Context will make it clear which of the two senses is intended on a given occasion. The paper is concerned primarily with the semantic/pragmatic aspects of the Oceanic possessive systems. However, to set the stage, some formal aspects will have to be considered first. This is the subject of section 2. Section 3 is concerned with the kinds of semantic/pragmatic distinctions expressed by the different types and subtypes of possessive construction. The phenomenon of fluidity, the ability of one and the same noun to occur in the PM position of more than one type or subtype of possessive construction, will be discussed in section 4. The existence of exceptions to the general patterns will be considered in section 5. Section 6 is concerned with two views of the Oceanic possessive systems, as a noun-class system and as a relational system, and it will be argued there that the Oceanic possessive systems are basically relational in the sense that, by and large, the choice of a possessive construction depends on the nature of the relation between the PM and the PR entities. Section 7 will offer some remarks on the motivation behind the development of a system with multiple types and subtypes of possessive construction, and the last section provides a summary. The languages referred to here and their (approximate) locations are given in Table 1; cf. Map 1.
Attributive possessive constructions in Oceanic Table 1. Languages referred to and their locations Language
Location
’Ala’ala Anejom Araki Bali-Vitu Banoni Cèmuhî Erromangan Fijian (Standard) Gapapaiwa Hawaiian Hoava Houaïlou Iaai Kairiru Kilivila Kokota Kosraean Kwaio Labu Lenakel Lenkau Lolovoli Lou Manam Mussau Nalik Niuean Paamese Pohnpeian Pukapukan Puluwatese Rotuman Tamambo Tobati Toqabaqita Ulithian Vinmavis Wayan Zabana
Papua New Guinea Vanuatu Vanuatu Papua New Guinea Papua New Guinea New Caledonia Vanuatu Fiji Papua New Guinea Hawai’i (Polynesia) Solomon Islands New Caledonia New Caledonia Papua New Guinea Papua New Guinea Solomon Islands Kosrae (Federated States of Micronesia) Solomon Islands Papua New Guinea Vanuatu Papua New Guinea Vanuatu Papua New Guinea Papua New Guinea Papua New Guinea Papua New Guinea Niue (Polynesia) Vanuatu Pohnpei (Federated States of Micronesia) Pukapuka (Polynesia) Puluwat (Federated States of Micronesia) Rotuma (Fiji) Vanuatu Papua (western New Guinea, Indonesia) Solomon Islands Ulithi, Fais (Federated States of Micronesia) Vanuatu Fiji Solomon Islands
251
Map 1.
The boundaries of the Oceanic subgroup. (From: The Oceanic languages, John Lynch, Malcolm Ross and Terry Crowley, 2002, p. 5 [Map 1.2 there], Richmond, Surrey: Curzon. Reproduced by permission of Taylor & Francis Books UK.)
252 Frantisek Lichtenberk
Attributive possessive constructions in Oceanic
2.
253
The formal aspects of Oceanic possessive constructions
With over 400 languages in the Oceanic group, it is impossible to do full justice here to the possessive systems found there, and the discussion will be restricted to the major patterns. 2.1.
The typical Oceanic patterns
2.1.1. Direct and indirect possessive constructions The typical system consists of two basic types of possessive construction, direct and indirect. In the direct type, the PM noun carries a possessive affix encoding or cross-referencing the PR: (4)
Lolovoli hava-da family-NONSG(INCL):POSS ‘our family’
(Hyslop 2001: 169)
See also (1) from Manam in section 1. In the indirect type, it is a possessive classifier 3 rather than the PM noun that carries a possessive affix: (5)
Lolovoli no-da hala POSS.CLF-NONSG(INCL):POSS visitor ‘our visitor’
(Hyslop 2001: 180)
See also (2) and (3) from Manam in section 1. 3
Considering the elements that carry possessive suffixes in indirect possessive constructions to be (possessive) classifiers is the usual approach adopted in recent descriptive work. However, Palmer and Brown (2007) argue that in the Kokota language and possibly some other languages those elements are, in fact, “generic nouns” (p. 208), and that it is these nouns that head possessive constructions. However, this analysis is not without problems. Here, the standard analysis is retained, considering those elements to be classifiers. Since the present study is concerned primarily with the semantic/pragmatic properties of Oceanic possessive constructions, the question of whether in a given language those elements form a morphosyntactic category of their own or whether they are perhaps a subcategory within the category of nouns is not of primary importance here.
254 Frantisek Lichtenberk With very few exceptions, the possessive affixes are suffixes, as in (1)– (5) above. A few languages have possessive prefixes. However, the prefixes normally exist in addition to possessive suffixes and their use is restricted in various ways. For example, in Western Fijian dialects the possessive prefixes are used only in direct possessive constructions and only when the PM–PR relation is other than kinship, such as ‘my blood’ (Geraghty 1983). It should also be noted at this point that the “possessive” affixes have, in some languages, functions other than indexing the PR in a possessive construction. For example, in Toqabaqita the same set of suffixes is used with one class of transitive verbs to index the direct object, and with certain verb-phrase internal particles to index the subject. The PR may be encoded by a noun phrase, in which case the typical pattern is for the PR noun phrase to be cross-referenced by means of a possessive affix either on the PM noun if the possessive construction is of the direct type, or on the possessive classifier if the possessive construction is of the indirect type (see section 2.1.2 for discussion of cross-referencing). Examples (6) and (7) from Hoava illustrate: (6)
sa
bele-na sa boko tail-3SG:POSS ART:SG pig ‘the pig’s tail’
ART:SG
(7)
a-na POSS.CLF-3SG:POSS
(Davis 2003: 98)
napo sa koburu drink ART:SG child
‘the child’s drink’
(Davis 2003: 102)
The ordering of the expressions of the PM, the PR and the possessive classifier varies from language to language: for example, (classifier) PM (PR) (Hoava), (PR) PM (classifier) (Manam), PM (classifier) (PR) (Anejom). 2.1.2. Cross-referencing of the PR With respect to PR phrases that are lexical (rather than pronominal; but see the Vinmavis example (24) in section 2.1.3), three basic types of crossreferencing can be distinguished: full, partial and construct. In full crossreferencing the possessive affix cross-references the PR both for person and for number. Besides singular and plural, many Oceanic languages also have a dual number, and some also have a trial or paucal number. A singular/ plural/dual/paucal system with full cross-referencing is found in (Standard)
Attributive possessive constructions in Oceanic
255
Fijian. This is illustrated in the two pairs of examples in (8)–(11) for the singular and the paucal numbers. In each pair, the same possessive suffix is used whether or not there is a PR phrase present. (8)
na
no-na
ART POSS.CLF-3SG:POSS
vale house
‘his house’ (9)
na
no-na
ART POSS.CLF-3SG:POSS
(Milner 1972: 22) vale na t raga house ART chief
‘the chief’s house’ (10) na
no-dratou
ART POSS.CLF-3PC:POSS
‘their (few) canoe’
(Milner 1972: 23) waqa canoe (Milner 1972: 22)
(11) na
no-dratou waqa na cauravou ART POSS.CLF-3PC:POSS canoe ART young.man ‘the (few) young men’s canoe’ (Milner 1972: 22)
Of the three types of cross-referencing systems, the full cross-referencing system is the most common one in Oceanic. In partial cross-referencing, the PR is cross-referenced only for person, not for number. Partial cross-referencing is found in, for example, Kwaio. When there is no PR noun phrase, there is a singular–plural (and dual) distinction in the third person: (12) ’i’i-na tail-3SG:POSS ‘its tail’
(Keesing 1985: 113)
(13) falai-ga head-3PL:POSS ‘their heads’
(Keesing 1985: 113)
However, when there is a PR noun phrase present, the “singular” possessive suffix must be used even if the PR is plural, as explicitly stated by Keesing (1985: 107): (14) lata-na wela name-3SG:POSS child ‘the child’s name’
(Keesing 1985: 107)
256 Frantisek Lichtenberk (15) lata-na ta’a name-3SG:POSS people ‘the people’s names’
(Keesing 1985: 107)
The system of partial cross-referencing is not common. It is found in a group of closely related languages spoken in the southeast Solomon Islands; but see also further below for Erromangan. In construct cross-referencing, a special “construct” affix is used on the PM noun or on the possessive classifier to cross-reference the PR, but only if there is a PR phrase present. This is illustrated by the following set of examples from Anejom. In (16), without a PR phrase, the possessive classifier carries the third person singular possessive suffix -n, while in (17), with a PR phrase, the possessive classifier carries the construct suffix -i: (16) neto lida-n sugarcane POSS.CLF-3SG:POSS ‘his/her sugarcane’
(Lynch 2000: 60)
(17) nade-n lida-i inhalav breast-3SG:POSS POSS.CLF-CONSTR baby ‘the baby’s breast’
(Lynch 2000: 60)
(In [17] the possessive suffix -n on nade ‘breast’ indexes the baby’s mother, not the baby; the breast is “a source of milk to be sucked by the baby”, hence the possessive classifier lida, “used with nouns whose referents are things one sucks the juice out of, but without consuming the flesh in any way” [Lynch 2000: 60].) The same construct suffix is used when the PR noun phrase is plural. elpu-Uje (18) injap um a-i sea POSS.CLF-CONSTR PL-Uje ‘the Uje people’s sea’
(Lynch 2000: 61)
The construct suffix is also used on the PM noun in direct possessive constructions: (19) risi-i di? mother-CONSTR who? ‘whose mother?’
(Lynch 2000: 58)
Construct cross-referencing systems are common in languages of Vanuatu and Micronesia. It should be noted, however, that the term “construct” pos-
Attributive possessive constructions in Oceanic
257
sessive affix is interpreted in different ways by different analysts. Thus, for example, Crowley (1998) uses the term to refer to a possessive suffix in Erromangan that functions as the third person singular possessive suffix when there is no PR phrase, as in (20) below, but which is also used to index the PR when there is a PR phrase, regardless of the grammatical number, as in (21) and (22). This is comparable to the system referred to further above as partial cross-referencing with respect to Kwaio. (20) retpo-n wife-3SG:POSS ‘his wife’
(Crowley 1998: 52)
(21) nompu-n natmonuc head-CONTR/3SG:POSS chief ‘the chief’s head’
(Crowley 1998: 172)
(22) nompu-n ovatmonuc head-CONSTR/3SG:POSS PL:chief ‘the chiefs’ heads’
(Crowley 1998: 172)
In the present study, the term “construct” possessive affix is restricted to cases either where the form of the affix is unique, unlike that of any of the other possessive affixes, as in Anejom ((17)–(19) above), or where the suffix need not index any of the features of the PR phrase, for which see (24) from Vinmavis in section 2.1.3. Finally, in at least one language PR phrases are not cross-referenced at all, even though possessive affixes are used on PM nouns and on possessive classifiers in the absence of a PR phrase. This is the case in Cèmuhî (Rivierre 1980). 2.1.3. Type of PR phrase The type of possessive construction required may depend on the type of PR phrase. In a number of languages possessive constructions with pronominal PR phrases exhibit idiosyncratic properties, although this need not apply to pronouns of all numbers and persons. For example, Vinmavis has a possessive suffix that functions specifically to index third person singular PRs, but it also functions as a construct suffix with all the other pronouns as PRs:
258 Frantisek Lichtenberk (23) netal-n leg-3SG:POSS ‘his/her leg’
(Crowley 2002: 642)
(24) netal-n get leg-CONSTR 1PL(INCL) ‘our legs’
(Crowley 2002: 643)
The same suffix is also used with lexical PR phrases: (25) netal-n matoro leg-CONSTR old.man ‘the old man’s leg’
(Crowley 2002: 642)
Similar use of the third person singular possessive suffix as a construct suffix with plural independent pronouns as PR phrases is found in Lou and Lenkau (Ross 1988: 332). In some languages independent personal pronouns cannot form a PR phrase. The PR can be indexed only by means of a possessive affix. BaliVitu4 is one such language: example (26a) without a pronominal PR phrase is grammatical, while (26b) with a pronominal PR is not: (26) a. a
lima-ma hand-2SG:POSS ‘your hand’
ART
b. *a
lima-ma oho hand-2SG:POSS 2SG (‘your hand’) (Ross 2002a: 370)
ART
In a few languages there are differences in the structure of possessive constructions depending on whether the PR noun phrase is common or proper. In Iaai, a common-noun PR in a direct possessive construction is crossreferenced on the PM noun by means of a possessive suffix, but there is no cross-referencing of proper-noun PRs (Ozanne-Rivierre 1976). And in Fijian, constructions with singular proper-noun PRs are different from those used with common-noun and plural proper-noun PRs (Milner 1972). This concludes the survey of the typical pattern of Oceanic possessives constructions with a distinction between direct and indirect constructions, with relatively minor variations on the basic pattern in some languages. 4
Ross (2002a) calls the language “Bali-Vitu”, but van den Berg and Bachet say in a grammar of Vitu that “Ross’s sketch is primarily a description of the Bali variety” (2006: 2).
Attributive possessive constructions in Oceanic
259
Next we will briefly consider some possessive systems that are different from the typical pattern in more significant respects. 2.2. The Polynesian pattern In the Polynesian languages the direct-indirect system of possessive constructions has been replaced by a different binary pattern. This is also, to some degree, true of the Rotuman language, a sister of the Polynesian group. For convenience, I will refer to the pattern discussed in what follows as the “Polynesian pattern”. In the Polynesian pattern, the direct possessive construction type is absent, apart from some lexical exceptions. Instead, there is a binary system that is customarily referred to in Polynesian linguistics as A-possession and O-possession on the basis of the vowels found in the two sets of possessive marking. Discussions of possessive constructions in several Polynesian languages (and in Rotuman) can be found in Fischer (2000), together with an overview of the Polynesian pattern by Clark (2000). The semantic/pragmatic aspects of the Polynesian pattern are briefly discussed in section 3.4 below. Examples (27) and (28) from Hawaiian illustrate A-possession and Opossession, respectively: (27) ka
pahi a Kimo knife POSS Kimo ‘Kimo’s knife’
ART
hale o Kimo house POSS Kimo ‘Kimo’s house’
(Wilson 1976: 29)
(28) ka
ART
(Wilson 1976: 29)
Within the basic A-O constrast, each Polynesian language has a number of subtypes of possessive construction. In the Hawaiian examples above, the possessive markers are formally prepositions. In another type of construction, the possessive markers are fused with an article; see (29) and (30) from Pukapukan: (29) t-a-ku tama ART-POSS-1SG:POSS child ‘my child’
(Salisbury 2002: 172)
(30) t-o-ku vaka ART-POSS-1SG:POSS canoe ‘my canoe’
(Salisbury 2002: 171)
260 Frantisek Lichtenberk The article t- signifies definiteness and singular number of the PM. Another subtype of possessive construction is usually referred to as “irrealis possession”: “the intention or anticipation that something will be possessed” (Clark 2000: 262). The possessive marker is added to an element m-, which, according to Clark, continues an earlier irrealis or optative marker. The variety of possessive constructions even in a single Polynesian language can be quite considerable, and the discussion above is no more than basic. The collection of articles edited by Fischer (2000) is a useful reference where more detail on some of the Polynesian languages can be found.
2.3. Absence of the indirect system of possessive classifiers In the Polynesian languages it is the direct possessive type that is absent. There are also languages where it is the indirect type that is absent. This is the case in Toqabaqita and its close relatives. The possessive construction in (31) corresponds to the direct type in other languages, but for reasons to become clear presently, it can also be referred to as suffixing, because the PR is cross-referenced on the PM noun by means of a possessive suffix: (31) qaba-na wela hand-3SG:POSS child ‘the child’s hand(s)’ In the other type, there is no possessive classifier and no indexing of the PR on the PM noun. This type of construction can be referred to as bare. (32) fanga wela food child ‘the child’s food’ The semantic/pragmatic aspects of Toqabaqita possessive constructions are discussed in sections 3.4 and 5. A possessive system with a contrast between a direct/suffixing construction and a bare construction is also found in Kairiru (Papua New Guinea) (Wivell 1981). For some remarks on Kairiru see section 5 below.
Attributive possessive constructions in Oceanic
261
2.4. Some other types of possessive system There are some other types of possessive system found in Oceanic, but as these will not figure in the discussion of the semantics/pragmatics of possessive constructions, they are mentioned here only very briefly. In some languages prepositions are used in possessive constructions, the complement of the preposition being the PR noun phrase. Discussion of possessive prepositions in some Oceanic languages can be found in Hooper (1985). The usual case is for possessive prepositions to be part of a larger system of possessive constructions. Possessive prepositions are found in Polynesian languages (see section 2.2 for Hawaiian), where they are part of the overall A–O possessive contrast and have a classificatory function, not unlike the possessive classifiers found in the typical Oceanic pattern. Finally, there are a few Oceanic languages that have only one basic type of possessive construction: there is no contrast comparable to those between direct and indirect, suffixing and bare, or A and O types of construction. Labu is one such language (Siegel 1984), and Tobati is another (Donohue 2002). Niuean, a Polynesian language, also has basically just one type of possessive construction, although, according to Massam and Sperlich (2000), traces of the Polynesian A–O possessive contrast do exist there. 3.
The semantics/pragmatics of possessive systems with multiple types of construction
3.1. Introduction We can now turn our attention to the vast majority of Oceanic languages that have possessive systems of multiple construction types and subtypes, and consider their use. Is their use governed strictly lexically, each noun having to be specified for which type of possessive construction it occurs in the PM position? Or are there some general patterns that determine the use of the various types? Once again, given the large number of languages involved, the answer is not simple. There are clear patterns, but there are also exceptions to these patterns. In this section and in section 4, the focus will be on the patterns; the existence of exceptions will be considered in section 5. The basic, overall pattern is that the choice of a possessive construction depends on the relation between the referents of the PM and the PR phrases. Since the notion of relation is central to the understanding of how the systems of Oceanic possessive constructions (normally) operate, some discussion of the notion is in order.
262 Frantisek Lichtenberk 3.2. Different kinds of the notion of relation with respect to possessive constructions There are (at least) three different types of the notions of “relation” and “relational” that are relevant to the present discussion. One is the idea that possessive constructions are relational. There are two entities which stand or are put in a certain relation to each other, one as PM and the other as PR. As is well known, the number of possible relations expressed by possessive constructions, while not open-ended, is quite large. Besides ownership (PR owns PM: ‘my money’, the money that belongs to me), some other relations are: use (PR uses PM without necessarily owning it: ‘my bus’, the bus I will take), control (PR has control over PM without owning it or using it: ‘my office’, the office I am in charge of), manufacture (PM is made by PR: ‘my cake’, the cake I baked), kinship (PM is a kin of PR’s: ‘my sister’), part of a whole (PM is part of PR: ‘my head’, the head which is part of my own body), and many others (see, for example, Langacker 1995: 56–57). This I take to be an uncontroversial sense of the notion of relation. Another sense of the terms “relation” and “relational” has to do with the fact that certain concepts are inherently relational. Nouns that express inherently relational concepts are sometimes referred to as “relational nouns”, such as mother (see, for example, Barker 1995; Partee 1997; Partee and Borschev 2003). When inherently relational nouns occur in the PM position in a possessive construction, the type of relation usually expressed involves inalienable possession, as in my mother. When a noun that is not inherently relational occurs in the PM position, the type of relation usually expressed involves alienable possession, as in my knife. Relational nouns tend to strongly favour a certain kind of relation between the PM and the PR. The relation is intrinsic to the meaning of a relational noun. Barker (1995) also uses the term “lexical possession” for this. Thus with my mother (with the core meaning of mother) there is an intrinsic relation of kinship. With nonrelational nouns, on the other hand, there is typically no intrinsic relation between the PM and the PR, and a variety of relations are freely available. Thus with my knife the relation may be one of ownership (the knife I own), use (the knife I use without owning it), manufacture (the knife I made), etc. In such cases, the relations can be said to be “extrinsic” (Barker 1995). The notions of inalienable and alienable possession and intrinsic and extrinsic relations will be relevant in later discussion. I take the notions of inherently relational concepts and relational nouns also to be uncontroversial. It is the third sense of “relation(al)” that has enjoyed some controversy in Oceanic linguistics with respect to possessive constructions. This is the
Attributive possessive constructions in Oceanic
263
idea that the choice of a possessive construction is determined by, or sensitive to, the nature of the relation between a PM entity and its PR. In principle, it is possible to recognize two views concerning the choice of possessive constructions in Oceanic, which will be referred to as “noun-class based” and “relation-based”, respectively. A useful discussion of the two views can be found in Pawley and Sayaba (1990), and I will consider their views in more detail in section 6. On the noun-class based view, the nouns of a language with multiple types of possessive construction fall into a number of classes depending on which of the possessive constructions they select when occurring in the PM position. For example, Milner (1972: 65) speaks of “gender” in Fijian, “a grammatical category of four classes”, which he terms “neutral”, “edible”, “drinkable” and “familiar”: a given noun (or “base” in Milner’s system of morphosyntactic categories) belongs in a certain class depending on which type of possessive constructions it selects. At the same time, however, Milner makes it clear that the classes have a semantic underpinning. For example, of the “edible” class he says (p. 66) that it includes nouns that denote “[a]rticles of solid food considered from the point of view of consumption (i.e. as distinguished from planting, selling, etc.)”. Furthermore, he notes that there are nouns that belong in more than one class, or in our terminology that they exhibit fluidity (section 4). François (2002) says about the Araki language that it has two basic categories, inalienable nouns and alienable nouns, although his classification has also partly to do with whether a given noun takes a possessive suffix (in a direct possessive construction) or not. However, François also notes that the assignment of nouns to classes is not always straightforward and that “most nouns can almost freely shift from one pattern to the other” (François 2002: 48). And for Gapapaiwa, McGuckin (2002: 303) identifies three classes of nouns, “each with its own set of possession markers”, but then goes on to say (p. 304) that “[t]hese possession classes do not correspond to fixed noun classes, as the same noun can occur in more than one possessive category”. A common theme that emerges from those studies that postulate noun classes on the basis of possessive constructions is that the languages in question exhibit fluidity, the ability of a noun to occur in the PM position of more than one type of possessive construction. On the relation-based view, the choice of a possessive construction depends on the kind of relation that holds between the PM and the PR. (For discussions of the relation-based view see Pawley 1973; Pawley and Sayaba 1990; Lynch 1973, 1982; Lichtenberk 1983b, 1985.) For example, if the PM is part of the PR’s own body, the direct possessive construction is (typically) used; if the PM is an article of food for the PR, an indirect possessive
264 Frantisek Lichtenberk construction with a certain possessive classifier is used; if the PM is an item of drink for the PR, an indirect possessive construction with a different possessive classifier is used; and so on. (It is for this reason that the possessive classifiers are termed “relational classifiers” in Lichtenberk 1983b.) Some of the detailed studies of possessive constructions in individual languages that adopt the relation-based view do, however, mention the existence of exceptions. We can now consider the semantic/pragmatic properties of the different types of possessive construction.
3.3. Direct possessive constructions In direct possessive constructions, the PM noun carries an affix that indexes the PR. Direct possessive constructions are overwhelmingly used to express inalienable possession, where the PM noun is inherently relational. This does not mean, however, that all types of inalienable possession are expressed by means of the direct construction (see section 5). There are several subtypes of inalienable possession that are normally expressed by the direct construction. These are discussed in A–I below. A. Parts of a whole, body parts. Included here are also concepts such as body and integral contents of a PR, such as blood (in the PR’s body) and juice (e.g. of fruit): Paamese (33) vati-n head-3SG:POSS ‘his/her head’
(Crowley 1996: 389)
Toqabaqita (34) suul-a fa qota juice-3SG:POSS CLF areca.nut ‘juice of an areca nut (being chewed)’ (fa is a “numeral” classifier used in noun phrases referring to fruit and certain other entities, not a possessive classifier)5
5
As (34) shows, the use of the classifier fa is not dependent on the presence of a numeral in the noun phrase.
Attributive possessive constructions in Oceanic
265
B. Natural bodily products that emanate from the PR’s body, and other products of physical bodies: for example, ‘tears’, ‘sweat’, ‘urine’, ‘faeces’, ‘semen’, ‘voice, sound’ (produced by the PR), ‘breath’, ‘smell/scent’ (exuded by the PR); and also ‘shadow, shade’ (cast by the PR), ‘reflection (of the PR, e.g. in water), ‘picture’ or some other representation of the PR, all of which are often part of a polysemy: Manam (35) boro tae-di pig faeces-3PL:POSS ‘pigs’ excrements’
(Lichtenberk 1983a: 279)
Tamambo (36) nunu-ku photo/reflection/picture/shadow-1SG:POSS ‘my photo/reflection/picture/shadow’ (a likeness of me) (Jauncey 1997: 229) C. Entities, matter on the surface of the PR’s body. Included here are concepts such as sores, dirt, tattoes, clothing (especially, though not necessarily, when being worn by the PR), and parasites such as lice: Banoni (37) kipi-na-i moono dirt-3SG:POSS-ART girl ‘the girl’s dirt’ Lolovoli (38) tatai-ne tattoo-3SG:POSS ‘her tattooes’ D.
(Lynch and Ross 2002: 445, from Lincoln 1976)
(Hyslop 2001: 171)
Mental organs, states, and products of mental processes:
Kilivila (39) nano-gu mind-1SG:POSS ‘my mind’
(Senft 1986: 45)
266 Frantisek Lichtenberk Lolovoli (40) domi-mu thought-2SG:POSS ‘your thoughts’ E.
(Hyslop 2001: 172)
Attributes such as the PR’s shape, size, name and age:
Nalik (41) a nounau-naande ART shape-3PL:POSS ‘their shape’ (also ‘their interest in something’)
(Volker 1998: 130)
Puluwatese (42) yii-e-mw age-SV-2SG:POSS ‘your age’
(Elbert 1972: 283)
F. Spatial and temporal relations. The possessive affixes are added to spatial (and temporal) prepositions or to what is sometimes referred to as relator nouns. Such prepositions and relator nouns usually derive historically from nouns that designate spatial aspects of objects, especially human bodies, and certain of their parts (such as the face or the back) (Bowden 1992). Toqabaqita (43) qi ninima-ku at beside-1SG:POSS ‘beside me’ Lolovoli (44) Lo tagu-i bongi gai-vesi …. LOC behind-CONSTR day NUM-four ‘After four days ….’
(Hyslop 2001: 176)
G. Kinship categories and certain other categories of social/cultural relations. The latter categories include concepts such as ‘friend’ and ‘partner’ (e.g. trading partner). Kilivila (45) ina-si mother-3PL:POSS ‘their mother’
(Senft 1986: 140)
Attributive possessive constructions in Oceanic
Iaai (46) ihumwi- friend-3SG:POSS ‘his/her friend’
267
(Ozanne-Rivierre 1976: 157)
H. The PR is a Patient, Theme or Stimulus (of emotion or of sensory perception). In a number of languages the direct construction is used when the PR has one of these roles in a transitive situation: Patient or Theme acted on by Agent, or Stimulus perceived by Experiencer. Often in such cases the PM phrase is, or contains, a nominalization of the corresponding verb. Such possession is sometimes referred to as “passive” (see, for example, Lynch 2001), as opposed to “active” possession (section 4 below). Manam (47) udi tanom-a-di banana plant-NMLZ-3PL:POSS ‘the planting of the bananas’
(Lichtenberk 1983a: 251)
Toqabaqita (48) riki-la-na wane baa look.at-NMLZ-3SG:POSS man that ‘that man’s appearance/look/mien’ (i.e., the way the man appears to be to others, how others see him) (lit.: ‘that man’s looking-at’) The PM need not be a nominalization; nevertheless, the PR has the role of a Patient, Theme or Stimulus in the associated event: Kokota (49) mereseni-na mheke medicine-3SG:POSS dog ‘medicine for dogs’
(Palmer 2002: 506)
In some other languages passive possession is expressed by means of an indirect possessive construction, using a classifier (see [53] in section 3.4). I. Emphatic pronominal forms. Such forms often carry the significance of ‘by oneself’, ‘on one’s own’: Paamese (50) Inau nakanian s so-k. 1SG 1SG:REAL:eat self-1SG:POSS ‘I ate by myself.’
(Crowley 1996: 407)
268 Frantisek Lichtenberk Manam (51) Rube-gu u-yalale. alone-1SG:POSS 1S:REAL-go ‘I went alone, by myself.’
(Lichtenberk 1983a: 307)
3.4. Classifier systems (indirect possessive constructions) As also mentioned in section 7, it is generally assumed that Proto Oceanic had (at least) three formally distinct possessive classifiers, and we can take the tripartite system as our starting point. Some languages have a tripartite system; in some, the number of classifiers has been reduced to two, or one, or none; and in some others it has been expanded beyond three, sometimes considerably. (Standard) Fijian has three possessive classifiers: a “food / passive” classifier ke-, used when the PM is an item of food for the PR (but see below) or another entity metonymically related to food (such as places where food is grown, and containers of food for the PR to eat), and it also signals passive possession, where the PR is a Patient, Theme or Stimulus, rather than an Agent or Experiencer in the associated situation (see category H in section 3.3, and examples (85)–(88) and the accompanying discussion in section 4); a “drink” classifier me-, used when the PM is an item of drink for the PR (but see below) or another entity metonymically related to drink (such as containers of drink for the PR to drink); and a “general” classifier no-/ne-, used when none of the other classifiers nor the direct construction are called for. Note that the food category subsumes only solid food; food that is runny, mushy, juicy, suckable is included in the drink category (see Pawley and Sayaba 1990 for Wayan, one of the Fijian languages). However, tobacco (for smoking) is in the food category. Examples (52) and (53) illustrate food and passive possession, respectively: (52) na
ke-da
vei-niu group-coconut ‘our coconut plantation (which we eat from)’
ART POSS.CLF-PL(INCL):POSS
(53) ke-mu i-roba6 POSS.CLF-2SG:POSS NMLZ-slap ‘your slap (you are slapped)’ 6
Geraghty (1983) calls the i- prefix a “preformative”.
(Pawley 1973: 162)
(Geraghty 1983: 249)
Attributive possessive constructions in Oceanic
269
Example (54) illustrates the drink classifier and (55) the general one: (54) na
me-mun
t
ART POSS.CLF-2PL:POSS tea ‘your tea’
(55) na
no-mu
ART POSS.CLF-2SG:POSS
(Milner 1972: 66)
waqa canoe
‘your canoe’
(Milner 1972: 65)
In Manam the number of possessive classifiers has been reduced to two. This is due to the merger of the food and the drink categories into one “alimentary” category. The classifier ana- (also an- and anan-) is used when the PM is an item of food or drink for the PR or an entity metonymically related to such, for example, gardens where food is grown, implements used to obtain or eat food, containers of food and drink for the PR (including personal baskets used to hold tobacco, areca nuts, lime and betel pepper for chewing, and bottles). The other classifier is a general one, used in all cases other than those that call for the alimentary one or for the direct construction. Examples (56) and (57) illustrate the alimentary classifier, and (58) the general one: (56) ulu ana-mi breadfruit POSS.CLF-2PL:POSS ‘your breadfruit (to eat)’
(Lichtenberk 1983a: 291)
(57) botoli ana-gu bottle POSS.CLF-1SG:POSS ‘my bottle’ (drink container)
(Lichtenberk 1983a: 293)
(58) tamoata asi ne-ø man bushknife POSS.CLF-3SG:POSS ‘the man’s bushknife’
(Lichtenberk 1983a: 294)
There are languages that have only one possessive classifier. In one respect, such forms are unlike possessive classifiers in other languages, because they do not contrast with other classifiers. Nevertheless, the term “classifier” is retained here for two reasons. First, systems with single classifiers are historical reductions of systems with multiple classifiers. And second, constructions with a simple classifier do contrast with direct possessive constructions, without a classifier.
270 Frantisek Lichtenberk Houaïlou is one language with a contrast between a direct possessive construction and a single indirect one. The sole possessive classifier is used whenever the direct construction is not appropriate (and vice versa). It is used even with some (but not all) kinship terms. Examples (59) and (60) illustrate the classifier: (59) d v i-vu 7 garden POSS.CLF-1DU(EXCL):POSS ‘our garden’ (60) p vaa i- a father POSS.CLF-1SG:POSS ‘my father’
(Leenhardt 1932: 192)
(La Fontinelle 1976: 300)
A possessive system with an opposition between a direct construction and an indirect construction with only one possessive classifier is also found in ’Ala’ala (Ross 2002b). A sole possessive classifier contrasts only with the direct possessive construction, and so, strictly speaking, is not necessary, provided the contrast is made in some other way. This is the case in Toqabaqita, which has no possessive classifiers whatsoever. The contrast is between a direct/suffixing construction, where the PM noun carries a possessive suffix, and a bare construction, where there is no indexing of the PR on the PM noun. In some of its aspects the bare construction is analogous to the system of indirect constructions in other languages (but see section 5). (61) fanga nia food 3SG ‘his/her food’ (62) biqu wane baa house man that ‘that man’s house’ (Nia in [61] is the third person singular independent pronoun.) A different kind of reduction in the number of possessive classifiers has taken place in the Polynesian languages, with their binary systems of Apossession and O-possession (and no direct possessive construction). The 7
Leenhardt (1932) gives the form as dov xivu; the representation d v i-vu is in accordance with La Fontinelle (1976).
Attributive possessive constructions in Oceanic
271
details of the use of the two constructions vary from language to language. Wilson (1982) has put forward an “Initial Control Theory” to account for the uses of A- and O- possession in Polynesian: the choice of a possessive construction is determined by whether or not the PR has control over the initiation of the relation. If the PR does have control, A-possession is used; if the PR does not have control, O-possession is used. This is illustrated for Hawaiian in the next pair of examples. In (63) the PR initiates the relation by having the child, and so A-possession is used. On the other hand, in (64) the PR does not initiate the relation to his parent, and so O-possession is used. (63) k- -na keiki ART-POSS.CLF-3SG:POSS child ‘his child’
(Wilson 1982: 19)
(64) k-o-na makua ART-POSS.CLF-3SG:POSS parent ‘his parent’
(Wilson 1982: 19)
We can now turn to languages in which the number of possessive classifiers is greater than the three-member system with a food–drink–general contrast. Moderate expansions are found in some of the languages of Vanuatu. Lolovoli has a food classifier, a drink classifier, a general classifier and a classifier for “natural or valued object possession” (Hyslop 2001: 176). The latter classifier, whose form is bula, is mainly used to express ownership of animals and crops: (65) bula-na boe POSS.CLF-3SG:POSS pig ‘his pig’
(Hyslop 2001: 178)
A more complex classifier system exists in Lenakel, which has a food classifier, a drink classifier, a plant classifier (for plants planted by the PR), a general classifier, and a classifier for locations (occupied by the PR). The latter classifier is only optional, and the general one may be used instead. The location classifier is illustrated in (66): (66) tn iimwa-nil-lau land POSS.CLF-3NONSG-DU ‘their homeland’
(Lynch 1978: 81)
272 Frantisek Lichtenberk In some languages the original classifier system has undergone great expansion. This is the case in most Micronesian languages. (Lee 1975: 111) gives a list of 19 “commonly used classifiers” for Kosraean. The list includes classifiers for transportation; land and shelters; plants; tools, pets and toys; drink; several classifiers for food; several classifiers for kinship relations; and several classifiers for decorations. And there is a general classifier. Example (67) contains the tool/pet/toy classifier: (67) mos nuhti-k breadfruit POSS.CLF-1SG:POSS ‘my breadfruit for toy’
(Lee 1975: 117)
For Pohnpeian, Rehg (1981) gives a list of 21 classifiers, which is not exhaustive. In fact, Rehg (1981: 179) says that “how many [possessive classifiers] there are in Ponapean [Pohnpeian] is difficult to determine”. Quite a few of the classifiers in the Micronesian languages are transparently related to nouns, and one finds cases of repeaters (Aikhenvald 2000), where a classifier is used with a noun from which it has developed through grammaticalization. The classifier and the source noun have identical or very similar forms. For example, Ulithian has a vehicle classifier of the form waa, which is used when the PM serves as a means of transportation, such as a ship, a bicycle, a plane or a canoe, for the PR. The word for ‘canoe’ too is waa: (68) waa-yire POSS.CLF-3PL:POSS ‘their canoe’
waa canoe (Sohn and Bender 1973: 268)
Outside of Micronesia a large system of possessive classifiers (including repeaters) is found in Iaai (New Caledonia). Besides a general classifier, there are classifiers for food, drink, chewable food, game killed or caught in hunting or fishing, voice and sounds, land and various products, boats, and so on. According to Ozanne-Rivierre (1976: 189), the set of Iaai possessive classifiers is open, and a comprehensive list is difficult, if not impossible, to establish. Mussau (Papua New Guinea) too has a relatively large set of possessive classifiers. Ross (2002c) lists nine of them, but the list is not necessarily exhaustive.
Attributive possessive constructions in Oceanic
273
4. Fluidity in the possessive systems A feature of Oceanic possessive systems regularly commented on in grammars is the fact that some nouns can occur in the PM position of more than one type of possessive construction. The phenomenon is sometime referred to as “overlap” (following Lynch 1973). Here the term “fluidity” will be used (pace Nichols [1992]). How fluid a possessive system is varies from language to language, but fluidity is by no means uncommon. There is fluidity between the direct possessive construction type and an indirect/classifier construction type, and there is fluidity between different indirect/classifier constructions. Sometimes the fluidity has to do with different senses of a noun. Such cases are illustrated first, starting with fluidity between the direct and the indirect construction types. In (69), from Tamambo, nunu has, among others, the sense of ‘pictorial representation of (the PR)’, while in (70) it has the sense of ‘object that carries a pictorial representation of something (not necessarily of the PR)’: (69) nunu-ku photo/reflection/picture/shadow-1SG:POSS ‘my photo/reflection/picture/shadow’ (a likeness of me) (Jauncey 1997: 229) (70) no-ku nunu POSS.CLF-1SG:POSS photo/picture ‘my photo(s)/picture(s) that belong(s) to me’
(Jauncey 1997: 229)
The different senses of a noun may also call for the use of different possessive classifiers, as in (71) and (72) from Araki. In (71), where the sense of the PM noun is ‘pig’, it is the “economic possession” classifier pula- that is used, while in (72), where the sense of the same PM noun is ‘pork’, it is the food classifier ha- that is used: (71) pula-ku po POSS.CLF-1SG:POSS pig ‘my pig (I breed)’
(François 2002: 100)
(72) ha-ku po POSS.CLF-1SG:POSS pig ‘my piece of pork (to eat)’
(François 2002: 100)
274 Frantisek Lichtenberk And in (73) and (74) from Lolovoli there is a contrast between the food and the drink possessive constructions, depending on whether the PM noun has the sense of citrus fruit to eat or citrus juice to drink: (73) ga-ku moli POSS.CLF-1SG:POSS citrus ‘my orange/pomelo (for me to eat)’
(Hyslop 2001: 185)
(74) me-ku moli POSS.CLF-1SG:POSS citrus ‘my orange/pomelo juice’
(Hyslop 2001: 185)
However, there is fluidity even when polysemy is not involved. In some cases, the use of different possessive constructions has to do with different referents or different kinds of referents. Examples (75) and (76) from Fijian illustrate. In both cases the PM noun has the sense of mango as fruit, but when a mango is green, unripe, it is “eaten” and the food classifier is called for, but when a mango is ripe and juicy, it is “sucked” when being consumed, and the drink classifier is called for: (75) na
ke-na maqo ART POSS.CLF-3SG:POSS mango ‘his mango for eating (i.e. green mango)’
(Pawley 1973: 168)
(76) na
me-na maqo ART POSS.CLF-3SG:POSS mango ‘his mango for sucking (i.e. ripe, juicy mango)’
(Pawley 1973: 168)
And in (77)–(79) from Manam there is a three way contrast with the noun ‘head’ in the PM position, and in each case the sense of the noun is that of a body part. However, in (77) the head is part of the PR’s own body and so the direct construction is used; in (78) the head is food for the PR and so the alimentary classifier construction is used; and in (79) the head is neither part of the PR’s own body nor food for him/her, and it is the general classifier construction that is used: (77) paana-gu head-1SG:POSS ‘my head (part of my body)’
(Lichtenberk 1983a: 302)
Attributive possessive constructions in Oceanic
(78) paana ana-gu head POSS.CLF-1SG:POSS ‘my head to eat (e.g. a fish head)’
275
(Lichtenberk 1983a: 302)
(79) paana ne-gu head POSS.CLF-1SG:POSS ‘my head (e.g. a head I found, cut off, or a head I will give my dog to eat)’ (Lichtenberk 1983a: 302) Fluidity is found even if the PM referent is one and the same entity but is conceptualized differently. As shown in (75) and (76) above from Fijian, the noun ‘mango’ can occur in the PM position with the food or the drink classifier. However, one and the same mango can serve a purpose other than being for the PR’s consumption, for example as something to be sold, in which case the general possessive classifier is required, and the distinction between a mango as food and a mango as “drink” disappears: (80) na
no-na
maqo mango ‘his mango (as property, e.g., which he is selling)’ (Pawley 1973: 168)
ART POSS.CLF-3SG:POSS
In Manam the noun for ‘grass-skirt’ occurs as PM in the direct construction when a grass-skirt is being worn by the PR at the relevant time, but when the same grass-skirt is not being worn, it is the indirect construction with the general classifier that is called for: (81) baligo-gu grass.skirt-1SG:POSS ‘my grass-skirt (when I am wearing it)’
(Lichtenberk 1983a: 301)
(82) baligo ne-gu grass.skirt POSS.CLF-1SG:POSS ‘my grass-skirt (when I am not wearing it)’ (Lichtenberk 1983a: 301) In the examples just given, the PR is (or may be) one and the same person. There is a different kind of fluidity, where one and the same entity is referred to by means of different possessive constructions because of different perspectives due to different PRs. For example, one and the same woman may be one person’s wife and another person’s sister. In Kosraean, two different classifiers are used:
276 Frantisek Lichtenberk (83) muhtwacn kiyuh-k woman POSS.CLF-1SG:POSS ‘my wife’
(Lee 1975: 118)
(84) muhtwacn wiyuh-k woman POSS.CLF-1SG:POSS ‘my sister’
(Lee 1975: 118)
Kosraean, which has a large set of possessive classifiers (section 3.4), has several kinship classifiers. Example (83) contains the classifier for mothers and wives, and (84) one of the classifiers for siblings. Fluidity having to do with PR perspective is common in the opposition between passive and active possession. In passive possession the PR is a Patient, Theme or Stimulus in the relevant situation (section 3.3), and correspondingly in active possession the PR is an Agent or Experiencer. One and the same state of affairs can be encoded from the perspective of the Patient/Theme/Stimulus or that of the Agent/Experiencer. This is the case in (85) and (86) from Fijian, with the food/passive and the general classifiers, respectively: (85) ke-mu i-vacu POSS.CLF-2SG:POSS NMLZ-punch ‘your punch (you receive)’
(Schütz 1985: 462)
(86) no-mu i-vacu POSS.CLF-2SG:POSS NMLZ-punch ‘your punch (you give)’
(Schütz 1985: 462)
In the Fijian examples the formal contrast is between two indirect constructions. In (87) and (88) from Manam the contrast is between a direct construction and an indirect construction. The PM in the direct construction is a nominalization of the verb nanari-t-a ‘tell a story about’. (87) nanari-t-a-a-gu tell.story-THC-TRANS-NMLZ-1SG:POSS ‘my story (story about me)’
(Lichtenberk 1983a: 303)
(88) nanari ne-gu story POSS.CLF-1SG:POSS ‘my story (e.g. one that I invented, told or like)’ (Lichtenberk 1983a: 303)
Attributive possessive constructions in Oceanic
277
5. Exceptions in the use of possessive constructions As discussed in sections 3 and 4, there are clear semantically/pragmatically based patterns in the use of the possessive constructions. However, descriptions of some languages, especially more detailed descriptions, also mention cases where the choice of a possessive construction for a given PM does not follow the general patterns in that language. (Exceptions, or apparent exceptions, to the use of classifiers are not unusual in classifier systems in general; see the quote from Aikhenvald 2000: 82 in section 6.) Some of these cases are genuine exceptions, in the sense that there is no explanation available for their existence. In other cases, however, the exceptions are systematic, and their existence can be accounted for. In the discussion that follows, the focus will be on cases that involve distinctions between the direct and the indirect possessive constructions, specifically cases where the direct construction would be expected but is not used. The cases will be divided into two broad categories, one having to do with kinship terms, and one having to do with body-part terms and terms for concepts having close association with the body. In Anejom some kinship terms, such as ‘father’, ‘grandparent’ and ‘wife’, occur in the direct construction, while others occur in one of two indirect constructions: for example, ‘husband’ takes the general classifier, and ‘sister’ and ‘brother’ take the passive-possession classifier. Lynch (2000) does not attempt to provide an explanation for the use of the indirect constructions. In Houaïlou the term for ‘mother’ occurs in the direct construction, but the term for ‘father’ occurs in the only type of indirect construction. La Fontinelle (1976) offers no explanatory comment. On the other hand, in some languages the use of more than one type of possessive construction is said to be (at least partially) pragmatically motivated. In Kairiru, of the 18 kinship terms recorded by Wivell (1981) one half occurs in the direct/suffixing construction and the other half in another, bare construction, where the PM carries no indexing of the PR. The latter construction is also used to express alienable possession. Wivell (1981: 54) characterizes the distinction thus: “Those terms that were inalienable were the ones that expressed important kin relations within the social structure, while the alienable ones were those that expressed fairly unimportant roles.” The former category includes the terms for ‘father’, ‘mother’ and ‘child’, and the latter category the terms for ‘mother’s brother’s wife, husband’s sister’s daughter’ and ‘great grandparent, great grandchild’. In Gapapaiwa, according to McGuckin (2002: 304), “[k]in terms for persons who are peers or subordinates are possessed in direct constructions,
278 Frantisek Lichtenberk while terms for persons in authority over ego require an indirect alienable construction”, for example, ‘spouse’ and ‘mother’, respectively. In some languages the exceptional treatment of some kinship terms is due to lexical replacement. In such cases, terms that are relatively new to the language do not (necessarily) select the direct construction. For example, in Toqabaqita the term for ‘mother’ occurs in the direct construction but the term for ‘father’ does not: (89) thaina-ku mother-1SG:POSS ‘my mother’ (90) maka nau father 1SG ‘my father’ Thaina continues Proto Oceanic *tina ‘mother’,8 while maka is a lexical innovation. There is an archaic term for ‘father’, thaama, which continues Proto Oceanic *tama ‘father’, and it occurs in the direct construction. Similarly, the term for ‘child’, wela, is a lexical replacement (cf. Proto Oceanic *natu) and occurs in the bare/non-suffixing construction, not in the direct / suffixing construction. In Wayan there are two types of possessive construction in which kinship terms occur as PMs, and, according to Pawley and Sayaba (1990: 158), “[t]here are no clear semantic grounds for the split”. However, Pawley and Sayaba point out that one of the two structures is an innovation unique to Western Fijian. It was originally used for part-of-whole terms, some of which later acquired kinship meanings. It is this class that is open, admitting new kinship terms, while the other class is closed. Exceptions are also found in some languages with body-part terms and terms having to do with the body. In a detailed study of inalienable possession in Paamese, Crowley (1996) concludes that the choice of a possessive construction is not fully predictable on semantic grounds. Most nouns that refer to internal organs occur in an indirect construction, and Crowley (1996: 398) offers an explanation: “Internal organs are the kind of things that would normally only be directly observed when there is a dead body that has been opened up. The possession of these items by the butcherer of the animal is clearly transferrable and so there is an alienable relationship 8
The Proto Oceanic reconstructions are from Lynch et al. (2002).
Attributive possessive constructions in Oceanic
279
between him and the body parts, in contrast to the formerly inalienable relationship between the animal and the body parts when it was still alive.” Note, however, that the indirect construction is also used when the internal organ is part of the PR’s, for example, an animal’s, own body, and so the implication in what Crowley says is that the use of the indirect construction expressing the alienable nature of the relation between an animal’s internal organ and the butcherer has been extended to the relation when the organ is part of the PR’s own body. Be that as it may, the fact remains that the noun that refers to a bird’s crop occurs in the direct construction, not the indirect one, and there is no obvious explanation for that. Outside of the area of internal organs, there is no explanation for the fact that the noun for ‘shin’ occurs in the indirect construction although the noun for ‘lower leg’ occurs in the expected direct construction. A different kind of exceptional behaviour of body-part terms is found in Toqabaqita. Such nouns occur in the direct/suffixing construction unless they are in the scope of a modifier other than the PR, such as a verb,9 a numeral or a demonstrative. Compare (91), where the noun ‘eye’ occurs in the direct/ suffixing construction, and (92) and (93), where the same noun occurs in the bare/non-suffixing construction: (91) maa-ku eye-1SG.POSS ‘my eye(s)’ (92) maa mauli nau eye be.on.left.side 1SG ‘my left eye’ (93) maa nau naqi eye 1SG this ‘this eye of mine’ The bare/non-suffixing construction is also used to express alienable possession; see (61) and (62) in section 3.4. There is no difference in inalienability between ‘my eye(s)’ on the one hand and ‘my left eye’ and ‘this eye of mine’ on the other. As argued in Lichtenberk (2005b), the relevant factor is that of PM individuation. When the PM is not in the scope of a modifier other than the PR, it is not individu9
In Toqabaqita, verbs can directly modify nouns.
280 Frantisek Lichtenberk ated vis-à-vis the PR; it is viewed basically as an aspect of the PR. On the other hand, through specification the PM is individuated, given more identity with respect to the PR, and there the same construction is used that serves to express alienable possession, where the PM is always individuated with respect to the PR. 6. The relational basis of the Oceanic possessive systems In a detailed study of the system of attributive possessive constructions in Wayan (a Western Fijian language) Pawley and Sayaba (1990) ask the following question: with respect to possessive marking, is the system one of noun classes or is it relational? The conclusion they reach is that it is a mixture of both: “Certain nouns belong to strict and semi-arbitrary noun classes, for purposes of possessive-marking, others show marking consistently following semantic principles.” (p. 168). They suggest that their findings apply in their basics also to (Standard) Fijian. And the conclusions they reach with respect to Wayan and Fijian may be of relevance to Oceanic in general. In the present study, on the other hand, it has been argued that the Oceanic systems are, on the whole, based on semantic/pragmatic principles, the crucial factor being the nature of the relation between the PM and the PR, even though there are also genuine exceptions. There is then some arbitrariness, but such cases are exceptions against the backdrop of semantically/pragmatically motivated systems. While certain of Pawley and Sayaba’s conclusions concerning Wayan are applicable to other Oceanic languages, there is some clarification that is in order. They use the term “relational analysis” (or “hypothesis”) to identify an approach to Oceanic possessive constructions in which the choice of a construction type is viewed as having to do with the nature of the relation between the PM and the PR. (They mention the existence of several variants of the relational analysis.) They contrast the relational analysis with a nounclass analysis, according to which each noun belongs to a certain class depending on the type of possessive construction it selects, and characterize it in this fashion: We take the relational hypothesis to entail not only the claim that: (i) possessive marking is determined by the semantic relation holding between possessed and possessor, but that (ii) this relation is not constant for all situations. That is, for any noun, speakers have some choice of possessive marker, constrained only by their imaginations or belief systems. Any constraints which can not be readily accounted for in these terms must be regarded as grammatical not semantic constraints. (Pawley and Sayaba 1990: 169)
Attributive possessive constructions in Oceanic
281
In the present study, the Oceanic possessive systems have been characterized as basically relational in nature: exceptions apart, the choice of the type of possessive construction depends on the nature of the PM–PR relation. This corresponds to point (i) in the quote from Pawley and Sayaba. Their point (ii) has to do with fluidity. Fluidity is, of course, evidence of the relational nature of the system: one and the same noun occurs in the PM position of different types of possessive construction depending on the nature of the relation between the PM and the PR. Pawley and Sayaba say that according to the relational hypothesis “for any noun, speakers have some choice of possessive marker, constrained only by their imaginations or belief systems” (see the quote above; emphasis added here). I am not aware of any such strong version of the relational analysis, according to which any noun in a given language exhibits, in principle, fluidity. (See also further below on fluidity and rigidity.) Importantly, Pawley and Sayaba’s point (i) is independent of their point (ii). Fluidity is not necessary for a system to be relational. Even if no noun exhibited fluidity, where each noun occurred in only one type of possessive construction and so could be said to belong in a certain noun class, the overall system could still be relational if each noun class was defined by the nature of the relation between the PM and the PR. An unstated assumption in Pawley and Sayaba’s approach is that noun classes are essentially arbitrary. This, however, is not the case. As Aikhenvald (2000: 21) points out: “There is always some semantic basis to the grouping of nouns into classes, but languages vary in how much semantic transparency there is.” And: “In languages with purely semantic assignment the class of a noun can be inferred from its meaning.” (p. 22). Thus, noun classes are semantically motivated to various degrees, even though in a noun-class system “[e]ach noun … belongs to one (or occasionally more than one) class(es)” (Aikhenvald 2000: 21). Classifier systems (for example, numeral-classifier systems) too are based on semantic/pragmatic principles. Sometimes there is fluidity, but there are also exceptions: “The choice of a classifier is usually semantically transparent; in some cases, however, the semantic link between a noun classifier and a noun is not obvious.” (Aikhenvald 2000: 82). It is not lack of fluidity that counts as an exception. Only those nouns that require a possessive construction that goes against the general pattern are exceptional. Such exceptions apart, the Oceanic possessive systems are basically relation-based, regardless of the degree of fluidity they permit. One more remark on fluidity is in order. Descriptions of Oceanic languages that do comment on fluidity (and most of them do), say explicitly or imply that certain nouns do not have fluidity, that they can occur only in one
282 Frantisek Lichtenberk type of possessive construction. The question that needs to be asked is to what extent such lack of fluidity is lexical/grammatical and to what extent it is semantic/pragmatic. It may be that certain kinds of fluidity do not occur because the kind of semantic/pragmatic relation that a given type of possessive construction expresses would be highly implausible or even absurd for a certain pairing of a PM and its PR. Thus, it is unlikely that the noun for ‘father’ would occur in the PM position in the food or the drink possessive construction. But is this a grammatical constraint or a matter of pragmatics? On the other hand, less than one-hundred-percent fluidity does not in itself mean that (genuine exceptions apart) the possessive system in a given language does not have a relational basis. At present, we have mostly only brief comments on creativity with respect to possessive fluidity in individual languages; what is needed is indepth studies. It may turn out that there is more fluidity than meets the eye. 7. Emergence of the possessive classifier system It is generally agreed that Proto Oceanic had a direct possessive construction and (at least) three possessive classifiers used in indirect constructions: a food, a drink and a general one (Lichtenberk 1985, Lynch et al. 2002). Lynch et al. (2002) suggest that Proto Oceanic may have had other possessive classifiers, including another, large-member set of classifiers, not unlike those found in most of the Micronesian languages, Iaai and Mussau (section 3.4). However, although such large-member systems are found in different primary subgroups of Oceanic, it is quite likely that those are later, independent developments from the more restricted, probably tripartite, system. This kind of development is quite natural. Possessive classifiers may have begun to develop shortly before the Proto Oceanic stage, but the historical evidence is unequivocal that earlier there had been only one basic type of possessive construction, which was of the direct/suffixing type. The following question then arises: why did a system of possessive classifiers develop for alienable possession but not for inalienable possession? I have discussed this issue elsewhere (Lichtenberk 2005b) and so only a brief synopsis will be given here, together with new, albeit indirect evidence. In inalienable possession there is typically a highly salient relation between a PM and its PR: a kinship relation, a part-whole relation, or some other kind of intrinsic relation (for example, Barker 1995). The interpretation of the relation as the salient one is quite stable across different contexts (although it can be overridden). In alienable possession,
Attributive possessive constructions in Oceanic
283
on the other hand, there is often no such strongly salient, context-stable relation, and correspondingly the referents of one and the same PM phrase can stand in various relations to their PRs. This was illustrated in the discussion of fluidity in section 4, in particular by the “mango” examples (75), (76) and (80) from Fijian: a mango as an item of food, or as an item of “drink” or as property to be sold. In alienable possession, the interpretation of the relation between a PM and its PR is highly variable across contexts. Possessive classifiers specify more closely the nature of the PM–PR relation. The development of possessive classifiers for alienable possession is well motivated because of the variability in the PM–PR relation. On the other hand, in inalienable possession there is no such strong motivation because of the presence of a highly salient, default relation between the PM and the PR. The default interpretation of linguistic constructions needs no overt marking (cf. Haiman 1985; Dixon 1994; Croft 2001 [2002]). On the other hand, there is motivation for there being an overt marker of a nondefault interpretation. For possessives, this is illustrated by the next pair of examples from Manam. The default interpretation of ‘X’s skin’ is for the skin to be part of the PR’s own body, as in (94), where the direct construction is used. However, if the intended interpretation is not the default one, this is signalled by means of a classifier, as in (95), in this case the alimentary classifier: (94) usi-gu skin-1SG:POSS ‘my skin (the skin of my body)’
Lichtenberk, field notes
(95) usi ana-gu skin POSS.CLF-1SG:POSS ‘my skin (for me to eat, e.g. chicken skin)’
Lichtenberk, field notes
To test the hypothesis that there is a difference in the presence of a highly salient, context-stable interpretation in inalienable possession, with intrinsically relational nouns, on the one hand, and in alienable possession with nouns that are not intrinsically relational, on the other, a set of experiments was performed with native speakers of English (Lichtenberk et al. 2004). Early in the history of Austronesian there was only one type of attributive possessive construction, without a distinction between inalienable and alienable possession. Similarly, English does not have a formal distinction between inalienable and alineable possession in its attributive possessive constructions, and so provides a good testing ground for the hypothesis.
284 Frantisek Lichtenberk In the experiment the subjects were presented with several sets of possessive noun phrases, some of which had inherently relational nouns as PMs, for example his children, and others had nouns as PMs that are not inherently relational, such as her cookies. For each stimulus the subjects were asked to give one interpretation of the relation between the PM and the PR. The results of the study convincingly demonstrated the existence both of a PM effect and of a PR effect. The PM effect has to do with the fact that the relational nouns as PMs elicited a restricted range of interpretations of the PM–PR relations, while the non-relational nouns elicited a broader range of interpretations. Furthermore, with the relational nouns there was always one interpretation that was clearly dominant, while such strong dominance was not found with the non-relational nouns. Thus, the interpretation of his children was in terms of a kinship relation, while her cookies elicited a variety of interpretations: the cookies she owns, the cookies she made, the cookies she bought, the cookies she will eat. The PR effect has to do with the fact that the nature of the PR had a greater effect on the interpretation of the PM–PR relation with PMs that are not inherently relational, while with the inherently relational nouns as PMs the interpretations were quite stable. Thus, for example, both for the soldier’s legs and for the general’s legs the interpretation was uniformly that of the legs being part of the soldier’s or the general’s own body. On the other hand, for the soldier’s regiment the dominant interpretation was that of the regiment the soldier is a member of, while for the general’s regiment the dominant interpretation was that of the regiment the general is in charge of. The PR effect is stronger with non-relational nouns than with relational nouns because with non-relational nouns there is typically no intrinsic, salient relation between the PM and the PR. It is true that a certain kind of PR is likely to favour a certain kind of interpretation, but such contextual factors are absent or attenuated with “neutral” PRs, such as possessive determiners, for example, her cookies.10 Although the study was done on English, the assumption is that very much the same cognitive factors operate in other languages and that they operated in the history of Austronesian, when the system of possessive classifiers began to develop. Obviously, a system of possessive classifiers does not have to develop (after all, they are not common in the languages of 10
William B. McGregor has suggested (pers. comm., 11 August 2007) that the gender of the PR might be relevant in some cases; cf. her cookies and his cookies. The study discussed here did not take gender into account.
Attributive possessive constructions in Oceanic
285
the world), but if they do develop they are more likely to develop, at least initially, for alienable possession, with PM nouns that are not inherently relational, than for inalienable possession, with PM nouns that are inherently relational. (Some Micronesian languages do have possessive classifiers for different kinship categories, as mentioned and illustrated for Kosraean in section 4 above), but those languages have large sets of possessive classifiers, which are later developments, postdating the emergence of the first, restricted set.)
8. Summary and conclusion With very few exceptions, Oceanic languages have more than one type of attributive possessive construction. In the typical system, there is a distinction between a direct possessive construction, where the PM noun carries affixes that index the PR, and more than one subtype of indirect construction, where the possessive affixes are attached to a possessive classifier. The direct construction type is strongly associated with inalienable possession, where there is an intrinsic link between the PM and the PR. There are, however, also language specific exceptions where certain PMs take an indirect construction rather than the direct one. With some exceptions, the choice of a possessive construction depends on the nature of the relation between the PM and the PR. This relational nature of the choice of possessive construction is particularly strongly evidenced by fluidity in the possessive systems. Sometimes such fluidity is due to the polysemy of a noun, but in some cases there is fluidity without polysemy, and the choice of a possessive construction depends on the pragmatics of the situation (for example, whether or not an article of clothing is being worn at the relevant time). When there is no fluidity, one can say that the given noun belongs in a certain class (the classes being established on the basis of the type of possessive construction used), but such classes are themselves by and large semantically/pragmatically grounded in the nature of the PM–PR relation. To say that the Oceanic possessive systems are relational in nature and semantically/pragmatically motivated does not mean that the choice of a possessive construction is always predictable, even disregarding genuine exceptions. As pointed out in section 3.4, the type of possessive construction that a noun selects may be based on metonymy. Thus, in Manam the noun for ‘garden’ selects the alimentary (food and drink) possessive classifier, because gardens are places where food is grown (Lichtenberk 1983a).
286 Frantisek Lichtenberk However, not in all languages does the noun ‘garden’ select the food or alimentary classifier. For example, in Anejom it selects the general classifier or the customary-possession classifier, even though the language has a food classifier (Lynch 2000); and in Zabana it selects the general classifier, even though the language has an alimentary classifier (Fitzsimons 1989). Metonymy is language/culture specific. A system may be sematically/pragmatically motivated, but that does not mean that everything is predictable. In Oceanic, possessive classifiers are used (again with some exceptions) to express alienable possession. The development of possessive classifiers for alienable possession was motivated by the fact that with nouns that are not inherently relational there is typically no highly salient, context-stable, relation between the PM and the PR. A classifier specifies more closely the type of the relation. There are specific classifiers, such as food and drink. There is also a general classifier (provided a language has more than one classifier), which only signifies that the relation is not any one of the more specific types. In inalienable possession, where the PM noun is intrinsically relational, there is normally a highly salient, context-stable kind of relation between the PM and the PR, and the development of classifiers there is much less motivated. While the development of a system of possessive classifiers for alienable possession was motivated, it is also a fact that the original system of (at least) three classifiers has been simplified in some languages or has disappeared altogether. While cognitive factors may motivate the existence of a grammatical construction or contrast, they do not determine their existence, and such factors may be overridden by other kinds of development. Acknowledgement I am grateful to William B. McGregor for detailed comments on an earlier version of this paper. References Aikhenvald, Alexandra Y. 2000 Classifiers: A Typology of Noun Categorization Devices. Oxford: Oxford University Press. Barker, Chris 1995 Possessive Descriptions. Stanford: CSLI Publications.
Attributive possessive constructions in Oceanic
287
Benthem, Johan van and Alice ter Meulen 1997 Handbook of Logic and Language. Amsterdam: Elsevier; and Cambridge, MA: MIT Press. Bowden, John 1992 Behind the Preposition: Grammaticalisation of Locatives in Oceanic Languages. Canberra: Pacific Linguistics. Chappell, Hilary and William B. McGregor (eds.). 1996 The Grammar of Inalienability: A Typological Perspective on Body Part Terms and the Part-Whole Relation. Berlin /New York: Mouton de Gruyter. Clark, Ross 2000 Possessive markers in Polynesian languages. In Steven R. Fischer (ed.), 258–268. Croft, William 2001 Radical Construction Grammar: Syntactic Theory in Typological Perspective. [Reprinted in 2002 with corrections]. Oxford: Oxford University Press. Crowley, Terry 1996 Inalienable possession in Paamese grammar. In Hilary Chappell and William B. McGregor (eds.), 383–432. Crowley, Terry 1998 An Erromangan (Sye) Grammar. Honolulu: University of Hawai‘i Press. Crowley, Terry 2002 Vinmavis. In John Lynch, Malcolm Ross and Terry Crowley (eds.), 638–649. Daniel, Michael 2005 Understanding inclusives. In Elena Filimonova (ed.), 3–48. Davidson, Jeremy H.C.S. (ed.). 1990 Pacific Island Languages: Essays in Honour of G. B. Milner. London: School of Oriental and African Studies, University of London; and Honolulu: University of Hawai‘i Press. Davis, Karen 2003 A Grammar of the Hoava Language, Western Solomons. Canberra: Pacific Linguistics. Dixon, R. M. W. 1994 Ergativity. Cambridge: Cambridge University Press. Donohue, Mark 2002 Tobati. In John Lynch, Malcolm Ross and Terry Crowley (eds.), 186– 203. Elbert, Samuel H. 1972 Puluwat Dictionary. Canberra: Pacific Linguistics. Filimonova, Elena (ed.). 2005 Clusivity: Typology and Case Studies of the Inclusive–Exclusive Distinction. Amsterdam /Philadelphia: John Benjamins.
288 Frantisek Lichtenberk Fischer, Steven R. (ed.). 2000 Possessive Markers in Central Pacific Languages. Special issue of Sprachtypologie und Universalienforschung, Language Typology and Universals, 53. Fitzsimons, Matthew 1989 Zabana: a grammar of a Solomon Islands language. Unpublished MA thesis, University of Auckland. Frajzyngier, Zygmunt, Adam Hodges and David S. Rood (eds.) 2005 Linguistic Diversity and Language Theories. Amsterdam /Philadelphia: John Benjamins. François, Alexander 2002 Araki: A Disappearing Language of Vanuatu. Canberra: Pacific Linguistics. Geraghty, Paul A. 1983 The History of the Fijian Languages. Honolulu: University of Hawai‘i Press. Haiman, John 1985 Natural Syntax: Iconicity and Erosion. Cambridge: Cambridge University Press. Halim, Amran, Lois Carrington and S. A. Wurm (eds.) 1982 Papers from the Third International Conference on Austronesian Linguistics, Vol. 1: Currents in Oceanic. Canberra: Pacific Linguistics. Hooper, Robin 1985 Proto-Oceanic *qi. In Andrew Pawley and Lois Carrington (eds.), 141–167. Hyslop, Catriona 2001 The Lolovoli Dialect of the North-East Ambae Language, Vanuatu. Canberra: Pacific Linguistics. Jauncey, Dorothy 1997 A grammar of Tamambo, the language of western Malo, Vanuatu. Unpublished PhD thesis, Australian National University. Keesing, Roger M. 1985 Kwaio Grammar. Canberra: Pacific Linguistics. La Fontinelle, Jacqueline de. 1976 La Langue de Houaïlou (Nouvelle-Calédonie): Description Phonologique et Description Syntaxique. Paris: Société d’Études Linguistiques et Anthropologiques de France. Lang, Ewald, Claudia Maienborn and Cathrine Fabricius-Hansen (eds.). 2003 Modifying Adjuncts. Berlin/New York: Mouton de Gruyter. Langacker, Ronald W. 1995 Possession and possessive constructions. In John R. Taylor and Robert E. MacLaury (eds.), 51–79. Lee, Kee-dong (with the assistance of Lyndon Cornelius and Elmer Asher) 1975 Kusaiean Reference Grammar. Honolulu: University Press of Hawai‘i.
Attributive possessive constructions in Oceanic
289
Leenhardt, Maurice 1932 Documents Néo-Calédoniens. Paris: Institut d’Ethnologie, Université de Paris. Lichtenberk, Frantisek 1983a A Grammar of Manam. Honolulu: University of Hawai‘i Press. Lichtenberk, Frantisek 1983b Relational classifiers. Lingua 60: 147–176. Lichtenberk, Frantisek 1985 Possessive constructions in Oceanic languages and in Proto-Oceanic. In Andrew Pawley and Lois Carrington (eds.), 93 –140. Lichtenberk, Frantisek 2005a Inclusive–exclusive in Austronesian: an opposition of unequals. In Elena Filimonova (ed.), 261–289. Lichtenberk, Frantisek 2005b Inalienability and possessum individuation. In Zygmunt Frajzyngier, Adam Hodges and David S. Rood (eds.), 339–362. Lichtenberk, Frantisek, Jyotsna Vaid and Hsin-Chin Chen 2004 Alienable/inalienable possession in grammar and in the lexicon: evidence from possessive phrase interpretation in English. Paper presented at the 4th International Conference on the Mental Lexicon, University of Windsor, 30 June – 3 July 2004. Lincoln, Peter C. 1976 Describing Banoni, an Austronesian language of Southwest Bougainville. Unpublished PhD thesis, University of Hawai‘i. Lynch, John 1973 Verbal aspects of possession in Melanesian languages. Oceanic Linguistics 12: 69–102. Lynch, John 1978 A Grammar of Lenakel. Canberra: Pacific Linguistics. Lynch, John 1982 Towards a theory of the origin of the Oceanic possessive constructions. In Amran Halim, Lois Corrington and S.A. Wurm (eds.), 243– 268. Lynch, John 2000 A Grammar of Anejom. Canberra: Pacific Linguistics. Lynch, John 2001 Passive and food possession in Oceanic languages. In Andrew Pawley, Malcolm Ross and Darrell Tryon (eds.), 193–214. Lynch, John and Malcolm Ross 2002 Banoni. In John Lynch, Malcolm Ross and Terry Crowley (eds.), 440–455. Lynch, John, Malcolm Ross and Terry Crowley (eds.) 2002 The Oceanic Languages. Richmond, Surrey: Curzon.
290 Frantisek Lichtenberk Massam, Diane and Wolfang Sperlich 2000 Possession in Niuean. In Steven R. Fischer (ed.), 281–292. McGuckin, Catherine 2002 Gapapaiwa. In John Lynch, Malcolm Ross and Terry Crowley (eds.), 297–321. Milner, George B. 1972 Fijian Grammar. (3rd ed.). Suva, Fiji: Government Press. Nichols, Johanna 1992 Linguistic Diversity in Space and Time. Chicago and London: University of Chicago Press. Ozanne-Rivierre, Françoise 1976 Le Iaai: Langue Mélanésienne d’Ouvéa (Nouvelle-Calédonie). Phonologie, Morphologie, Esquisse Syntaxique. Paris: Société d’Études Linguistiques et Anthropologiques de France. Palmer, Bill 2002 Kokota. In John Lynch, Malcolm Ross and Terry Crowley (eds.), 498–524. Palmer, Bill and Dunstan Brown 2007 Heads in Oceanic indirect possession. Oceanic Linguistics 46: 199 – 209. Partee, Barbara H. 1997 Genitives – a case study. Appendix B to Theo M. V. Janssen (1997), Compositionality. In Johan van Benthem and Alice ter Meulen (eds.), 464–470. Partee, Barbara H. and Vladimir Borschev 2003 Genitives, relational nouns, and argument-modifier ambiguity. In Ewald Lang, Claudia Maienborn and Cathrine Fabricius-Hansen (eds.), 67–112. Pawley, Andrew 1973 Some problems in Proto-Oceanic grammar. Oceanic Linguistics 12: 103–188. Pawley, Andrew and Lois Carrington (eds.). 1985 Austronesian Linguistics at the 15th Pacific Science Congress. Canberra: Pacific Linguistics. Pawley, Andrew, Malcolm Ross and Darrell Tryon (eds.). 2001 The Boy from Bundaberg: Studies in Melanesian Linguistics in Honour of Tom Dutton. Canberra: Pacific Linguistics. Pawley, Andrew and Timoci Sayaba 1990 Possessive-marking in Wayan, a Western Fijian language: noun class or relational system? In Jeremy H. C.S. Davidson (ed.), 147–171. Rehg, Kenneth L. (with the assistance of Damian G. Sohl) 1981 Ponapean Reference Grammar. Honolulu: University of Hawai‘i Press.
Attributive possessive constructions in Oceanic
291
Rivierre, Jean-Claude 1980 La Langue de Touho: Phonologie et Grammaire du Ce mu hi (Nouvelle-Calédonie). Paris: Société d’Études Linguistiques et Anthropologiques de France. Ross, Malcolm 1988 Proto Oceanic and the Austronesian Languages of Western Melanesia. Canberra: Pacific Linguistics. Ross, Malcolm 2002a Bali-Vitu. In John Lynch, Malcolm Ross and Terry Crowley (eds.), 362–386. Ross, Malcolm 2002b ’Ala’ala. In John Lynch, Malcolm Ross and Terry Crowley (eds.), 347–361. Ross, Malcolm 2002c Mussau. In John Lynch, Malcolm Ross and Terry Crowley (eds.), 148–166. Ross, Malcolm, Jeff Siegel, Robert Blust, Michael A. Colburn and W. Seiler (eds.). 1984 Papers in New Guinea Linguistics no. 23. Canberra: Pacific Linguistics. Salisbury, Mary C. 2002 A grammar of Pukapukan. Unpublished PhD thesis, University of Auckland. Schütz, Albert J. 1985 The Fijian Language. Honolulu: University of Hawai‘i Press. Senft, Gunter 1986 Kilivila: The Language of the Trobriand Islanders. Berlin / New York: Mouton de Gruyter. Siegel, Jeff 1984 Introduction to the Labu language. In Malcolm Ross, Jeff Siegel, Robert Blust, Michael A. Colburn and W. Seiler (eds.), 83–157. Sohn, Ho-min and B. W. Bender 1973 A Ulithian Grammar. Canberra: Pacific Linguistics. Taylor, John R. and Robert E. MacLaury (eds.). 1995 Language and the Cognitive Construal of the World. Berlin /New York: Mouton de Gruyter. van den Berg, René and Peter Bachet 2006 Vitu Grammar Sketch. Ukarumpa, Papua New Guinea: Summer Institute of Linguistics. Volker, Craig A. 1998 The Nalik Language of New Ireland, Papua New Guinea. New York: Peter Lang. Wilson, William H. 1976 The O and A possessive markers in Hawaiian. Unpublished MA thesis, University of Hawai‘i.
292 Frantisek Lichtenberk Wilson, William H. 1982 Proto-Polynesian Possessive Marking. Canberra: Pacific Linguistics. Wivell, Richard 1981 Kairiru grammar. Unpublished MA thesis, University of Auckland.
Possessive clauses in East Nusantara, the case of Tidore Miriam van Staden
1. Introduction Discussion of the origin of attributive possessive constructions have so far typically focussed on a number of different possible underlying schemas (Heine 1997). One possible schema that has been understudied is one where the attributive construction is directly derived from the predicative possessive construction, involving a lexical predicate (verb) whose meaning may be glossed as ‘have’ or ‘belong’. Heine (1997: 29) notes in passing that Tok Pisin, the national language of Papua New Guinea, is an example of this type (cf. Verhaar 1995), but apparently the schema is deemed to be crosslinguistically uncommon enough to warrant exclusion from the discussion. Possessive constructions that involve markers that are related to either subject or object markers in the clause are likewise ignored, although this is found more commonly cross-linguistically (e.g. in Yaqui Jelinek and Escalante 1988: 412ff., and in Hungarian de Groot 1983, see also Chappell and McGregor 1996; Allen 1964). In East Nusantara, a linguistic area in eastern Indonesia and Timor (Klamer et al. 2008), we find both possibilities. In several languages, including the local Malay varieties, the possessive construction involves a ligature that is derived from a verb meaning ‘have’. And in various cases these constructions have retained some clausal characteristics, e.g. in Buru (Grimes 1991). In other languages, and these include Papuan languages such as Abui (Kratochvíl 2007), Maybrat (Dol 1999) and Inanwatan (de Vries 2004), the possessive construction is made up of a possessum head with a subject or object marker cross-referencing the possessor. The Papuan language Tidore is an apparent mix between these two types. Like Austronesian Buru and the local Malay variety North Moluccan Malay it has a construction that displays clausal properties. But it is rather different in that it uses a possessor prefix that does not correspond to the subject prefix, and at the same time there is no traceable verbal element, like in various other Papuan languages. The result is a basic split in the system of argument cross-referencing with two types of first argument markers on the predicate:
294 Miriam van Staden “actor”, corresponding to the usual S and A roles in an accusative language, as in (1), and “possessor”, as in (2): 1 (1)
Ngori to-fayaa. 1SG.N 1SG.F.A-woman ‘I am a woman.’
(2)
Ngori ri-fayaa. 1SG.N 1SG.N.POS-woman ‘I have a wife; my wife.’
The former occur typically on verbs, and the latter typically on nouns. But neither do so exclusively. It is not the case, therefore, that the distribution of the prefixes is given by the word class of the host. Of course, the likelihood that the semantics of a verb call for an agent or undergoer participant rather than a possessor participant may be greater and at the same time possessor participants will be linked most typically with possessed goods expressed in nominal predicates (cf. Taylor 1989, 1996 on the prototypical properties of possession as an experiential gestalt, which typically involves a possessum that is a ‘specific concrete thing (usually inaninmate)’.) Yet these are facts of life rather than language, and indeed both types occur, as will be demonstrated. Clearly the gloss ‘A’ for ‘actor’ may be somewhat misleading as it would appear to exclude undergoer arguments in intransitive predications, but these, too, align with transitive A: (3)
piga yo-peka tora si suka plate 3NH.A-fall downwards first break ‘The plate falls and breaks.’
Here it is a label only to distinguish it from the ‘possessor’ participant. Tidore is furthermore interesting because it appears to lack a distinction between attributive and predicative possession. In attributive possession the possessor functions as an attribute in the noun phrase, whereas in predica1
The following abbreviations are used: A – actor, first argument of actor predication; CAUS – causative; CTRF – centrifugal; – CTRP – centripetal; DIST – distal; EMPH – emphatic; F – feminine; GEN – generic; H – human; INAL – inalienable; LOC –locative; M – masculine; N – neutral; NH – non-human; NM – noun marker; NOM – nominalisation; O – object; PL – plural; POS – possessor, first argument of possessive predication; PRED – predicate; PROX – proximal; S – subject; SG – singular; U – unmarked (feminine+non-human); and <> – unsegmented complex.
Possessive clauses in East Nusantara, the case of Tidore
295
tive possession the possessum and the possessor are in a predicational configuration in which either may be the predicate or the argument (see also the introduction to this volume). It has been claimed that all languages distinguish these two configurational relations (Heine 1997: 26). Tidore appears to be a counter-example. 2. Word classes and structural properties of Tidore Before embarking on the study of the possessive construction, it is important to first detail some aspects of Tidore grammar. One of the most notable properties of Tidore is that although word classes may be established, these classes play only a minor role in the syntax. A morphological nominalisation process distinguishes verbs on the one hand from nouns and adjectives on the other. And adjectives are a separate category because they can be noun attributes, unlike verbs. This is then also the only syntactic process where category appears relevant. Otherwise, nouns and verbs can be used to refer when they are arguments and to predicate. Take for instance a categorial noun like kolano ‘king’. This noun may be inserted in an argument position, it may have a noun attribute, and may be quantified, as in (4). But at the same time, it can also function as the predicate in a clause, in which case it will have the first argument ‘actor’ cross-reference prefix, as in (5). Here kolano predicates a property (a function) of an entity; it does not serve to identify a referent. Referential nominal predications are also possible, but then the predicate is a noun phrase, as in (6), rather than a bare noun. Now the noun cannot host a cross-reference marker: (4)
Kolano simo=ge i-gonyihi toma Kota ma-Bopo. king old=GEN 3SG.M.POS-place LOC Kota ma-Bopo ‘The old king’s place is at Kota ma-Bopo.’
(5)
Una wo-kolano toma Arab. 3SG.M 3SG.M.A-king LOC Arabia ‘He is king in Arabia.’
(6)
Una=ge kolano rimoi toma Arab isa. 3SG.M king one LOC Arabia landwards ‘He is a king from Arabia.’
One of the most curious examples showing that virtually anything may be the predicate requires a little explanation. In example (7), the interjection
296 Miriam van Staden joo is used as the main predicate in a sentence. There are two forms for ‘yes’, formal joo and informal oe. During conversations, listeners are expected to signal to the speaker that they are still listening by saying joo! or oe! at regular intervals. In the example below, the speaker comments on my own acquisition of this discursive practice. The modifier saki ‘tasty’ is also used to express satisfaction with non-food items and bahaya ‘dangerous’ is a loan from Malay that is used to express ‘extreme’ qualities: (7)
Mina mo-joo=ge saki bahaya! 3SG.F 3SG.F.A-polite.yes=GEN tasty dangerous ‘She says polite ‘yes’ in a very good way. (lit. ‘she yesses dangerously tasty!’)’
This flexibility is a property of the North Halmahera language family. In fact, among its relatives, Tidore is one of the least flexible. In various descriptions we find statements like the following: The question, what constitutes a verb in Tabaru, can only be determined through use, from which it must appear whether they can occur with the personal pronouns [i.e. the prefixes, mvs]. (Fortgens 1928: 357)
Among the words that occur as nouns and verbs are the following from Tabaru: toimi ‘bow, to shoot with bow’; gumaa ‘fish hook, to angle’; ngowaka ‘child, give birth’; singina ‘heart, think’; mede ‘moon, walk in the moonlight (for hunting)’. And on Tobelo Hueting writes: Almost all nouns, adjectives, stative verbs, even adverbs can occur as verbs or in a verbal form. This gives the language as a whole a verby character. (Hueting 1936: 361)
In Galela adjectives take person marking also when used attributively. Attributive adjectives show reduplication of the first syllable, but otherwise are inflected like predicatively used adjectives, as in the b. example (van der Veen 1915: 82; van Baarda 1908: 35–36): 2 (8)
2
Galela a. A-wi-dohu i-lamo 3N.S-3SG.M.O-foot 3N.S+3SG.M.O-big ‘His foot is big.’
Van Baarda calls these forms “participles”.
Possessive clauses in East Nusantara, the case of Tidore
297
b. A-wi-dohu i-la-lamo 3N.S-3SG.M.O-foot 3N.S+3SG.M.O-RED-big ‘His big foot.’ The Northeast Halmahera languages not only have inflected nouns and adjectives, but even number expressions with human referents require a predicative construction with an inflected numeral: (9)
Yanau (i-)wi-moi. man (3PL.SU-)3SG.M.OBJ-one ‘One man.’
Word classes, including (sub)classes of verbs and adjectives are nevertheless distinguishable, cf. Hueting (1936: 366 ff.) for Tobelo, chiefly by looking at the category changing morphology. Only adjectives and verbs may be nominalised by means of the N-prefix. Conversely, verbs and adjectives are not commonly used as referring expressions. But we do find examples like (10) and (11) in which a verb is embedded in a possessive construction, or in which a clause (12) functions as an argument in another clause: (10) Ona gahi ena ma-kia. 3PL make 3NH 3NH.POS-marry ‘They prepared the wedding.’ (11) Una i-karja duga gahi gii jira ifa jira 3SG.M 3SG.M.POS-work only make ‘person’ bad don’t bad wako koliho. return go.back ‘His only occupation was (saying) “don’t do anything evil to anyone, or the evil will come back (to you)”.’ (12) Difutu bolo modiri si ngona no-manyasal haro. tomorrow or day.after.tomorrow first 2SG 2SG.A-regret arrive ‘Tomorrow or the day after you will regret this. (lit. “you regret” arrives)’ Because categories are flexible in Tidore, it is important to keep functional labels distinct from categorial ones. The label predicate will therefore be used to refer to the functional notion of expressing a property or relation, i.e. the typical function of a verb category, and the term ‘referring expres-
298 Miriam van Staden sion’ is used to refer to the functional property of referring to some entity, i.e. the typical function of a noun (phrase). Similarly, ‘attribute’ is the functional label for the adjective in its typical function modifying the noun in the noun phrase (Hengeveld 1992). 3.
The expression of possession in Tidore
3.1. The possessive construction There exists a vast literature on what constitutes ‘possession’, and as a consequence, what should be typed a ‘possessive construction’. Possessing fingers is different from possessing brothers and different again from possessing boats or houses or books. The linguistic categories in a language reveal how different kinds of possessive relationships are grouped and distinguished from other kinds of possessive relationships, or ‘relationships of association’ as they may also be called. In Tidore there is a single construction that expresses relations of ownership, kinship, association and certain topological relations such as body parts, parts of objects, and relational nouns. This construction will be referred to as the ‘possessive construction’. The possessive construction consists minimally of a ‘possessum’ entity that is typically, but not exclusively, a noun, with a possessive prefix that cross-references the possessor. The possessive prefix is obligatory. It is very different in this respect from the actor prefix that is optional in most contexts. The possessor may be further expressed in a noun phrase preceding the possessum: (13) Una i-fola yo-ruba rai. 3SG.M 3SG.M.POS-house 3NH-fall.apart already ‘His house is falling apart.’ (14) Cole ma-giba yo-foluji. bra 3NH.POS-strap 3N.A-come.loose ‘The bra strap has come loose.’ The possessor noun phrase may be internally complex. It may, for instance, consist of a possessive construction: (15) Ica mi-ngofa na-guru. Ica SG.F.POS-child 3PL.POS-teacher ‘The teacher of Ica’s children.’
Possessive clauses in East Nusantara, the case of Tidore
299
As common in the world’s languages, in Tidore, too, the possessive construction is used to express benefactive/malefactive relations (Lichtenberk 1985, 2002; Margetts 2004). The possessive construction then follows the transitive verb and the possessor is the beneficiary of the action described by the main predicate and the possessum. (16) Mina mo-foli Ica mi-saloi rimoi. 3SG.F 3SG.F.A-buy Ica 3SG.F.POS-basket one ‘She bought Ica a basket.’ Note that for the expression of a beneficiary it is also possible to have a double object construction (see van Staden 2000: 222–223 for details): (17) Mina mo-foli Ica saloi rimoi. 3SG.F 3SG.F.A-buy Ica basket one ‘She bought Ica a basket.’ Like most East Nusantara languages, but unlike some of the related North Halmahera languages, e.g. Tobelo (Holton 2003), Tidore distinguishes alienable from inalienable possession. The latter unsurprisingly includes the expression of body parts, kinship terms, locative nouns and expressions of part-whole relations (Chappell and McGregor 1996), but also items that are ‘baptized’, i.e. ceremonially attached to a possessor, such as ‘boats’, ‘houses’, and ‘names’ (van Staden 2006). The group of alienables include notions of non-permanent or incidental ownership as in ‘my bag’, and association, such as ‘his photograph’ (the photograph in which he appears). The difference between the two is that in the case of inalienable possession the possessive marker may in all cases be reduced to the 3NH marker mawhereas in alienable possession, the possessive prefix must cross-reference person and number of the possessor. The formal difference between alienable and inalienable possession is thus slight and I consider them to be variations on the possessive construction, rather than different construction types. The possessive construction as a whole is typically an argument in a clause. It may then be preceded by a preposition: (18) Ngoto lahi [se mansia miskin romoi ena=ge 3NH=there 1SG.N.A ask.for ADD people poor one na-ngofa rimoi=ge] la ngoto kia. 3PL.POS-child one=there in.order.to 1SG.N.A marry ‘I will ask a poor person’s child to marry me (lit. ‘so that I marry’).’
300 Miriam van Staden As this example shows it can be quantified also. Whether the quantifier has scope over the possessum only or over the entire construction is ambiguous. Tidore has no opposition between ‘their three children’ and ‘three of their children’. Adjectival modification of the possessive construction has so far not been attested in discourse; although in elicitation it is deemed possible: (19) ?Ona na-ngofa jang rimoi. 3PL 3PL.POS-child beautiful one ‘Their beautiful child; a beautiful child of theirs.’ We will return to modification and quantification in section 5.1, in which implications for the analysis of possessive constructions are discussed. 4. Other means of expressing relationships of possession Although the focus of this chapter is on the possessive construction, this is not the only way to express relationships of possession or association. There are four additional constructions that may express relations of possession. All four are predicative possessive constructions. Three are conceptually grounded in other domains: the ‘Topic’ construction may be rendered as ‘(as for) X, Y’, the ‘existential’ locative construction expresses “X exists (at) Y” and the company construction gives “X is with Y” (Heine 1997: 47). The fourth construction is rather interesting, involving the dummy possessum element due and may be seen as a variation on the Topic construction combined with the possessive construction. The Topic construction is a juxtaposition of a possessor noun phrase followed by a possessum noun phrase, as in (20). The existential construction is rather similar but has the existential verb sema ‘to be, exist’ (21). The companion construction is again superficially highly similar but now the possessum noun phrase is preceded by preposition soma ‘additional participant’ as shown by (22): (20) Fola=ge/ ngora biasa ua. house=there door normal NEG ‘That house does not have a normal door.’ (lit. ‘The house, no normal doors’) (21) Mina sema ngofa rai. 3SG.F be child already ‘She already has children.’ (lit. ‘She, children already exist’)
Possessive clauses in East Nusantara, the case of Tidore
301
(22) Mina soma raa yang. 3SG.F ADD husband not.yet ‘She does not have a husband yet.’ (lit. ‘She not yet with husband’) An important difference between the Topic construction and the other two is that the prosodic contour of the Topic construction shows a rise in pitch on the Topic followed by a sharp drop, indicated by the slash in the language line, whereas the other two have a regular declarative clause contour. Although they are similar in form and appear to have a similar meaning, sema ‘be’ and soma ‘additional participant’ must be distinguished. In most contexts, they are in complementary distribution: (23) Mina sema/ *soma bolo ua? 3SG.F be ADD or NEG ‘Is she (here/there)?’ (24) Oyo hula soma / *sema igo ma-ake. eat sago ADD be coconut 3NH.POS-water ‘Eat sago with coconut juice.’ Preposition soma is part of a paradigm of four adpositions: soma and se precede company or instrument phrases. The former is used before non-human denoting nouns, the latter before human denoting nouns. Its etymology is unclear. It is possible that the collocation of te and se before non-human denoting nouns preceded by ma- led to this reanalysis.3 But it is equally possible that soma is derived from coma ‘to add’, e.g. for adding spices to a dish. Existential predicate sema similarly lacks obvious cognates in the 3
The other North Halmahera languages have a morpheme re (Sahu, Visser and Voorhoeve 1987) or de (e.g. in Galela, van Baarda 1895) which, like se in Tidore, functions both as a conjunction and as a preposition on (some) oblique arguments. It is also found in expressions of possession, for example in Galela where it may only be used with inanimate possessors ‘to indicate that it has something’ (van Baarda 1895: 246): (i) O kurisi de ma gogocoho-ka NM chair with 3SG.POS armrest-DIR ‘The chair with armrests.’ or ‘The chair has armrests.’ In this use it resembles the Tidore preposition soma. However, there is no regular sound correspondence between Sahu r, Galela d and Tidore s. In fact, according to Wada (1980: 503), Tidore should have had y instead.
302 Miriam van Staden other North Halmahera languages. Instead, some of the related North Halmahera languages have an existential predicate ka, which in Tidore is found only as a bound morpheme in the locative predicates (e.g. ka-tai ‘be located seawards’, ka-tau ‘be located upwards’, etc.) and in the negative existential predicates ka-ua ‘be not’, ka-yang ‘be not yet’ and ka-rewa ‘be no more’. The fourth construction type involves the bound morpheme due ‘possession(s)’ that takes the position of the possessum. It is considered a ‘dummy’ possessum element since it does not occur outside the possessive construction. In function it resembles the predicative possessor constructions, as found in English when the possessor is the predicate and the possessum its argument (‘the red one is mine’) or vice versa (‘mine is the red one’). This construction type will be referred to as ‘generic possession’, after the ‘generic’ element due. But Tidore has no set of independent possessive pronouns that could function as the predicate. Instead, the equivalent in Tidore involves non-verbal predication with two NPs, one of which is a possessive construction built on a ‘dummy’ possessum due. This possessive construction may be the argument of the construction, as in (25), where the predicate gives the identity of the possessed entity, or it may be the predicate, as in (26): (25) Mina mi-due ena ma-bulo. 3SG.F 3SG.F.POS-possess 3NH NOM-white ‘Hers is the white one.’ (26) Ena=re laba ma-due se ena=re saihuu 3NH=PROX monkey 3NH.POS-possess and 3NH.here captain ma-due. 3NH.POS-possess ‘This is the monkey’s and this is the captain’s.’ The two sentences answer different questions. In the first example, (27), the utterance is a response to the question which one she owns, while in the second, (28), it is an answer to the question who owns (or should own, as in this case) the entity: (27) Mina mi-due mbe? 3SG.F 3SF.POS-possess which.one ‘Which one is hers?’
Possessive clauses in East Nusantara, the case of Tidore
303
(28) Ngge nage na-due? this.one who 3PL.POS-possess ‘Whose is this one?’ These alternative ways of expressing possessive relations are illustrated here with a purpose. Not only do they give a more complete picture of the ways of expressing possession in Tidore, they also show that all alternative strategies to the possessive construction described in §3.1 involve predications. In the following section I will show that the possessive construction itself is also a case of predicative possession, and that in fact, Tidore has no true attributive possessive construction. In the remainder of the paper I will then try to find explanations for this highly unusual situation. 5. The possessive construction as a predication Heine’s study of possession led him to conclude that in all known languages clear differences show up between ‘attributive possession’ and ‘predicative possession’ (1997: 26). In the former, the possessor and possessum are generally contained in a noun phrase and the possessor functions as an ‘attribute’ of the possessum; in the latter, the relation between the possessum and the possessor is expressed in a predication. In the former possession would be presupposed, in the latter possession is asserted. In terms of Hengeveld and Mackenzie’s (2008) framework, following Dik (1989, 1997), it could be said that in the former case the possessor further restricts the potential set of referents designated by the possessum (from ‘all possible entities X’ to ‘the entities X belonging to Y’) while in the latter a relation of possession is predicated of the possessor and possessum arguments. The expression of both types may rely on the same conceptual structure and even on the same linguistic form, but ‘clear differences’, according to Heine, may always be found. Tidore is an exception. The evidence presented here shows that one construction is used whether possession is presupposed or not. The possessive construction in Tidore is a type of predicative possession. But this predication is regularly embedded as an argument of another predicate in which case the entire construction may used referentially. This is the typical case where possession is presupposed, i.e. where an attributive possessive construction would be expected. In brief, the structure is: [Y pos-X] for ‘Y has X’, which is embedded as an argument into [[Y has-X] Verb] ‘Y’s X Verbs’.
304 Miriam van Staden 5.1. The arguments The arguments in favour of the analysis of possessive constructions as predications are the following: 1) Possessive constructions can constitute independent utterances; 2) Embedded possessive constructions may have either a referential or a predicative reading; 3) The heads of possessive constructions can be preceded by focus marker kama; 4) Possessive constructions and other clauses, but not noun phrases, can follow conjunction untuk; 5) Adjectives and verbs otherwise do not occur as heads of referring expressions; the only time they occur in a typically nominal position is as heads of possessive constructions. The first argument is more appropriately a condition for a clausal analysis rather than evidence, since non-clausal utterances also exist. The second, third and fourth arguments constitute evidence, however, while the last argument may be considered supporting evidence. Each will be discussed in turn. 5.1.1. Independent utterances If possessive constructions are predications then they should be able to constitute independent clauses. And indeed they are. Possessive constructions need not be embedded in a larger structure but they may predicate possession on their own. As such they may have their own temporal modal and aspectual modifiers, and they may be questioned like other clauses, as in (33):4 (29) Ua
si, ngona yuke dadi ngona ni-due. first 2SG earlier so 2SG 2SG.POS-possess ‘No, you were first so it is yours.’
NEG
4
Yes /no questions in Tidore are formed by adding ‘or no’ to the questioned clause. In conjunction with aspectual modifiers rai ‘already’ and moju ‘still’, their negative counterparts are used, so rai bolo yang ‘yet or not yet’ (as in example (33)), moju bolo rewa ‘X still or no more’.
Possessive clauses in East Nusantara, the case of Tidore
305
(30) Fayaa rimoi ngge toma fola ma-doya ngge, mina woman one 3NH.GEN LOC house INAL-inside 3NH.GEN 3SG.F mi-cahaya. 3SG.F.POS-shine ‘There is a woman in the house, she is a shining beauty (lit. has brightness).’ rewa rai, […] Jailolo ena-re i-sultan Jailolo 3NH-PROX 3SG.M.POS-sultan not.anymore already ‘Then later (lit. towards the present) Jailolo no longer had a sultan, […].’
(31) Ino
CTRP
(32) una wo-rasa paha rea karna una 3SG.M 3SG.M.A-feel bear not.anymore because 3SG.M ua i-daera 3SG.M.POS-area NEG ‘He felt he couldn’t take it any more because he did not have a piece of the land.’ rai bolo yang? (33) Mina mi-ngofa 3SG.F 3SG.F.POS-child already or not.yet ‘Has she got children? (lit. she has child already or not yet?)’ In discussions with native speakers, these sentences are never felt to be incomplete or to require a context as is often the case with non-clausal utterances. It is notable that when used as an independent utterance the possessive construction typically has a modal or aspectual modifier. In the context of Tidore this may be an important fact since historically the modal and aspectual modifiers are intransitive verbs, taking as their only argument the event description that is in their scope. In some of the related North Halmahera languages they still carry person marking like other verbs, and even in Tidore some of the modal and aspectual modifiers regularly function as main predicates, e.g. rewa ‘not anymore’ (cf. example (31) above): (34) Gula=re yo-rewa rai. sugar=PROX 3NH.A-not.anymore already ‘The sugar is all gone.’ In present-day Tidore, the placement of modal and aspectual modifiers is always immediately after the unit they scope over, e.g. the following examples in which the predicate is underlined and the scope unit in brackets:
306 Miriam van Staden (35) Turus [gahi ena ma-mammam] dadi. then make 3NH 3NH.POS-sweets able.to lit.: ‘Then (make its pudding) is possible/happened.’ (36) [[Mina mo-wako] maya] rewa. 3SG.F 3SG.F.A-return may not.anymore lit.: ‘((She return) may) not anymore.’ Possessive constructions are no different. What is interesting is that in many cases there are two readings when the possessive construction has a modal or aspectual modifier: (37) una i-ngofa nau-nau duga rimoi 3SG.M 3SG.M.POS-child RED-male only one 1. ‘He has one son only.’ 2. ‘His male children number just one.’ In the first reading it is the possessive relation that is predicated, but in the second possession of sons is presupposed and it is just number that is predicated. The former reading, without the presupposition, is expected when the predicate is the possessum; the latter reading is expected if the possessive construction is a referential phrase and the predicate is duga rimoi ‘only one’.5 A similar situation arises with numerals: (38) Mina mi-ngofa rimoi reke. 3SG.F 3SG.F.POS-child one cry 1. ‘One of her children is crying.’ 2. ‘Her one child is crying.’ Tidore has no distinction between ‘one of her children’ and ‘her one child’. In the first reading the numeral has scope over the entire possessive construction (cf. the paraphrase ‘as for her children, one is crying’), and in the second it scopes over just the possessum (cf. she has one child; it cries).
5
A reading in with naunau as the predicate is excluded here on prosodic grounds because there is no drop in pitch after i-ngofa and on syntactic grounds because in that case it could not have be followed by duga rimoi ‘only one’. But it is in theory possible to say: (ii) Una i-ngofa [nau-nau.]pred 3SG.M 3SG.M.POS-child RED-male ‘His child is a boy.’
Possessive clauses in East Nusantara, the case of Tidore
307
The former is therefore compatible with a non-presupposed or predicative reading and the latter with a presupposed or attributive one.
5.1.2. Complement taking predicates When possessive constructions are embedded as complements of verbs that take both clausal and noun phrase complements, different readings are often available: (39) Una wo-hoda mina mi-ngofa. 3SG.M 3SG.M.A-see 3SG.F 3SG.F.POS-child 1. ‘He sees her children.’ 2. ‘He sees (that) she has children.’ It is doubtful that this points at a ‘clear distinction’ between an attributive and predicative construction in Tidore. Instead of structural ambiguity, the different readings are the result of differences in the informational status of possessive relation as either presupposed or asserted. What it does show clearly is that predicative readings are available.
5.1.3. Negative focus with kama Perhaps the most striking property of the possessive construction is that the possessum behaves like a verbal predicate with respect to the placement of the negative focus particle kama. Tidore negative particles ua ‘not, no’, rewa ‘not anymore’, ifa ‘don’t’, yang ‘not yet’ all occur in clause final position. When focus is on the main predicate, a particle kama may directly precede it. This applies equally to verbal and nominal predicates: (40) Ngom kama hoda mina rewa. 1PL.EX NEG.FOC see 3SG.F not.anymore ‘We don’t see her anymore.’ (41) Una kama kolano ua. 3SG.M NEG king NEG ‘He was not a king.’ In this last example, the noun kolano ‘king’ predicates a property (function, profession) of some person (recall example 5). Because the negative parti-
308 Miriam van Staden cles always occur clause finally, negation in complex clauses is ambiguous. An important use of kama is to disambiguate: (42) Ngom kama baso mina mo-nyanyi 3SG.M NEG.FOC hear 3SG.F 3SG.F.A-sing ‘We have not heard her singing yet.’
yang. not.yet
(43) Ngom baso mina kama mo-nyanyi yang. 3SG.M hear 3SG.F NEG.FOC 3SG.F.A-sing not.yet ‘We hear that she hasn’t sung yet.’ The focus particle cannot precede noun phrase arguments, as (44) shows, but in clauses containing a possessive construction kama may occur before the possessum, as in (45) to (47): (44) *Ngom baso kama mina yang. 3SG.M hear NEG.FOC 3SG.F not.yet intended reading: ‘We have not heard her yet.’ (45) Ngoto hoda mina kama mi-ngofa ua. 1SG.N.A see 3SG.F NEG 3SG.F.POS-child NEG ‘I saw she has no children.’ (lit. ‘I saw she was without children’) (46) Mina kama mi-oli fugo ua. 3SG.F NEG 3SG.F.POS-voice come.out NEG ‘Her voice did not come out.’ (lit. ‘she does not have a voice that comes out) i.e. she did not speak’) (47) Una kama i-yohu gola ua. 3SG.M NEG 3SG.M.POS-leg hurt NEG ‘His leg does not hurt.’ (lit. ‘he does not have a leg that hurts’) There is one restriction on this type of negation. If the semantics of the matrix predicate presupposes the existence of an entity as, for instance the undergoer, and if this entity is is the possessum, internal negation is impossible. In other words, it is nonsensical to express an action affecting an undergoer, when this undergoer is then claimed to be non-existent: (48) *Ngoto cako mina kama mi-ngofa ua. 3SG.F NEG 3SG.F.POS-child NEG 1SG.N.A hit ??? ‘I am hitting the child she does not have.’
Possessive clauses in East Nusantara, the case of Tidore
309
(49) *Ngoto jau mina kama mi-gia ua. 1SG.N.A hold 3SG.F NEG 3SG.F.POS-hand NEG ???‘I hold her non-existing hand.’ This is, however, a semantic and not a structural restriction. Informants stated that these utterances were nonsense as it would be impossible to ‘hit non-existing children’, or ‘hold non-existing hands’. These facts all point to an analysis of the possessed noun as a clausal predicate with a possessor argument.
5.1.4. Complements of untuk Similar evidence comes from the type of complement that loan conjunction untuk takes. As we saw earlier, beneficiaries may be expressed through the possessive construction. An alternative to this construction is the use of a purpose clause introduced by the Malay loan word untuk ‘for’. In Standard Indonesian this is a preposition directly marking the beneficiary, as in (50). In Tidore, the complement of untuk cannot be a noun phrase, hence the ungrammaticality of (51), but must be clausal, as in (52): Indonesian (50) Apa surat ini untuk saya? what letter this for 1SG ‘Is this letter for me?’ (51) *Mina foli saloi rimoi untuk Ica. 3SG.F buy basket one in.order.to Ica (52) Mina isa rasi untuk gahi 3SG.F landwards first in.order.to make mina mi-raa=ge ngam. 3SG.F 3SG.F.POS-husband=GEN food ‘She goes home (lit. ‘landwards’) now to make her husband a meal.’ Yet, possessive constructions are possible complements, as shown in (53). In fact, generic -due must be used if the possessed item is not further specified, compare (54) to (51) above. Interestingly, even in the North Moluccan Malay spoken on Tidore a clausal complement is preferred to the nominal
310 Miriam van Staden phrase, as in (55) (see section 6 for more cases of analogy in structure between Tidore and North Moluccan Malay): 6 (53) Ua
si ngona ngofa no-nau, ge ngan first 2SG child C.RED-male there 2SG sa-dia untuk ngori ri-diali. CAUS-NOM.leave.behind for 1SG.N 1SG.POS-NOM.replace ‘Otherwise, if you have a boy, then you leave it for my replacement (i.e. to replace me).’ NEG
(54) Mina foli saloi rimoi untuk Ica mi-due. 3SG.F buy basket one in.order.to Ica 3SG.F.POS-possess. ‘She bought a basket for Ica (to have).’ (55) Ini untuk kita punya! this for 1SG have ‘This one is for me (to have)!’ This demonstrates that the possessive construction groups with clauses and not with noun phrases. 5.1.5. Verbs and adjectives as heads If possessive constructions are treated as clausal predications with a ‘possessum’ main predicate, it is not surprising that the possessum could also be a verb or adjective. While verbs and adjectives are extremely uncommon as heads of referring expressions (i.e. noun phrases with ‘non-nominal heads’ are virtually non-existent) they do occur as heads of possessive constructions. (56) Mina mi-jang. 3SG.F 3SG.F.POS-beautiful ‘Her beauty.’
6
In formal contexts Standard Indonesian may interfere with North Moluccan Malay.
Possessive clauses in East Nusantara, the case of Tidore
311
(57) Una tagi pana namo nde una i-oyo se 3SG.M go shoot bird 3NH.PROX 3SG.M 3SG.M.POS-eat and yuru ake […] te simo nange. drink water LOC old just.now ‘He (who) went shooting birds, he had food and drink with this old man just mentioned.’ The fact that adjectives and verbs occur in the possessive construction in a syntactic position typically associated with nouns, but not as heads of noun phrases, may be explained by the predicative rather than referential function that the head of the possessive construction has. The reason that the vast majority of possessums are nominal is semantic rather than structural: what is possessed is typically a physical object expressed by a noun. There is no need to assume nominalisation in these cases. In actor clauses, the main predicate is typically verbal, although nominal and adjectival predicates are also attested, while in possessor-clauses the main predicate is typically nominal, although both adjectival and verbal predicates also occur. Contrast this to the nominalisation process involving the prefix ma (also for 3NH.POS): 7 (58) Mina ma-jang. 3SG.F NOM-beautiful ‘She is a beauty.’ (59) Mina ma-din kabaya. 3SG.F NOM-sew dress ‘She is a seamstress.’ Here, there is evidence of derivation, since the ma-prefixed stem takes actor marking: (60) Mina mo-ma-din kabaya. 3SG.F 3SG.F.A-NOM-sew dress ‘She sews dresses for a living.’ 7
Again in North Moluccan Malay we find the literal equivalent: (iii) Mina ma- jang sado! Tidore de pe cantik sampe! North Moluccan Malay 3SG(.F) POS beautiful until ‘She is such a beauty!’
312 Miriam van Staden The difference in meaning between the actor and possessive constructions is that with ma- the property is no longer transient but permanent. Thus (62) is unacceptable. (61) Ngire mina jang sodooo! last.night 3SG.F beautiful until ‘Last night she was very beautiful!’ (62) *?Ngire mina ma-jang sodooo! last.night 3SG.F NOM-beautiful until 5.2. Against a clausal analysis So far the arguments supported the analysis of the Tidore possessive construction as basically ‘clausal’. However, although historically this may be the case, and synchronically also the construction shows clausal properties, there are several ways in which possessive constructions differ from actorclauses and these could be taken as counter-arguments against this clausal analysis. First of all, unlike typical actor clauses, possessive constructions do not take aspectual or temporal modifiers when functioning as arguments of verbs. And as independent utterances they are never modified by modal auxiliaries such as dadi ‘be able to’, or maya ‘want to’: (63) *Mina mi-ngofa dadi ua. 3SG.F 3SG.F.POS-child able.to NEG intended reading: ‘She is unable to have children.’ Instead, a verbal predicate is called on: (64) Mina mo-dahe ngofa dadi ua. 3SG.F 3SG.F.A-find child able.to NEG ‘She cannot have children.’ Furthermore, the distribution of actor and possessor prefixes is not perfectly parallel. Actor prefixes are always optional. There are situations in which actor prefixes cannot occur on an auxiliary, but wherever they are found in texts native speakers consider the utterances equally good without the prefix. Possessor prefixes, on the other hand, are obligatory. And although ac-
Possessive clauses in East Nusantara, the case of Tidore
313
tor prefixes occur also on nominal predicates, as was illustrated in section 2, this is only possible when the predicate is a bare noun. Once the predicate is a full noun phrase it can no longer have an actor prefix, recall examples (5) and (6), repeated here for convenience: (5)
Una wo-kolano toma Arab. 3SG.M 3SG.M.A-king LOC Arabia ‘He is king in Arabia.’
(6)
Una=ge kolano rimoi toma Arab isa. 3SG.M king one LOC Arabia landwards ‘He is a king from Arabia.’
However, in possessive constructions, the possessum is a noun phrase that has the possessor prefix. This lack of parallelism could be seen as counterevidence. Finally, if the possessive is analysed as a clause, one might expect that the assertion reading (i.e. the predicative reading) should be primary in complementation constructions of the type discussed in § 5.1.2 above. But in fact it is the presupposition reading that is the unmarked one, perhaps suggesting that the construction is a noun phrase. But this could also be the result of the typical use of the construction expressing presupposed possession in conjunction with the likelihood that in cases in which the possessive construction follows a verbal predicate that may take both clausal and nominal complements the second interpretation is typically intendend.
5.3. Evaluation The various arguments that were presented show that Tidore lacks a distinction between an attributive and predicative possessive construction. The possessive construction has clausal properties also when possession is presupposed. The result is that in Tidore there is a split in the system of argument marking on the predicate: transitive predicates cross-reference the first argument by means of a ‘subject’ or ‘actor’ prefix, as do intransitive predicates that describe actions, states, experiences, etc., except those predicates that give a possessive relation, because these take a ‘possessor’ first argument. This gives the regular oppositions illustrated in the introduction, and repeated here:
314 Miriam van Staden (65) Ngori to-fayaa. 1SG.N 1SG.N.A-woman ‘I am a woman.’ (66) Ngori ri-fayaa. 1SG.N 1SG.N.POS-woman ‘I have a wife; my wife.’ Typologically, the Tidore system is rather unusual. Possessive markers that are identical to subject markers are found in some languages, but to find what appears to be a split in the system of predicate cross-referencing depending on whether the relation between the argument and the predicate is one of actor or undergoer cross-referencing on the one hand versus possessor marking on the other appears truly rare. Also, as pointed out, not to find a clear distinction between attributive and predicative possessive constructions is exceptional. In the next sections, possible historical explanations for this synchronically ‘odd’ possessive construction are explored. One possibility that is examined is that the Tidore system may be copied from North Moluccan Malay through a process called ‘metatypy’ (Ross 2006). I also investigate whether other languages in the same linguistic area display constructions that are anything like what we find in Tidore and how Tidore compares to genealogically related languages of the North Halmahera Family and the Papuan languages of the Bird’s Head. It is shown that this last comparison gives one possible account for the synchronically rather unusual Tidore construction. 6. North Moluccan Malay Throughout Eastern Indonesia, different varieties of Malay have served as the lingua franca for many centuries and despite education in Standard Indonesian, their importance has increased rather than decreased during the past sixty years. The specificities differ from place to place so that ‘Ambonese Malay’ is distinguished from ‘Kupang Malay’ (spoken on Timor), ‘North Moluccan Malay’ and ‘Irian Malay’, but the general patterns are the same in all these varieties that together are referred to as ‘Eastern Indonesian Malay’. The possessive construction shows the same pattern throughout the area. Unlike Malay varieties in other parts of Indonesia and Malaysia, and unlike the standardized national variety ‘Indonesian’, the order in the possessive construction is possessor-possessum. Furthermore, the construction
Possessive clauses in East Nusantara, the case of Tidore
315
contains a possessive ligature that is derived from the verb punya ‘have’: pu in Ambonese Malay (Tjia 1997; van Minde 1997), pe in Irian and North Moluccan Malay. Where the varieties of Eastern Indonesian Malay differ is whether punya may be used in full in attributive possessive constructions, and whether this construction is also used to express benefaction. The claims in this section apply only the variety spoken on Tidore, although following native speaker practice it is referred to as North Moluccan Malay which would cover all of the North Moluccas. Where languages are in prolonged contact, the systems may converge to a greater or lesser extent. The diachronic process underlying such syntactic convergence of systems has been referred to as ‘metatypy’ (Ross 1996, 1997, 2001, 2006). In metatypy ‘the syntax of one of the languages of a bilingual speech community is restructured on the model of the syntax of the speaker’s other language’ (Ross 2006). The result is very similar ways of speaking in languages that are in contact. In the North Moluccas, in particular in Ternate, Tidore and Taba, this has occurred on a large scale. We find it for instance in the use of spatially loaded terms in the regional Malay varieties (Bowden 2005). The North Moluccan Malay expressions appear to be literal copies from the indigenous varieties: (67) Dong pi lao di=lao.8 North Moluccan Malay Ona tagi hoo ka-tai. Tidore 3PL go sea(wards) ‘locative’9=sea(ward) ‘They went seawards to a location seawards.’ In the expression of possession the same similarities in patterning are found: (68) Dong pe anak so tiga. North Moluccan Malay Ona nangofa range rai. Tidore 3PL 3PL.POS-child already three already ‘They already have three children.’
8
9
Note that Tidore has distinct morphemes for ‘sea’ (ngolo), moving seawards (hoo) and seaward location (=tai), where Malay has only lao ‘sea’, showing that Tidore was the model for North Moluccan Malay rather than vice versa. In North Moluccan Malay this locative corresponds to the locative preposition in Standard Indonesian; in Tidore the locative is a predicate deriving prefix.
316 Miriam van Staden (69) Turus potong de =pe kapala. Turus tola ena madofolo. then cut/chop 3NH 3NH.POS head ‘Then cut off its head.’
North Moluccan Malay Tidore
Depending on the focus of the construction possessor third person pronoun dia can be reduced to de to which the ligature pe cliticises in (69), further obscuring the original verbal element and the original verbal construction. In attributive possession, the possessive ligature in North Moluccan Malay is typically the reduced form pe, but at least on Tidore, the full form of the verb punya is also found in attributive possessive expressions. In example (70) the North Moluccan Malay possessive construction is a code switch used to translate the Tidore word dano ‘grandchild’. The remainder of the sentence is also in North Moluccan Malay: North Moluccan Malay (70) Dano-dano raja-raja punya ana cucu itu kalau fam RED-grandchild RED-king have child grandchild that if family Alting itu Blanda. Alting that Dutch ‘(As for) the grandchildren, the grandchildren of the kings, if they are called Alting then they are Dutch.’ In the beneficiary expressions, too, the choice is between the full verb and the reduced form: North Moluccan Malay (71) Ekal, ambil ci Miriam punya/pe kursi ka=mari. Ekal fetch sister Miriam have chair DIR =here ‘Ekal, fetch Miriam a chair.’ In predicative possession only the verb may not be reduced, giving a formal distinction between the two functions. And in generic possession also the full form of the verb is obligatory, but the similarity in structure to Tidore is again striking: (72) Yang kita punya puti. Ngori ri- due bulo. REL 1SG 1SG.POS-have white ‘Mine is/are white.’
North Moluccan Malay Tidore
Possessive clauses in East Nusantara, the case of Tidore
317
The question that this parallel raises is whether perhaps the Tidore construction was reanalysed as clausal on the basis of its similarity to the Malay possessive construction with its obvious clausal history. But the opposite is also possible. Perhaps the North Moluccan Malay construction evolved under the influence of the clausal constructions in the indigenous languages of the area. A third possibility is of course that the two happen to be similar synchronically, but have independent histories. There are two arguments against the first scenario. First, clausal possessive constructions are a characteristic of all North Halmahera languages as will be demonstrated in the next sections. Furthermore, such constructions are found throughout East Nusantara. Second, the North Moluccan Malay possessive construction has the possessor-possessum order which is unusual for an Austronesian language. Malay dialects spoken in western Indonesia have the usual possessum-possessor order. But this order is common for both Papuan and Austronesian languages in East Nusantara. It appears that the indigenous Papuan languages influenced the Austronesian languages in this area. If the clausal analysis in Tidore is a case of metatypy, we would need to assume that North Moluccan Malay first took the order possessor-possessum from languages like Tidore, and then introduced a clausal construction to express this relation of possession. And this construction was then subsequently borrowed back into Tidore. Complex scenarios like these should not always be ruled out, but it would seem more likely that the influence has been the other way round. The North Moluccan Malay construction is based on the indigenous languages of the area in which it is spoken. For the same reason it is unlikely that the Malay construction developed completely independently from the indigenous languages of the area. In the following two sections the expression of possession in East Nusantara as the linguistic area that Tidore forms part of, and in the North Halmahera Family will be explored in order to examine to what extent the Tidore construction is a product of its neighbours or ancestors. 7. East Nusantara East Nusantara is roughly the area east of Bali and Sulawesi up to and including the Bird’s Head of Papua and perhaps even further East along the North coast of New Guinea (van Staden and Reesink 2008; Klamer et al. 2008; Klamer 2003; Baird 2002). This area is characterized historically by high mobility of people and intense contact between people and languages. It is a complex patchwork of languages. Here different branches of Aus-
318 Miriam van Staden tronesian meet as well as a large number of Papuan languages for which genetic relations cannot or only with great reservations be established (Wurm 1982; Foley 1986; Reesink 2000, 2002, 1998). A West Papuan Phylum has been proposed (Voorhoeve 1987, 1987) relating the North Halmahera languages, including Tidore, to several of the languages of the Bird’s Head, Papua. But Reesink (2000) has proposed that this phylum is made up of six language families: 1. West Bird’s Head with North Halmahera, three isolates: 2. Abun, 3. Maybrat, 4. Mpur, and two small families in the Eastern Bird’s Head: 5. Meyah and Sougb, and 6. Hatam and Mansim. In addition, the South Bird’s Head languages again form a separate family, as do the Papuan languages of Alor-Timor-Pantar, such as Abui and Adang. In addition to these Papuan languages, we also find languages belonging to two different branches of Austronesian: Central Malayo-Polynesian (e.g. Buru) (Blust 1993) and South Halmahera West New Guinea (Biak), which together with the Oceanic languages make up the Eastern Malayo-Polynesian subgroup (e.g.Tryon 1995: 34).
Figure 1. East Nusantara. Papuan languages are found in Papua and in the encircled areas; Austronesian languages are spoken elsewhere in the archipellago. Map adapted from Klamer, et al. (2008)
Possessive clauses in East Nusantara, the case of Tidore
319
Possessive constructions in East Nusantara are relatively well-studied because they exhibit a number of cross-linguistically interesting features. First, the Austronesian languages in East Nusantara show an unexpected word order in attributive possessive constructions. In spite of their general SVO typology, the order in the possessive construction is typically possessorpossessum. This is most likely the result of language contact with neighbouring Papuan languages that are at least originally SOV and also have possessor-possessum order (Klamer et al. 2008). Another characteristic of East Nusantara is that many languages formally distinguish alienable and inalienable possession. Furthermore it is a striking feature that for the expression of at least one of these two possessive relations, the marker of possession is often derived from a subject or object prefix. Again, this is found alike in Austronesian and Papuan languages, and again there is evidence that the latter have influenced the former in this respect. Several languages have a verbal element in at least one of the attributive possessive constructions. For example in Biak, an Austronesian language of Papua, the construction that expresses alienable possession involves a complex ‘ligature’ that comprises of an element ve ‘POS’ traceable to a verb ‘to possess’, which is inflected for subject (possessor) and followed by a complex ‘article’ that expresses the possessum (Van den Heuvel 2006: 229ff.): Biak (73) … ko-swar min ko-ve=s-i 1PL.INC-love member 1PL.INC-POS=3PL .ANIM-SPC ‘… we love our neighbours.’ (Van den Heuvel 2006: 437) Buru has a possessive construction which ‘mirrors the structure of an active transitive clause with a pre-verbal Actor, the possessive word ( verb) and a post-verbal Undergoer’ (Grimes 1991: 278). The possessive word has an applicative suffx -k to indicate ‘a definite pronominal object’ (p. 229). The possessive word even accepts certain valence changing verbal prefixes: Buru (74) Kawasan p-em-nake-k geba rua ute tinge […] CAUS-STAT-3SG.POSS-k person two DAT 3SG […]. head ‘The village head put two people at his disposal […]’ As in Tidore, this one construction may be interpreted predicatively and attributively. Grimes summarises this as follows:
320 Miriam van Staden When the possessive construction functions predicatively, the possessive word is the nucleus of the clause […], behaving like a transitive verb with a subject and object. Used predicatively, the possessive word is the head of the construction. When the possessive construction functions as a nominal argument in a clause, the possessive word functions as a possessive pronoun, being a pre-head modifier to the head noun. Thus, whether the possessive word is behaving verbally or nominally, or whether it is functioning as the head of a predicative possessive construction or as the modifier of a possessive NP depends on its distribution. (Grimes 1991: 282)
More generally the possessor affixes correspond to either the subject or the object markers, also when no verbal element is found. For instance in Maybrat the possessor is expressed by a subject prefix in inalienable possessive constructions: Maybrat (75) Fnia m-ao. woman 3U-foot ‘The woman’s foot.’
(Dol 1999: 149)
And other Papuan languages of the Bird’s Head such as Meyah (Gravelle 2004: 171), Moi (Menick 1995: 65), Mpur (Odé 2002: 62) and Hatam (Reesink 1999: 49) are identical. The Southern Bird’s Head language Inanwatan has two highly similar sets of prefixes for inalienable possession (de Vries 2004: 29) and object marking (de Vries 2004: 36) although the two differ in certain morpho-phonological properties. As in the North Halmahera languages, the possessive markers in Inanwatan have optional arguments that the object markers lack. In Adang and Abui, two Alor-Timor-Pantar languages, we find that person prefixes marking possessor on nouns also cross-reference objects, though not subjects. In Adang only a particular subset of verbs is inflected: Adang (76) N-e na-fel mi habu. 1SG-GEN 1SG-ear COMP wide ‘My ears are wider (e.g. than yours)’ Adang (77) Sa na-tan. 3SG.NOM 1SG-ask ‘S/he asked me.’
(Haan 2001: 134)
(Haan 2001: 46)
Possessive clauses in East Nusantara, the case of Tidore
321
But in Abui there are even two sets of possessor prefixes that correspond to two sets of second argument markers. The prefix for inalienables in (78a) corresponds to the ‘patient’ marker in (78b), and the prefix for alienables in (79a) corresponds to the ‘locative’ category in (79b), which includes beneficiaries and ‘themes’ and which may be thought of as locations (Kratochvíl in progress). Abui (78) a. Na-min 1SG.INAL-nose ‘My nose.’ b. Me na-dak-e. come 1SG.PAT-clutch-IPFV ‘Come and hold me tight.’ Abui (79) a. Ne-wil. 1SG.AL-child ‘My child.’ b. Simon ne-l to-ha-loi. Simon 1SG.LOC-give DISTR.REC-3ii.PAT-chase ‘Simon chases me.’ And further east in the Cenderawasih Bay inalienable nouns in Yawa have a prefix identical to the undergoer prefixes used on transitive verbs and stative verbs. Yawa (80) In-aneme. 1SG.OBJ-hand ‘My hand.’
(Jones 1986; see also Klamer et al. 2008)
Summarising, in a number of East Nusantara languages, the possessive construction involves an (erstwhile) verb, but in the majority the possessor is marked directly on the possessum by means of an affix that is also used to mark either subject or object in the clause. Where Tidore differs is that the possessor marker does not correspond to the subject marker (and there is no object marker) and at the same time the possessive construction displays clausal properties even though there is no trace of an erstwhile verb, as in Buru and Biak.
322 Miriam van Staden 8. The North Halmahera family In the final part of this chapter I compare the Tidore possessive construction to the closely related languages of the North Halmahera family. The first obvious question is whether a clear distinction between attributive and predicative possession can be made in these languages, whether the possessive constructions in these languages, too, appear to be clausal. But the other North Halmahera languages are also interesting because they appear to be rather conservative compared to Tidore in a number of respects, including the systems of person cross-referencing on both verbal predicates and possessum noun phrases. It is argued here that these systems of crossreferencing have some important clues to a possible account of the presentday rather unique distinction between actor prefixes and possessor prefixes in Tidore. The North Halmahera family is usually presented as consisting of ten languages.10 However, six of these are so closely related that they may be considered dialects of one language ‘Northeast Halmaheran’ (Voorhoeve 1988): Galela, Tobelo, Tobaru, Pagu, Loda, Modole, and similarly Ternate and Tidore are dialects of one language (cf. Figure 2). The Northeast Halmahera languages are the most archaic in terms of retention of sounds from proto-Halmaheran (Wada 1980), SOV word order and postpositions, use of articles and noun class markers, the spatial deictic system, and also the occurrence of object prefixes, which in Tidore as well as in Ternate and West10
The data for this section were taken from various sources, most of them dating from the early 20th century, but some more recent studies are also available. Where the reports are in disagreement I give specific references. For details on Galela see Van Baarda (1908) and Shelden (1998, 1991), on Tobelo see Hueting (1936) and Holton (2006, 2003), on Pagu Wimbish (1991) and on Tabaru Fortgens (1928). For comparative studies see in particular Van der Veen (1915); Capell (1975); Voorhoeve (1988, 1987, 1987); and Wada (1980). I have omitted from the presentation some details on phonetic realization, e.g. differences between long and short vowels and stress where they are non-pertinent to the present discussion. Galela has a distinction between a dental /d/ and a retroflex // which are represented by the single grapheme d in this paper. I have changed the glosses of some of the examples to align the terminology or I have added glosses where the original sources (usually the older ones) do not give them. Also, most older sources as well as some of the more recent ones (e.g. Wimbish) give the verbal markers and the possessive markers as separate words, but there is enough morpho-phonological evidence to warrant their presentation as prefixes in all North Halmahera languages.
Possessive clauses in East Nusantara, the case of Tidore
323
Makian (van der Veen 1915) have been lost. The possessive constructions in all the North Halmahera languages show strong resemblance, although again the Northeast Halmahera languages appear to be more conservative than Tidore.
Figure 2. The North Halmahera languages
8.1. Clausal analysis in the North Halmahera languages The first question is to what extent the possessive constructions in the other North Halmahera languages have clausal properties like Tidore. Although this issue has not been addressed for any of the languages in the literature, text material shows that the possessive construction may be used predicatively in at least several different North Halmahera languages, e.g. in Tobelo:
324 Miriam van Staden (81) Ma-koano ngo-i-hekata ya-tumidi, NM-king NM-3SG.M.POS-wife 3N.SU+3PL.OBJ-seven ya-butanga manga ngohaka o nauru ya-butanga. NM male 3N.SU+3PL.OBJ-six 3N.SU+3PL.OBJ-six MANGA child ‘The king had seven wives, six (of them) had six sons.’ (Hueting 1936: 373) Because the possessed item is female, the possessive pronoun is ngo-i rather than a-i cf. also (Hueting 1936: 353). Like other clausal arguments the possessor noun phrase may be postposed as in the Pagu example (82): Pagu (82) ma kolan awi-ngoak mosoles dauk ART king 3M.POSS- child bachelorette there:south m-o-matetengo, ami-penjagaan o macan de o garuda … 3SG.F-SU-one 3SG.F.POSS-guard ART tiger and ART eagle ‘The king’s only daughter is there, and she has a tiger and an eagle for guards.’ (Wimbish 1991: 136) Furthermore, the possessive constructions are frequently translated as having two readings: a predicative and an attributive one, as in Tobelo o tau mahailoa ‘a beautiful house (the house has beauty), the beauty of the house’ and o gakana ma-doto ‘a sharp knife, the sharpness of the knife’ (Hueting 1936: 368). Also in support of a clausal analysis is the occurrence of adjectives in this construction and the ensuing ambiguity between a phrasal and a clausal reading as in the following Galela example: (83) Awi-lamo. 3SG.M.POS-big ‘His bigness; how big he is!’
(van Baarda 1908: 38)
There is, however, no data on possessive constructions in complementation or on the extent to which possessive constructions pattern with clauses or noun phrases of the kind described for Tidore (e.g. the complementation of untuk). This means that the data so far show that as in Tidore there is no clear distinction between a predicative and attributive possessive construction, but whether the constructions in the other North Halmahera languages also pattern with clauses rather than noun phrases is still unclear.
Possessive clauses in East Nusantara, the case of Tidore
325
8.2. Personal and possessive pronouns The Northeast Halmaheran languages and Sahu all have pronoun inventories that are very much like the pronouns in Tidore and Ternate. The personal pronouns are without exception historically morphologically complex as shown in Table 1 where the roots of the pronouns are underlined: Table 1. Personal pronouns in Tidore and Tobelo (Northeast Halmahera) singular
plural
Tidore
Tobelo
Tidore
Tobelo
1 EXC 1 INCL
ngo-ri –
ngo-hi –
ngo-m ngo-ne
ngo-mi ngo-ne
2
ngo-na
ngo-na
ngo-n
ngi-ni
3M 3F
u-na mi-na
u-nanga mi-nanga
o-na
o-nanga
3 NH
e-na
e-nanga
e-na
e-nanga
The suffix na(nga) found with the third person pronouns has its origin in the demonstratives, cf. re-na(nga) ‘here’ and ge-na(nga) ‘there’. It may be typed as a ‘deictic pointer’. First and second person have a prefixing element ngothat also occurs before women’s names or kin terms in all North Halmahera languages, e.g. ngo-Desi ‘Desi’ (see e.g. van der Veen 1915: 194; van Staden 2000: 92; Visser and Voorhoeve 1987: 53). In Sahu, ngo is in opposition with a marker a before nouns referring to men, but this is not found in the other languages. Its presence on first and second person pronouns could be explained if ngo- is considered a kind of vocative element used to refer to those present in the discourse. However, the noun markers ngo- (and a- in Sahu) is not used before proper names or kinship terms when addressing a person (Visser and Voorhoeve 1987: 36). Where Tidore and Ternate (cf. Hayami-Allen 2001: 136–137) are rather different from the other languages is that they lack a set of possessive pronouns, which can be formed morpho-syntactically in the other members of the North Halmahera family. In the Northeast Halmahera languages as well as in Sahu the personal pronouns may be preceded by a specific marker to give what appear to be possessive pronouns:
326 Miriam van Staden Pagu ai-kadu dumoi, w-a-gogon-oka (84) Oli to-una so POS-3SG.M 3SG.M.POS-sack one 3SG.M-NH-hide-NFUT ‘So he hid his sack (lit. so his sack, he hid it).’ (Wimbish 1991: 158) Tobelo nenanga? (85) To-munanga ami-gakana POS-3SG.F 3SG.F.POS-knife this ‘Is this her knife?’
(Hueting 1936: 353)
In the Northeast Halmahera languages, this marker is to for all persons, with the exception of the 2PL form in Tobelo which takes ti. In Sahu the marker is tV, where V assimilates to the first vowel of the pronoun, e.g. to ngoi ‘my, mine’, ta ngana ‘your(s)’, ti ngini ‘your(s) (plural)’, but is /o/ before /u/, i.e. to mungana ‘her(s)’ (Visser and Voorhoeve 1987: 53). This possibility is completely absent in Tidore and Ternate. Van Baarda (1908: 69–70) considers to a preposition ‘of, for’, e.g. tongohi ‘mine or for me’, but most authors treat it as a prefix, e.g. Wimbish on Pagu (1991: 139–140), or give it as part of the base form as Hueting does for Tobelo (1936: 353). In Sahu the possessor marker preceeds the noun markers a and ngo, as in example (86), supporting rather the prepositional analysis, in which case it may be considered a semantic marker. It indicates that the noun phrase has the role of possessor: Sahu (86) Wala ge t-a Salaka house that POS-NM Salaka ‘That house is Salaka’s’
(Visser and Voorhoeve 1987: 54)
This particular prefix or preposition is rather interesting for two reasons. First, it raises the question whether perhaps this reveals an underlying locative schema for the possessive constructions in North Halmahera. Recall that Tidore has the locative prepositions te/toma and these could perhaps be related to this possessive to. But the second reason is the more intrigueing one. Perhaps these constructions with to reveal something about the nature of the possessive construction in the North Halmahera languages. If they form possessive pronouns, do these occur attributively or predicatively? Or is there no distinction here, as in the Tidore possessive construction? And if Van Baarda is correct and to is a semantic role marker rather than a possessive pronoun deriving prefix, does this support the analysis of possessive constructions as predications?
Possessive clauses in East Nusantara, the case of Tidore
327
To start with the first question on the relation between locative prepositions te and toma in Tidore and possessive to, it is indeed tempting to see the two as related. This would then suggest an instance of locative-topossessive grammaticalisation (Heine 1997). However, the status and history of such locative prepositions in the North Halmahera languages is unclear. Sahu has a preposition toma11 but its distribution has not been described and it is not in opposition with te but with ra (John Severn, pers. comm.). A form to occurs infrequently in Tabaru (Fortgens 1928: 389–390) to derive verbs with a locative component, e.g. toobiri to arrive by night, and Fortgens suggests that the many names of languages beginning in to (Togutil, Tobelo, Tabaru and even Tidore, regularly pronounced as Todore) may also find their origin in a to- prefixed root. But signs of a locative preposition to are completely absent in other Northeast Halmahera languages such as Pagu, and are generally unexpected in SOV languages. In relation to the second question, pronouns preceded by to may occur dependently and independently. In other words, they may occur before noun phrases as apparent attributes (my, your, etc.), or they may constitute noun phrases on their own (mine, yours, etc.) as in the following example: Tobelo (87) to-ngohi ahi-tiwi nenanga? koali, to-ngohi. POS-1SG 1SG.POS-money this no POS-1SG ‘Is this my money? No, (it’s) mine.’ (Hueting 1936: 353) In all the North Halmahera languages that have to/tV it also occurs on the question word particle, as in (88): Tobelo (88) to-nago-nanga manga-tau nenanga? pos-who-3P 3P.POS-house this ‘Whose house is this?’
(Hueting 1936: 355)
The marker is, however, not obligatory – at least not in all languages and in all contexts. What precisely determines its presence is still not entirely clear. In Galela it is absent in contrastive expressions and after conjunctions or adverbs (van Baarda 1908: 70), but not in Pagu and Loda. And in Pagu it is 11
Visser and Voorhoeve (1987) do not mention it in their sketch grammar although it occurs in one of the examples in contrast to re the preposition for company / instrument (cf. Tidore se) (p.56).
328 Miriam van Staden absent when the possessor pronoun is preceded by the company/instrument preposition de ‘and, with’, as in (89): Pagu (89) … wa-make nage de ami lomang-oka. 3M.SG.A:3NH.O-see that with 3SG.F.POS name-N.FUT ‘… he saw that it had her name (lit. he saw it already with her name).’ (Wimbish 1991: 133) Also in assertion of possession the pronouns again do not always take to (90), although they may (91): Pagu (90) Ei ngoi ai-ngoak iwa. hey 1SG 1SG.POS-child gone ‘Hey, I don’t have any children.’
(Wimbish 1991: 135)
Pagu (91) To-ngoi naga alat-oka. POS-1SG ‘is’ tool-NFUT ‘my tool is already here.’ (lit. mine is tool-already) (Wimbish 1991: 135) Another question is whether this prefix or preposition is part of the pronominal paradigm or whether it is an element that occurs in other environments as well. In Sahu, it occurs with pronouns as well as nouns to ‘emphasize the possessorship’ (Visser and Voorhoeve 1987: 53), and although the other studies do not make explicit reference to the occurrence of to outside the pronominal paradigm, textual examples show that this is possible in at least several of the Northeast Halmahera languages, as e.g. in the following example from Tobelo: Tobelo (92) Nenanga ma-hangaji ai-ja? Koali, this 3NH.POS-district.head 3SG.M.POS-fishing.net NEG, to ma-uku. POS 3NH.POS-hukum ‘Is this the district head’s fishing net? No, the hukum’s.’ (Hueting 1936: 353) Clearly, on noun phrases to is not obligatory as (93) shows:
Possessive clauses in East Nusantara, the case of Tidore
329
Pagu mosoles dauk (93) Ma kolan awi-ngoak ART king 3SG.M.POS-child bachelorette there:south m-o-matetengo, ami-penjagaan o macan de o garuda. 3SG.F.POS-guard ART tiger and ART eagle 3SG.F-S-one ‘The king’s only daughter is there, and she has a tiger and an eagle for guards.’ (Wimbish 1991: 136) Summarising, all North Halmahera languages except for Tidore and Ternate have a marker to that occurs before the personal pronouns, and at least in some languages it also occurs before possessor noun phrases. In most languages the distribution of to is restricted at least in part by the occurrence of adpositions and adverbs, and in Sahu discourse considerations also play a role in its distribution. It appears then that to does not so much derive possessive pronouns as mark the semantic role of the possessor in the possessive construction. This ties in with a predicative analysis for these languages also. The marker distinguishes possessor ‘first arguments’ from other first arguments as a distinct semantic role in the clause, giving the possessive relation as a special event type. But is there further support for predicative possession also in the other North Halmahera languages? And how does the possessive construction in these languages shed any light on the Tidore situation? In order to address this question it is necessary to examine the possessive prefixes and the subject and object markers in the North Halmahera languages in some detail.
8.3. Possessive and other person prefixes in the North Halmahera languages The possessive prefixes in the North Halmahera languages all bear close resemblance. The differences that we do find are, however, rather significant. Also the differences in the systems of verbal cross-reference marking between Tidore on the one hand and the Northeast Halmaheran languages on the other is important and may shed light on the curious synchronic situation in Tidore. In this section I explore the possible origin of the possessive markers in Tidore as related to the object cross-reference markers that are found in the Northeast Halmaheran languages but not in Tidore. Table 2 gives an overview of the possessive prefixes in a number of North Halmahera languages. The bracketed elements are optional, often characteristic of informal, spoken language.
330 Miriam van Staden Table 2. Possessive marking in several North Halmahera languages possessive prefixes Tidore
Tobelo
Tabaru
Galela
Sahu
singular 1 2 3M 3F 3 NH
ri ni i mi maa
ahi ani ai ami ma
ai (a)ni (ani) wi (a)mi ma
(a)i (a)ni (a)wi (a)mi ma
(a)ri (a)ni ai (a)mi ma
plural 1 EXC 1 INCL 2 3 H.PL 3 NH
mi na ni na ma
mia nanga nia manga ma
mi(a) minga nanga ni(a) ninga manga ma
amia nanga ania manga ma
minga nanga ninga ma(nga) ma
a
ma is incongruous in all aspects. It is likely that it has its origins outside the possessive paradigm, possibly as a stative verb marker (see Holton 2006) and then saw its distribution widened to include relations of associated and relational possession (and reflexivity). This use then developed into the distinction between human plural and non-human possessors, which is otherwise absent in the paradigms of person affixation.
Table 3. Pronouns, subject and object markers in the North Halmahera languages Pronouna
subject prefix
object prefixb
singular 1 2 3M 3F 3 NH
ri/hi na u mi e
tonowomo-
(h)ini(w)imia/ia/ya/yo-
plural 1 EXC 1 INCL 2 3 H.PL 3 NH
m/mi ne n/ni o/a e
mip/h/w/foniyoyo-
minania-/yo-/kia/ ia/ya/yo-
a b
This column gives the roots from the forms in Table 1. Slashes indicate different forms in different languages. Not in Tidore, Ternate and West-Makian.
Possessive clauses in East Nusantara, the case of Tidore
331
8.3.1. Prefix aIf we compare the forms in Table 2 to the pronominal roots, subject and object prefixes in Table 3, it transpires that it is in particular the addition of the prefix a- that distinguishes the possessive prefixes from the object prefixes and to a lesser extent from the roots of the personal pronouns. Van der Veen (1915: 195) following Schmidt (1895: 134) indeed suggests that the possessive forms have been derived from the object markers, which would in turn have been derived from the personal pronouns (van der Veen 1915: 47; Schmidt 1895: 138). We return to this issue below. Clearly suffixing -a on the plural forms is not related to the prefix a-, as the Galela paradigm shows where the two co-occur, e.g. 1EXC a-mi-a. Van der Veen (1915: 195–196) and Schmidt (1895) suggest that the suffix nga on the third person plural derives from an ‘old plural marker’. But on the other plural forms minga ‘1PL.EXC’ and ninga ‘2PL’ it would occur by ‘false analogy’ because the root morphemes are already distinct from the singular forms and do not need the nga to disambiguate. But, it is as likely that if nga is indeed an old pluralizer, this has been reduced to a in most of the forms, including mia and nia, but survived in those forms where reduction would yield homonymy. In a number of languages the prefix a- in the possessive prefixes is optional, as indicated by the brackets, and in some languages it is omitted altogether. Its origin is unclear. Van der Veen (1915: 195) again follows Schmidt (1895) treating it as deriving from the masculine and neuter noun marker o because the two would not co-occur, and because whenever the possessor is female, a- is replaced by ngo, which is otherwise the feminine counterpart of o. This analysis cannot be correct. First, the position of the article is before the feminine noun marker, while the possessive prefix follows it. Compare Tobelo ani ‘2SG.POS’ before a masculine possessum versus ngoni ‘2SG.POS’ before a feminine possessum (Hueting 1936: 353). Furthermore, in several languages, including Galela and Tobelo, the article o and the female noun marker ngo do co-occur, although the article and a do not, e.g. o ngo Luri ‘the woman Luri’ in Tobelo (Hueting 1936: 353) and o ngo-meme ma-roka ‘mother’s husband’ in Galela (van Baarda 1908: 30). And finally there is no further support for a sound change from a to o. The absence of the article before possessum noun phrases is then more likely the result of phonological elision. It is notable, however, that this a- is related to the -a- that is found in the object.
332 Miriam van Staden 8.3.2. Possessor and object markers Returning to the similarity of the possessor prefixes to the object prefixes, the analogy between the forms suggests a historical derivation of the former from the latter. This would tie in with the general typology of the languages of East Nusantara. As we have seen, in many languages in this area the possessive construction involves markers derived from either the subject or the object verbal cross-reference markers. At the same time, this historical derivation would provide an explanation for why the possessive constructions in these languages have clausal properties. However, the evidence for derivation from object prefixes is not immediately straightforward. The possible origin of the possessive marker proposed in this section is tentative and requires a closer study of the history of the languages of North Halmahera as well as a closer examination of the systems of verbal crossreferencing in the Northeast Halmahera languages to substantiate the hypothesis put forward here. First, we need to examine what it is about the object markers that makes them resemble possessive markers more than the other markers. Indeed, if we compare just the roots of the personal pronouns, then all verbal crossreference prefixes and the possessive markers resemble each other since they all derive from the personal pronouns. The reason the possessor prefixes resemble the object markers more than the subject markers is because of the irregular subject markers for 1SG (to) and 1PL.INCL (po), where the object markers are not irregular, and because the subject prefixes end in o rather than i. But the inclusive/exclusive opposition in the North Halmahera languages is generally attributed to language contact with surrounding Austronesian languages (Klamer et al. 2008). The irregularity of the 1PL.INCL form may then be the result of a later introduction of the forms into the system. The irregular 1SG form is still unexplained, but it is notable that the same form is found throughout the Papuan languages of the West Papuan Phylum, including Maybrat, Moi, and Abun. The final vowel of the subject prefix is likely a separate morpheme. This has been argued, for instance for Pagu (Wimbish 1991). The forms in -o- indicate that there is only a subject or that the object is either unspecified or human. As such it is in opposition with a prefix ending in -a- , which indicates that the object is non-human, directly affected by the verb and specified or definite (i.e. it occurs with canonical transitive verbs).
Possessive clauses in East Nusantara, the case of Tidore
333
Pagu (94) t-o-sano 1SG-SU-ask ‘I ask (something)’
(Wimbish 1991: 46)
Pagu (95) t-a-sano 1SG-NH-ask ‘I ask a specific question’
(Wimbish 1991: 46)
This means that if we disregard the inclusive form and consider the endings in -o- to be separate morphemes, only the aberrant first person singular subject prefix is different from the object prefix and the possessive marker. So then the question is whether the possessive prefix would be derived from the object cross-reference marker or whether it could have derived directly from the pronominal roots. Yet there is another link to the object cross-referencing paradigm. In the North Halmahera languages, the possessive markers typically had a prefix aand it is possible that this prefix is the same a- as the one occurring before the object markers when these refer to referents that are non-human, directly affected by the verb and specified or definite. One reason to assume this is that at least according to Wimbish’ account, the Pagu language does not allow for the possibility of having human denoting objects that are affected by the event, and the other Northeast Halmaheran languages appear to be similar in this respect. In the following paragraphs, I will try to make a plausible case that the semantic role of possessor could be just this: a human object that is affected by the event. If that is the case, then the generalisation would be that all affected objects take a and all unaffected objects have o, but how does that explain that there are no transitive verbs that have affected human objects? The hypothesis is then that there is a restriction on participant structure which stipulates that each predicate may have only one argument that is actively involved in the event, either as an agent or as a patient. This means that if there is a human agent, the patient cannot be an affected human. But if there is no agent, then this leaves the possibility of having an affected human object. And this is the possessive construction: a predication without a subject, but with an affected human object. The argumentation is as follows. First, we follow Wimbish and assume that the subject prefixes are indeed historically bi-morphemic, consisting of the pronominal roots followed by o in intransitives of when the second argument is either human or not affected by the predicate (“oblique”) as in Table 4.
334 Miriam van Staden Table 4. Morphological analysis of subject markers singular
plural
person ‘intransitive’ fused root marker form
person root
1EXC 1INC
h/ri
o
to
mi ngaa n/f/w/p/h
o o
mi/miyo fo
2
ni
o
no
ni
nga
o
ni/niyo
3M 3F 3NH
u/w mi e
o o o
wo mo yo
i/y
nga
o
yo/i/ya
a
pluralizer ‘intransitive’ fused marker form
nga is the form of the pluralizer that survives today, but it is actually more likely that it was ka or /a, as evidenced by the various languages that have a /k/ in the 3pl object prefixes.
For the irregular plural forms we assume that the old pluralizer, now lost in the subject prefixes but still visible in the possessive markers in several languages, prevented the kind of fusion that took place with the singular forms, explaining their irregular endings in –i, as in Tidore, and the forms miyo and niyo in the Northeast Halmahera languages. These subject markers followed by o are again followed by the pronominal roots cross-referencing the object when this is human. Fusion of subject and object prefix takes place in a number of cases, reducing for instance the 3NH object pronoun e to ‘zero’. In the case of a non-human affected object (direct object), the marker is not o but a. Otherwise the process is identical, giving an identical fused paradigm with the one presented in Table 4. This account gives a ‘defective paradigm’ as in Table 5 as it appears to exclude the possibility of affected human arguments: Table 5. -o- and -a- marking in Pagu
single argument unaffected non-human object unaffected human object affected non-human object affected human object
-o-
-a-
+ + + – ?
– – – + ?
Possessive clauses in East Nusantara, the case of Tidore
335
But what if the initial a- on the possessive markers is reinterpreted as a marker of affected human objects? In that case the constructions are like experiential constructions to the extent that there is no subject. This could be explained as a constraint on the number of verbal arguments that may be directly involved in the event. Only one argument may be involved either as an agent (subject) or as a possessor (object). If there is a subject, the interpretation of the predication is ‘eventive’ with possibly an indirectly affected human object, or a non-human object which is never as involved as a human argument. In experiential constructions, the subject is a third person plural ‘dummy’ and the non-affected object is the experiencer. When there is no subject, the interpretation of the predication will be ‘possessive’, with the object expressing the possessor. There is one small hitch in Galela. In Tabaru and in Pagu, the subject markers for 1EXC and 2PL are always miyo and niyo, regardless of whether there is an unaffected non-human object or not. However, in Galela there is an opposition, so that mi and ni are found with intransitive verbs and miyo and niyo with transitive verbs that take unaffected non-human objects (and of course miya and niya with affected non-human objects).12 One possibility is that this is evidence for different paradigms for transitive and intransitive verbs, whereby neutralisation of intransitive subject and transitive subject followed by an unaffected 3NH object has taken place for all persons and numbers, except for these two. However, I propose that the opposite has occurred: in Galela the intransitive forms miyo and niyo have been reanalysed as bimorphemic mi-yo and ni-yo in which the first element expresses subject and the last object. In intransitive predications, the ‘object’ prefix was then deleted. Tabaru and Pagu then give evidence of the original system. What this means is that there is a way in which the similarity between object prefixes and possessor markers in the North Halmahera languages can be accounted for historically. This links the North Halmahera languages to many other languages of East Nusantara, and it may explain the clausal properties that possessive constructions in Tidore have. In various ways the possessor markers are more archaic: they have retained the affected object marker a-, but also the pluralizer -a still shows in most North Halmahera languages, though not in Tidore.
12
Sheldon (1991) discusses only miya and niya, but Van Baarda also mentions the other two forms.
336 Miriam van Staden 9. Conclusion Tidore like all North Halmahera languages is ‘verby’ in nature. Although categories may be established the language allows virtually all categories to fulfil the function of main predicate in a clause. What are categorially nouns or adjectives or even interjections may occur as predicates in clauses, in which case they receive first argument cross-referencing. There is a tendency to see such processes as instances of nominalisation or denominalisation, but this is unnecessary. In fact, it obscures the observation that syntactic templates are not defined in terms of categories but in terms of functions. Possessive constructions predicate a possessive relation, regardless of whether the construction as a whole is used to refer or to ascribe a relation. This is deferred to pragmatics as the way in which a speaker employs the construction in discourse. The result in Tidore is a split in the system of argument cross-referencing distinguishing actor arguments from possessive arguments. The first type comprises all the traditional first argument types from controller of activity to experiencer or theme (cf. Dik 1997: 56); the second type is the possessor first argument. A second result of this analysis is that there is no clear distinction between attributive and predicative possession in Tidore. Finally, the present analysis of the Tidore system explains the clausal properties that the construction has. Although this construction is typologically rather unique, in this paper it was shown that Tidore does not stand alone among its relatives, but that it is in fact an areal pattern to have possessor markers that correspond to clausal argument markers and even the clausal properties of possessive constructions are shared in other unrelated languages of the same linguistic area.
References Allen, W. Sidney 1964 Transitivity and possession. Language 40 (3): 337–343. van Baarda, M. J. 1895 Woordenlijst Galelareesch-Hollandsch. Met Ethnologische Aanteekeningen, op de Woorden, die Daartoe Aanleiding Gaven. Den Haag: Koninklijk Instituut voor Taal-, Land- en Volkenkunde van Nederlands Indië. van Baarda, M. J. 1908 Leidraad bij het Bestudeeren van ’t Galela’sch Dialect, op het Eiland Halmaheira. Den Haag: Nijhof.
Possessive clauses in East Nusantara, the case of Tidore
337
Baird, Louise 2002 A grammar of Keo: an Austronesian language of East Nusantara. PhD thesis, Australian National University. Bowden, John 2005 Language contact and metatypic restructuring in the directional system of North Maluku Malay. Concentric: Studies in Linguistics 31 (2): 133–158. Capell, Arthur 1975 The West Papua Phylum: General, and Timor and the areas further West. In New Guinea Area Languages and Language Study. Volume 1: Papuan languages and the New Guinea Linguistic Scene, Stephen A. Wurm (ed.), 667–716. Canberra: Pacific Linguistics. Chappell, Hilary and William B. McGregor (eds.) 1996 The Grammar of Inalienability: A Typological Perspective on Body Part Terms and the Part-Whole Relation. Berlin /New York: Mouton de Gruyter. Dik, Simon C. 1989 The Theory of Functional Grammar. Part I: The Structure of the Clause. Dordrecht: Foris. Dik, Simon C. 1997 The Theory of Functional Grammar. Part I: The Structure of the Clause. Second, revised edition, edited by Kees Hengeveld. Berlin / New York: Mouton de Gruyter. Dol, Philomena H. 1999 A Grammar of Maybrat: A Language of the Bird’s Head, Irian Jaya, Indonesia. PhD thesis, Leiden University. Foley, Willam A. 1986 The Papuan Languages of New Guinea: Cambridge Language Surveys. Cambridge: Cambridge University Press. Fortgens, J. 1928 Grammaticale aantekeningen van het Tabaroesch, Tabaroesche volksverhalen en raadsels. Bijdragen tot de Taal-, Land- en Volkenkunde 84: 300 –544. Gravelle, Gilles 2004 Meyah: an east Bird’s Head language of Papua, Indonesia. PhD thesis, Vrije Universiteit. Grimes, Charles E. 1991 A grammar of Buru. PhD thesis, Australian National University. de Groot, Casper 1983 On non-verbal predicates in Functional Grammar: the case of possessives in Hungarian. In Advances in Functional Grammar, Simon C. Dik (ed.), 93–122. Dordrecht: Foris. Haan, Johnson Welem 2001 The grammar of Adang: a Papuan language spoken on the Island of Alor East Nusa Tenggara – Indonesia. PhD thesis, University of Sydney.
338 Miriam van Staden Hayami-Allen, Rika 2001 A descriptive study of the language of Ternate, the Northern Moluccas, Indonesia. PhD thesis, University of Pittsburgh. Heine, Bernd 1997 Possession: Cognitive Sources, Forces and Grammaticalization. Cambridge: Cambridge University Press. Hengeveld, Kees 1992 Parts of speech. In Layered Structure and Reference in a Functional Perspective, Michael Fortescue, Peter Harder and Lars Kristoffersen (eds.), 29–55. Amsterdam: John Benjamins. Hengeveld, Kees and J. Lachlan Mackenzie 2008 Functional Discourse Grammar. Oxford: Oxford University Press. Holton, Gary 2003 Tobelo. Languages of the World /Materials, Vol. 328. Munich: Lincom Europa. Holton, Gary 2006 The relational noun marker in Tobelo (Northeast Halmaheran). Tenth International Conference on Austronesian Linguistics (10-ICAL). Puerto Princesa City, Palawan, Philippines. Hueting, A. 1936 Iets over de spraakkunst der Tobeloreesche taal. Bijdragen tot de Taal-, Land- en Volkenkunde 94 (3): 295–407. Jelinek, Eloise and Fernando Escalante 1988 ‘Verbless’ possessive sentences in Yaqui. In In Honor of Mary Haas: From the Haas Festival Conference on Native American Linguistics, William Shipley (ed.), 411–429. Berlin/New York: Mouton de Gruyter. Jones, Linda K. 1986 The question of ergativity in Yawa, a Papuan language. Australian Journal of Linguistics 6: 37–55. Klamer, Marian A. F. 2003 ‘Report’ constructions in Kambera (Austronesian). In Reported Discourse: A Meeting Ground for Different Linguistic Domains, Tom Güldemann and Manfred von Roncadorm (eds.), 323–340. Amsterdam: John Benjamins. Klamer, Marian A. F., Ger P. Reesink and Miriam van Staden 2008 East Nusantara as a Linguistic Area. In From Linguistic Areas to Areal Linguistics, Pieter Muysken (ed.), 95–149. Amsterdam /Philadelphia: Benjamins. Kratochvíl, Frantiek 2007 A grammar of Abui: A Papuan Language of Alor. PhD Thesis, Leiden University.
Possessive clauses in East Nusantara, the case of Tidore
339
Lichtenberk, Frantisek 1985 Possessive constructions in Oceanic languages and in Proto-Oceanic. In Austronesian Linguistics at the 15th Pacific Science Congress, Andrew Pawley and Lois Carrington (eds.), 93–140. Canberra: Pacific Linguistics. Lichtenberk, Frantisek 2002 The possessive-benefactive connection. Oceanic Linguistics 41 (2): 339–474. Margetts, Anna 2004 From implicature to construction: emergence of a benefactive construction in Oceanic. Oceanic Linguistics 43 (2): 445–468. Menick, Raymond 1995 Moi: a language of the West Papuan Phylum. a preview. In Tales from a Concave World: Liber Amicorum Bert Voorhoeve, Connie Baak, Dick van der Meij and Mary Bakker (eds.), 55 –73. Leiden: Projects Division, Department of Languages and Cultures of South-East Asia and Oceania. van Minde, Don 1997 Malayu Ambong: Phonology, Morphology, Syntax. Leiden: CNWS. Odé, Cecilia 2002 A sketch of Mpur. In Languages of the Eastern Bird’s Head, Ger P. Reesink (ed.), 45–107. Canberra: Pacific Linguistics. Reesink, Ger P. 1998 The Bird’s Head as a Sprachbund. In Perspectives on the Bird’s Head of Irian Jaya, Indonesia, Jelle Miedema; Cecilia Odé and Rien A.C. Dam (eds.), 603–642. Amsterdam: Rodopi. Reesink, Ger P. 1999 A Grammar of Hatam. C–146. Pacific Linguistics. Canberra: ANU. Reesink, Ger P. 2000 West Papuan languages: roots and development. Paper presented at Papuan Pasts, Canberra. Reesink, Ger P. (ed.) 2002 Languages of the Eastern Bird’s Head. Canberra: Pacific Linguistics. Ross, Malcolm 1996 Contact-induced change and the comparative method: cases from Papua New Guinea. In The Comparative Method Reviewed: Regularity and Irregularity in Language Change, Mark Durie and Malcolm Ross (eds.), 180–217. New York: Oxford University Press. Ross, Malcolm 1997 Social networks and kinds of speech community events. In Archaeology and Language, Roger M. Blench and Matthew Spriggs (eds.), 209–261. London: Routledge.
340 Miriam van Staden Ross, Malcolm 2001 Contact-induced change in Oceanic languages in North-West Melanesia. In Areal Diffusion and Genetic Inheritance, Alexandra Y. Aikhenwald and Robert M.W. Dixon (eds.), 134–166. Oxford: Oxford University Press. Ross, Malcolm 2006 Metatypy. In The Encyclopaedia of Language and Linguistics, Vol. 8, Keith Brown (ed.), 95–99. New York: Elsevier. Schmidt, Pater Wilhelm 1895 Die Sprachlichen Verhältnesse von Deutsch-Neuguinea. Zeitschrift für Afrikansiche, Ozeanische und Ostasiatische Sprachen II (4). Shelden, Howard 1991 Galela pronominal verb prefixes. In Papers in Papuan Linguistics, Tom Dutton (ed.), 161–175. Canberra: Pacific Linguistics. Shelden, Howard 1998 Transitivity and Galela pronominal reference. In SIL Electronic Working Papers SILEWP 1998-005. (http://www.sil.org/silewp/ 1998/005/) Dallas, Texas: SIL International. van Staden, Miriam 2000 Tidore: A linguistic description of a language of the North Moluccas. PhD thesis, Leiden University. van Staden, Miriam 2006 The body and its parts in Tidore, a Papuan language of Eastern Indonesia. Language Sciences 28: 323–343. van Staden, Miriam and Ger P. Reesink 2008 Serial verb constructions in a linguistic area. In Serial Verb Constructions in Austronesian and Papuan Languages, Gunter Senft (ed.), 17–54. Canberra: Pacific Linguistics. Taylor, John R. 1989 Possessive genitives in English. Linguistics 27: 663–686. Taylor, John R. 1996 Possessives in English: An Exploration in Cognitive Grammar. Oxford: Oxford University Press. Tjia, Johnny 1997 Verb serialisation in Ambonese Malay. MA thesis, University of Oregon. Van den Heuvel, Wilco 2006 Biak. Description of an Austronesian language of Papua. PhD thesis, Vrije Universiteit. van der Veen, Hendrik 1915 De Noord-Halmahera'se taalgroep tegenover de Austronesiese talen. Leiden: van Nifterik.
Possessive clauses in East Nusantara, the case of Tidore
341
Verhaar, John W. M. 1995 Toward a Reference Grammar of Tok Pisin: An Experiment in Corpus Linguistics. Oceanic Linguistics Special Publication, Vol. 26. Honolulu: University of Hawai’i Press. Visser, Leontine E. and C. Lambertus Voorhoeve 1987 Sahu-Indonesian-English Dictionary and Sahu Grammar Sketch. Verhandelingen van het Koninklijk Instituut voor Taal-, Land- en Volkenkunde, Vol. 126. Dordrecht: Foris. Voorhoeve, C. Lambertus 1987 The Non-Austronesian languages in the North Moluccas. In Halmahera dan Raja Empat sebagai Kesatuan majemuk, E.K. M. Masinambow (ed.). Jakarta: Lembaga Ekonomi dan Kemasyarakatan Nasional, Lembaga Ilmu Pengetahuan Indonesia. Voorhoeve, C. Lambertus 1987 The masked bird: Linguistic relations in the Bird’s Head area. In Peoples on the Move, Paul Haenen and Jan Power (eds.), 78–101. Nijmegen: Centre for Australian and Oceanic Studies. Voorhoeve, C. Lambertus 1988 The languages of the North Halmahera Stock. Papers in New Guinea Linguistics 26: 181–209. Canberra: Pacific Linguistics. de Vries, Lourens J. 2004 A Short Grammar of Inanwatan, an Endangered Language of the Bird’s Head of Papua, Indonesia. Canberra: Pacific Linguistics. Wada, Yuiti 1980 Correspondence of consonants in North Halmahera languages and the conservation of archaic sounds in Galela. In The Galela of Halmahera: A Preliminary Survey, Naomichi Ishige (ed.), 497–529. Senri Ethnological Series, Vol. 7. Osaka: National Museum of Ethnology. Wimbish, Sandra Gay 1991 An Introduction to Galela through the Analysis of Narrative Discourse. Ann Arbor: University of Michigan Press. Wurm, Stephen A. 1982 Papuan Languages of Oceania. Ars Linguistica, Vol. 7. Tübingen: Gunter Narr Verlag.
Possessive expressions in the Southwestern Amazon Hein van der Voort
1. Introduction The Southwestern region of the Amazon river basin, which in the present article is understood to be the region that covers the Bolivian and Brazilian sides of the Guaporé, Mamoré and Upper Madeira rivers, is one of the world’s linguistically most diverse places. The Southwestern Amazon, here also referred to as the Guaporé region, harbours representatives of seven different linguistic stocks, including Arawak, Chapacura, Macro-Jê, Nambikwara, Pano, Tacana and the majority of the branches of the Tupi linguistic stock. Furthermore twelve language isolates or unclassified languages are spoken in the region. The majority of the approximately 50 languages of the region are highly endangered with extinction, about half of them having fewer than 50 speakers and a third ten or fewer. During the past two decades, a considerable number of initiatives to document and describe these languages have fortunately been taken. This article is meant to give a preliminary survey of the different ways adnominal possession is expressed across a sample set of languages from the Southwestern Amazon (see Map 1). In section 2 I will sketch the types of adnominal possessive expression found in Kwaza, a language isolate spoken by approximately 25 people. In section 3 I will discuss the similarities with possessive expressions in other languages of the region: Aikanã (isolate), Arikapú (Macro-Jê), Baure (Arawak), Kanoê (isolate), Latundê (Nambikwara), Mekens (Tupí) and Wari’ (Chapacura). The final section 4 contains a table summarising the different possessive strategies. In view of the fact that the languages discussed here belong to different families, and even include three isolates, explanations for certain similarities are probably found in areal diffusion. This article deals mainly with adnominal possession, which can be described as a grammatical construction involving a modifying noun and a head noun that expresses a possessive relationship between a possessor (PR) and a possessum (PM), respectively. It contrasts with predicative possession and external possession. Predicative possession constructions involve the verb ‘to have’, or another verbal copula, and express a possessive
344 Hein van der Voort relationship between a possessor argument and a possessum argument (e.g. Stassen 2006, 2009). External possession concerns an understood possessive relationship between entities that represent distinct arguments of a (any) verb that do not (necessarily) carry possessive marking, as in ‘He kissed her on the cheek.’ (e.g. Payne and Barshi 1999). Predicative possession will be dealt with only briefly, when relevant (sections 2.4, 3.1.2, 3.3.2), and external possession not at all, since they are not of central importance here and merit discussion in separate articles.
Map 1. Location of languages discussed (© Willem Doelman).
The possessive relationship is prototypically one of ownership, but it can also metaphorically express body part, kinship, availability, location, control and other more abstract relationships (e.g. Herslund and Baron 2001; Koptjevskaja-Tamm 2006; Lichtenberk this volume). Adnominal possession is often called ‘attributive’ possession (e.g. Baron et al. 2001), but in this article the term ‘attributive’ will for practical reasons be reserved for a modifying relationship between a head noun and a dependent modifying noun. Although the term ‘genitive’ can be used in a wider sense than just case marking (Dryer 2007), it will be avoided here.
Possessive expressions in the Southwestern Amazon
345
According to Nichols (1992: 71) South American languages show a preference for ‘head-marking’, which is confirmed for the Guaporé region (Crevels and van der Voort 2008: 171). However, the Guaporé languages do not unambiguously reflect Dixon and Aikhenvald’s (1999: 8) observation regarding Amazonian languages that possession is typically marked on the PM. As shown in section 4, three of the eight languages discussed in the present article canonically mark possession on the PR rather than on the PM. The issue of head- and dependent-marking is not a central concern in this article and it will be touched upon only peripherally, in sections 2.2.1 and 2.3. In this article I will approach possession primarily from the perspective of form rather than of function. Hence, I will focus on the forms that are dedicated to the expression of the possessive function, rather than elaborate on the range of different functions that possessive forms and constructions might have. Nevertheless, in order to analyse the origin and development of specific possessive forms in the little-studied languages under discussion, it will now and then be necessary to expand on alternative functions, usually in separate subsections (e.g. § 3.1.1, § 3.2.2, § 3.4.2). Also, when languages have no “dedicated” possessive expressions, such as Mekens and Arikapú, it will be necessary to explain how the possessive function is fulfilled. 2. Possessive constructions in Kwaza Kwaza, which has been studied and described since the mid 1990s by the author (e.g. van der Voort 2004), is a morphologically complex language with a variable word order, although there is a clear tendency towards headfinal structures. The grammatical categories of Kwaza are verbs, nouns, adverbs and particles. Most of the grammatical complexity is found in verbal morphology, whereas nominal morphology is relatively simple. However, the highly productive nominalisation possibilities enable the creation of morphologically very complex nouns. The morphology is mainly suffixing. The verb is obligatorily inflected for person and mood, although morphological ellipsis is attested under specific circumstances. The morphological structure of the verb is: root + optional derivation + obligatory person inflexion + obligatory mood inflexion, as shown in this example: 1, 2 1
The Kwaza consonants /c/ and /x/ are pronounced as IPA retracted [t] and [s] respectively. The vowel /y/ is pronounced somewhere between IPA [] and []. Main word stress is marked by an apostrophe ['] preceding the stressed syllable.
346 Hein van der Voort (1)
cari-'nã-da-ki shoot-FUT-1S-DEC ‘I will kill (it, e.g. the game animal).’
The verb that is unmarked for person is interpreted as having a third person subject. Overt expression of arguments as pronouns or other independent constituents is not obligatory. Nouns are not obligatorily inflected, there is no nominal number marking, and the animate object case is not always applied. The distinction between inclusive and exclusive person reference is found in both the pronominal and the bound person marking paradigms. There are several basic moods, including declarative, interrogative, volitive, imperative, negative imperative, etc. There are several moods for subordinated adverbial clauses and there is a cosubordinated mood for clause chains. Different nominalisers are used for attributive and complement clauses. In the following subsections I will sketch the different possessive strategies in Kwaza and analyse their morphosyntactic properties. 2.1. The canonical adnominal possessive construction There are no specific possessive pronouns or inflexions in Kwaza. The usual way to express an adnominal possessive relationship in Kwaza requires derivation of the PR noun by the morpheme -dy-, followed by the nominalising morpheme -h . This latter morpheme is probably the most frequently occurring morpheme of Kwaza. As will be seen in § 2.2.1 and § 2.2.2, -h functions among others in attributive clause formation and as a 2
The following abbreviations are used: ART – article, AUX – auxiliary, BEN – benefactive (verbal derivation), BER – beneficiary, C – coreferent, CAU – causative, CD – directional classifier, CL – classifier, COL – collective, COMIT – comitative, COP – copula, CSO – cosubordination, DEC – declarative, DEM – demonstrative, DET – detrimental, DR – directional, DS – different subject, F – feminine, FUT – future, IMPF – imperfective, INCL – inclusive, INFL – Wari’ ‘inflection’, INT – interrogative, LOC – locative case, M – masculine, N – neuter, NOM – nominaliser, OBL – oblique case, P – plural, PM – possessum; POS – possessive, POT – potential, PR – possessor; PROX – proximate, REF – referential, REFL – reflexive, S – singular, VOL – volitive. The numerals 1, 2, and 3 indicate respectively, first person singular, second person, and third person. The hyphen (-) indicates a morphemic boundary; the equal sign (=) indicates composition or a clitic boundary; and the period (.) separates semantic units in a portmanteau morpheme.
Possessive expressions in the Southwestern Amazon
347
semantically neutral classifier. Other ways to express possessives in Kwaza are less productive. The most frequently attested constituent order in possessive constructions is PR–PM: (2)
tawi'wi-dy-h a'xy Tavivi-POS-NOM house ‘Tavivi’s house’
Personal pronouns can also occur as PRs: (3)
'xyi-dy-h 'kopu you-POS-NOM cup ‘your cup’
The only exception to the pattern described here is the first person inclusive pronoun txana, in the possessive use of which omission of the morpheme -dy- is preferred, i.e. txana-h ‘ours (INCL)’ rather than txana-dy-h . Headless possessive constructions, in which an explicit PM is absent, are frequently attested in Kwaza: (4)
na-'ay-h -dy-h
i'si-ki die-DEC ‘That person’s (son) has died.’
PROX-that-NOM-POS-NOM
Note that the PR in the above example is a demonstrative expression. Since demonstrative roots are verbal, the structure of demonstrative expressions requires a nominaliser, hence the repeated occurrence of -h in a demonstrative PR. The use of bare or derived nominal stems as predicates (called ‘zeroverbalisation’ in van der Voort 2004) is very productive in Kwaza. It may also involve (headless) possessive constructions in different ways: (5)
'si-dy-h -ki ('kopu) I-POS-NOM-DEC cup ‘(The cup) is mine/it is my (cup).’ 3
(6)
'si-dy-h 'kopu-ki I-POS-NOM cup-DEC ‘The cup is mine./It is my cup.’
3
Reverse constituent order is also possible.
348 Hein van der Voort 2.2. Morphological characteristics of the canonical possessive construction It is not so clear whether the morpheme -dy- should be considered derivational or inflexional. The morpheme is inflexional according to the criterion that it does not change the semantics of the word to which it attaches. Rather, it indicates the syntactic function of the word in the clause. Also the characteristic position of stress, on the ultimate syllable of the (derived) root, suggests that -dy- belongs to the inflexional layer of the word. However, the fact that the morpheme -dy- does not form part of a paradigm suggests that it may be a derivational affix. In addition, the obligatory presence of the nominaliser -h suggests that -dy- has caused a category change of the preceding noun. This adds further support to the derivational analysis. The nominalising morpheme -h is one of the most productive and frequently occurring morphemes of Kwaza. It typically creates deverbal nouns that can function as arguments and attributes. Besides functioning as a straightforward nominaliser, it occurs with temporal and aspectual effects in recursively derived verbal constructions. Also it is often used as a semantically neutral classifier, in case a specific classifier is not employed (this will be explained in §2.2.2). The possessive construction may actually be regarded as one of the least expected habitats of -h . The obligatory occurrence of this element in possessive constructions may be a reason to consider the combination -dy-h as lexicalised. Consider the following examples: (7)
'si-dy-h I-POS-NOM ‘mine’
(8)
*si-dy-ki I-POS-DEC ‘It is mine.’
If the element -h really does have a nominalising function under all circumstances, example (8) would be grammatical. In the following subsections, I will discuss the elements -h (§ 2.2.1, § 2.2.2) and -dy- (§2.2.3), in order to better understand the combination -dy-h .
Possessive expressions in the Southwestern Amazon
349
2.2.1. Similarities with the attributive construction The possessive construction has similarities with the attributive construction. On the structural level, attributive modification of nouns is expressed by juxtaposition of nominal constituents, as shown by the following example:4 (9)
ywy'nw duma'ru tree ladle ‘big wooden spoon’
In terms of parts of speech membership, attributively modifying lexemes in Kwaza are mostly verbs, for example: awy- ‘to be cold’, arwa- ‘to be new’, atsile- ‘to be heavy’, wotsu- ‘to be skinny’. Even quantitative expressions such as aky- ‘to be two’ (as in example (13)) are verbs. Hence, on the morphological level, attributive modification of nouns usually involves nominalisation: (10) a'xy arwa-'h house new-NOM ‘new house’ In fact, the nominalised verb can be morphologically very complex and contain derivational morphemes such as classifiers, and person inflexion. Therefore, no fundamental distinction can be drawn between relative clauses and attributive constructions. The contrast between the following examples shows that replacement of the word-final verbal mood marker by a nominaliser is enough to create a relative clause (see also van der Voort 2006): (11) atxitxi'n bar -'ri-da-ki pancake heat-CL:flat-1S-DEC ‘I baked the maize pancake.’
4
Even though the relatively rare modifier-noun word order could help support an analysis of (9) as a (classifying) compound, I chose to regard it as an attributive construction since, in contrast to verbal compounds, full nominal compounds are very rare in Kwaza. Furthermore, Kwaza classifiers are suffixes and cannot occur as free elements (van der Voort 2004: 128ff.).
350 Hein van der Voort (12) atxitxi'n bar -'ri-da-h pancake heat-CL:flat-1S-NOM ‘maize pancake which I baked’ Also, headless attributive constructions, including headless relative clauses, are frequently attested: (13) (a'le) aky-'h axe two-NOM ‘the two (axes)’ Since possessive constructions also contain a nominaliser, one might wonder whether they can be distinguished from attributive constructions at all. There are good reasons to assume there is a fundamental distinction, however. In the first place, there is a morphological difference between possessive and attributive constructions. Even though the possessive construction requires the use of a nominaliser, the preceding possessive morpheme -dydoes not appear to produce verb roots. As examples (7) and (8) above show, the nominaliser cannot be replaced by a mood marker. By contrast, morphologically complex attributive constructions such as (12) are probably derived from predicative constructions such as (11) in which a mood marker occurs instead of a nominaliser. Second, there is a marked difference in constituent order. As mentioned in section 2, in spite of the attested syntactic variability, the language tends to prefer head-final structures. Consequently, the great majority of the possessive constructions are head-final, whereas attributive constructions tend to be head-initial. An explanation of this unexpected attributive order is probably found in the predicational origin of the relative clause. Its formation is a purely morphological operation: a nominaliser is applied to a headfinal predicative expression like (11), which, without change of constituent order, results in a head-initial attributive expression like (12). Only in the case of underived noun-noun attributive constructions, as in example (9), head-final order is attested with some frequency. For the reasons discussed in this and the previous section, it is doubtful whether the possessive construction is a subtype of the attributive construction; moreover, it is possible that the element -h in the combination -dy-h does not function as a nominaliser. Indeed it may well be synchronically unanalysable.
Possessive expressions in the Southwestern Amazon
351
2.2.2. The occurrence of classifiers As mentioned above, the nominaliser -h can function as a neutral classifier. Conversely, classifiers can also have nominalising properties and can be considered as bound derivational morphemes. Kwaza classifiers are described and analysed exhaustively in van der Voort (2004). In this subsection I will discuss their behaviour in possessive constructions and their consequences for the analysis of -dy-h . Kwaza has an enormous arsenal of classifiers (over a hundred), that have a very wide distribution and that express semantic characteristics including shape, substance, texture, part-whole relations, direction and sex. Their actual meanings can range from highly abstract, such as ‘thing’ to uniquely specific, such as ‘spoon’. In principle, all classifiers have the same distribution. Moreover, in almost every place where the nominaliser -h can be found, a classifier can replace it. Hence, classifiers can occur in attributive, numeral, demonstrative, interrogative and adverbial expressions, and they can furthermore be attached to nouns and incorporated in verbs. They can also occur in possessive expressions. In Kwaza, classifiers describe properties of their referents and they may be used anaphorically. Note the following examples that contrast the use of a specific classifier with the use of -h as a neutral classifier: (14) ha-ha-'ro-ki clean-clean-CL:vessel-DEC ‘He washes pans.’ (15) ha-ha-'h -ki clean-clean-NOM-DEC ‘He washes things.’ In attributive constructions, a specific classifier can occur instead of -h : (16) (a'xy) aky-'xy (house) two-CL:house ‘two houses’ (17) aky-'h two-NOM ‘two (ones)’
352 Hein van der Voort The following examples show the contrast between a predicate and a headless relative clause that is nominalised by a specific classifier: (18) a'xy-dy-'xa-tsy-tse house-CAU-2S-POT-DEC ‘You are going to make a house.’ (19) a'xy-dy-'xa-tsy-ka'n a'w i-da-ki house-CAU-2S-POT-CL:oblong see-1S-DEC ‘I saw boards for you to make a house with.’ Also in possessive constructions classifiers can occur instead of the element -h . In principle all classifiers can occur in this position and there is no specific subcategory of classifiers restricted to possessive constructions exclusively. In addition, the possessive constructions in which classifiers occur are canonical and do not represent a separate strategy to express possession. Thus, Kwaza does not have the type of possessive classifiers found in many Austronesian languages – see e.g. Lichtenberk (this volume) and Crowley (1995). Consider the following example: (20) (a'xy) 'si-dy-xy house I-POS-CL:house ‘my house’ Since the classifier -xy in this example has the specific semantic content ‘house’ (it is indeed etymologically related to the noun a'xy ‘house’), its referent can be omitted without loss of semantic content, similar to headless relative clauses and the pro-drop possibility of predicates.5 Compare the following to example (2), with which no difference in meaning was attested: (21) tawi'wi-dy-xy Tavivi-POS-CL:house ‘Tavivi’s house’ The next example contains a headless possessive construction modified by a relative clause, with the specific classifier -xy acting almost like a crossreference morpheme: 5
Overt pronominal expression of arguments was attested to have a disambiguating or emphatic effect. The pragmatic consequences of full versus reduced expression of the PM have not been investigated.
Possessive expressions in the Southwestern Amazon
353
(22) 'si-dy-xy haka-xy-'na 'a-xa-ki I-POS-CL:house old-CL:house-LOC exist-2S-DEC ‘You live in my old house.’ Many classifiers characterise or refer to specific entities for which no independent monomorphemic noun is available. Such classifiers can be attached to the noun formative root e- to form an independent noun (see also § 2.4). They can also occur in other expressions and replace the element -h . No difference in meaning was attested between the following examples: (23) o'lu-dy-h e-'kai curassow-POS-NOM Ø-CL:leg ‘leg of a curassow bird’ (24) o'lu-dy-'kai curassow-POS-CL:leg ‘leg of a curassow bird’ Many body-part classifiers are of this type and probably just as many are not. Furthermore, a number of classifiers that have a very general referential scope and that are not etymologically related to a full noun can also occur with the empty root e-. The following examples illustrate the classifiers -ro ‘container, vessel, cup’ and -ri ‘flat object’: (25) 'si-dy-ro 'dai-xa-ki I-POS-CL:vessel grab-2S-DEC ‘You took my cup.’ (26) e-'ro Ø-CL:vessel ‘container’, ‘radio’ 6 (27) a'r i-dy-h e-'ri tapir-POS-NOM Ø-CL:flat ‘the tapir’s liver’
6
The second interpretation of this example is an illustration of (the frequently attested) lexicalisation of classifiers in combination with other elements.
354 Hein van der Voort (28) a'r i-dy-'ri tapir-POS-CL:flat ‘the tapir’s liver’7 The most general classifier is the semantically neutral nominaliser -h , both in possessive constructions and in any other construction that can feature a classifier. As shown above, classifiers can have nominalising properties and they can also replace -h in possessive constructions. For these reasons, the combination -dy-h (discussed in §2.1 and §2.2) should not be regarded as a single fossilised possessive morpheme. It is a partially lexicalised combination in the sense that it functions as an idiom on the morphological level. But it is also still transparent in that the constituting elements have preserved their original functions and permutability to a certain extent. This kind of semiproductive morphological patterning represents a phenomenon also attested in other parts of the grammar.
2.2.3. Homophonous forms of -dyIn view of the special position that possessive -dy- occupies in the general make-up of Kwazá nominal morphology, it is necessary to consider other morphemes with which it is homophonous. There is another form -dy- that most likely originates in the possessive morpheme. It is used in a fixed combination with the adverbial nominaliser -nãi, with the specific meaning ‘language of’. This construction is adverbial and no possessed head can be identified on the semantic level, as the following examples show: (29) 'tyka-dy-nãi jã'si-ta Mekens-POS-NOM hear-CSO ‘they understood the Mekens language’ (30) 'xyi-dy-nãi are'ta-da-m you-POS-NOM know-1S-VOL ‘I want to learn your language.’ 7
In some rare cases the omission of the possessive morpheme may not seem to change the meaning, as in ar i-'ri ‘the tapir’s liver’. The structure of ar i-'ri is fundamentally different from (28), however, since it (ar i-'ri) represents a classified nominal comparable to a nominal compound. A similar alternative for example (24), *olu-'kai, was rejected by the consultant.
Possessive expressions in the Southwestern Amazon
355
One of the reasons to consider -dy- as originally possessive here is its similar behaviour with regard to the first person inclusive pronoun. The expression txana-nãi ‘our language’ is preferred to txana-dy-nãi. Remember that a similar preference with regard to the canonical possessive was mentioned in the paragraph under example (3). An additional morpheme that looks suspiciously like the canonical possessive is the comitative case suffix -dyn . Consider example (31): (31) jere'xwa-dyn hyhy'rwa-da-ki jaguar-COMIT move-1S-DEC ‘I’m walking with the dog8 (by my side).’ The comitative belongs with the beneficiary -du, instrumental -ko and locative -na to the semantic cases of Kwaza; these do not mark obligatory grammatical arguments but optional “oblique” arguments. There are both formal and semantic similarities with the canonical possessive construction, but there are also important differences. Even though there could be an etymological relationship with the possessive morpheme -dy-, and perhaps with the verbal derivational reflexive morpheme -n -, the comitative morpheme -dyn is not further analysable synchronically. And even though in certain languages the comitative can mark a PM (e.g. McGregor 2001: 81; Stassen 2006: 771) nothing like that has been observed in Kwaza. Finally, the possessive morpheme -dy- is homophonous with the causative/benefactive morpheme -dy- and with the different subject marker -dy-. It is possible that they are etymologically related, but it is difficult to demonstrate this empirically. Being mainly verbal suffixes, their distribution is to a large extent complementary with that of possessive -dy-. However, their very different specific functions, combined with their very limited and specific distributions are sufficient reason to consider of each of these items synchronically as separate morphemes.
2.3. An alternative adnominal possessive Kwaza has no clear inflexional paradigms for possession according to person or other categories. Instead, person and number are distinguished through pronouns, as seen in the canonical possessive construction presented above. 8
In Kwaza as well as in various other languages of the region, the words for ‘jaguar’ and ‘dog’ are identical.
356 Hein van der Voort Nevertheless, Kwaza has an alternative possessive construction that is used frequently, although its distribution is restricted. Whereas the canonical possessive is marked on the PR, with (neutral or specific) reference to the semantic class characteristics of the PM, the alternative possessive is marked on the PM, with reference to a third person PR. It is created by attaching the suffix -tjate directly to a nominal PM. Note the following example, which is based on the independent noun mã ‘mother’: (32) mã-tja'te mother-3.POS ‘his mother’ The possessive morpheme -tjate does not form part of a paradigm and it refers exclusively to a third person PR. It has no other uses and there are no other head-marking possessive strategies to express different persons. The morpheme -tjate is not further analysable. It may be a borrowed morpheme, as will be discussed in (§3.1.3). Overt expression of the dependent PR is rare in the alternative possessive construction, but it is attested: (33) hu'ri en -tja'te-(na) ojabu'ru-da-h -ki paca foraging.place-3.POS-LOC arrive.thither-1S-NOM-DEC ‘I arrived at the foraging place of the pacas.’ Herslund and Baron (2001: 14) discuss the phenomenon that languages make use of different adnominal possessive constructions to distinguish different kinds of possessive relationships, such as alienable vs. inalienable. Rijkhoff (this volume) mentions a special definiteness effect for third person PR markers in Uralic and Turkic languages. In Kwaza, however, no such distinctions were attested between the alternative and the canonical possessive in the third person. The following contrasted examples are semantically identical: (34) kanwã-tja'te canoe-3.POS ‘his canoe’ (35) ' -dy-h he-POS-NOM ‘his canoe’
ka'nwã canoe
Possessive expressions in the Southwestern Amazon
357
No productive strategies are attested to distinguish between reflexive and non-reflexive possession. The different adnominal possessive constructions do not seem to contribute to such a distinction at any rate: (36) tsu'ty-si'ki
-dy-'h 'bu-ki head-CL:skin he-POS-NOM put-DEC ‘He uses his own hat.’ (37) tsu'ty-si'ki-tja'te 'bu-ki head-CL:skin-3.POS put-DEC ‘He uses his own hat.’ In the following example, the object can refer either to the wife of the third person subject or to the wife of another third person: (38) etay-tja'te 'ta-ta woman-3.POS talk-CSO ‘hei said to hisi/j wife’ Recursive embedding of possessives is attested, though very rarely. There is one example of an alternative possessive construction inside a canonical possessive construction (see also van der Voort 2004: 202): (39) h dj -tja'te-dy-h eto'hoi older.brother.female-3.POS-POS-NOM child ‘child of older brother of female’9 (lit. ‘her older brother’s child’) Furthermore, there are a handful of attestations of a canonical possessive inside an (redundant) alternative possessive construction: (40) dodo'txi-dy-'h e'tay-tja'te latex-POS-NOM woman-3.POS ‘Latex-man’s wife’ The next example is rather exceptional because the head of a canonical possessive has two referents: 9
This expression was equated by the informant with the more general kinship terms kore ‘son of brother or sister’ and koretay ‘daughter of brother or sister’.
358 Hein van der Voort (41) 'xyi-dy-h a'ha-tjate you-POS-NOM father-3.POS ‘father of your father (i.e. your grandfather)’ (lit. ‘your father’s’)
2.4. The dummy root e- and the verb ‘to have’ As mentioned in the previous section, many languages have different adnominal possessive constructions with different semantic/pragmatic uses. In Kwaza, no productive strategies are encountered that express a distinction between alienable and inalienable possession. The semantically empty noun-formative dummy root e- discussed towards the end of § 2.2.2 might represent a historical remnant of an inalienable construction, since it occurs often with part-whole expressions that may involve obligatory possession. This must probably be regarded today as a regularity in the lexicon and not as a productive morphosyntactic strategy (assuming e- ever had anything to do with alienability distinctions). Even granted that alienability distinctions are culture- or language-specific (Chappell and McGregor 1995), no consistently observed criteria for such a distinction in Kwaza could be identified. Besides expected inalienable PMs such as body parts and kinship relations, the e- root occurs also with expressions for items that one would not expect to count as inalienable. Furthermore, not all entities that one would expect to count as inalienable PMs require e-. As an example, the body part noun tsuty ‘head’ does represent a likely candidate for inalienable concepts even though it does not contain the empty root. By contrast, the noun e-m ‘liquid’ does not seem to represent an inalienable or obligatorily possessed concept in spite of the empty root e-. The following example refers to a leg that may be cut off or not: (42) e-'kai-tja'te Ø-CL:leg-3.POS ‘his leg’ Another analysis of the dummy root e- relates it to the verb e- ‘to have’, with which it is homophonous. The verb e- ‘to have’ is used to express predicative possession in Kwaza, as in: (43) si eto'hoi aky-'h e'mã 'e-da-ki I child two-NOM three have-1S-DEC ‘I have three children.’
Possessive expressions in the Southwestern Amazon
359
This verb occurs also in specific habitual constructions, where it can be literally interpreted as an abstract predicative possessive: (44) duture-xwa'nã ja-'nãi e-'re eat-NOM have-INT pig-CL:flesh ‘Does she eat pork?’ (lit. ‘Does she have the eating of pork?’) The relationship between the verb ‘to have’, possession and aspect has been discussed in typological work by among others Benveniste (1960), Heine (1997) and Seiler (1977). To a very limited extent, the verb e- ‘to have’ is also used as an existential verb. The normal existential verb in Kwaza is a- ‘to be’, which in addition functions in another specific habitual construction. As it happens, there is also a semantically empty adverb formative root a-. This dummy root is used to create adverbs out of bound verbal derivational directional markers. Since the two roots e- and a- both can function as semantically empty formative dummy roots and as semantically very abstract verb roots, their parallel distributions suggest a relationship between their different functions. Further discussion of the dummy roots would take us too far from the present topic and merits a separate article. The reason for mentioning their parallelism here is to show that the empty noun formative e- and the verb root e- ‘to have’ should probably be considered as the same element. The dummy root e- or i- is also encountered (without the meaning ‘to have’) in Kanoê (§3.2), Latundê (§3.3), Baure (§ 3.4) and other languages of the Guaporé region. It was discussed as an areal feature in Crevels and van der Voort (2008: 167–168) and van der Voort (2005: 397–398). 3.
Possessive constructions in neighbouring languages
The previous section is a relatively detailed case study of Kwaza (isolate). In the present section I will discuss similarities and differences with possessive constructions in Aikanã (isolate), Kanoê (isolate), Latundê (Nambikwara), Baure (Arawak), Wari’ (Chapacura), Mekens (Tupi) and Arikapú (Macro-Jê), in that order. 3.1. The canonical possessive construction in Aikanã The Aikanã (also referred to as Massaká, Tubarão or Huari) are the traditional neighbours of the Kwaza. Their language is probably also an isolate,
360 Hein van der Voort although there are some lexical and structural similarities with its neighbouring languages. Aikanã is spoken by about 150 persons. Since the mid 1980s the language was studied and was partially described by the Brazilian linguist Ione Vasconcelos (2003) and by an American team of linguists led by Leanne Hinton (ed. 1993), and it is presently being studied by the author. Aikanã is a morphologically highly complex language. Like Kwaza, most of the morphological complexity resides in the verb. Furthermore it has similar classifying and directional suffixes, valency-changing suffixes, a distinction between future and non-future tense, person and mood inflexion, animate object case marking, and no nominal number inflexion. Unlike Kwaza, but similar to another Rondônian isolate Kanoê (§ 3.2), Aikanã has several different verbal inflexion classes, some of which require argument prefixes rather than suffixes. Also, it lacks an inclusive-exclusive distinction. With respect to possessive expressions, Aikanã shows both differences and similarities to Kwaza. Unlike Kwaza, Aikanã has a (partially transparent) set of specific possessive pronouns. Some of the possessive pronouns contain the element -z , which probably originates from the productive beneficiary case suffix that also has a genitive function.10 Furthermore, a specific verbal benefactive inflexional paradigm exists for beneficiary and possessive objects. In the following table the different Aikanã pronouns and benefactive inflexions are listed: Table 1. Aikanã pronouns and benefactive inflexions personal
possessive
benefactive
1S 2S 3S
hi'sa h 'zã kaj'ne, ka'ria
txü'txü h 'z kajne'z , kari'z
-ku-, -k -u(a)-w(e)-
1P 2P 3P
sa'te h zã('za) kaj'ne?ene, ka'ri?ene
sate'z h z 'za kari?ene'z
-k jã-uãjã-we?eje-
a
The first form kaj'ne ‘he, she, it’ is a third person pronoun. The second form ka'ri ‘that, that one’ is a demonstrative pronoun that is often used as a third person pronoun.
10
Aikanã /z/ is usually pronounced as IPA [] and in nasal environments sometimes as [n]. The vowel /ü/ is pronounced as IPA [y].
Possessive expressions in the Southwestern Amazon
361
The possessive and benefactive are complementary constructions in Aikanã. Both involve the suffix -z . As illustrated in § 3.1.1 below, -z marks the beneficiary object in Aikanã verb phrases. Inside noun phrases it marks a possessive relationship. The following example illustrates its possessive use: (45) 'masio-z ha'ne kãwã-' Marcio-BER water that’s.it-DEC ‘It is Marcio’s river.’ The suffix -z can also be recognised in the possessive pronominal forms in Table 1, where it is a fossilised element in the second person and transparent in the third person and first person plural. In possessive constructions, the possessive pronouns do not receive an extra beneficiary suffix -z , whether they contain the (fossilised) element -z or not: (46) txü'txü ha'ne kãwã-' my water it’s-DEC ‘It is my river.’ (47) 'txütxü 'kapedi'ka my shoulder ‘my shoulders’ (48) 'h z 'kapedi'ka your shoulder ‘your shoulders’ Application of -z to the first person singular possessive pronoun is characteristic of children’s speech: txütxü-z ‘mine’ (lit. ‘of my’). 3.1.1. The benefactive and possessive pronouns Beneficiary objects are marked both by benefactive object inflexion (cf. Table 1) of the verb, the beneficiary case marker on the relevant noun, and, optionally, an oblique case marker: (49) aikanã-'ene-z 'pürü-'ka-pa-we?eje-' Aikanã-COL-BER work-1S-CL:big-3P.BEN-DEC ‘I worked for the Aikanã.’
362 Hein van der Voort (50) ma'new-z -(je) 'pürü-'txa-pa-we-' Manoel-BER-OBL work-1P-CL:big-3S.BEN-DEC ‘We worked for Manoel.’ When the beneficiary object is a first or second person, possessive pronouns with obligatory oblique marking (rather than personal pronouns with beneficiary marking) are (optionally) employed: (51) h z -'je pürü-'ka-pa-ua-' your-OBL work-1S-CL:big-2S.BEN-DEC ‘I worked for you.’ (52) ü'?ü-ka-ua- h z -'je kaw-me-i-'za keep-1S-2S.BEN-DEC your-OBL eat-2S-NOM-COL ‘I kept for you things for you to eat.’ (53) txütxü-'je pürü-me-'pa-k - my-OBL work-2S-CL:big-1S.BEN-DEC ‘You worked for me.’ Instead of a benefactive construction involving possessive pronouns, one might rather expect a more transparent benefactive construction with beneficiary-marked personal pronouns. The fact that the latter is not attested is an additional indication that the possessive is related to or identical with the benefactive. The morpheme -z ‘BENEFICIARY’ might just as well be glossed as ‘GENITIVE’, or, conversely, the possessive pronouns as beneficiary pronouns. An even better descriptive solution may be to use a term that covers both the concepts possessive and beneficiary. This would be in accordance with the spirit of Gil’s (2001) ‘macrofunctional’ approach, which seeks to eliminate artificial ambiguity between Eurocentrically defined functions that are not actually distinguished from one another in the language. For the time being a compromise is found in the consistent use of ‘BENEFICIARY’ as a morphological gloss and ‘possessive’ to characterise the pronouns.11
11
It is unclear whether the suffix -z is related either with the Kwaza possessive suffix -dy- or with the Kwaza beneficiary suffix -du.
Possessive expressions in the Southwestern Amazon
363
3.1.2. Predicative possession and the benefactive In predicative possessive constructions, the expression of the relationships may involve both possessive and personal or demonstrative pronouns and benefactive verb inflexions: (54) 'hiba 'kari-we- this that-3S.BEN-DEC ‘This is his.’ (55) 'kari-k jã- that-1P.BEN-DEC ‘It’s ours.’ Some possessive relations can be expressed in alternative ways. The semantic difference between them is not yet clear: (56) 'hiba txütxü-ku- this my-1S.BEN-DEC ‘This is mine.’ (57) 'hiba 'kari-ku- this that-1S.BEN-DEC ‘This is mine.’ (58) 'hiba 'h z -ua- this your-2S.BEN-DEC ‘This is yours.’ (59) 'hiba 'kari-ua- this that-2S.BEN-DEC ‘This is yours.’ In example (56) the possessive pronoun is used as a verb root. The example could be interpreted literally as: ‘This is mine for me.’. In (57) a personal pronoun forms the root of the predicate: ‘This is (it) for me.’. There is a similar difference between (58) and (59). In the following set of examples, the first expression is less common, since, as the consultant explained, a wife is a human being:
364 Hein van der Voort (60) deti'a 'txütxü-ku- woman my-1S.BEN-DEC ‘The woman is mine.’ (61) deti'a 'kari-ku- woman that-1S.BEN-DEC ‘That is my wife.’ A similar distiction is suggested by the translations of the next set of examples: (62) wãwã'? 'txütxü-ku- child my-1S.BEN-DEC ‘The child is mine.’ (63) wãwã'? 'kari-ku- child that-1S.BEN-DEC ‘That is my son/daughter.’ It could be that the alternative versions of these examples involve animacy distinctions. Additional field research in the future will hopefully improve our understanding of the semantic or pragmatic distinction between these expressions. 3.1.3. The alternative possessive construction in Aikanã In addition to the canonical possessive constructions, Aikanã also has an alternative way to express the possessive. This alternative possessive is similar to the Kwaza alternative possessive discussed in § 2.3. Even the morpheme involved, -deri ‘third person PR’, is similar both in form and meaning to the equivalent Kwaza morpheme -tjate. Note the following construction: (64) kura-de'ri husband-3.POS ‘her husband’ This alternative possessive is attested very frequently, but again, it only exists for a third person PR and it does not seem to be determined by features of alienability or animacy. Among the languages of the Guaporé region, only
Possessive expressions in the Southwestern Amazon
365
Aikanã and Kwaza were found to have such an extra-paradigmatic third person PR morpheme. The construction could be considered a shared irregularity on the structural level. As was briefly discussed in van der Voort (2005: 401), this is a possible indicator of a genealogical relationship between the languages. It is perhaps even more likely that it represents an instance of (ancient) borrowing. 3.2. Possessive constructions in Kanoê Kanoê (also referred to as Kapixana) is a language isolate whose speakers used to be traditional neighbours of the Kwaza, Aikanã and Mekens. With five remaining speakers Kanoê is on the brink of extinction. The language has been studied and described since the early 1990s by Bacelar (e.g. 2004), on whose work this section is based. It is a morphologically very complex language, especially with regard to verbs. It shows some lexical and structural similarities with the languages of the region, in particular with Kwaza and Aikanã, as discussed in van der Voort (2005). Kanoê differs from Kwaza and is similar to Aikanã in that it has different inflexional verb classes and that it employs both person prefixes and suffixes. Another resemblance with Aikanã is that Kanoê has a set of possessive pronouns in addition to personal pronouns and that it lacks an inclusive-exclusive distinction. The Kanoê pronouns are listed in Table 2. Table 2. Kanoê pronouns (from Bacelar 2004: 143, 145) personal
possessive
1S 2S 3S
aj m
oj
ña, jã pja ojo
1P 2P 3P
ajte m te ojte
jato pjato ojoto
As in Aikanã, some of the possessive pronouns in Kanoê are transparent. They contain the possessive morpheme -o, which in the plural pronouns seems to have changed or replaced the vowel of the collective morpheme -te. The PR in a noun phrase may be represented by a possessive pronoun: 12 12
The Kanoê vowel /y/ is pronounced as IPA [].
366 Hein van der Voort (65) pja tyj your house ‘your house’
(Bacelar 2004: 146)
The possessive morpheme -o itself is productively used in full noun phrases, such as: (66) nañu-o kyj wasp-POS sting ‘the wasp’s sting’
(Bacelar 2004: 99)
Just as in Aikanã, PRs that contain the possessive morpheme (-o) represent morphologically independent nouns and no further marking is required of either the PR or the PM. Kanoê does not have an alternative possessive construction like the Kwaza (cf. §2.3) and Aikanã (cf. §3.1.3) expressions that involve just the third person singular. 3.2.1. Kanoê classifiers in possessive constructions Similar to Kwaza (as in examples (23), (26), (27) and (42)), Kanoê has a large subset of classifiers that can be turned into full nouns by the semantically empty noun formative root i-. The empty root occurs mainly with ‘inalienable’ items such as body part classifiers. In possessive constructions, these classifiers build equivalent alternative expressions. Either the classifier combined with the empty root functions as an independent PM, as in example (67), or the empty root is omitted and the classifier is attached as a bound morpheme to the PR, as in example (68). (67) ytse-o i-katsi tree-POS Ø-CL:root ‘root of tree’
(Bacelar 2004: 100)
(68) ytse-o-katsi tree-POS-CL:root ‘root of tree’
(Bacelar 2004: 100)
This set of alternative expressions is strikingly similar to the set of alternative espressions in Kwaza, illustrated by examples (23) and (24). An important difference from Kwaza, however, is that in Kanoê the classifier does not
Possessive expressions in the Southwestern Amazon
367
attach to an incomplete possessive stem, replacing a nominaliser. Instead, it is affixed to a complete PR form, which does not require a classifier. In fact, the Kanoê construction looks rather like a compound, since a third alternative is the following: (69) ytse-katsi tree-CL:root ‘tree root’
(Bacelar 2004: 100)
This form resembles Kwaza noun-classifier combinations, such as tsu'ty-si'ki ‘hat’ in examples (36) and (37), which, however, are not very productive. It may be that the subset of classifiers which combine with the empty root should instead be analysed as bound noun roots and the empty root i- as a noun-formative prefix. 3.2.2. The benefactive and possessive marking When used in combination with the oblique case suffix -ni, the Kanoê possessive suffix -o has a beneficiary (70) or comitative (71) sense. This was attested only with the first and second person singular personal pronouns and it notably does not involve possessive pronouns: (70) mi pi-tej-ja-ry-nu mi-tsi piña aj-o-ni you 2-transport-DR-REFL-FUT 2-INT pineapple I-POS-OBL ‘Will you bring pineapple with you for me?’ (Bacelar 2004: 235) (71) aj uræ ipæ õ-e-re mi-o-ni I pig kill 1-DEC-AUX you-POS-OBL ‘I hunt pigs with you.’
(Bacelar 2004: 242)
There are some similarities between the benefactive constructions in Kanoê (70) and Aikanã (§3.1.1). Both may involve a combination of a possessive element and an oblique case marker. Then again, they are different in that in Aikanã a possessive pronoun is used, whereas in Kanoê a more analytic combination of a possessively marked personal pronoun is used in the benefactive construction. Furthermore, Aikanã and Kanoê differ also with respect to beneficiary arguments in general: in Aikanã the PR and the beneficiary argument are marked by the same morpheme -z ; in Kanoê beneficiary objects are usually unmarked, just like dative, recipient and goal objects (Bacelar 2004: 234).
368 Hein van der Voort 3.3. Possessive constructions in Latundê Lakondê and Latundê are dialects of a language that belongs to the northern branch of the small Nambikwara language family. The language, henceforth referred to as Latundê, has about 20 speakers (there is only one surviving speaker of the Lakondê dialect). It has been described by Telles (2002a), on whose work this section is based. Like many languages of the Guaporé region, Latundê is a morphologically complex language. The possessive constructions of Latundê, which are also discussed in a separate article by Telles (2002b), are different from those in the language isolates discussed above. Latundê has a full paradigm of possessive prefixes, which are attached to the nominal PM. The prefixes are reduced versions of the personal pronouns, listed in Table 3 below: Table 3. Latundê personal pronouns and possessive prefixes (from Telles 2002a: 150, 156 and 2002b: 158)
1S 1P 2 3S 3P a a
personal
possessive
ta ja n h wa ja hãja a wja
ta n hwa hãj-, nã-, ãa w-, ã-
The possibility of distinguishing between the singular and plural of the third person possessive in Latundê does not exist in Lakondê (Telles 2002a: 157).
Independent nouns in isolation in Latundê, including possessed nouns, have to end in the referential suffix -te or -tu, which produces a citation form:13 (72) wa-sih-te 2.POS-house-REF ‘your house’
(Telles 2002a: 156)
In Latundê, possession can be expressed either in a synthetic manner, employing a possessive prefix, or in an analytic manner, using a full personal pronoun. 13
In spontaneous speech the referential suffixes also play a not yet fully understood role in the marking of pragmatic status (Telles: 2002a: 215).
Possessive expressions in the Southwestern Amazon
369
(73) ta-najn-kin n-te 1S.POS-head-CL:round-REF ‘my head’
(Telles 2002a: 163)
(74) taja najn-kin n-te I head-CL:round-REF ‘my head’
(Telles 2002a: 163)
The semantic value of the alternatives is identical, although the analytic expression tends to occur more frequently in predicative constructions than the synthetic one. Compounding is a productive word-formation process in Latundê. In Telles (2002a,b), a distiction is made between ‘non-genitive’ and ‘genitive’ compounds, as in the following contrasting examples and their translations: (75) kejajn-tawn-te peccary-tail-REF ‘peccary tail’ (a kind of tail)
(Telles 2002a: 159)
(76) kejajn-ã-tawn-te peccary-3.POS-tail-REF ‘the tail of the peccary’
(Telles 2002a: 159)
In the ‘genitive’ compound, the third person PR prefix ã- occurs in intermediate position, coincidentally resembling the Kanoê PR morpheme -o in example (68). The referential suffix indicates that the preceding roots represent a single (composite) stem. Latundê also has a noun-formative empty root -, that can be used to turn classifiers into independent nouns (Telles 2002a: 96), and which is similar to Kwaza e- (see §2.4) and Kanoê i- (see § 3.2.1). Possessive constructions in other Nambikwara languages, with which there are certain differences, are discussed in Araujo (2004) and Kroeker (2001).
3.3.1. Alternative possessive Again, there are alternative ways to express possession, since there may be a reason to consider the third person possessive morpheme ã- as a special case, different from the others. Note that, in addition to full nouns, Latundê
370 Hein van der Voort classifiers can also be directly preceded by a possessive prefix.14 In this type of construction the third person possessive morpheme ã- can be inserted between the possessive prefix and the classifier, without any known consequences for the meaning. Compare the following examples: (77) ta-ni-tu 1S.POS-CL:hemispheric-REF ‘my pan’
(Telles 2002a: 194)
(78) ta-ã-ni-tu 1S.POS-3.POS-CL:hemispheric-REF ‘my pan’
(Telles 2002a: 194)
The canonical possessive construction in Latundê is of the same type as the alternative construction in Kwaza and Aikanã. These constructions are characterised by PR marking on the PM. The difference is that in Latundê there is a full paradigm, whereas in Kwaza and Aikanã it features only a third person morpheme that does not belong to a paradigm. However, also in Latundê the third person possessive behaves differently from the other possessive morphemes. 3.3.2. A dummy possessum In predicative possessive constructions, a special semantically empty root nãn (Telles 2002a: 160) may substitute an overt noun if its referent is understood from the context: (79) ta-hu -kah-tãn 1.POS-bow-CL:long-IMPF ‘it was my bow’
(Telles 2002a: 161)
(80) ta-nãn-kah-tãn 1.POS-Ø-CL:long-IMPF ‘it was mine (my bow)’
(Telles 2002a: 161)
14
Although Latundê has a small set of classifiers, they have a similar distribution and nominalising potential as in Kwaza. Some classifiers are even identical in form with those in Kwaza, Kanoê and Aikanã, which is probably due to areal diffusion (van der Voort 2005: 395–397).
Possessive expressions in the Southwestern Amazon
371
The empty root nãn was encountered only in possessive constructions. It should probably be regarded as verbal rather than nominal, because the nominal referential suffix -te cannot be attached to it unless a classifier is inserted: (81) ta-nãn-saw-te 1.POS-Ø-CL:liquid-REF ‘my liquid (e.g. coffee)’
(Telles 2002a: 162)
(82) *ã-nãn-te 3.POS-Ø-REF ‘his (e.g. hand)’
(Telles 2002a: 162)
3.4. Possessive constructions in Baure Baure is a language that belongs to the southern branch of the Arawak language family. It is spoken by about 60 elderly people in the Bolivian lowlands. It has been described by Danielsen (2007), on whose work this section is based. Like most languages of the region, the morphology of Baure is rather complex and shows polysynthetic traits. It has several characteristics in common with other non-Arawak languages, such as classifiers and directionals. There is no inclusive-exclusive distinction in the person reference system. There is a gender distinction for the third person singular. The system is based on person proclitics that are attached both to verbs and nouns. When attached to verbs, the proclitics refer to a subject argument, when attached to nouns they refer to a PR. In addition, there is a possessive suffix -no that is homophonous with a nominaliser. As Danielsen (2007: 119) explains, two different kinds of nouns are distinguished with regard to the expression of possession. Optionally possessed nouns receive a person proclitic and usually the derivational possessive suffix -no, as shown by the following examples: (83) yaki fire ‘fire’
(Danielsen 2007: 87)
(84) ni=yaki-no 1S=fire-POS ‘my fire’
(Danielsen 2007: 87)
372 Hein van der Voort Obligatorily possessed nouns receive only a person proclitic: (85) e-ser Ø-tooth ‘a tooth of someone’
(Danielsen 2007: 120)
(86) ni=ser 1S=tooth ‘my tooth’
(Danielsen 2007: 120)
The element e- is analysed by Danielsen (2007: 119) as an ‘unspecified PR’ prefix, because its application to obligatorily possessed noun roots is required when the PR is unknown. Note, however, that it is not fully productive, and with some roots it occurs lexicalised with a specific unpredictable meaning.15 A similar element occurs in the other languages discussed above, where it is often, but not always, attested with body parts, and where its distribution is not strictly determined by ‘inalienability’ distinctions. Just as in these languages, the element e- in Baure should perhaps be regarded as a semantically empty noun-formative root, which is an areal feature of the region (see also § 2.4). In many respects, the bound noun roots behave like other classifiers in Baure (Danielsen 2007: 142 ff.), and similar to the body part classifiers in Kwaza (§ 2.2.2) and Kanoê (§ 3.2.1). In addition to the above two kinds of nouns, there is a set of nouns that cannot be possessed, which includes astronomical bodies and animals.16 Danielsen (2007: 124) nevertheless points out that it is possible to express possession in the case of domestic animals. This involves a construction where the PM is preceded by an ‘inalienably possessed noun’: (87) kove’ dog ‘dog’
(Danielsen 2007: 124)
(88) ni=per kove’ 1S=domestic.animal dog ‘my dog’
(Danielsen 2007: 124)
15 16
Danielsen (2007: 120) mentions the existence of several lexicalised combinations, such as -waki ‘palm of the hand’, ewaki ‘forked branch of tree’. Consequently, these nouns are not attested with person proclitics, except in certain lexicalised combinations.
Possessive expressions in the Southwestern Amazon
373
As a bound noun root, the generic noun per ‘domestic animal’ can function as a classifier. Consequently, the above construction is very similar to the possessive classifier constructions discussed for Gran Chaco languages by Fabre (2007) and Messineo and Gerzenstein (2007) and the ‘indirect’ possessive construction described by Lichtenberk (this volume) for Oceanic languages. The generic noun per also occurs in other constructions than the above.
3.4.1. Expression of the possessor Possessive NPs may consist of an inflected PM and a PR:17 ro=wer18 to kotis DEM.M 3SM=house ART lizard ‘the house of the lizard’
(89) te
(Danielsen 2007: 125)
In order to form personal pronouns, person proclitics are attached to the element -ti’, which is etymologically related to the feminine demonstrative root ti. Possessive pronouns are subsequently created by the attachment of an additional suffix -ro (or its phonotactically determined variant -r), which may be related to the possessive suffix -no (Danielsen pers.comm.). Hence, the structure of the first person possessive pronoun, for example, is ni-ti-ro ‘my’. The segmentation of possessive pronouns is ignored in the examples: (90) ritiro-wo na’ her-COP egg ‘She has got eggs.’ (lit. ‘The eggs are hers.’)
(Danielsen 2007: 203)
Note that possessive NPs may also involve possessive pronouns, albeit rarely, as shown in example (91). (91) ritir na’ her egg ‘her eggs’ 17 18
(Danielsen 2007: 203)
Baure // represents IPA []. The glide /y/ is pronounced as IPA [j]. The root -wer ‘house’ belongs to a small set of nouns that cannot undergo derivation. Therefore, the possessive suffix -no cannot be applied here, even though the referent might be regarded as alienable.
374 Hein van der Voort 3.4.2. Possessive pronouns in benefactive constructions It is noteworthy that possessive pronouns are used in analytic renderings of benefactive expressions. The synthetic benefactive construction involves a benefactive morpheme on the verb and extra person inflexion for the beneficiary argument, as in (92). The analytic benefactive construction is not realised through verbal morphology but by an extra personal pronoun preceded by a possessive pronoun, both referring to the beneficiary argument, as in (93). (92) ni=ay-ino-wo=pi=ro 1S =desire-BEN-COP =2S =3SM ‘I wish it for you.’
(Danielsen 2007: 16)
(93) ni=aya-wo=ro pitir piti’ 1S =desire-COP =3SM your you ‘I wish it for you.’
(Danielsen 2007: 16)
According to Danielsen (2007: 16, 320), the analytic benefactive construction may be a recent development caused by language obsolescence. She notes that the speakers of Baure do not regard the construction as the result of Spanish influence. 3.5. Possessive constructions in Wari’ Wari’ (also known as Pakaa Nova) is a language that belongs to the small Chapacura language family of western Brazil and northern Bolivia. It has close to 2000 speakers and it is described in Everett and Kern (1997), on whose work this section is based. Wari’ has a grammaticalised gender system, like certain other languages of the region. The language is syntactically rather than morphologically complex. Person marking is realised through particles that Everett and Kern refer to as nominal and verbal inflexional clitics (NICs and VICs). Inclusive and exclusive first person plural and third person feminine and masculine are distinguished. According to Everett and Kern (1997: 311) there are no possessive pronouns. The pragmatically unmarked constituent order of Wari’ is VOS. Possessive constructions in Wari’ involve PR-marking on the PM, like in Latundê. The possessed noun agrees through an inflexional clitic with person, number and third person gender of the PR. There are three different
Possessive expressions in the Southwestern Amazon
375
types of possessive marking, corresponding to different classes of nouns; (a) inalienably possessed nouns are inflected by attachment of special possessive suffixes;19 (b) possessable nouns are inflected by NICs, i.e. particles that consist of the element ne- combined with variants of the possessive suffixes; (c) inalienably possessed kinship terms are inflected by VICs. The following examples show the respective different possessive expressions:20 (94) xiri-con Xijam house-3SM.POS Xijam ‘Xijam’s house’
(Everett and Kern 1997: 150)
(95) pije’ nequem Hatem child POS.3SF Hatem ‘Hatem’s child’
(Everett and Kern 1997: 150)
(96) co-xa’ ca Xijam INFL-younger.sibling.my 3SM Xijam ‘Xijam’s younger sibling’
(Everett and Kern 1997: 376)
The third construction is limited to a specific set of kinship terms that require the verbal ‘inflection’ prefix co-. The kinship term always has a first person singular possessive interpretation. It is followed by a VIC because, as explained by Everett and Kern (1997: 375–377), it has undergone zeroderivation from an original verbal clause. The literal interpretation of (96) is therefore quotative: ‘Xijam he (says) “my younger sibling”’. The construction is grammaticalised and not transparent to the speakers of Wari’. Also in other parts of Wari’ grammar, the quotative is used in other functions than to quote direct speech, such as future tense. Note that in all constructions an independent noun referring to the PR can be omitted if understood from context, as in e.g. xiri-con ‘his house’ (cf. example 94), and in:
19
20
Wari’ has two noun classes. The class of inalienable nouns is characterised by the fact that the citation form ends in -xi ‘1P.INCL’. The class of possessable nouns does not have this characteristic (Everett and Kern 1997: 3). From the examples below it appears that the noun ‘house’ belongs to the inalienable class, whereas the noun ‘child’ belongs to the possessable class. Wari’ /x/ represents IPA [(t)∫].
376 Hein van der Voort (97) wina-hu’ head-2P.POS ‘your head’
(Everett and Kern 1997: 236)
In the absence of adjectives in Wari’, attributive concepts are often expressed by nouns in possessive constructions: (98) wijima-in xirim smallness-3N.POS house ‘small house’ (lit. ‘the house’s smallness’) (Everett and Kern 1997: 152)
3.6. Possessive constructions in Mekens (Sakurabiat) The language of the Mekens (also known as Sakurabiat) belongs to the Tuparí branch of the Tupí language family. It has about 25 speakers. The language is described in Galucio (2001), on which work this section is based. Mekens is not a morphologically highly complex language, although in general, verbs are more complex than nouns. Mekens morphology involves inflexional argument prefixes and derivational valency changing prefixes, whereas for tense, mood, aspect and category change, suffixes are employed. Inclusive, exclusive first person plural and coreferential third persons are distinguished both in the pronominal and person marking systems. There are personal and reflexive pronouns, but no possessive pronouns. No special possessive morphemes are used in Mekens, and possessive full noun phrases consist of juxtaposed nouns. Anaphorically expressed PRs are indicated by inflexional person prefixes on the PM. A PR can never be expressed by a full pronoun. Somewhat similar to Baure and Wari’, possessive constructions are organised in different ways, depending on the (language-specific) alienability of the PM. In Mekens, inalienable nouns and kinship terms are always possessed, which is indicated by a person prefix or by a preposed noun: (99) e-pisa 2S-liver ‘your liver’
(Galucio 2001: 32)
Possessive expressions in the Southwestern Amazon
(100) sakrap okway spider.monkey tail ‘spider monkey’s tail’
377
(Galucio 2001: 32)
If there is no overt PR noun, third person PRs are expressed by the prefix ibefore consonants and by the prefix s- before vowels: (101) i-piso 3S-foot ‘his foot’
(Galucio 2001: 35)
(102) s-anp 3S-head ‘his head’
(Galucio 2001: 36)
When alienable nouns occur possessed, this is indicated in the same way as the inalienable situation, except for a subclass of nouns which require prefixation of a t.21 The word ek ‘house’ belongs to this subclass: (103) o-tek 1S-house ‘my house’
(Galucio 2001: 33)
(104) o-top tek 1S-father house ‘my father’s house’
(Galucio 2001: 33)
A third set of nouns does not accept person prefixes, but needs to be juxtaposed to an inflected obligatorily possessed (inalienable) noun of general semantic content that functions as a classifier. The word apara ‘banana’ belongs to this subclass: (105) o-iko apara 1S-food banana ‘my banana’
21
(Galucio 2001: 33)
This t is not an affix. As Galucio (2001: 32) observes, its occurrence is neither phonologically nor morphologically predictable, and its prefixation to certain nouns under possession represents a regularity in the lexicon.
378 Hein van der Voort (106) *o-apara 1S-banana ‘my banana’
(Galucio 2001: 33)
The latter construction reminds one of the ‘indirect’ possessive construction described by Lichtenberk (this volume) for Oceanic languages. The difference is that the Oceanic construction involves a special possessive classifier, whereas in Mekens a general classifying noun with a wider functional and distributional domain is used. The construction is also very similar to the one described for Baure in §3.4. In Mekens, a reflexive PR, i.e. a PR that is coreferent with the subject of the sentence, is expressed for the third person singular and plural, by distinct person markers. The distinction is a consequence of the switch-reference marking system of Mekens. The following examples contrast two third person singular forms: (107) i-tek 3S-house ‘his house’
(Galucio 2001: 76)
(108) se-tek 3S.C-house ‘his (own) house’
(Galucio 2001: 76)
3.7. Possessive constructions in Arikapú Arikapú is a language that belongs to the small Jabutí branch of the MacroJê language family. It is spoken by only one elderly person. The language has been studied by the author since 2001, and it was recently classified in cooperation with Eduardo Ribeiro (Ribeiro and van der Voort 2010). Its traditional neighbouring languages belong mainly to the Tuparí family. Like the Chapacura and Tupí languages, Arikapú is morphologically a rather simple language although there are some nominal and verbal suffixes. Person marking is realised by agreement prefixes, that can be used in addition to personal pronouns. The distribution of person marking obeys an ergative pattern. Arguments are indicated through the application of person prefixes on the verb. Possession is indicated by the same prefixes, when applied to PM nouns, referring to the PR:
Possessive expressions in the Southwestern Amazon
379
(109) a-kan 2-hat ‘your hat’ (110) i-n 'r 1S-house ‘my house’ Juxtaposed nouns and (rare) pronoun-noun combinations have a possessive interpretation. The constituent order is PR–PM: (111) 'nw 'kaj tapir penis ‘tapir’s penis’ (112) i'h pr 'm I plate ‘(It is) my plate.’ Some examples suggest that a PR that is coreferential with the subject of a sentence can be distinguished by the third person marker ta-. This person marker is otherwise applied only to intransitive verbs: (113) ta-n 'r-n ta-n 't 3-place-LOC 3-sleep ‘He’s sleeping in his own house.’ (114) paku' ta-'txi ã'rã-wiro woman 3-mother see-FUT ‘The woman went to see her mother.’ The Arikapú coreferential feature does not seem to be part of a switchreference marking system as in Mekens. Arikapú shares the coreferential feature with other Macro-Jê languages (Ribeiro and van der Voort in press, Rodrigues 1986: 55).
380 Hein van der Voort 3.7.1. Impersonal reference in possessive constructions The prefixes for first person singular (i-) and first person plural (txi-)22 are also used as third person or impersonal prefixes, especially in citation forms. These two prefixes seem to function often as a dummy PR with inalienable concepts as well as a dummy argument with verbs. When the PR or the third person object is expressed by a full noun, the impersonal prefixes are absent. The following examples contrast these constructions: (115) i-'k 1S-skin ‘(its) paper’, ‘my skin’ (116) patxi='k tobacco=skin ‘cigarette paper’ (117) 'txi-ku'ju 1P-wing ‘our/one’s/its wing’ (118) 'aro='kuju guan=wing ‘wing of a guan bird’ Example (115) and (117) show that the absence of a (pro-)noun may allow for grammatical ambiguity, since in principle, various interpretations of the person prefixes are possible. Alternatively, examples (116) and (118) could perhaps be analysed as attributive compounds rather than as possessive constructions, or maybe even as classifying derivations, since the forms k ‘skin’ and kuju ‘wing’ were never attested as free nouns. Unlike many other Macro-Jê languages (Rodrigues 1999: 191) and unlike Baure (§ 3.4) and Mekens (§ 3.6), Arikapú does not have generic classifying nouns to distinguish between alienable and inalienable possession.
22
Arikapú /x/ represents IPA [].
Possessive expressions in the Southwestern Amazon
381
4. Final remarks The languages discussed in this article form only a small subset of the entire number of languages of the Southwestern Amazon, and they mainly represent the language diversity on the Brazilian side of the Guaporé River. Kwaza is taken as a standard of comparison and is presented in more depth than the other languages, because of the sources, data and experiences available to me. The purpose of this article is to provide a first impression of the different structural possibilities for the expression of adnominal possession in the Southwestern Amazon region, and to explore possible areal linguistic relationships. Two main types of adnominal possessive strategies were encountered, both expressed in a range of different varieties. Type I is characterised by the occurrence of a general possessive element that is attached to the PR. It forms the standard possessive construction in Aikanã, Kanoê and Kwaza. It could be considered as a ‘genitive’ construction, but I have not used this term in order to avoid possible association with case marking, which is different in these languages. Type II is characterised by the attachment of PR agreement morphemes or person markers to the PM. This is the standard possessive construction in Arikapú, Baure, Latundê, Mekens and Wari’. An overview of the different characteristics of possessive expressions is presented in the Table 4:
+ +
+ + + +
+
Kwaza
Latundê
Wari’
Arikapú
(+)
Aikanã
possessive morpheme (I) PR agreement (II) person agreement (II) possessive pronoun personal pronoun juxtaposed nouns PR-PM order (in)alienability PM reference by classifier possessive classifier benefactive effects coreferential possessive
Kanoê
a b c d e f g h i j k l
Mekens
Baure
Table 4. Possessive features in the Southwestern Amazon
+
+ (+)
+ (+)
+
+
+ +
+ + +
+ + + (+)
+ +
+ (+) +
+
+
+
+
+
+
+
+
+ +
+
382 Hein van der Voort Some observations should be made about these characteristics. With regard to characteristics (a), (b) and (c), both Aikanã and Kwaza have an additional type II construction. This alternative possessive has limited applicability in that it only agrees with the third person singular. Note that in Latundê the canonical possessive is of type II, although the third person singular possessive behaves in a special way too. Furthermore there is partial overlap in Baure because a possessive morpheme can be attached to the PM in addition to PR agreement. In Table 4 a distinction is made between two varieties of PR marking on the PM (i.e. type II): characteristic (b) concerns agreement marking that is used exclusively in possessive constructions, whereas characteristic (c) concerns person marking that is also used in other constructions, such as verbal cross-reference. With regard to (e), both Arikapú and Latundê have an additional third type of possessive, in which juxtaposition of a personal pronoun and a noun can have a PR-PM interpretation. In Arikapú this is correlated with the possibility of juxtaposing nouns with a possessive interpretation, as mentioned in (f). Only Aikanã, Baure and Kanoê have a special set of possessive pronouns that differs from their personal pronouns, as mentioned in (d). Wari’ is the only exception to the predominant constituent order PR– PM, mentioned in (g). Its constituent order is PM–PR. With regard to (h), half of the languages display (in)alienability effects. However, only Baure, Mekens and Wari’ make a real distinction between obligatorily possessed (in the sources often called ‘inalienable’) and nonobligatorily possessed (often ‘alienable’) nouns. In those languages, the inalienable possessive constructions are less complex than the alienable, as is typologically expected (e.g. Chappell and McGregor 1995: 4, Heine 1997: 172ff., Nichols 1992: 117). Arikapú and Kanoê do not make a clear distinction in the sense that it is not expressed through different possessive constructions. Characteristic (i) concerns several languages in which the PM can be expressed by a classifier. Characteristic (j) is about a specific construction of limited applicability in Baure and Mekens, in which a PR can be expressed by an inflected classifying noun that is juxtaposed to the PM noun. The construction strongly resembles the possessive classifier construction described for Austronesian languages by Lichtenberk (this volume). Elsewhere in South America, the possessive classifier construction was attested in the languages of the Gran Chaco region of Argentine, Bolivia and Paraguay. It is discussed by Fabre (2007), who considers it an areal feature. Interestingly, he includes Chiqui-
Possessive expressions in the Southwestern Amazon
383
tano (possibly a Macro-Jê language) in his discussion. Chiquitano not being remote from Baure, the areal feature discussed by Fabre may have spread all the way into Rondônia. Characteristic (k) concerns three languages of the region where there is a relationship between possessive and benefactive expressions: Aikanã, Baure and Kanoê. Although the constructions differ widely between these languages, their existence might be the result of areal diffusion. Also the close formal resemblance between the possessive and beneficiary suffixes in Kwaza is evident, as is the fact that the genitive and dative in Cavineña are expressed by the same morpheme (Guillaume 2008: 521). Universal tendencies of grammaticalisation may have reinforced a relationship between possessive and benefactive expressions. Heine and Kuteva (2002: 55) list several languages (Modern Arabic, Baka of Cameroon, Nigerian Pidgin English, Tayo Creole French of New Caledonia) in which a benefactive preposition has developed into an adnominal possessive marker. The significance of the parallel distribution of feature (d) concerning the existence of dedicated possessive pronouns is unclear. Characteristic (l) is about the grammaticalised distinction of a PR that is coreferent with the subject of the sentence. The Arikapú coreferential feature is limited to possessive constructions, whereas in Mekens it follows from a general switch-reference marking faculty of the language. The parallel distribution of feature (f) concerning the possessive interpretation of juxtaposed nouns is probably coincidence. As far as we know, the languages discussed in this article are all genealogically unrelated. Three are language isolates and five represent different linguistic families (or ‘stocks’). This diversity seems to be reflected in Table 4 above, although the languages are too few to draw firm conclusions. Nevertheless, Aikanã, Kanoê and Kwaza are clearly similar with regard to characteristic (a). Since these languages, being isolates, are otherwise very different, this similarity may be due to areal diffusion. One could furthermore imagine that the alternative possessive construction of Aikanã and Kwazá in (b) represents an older canonical construction that was replaced. Also the feature (j) may have been subject to areal diffusion, involving Baure and Mekens. Feature (k) shared by Aikanã, Baure and Kanoê could reflect a universal tendency, although it could also be the result of areal diffusion, possibly reinforced by universal tendencies. Possession in the other languages of the region deserves to be investigated as well. Several other thorough descriptions of previously underdocumented languages, besides the ones cited here, have appeared recently, such as e.g. van Gijn (2006) on Yurakare, Haude (2006) on Movima and
384 Hein van der Voort Sakel (2004) on Mosetén, which all are Bolivian isolates, and Guillaume (2008) on Cavineña, a Tacanan language. Furthermore, work on other members of the language families represented here, in particular the Tupí languages, needs to be considered. The study of this corpus may shed additional light on the extent of areal diffusion of specific possessive strategies recurring in this article, or may lead to the discovery of others. Finally, possession in other parts of South America deserves investigation. Especially the Amazonian languages still suffer from a lack of attention in general typological work, whereas their impressive genealogical diversity makes them an important testing ground for universal aspects of grammar. Some of these languages employ rare or unique possessive strategies, like the abovementioned isolate Movima, in which inalienable and predicative possession are expressed by reduplication (Haude 2006: 238ff.). Acknowledgements The Kwaza, Aikanã and Arikapú data in this article are from my own field work in Rondônia during extended periods between 1995 and 2009. Without the help of my language consultants Mario Kwazá, Luiz Aikanã, Manoel Aikanã and Nazaré Arikapú this work would not have been possible. I also want to acknowledge the Netherlands Organization for Scientific Research (NWO) for having generously financed my study of Kwaza, Aikanã and Arikapú during many years. In addition, I would like to express my deep gratitude to the indigenous communities in Rondônia for their hospitality. I am furthermore very grateful to Willem Doelman for providing an excellent map. Finally, I owe many thanks to Swintha Danielsen, Mily Crevels, Rik van Gijn, Olga Krasnoukhova, Bill McGregor, Pieter Muysken, Jan Rijkhoff and Leon Stassen for their valuable comments. Data from other languages cited here are from the work by colleagues, who do not necessarily agree with my analyses, and I apologise for sometimes having strayed from theirs. All errors are mine.
Possessive expressions in the Southwestern Amazon
385
References Araujo, Gabriel Antunes de 2004 A grammar of Sabanê, a Nambikwaran language. PhD thesis. Utrecht: Netherlands Graduate School of Linguistics (LOT). (also available on http://www.lotpublications.nl/publish/articles/000919/bookpart.pdf) Bacelar, Laércio 2004 Gramática da língua Kanoê. PhD thesis, Nijmegen: Katholieke Universiteit Nijmegen. (also available on http://webdoc.ubn.kun.nl/mono/ b/bacelar_l/gramdalik.pdf) Baron, Irène, Michael Herslund and Finn Sørensen (eds.) 2001 Dimensions of Possession, Typological Studies in Language 47. Amsterdam /Philadelphia: John Benjamins. Benveniste, Émile 1960 “‘Être’ et ‘avoir’ dans leurs fonctions linguistiques”. Bulletin de la Société de Linguistique, 55 (1): 113–134. [Reprinted in: Problèmes de linguistique générale, Paris: Gallimard, 1966, 187–207.] Brown, Keith (ed.) 2006 Encyclopedia of Languages and Linguistics, Second edition, Vol. 9. Oxford: Elsevier. Chappell, Hilary and William B. McGregor 1995 Prolegomena to a theory of inalienability. In Hilary Chappell and William B. McGregor (eds.), 3–30. Chappell, Hilary and William B. McGregor (eds.) 1995 The Grammar of Inalienability: A Typological Perspective on Body Part Terms and the Part-Whole Relation, Empirical Approaches to Language Typology 14. Berlin / New York: Mouton de Gruyter. Crevels, Mily and Hein van der Voort 2008 The Guaporé-Mamoré region as a linguistic area. In From Linguistic Areas to Areal Linguistics, Pieter Muysken (ed.), 151–179. Studies in Language Companion Series 90. Amsterdam/ Philadelphia: John Benjamins. Crowley, Terry 1995 Inalienable possession in Paamese grammar. In Hilary Chappell and William B. McGregor (eds.), 383–432. Danielsen, Swintha 2007 A Grammar of Baure. Indigenous Languages of Latin America (ILLA), Vol. 6. Leiden: CNWS Publications. Dixon, Robert M. W. and Alexandra Y. Aikhenvald 1999 Introduction. In Robert M.W. Dixon and Alexandra Y. Aikhenvald (eds.), 1–21. Dixon, Robert M. W. and Alexandra Y. Aikhenvald (eds.) 1999 The Amazonian Languages, Cambridge Language Surveys. Cambridge: Cambridge University Press.
386 Hein van der Voort Dryer, Matthew S. 2007 Noun Phrase structure. In Language Typology and Syntactic Description, Volume II: Complex Constructions, Timothy Shopen (ed.), 151–205. Second edition. Cambridge: Cambridge University Press. Everett, Daniel L. and Barbara Kern 1997 Wari’: The Pacaas Novos Language of Western Brazil, London: Routledge. Fabre, Alain 2007 Morfosintaxis de los clasificadores posesivos en las lenguas del Gran Chaco (Argentina, Bolivia y Paraguay). UniverSOS. Revista de Lenguas Indígenas y Universos Culturales 4: 67–85. Galucio, Ana Vilacy 2001 The morphosyntax of Mekens (Tupi). PhD thesis, University of Chicago. Gijn, Rik van 2006 A grammar of Yurakaré. PhD thesis, Radboud Universiteit Nijmegen. (Avail. on http://webdoc.ubn.ru.nl/mono/g/gijn_e_van/gramofyu.pdf) Gil, David 2001 Escaping Eurocentrism: fieldwork as a process of unlearning. In Linguistic Fieldwork, Paul Newman and Martha Ratliff (eds.), 102–132. Cambridge: Cambridge University Press. Guillaume, Antoine 2008 A Grammar of Cavineña. Mouton Grammar Library 44. Berlin /New York: Mouton de Gruyter. Haude, Katharina 2006 A Grammar of Movima. PhD thesis, Radboud Universiteit Nijmegen. (Avail. on http://webdoc.ubn.ru.nl/mono/h/ haude_k/gramofmo.pdf) Heine, Bernd 1997 Possession: Cognitive Sources, Forces, and Grammaticalization. Cambridge Studies in Linguistics 83. Cambridge: Cambridge University Press. Heine, Bernd and Tania Kuteva 2002 World Lexicon of Grammaticalization. Cambridge: Cambridge University Press. Herslund, Michael and Irène Baron 2001 Introduction: Dimensions of possession. In Baron et al. (eds.), 1–25. Hinton, Leanne (ed.) 1993 Aikana Modules: A class report on the fieldnotes of Harvey Carlson. Typescript, 158 pp. Berkeley: University of California. Koptjevskaja-Tamm, Maria 2006 Possession, Adnominal. In Keith Brown (ed.), 765–769.
Possessive expressions in the Southwestern Amazon
387
Kroeker, Menno 2001 A descriptive grammar of Nambikuara. International Journal of American Linguistics, 67 (1): 1–87. Lichtenberk, Frantisek this vol. Attributive possessive constructions in Oceanic. McGregor, William B. 2001 The verb HAVE in Nyulnyulan languages. In Baron et al. (eds.), 67–98. Messineo, Cristina and Ana Gerzenstein 2007 La posesión en dos lenguas indígenas del Gran Chaco: toba (guaycurú) y maká (mataguayo). Línguas Indígenas Americanas (LIAMES) 7: 61–79. Nichols, Johanna 1992 Linguistic Diversity in Space and Time, Chicago: University of Chicago Press. Payne, Doris L. and Immanuel Barshi (eds) 1999 External Possession. Typological Studies in Language 39. Amsterdam / Philadelphia: John Benjamins. Ribeiro, Eduardo and Hein van der Voort in press Nimuendajú was right: The inclusion of the Jabuti language family in the Macro-Jê stock. International Journal of American Linguistics 76 (4). Rijkhoff, Jan. this vol. On the co-variation between form and function of adnominal possessive modifiers in Dutch and English. Rodrigues, Aryon Dall’Igna 1986 Línguas brasileiras: Para o conhecimento das línguas indígenas, São Paulo: Edições Loyola. Rodrigues, Aryon Dall’Igna 1999 Macro-Jê. In Robert M.W. Dixon and Alexandra Y. Aikhenvald (eds.), 165–206. Sakel, Jeanette 2004 A Grammar of Mosetén. Mouton Grammar Library 33. Berlin /New York: Mouton de Gruyter. Seiler, Hansjakob 1977 On the semantico-syntactic configuration ‘Possessor of an act’. In Sprache und Sprachen: Gesammelte Aufsätze, Hansjakob Seiler (ed.), 169–186. München: Wilhelm Fink. Stassen, Leon 2006 Possession, Predicative. In Keith Brown (ed.), 769 –773. Stassen, Leon 2009 Predicative Possession. Oxford Studies in Typology and Linguistic Theory. Oxford: Oxford University Press.
388 Hein van der Voort Telles, Stella 2002a Fonologia e gramática Latundê/ Lakondê. PhD thesis, Vrije Universiteit Amsterdam. Telles, Stella 2002b Construções possessivas em Latundê. In Línguas Indígenas Brasileiras: Fonologia, gramática e história, Ana Suelly Arruda Câmara Cabral and Aryon Dall’Igna Rodrigues (org.), 157–163. Atas do I. Encontro Internacional do GTLI, tomo II. Belém: Editora Universitária UFPA. Vasconcelos, Ione 2003 Aspectos fonológicos e morfofonológicos da língua Aikanã. PhD thesis, Maceió: Universidade Federal de Alagoas. Voort, Hein van der 2004 A Grammar of Kwaza. Mouton Grammar Library 29. Berlin /New York: Mouton de Gruyter. Voort, Hein van der 2005 Kwaza in a comparative perspective. International Journal of American Linguistics, 71 (4): 365–412. Voort, Hein van der 2006 Construções atributivas em Kwazá. Boletim do Museu Paraense Emílio Goeldi (Ciências humanas), 1 (1): 87–104. (Also available on http://marte.museu-goeldi.br/seer/index.php/boletimhumanas/article/ viewFile/5/30)
Possession in the visual-gestural modality: How possession is expressed in British Sign Language Kearsy Cormier and Jordan Fenlon
1. Introduction All known languages have some way of expressing possession, and signed languages are no exception. The primary question we pose in this chapter is: Is possession expressed differently in signed languages due to the use of the visual-gestural (rather than the aural-oral) modality, or are patterns relating to possession essentially the same for signed and spoken languages? Our chapter begins with background about British Sign Language (BSL), followed by an overview of the pronominal system of BSL. We then move on to look at attributive and predicative possession in BSL and finally an adjectival predicate of predisposition in BSL closely related to the possessive pronoun. Although this chapter is primarily an overview of possession in BSL, we will also include observations on other signed languages, such as American Sign Language (ASL), where applicable. 2. British Sign Language BSL is the natural signed language used by the deaf community in the United Kingdom. BSL has its own phonological, morphological and syntactic structure. Phonologically, lexical signs in BSL are made up of four primary parameters: handshape, movement, location (i.e. place of articulation) and palm/finger orientation. Every sign is specified for each of these parameters, which are phonologically contrastive (i.e. minimal pairs can be identified with each parameter). Clearly, BSL is neither simply an elaborated gestural system nor a manual code based on English. Nonetheless, because BSL is a minority language within the United Kingdom, there is strong and constant contact between BSL and English and thus BSL does borrow elements from English. Having said that, we will point out instances where there are possessive constructions within signed languages which are known borrowings from the surrounding spoken language.
390 Kearsy Cormier and Jordan Fenlon The examples in this chapter come from a variety of sources. Some are from or based on elicited data, some are from broadcast television programmes, some are from an annotated corpus of BSL stories and fables (Woll, Sutton-Spence, and Waters 2004), and some are from naturalistic video-recorded conversation. In all cases, examples are from or based on signing from both native and non-native signers. To be representative of the community, it is important to include data from both native and non-native signers, since only a very small percentage of the British Deaf community, generally considered to be 5–10%, are native signers born to deaf parents, and thus non-native signers (deaf with hearing and usually non-signing parents) make up the vast majority of the typical deaf population.1 3. Overview of pronouns and determiners One of the striking characteristics about signed languages is the way that signers use the space around them for referential purposes. Signers refer to a person/object physically present within a discourse situation simply by pointing to him/her/it. For referents who are not present, the signer establishes a location in space for the referent by pointing to that location. These pointing signs acts as pronouns and are glossed here as PRO-1 (first person) or PRO-non1 (for non-first person) – see Figures 1a and 1b.2 The first person singular pronoun consists of a point to the signer’s own chest. 1
2
The proportion of native signers in the UK is generally considered to be approximately 5–10%, following statistics reported in Australia and the United States (Deaf Society of New South Wales 1998; Mitchell and Karchmer 2004). Following conventions in the sign language literature, glosses are given for signs using all caps. Signs which require more than one English word for translation are glossed with English words separated by hyphens (e.g. ICE-CREAM). Fingerspelled words are indicated with hyphens in between letters (e.g. J-O-E) – see also footnote 9. Pronouns are glossed as PRO-1 for first person and PROnon1 for non-first person. Possessive pronouns are glossed as POSS-1 for first person and POSS-non1 for non-first person. For ease of exposition, the dual and number-incorporated pronouns will be notated with a simple English gloss instead (e.g. TWO-OF-US, THREE-OF-THEM). Within examples, when indexing different non-first person locations, PRO-2/POSS-2 is used for forms which index the location of the addressee. Indices such as a, b, c (e.g. PRO-a, POSS-a) are used to index non-addressed participants. Unless otherwise noted, where indices occur with nouns (e.g. BOX-a), the noun sign may have been established in that locus previously in the discourse, or some sign within the example establishes
Possession in the visual-gestural modality 391
Figure 1a. PRO-1 ‘me’
Figure 1b. PRO-non1 ‘him/her/it’
3.1. Person There has been a fair amount of debate within the sign language literature regarding the issue of person marking on pronouns in signed languages. In the early days of sign language research, researchers assumed a threeperson system analogous to those found in spoken languages (Friedman 1975; Klima and Bellugi 1979; Padden 1983, 1990). A three-person system is problematic, however, because there is no listable set of location values in the signing space to which a non-first person pronoun may point, for addressee or non-addressed participants. That is, there is no single location that may act as a morpheme to indicate second or third person. To address this issue, some researchers have taken the view that sign language pronouns do not exhibit person marking at all and that locations associated with pronouns instead act as variables (‘loci’) whose content comes from discourse (Cormier, Wechsler, and Meier 1999; Lillo-Martin and Klima 1990). More recently, most sign linguists subscribe to a two-person system as proposed by Meier (1990) for ASL. Such an analysis recognises the ‘listability problem’ (Rathmann and Mathur 2002) of multiple second / third person location values while at the same time recognising the special status the noun at that locus. Where repetitions of movement occur, the number of path movements is indicated with ‘+’ (e.g. POSS-non1++ has two movement paths). In English translations of examples, for ease of exposition, different gendered pronouns (e.g. he vs. she) are used to distinguish different referents, though BSL pronouns do not mark gender.
392 Kearsy Cormier and Jordan Fenlon of first person, for which there is only one location (the signer’s chest). This two-person system has been assumed by other researchers for ASL and other signed languages (Engberg-Pedersen 1993; Farris 1994, 1998; Liddell 2003), including BSL (Cormier 2007), and will be assumed in this chapter as well.3 3.2. Number In terms of number marking, BSL has singular, dual, number-incorporated and plural pronouns, as shown in Table 1. Table 1.
Personal pronouns in BSL
Singular Dual Number-incorporated (3/4/5) Plural
First person
Non-first person
PRO-1 (‘I’) TWO-OF-US 3/4/5-OF-US PRO-1pl (‘we’)
PRO-non1 (‘you/he/she/it’) TWO-OF-YOU/THEM 3/4/5-OF-YOU/THEM PRO-non1pl (‘they’)
The primary reason to distinguish dual pronouns from number-incorporated pronouns that represent three, four or five referents in BSL is the way these pronouns index their referents. The dual pronoun behaves as a singular pronoun by indexing the location associated with each referent. Specifically, the dual pronoun moves between the two locations associated with each referent. If one of the indexed locations is the location of the signer’s own body, then this form may be glossed as TWO-OF-US (shown in Figure 2); if not then it may be glossed as TWO-OF-YOU (if the location of the addressee is included) or TWO-OF-THEM. The number-incorporated pronouns (e.g. THREE-OF-US shown in Figure 3), on the other hand, only index a single general location for all three, four or five referents; these pronouns have a circular movement which may or may not encompass the locations associated with each individual referent.
3
Not all sign language researchers subscribe to a two-person analysis for signed languages – e.g. Berenz (2002) argues for a three-person system for Brazilian and American Sign Languages, as do Alibasic, Ciciliani and Wilbur (2006) for Croatian Sign Language.
Possession in the visual-gestural modality 393
Figure 2. TWO-OF-US
Figure 3. THREE-OF-US
The non-first person plural pronoun (PRO-non1pl) is similar to the numberincorporated pronouns in that it only generally indexes the locations of its referents. This pronoun is a pointing sign which has a sweeping movement across a span of space associated with the referents. To index the locations associated with each of more than two referents, signers use a sequence of singular and/or dual (indexing) pronouns, as in (1). (1)
PRO-a PRO-b PRO-c ALL LIKE CAT ‘That person, that one and that one all like cats.’
The first person plural pronoun PRO-1pl is produced with an index handshape at the signer’s chest but, instead of pointing directly at the signer’s chest, in the plural form the index finger traces a small circular movement just in front of the chest. This form is the least indexic of all the pronouns – i.e. the other pronouns point to the location(s) associated with their referents. The first person plural pronoun only indexes (points to) the signer’s chest and does not index the other referents.
394 Kearsy Cormier and Jordan Fenlon 3.3. Exclusive pronouns Cormier (2007) has investigated whether BSL has an inclusive/exclusive distinction within first person plural pronouns.4 The findings indicated that within first person plurals in BSL, the citation forms of these pronouns can be used in either inclusive or exclusive contexts. The citation form for the first person plural pronoun as shown in Figure 4a is produced at the centre of the signer’s chest (though the location can change to the right or left of the signer’s chest to index location of the group, as described above). BSL also has variants of plurals which are exclusive forms. When a referent is excluded, the signer displaces the plural pronoun to the right or left side of his/her chest as shown in Figure 4b. The excluded referent(s) may be the addressee(s) or any other non-first person referent(s) which is/are salient in the discourse. The location of the pronoun gives no information about the perceived location of the excluded referents, only that they are in a location other than that of the included referents.
Figure 4a. PRO-1pl (citation form) ‘we’
4
Figure 4b. PRO-1pl (displaced) ‘we’ (excluding someone salient in the discourse)
A similar system of exclusive pronouns has also been identified in ASL – for more see Cormier (2005, 2007).
Possession in the visual-gestural modality 395
Exclusive forms were identified with first person plurals (PRO-1pl) and with first person number-incorporated plurals (3/4/5-OF-US). Because the first-person dual pronoun indexes the location of each referent, no grammatical exclusive form is proposed for the dual pronoun, only plurals and number-incorporated forms. 3.4. Word order and pro-drop Syntactically pronouns in BSL occur in situ in the position of the subject or the object, as in (2) and (3).5 Morphological case systems have not been identified for any known signed languages (Alibasic Ciciliani and Wilbur 2006), so pronouns do not vary in form based on whether they occur in subject or object position. (2)
PRO-1 LIKE ICE-CREAM ‘I like ice-cream.’
(3)
BOSS LIKE PRO-1 ‘The boss likes me.’
Pro-drop (with either the subject or object pronoun) is common in BSL and other signed languages, particularly with singular pronouns, when the referent is retrievable from context, as in (4). With plural pronouns, studies of ASL and Australian Sign Language (Auslan) have shown that first person plural pronouns are dropped more often than first person singular pronouns (Schembri and Johnston 2006; Wulf, Dudis, Bayley, and Lucas 2002). With double agreement verbs (a type of verb in which the start and end location of the verb or the direction in which the verb is facing reflects the subject / source and object/goal of the verb, respectively), pro-drop is usually the rule rather then the exception, particularly with objects, as in (5). 5
Word order in BSL like other signed languages is quite variable on the surface. Grammatical processes like topic marking make many different word orders possible. There have been attempts by sign language researchers to posit a basic default order from which other orders are derived. Such researchers have claimed that the basic word order for ASL is SVO (Fischer 1975; Neidle, Kegl, MacLaughlin, Bahan, and Lee 2000). There has not been as much work on word order in BSL but it appears that BSL is also, underlyingly at least, SVO, although see Deuchar (1983) for a different perspective.
396 Kearsy Cormier and Jordan Fenlon (4)
LIKE ICE-CREAM ‘I like ice-cream.’
(5)
BEEN a-HELP-1 ‘He helped me.’
Pronoun copy is also common, whereby the pronoun in situ is then copied and produced again at the end of the clause as in (6). (6)
PRO-1 LIKE ICE-CREAM PRO-1 ‘I like ice cream.’
This section has provided an overview of the pronominal system of BSL, specifically personal pronouns. This is relevant for considering possessive pronouns in BSL, to which we turn next. 4.
Possession in BSL
As in spoken languages, signed languages exhibit both attributive and predicative possession. We first look at attributive possession, expressed primarily by possessive pronouns. 4.1.
Attributive possession
The singular and plural possessive pronouns in BSL behave very similarly to the singular and plural personal pronouns, respectively. The primary difference is the handshape. While the personal pronoun PRO uses an ‘index’ handshape (extended index finger similar to the handshape used in a pointing gesture), the possessive pronoun POSS uses a fist handshape with the thumb alongside the index finger or wrapped across the other fingers. For the first person form, the back of the fingers contact the signer’s chest, and for the non-first person form the back of the fingers face outward away from the signer, toward the location associated with the referent. 4.1.1. Number Table 2 shows the BSL possessive pronouns by person and number.
Possession in the visual-gestural modality 397 Table 2. Possessive pronouns in BSL
Singular or collective plural Plural
First person
Non-first person
POSS-1 ‘my’ POSS-1pl ‘our’
POSS-non1 ‘your/his/her/its’ POSS-non1pl ‘their’
The singular forms are shown in Figures 5 and 6. Singular possessive pronouns representing the possessor (PR) may occur on their own with the possessum (PM) noun (e.g. POSS-a CAR ‘his car’) or in apposition with a personal pronoun (e.g. PRO-a POSS-a CAR ‘his car’). (See § 4.1.2 for the use of POSS with nominal PRs.)
Figure 5. POSS-1 ‘my’
Figure 6. POSS-non1 ‘your’
Plural possessive pronouns use the same handshape as the singular possessive forms. The non-first person plural possessive pronoun POSS-non1pl ‘their’ is very similar to the pronoun PRO-non1pl ‘them’ (with a sweeping motion across the locations associated with the referents) but uses a fist handshape, as shown in Figure 8. The first person plural possessive pronoun POSS-1pl ‘our’ as shown in Figure 7 is the same as the first person plural pronoun (PRO-1pl ‘we’: circular motion at the chest) but has a fist handshape; the back of the fingers remain facing the signer during the production of the sign.
398 Kearsy Cormier and Jordan Fenlon
Figure 7. BSL POSS-1pl ‘our’
Figure 8. BSL POSS-non1pl ‘their’
With plurals, signers may use a combination of possessive pronoun(s) with multiple singular pronouns, in which case the personal pronouns typically precede the possessive pronoun(s) as in (7). (7)
PRO-a PRO-b PRO-1 POSS-1pl PARTY TONIGHT ‘Our party (the party of him, her and me) will be tonight.’
Possession in the visual-gestural modality 399
To specify distribution of possession, signers may use a sequence of singular possessive pronouns as in (8) – cf. collective possession as in (12a). (8)
POSS-a POSS-b POSS-1 TELEVISON BUY SAME SHOP ‘We (he, she and I) bought our televisions from the same shop.’
Although these plural forms do exist and are used, it is also possible (and quite common) for BSL signers to use the singular possessive instead of the plural with plural referents, both in first and non-first person, for collective plurals, as shown in (9a). If the number of referents is not already known or not clear in context, then a plural personal pronoun or other quantifier may accompany the singular possessive; in this case, again, the pronoun/quantifier generally precedes the possessive but can occur afterwards. In (9b) and (9c), the non-first person possessive is directed towards the same general location as the quantifier/pronoun which occurs in the same utterance.6 In (9d) and (9e), the singular first person possessive is, as always, directed towards the signer’s chest. If a plural pronoun or quantifier is used in combination with a possessive, the data and intuitions from deaf informants indicate that the possessive must be singular and cannot be plural, as shown in (9f). (9) a. WHAT ABOUT PARENTS MAKE POSS-non1-emp DECISION7 ‘What about parents making their own decision?’ b. THERE PEOPLE ALL-OF-THEM POSS-non1 BELIEF DIFFERENT ‘There, people have different beliefs.’ c. NOW WANT FOCUS POSS-non1 PERSONALITY TWO-OF-YOU ‘Now I want to focus on your (both of your) personalities.’ d. THREE-OF-US POSS-1 HOUSE SELL ‘The house belonging to the three of us is for sale.’ e. TWO-OF-US HOUSE POSS-1 SELL ‘The house belonging to the two of us is for sale.’ f. *THREE-OF-US POSS-1pl HOUSE SELL ‘The house belonging to the three of us is for sale.’ 6
7
With non-first person forms, singular collective possessive pronouns are directed towards a location in space that appears approximately equidistant between the multiple locations associated with the referents (if dual or plural), or the same general location of the pronoun (if a number-incorporated pronoun). This particular example is an emphatic use of the possessive, as described in § 4.1.5.
400 Kearsy Cormier and Jordan Fenlon The use of singular possessive pronouns as a way of marking plural possession in a collective context should not be surprising, as the nature of collective plurals crosslinguistically is that they act as singulars. However, these singulars used as collectives are more interesting when considering their indexicality (i.e. the extent to which they point to or index the locations associated with their referents). While the personal pronouns in these combinations may be providing indexic information (e.g. the dual and plural forms which point to more than one location), the collective plural possessive forms in these constructions have lost much of their indexicality (compared to truly plural possessives) in that they only index a single location. As noted above, the first person plural personal pronoun PRO-1pl (as well as the first person plural possessive pronoun POSS-1pl) are already less indexic than their singular counterparts. Because the collective plural form is the same as the singular form with the same degree of indexicality, these collective plurals are in fact not marking number at all but only person (first or non-first). Thus when a plural personal pronoun occurs which marks number, as in (9f), an additional plural possessive form is not acceptable. For distributive readings (e.g. multiple objects that are each possessed separately by different people, as opposed to the default collective reading – the single object that is possessed together by more than one person), signers use multiple possessive pronouns directed to the location of each referent, either juxtaposed (10a) or with multiple PM nouns (10b). (10) a. POSS-2 POSS-1 LAND CONTINUE SELL ‘Your land and my land are still for sale.’ b. POSS-2 LAND POSS-a LAND POSS-1 LAND CONTINUE SELL ‘Your land, his land and my land are still for sale.’ Personal pronouns may also be used in addition to the possessives as in (11). These lists are not necessarily exhaustive, so that (11) may indicate either precisely three land owners or it may indicate a number larger than three. (11) POSS-a POSS-b POSS-1 LAND PRO-a PRO-b PRO-1 WILL SELL ‘His land, her land and my land’/‘The land that each of us owns will be sold.’
Possession in the visual-gestural modality 401
4.1.2. Word order and pro-drop As seen in the examples above, possessive pronouns in BSL occur before the noun they modify (12a), and optionally may be copied immediately after the noun as well, as in (12b). Possessive pronouns may also occur only after the nouns they modify (e.g. TELEVISION POSS-1pl NEW), though the patterns in (12a) and (12b) appear to be preferred. (12) a. POSS-1pl TELEVISION NEW ‘Our television is new.’ b. POSS-1pl TELEVISION POSS-1pl NEW ‘Our television is new.’ With a nominal PR, the PR, possessive pronoun and PM are placed in apposition as in (13). The possessive pronoun usually follows the PR, such that the typical order is PR POSS PM.8 This same ordering is identified for nominal PRs in ASL, Croatian Sign Language, and Austrian Sign Language (Pichler et al. 2008). (13) JOHN POSS-non1 CAR NICE ‘John’s car is nice’ Similarly to pro-drop with personal pronouns, possessive pronouns are not required if the PR can be inferred from context as shown in (14). (14) NAME WHAT? ‘What is your/his/her name?’
8
It is possible for the possessive pronoun to precede the possessor in both BSL and ASL (e.g., POSS-non1 JOHN CAR NICE ‘John’s car is nice’). MacLaughlin (1997) shows that this construction (POSS PR PM) can only be used in ASL when the possessor in question has been established previously (i.e. it cannot be used with indefinite possessors) and that this construction is strongly preferred in topic position. According to Neidle, Kegl, MacLaughlin, Bahan and Lee (2000), “the proper analysis of this construction remains something of a mystery” (p. 182). We only identified one token of this ordering in our corpus of BSL data; by far the most frequent ordering was PR POSS PM.
402 Kearsy Cormier and Jordan Fenlon 4.1.3. Abstract PMs Attributive possessive pronouns can be used with either concrete entities as in many of the examples above, or with abstract concepts, such as ‘time’, ‘view’, or ‘fault’, as in (15) below. (15) a. COMPARE POSS-1 TIME NOTHING ‘Compared to my time there was nothing.’ b. POSS-1 VIEW ‘That’s my view.’ c. BUT POSS-non1 FAULT ‘But it was her fault.’ 4.1.4. Exclusive possessive pronouns First person plural possessive pronouns in BSL also have an exclusive form, similar to the exclusive personal pronouns described in § 3.3. The exclusive form for BSL shown in Figure 9 below is the same as the citation form as in Figure 7 above but is displaced to the signer’s left or right side, and this form may exclude any salient referent.
Figure 9. POSS-1pl (displaced) ‘our’ (belonging to us, excluding someone salient in the discourse)
Possession in the visual-gestural modality 403
4.1.5. Emphatic possessive pronouns Possessive pronouns may be made emphatic by producing the possessive with a sharp sweeping movement towards the location associated with the referent, with either one or two hands, as shown in Figures 10a and 10b and in (16).
Figure 10a. POSS-non1-emp (one hand)
Figure 10b. POSS-non1-emp (two hands)
404 Kearsy Cormier and Jordan Fenlon (16) a. BUY CAR YESTERDAY. NICE POSS-1-emp CAR. FREE DRIVEABOUT. ‘I bought a car yesterday. It’s nice to have my own car. I’m free to drive anywhere.’ b. WHAT ABOUT PARENTS MAKE POSS-non1-emp DECISION ‘What about parents making their own decision?’ c. POSS-non1-emp BUSINESS MEDIA, WOW ‘Your own business in media – wow!’ d. PRO-1 POSS-1-emp ROOM ‘I had my own room.’ 4.1.6. Morphological marking of possession Possession in signed languages that have been documented to date is generally marked using possessive pronouns rather than morphologically. BSL does not have an equivalent to the English -’s possessive marker in widespread usage today. However, it is reported that older BSL signers used (and some still use) a form which consists of a modified form of the fingerspelled letter -S- cliticised to the end of the noun PR. Auslan, a language historically related to BSL, has retained this form in both older and younger signers, shown below in Figure 11 (Branson, Toms, Bernal, and Miller 1995; Johnston and Schembri 2007).9
9
Fingerspelling systems within signed languages are based on the written alphabet of the surrounding hearing community and are used for various purposes, including proper names and other concepts for which a native lexical sign may not exist. Auslan is historically related to BSL; Auslan, BSL and New Zealand Sign Language are generally considered to be dialects of the same language (Johnston 2003; McKee and Kennedy 2000), and all three languages share a common twohanded fingerspelling system.
Possession in the visual-gestural modality 405
Figure 11. Auslan MOTHER’S (MOTHER + ’S) (Reprinted with permission from T. Johnston & A. Schembri (2007). Australian Sign Language: An introduction to sign language linguistics. Cambridge: Cambridge University Press.)
ASL also has a similar affixal ’S possessive marker. Pichler et al. (2008) note than although this form was likely borrowed from Signed English systems (sign systems which are not natural signed languages but are based on English morphology and syntax), it is judged acceptable by ASL signers, particularly in utterances with multiple possessors, as in FATHER ’S BROTHER ’S WIFE. This marker (glossed by Pichler et al. as APOSTROPHE-S) consists of a modified form of the letter -S- from the fingerspelling system used by ASL signers. Unlike English, these affixal possessive markers are not obligatory for expressing possession with nominal PRs in Auslan and ASL, and are used along with other types of possession marking as described in this chapter. 4.1.7. Inalienable possession It has been suggested that BSL systematically encodes inalienable possession (Fenlon and Cormier 2006; Sutton-Spence and Woll 1999). Inalienable possession refers to nouns that are inherently linked to the PR (such as kinship terms and body-part nouns) as opposed to alienable nouns which are
406 Kearsy Cormier and Jordan Fenlon separable from the PR (e.g., a book). This distinction is encoded in BSL by using a form identical to the personal pronoun PRO in place of the possessive pronoun POSS with inalienable nouns; see (17). (17) a. PRO-1 MOTHER ‘my mother’ b. PRO-1 LEG ‘my leg’ c. POSS-non1 BOOK ‘your book’ Inalienable possession in spoken languages is strongly associated with two semantic categories: kinship terms and body-part nouns. However, this feature is not exclusive to these categories and has been shown in some languages to extend to personal possessions, clothing and physical and mental characteristics (Cooper 2002). In keeping with this pattern, BSL uses a form identical to the personal pronoun PRO in place of the possessive pronoun POSS with kinship terms and body part nouns as well as NAME as in (18a) and other personal possessions such as HOUSE as in (18b). (18) a. PRO-1 NAME J-O-E ‘My name is Joe.’ b. FRIEND STAY PRO-1 HOUSE PRO-1 ‘Friends stay at my house.’ However, the issue of whether BSL encodes inalienable possession grammatically can be called into question. Firstly, although personal pronouns are used with inalienable phenomena such as body-part nouns, kinship terms and NAME, it appears that signers show a preference for the possessive pronoun POSS (rather than the personal pronoun PRO) with kinship terms. Secondly, it is unclear in examples such as (18a) whether NAME can be classed as a noun or a verb. If it is a verb, then the PRO is functioning as a personal pronoun (e.g. I am called Joe) as opposed to a possessive pronoun (e.g. My name is Joe). Despite problems with claiming that BSL grammaticises inalienability, patterns of inalienability in BSL are consistent with patterns across spoken languages. Specifically, BSL fits the crosslinguistic pattern in which kinship terms and body-part nouns are inalienable in languages that have the distinction. It has also been shown crosslinguistically that the existence of
Possession in the visual-gestural modality 407
other semantic categories marked as inalienable in a given language (beyond kinship and body part terms) implies that body parts and/or kinship terms will also be marked as inalienable (Kliffer 1996; Nichols 1988, 1992); this generalisation fits BSL as well. Finally, in some spoken languages (such as Navajo) the alienable/inalienable distinction is optional as it appears to be in BSL as well (Nichols 1992). In addition, our BSL data indicates that it is possible to express inalienable possession with overt nominal PRs through the juxtaposition of the PR and PM, as in (19). Similar observations have been made for ASL, Croatian Sign Language, and Austrian Sign Language in Pichler et al. (2008) where the construction PR PM is strongly preferred over PR POSS PM for inalienable constructions. (19) a. BABY EAR ‘the baby’s ear’ b. WOMAN SON ‘the woman’s son’ To our knowledge, the use of two distinct forms marking possession (where one encodes inalienable possession and the other marks alienable possession, as in (17) above) has not been identified in any other signed language in a possessive construction without an overt nominal PR. Use of the personal pronoun instead of the possessive pronoun has been noted for some other signed languages – specifically, ASL, Croatian Sign Language, and Austrian Sign Language – but the claim with these signed languages is that the personal pronoun PRO may optionally be used instead of POSS in any possessive construction, not just with inalienable nouns (Pichler et al. 2008). 4.1.8. Spatial marking In addition to the use of possessive pronouns, BSL can also associate PMs with PRs by changing the location of some noun signs directly within the signing space. The primary criterion for such spatial marking of noun signs is that the place of articulation of the sign must be the neutral space in front of the signer, not a location on the body; the sign may then be located at a particular location in space associated with some referent. It is not entirely clear if this spatially marks PMs only for possession (‘my house’, ‘your house’) or if the marking is at particular locations which just so happen to correspond to locations associated with PRs (‘the house here/associated
408 Kearsy Cormier and Jordan Fenlon with me’, ‘the house there/associated with you’). In (20a) and (20b), the sign HOUSE is produced at two different locations. Note the first person pronoun PRO-1 is used as a possessive in (20a), as in (18b) above, and thus the spatial marking of HOUSE at the first person locus is redundant, while in (20b) both tokens of the pronoun PRO-1 refer to the subject of the clause, not to the PR, so the only indication of the PR here is the non-first person spatial marking of the noun HOUSE. (20) a. FRIEND STAY PRO-1 HOUSE PRO-1 … ‘Friends would stay at my house …’ (HOUSE is signed close to the signer chest.) b. INVITE SAME CHANGE STAY PRO-1 GO HOUSE PRO-1 GO ‘Then they would invite me to go to their house.’ (HOUSE is signed away from the signer chest) The grammaticisation of locative markers into possession markers is well established for many spoken languages (Heine 1997), and it may be that this same process is occurring/has occurred in BSL. Further evidence for the strong connection between locatives and possession comes from Pichler and colleagues (2008), who have found that the predicative possessive in Austrian Sign Language is expressed by using a sign which appears to be derived from the sign DA meaning ‘here’. We now turn to predicative possession in BSL.
4.2. Predicative possession There are two primary ways to mark predicative possession in BSL – HAVE and BELONG. These two types are attested crosslinguistically amongst spoken languages and as in other languages differ in whether the PR (as with HAVE) or the PM (as with BELONG) is the subject or topic of the sentence (Heine 1997; Herslund and Baron 2001). 4.2.1. Predicative expression of possession and existence The lexical verb HAVE in BSL (shown below in Figure 12) can be used to indicate possession. The PR, if overtly expressed, acts as subject of the sentence. HAVE (glossed here as HAVEposs) can be used with any type of alienable possession, as in (21).
Possession in the visual-gestural modality 409
Figure 12. HAVEposs/ex
(21) a. JOHN HAVEposs CAR ‘John has a car.’ b. BOOK HAVEposs ‘I have a book.’ As with many other languages, the same lexical item HAVE which is used for possession can also be used for existence (Heine 1997), glossed in (22) as HAVEex. (22) a. GOD HAVEex ‘There is a God.’ b. HAVEex GOSSIP ‘There was gossip.’ BSL also has signs for negative possession and negative existence. For negation of possession, the sign HAVE-NEGposs is phonologically related to the sign HAVE, with what Brennan (1992) described as a negative affix occurring on some BSL signs (forearm rotation accompanied by an opening of the hand; see Figure 13). This sign is used for possession only (23a), not existence (23b). There are other variants of this sign also meaning ‘not have’, including one which begins with two 5-hands palm down and ends with both hands pronated so that the palms are facing upwards.
410 Kearsy Cormier and Jordan Fenlon
Figure 13. HAVE-NEGposs
(23) a. CAR HAVE-NEGposs ‘I/someone doesn’t have a car.’ b. *GOD HAVE-NEGposs ‘There is no God.’ The sign NOT-HAVE, a sign formationally unrelated to HAVE (shown in Figure 14) can be used to negate either possession (examples (24a) and (24b)) or existence ((24b) and (24c)). A more general negation sign NOTHING (as shown in Figure 15) can also be used predicatively to negate either possession (25a) or existence (25b), in addition to its (pro)nominal usage (25c).
Figure 14. NOT-HAVEposs/ex
Possession in the visual-gestural modality 411
(24) a. PRO-1 CAR NOT-HAVEposs ‘I don’t have a car.’ b. CAR NOT-HAVEposs/ex ‘There is no car.’/‘I/someone doesn’t have a car.’ c. GOD NOT-HAVEex ‘There is no God.’
Figure 15. NOTHING
(25) a. WILLIAM NOTHING HUSBAND ‘William didn’t have a husband.’ b. BISCUIT NOTHING ‘There are no biscuits’ c. OPEN-CUPBOARD BUT NOTHING ‘(I) opened the cupboard but there was nothing there.’ Interestingly, according to Pichler et al. (2008), in ASL, Croatian Sign Language, and Austrian Sign Language, the lexical sign which is used for predicative possession is the same as the lexical sign used for existence (HAVE, IMATI, and DA, respectively), and the same is true for negative possession and negative existence (NONE, NEMATI, and KEIN (DA), respectively). Kristoffersen (2003) and Arik (2008) note the same for Danish Sign Language and Turkish Sign Language respectively – i.e. that the same lexical items are used both for possession and existence. The fact that expression of possession and existence patterns differently in BSL from these other signed languages reminds us that, although there may be an underlying
412 Kearsy Cormier and Jordan Fenlon and historical relationship between possession, existence, and locatives (Freeze 1992), these constructions do surface differently in some languages. We return to the relationship between possession and locatives in § 6.
4.2.2. BELONG BSL has another predicative possessive form which foregrounds the PM rather than the PR. The sign BELONG makes reference to the location of not only the PR (as possessive pronouns do) but also the PM. This form uses the same fist handshape as the possessive pronouns. But rather than simply directing the hand towards the PR referent, BELONG begins at the location associated with the PM noun (the person or object being possessed, L1 in Figure 16) and ends with the location associated with the PR (L2 in Figure 16). Thus whether the PM is named before BELONG as in (26a) or after BELONG as in (26b), in each case, the location of BELONG begins at the location associated with the PM. It is not acceptable for BELONG to begin with the location of the PR and end with the location of the PM, as in (26d).
Figure 16. BSL a-BELONG-b
(26) a. BOOK-a a-BELONG-b TEACHER-a ‘That book is the teacher’s.’ b. a-BELONG-2 BEDROOM-a? ‘Is that your bedroom? c. a-BELONG-1 (where a is the location already associated with a box) ‘That box is mine’
Possession in the visual-gestural modality 413
d. *1-BELONG-a ‘That box is mine’ In (26), the referents ‘book’, ‘teacher’, and ‘bedroom’ may have been set up in space already within the discourse, or they may not have been. If they have not, the sign BELONG establishes the loci for each. In example (26c), either the location of the box has already been established in space previously in the discourse, or the box is physically present during the utterance. The verb BELONG falls into the category of verbs in BSL known as agreement verbs, described in § 3.4 above, in which the verb begins at the location associated with the subject and ends at the location associated with the object. Agreement verbs exist in all known signed languages. However, a predicative possession form like BELONG in BSL which acts as an agreement verb has not to our knowledge been identified in any other sign language outside of the British Sign Language family.
4.2.3. Possessive pronouns In addition to attributive possession, possessive pronouns may also be used predicatively as below in (27a) and (27b). The short, sharp movement of predicative possessive pronouns towards the PR is often reiterated, possibly due to its phrase final position toward the location associated with the referent (cf. Nespor and Sandler 1999). This form may be one-handed or twohanded, as shown in Figures 17a and 17b.
Figure 17a. POSS-non1++
414 Kearsy Cormier and Jordan Fenlon
Figure 17 b. POSS-non1 (2hands)++
(27) a. SAW SOMEONE TAKE CUP. BEEN TOLD POSS-1++ ‘I saw someone take my cup. I had told him it was mine.’ b. CAR POSS-non1++ ‘The car is yours.’
5. Adjectival predicate of predisposition The predicative possessive pronoun has also evolved into an adjectival predicate indicating that the ‘possessor’ has a predisposition to act a certain way, or that a certain characteristic is typical of the ‘possessor’. The term ‘predispositional’ follows the use of the term predispositional aspect coined by Klima and Bellugi (1979) to describe aspectual modulation on some ASL adjectival predicates to indicate a predisposition to a certain characteristic (e.g. ASL MISCHEVIOUS vs. MISCHIEF-PRONE, where the sign MISCHIEF-PRONE is the same as MISCHEVIOUS but is twohanded and has a reduplicated circular path movement instead of only the hand-internal movement of MISCHEVIOUS). Examples of the BSL adjectival predicate of predisposition BE-TYPICAL-OF, identical in form to the predicative possessive pronouns shown above in Figure 17, are given in (28a) and (28b).
Possession in the visual-gestural modality 415
(28) a. RABBIT BOUNCE-ALONG LAUGH, SLOW IS-TYPICAL-OFnon1++ PRO-1 FAST ‘The rabbit bounced along laughing and said “It’s your way to be slow, I am fast.” ’ b. WHY IS-TYPICAL-OF-non1(2hands)++ ALWAYS FAR STILL CRAWL ‘Why is it always your way, even though it is far, to still crawl along?’ BSL also has a compound sign which has combined the adjectival predicate of predisposition with the sign ALWAYS to resulting in an adverbial form meaning ‘typically’, glossed in (29) as ALWAYS-TYPICAL and shown in Figure 18. In ALWAYS-TYPICAL, the repeated movement of the predispositional marker is lost. This pattern of phonological reduction of movement in the surface form is common in compound signs in signed languages (Klima and Bellugi 1979; Sutton-Spence and Woll 1999).
Figure 18. ALWAYS-TYPICAL
(29) JONATHAN ALWAYS-TYPICAL-a LATE ‘Jonathan is always late.’ The predicative possessive pronoun in ASL can also be used predispositionally, either alone (BE-TYPICAL-OF, formationally identical to POSS++) or compounded with the verb TEND (TEND-POSS). This form TENDPOSS, glossed by Klima and Bellugi (1979) as TEND^(HIS), is used with
416 Kearsy Cormier and Jordan Fenlon predicates to indicate predisposition to a certain activity or characteristic. As with BSL ALWAYS-TYPICAL, the repeated movement of the predispositional marker BE-TYPICAL-OF in the ASL compound TEND-POSS is lost. These predispositional markers in both BSL and ASL are very likely examples of grammaticisation of a possessive pronoun to an aspectual marker denoting predisposition. Typologically grammaticisation of possession into aspect marking in spoken languages is very common (Heine 1997). However, this grammaticisation path typically occurs with verbal markers of possession which then become verbal aspectual markers (e.g. English ‘have’). With BSL (and ASL), what we see instead is pronominal markers of possession (POSS ‘his/her/its’) becoming adjectival predicates of predisposition (BETYPICAL-OF). The BSL and ASL verbs HAVE do not have any aspectual function at all and only mark possession or existence (as described in §4.2.1).
6. Discussion and conclusion In this chapter we have seen that British Sign Language exhibits many of the same patterns of possession as spoken languages and in similar ways. BSL has both attributive and predicative possession. BSL exhibits patterns consistent with those in spoken languages with inalienable possession. The lexical item HAVE is used for both possession and existence, which is another pattern seen in many spoken languages. BSL also has a split between HAVE and BELONG (foregrounded PR vs. foregrounded PM) shared by many spoken languages. When we turn to looking at differences across signed languages regarding possession, these are generally the types of differences that we find across spoken languages – i.e. those features which are subject to parametric variation. Some languages like Australian Sign Language (Auslan), ASL and English have a possessive clitic which attaches to nouns; some languages like Danish (and, it seems, present-day BSL) do not. Some spoken languages mark inalienability while others do not; there is evidence that BSL and some other signed languages mark inalienability (optionally at least) though it seems in different ways. So it is clear there are many similarities between possession in signed languages and spoken languages. The differences seem to be strongest when considering the locative nature of many BSL signs. Across spoken languages, there is a strong link between possession, existentials and location
Possession in the visual-gestural modality 417
(Freeze 1992; Heine 1997; Herslund and Baron 2001).10 Specifically, in many spoken languages, the PM in a possessive construction is encoded as subject while the PR is encoded as a locative (e.g. dative) argument. In signed languages, signs such as pronouns, some nouns, agreement verbs, demonstratives, etc. are inherently locative. As we have seen in this chapter, many signs denoting (and/or grammaticised from) possession – e.g. possessive pronouns, nouns which can be spatially located, BELONG as an agreement verb, the predispositional markers BE-TYPICAL-OF and ALWAYS-TYPICAL – are spatially modified. This use of space is a natural product of the visualgestural modality and is perhaps the most striking difference between signed and spoken languages when it comes to possession (and also pronominal/ nominal reference and agreement in general), and is a good reason to consider signed languages when looking at the marking of possession across different languages. Indeed, typological studies which aim to cover phenomena across a wide variety of languages often neglect to include signed languages. This oversight is unfortunate because signed languages have much to offer language typologists. The features and categories that all known spoken and signed languages have in common are candidates for universal grammar, while those categories that differ across signed and spoken languages – particularly those which are thought to be universal amongst spoken languages – can help us tease apart those features which are truly inherent to human language versus those which may not be universal after all. In this chapter we have seen that while possession is likely a semantic category that is expressed in all human languages, language modality certainly does shape the way in which it is expressed. Acknowledgements We would like to acknowledge the Possession in Sign Languages project headed by Ulrike Zeshan; some of the elicitation materials used to collect some of the data given in this chapter were created by Zeshan specifically for that project, originally based at the Max Planck Institute for Psycholinguistics. We are also grateful to the participants from whom we collected data, and to Mark Wheatley, June Fenlon, Helen Earis, and Sandra Smith for modeling for the example figures in the chapter. We thank Rachel Sutton10
Though see Payne (this volume) for arguments against the “possession is location” view.
418 Kearsy Cormier and Jordan Fenlon Spence for assistance in collecting, analysing and discussing data presented in this chapter, Bencie Woll for her input particularly on inalienable possession, and Frances Elton and Clark Denmark for helpful discussions of some of the signs and patterns described in this chapter. This work was supported by the Economic and Social Research Council of Great Britain (Grant RES620-28-6001) Deafness, Cognition and Language Research Centre (DCAL) and also by the Centre for Deaf Studies (University of Bristol). Finally, we thank William McGregor and Adam Schembri for helpful comments on drafts of this chapter. Any errors that may remain are, of course, our own.
References Alibasic Ciciliani, Tamara and Ronnie B. Wilbur 2006 Pronominal system in Croatian Sign Language. Sign Language and Linguistics 9 (1–2): 95–132. Arik, Engin 2008 Locatives, existentials and possessives in Turkish Sign Language (TID). Paper presented at the Annual Meeting of the Linguistic Society of America, Chicago, IL. Berenz, Norine 2002 Insights into person deixis. Sign Language and Linguistics 5 (2): 203–227. Branson, Jan, Jennifer Toms, Brian Bernal and Don Miller 1995 The history and role of fingerspelling in Auslan. In Sign Language Research 1994: Proceedings of the Fourth European Congress on Sign Language Research, Munich, September 1–3, 1994, Heleen Bos and Trude Schermer (eds.), 53–67. Hamburg: Signum Press. Brennan, Mary 1992 An introduction to the visual world of BSL. In Dictionary of British Sign Language/English, David Brien (ed.), 1–133. London: Faber and Faber. Cooper, W. Robert 2002 Inalienable possession in Finnish and English: the use of possessive pronouns/suffixes with nouns designating part of the body. Helsinki English Studies 2. Cormier, Kearsy 2005 Exclusive pronouns in American Sign Language. In Clusivity: Typology and Case Studies of Inclusive-Exclusive Distinction, Elena Filimonova (ed.), 241–268. Amsterdam: John Benjamins.
Possession in the visual-gestural modality 419 Cormier, Kearsy 2007 Do all pronouns point? Indexicality of first person plural pronouns in BSL and ASL. In Visible Variation: Comparative Studies on Sign Language Structure, Pamela M. Perniss, Roland Pfau and Markus Steinbach (eds.), 63–101. Berlin /New York: Mouton de Gruyter. Cormier, Kearsy, Stephen Wechsler and Richard P. Meier 1999 Locus agreement in American Sign Language. In Lexical and Constructional Aspects of Linguistic Explanation, Gert Webelhuth, JeanPierre Koenig and Andreas Kathol (eds.), 215–229. Stanford, CA: CSLI Press. Deaf Society of New South Wales 1998 Hands up NSW: A Profile of the Deaf Community of NSW. Sydney: Deaf Society of New South Wales. Deuchar, Margaret 1983 Is British Sign Language an SVO language? In Language in Sign: An International Perspective on Sign Language, Jim Kyle and Bencie Woll (eds.), 69 –76. London: Croom Helm. Engberg-Pedersen, Elisabeth 1993 Space in Danish Sign Language. Hamburg: Signum Press. Farris, Michael A. 1994 Sign language research and Polish Sign Language. Lingua Posnaniensis 36: 13–36. Farris, Michael A. 1998 Models of person in sign languages. Lingua Posnaniensis 40: 47–59. Fenlon, Jordan and Kearsy Cormier 2006 Inalienable possession in British Sign Language. Paper presented at the Ninth International Conference on Theoretical Issues in Sign Language Research, Universidade Federale de Santa Catarina, Florianopolis, SC, Brazil, 6–9 December 2006. Fischer, Susan D. 1975 Influences on word order change in American Sign Language. In Word Order and Word Order Change, Charles N. Li (ed.), 1–25. Austin: University of Texas Press. Freeze, Ray 1992 Existentials and other locatives. Language 68: 553–595. Friedman, Lynn 1975 Space and time reference in American Sign Language. Language 51 (4): 940–961. Heine, Bernd 1997 Possession: Cognitive Sources, Forces, and Grammaticalization. Cambridge: Cambridge University Press.
420 Kearsy Cormier and Jordan Fenlon Herslund, Michael and Irène Baron 2001 Introduction: Dimensions of possession. In Dimensions of Possession, Irène Baron, Michael Herslund and Finn Sørensen (eds.), 1–25. Amsterdam: John Benjamins. Johnston, Trevor 2003 BSL, Auslan and NZSL: Three signed languages or one? In Proceedings of the Seventh International Conference on Theoretical Issues in Sign Language Research, Anne Baker, Beppie van den Bogaerde and Onno Crasborn (eds.), 47–69. Hamburg: Signum Verlag. Johnston, Trevor and Adam Schembri 2007 Australian Sign Language: An Introduction to Sign Language Linguistics. Cambridge: Cambridge University Press. Kliffer, Michael D. 1996 Commonalities of French and Mandarin inalienable possession. Language Sciences 18 (1–2): 53–69. Klima, Edward and Ursula Bellugi 1979 The Signs of Language. Cambridge, MA: Harvard University Press. Kristoffersen, Jette Hedegaard 2003 Existence, location and possession and the order of constituents in Danish Sign Language. In Crosslinguistic Perspectives in Sign Language Research: Selected Papers from TISLR 2000, Anne Baker, Beppie van den Bogaerde and Onno Crasborn (eds.), 131–139. Hamburg: Signum Verlag. Liddell, Scott K. 2003 Grammar, Gesture and Meaning in American Sign Language. Cambridge: Cambridge University Press. Lillo-Martin, Diane, and Edward Klima 1990 Pointing out differences: ASL pronouns in syntactic theory. In Theoretical Issues in Sign Language Research, Vol. 1: Linguistics, Susan D. Fischer and Patricia Siple (eds.), 191–210. Chicago: University of Chicago Press. MacLaughlin, Dawn 1997 The structure of determiner phrases: evidence from American Sign Language. PhD thesis, Department of Linguistics, Boston University. McKee, David and Graeme Kennedy 2000 Lexical comparisons of signs from American, Australian, British and New Zealand Sign Languages. In The Signs of Language Revisited: An Anthology to Honor Ursula Bellugi and Edward Klima, Karen Emmorey and Harlan Lane (eds.), 49–76. Mahwah, NJ: Lawrence Erlbaum Associates. Meier, Richard P. 1990 Person deixis in ASL. In Theoretical Issues in Sign Language Research, volume 1: Linguistics, Susan D. Fischer and Patricia Siple (eds.), 175–190. Chicago: University of Chicago Press.
Possession in the visual-gestural modality 421 Mitchell, Ross E. and Michael Karchmer 2004 Chasing the mythical ten percent: parental hearing status of deaf and hard of hearing students in the United States. Sign Language Studies 4 (2): 138–163. Neidle, Carol, Judy Kegl, Dawn MacLaughlin, Ben Bahan and Robert Lee 2000 The Syntax of American Sign Language. Cambridge, MA: MIT Press. Nespor, Marina and Wendy Sandler 1999 Prosody in Israeli Sign Language. Language and Speech 42 (2–3): 143–176. Nichols, Johanna 1988 On alienable and inalienable possession. In In Honor of Mary Haas: From the Haas Festival Conference on Native American Linguistics, W. Shipley (ed.), 557–609. Berlin: Mouton de Gruyter. Nichols, Johanna 1992 Linguistic Diversity in Space and Time. Chicago: University of Chicago Press. Padden, Carol A. 1983 Interaction of morphology and syntax in American Sign Language. PhD thesis, Department of Linguistics, University of California at San Diego. Padden, Carol A. 1990 The relation between space and grammar in ASL verb morphology. In Sign Language Research: Theoretical Issues, Ceil Lucas (ed.), 118–132. Washington, DC: Gallaudet University Press. Payne, Doris this vol. Is possession mere location? Contrary evidence from Maa. Pichler, Deborah Chen, Katharina Schalber, Julie Hochgesang, Marina Milkovic, Ronnie B. Wilbur, Martina Vulje et al. 2008 Possession and existence in three sign languages. In Sign Languages: Spinning and Unraveling the Past, Present and Future. TISLR9, Fortyfive Papers and Three Posters from the 9th Theoretical Issues in Sign Language Research Conference, Ronice Muller de Quadros (ed.), 440–458. Petropolis /RJ. Brazil: Editora Arara Azul. Rathmann, Christian and Gaurav Mathur 2002 Is verb agreement the same cross-modally? In Modality and Structure in Signed and Spoken Languages, Richard P. Meier, Kearsy Cormier and David Quinto-Pozos (eds.), 370–404. Cambridge: Cambridge University Press. Schembri, Adam and Trevor Johnston 2006 Sociolinguistic variation in Australian Sign Language Project: Grammatical and lexical variation. Paper presented at the 9th International Conference on Theoretical Issues in Sign Language Research, Universidade Federale de Santa Catarina, Florianopolis, SC, Brazil, 6–9 December 2006.
422 Kearsy Cormier and Jordan Fenlon Sutton-Spence, Rachel and Bencie Woll 1999 The Linguistics of British Sign Language. Cambridge: Cambridge University Press. Woll, Bencie, Rachel Sutton-Spence and Dafydd Waters 2004 ECHO data set for British Sign Language (BSL). London: Department of Language and Communication Science, City University. http://www.let.ru.nl/sign-lang/echo. Wulf, Alyssa, Paul Dudis, Robert Bayley and Ceil Lucas 2002 Variable subject presence in ASL narratives. Sign Language Studies 3 (1): 54 –76.
Contributors
Kearsy Cormier Deafness Cognition and Language Research Centre (DCAL) University College London 49 Gordon Square London WC1H 0PD UK
Mirjam Fried Ústav pro jazyk esk AV R Letenská 4 118 51 Praha 1 – Malá Strana Czech Republic [email protected]
[email protected]
Liesbet Heyvaert Department of Linguistics
Kristin Davidse Department of Linguistics
University of Leuven Blijde-Inkomststraat 21 PO Box 3308 3000 Leuven Belgium
University of Leuven Blijde-Inkomststraat 21 PO Box 3308 3000 Leuven Belgium [email protected] Sonja Eisenbeiß Department of Language and Linguistics
[email protected] Frantisek Lichtenberk Department of Applied Language Studies and Linguistics
University of Essex Colchester C04 3SQ UK
University of Auckland Private Bag 92019 Auckland New Zealand
[email protected]
[email protected]
Jordan Fenlon Deafness Cognition and Language Research Centre (DCAL)
Ayumi Matsuo The School of English Literature, Language and Linguistics
University College London 49 Gordon Square London WC1H 0PD UK
Jessop West 1 Upper Hanover Street Sheffield S3 7RA UK
[email protected]
[email protected]
424 Contributors William B. McGregor Afdeling for Lingvistik – Institut for Antropologi, Arkæologi og Lingvistik Aarhus Universitet Building 1410 Ndr. Ringgade 8000 Århus C Denmark
Miriam van Staden Theoretical Linguistics Faculty of Humanities University of Amsterdam Spuistraat 210 1012 VT Amsterdam The Netherlands
[email protected]
[email protected] [email protected]
Doris Payne Department of Linguistics
Hein van der Voort Afdeling Taalwetenschap
University of Oregon Eugene, OR 97403 USA
Radboud Universiteit Nijmegen Postbus 9103 6500 HD, Nijmegen The Netherlands
[email protected]
[email protected] Jan Rijkhoff Afdeling for Lingvistik – Institut for Antropologi, Arkæologi og Lingvistik
Peter Willemse UFR des Langues Etrangères Appliquées
Aarhus Universitet Building 1410 Ndr. Ringgade 8000 Århus C Denmark
Université Lille 3 Charles-de-Gaulle 14 Place Bodart Timal BP 447 59058 ROUBAIX cedex 01 France
[email protected]
[email protected]
Ingrid Sonnenstuhl Institute for Language and Information University of Düsseldorf Universitaetsstrasse 1 40225 Düsseldorf Germany [email protected]
Index of subjects
abstract nouns, 67, 225–226 abstract possession, 181, 359 active possession, 276 actor prefixes, 295, 298, 312–313, 322 adjectival predicates of predisposition, 389, 414–416 adjectives, 54–55, 94, 121, 214n, 295– 298, 304, 310–312, 336 classifying, 63–64, 67 relational, 64n adjuncts, see satellites adnominal possession (see also attributive possession), 2–5, 8, 51–97, 144, 150, 343–384 in child language, 155–179, 185–186, 188–189, 193, 196–199 adverbs, 52, 345, 359 adverb formatives, 359 affectedness, 6, 192, 197, 214–216, 219– 221, 227, 229–233, 235, 239–240, 242, 244–245 affected objects, 333–335 affected possessors, 214–246 passim affective construal, 243 affective verbs, 227, 230, 236 Agent – see agent role agent role, 100, 110, 112, 119, 123, 180, 224, 243, 267–268, 276, 294, 333 agentive participant, 231 alienable noun, 377, 382 alienable possession, 7, 262, 277, 286 vs. inalienable possession, 9, 262– 263, 282–285, 299, 319, 320, 356, 358, 375, 381–382, 406–407 alienable possessum, 25, 42, 44, 192, 216 alienability, 215–216, 221, 225, 231, 364, 376 anchor – see referential anchor animate, 1, 82, 107, 109, 158, 197, 223
applicative, 123, 319 areal diffusion, 343, 370n areal features, 359, 372, 383–384 areal patterns, 336, 381 association (see also bridging), 22–24, 37–38, 298–300 attributive compound, 380 attributive construction, 344, 349–351 attributive possession (see also adnominal possession), 2–3, 6, 9, 97, 213– 214, 217, 231–234, 238, 239n, 246, 249–286, 293, 316, 319, 344, 396– 408, 416 non-universality of contrast with predicative possession, 7–8, 294, 303–314, 324, 336 benefactive, 8, 123, 299, 355, 360–364, 367, 374, 381, 383 beneficiary, 109n, 232, 233, 299, 309, 316, 355, 360–362, 367, 374 body part classifiers, 353, 366, 372 body part nouns/terms, 40–41, 117n, 153, 175, 195–196, 197, 216, 219, 221, 225n, 227, 230, 242, 264, 274, 277, 278–279, 299, 358, 405, 406–407 body product terms, 265 bridging, 20–25, 36, 37, 39–41 causative (see also benefactive), 8, 355 classifiers (see also modifier, classifying and possessive classifiers), 8–9, 264, 281, 347–348, 349n, 351–354, 366– 367, 370–373, 377 comitative, 355, 367 compounds, 66, 349n, 369, 380, 415 Construction Grammar, 6, 108, 217, 234– 235, 246 constructional gestalt, 245
426 Index of subjects constructional networks, 213, 217, 238– 240, 246 context, 6, 20–21, 23, 113, 226–227, 250, 283, 286 coreference (see also discourse status, co-referential), 17 corpora, 13n, 27, 44n, 45, 134, 154, 202, 213, 390 dative arguments, 192, 194, 197, 232 dative constructions, 192, 239 dative marking, 156–157, 183, 189– 191, 195–196, 198, 214, 222, 227, 231–-232, 236, 239, 245 dative object, 190–192 Dative of Interest (DI), 6, 233, 236–237, 239 dative possessor, 214, 222, 227, 231–232, 235–236, 245 definite, 13–28 passim, 37–47 passim, 54–55, 57, 59, 79–89 passim, 93, 181–182, 187, 332 dependency reversal, 59, 76n direct possessive construction, 6–7, 253– 254, 256, 258–260, 263–267, 269– 270, 273, 276–277, 279, 282, 285 direct prefixes, 121 discourse context, 3, 15n, 19, 22, 30–32, 37, 42, 47–48, 240 discourse status anchored, 27, 41–44, 47 coreferential, 27, 28 –31, 36, 379, 382–383 given, 3, 13, 17–20, 26–28, 41, 47 inferable, 25–27, 36–47 initial mention, 15, 24, 26, 28n, 46 new, 3, 13, 17–20, 24–28, 31, 37–38, 41–42, 45–47 non-initial, 24, 46 text referent, 27, 32–36, 41 discourse functions, 20, 47–48 double object construction, 299 dummy, 9, 300, 302, 335, 358–359, 370– 371, 380
existence, 3–4, 80, 109, 111, 115–116, 133, 138, 181, 409–411, 416 existential constructions, 4, 82, 108n, 115, 128n, 133–134, 136–138, 300 existential predicates/verbs, 108, 120, 301–302, 359, 409–411 existential schema, 216 experiential gestalt, 216, 241, 245, 294 extension in child language, 174, 178–179, 188 metaphorical, 114n, 138 prototype, 244, 246 external possession constructions (EPCs) (see also affected possessors), 2, 5– 6, 144, 153–154, 189–192, 199, 214, 216, 230, 344 in child language, 192–198 external possessor – see external possession construction flexible parts-of-speech, 296–297 fluidity (of classifiers), 7, 250, 263, 273– 276, 281–282, 285 focus, 222–223 focus markers, 304, 307–308 frame, 22, 24, 234 cognitive, 217 interpretive, 234 frame participant, 235, 237, 239 Functional Grammar, 52 functional prototype, 217 functional space, 217, 238, 246 genitive compound, 369 construction, 6, 14, 56, 108, 150, 156, 162, 167, 171, 178, 381 non-determiner/non-referential, 63, 75 possessor, 214, 222, 238, 245 of quantification, 77 -s, see -s genitive schema, 217, 233 syntactic – see syntactic genitive goal schema, 217, 233
Index of subjects grammatical construction, 213, 217, 234, 239, 246 grammatical role, 152, 180, 183, 188, 216, 232n grammaticalization, 59, 61–62, 77, 124, 138, 272, 327, 374–375, 383 habitual possession relation, 175, 177 homophony, 354–355, 358, 371 identifiable (referent), 15n, 16, 19, 23, 55, 80, 82, 93, 95, 136 identification, 15–22, 47, 81, 93n Impersonal passive, 122, 133, 137 impersonal reference, 280 inalienable noun, 321, 375n, 376–377, 382, 407 inalienable possession, 7, 10, 40–41, 60, 117, 158, 192, 217, 219, 262, 264, 277–278, 284, 405–407, 416 inalienable possessum, 235, 239, 285, 321, 366 inalienability, 6, 66n, 215–216, 221, 231, 244, 279, 416 inanimate, 1, 110, 158, 216, 223, 301n inclusive-exclusive contrast, 332, 346, 394 inclusive pronouns, 249n, 347, 355, 374, 376 indefinite, 14, 16, 19, 21, 24–25, 30, 42– 44, 47, 54–55, 83–89, 92–93, 182– 183, 187 indirect object, 190–192, 232n indirect possession, 6–7, 253–254, 258, 263–264, 267–273, 275, 276–279, 282, 285, 373, 378 individuation, 279 indivisibility, 220 inflexion, 345–346, 348–349, 360–361, 363, 374–375 information flow, 217, 222 given, 47 new, 24
427
status, 307 structure, 222–223, 227–228, 235– 236 interpersonal level, 52, 53, 62 intrinsic relation, 262, 282, 285 inverse prefixes, 121–122, 130 involvement, 217, 219, 227 kin terms, 44, 145, 147, 159, 165, 167, 169–171, 199, 270, 277–278, 299, 325, 357n, 375–376, 405–407 kinship, 119, 153, 176, 221, 226, 229, 262, 266, 272, 282, 284, 298, 344, 358 as anchoring relation, 42–43 classifier, 276, 285 language contact, 315, 317, 319, 332, 389 layering in the noun phrase, 51–56, 60, 62– 90, 96 in constructions, 238, 245 levels – see interpersonal level, representational level linguistic areas, 293, 314, 317 location, 4, 52–55, 58, 62, 80, 120, 132, 134–135, 137, 180, 185–188, 321 relation to possession, 82, 107–108, 109–117, 137–138, 181, 416–417 predication of, 124–127 locative schema, 326 malefactive, 299 metaphor, 61, 109, 114n, 126, 138, 344 metaphorical constructions, 76 metaphorical Locative, 110, 126, 138 metatypy, 314–315, 317 modification relation, 51, 62, 66–67, 70, 74, 89–90, 94, 96 Modification construction, 231, 237–238 modifiers (see also satellites) adnominal, 60–61, 82–83, 97 aspectual, 305–306, 312 classifying, 51, 53, 55, 58, 61–62, 87, 96–97
428 Index of subjects discourse-referential, 53, 59 grammatical, 52–53 illocutionary, 53 localizing, 51, 53, 58, 79 proposition, 53 qualifying, 51, 53, 58, 61, 67, 71, 74, 93, 97 quantifying, 53, 56, 79 temporal, 312 modifying construction – see attributive construction nominalizations, 26, 45–46, 126, 224, 239, 267, 276, 295, 297, 311, 336, 345–352, 354, 367, 370n, 371 non-agentive, 236–237, 243 non-topicality, 38–39 noun formative, 353, 358–359, 366–367, 369, 372 noun phrase in child language, 154–155, 202–203 layered model of – see layering object prefixes, 319, 322, 330–335 operators, 52, 89 clause operators, 52, 54 NP operators, 54, 62, 77–79 part-whole relations, 21–22, 37, 41, 60, 107, 119, 131, 172, 177–178, 216, 242, 282, 299, 351, 358 passive possession, 276 possessibility hierarchy, 126, 221, 224– 225, 234–235, 238–239, 243, 245 possessive classifiers, 6–7, 9, 249, 253– 254, 256–257, 261, 264, 267–277, 281–286, 352, 373, 378, 381–382 possessive construal, 31, 216, 226, 245 possessive determiner, 13n, 14, 37, 47, 284 possessive frame, 236 –244 passim possessive prefixes, 254, 293, 298–299, 312–313, 320–321, 322, 329–333, 368–370, 372, 377–378, 380
possessive pronouns, 3, 9, 28, 59, 155– 163, 165–167, 169, 172, 178, 197– 198, 237, 320, 324–329, 360–363, 365, 367, 373–374, 381–383, 389, 390n, 396–408, 413–417 possessive prototype, 217, 238, 240–241, 244–246 possessive -s – see -s genitive possessive schema, 216, 245 predication, 3, 51, 62, 66–67, 70, 74, 89– 90, 94, 96–97, 117, 185–188, 295, 302–304, 310, 326, 333, 335 of existence, 115, 120 of location, 115, 124, 126, 132, 138 predicative possession, 2, 5, 7, 9–10, 107–108, 111–112, 114, 117, 119, 136, 144, 154, 179–189, 213, 294, 303, 316, 322, 329, 336, 343–344, 358–359, 363–364, 384, 389, 408– 414, 416 in child language, 182–189 prenominal possessive, 13–20, 25, 37n, 38, 40 prepositional phrases, 66, 84, 159, 195– 196, 198 prepositional possessive constructions, 5, 150, 173 pronouns in British Sign Language inalienable possession, 405–407 exclusivity, 394–395, 402 number, 392–393, 397–400 person, 390–392, 397–400 spatial marking, 390, 407–408 word order, 395–396, 400–402 prototypical ownership, 216, 234 prototypical possession/possessive relation, 158–159, 176, 198, 234 quotative markers, 56 reduplication, 10, 296, 384 reference point (see also referential anchor), 17–18, 21, 113, 117, 119
Index of subjects reference point construction, 2–3, 13, 17–21, 30, 47, 48, 113 referential anchor (see also reference point), 55, 80–81, 83, 95 relational classifier – see possessive classifier relative clauses, 24, 52, 55–56, 81–82, 222, 349–350, 352 representational level, 52–53 satellites, 52, 54–55, 58–59, 61–64, 66– 70, 77–83, 89, 91, 93–96 schemas, 23, 112, 152, 181–182, 293 scope, 52–53, 61–62, 77, 89, 96, 279, 300, 305–306 semantic roles, 107–111, 113, 114n, 123, 137, 232n, 236, 326, 329, 333 -s genitive, 5, 13n 27, 116, 143, 145, 147, 149–150, 156, 158–159, 164–172, 178, 196, 198–199, 404–405 situated possession, 6, 234–237, 239– 242, 245 subject prefixes, 293, 319–320, 333–334 switch-reference, 378–379, 383
429
syntactic constraints, 159, 170–171, 178, 199, 216–217, 230–231 syntactic functions, 224, 230, 235, 348 syntactic roles, 236 syntactic genitives, 239–240 ‘thematic’ dative, 232 thematic role, 190 Theme, 109–114, 116, 120, 123, 125, 135, 137, 267–268, 276, 321, 336 topic, 48, 112, 180, 408 topic construction, 117, 300–301 topic marking, 395n topical item, 38, 114n, 158, 236 topicality, 121, 126 topic-focus relations, 222 transitivity hierarchy, 230 unaffected object, 333–335 universals, 383–384, 417 verb semantics, 217, 227–230, 239, 243– 244, 246 verbal ligature, 8, 293, 315–316, 319
Index of languages
Abui, 293, 318, 320–321 Adang, 318, 320 ’Ala’ala, 270 Amazonian languages, 79, 343–384 American Sign Language (ASL), 9, 389, 391, 392n, 394n, 395, 401, 405, 407, 411, 414–416 Amharic, 115–116 Anejom), 254, 256–257, 277, 286 Araki, 263, 273 Aikanã, 8, 359–361, 364–367, 370, 381–383 Arikapú, 378–383 Australian languages, 79 Australian Sign Language (Auslan), 9, 395, 404–405, 416 Austrian Sign Language, 401, 407–408, 411 Austronesian languages, 76, 249–286, 293, 317–319, 332, 352, 382 Bali-Vitu, 258 Banoni, 265 Baure, 8, 371–374, 381–383 Biak, 319, 321 Buru, 293, 319, 321 British Sign Language (BSL), 9, 389– 417 Cavineña, 383 Cèmuhî, 257 Chadic languages, 78 Chiquitano, 383 Croatian Sign Language, 401, 407, 411 Czech, 6, 213–246 Danish Sign Language, 411 Dutch, 3–4, 51–97
East Nusantara languages, 7, 293–336 English, 2–3, 13–48, 51–97, 112, 118– 120, 138, 143, 155–156, 157–159 passim, 164–172 passim, 178, 199, 283–284, 389, 390n, 404–405, 416 Erromangan, 256–257 Ewe, 82-83 Fijian, 254–255, 258, 263, 268, 274, 276, 278, 280, 283 French, 111, 138 Galela, 56, 296–297, 301n, 322, 324, 327, 330–331, 335 Gapapaiwa, 263, 277–278 German, 5, 143–203 Germanic languages, 62, 78 Greek (Modern), 160, 164, 173, 178 Hausa, 67 Hawaiian, 259, 271 Hebrew, 160, 164–165, 173–174, 178 Hoava, 254 Houaïlou, 270, 277 Iaai, 258, 267, 272, 282 il-Chamus, 107; see also Maa Indonesian, 59, 309, 314–315 Italian, 16 Jabutí languages, 378 Jakaltek, 111, 115–116 Japanese, 5, 164, 172, 178, 192, 194, 196, 198 Kairiru, 260, 277 Kanoê, 359, 365–367, 369, 381–383 Kilivila, 265–266
Index of languages Kokota, 253n, 267 Komi (Southern Permyak dialect), 59 Kosraean, 272, 275–276 Kwaio, 255 Kwazá, 8–9, 345–360, 362n, 364–372 passim, 381–383 passim Lakondê, 368 Labu, 261 Latundê, 368–371, 381–382 Lenakel, 271 Lenkau, 258 Lithuanian, 65–66 Loda, 322, 327 Lolovoli, 253, 265–266, 271, 274 Lou, 258 Maa, 4, 107-138 Maasai, 107, 122; see also Maa Macro-Jê languages, 343, 378–380 Malay/North Moluccan Malay, 293, 296, 309–310, 311n, 314–317 Manam, 249–250, 254, 265, 267–269, 274–276, 283, 285 Mauritian Creole, 59 Maybrat, 293, 320, 332 Mekens, 345, 376–378, 381–383 Movima, 10, 384 Mussau, 272, 282 Nalik, 266 Nambikwara languages, 343, 368, 369 Nasioi, 60 Niuean, 261 Nyulnyul, 4, 108n, 117n Oceanic languages, 6–7, 249–286, 318, 373, 378 Paamese, 264, 267, 278–279 Pagu, 322, 324, 326–329, 332–335 Panare, 78
431
Pohnpeian, 272 Polynesian languages, 259–260, 261, 270–271, 318 Pukapukan, 259 Puluwatese, 266 Romance languages, 62 Rotuman, 259 Roviana, 65 Russian, 77 Sahu, 301n, 325–330 passim Sakurabiat – see Mekens Samoan, 55–56 Signed English, 405 Swedish, 65, 75, 164 Tamambo, 265, 273 Tidore, 7–8, 293–330 passim, 334–336 Tobati, 261 Tobelo, 296–297, 299, 323–331 passim Tok Pisin, 293 Toqabaqita, 254, 260, 266–267, 270, 278–279 Tuparí languages, 376, 378 Turkic languages, 59, 356 Turkish, 59–60 Turkish Sign Language, 411 Ulithian, 272 Uralic languages, 59, 356 Vinmavis, 254, 257–258 Wari’, 374–376, 381–382 Wayan, 268, 278, 280 Western Oceanic languages, 76–77 West-Makian, 322–323 Yawa, 321 Zabana, 286
Index of persons
Abbott, Barbara, 14 Ahland, Micahel, 115, 116n Aikhenvald, Alexandra, 272, 277, 281, 345 Alibasic Cicilliani, Tamara, 392n, 395 Araujo, Gabriel de, 369 Ariel, Mira, 22–23, 40 Arik, Engin, 411 Armon-Lotem, Sharon, 160, 164–165, 172–173 Bacelar, Laércio, 365–367 Baker, Carl, 145 Barker, Chris, 30, 42, 262, 282 Baron, Irène, 1, 107, 109, 144, 180–181, 188, 199, 344, 356, 408, 417 Barshi, Immanuel, 2, 6, 107, 189, 192, 214, 216, 218, 225, 344 Bates, Elizabeth, 145, 148–149 Bauer, Laurie, 66 Bayley, Robert, 395 Bellugi, Ursula, 391, 414–415 Berenz, Norine, 392n Berman, Ruth, 164–165, 173 Bernal, Brian, 404 Bertolo, Stefano, 145 Benveniste, Émile, 359 Berger, Andrea, 79 Biber, Douglas, 14 Blanche-Benveniste, Claire, 31 Blust, Robert, 318 Bohnacker, Ute, 164 Börjars, Kersti, 158 Borschev, Vladimir, 262 Bowden, John, 266, 315 Bowerman, Melissa, 143, 179, 181 Braine, Martin, 149 Brannon, Elizabeth, 79 Branson, Jan, 404
Brennan, Mary, 409 Brooks, Patricia, 149 Brown, Dunstan, 253 Brown, Gillian, 22, 31 Brown, Roger, 143, 164–165, 172–173 Bugenhagen, Robert, 80 Butler, Christopher, 52 Chafe, Wallace, 14, 22–23, 28, 119 Chappell, Hillary, 2, 7, 66n, 218, 221, 293, 299, 358, 382 Chomsky, Noam, 148 Chouinard, M. M., 149 Croft, William, 217–218, 283 Crowley, Terry, 257–258, 264, 267, 278– 279, 352 Chervel, André, 31 Clahsen, Harald, 154, 156, 164–165, 173 Clancy, Patricia, 164–165, 173 Clark, Eve, 82, 109–110, 112, 115–116, 138, 148–149 Clark, Ross, 259–260 Claudi, Ulrike, 83 Comrie, Bernard, 59, 77, 109n Cooper, W. Robert, 406 Corina, David, 148 Cormier, Kearsy, 9, 389, 391–392, 394, 405, 423 Cowie, F., 148 Craig, Colette, 111 Crain, Stephen, 151, 160 Crevels, Mily, 345, 359 Danielsen, Swintha, 371–374 Davidse, Kristin, 2–3, 13, 423 Declerck, Renaat, 16, 21 DeLancey, Scott, 107, 109–110, 114 Denison, David, 158 Dik, Simon, 52, 54, 93, 97n, 303, 336
Index of persons Dixon, Robert, 67, 78, 283, 345 Donohue, Mark, 261 Du Bois, John, 2, 13–15, 22–24, 40 Dudis, Paul, 395 Eisenbeiß, Sonja, 4–5, 10, 143–202 passim, 423 Emmott, Catherine, 31 Engberg-Pedersen, Elisabeth, 392 Englebretson, Robert, 59 Everett, Daniel, 79, 374 –375 Fabre, Alain, 373, 382–383 Farris, Michael, 392 Fenlon, Jordan, 9, 389, 405, 423 Fischer, Steven, 259 –260 Fischer, Susan, 395n Fitzsimons, Matthew, 286 François, Alexander, 263 Frajzyngier, Zygmunt, 78 Fraurud, Kari, 23 –24, 42, 59 Freeze, Ray, 107, 412, 417 Friedman, Lynn, 391 Galasso, Joseph, 164–165, 173 Galucio, Ana, 376–378 Garrod, Simon, 23 Geraghty, Paul, 254, 268n Gerzenstein, Ana, 373 Giegerich, Heinz, 66 Gil, David, 78n, 362 Gildea, Spike, 78 Givón, Talmy, 78, 81 Goldberg, Adele, 6, 108, 145, 152 Greenberg, Joseph, 122 Grinevald, Collette, see Craig, Collette Gruber, Jeffrey, 109–110 Guasti, Maria, 151–152 Guillaume, Antoine, 383–384 Guillemin, Diana, 59 Gundel, Jeannette, 14 Gunkel, Lutz, 64n Gutierrez, Salome, 128n
433
Haiman, John, 283 Halliday, Michael, 15, 21, 32, 63 Hamaya, Mitsuyo, 120n, 121 Harbert, Wayne, 145, 159 Hasan, Ruqaiya, 15, 21, 32 Haspelmath, Martin, 78–79, 83, 144 Haude, Katharina, 10, 383–384 Hawkins, John, 20–22, 79 Hedberg, Nancy, 14 Heine, Bernd, 1, 7, 56, 61–62, 83, 97, 111–112, 118n, 138, 144, 153, 180– 181, 213–214, 216, 233, 293, 295, 300, 303, 327, 359, 382–383, 408– 409, 416–417 Hengeveld, Kees, 52, 56, 62n, 93, 298, 303 Herslund, Michael, 1, 107, 109, 144, 180– 181, 188, 199, 344, 356, 408, 417 Hinton, Leanne, 360 Hooper, Robin, 261 Huddleston, Rodney, 14 Hyams, Nina, 151 Hyslop, Catriona, 271 Jackendoff, Ray, 109 Jacobs, Peter, 120n, 121 Jäger, Gerhard, 158 Johnston, Trevor, 395, 404 Jordan, Kerry, 79 Karchmer, Michael, 390n Keesing, Roger, 255 Kegl, Judy, 395n, 401n Keizer, Evelien, 93 Kemmer, Suzanne, 113 Kennedy, Graeme, 404n Kern, Barbara, 374–375 Kiparsky, Paul, 148 Kirk, Arthur, 78 Kliffer, Michael, 407 Klima, Edward, 391, 414–415 König, Ekkehard, 144 König, Christa, 121
434
Index of persons
Koptjevskaja-Tamm, Maria, 75, 344 Kristoffersen, Jette Hedegaard, 411 Kroeker, Menno, 369 Kuna, Branko, 117n Küntay, Aylin, 149 Kuteva, Tania, 61–62, 83, 383 La Fontinelle, Jacqueline de, 270n, 277 Langacker, Ronald, 2, 14, 15n, 17–18, 31, 47, 113, 117, 118n, 262 LaPolla, Randy, 113, 119 Lee, Kee-dong, 272 Lee, Robert, 395, 401n Leenhardt, Maurice, 270 Lehmann, Christian, 81 Lichtenberk, Frantisek, 7, 249, 263–264, 269, 275, 279, 282–283, 285, 299, 344, 352, 373, 378, 382, 423 Liddell, Scott, 392 Lillo-Martin, Diane, 391 Lucas, Ceil, 395 Lust, Barbara, 151 Lynch, John, 249n, 256, 263, 267, 273, 277, 282, 286 Lyons, Christopher, 2, 13–14, 16, 20–21 Lyons, John, 61, 80, 82, 109, 114n MacLaughlin, Dawn, 395, 401n MacWhinney, Brian, 145, 148 –149, 173 Malchukov, Andrej, 59 Marcus, Gary, 146, 148–149 Marinis, Theodoros, 143, 160, 164, 173 Martin, James, 2, 13, 15–16, 21, 47 Massam, Diane, 261 Mathur, Gaurav, 391 Matsuo, Ayumi, 4–5, 10, 143, 194, 423 McCawley, James, 97n McGregor, William, 1–2, 7, 9, 40, 66n, 79, 95n, 108n, 117n, 180, 194, 218, 221, 284n, 293, 299, 355, 358, 382, 424 McGuckin, Catherine, 263, 277 McKee, David, 404n
Meier, Richard, 391 Menges, Karl Heinrich, 59 Menn, Lise, 165 Messineo, Cristina, 373 Miller, Don, 404 Miller, Max, 146 Mills, Anne, 145, 156, 165, 168 Milner, George, 255, 258, 263 Mitchell, Ross, 390 Mithun, Marianne, 78 Moravcsik, Edith, 81 Mpaayei, John, 121 Neidle, Carol, 395n, 401n Nespor, Marina, 413 Neumann, Dorothea, 189, 192 Newman, Paul, 67n, 78 Newmeyer, Frederik, 97n Nichols, Johanna, 273, 345, 382, 407 Ozanne-Rivierre, Françoise, 258, 272 Padden, Carol, 391 Palmer, Bill, 253n Partee, Barbara, 262 Pawley, Andrew, 263, 268, 278, 280– 281 Payne, Doris, 2, 4, 6, 107, 120n, 121, 182, 189, 192, 214, 216, 218, 225, 344, 417n, 424 Penner, Zvi, 156, 164–165, 172 Peters, Ann, 165 Pichler, Deborah, 401, 405, 407–408, 411 Pinker, Steven, 143, 145, 151–152, 179, 181, 184, 199 Prince, Ellen, 79 Quirk, Randolph, 2, 13–14, 75, 89 Radford, Andrew, 143, 152, 153n, 164– 165, 173 Rasmussen, Kent, 120n Rathmann, Christian, 391
Index of persons Rehg, Kenneth, 272 Ribeiro, Eduardo, 378–379 Rijkhoff, Jan, 3, 51–52, 54, 55n, 59, 62, 78–79, 356, 424 Rivierre, Jean-Claude, 257 Rizzi, Luigi, 151 Rodrigues, Aryon, 379–380 Rosenbach, Anette, 2, 13–14, 57n, 158 Ross, Malcolm, 76, 258, 270, 272, 314–315 Ruff, Claudia, 157, 162–163, 165, 167, 172–173 Sakel, Jeanette, 384 Sandler, Wendy, 413 Sanford, Anthony, 23 Sayaba, Timocy, 263, 268, 278, 280–281 Scott, Alan, 158 Schembri, Adam, 395, 404 Seibt, Johanna, 55n Seiler, Hansjakob, 1, 56, 153, 180, 359 Shibatani, Masayoshi, 189 Siegel, Jeff, 261 Siewierska, Anna, 59 Slobin, Dan, 149, 151 Song, Jae Jung, 62n Sonnenstuhl, Ingrid, 4 –5, 10, 143, 202, 424 Sørensen, Finn, 107, 144, 180 –181, 188, 199 Sperlich, Wolfgang, 261 Stassen, Leon, 344, 355 Stenzel, Achim, 165 Stephany, Ursula, 160 Sutton-Spence, Rachel, 390, 405, 415 Takahashi, Hidemitsu, 33 Tauli, Valter, 59 Taylor, John, 1–2, 13, 17–19, 21, 28, 42, 47, 63n, 216, 240, 294 Telles, Stella, 368 –371
435
Tham, Shiao Wei, 113 Tomasello, Michael, 145, 148–149, 151– 152, 153n, 167 Tryon, Darrell, 318 Tsunoda, Tasaku, 1, 5, 192, 216, 225 Tucker, Archibald, 121 Vainikka, Anne, 154, 164 –165, 173 van Baarda, M. J., 296, 301n, 322n, 326, 327, 331 van den Berg, Helma, 109n van den Berg, René, 258n van der Voort, Hein, 8–9, 10, 343, 345, 347, 349, 351, 357, 359, 365, 370n, 378–379, 424 Van Valin, Robert, 113, 119 Varlokosta, Spyridoula, 160 Vasconcelos, Ione, 360 Velazquez-Castillo, Maura, 113, 117n von Garnier, Katharine, 78 Wagner, Klaus, 154, 173 Waters, Dafydd, 390 Wechsler, Stephen, 391 Weissenborn, Jürgen, 156, 164–165, 172 Wertheimer, Max, 109 Wexler, Kenneth, 151, 153n Wilbur, Ronnie, 392n, 395 Willemse, Peter, 2–3, 13, 19–20, 63n, 84n, 424 Williams, Edward, 56 Wilson, William, 271 Wivell, Richard, 260, 277 Woll, Bencie, 390, 405, 415 Wulf, Alyssa, 395 Wurm, Stephen, 318 Young-Scholten, Martha, 154 Yule, George, 22, 31 Zacharski, Ron, 14
